diff --git a/daemons/fenced/pacemaker-fenced.c b/daemons/fenced/pacemaker-fenced.c index 092d872d59..a199803501 100644 --- a/daemons/fenced/pacemaker-fenced.c +++ b/daemons/fenced/pacemaker-fenced.c @@ -1,1751 +1,1752 @@ /* * Copyright 2009-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include // PRIu32, PRIx32 #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #define SUMMARY "daemon for executing fencing devices in a Pacemaker cluster" char *stonith_our_uname = NULL; long stonith_watchdog_timeout_ms = 0; GList *stonith_watchdog_targets = NULL; static GMainLoop *mainloop = NULL; gboolean stand_alone = FALSE; static gboolean stonith_shutdown_flag = FALSE; static qb_ipcs_service_t *ipcs = NULL; static xmlNode *local_cib = NULL; static pe_working_set_t *fenced_data_set = NULL; static const unsigned long long data_set_flags = pe_flag_quick_location | pe_flag_no_compat | pe_flag_no_counts; static cib_t *cib_api = NULL; static pcmk__output_t *logger_out = NULL; static pcmk__output_t *out = NULL; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static struct { bool no_cib_connect; gchar **log_files; } options; static crm_exit_t exit_code = CRM_EX_OK; static void stonith_shutdown(int nsig); static void stonith_cleanup(void); static int32_t st_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { if (stonith_shutdown_flag) { crm_info("Ignoring new client [%d] during shutdown", pcmk__client_pid(c)); return -EPERM; } if (pcmk__new_client(c, uid, gid) == NULL) { return -EIO; } return 0; } /* Exit code means? */ static int32_t st_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; int call_options = 0; xmlNode *request = NULL; pcmk__client_t *c = pcmk__find_client(qbc); const char *op = NULL; if (c == NULL) { crm_info("Invalid client: %p", qbc); return 0; } request = pcmk__client_data2xml(c, data, &id, &flags); if (request == NULL) { pcmk__ipc_send_ack(c, id, flags, "nack", NULL, CRM_EX_PROTOCOL); return 0; } op = crm_element_value(request, F_CRM_TASK); if(pcmk__str_eq(op, CRM_OP_RM_NODE_CACHE, pcmk__str_casei)) { crm_xml_add(request, F_TYPE, T_STONITH_NG); crm_xml_add(request, F_STONITH_OPERATION, op); crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, pcmk__client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); send_cluster_message(NULL, crm_msg_stonith_ng, request, FALSE); free_xml(request); return 0; } if (c->name == NULL) { const char *value = crm_element_value(request, F_STONITH_CLIENTNAME); if (value == NULL) { value = "unknown"; } c->name = crm_strdup_printf("%s.%u", value, c->pid); } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_trace("Flags %#08" PRIx32 "/%#08x for command %" PRIu32 " from client %s", flags, call_options, id, pcmk__client_name(c)); if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(flags & crm_ipc_client_response); CRM_LOG_ASSERT(c->request_id == 0); /* This means the client has two synchronous events in-flight */ c->request_id = id; /* Reply only to the last one */ } crm_xml_add(request, F_STONITH_CLIENTID, c->id); crm_xml_add(request, F_STONITH_CLIENTNAME, pcmk__client_name(c)); crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname); crm_log_xml_trace(request, "ipc-received"); stonith_command(c, id, flags, request, NULL); free_xml(request); return 0; } /* Error code means? */ static int32_t st_ipc_closed(qb_ipcs_connection_t * c) { pcmk__client_t *client = pcmk__find_client(c); if (client == NULL) { return 0; } crm_trace("Connection %p closed", c); pcmk__free_client(client); /* 0 means: yes, go ahead and destroy the connection */ return 0; } static void st_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p destroyed", c); st_ipc_closed(c); } static void stonith_peer_callback(xmlNode * msg, void *private_data) { const char *remote_peer = crm_element_value(msg, F_ORIG); const char *op = crm_element_value(msg, F_STONITH_OPERATION); if (pcmk__str_eq(op, "poke", pcmk__str_none)) { return; } crm_log_xml_trace(msg, "Peer[inbound]"); stonith_command(NULL, 0, 0, msg, remote_peer); } #if SUPPORT_COROSYNC static void stonith_peer_ais_callback(cpg_handle_t handle, const struct cpg_name *groupName, uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len) { uint32_t kind = 0; xmlNode *xml = NULL; const char *from = NULL; char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from); if(data == NULL) { return; } if (kind == crm_class_cluster) { xml = string2xml(data); if (xml == NULL) { crm_err("Invalid XML: '%.120s'", data); free(data); return; } crm_xml_add(xml, F_ORIG, from); /* crm_xml_add_int(xml, F_SEQ, wrapper->id); */ stonith_peer_callback(xml, NULL); } free_xml(xml); free(data); return; } static void stonith_peer_cs_destroy(gpointer user_data) { crm_crit("Lost connection to cluster layer, shutting down"); stonith_shutdown(0); } #endif void do_local_reply(xmlNode *notify_src, pcmk__client_t *client, int call_options) { /* send callback to originating child */ int local_rc = pcmk_rc_ok; int rid = 0; uint32_t ipc_flags = crm_ipc_server_event; if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_LOG_ASSERT(client->request_id); rid = client->request_id; client->request_id = 0; ipc_flags = crm_ipc_flags_none; } local_rc = pcmk__ipc_send_xml(client, rid, notify_src, ipc_flags); if (local_rc == pcmk_rc_ok) { crm_trace("Sent response %d to client %s", rid, pcmk__client_name(client)); } else { crm_warn("%synchronous reply to client %s failed: %s", (pcmk_is_set(call_options, st_opt_sync_call)? "S" : "As"), pcmk__client_name(client), pcmk_rc_str(local_rc)); } } uint64_t get_stonith_flag(const char *name) { if (pcmk__str_eq(name, T_STONITH_NOTIFY_FENCE, pcmk__str_casei)) { return st_callback_notify_fence; } else if (pcmk__str_eq(name, STONITH_OP_DEVICE_ADD, pcmk__str_casei)) { return st_callback_device_add; } else if (pcmk__str_eq(name, STONITH_OP_DEVICE_DEL, pcmk__str_casei)) { return st_callback_device_del; } else if (pcmk__str_eq(name, T_STONITH_NOTIFY_HISTORY, pcmk__str_casei)) { return st_callback_notify_history; } else if (pcmk__str_eq(name, T_STONITH_NOTIFY_HISTORY_SYNCED, pcmk__str_casei)) { return st_callback_notify_history_synced; } return st_callback_unknown; } static void stonith_notify_client(gpointer key, gpointer value, gpointer user_data) { xmlNode *update_msg = user_data; pcmk__client_t *client = value; const char *type = NULL; CRM_CHECK(client != NULL, return); CRM_CHECK(update_msg != NULL, return); type = crm_element_value(update_msg, F_SUBTYPE); CRM_CHECK(type != NULL, crm_log_xml_err(update_msg, "notify"); return); if (client->ipcs == NULL) { crm_trace("Skipping client with NULL channel"); return; } if (pcmk_is_set(client->flags, get_stonith_flag(type))) { int rc = pcmk__ipc_send_xml(client, 0, update_msg, crm_ipc_server_event); if (rc != pcmk_rc_ok) { crm_warn("%s notification of client %s failed: %s " CRM_XS " id=%.8s rc=%d", type, pcmk__client_name(client), pcmk_rc_str(rc), client->id, rc); } else { crm_trace("Sent %s notification to client %s", type, pcmk__client_name(client)); } } } void do_stonith_async_timeout_update(const char *client_id, const char *call_id, int timeout) { pcmk__client_t *client = NULL; xmlNode *notify_data = NULL; if (!timeout || !call_id || !client_id) { return; } client = pcmk__find_client_by_id(client_id); if (!client) { return; } notify_data = create_xml_node(NULL, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_TYPE, T_STONITH_TIMEOUT_VALUE); crm_xml_add(notify_data, F_STONITH_CALLID, call_id); crm_xml_add_int(notify_data, F_STONITH_TIMEOUT, timeout); crm_trace("timeout update is %d for client %s and call id %s", timeout, client_id, call_id); if (client) { pcmk__ipc_send_xml(client, 0, notify_data, crm_ipc_server_event); } free_xml(notify_data); } /*! * \internal * \brief Notify relevant IPC clients of a fencing operation result * * \param[in] type Notification type * \param[in] result Result of fencing operation (assume success if NULL) * \param[in] data If not NULL, add to notification as call data */ void fenced_send_notification(const char *type, const pcmk__action_result_t *result, xmlNode *data) { /* TODO: Standardize the contents of data */ xmlNode *update_msg = create_xml_node(NULL, "notify"); CRM_LOG_ASSERT(type != NULL); crm_xml_add(update_msg, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(update_msg, F_SUBTYPE, type); crm_xml_add(update_msg, F_STONITH_OPERATION, type); stonith__xe_set_result(update_msg, result); if (data != NULL) { add_message_xml(update_msg, F_STONITH_CALLDATA, data); } crm_trace("Notifying clients"); pcmk__foreach_ipc_client(stonith_notify_client, update_msg); free_xml(update_msg); crm_trace("Notify complete"); } /*! * \internal * \brief Send notifications for a configuration change to subscribed clients * * \param[in] op Notification type (STONITH_OP_DEVICE_ADD, * STONITH_OP_DEVICE_DEL, STONITH_OP_LEVEL_ADD, or * STONITH_OP_LEVEL_DEL) * \param[in] result Operation result * \param[in] desc Description of what changed * \param[in] active Current number of devices or topologies in use */ static void send_config_notification(const char *op, const pcmk__action_result_t *result, const char *desc, int active) { xmlNode *notify_data = create_xml_node(NULL, op); CRM_CHECK(notify_data != NULL, return); crm_xml_add(notify_data, F_STONITH_DEVICE, desc); crm_xml_add_int(notify_data, F_STONITH_ACTIVE, active); fenced_send_notification(op, result, notify_data); free_xml(notify_data); } /*! * \internal * \brief Send notifications for a device change to subscribed clients * * \param[in] op Notification type (STONITH_OP_DEVICE_ADD or * STONITH_OP_DEVICE_DEL) * \param[in] result Operation result * \param[in] desc ID of device that changed */ void fenced_send_device_notification(const char *op, const pcmk__action_result_t *result, const char *desc) { send_config_notification(op, result, desc, g_hash_table_size(device_list)); } /*! * \internal * \brief Send notifications for a topology level change to subscribed clients * * \param[in] op Notification type (STONITH_OP_LEVEL_ADD or * STONITH_OP_LEVEL_DEL) * \param[in] result Operation result * \param[in] desc String representation of level ([]) */ void fenced_send_level_notification(const char *op, const pcmk__action_result_t *result, const char *desc) { send_config_notification(op, result, desc, g_hash_table_size(topology)); } static void topology_remove_helper(const char *node, int level) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; xmlNode *data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); fenced_unregister_level(data, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_DEL, &result, desc); pcmk__reset_result(&result); free_xml(data); free(desc); } static void remove_cib_device(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); } if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } rsc_id = crm_element_value(match, XML_ATTR_ID); stonith_device_remove(rsc_id, true); } } static void remove_topology_level(xmlNode *match) { int index = 0; char *key = NULL; CRM_CHECK(match != NULL, return); key = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); topology_remove_helper(key, index); free(key); } static void add_topology_level(xmlNode *match) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; CRM_CHECK(match != NULL, return); fenced_register_level(match, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_ADD, &result, desc); pcmk__reset_result(&result); free(desc); } static void remove_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if (match && crm_element_value(match, XML_DIFF_MARKER)) { /* Deletion */ int index = 0; char *target = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); if (target == NULL) { crm_err("Invalid fencing target in element %s", ID(match)); } else if (index <= 0) { crm_err("Invalid level for %s in element %s", target, ID(match)); } else { topology_remove_helper(target, index); } /* } else { Deal with modifications during the 'addition' stage */ } } } static void register_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); remove_topology_level(match); add_topology_level(match); } } /* Fencing */ static void fencing_topology_init(void) { xmlXPathObjectPtr xpathObj = NULL; const char *xpath = "//" XML_TAG_FENCING_LEVEL; crm_trace("Full topology refresh"); free_topology_list(); init_topology_list(); /* Grab everything */ xpathObj = xpath_search(local_cib, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } #define rsc_name(x) x->clone_name?x->clone_name:x->id /*! * \internal * \brief Check whether our uname is in a resource's allowed node list * * \param[in] rsc Resource to check * * \return Pointer to node object if found, NULL otherwise */ static pe_node_t * our_node_allowed_for(const pe_resource_t *rsc) { GHashTableIter iter; pe_node_t *node = NULL; if (rsc && stonith_our_uname) { g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node && strcmp(node->details->uname, stonith_our_uname) == 0) { break; } node = NULL; } } return node; } static void watchdog_device_update(void) { if (stonith_watchdog_timeout_ms > 0) { if (!g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) && !stonith_watchdog_targets) { /* getting here watchdog-fencing enabled, no device there yet and reason isn't stonith_watchdog_targets preventing that */ int rc; xmlNode *xml; xml = create_device_registration_xml( STONITH_WATCHDOG_ID, st_namespace_internal, STONITH_WATCHDOG_AGENT, NULL, /* stonith_device_register will add our own name as PCMK_STONITH_HOST_LIST param so we can skip that here */ NULL); rc = stonith_device_register(xml, TRUE); free_xml(xml); if (rc != pcmk_ok) { rc = pcmk_legacy2rc(rc); exit_code = CRM_EX_FATAL; crm_crit("Cannot register watchdog pseudo fence agent: %s", pcmk_rc_str(rc)); stonith_shutdown(0); } } } else if (g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) != NULL) { /* be silent if no device - todo parameter to stonith_device_remove */ stonith_device_remove(STONITH_WATCHDOG_ID, true); } } static void update_stonith_watchdog_timeout_ms(xmlNode *cib) { long timeout_ms = 0; xmlNode *stonith_watchdog_xml = NULL; const char *value = NULL; stonith_watchdog_xml = get_xpath_object("//nvpair[@name='stonith-watchdog-timeout']", cib, LOG_NEVER); if (stonith_watchdog_xml) { value = crm_element_value(stonith_watchdog_xml, XML_NVPAIR_ATTR_VALUE); } if (value) { timeout_ms = crm_get_msec(value); } if (timeout_ms < 0) { timeout_ms = pcmk__auto_watchdog_timeout(); } stonith_watchdog_timeout_ms = timeout_ms; } /*! * \internal * \brief If a resource or any of its children are STONITH devices, update their * definitions given a cluster working set. * * \param[in,out] rsc Resource to check * \param[in,out] data_set Cluster working set with device information */ static void cib_device_update(pe_resource_t *rsc, pe_working_set_t *data_set) { pe_node_t *node = NULL; const char *value = NULL; const char *rclass = NULL; pe_node_t *parent = NULL; /* If this is a complex resource, check children rather than this resource itself. */ if(rsc->children) { GList *gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { cib_device_update(gIter->data, data_set); if(pe_rsc_is_clone(rsc)) { crm_trace("Only processing one copy of the clone %s", rsc->id); break; } } return; } /* We only care about STONITH resources. */ rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); if (!pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { return; } /* If this STONITH resource is disabled, remove it. */ if (pe__resource_is_disabled(rsc)) { crm_info("Device %s has been disabled", rsc->id); return; } /* if watchdog-fencing is disabled handle any watchdog-fence resource as if it was disabled */ if ((stonith_watchdog_timeout_ms <= 0) && pcmk__str_eq(rsc->id, STONITH_WATCHDOG_ID, pcmk__str_none)) { crm_info("Watchdog-fencing disabled thus handling " "device %s as disabled", rsc->id); return; } /* Check whether our node is allowed for this resource (and its parent if in a group) */ node = our_node_allowed_for(rsc); if (rsc->parent && (rsc->parent->variant == pe_group)) { parent = our_node_allowed_for(rsc->parent); } if(node == NULL) { /* Our node is disallowed, so remove the device */ GHashTableIter iter; crm_info("Device %s has been disabled on %s: unknown", rsc->id, stonith_our_uname); g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { crm_trace("Available: %s = %d", pe__node_name(node), node->weight); } return; } else if(node->weight < 0 || (parent && parent->weight < 0)) { /* Our node (or its group) is disallowed by score, so remove the device */ int score = (node->weight < 0)? node->weight : parent->weight; crm_info("Device %s has been disabled on %s: score=%s", rsc->id, stonith_our_uname, pcmk_readable_score(score)); return; } else { /* Our node is allowed, so update the device information */ int rc; xmlNode *data; GHashTable *rsc_params = NULL; GHashTableIter gIter; stonith_key_value_t *params = NULL; const char *name = NULL; const char *agent = crm_element_value(rsc->xml, XML_EXPR_ATTR_TYPE); const char *rsc_provides = NULL; crm_debug("Device %s is allowed on %s: score=%d", rsc->id, stonith_our_uname, node->weight); rsc_params = pe_rsc_params(rsc, node, data_set); get_meta_attributes(rsc->meta, rsc, node, data_set); rsc_provides = g_hash_table_lookup(rsc->meta, PCMK_STONITH_PROVIDES); g_hash_table_iter_init(&gIter, rsc_params); while (g_hash_table_iter_next(&gIter, (gpointer *) & name, (gpointer *) & value)) { if (!name || !value) { continue; } params = stonith_key_value_add(params, name, value); crm_trace(" %s=%s", name, value); } data = create_device_registration_xml(rsc_name(rsc), st_namespace_any, agent, params, rsc_provides); stonith_key_value_freeall(params, 1, 1); rc = stonith_device_register(data, TRUE); CRM_ASSERT(rc == pcmk_ok); free_xml(data); } } /*! * \internal * \brief Update all STONITH device definitions based on current CIB */ static void cib_devices_update(void) { GHashTableIter iter; stonith_device_t *device = NULL; crm_info("Updating devices to version %s.%s.%s", crm_element_value(local_cib, XML_ATTR_GENERATION_ADMIN), crm_element_value(local_cib, XML_ATTR_GENERATION), crm_element_value(local_cib, XML_ATTR_NUMUPDATES)); if (fenced_data_set->now != NULL) { crm_time_free(fenced_data_set->now); fenced_data_set->now = NULL; } fenced_data_set->localhost = stonith_our_uname; pcmk__schedule_actions(local_cib, data_set_flags, fenced_data_set); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->cib_registered) { device->dirty = TRUE; } } /* have list repopulated if cib has a watchdog-fencing-resource TODO: keep a cached list for queries happening while we are refreshing */ g_list_free_full(stonith_watchdog_targets, free); stonith_watchdog_targets = NULL; g_list_foreach(fenced_data_set->resources, (GFunc) cib_device_update, fenced_data_set); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->dirty) { g_hash_table_iter_remove(&iter); } } fenced_data_set->input = NULL; // Wasn't a copy, so don't let API free it pe_reset_working_set(fenced_data_set); } static void update_cib_stonith_devices_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; char *reason = NULL; bool needs_update = FALSE; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); const char *shortpath = NULL; if ((op == NULL) || (strcmp(op, "move") == 0) || strstr(xpath, "/"XML_CIB_TAG_STATUS)) { continue; } else if (pcmk__str_eq(op, "delete", pcmk__str_casei) && strstr(xpath, "/"XML_CIB_TAG_RESOURCE)) { const char *rsc_id = NULL; char *search = NULL; char *mutable = NULL; if (strstr(xpath, XML_TAG_ATTR_SETS) || strstr(xpath, XML_TAG_META_SETS)) { needs_update = TRUE; pcmk__str_update(&reason, "(meta) attribute deleted from resource"); break; } pcmk__str_update(&mutable, xpath); rsc_id = strstr(mutable, "primitive[@" XML_ATTR_ID "=\'"); if (rsc_id != NULL) { rsc_id += strlen("primitive[@" XML_ATTR_ID "=\'"); search = strchr(rsc_id, '\''); } if (search != NULL) { *search = 0; stonith_device_remove(rsc_id, true); /* watchdog_device_update called afterwards to fall back to implicit definition if needed */ } else { crm_warn("Ignoring malformed CIB update (resource deletion)"); } free(mutable); } else if (strstr(xpath, "/"XML_CIB_TAG_RESOURCES) || strstr(xpath, "/"XML_CIB_TAG_CONSTRAINTS) || strstr(xpath, "/"XML_CIB_TAG_RSCCONFIG)) { shortpath = strrchr(xpath, '/'); CRM_ASSERT(shortpath); reason = crm_strdup_printf("%s %s", op, shortpath+1); needs_update = TRUE; break; } } if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } else { crm_trace("No updates for device list found in CIB"); } free(reason); } static void update_cib_stonith_devices_v1(const char *event, xmlNode * msg) { const char *reason = "none"; gboolean needs_update = FALSE; xmlXPathObjectPtr xpath_obj = NULL; /* process new constraints */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_CONS_TAG_RSC_LOCATION); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; /* Safest and simplest to always recompute */ needs_update = TRUE; reason = "new location constraint"; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpath_obj, lpc); crm_log_xml_trace(match, "new constraint"); } } freeXpathObject(xpath_obj); /* process deletions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { remove_cib_device(xpath_obj); } freeXpathObject(xpath_obj); /* process additions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpath_obj, lpc); rsc_id = crm_element_value(match, XML_ATTR_ID); standard = crm_element_value(match, XML_AGENT_ATTR_CLASS); if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } crm_trace("Fencing resource %s was added or modified", rsc_id); reason = "new resource"; needs_update = TRUE; } } freeXpathObject(xpath_obj); if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } } static void update_cib_stonith_devices(const char *event, xmlNode * msg) { int format = 1; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); switch(format) { case 1: update_cib_stonith_devices_v1(event, msg); break; case 2: update_cib_stonith_devices_v2(event, msg); break; default: crm_warn("Unknown patch format: %d", format); } } /*! * \internal * \brief Check whether a node has a specific attribute name/value * * \param[in] node Name of node to check * \param[in] name Name of an attribute to look for * \param[in] value The value the named attribute needs to be set to in order to be considered a match * * \return TRUE if the locally cached CIB has the specified node attribute */ gboolean node_has_attr(const char *node, const char *name, const char *value) { GString *xpath = NULL; xmlNode *match; CRM_CHECK((local_cib != NULL) && (node != NULL) && (name != NULL) && (value != NULL), return FALSE); /* Search for the node's attributes in the CIB. While the schema allows * multiple sets of instance attributes, and allows instance attributes to * use id-ref to reference values elsewhere, that is intended for resources, * so we ignore that here. */ xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, "//" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE "[@" XML_ATTR_UNAME "='", node, "']/" XML_TAG_ATTR_SETS "/" XML_CIB_TAG_NVPAIR "[@" XML_NVPAIR_ATTR_NAME "='", name, "' " "and @" XML_NVPAIR_ATTR_VALUE "='", value, "']", NULL); match = get_xpath_object((const char *) xpath->str, local_cib, LOG_NEVER); g_string_free(xpath, TRUE); return (match != NULL); } /*! * \internal * \brief Check whether a node does watchdog-fencing * * \param[in] node Name of node to check * * \return TRUE if node found in stonith_watchdog_targets * or stonith_watchdog_targets is empty indicating * all nodes are doing watchdog-fencing */ gboolean node_does_watchdog_fencing(const char *node) { return ((stonith_watchdog_targets == NULL) || pcmk__str_in_list(node, stonith_watchdog_targets, pcmk__str_casei)); } static void update_fencing_topology(const char *event, xmlNode * msg) { int format = 1; const char *xpath; xmlXPathObjectPtr xpathObj = NULL; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, "format", &format); if(format == 1) { /* Process deletions (only) */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); remove_fencing_topology(xpathObj); freeXpathObject(xpathObj); /* Process additions and changes */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } else if(format == 2) { xmlNode *change = NULL; int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; xml_patch_versions(patchset, add, del); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); if(op == NULL) { continue; } else if(strstr(xpath, "/" XML_TAG_FENCING_LEVEL) != NULL) { /* Change to a specific entry */ crm_trace("Handling %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "create") == 0) { add_topology_level(change->children); } else if(strcmp(op, "modify") == 0) { xmlNode *match = first_named_child(change, XML_DIFF_RESULT); if(match) { remove_topology_level(match->children); add_topology_level(match->children); } } else if(strcmp(op, "delete") == 0) { /* Nuclear option, all we have is the path and an id... not enough to remove a specific entry */ crm_info("Re-initializing fencing topology after %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else if (strstr(xpath, "/" XML_TAG_FENCING_TOPOLOGY) != NULL) { /* Change to the topology in general */ crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } else if (strstr(xpath, "/" XML_CIB_TAG_CONFIGURATION)) { /* Changes to the whole config section, possibly including the topology as a whild */ if(first_named_child(change, XML_TAG_FENCING_TOPOLOGY) == NULL) { crm_trace("Nothing for us in %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); } else if(strcmp(op, "delete") == 0 || strcmp(op, "create") == 0) { crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else { crm_trace("Nothing for us in %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); } } } else { crm_warn("Unknown patch format: %d", format); } } static bool have_cib_devices = FALSE; static void update_cib_cache_cb(const char *event, xmlNode * msg) { int rc = pcmk_ok; long timeout_ms_saved = stonith_watchdog_timeout_ms; bool need_full_refresh = false; if(!have_cib_devices) { crm_trace("Skipping updates until we get a full dump"); return; } else if(msg == NULL) { crm_trace("Missing %s update", event); return; } /* Maintain a local copy of the CIB so that we have full access * to device definitions, location constraints, and node attributes */ if (local_cib != NULL) { int rc = pcmk_ok; xmlNode *patchset = NULL; crm_element_value_int(msg, F_CIB_RC, &rc); if (rc != pcmk_ok) { return; } patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); pcmk__output_set_log_level(logger_out, LOG_TRACE); out->message(out, "xml-patchset", patchset); rc = xml_apply_patchset(local_cib, patchset, TRUE); switch (rc) { case pcmk_ok: case -pcmk_err_old_data: break; case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; break; default: crm_warn("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; } } if (local_cib == NULL) { crm_trace("Re-requesting full CIB"); rc = cib_api->cmds->query(cib_api, NULL, &local_cib, cib_scope_local | cib_sync_call); if(rc != pcmk_ok) { crm_err("Couldn't retrieve the CIB: %s (%d)", pcmk_strerror(rc), rc); return; } CRM_ASSERT(local_cib != NULL); need_full_refresh = true; } pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); if (timeout_ms_saved != stonith_watchdog_timeout_ms) { need_full_refresh = true; } if (need_full_refresh) { fencing_topology_init(); cib_devices_update(); } else { // Partial refresh update_fencing_topology(event, msg); update_cib_stonith_devices(event, msg); } watchdog_device_update(); } static void init_cib_cache_cb(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { crm_info("Updating device list from CIB"); have_cib_devices = TRUE; local_cib = copy_xml(output); pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); fencing_topology_init(); cib_devices_update(); watchdog_device_update(); } static void stonith_shutdown(int nsig) { crm_info("Terminating with %d clients", pcmk__ipc_client_count()); stonith_shutdown_flag = TRUE; if (mainloop != NULL && g_main_loop_is_running(mainloop)) { g_main_loop_quit(mainloop); } } static void cib_connection_destroy(gpointer user_data) { if (stonith_shutdown_flag) { crm_info("Connection to the CIB manager closed"); return; } else { crm_crit("Lost connection to the CIB manager, shutting down"); } if (cib_api) { cib_api->cmds->signoff(cib_api); } stonith_shutdown(0); } static void stonith_cleanup(void) { if (cib_api) { cib_api->cmds->del_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb); cib_api->cmds->signoff(cib_api); } if (ipcs) { qb_ipcs_destroy(ipcs); } crm_peer_destroy(); pcmk__client_cleanup(); free_stonith_remote_op_list(); free_topology_list(); free_device_list(); free_metadata_cache(); fenced_unregister_handlers(); free(stonith_our_uname); stonith_our_uname = NULL; free_xml(local_cib); local_cib = NULL; } static gboolean stand_alone_cpg_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { stand_alone = FALSE; options.no_cib_connect = true; return TRUE; } static void setup_cib(void) { int rc, retries = 0; cib_api = cib_new(); if (cib_api == NULL) { crm_err("No connection to the CIB manager"); return; } do { sleep(retries); rc = cib_api->cmds->signon(cib_api, CRM_SYSTEM_STONITHD, cib_command); } while (rc == -ENOTCONN && ++retries < 5); if (rc != pcmk_ok) { crm_err("Could not connect to the CIB manager: %s (%d)", pcmk_strerror(rc), rc); } else if (pcmk_ok != cib_api->cmds->add_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb)) { crm_err("Could not set CIB notification callback"); } else { rc = cib_api->cmds->query(cib_api, NULL, NULL, cib_scope_local); cib_api->cmds->register_callback(cib_api, rc, 120, FALSE, NULL, "init_cib_cache_cb", init_cib_cache_cb); cib_api->cmds->set_connection_dnotify(cib_api, cib_connection_destroy); crm_info("Watching for fencing topology changes"); } } struct qb_ipcs_service_handlers ipc_callbacks = { .connection_accept = st_ipc_accept, .connection_created = NULL, .msg_process = st_ipc_dispatch, .connection_closed = st_ipc_closed, .connection_destroyed = st_ipc_destroy }; /*! * \internal * \brief Callback for peer status changes * * \param[in] type What changed * \param[in] node What peer had the change * \param[in] data Previous value of what changed */ static void st_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data) { if ((type != crm_status_processes) && !pcmk_is_set(node->flags, crm_remote_node)) { /* * This is a hack until we can send to a nodeid and/or we fix node name lookups * These messages are ignored in stonith_peer_callback() */ xmlNode *query = create_xml_node(NULL, "stonith_command"); crm_xml_add(query, F_XML_TAGNAME, "stonith_command"); crm_xml_add(query, F_TYPE, T_STONITH_NG); crm_xml_add(query, F_STONITH_OPERATION, "poke"); crm_debug("Broadcasting our uname because of node %u", node->id); send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); } } static pcmk__cluster_option_t fencer_options[] = { /* name, old name, type, allowed values, * default value, validator, * short description, * long description */ { PCMK_STONITH_HOST_ARGUMENT, NULL, "string", NULL, "port", NULL, N_("Advanced use only: An alternate parameter to supply instead of 'port'"), N_("some devices do not support the " "standard 'port' parameter or may provide additional ones. Use " "this to specify an alternate, device-specific, parameter " "that should indicate the machine to be fenced. A value of " "none can be used to tell the cluster not to supply any " "additional parameters.") }, { PCMK_STONITH_HOST_MAP,NULL, "string", NULL, "", NULL, N_("A mapping of host names to ports numbers for devices that do not support host names."), N_("Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and 3 for node2") }, { PCMK_STONITH_HOST_LIST,NULL, "string", NULL, "", NULL, N_("Eg. node1,node2,node3"), N_("A list of machines controlled by " "this device (Optional unless pcmk_host_list=static-list)") }, { PCMK_STONITH_HOST_CHECK,NULL, "string", NULL, "dynamic-list", NULL, N_("How to determine which machines are controlled by the device."), N_("Allowed values: dynamic-list " "(query the device via the 'list' command), static-list " "(check the pcmk_host_list attribute), status " "(query the device via the 'status' command), " "none (assume every device can fence every " "machine)") }, { PCMK_STONITH_DELAY_MAX,NULL, "time", NULL, "0s", NULL, N_("Enable a base delay for fencing actions and specify base delay value."), N_("Enable a delay of no more than the " "time specified before executing fencing actions. Pacemaker " "derives the overall delay by taking the value of " "pcmk_delay_base and adding a random delay value such " "that the sum is kept below this maximum.") }, { PCMK_STONITH_DELAY_BASE,NULL, "string", NULL, "0s", NULL, N_("Enable a base delay for " "fencing actions and specify base delay value."), N_("This enables a static delay for " "fencing actions, which can help avoid \"death matches\" where " "two nodes try to fence each other at the same time. If " "pcmk_delay_max is also used, a random delay will be " "added such that the total delay is kept below that value." "This can be set to a single time value to apply to any node " "targeted by this device (useful if a separate device is " "configured for each target), or to a node map (for example, " "\"node1:1s;node2:5\") to set a different value per target.") }, { PCMK_STONITH_ACTION_LIMIT,NULL, "integer", NULL, "1", NULL, N_("The maximum number of actions can be performed in parallel on this device"), N_("Cluster property concurrent-fencing=true needs to be configured first." "Then use this to specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.") }, { "pcmk_reboot_action",NULL, "string", NULL, "reboot", NULL, N_("Advanced use only: An alternate command to run instead of 'reboot'"), N_("Some devices do not support the standard commands or may provide additional ones.\n" "Use this to specify an alternate, device-specific, command that implements the \'reboot\' action.") }, { "pcmk_reboot_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for reboot actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'reboot\' actions.") }, { "pcmk_reboot_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'reboot' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'reboot\' actions before giving up.") }, { "pcmk_off_action",NULL, "string", NULL, "off", NULL, N_("Advanced use only: An alternate command to run instead of \'off\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'off\' action.") }, { "pcmk_off_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for off actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'off\' actions.") }, { "pcmk_off_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'off' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'off\' actions before giving up.") }, { "pcmk_on_action",NULL, "string", NULL, "on", NULL, N_("Advanced use only: An alternate command to run instead of 'on'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'on\' action.") }, { "pcmk_on_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for on actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'on\' actions.") }, { "pcmk_on_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the 'on' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'on\' actions before giving up.") }, { "pcmk_list_action",NULL, "string", NULL, "list", NULL, N_("Advanced use only: An alternate command to run instead of \'list\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'list\' action.") }, { "pcmk_list_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for list actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'list\' actions.") }, { "pcmk_list_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'list\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'list\' actions before giving up.") }, { "pcmk_monitor_action",NULL, "string", NULL, "monitor", NULL, N_("Advanced use only: An alternate command to run instead of \'monitor\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'monitor\' action.") }, { "pcmk_monitor_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for monitor actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal.\n" "Use this to specify an alternate, device-specific, timeout for \'monitor\' actions.") }, { "pcmk_monitor_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'monitor\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'monitor\' actions before giving up.") }, { "pcmk_status_action",NULL, "string", NULL, "status", NULL, N_("Advanced use only: An alternate command to run instead of \'status\'"), N_("Some devices do not support the standard commands or may provide additional ones." "Use this to specify an alternate, device-specific, command that implements the \'status\' action.") }, { "pcmk_status_timeout",NULL, "time", NULL, "60s", NULL, N_("Advanced use only: Specify an alternate timeout to use for status actions instead of stonith-timeout"), N_("Some devices need much more/less time to complete than normal." "Use this to specify an alternate, device-specific, timeout for \'status\' actions.") }, { "pcmk_status_retries",NULL, "integer", NULL, "2", NULL, N_("Advanced use only: The maximum number of times to retry the \'status\' command within the timeout period"), N_("Some devices do not support multiple connections." " Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining." " Use this option to alter the number of times Pacemaker retries \'status\' actions before giving up.") }, }; void fencer_metadata(void) { const char *desc_short = N_("Instance attributes available for all " "\"stonith\"-class resources"); const char *desc_long = N_("Instance attributes available for all \"stonith\"-" "class resources and used by Pacemaker's fence " "daemon, formerly known as stonithd"); gchar *s = pcmk__format_option_metadata("pacemaker-fenced", desc_short, desc_long, fencer_options, PCMK__NELEM(fencer_options)); printf("%s", s); g_free(s); } static GOptionEntry entries[] = { { "stand-alone", 's', G_OPTION_FLAG_NONE, G_OPTION_ARG_NONE, &stand_alone, N_("Deprecated (will be removed in a future release)"), NULL }, { "stand-alone-w-cpg", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, stand_alone_cpg_cb, N_("Intended for use in regression testing only"), NULL }, { "logfile", 'l', G_OPTION_FLAG_NONE, G_OPTION_ARG_FILENAME_ARRAY, &options.log_files, N_("Send logs to the additional named logfile"), NULL }, { NULL } }; static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; context = pcmk__build_arg_context(args, "text (default), xml", group, "[metadata]"); pcmk__add_main_args(context, entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; crm_cluster_t *cluster = NULL; crm_ipc_t *old_instance = NULL; GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "l"); GOptionContext *context = build_arg_context(args, &output_group); crm_log_preinit(NULL, argc, argv); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { exit_code = CRM_EX_ERROR; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); goto done; } if (args->version) { out->version(out, false); goto done; } if ((g_strv_length(processed_args) >= 2) && pcmk__str_eq(processed_args[1], "metadata", pcmk__str_none)) { fencer_metadata(); goto done; } // Open additional log files pcmk__add_logfiles(options.log_files, out); crm_log_init(NULL, LOG_INFO + args->verbosity, TRUE, (args->verbosity > 0), argc, argv, FALSE); crm_notice("Starting Pacemaker fencer"); old_instance = crm_ipc_new("stonith-ng", 0); if (old_instance == NULL) { /* crm_ipc_new() will have already logged an error message with * crm_err() */ exit_code = CRM_EX_FATAL; goto done; } if (crm_ipc_connect(old_instance)) { // IPC endpoint already up crm_ipc_close(old_instance); crm_ipc_destroy(old_instance); crm_err("pacemaker-fenced is already active, aborting startup"); goto done; } else { // Not up or not authentic, we'll proceed either way crm_ipc_destroy(old_instance); old_instance = NULL; } mainloop_add_signal(SIGTERM, stonith_shutdown); crm_peer_init(); fenced_data_set = pe_new_working_set(); CRM_ASSERT(fenced_data_set != NULL); cluster = pcmk_cluster_new(); /* Initialize the logger prior to setup_cib(). update_cib_cache_cb() may * call the "xml-patchset" message function, which needs the logger, after * setup_cib() has run. */ rc = pcmk__log_output_new(&logger_out) != pcmk_rc_ok; if (rc != pcmk_rc_ok) { exit_code = CRM_EX_FATAL; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format log: %s", pcmk_rc_str(rc)); goto done; } pe__register_messages(logger_out); pcmk__register_lib_messages(logger_out); pcmk__output_set_log_level(logger_out, LOG_TRACE); fenced_data_set->priv = logger_out; if (!stand_alone) { #if SUPPORT_COROSYNC if (is_corosync_cluster()) { cluster->destroy = stonith_peer_cs_destroy; cluster->cpg.cpg_deliver_fn = stonith_peer_ais_callback; cluster->cpg.cpg_confchg_fn = pcmk_cpg_membership; } #endif // SUPPORT_COROSYNC crm_set_status_callback(&st_peer_update_callback); if (crm_cluster_connect(cluster) == FALSE) { exit_code = CRM_EX_FATAL; crm_crit("Cannot sign in to the cluster... terminating"); goto done; } pcmk__str_update(&stonith_our_uname, cluster->uname); if (!options.no_cib_connect) { setup_cib(); } } else { pcmk__str_update(&stonith_our_uname, "localhost"); crm_warn("Stand-alone mode is deprecated and will be removed " "in a future release"); } init_device_list(); init_topology_list(); pcmk__serve_fenced_ipc(&ipcs, &ipc_callbacks); // Create the mainloop and run it... mainloop = g_main_loop_new(NULL, FALSE); crm_notice("Pacemaker fencer successfully started and accepting connections"); g_main_loop_run(mainloop); done: g_strfreev(processed_args); pcmk__free_arg_context(context); g_strfreev(options.log_files); stonith_cleanup(); pcmk_cluster_free(cluster); pe_free_working_set(fenced_data_set); pcmk__output_and_clear_error(&error, out); if (logger_out != NULL) { logger_out->finish(logger_out, exit_code, true, NULL); pcmk__output_free(logger_out); } if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } diff --git a/doc/sphinx/Pacemaker_Development/components.rst b/doc/sphinx/Pacemaker_Development/components.rst index e14df26ad6..d4f24fa18f 100644 --- a/doc/sphinx/Pacemaker_Development/components.rst +++ b/doc/sphinx/Pacemaker_Development/components.rst @@ -1,489 +1,489 @@ Coding Particular Pacemaker Components -------------------------------------- The Pacemaker code can be intricate and difficult to follow. This chapter has some high-level descriptions of how individual components work. .. index:: single: controller single: pacemaker-controld Controller ########## ``pacemaker-controld`` is the Pacemaker daemon that utilizes the other daemons to orchestrate actions that need to be taken in the cluster. It receives CIB change notifications from the CIB manager, passes the new CIB to the scheduler to determine whether anything needs to be done, uses the executor and fencer to execute any actions required, and sets failure counts (among other things) via the attribute manager. As might be expected, it has the most code of any of the daemons. .. index:: single: join Join sequence _____________ Most daemons track their cluster peers using Corosync's membership and CPG only. The controller additionally requires peers to `join`, which ensures they are ready to be assigned tasks. Joining proceeds through a series of phases referred to as the `join sequence` or `join process`. A node's current join phase is tracked by the ``join`` member of ``crm_node_t`` (used in the peer cache). It is an ``enum crm_join_phase`` that (ideally) progresses from the DC's point of view as follows: * The node initially starts at ``crm_join_none`` * The DC sends the node a `join offer` (``CRM_OP_JOIN_OFFER``), and the node proceeds to ``crm_join_welcomed``. This can happen in three ways: * The joining node will send a `join announce` (``CRM_OP_JOIN_ANNOUNCE``) at its controller startup, and the DC will reply to that with a join offer. * When the DC's peer status callback notices that the node has joined the messaging layer, it registers ``I_NODE_JOIN`` (which leads to ``A_DC_JOIN_OFFER_ONE`` -> ``do_dc_join_offer_one()`` -> ``join_make_offer()``). * After certain events (notably a new DC being elected), the DC will send all nodes join offers (via A_DC_JOIN_OFFER_ALL -> ``do_dc_join_offer_all()``). These can overlap. The DC can send a join offer and the node can send a join announce at nearly the same time, so the node responds to the original join offer while the DC responds to the join announce with a new join offer. The situation resolves itself after looping a bit. * The node responds to join offers with a `join request` (``CRM_OP_JOIN_REQUEST``, via ``do_cl_join_offer_respond()`` and ``join_query_callback()``). When the DC receives the request, the node proceeds to ``crm_join_integrated`` (via ``do_dc_join_filter_offer()``). * As each node is integrated, the current best CIB is sync'ed to each integrated node via ``do_dc_join_finalize()``. As each integrated node's CIB sync succeeds, the DC acks the node's join request (``CRM_OP_JOIN_ACKNAK``) and the node proceeds to ``crm_join_finalized`` (via ``finalize_sync_callback()`` + ``finalize_join_for()``). * Each node confirms the finalization ack (``CRM_OP_JOIN_CONFIRM`` via ``do_cl_join_finalize_respond()``), including its current resource operation history (via ``controld_query_executor_state()``). Once the DC receives this confirmation, the node proceeds to ``crm_join_confirmed`` via ``do_dc_join_ack()``. Once all nodes are confirmed, the DC calls ``do_dc_join_final()``, which checks for quorum and responds appropriately. When peers are lost, their join phase is reset to none (in various places). ``crm_update_peer_join()`` updates a node's join phase. The DC increments the global ``current_join_id`` for each joining round, and rejects any (older) replies that don't match. .. index:: single: fencer single: pacemaker-fenced Fencer ###### ``pacemaker-fenced`` is the Pacemaker daemon that handles fencing requests. In the broadest terms, fencing works like this: #. The initiator (an external program such as ``stonith_admin``, or the cluster itself via the controller) asks the local fencer, "Hey, could you please fence this node?" #. The local fencer asks all the fencers in the cluster (including itself), "Hey, what fencing devices do you have access to that can fence this node?" #. Each fencer in the cluster replies with a list of available devices that it knows about. #. Once the original fencer gets all the replies, it asks the most appropriate fencer peer to actually carry out the fencing. It may send out more than one such request if the target node must be fenced with multiple devices. #. The chosen fencer(s) call the appropriate fencing resource agent(s) to do the fencing, then reply to the original fencer with the result. #. The original fencer broadcasts the result to all fencers. #. Each fencer sends the result to each of its local clients (including, at some point, the initiator). A more detailed description follows. .. index:: single: libstonithd Initiating a fencing request ____________________________ A fencing request can be initiated by the cluster or externally, using the libstonithd API. * The cluster always initiates fencing via ``daemons/controld/controld_fencing.c:te_fence_node()`` (which calls the ``fence()`` API method). This occurs when a transition graph synapse contains a ``CRM_OP_FENCE`` XML operation. * The main external clients are ``stonith_admin`` and ``cts-fence-helper``. The ``DLM`` project also uses Pacemaker for fencing. Highlights of the fencing API: * ``stonith_api_new()`` creates and returns a new ``stonith_t`` object, whose ``cmds`` member has methods for connect, disconnect, fence, etc. * the ``fence()`` method creates and sends a ``STONITH_OP_FENCE XML`` request with the desired action and target node. Callers do not have to choose or even have any knowledge about particular fencing devices. Fencing queries _______________ The function calls for a fencing request go something like this: The local fencer receives the client's request via an IPC or messaging layer callback, which calls * ``stonith_command()``, which (for requests) calls * ``handle_request()``, which (for ``STONITH_OP_FENCE`` from a client) calls * ``initiate_remote_stonith_op()``, which creates a ``STONITH_OP_QUERY`` XML request with the target, desired action, timeout, etc. then broadcasts the operation to the cluster group (i.e. all fencer instances) and starts a timer. The query is broadcast because (1) location constraints might prevent the local node from accessing the stonith device directly, and (2) even if the local node does have direct access, another node might be preferred to carry out the fencing. Each fencer receives the original fencer's ``STONITH_OP_QUERY`` broadcast request via IPC or messaging layer callback, which calls: * ``stonith_command()``, which (for requests) calls * ``handle_request()``, which (for ``STONITH_OP_QUERY`` from a peer) calls * ``stonith_query()``, which calls * ``get_capable_devices()`` with ``stonith_query_capable_device_cb()`` to add device information to an XML reply and send it. (A message is considered a reply if it contains ``T_STONITH_REPLY``, which is only set by fencer peers, not clients.) The original fencer receives all peers' ``STONITH_OP_QUERY`` replies via IPC or messaging layer callback, which calls: * ``stonith_command()``, which (for replies) calls * ``handle_reply()`` which (for ``STONITH_OP_QUERY``) calls * ``process_remote_stonith_query()``, which allocates a new query result structure, parses device information into it, and adds it to the operation object. It increments the number of replies received for this operation, and compares it against the expected number of replies (i.e. the number of active peers), and if this is the last expected reply, calls * ``request_peer_fencing()``, which calculates the timeout and sends ``STONITH_OP_FENCE`` request(s) to carry out the fencing. If the target node has a fencing "topology" (which allows specifications such as "this node can be fenced either with device A, or devices B and C in combination"), it will choose the device(s), and send out as many requests as needed. If it chooses a device, it will choose the peer; a peer is preferred if it has "verified" access to the desired device, meaning that it has the device "running" on it and thus has a monitor operation ensuring reachability. Fencing operations __________________ Each ``STONITH_OP_FENCE`` request goes something like this: The chosen peer fencer receives the ``STONITH_OP_FENCE`` request via IPC or messaging layer callback, which calls: * ``stonith_command()``, which (for requests) calls * ``handle_request()``, which (for ``STONITH_OP_FENCE`` from a peer) calls * ``stonith_fence()``, which calls * ``schedule_stonith_command()`` (using supplied device if ``F_STONITH_DEVICE`` was set, otherwise the highest-priority capable device obtained via ``get_capable_devices()`` with ``stonith_fence_get_devices_cb()``), which adds the operation to the device's pending operations list and triggers processing. The chosen peer fencer's mainloop is triggered and calls * ``stonith_device_dispatch()``, which calls * ``stonith_device_execute()``, which pops off the next item from the device's pending operations list. If acting as the (internally implemented) watchdog agent, it panics the node, otherwise it calls * ``stonith_action_create()`` and ``stonith_action_execute_async()`` to call the fencing agent. The chosen peer fencer's mainloop is triggered again once the fencing agent returns, and calls * ``stonith_action_async_done()`` which adds the results to an action object then calls its * done callback (``st_child_done()``), which calls ``schedule_stonith_command()`` for a new device if there are further required actions to execute or if the original action failed, then builds and sends an XML reply to the original fencer (via ``send_async_reply()``), then checks whether any pending actions are the same as the one just executed and merges them if so. Fencing replies _______________ The original fencer receives the ``STONITH_OP_FENCE`` reply via IPC or messaging layer callback, which calls: * ``stonith_command()``, which (for replies) calls * ``handle_reply()``, which calls * ``fenced_process_fencing_reply()``, which calls either ``request_peer_fencing()`` (to retry a failed operation, or try the next device in a topology if appropriate, which issues a new ``STONITH_OP_FENCE`` request, proceeding as before) or ``finalize_op()`` (if the operation is definitively failed or successful). * ``finalize_op()`` broadcasts the result to all peers. Finally, all peers receive the broadcast result and call * ``finalize_op()``, which sends the result to all local clients. .. index:: single: fence history Fencing History _______________ The fencer keeps a running history of all fencing operations. The bulk of the relevant code is in `fenced_history.c` and ensures the history is synchronized across all nodes even if a node leaves and rejoins the cluster. In libstonithd, this information is represented by `stonith_history_t` and is queryable by the `stonith_api_operations_t:history()` method. `crm_mon` and `stonith_admin` use this API to display the history. .. index:: single: scheduler single: pacemaker-schedulerd single: libpe_status single: libpe_rules single: libpacemaker Scheduler ######### ``pacemaker-schedulerd`` is the Pacemaker daemon that runs the Pacemaker scheduler for the controller, but "the scheduler" in general refers to related library code in ``libpe_status`` and ``libpe_rules`` (``lib/pengine/*.c``), and some of ``libpacemaker`` (``lib/pacemaker/pcmk_sched_*.c``). The purpose of the scheduler is to take a CIB as input and generate a transition graph (list of actions that need to be taken) as output. The controller invokes the scheduler by contacting the scheduler daemon via local IPC. Tools such as ``crm_simulate``, ``crm_mon``, and ``crm_resource`` can also invoke the scheduler, but do so by calling the library functions directly. This allows them to run using a ``CIB_file`` without the cluster needing to be active. The main entry point for the scheduler code is -``lib/pacemaker/pcmk_sched_allocate.c:pcmk__schedule_actions()``. It sets +``lib/pacemaker/pcmk_scheduler.c:pcmk__schedule_actions()``. It sets defaults and calls a series of functions for the scheduling. Some key steps: * ``unpack_cib()`` parses most of the CIB XML into data structures, and determines the current cluster status. * ``apply_node_criteria()`` applies factors that make resources prefer certain nodes, such as shutdown locks, location constraints, and stickiness. * ``pcmk__create_internal_constraints()`` creates internal constraints, such as the implicit ordering for group members, or start actions being implicitly ordered before promote actions. * ``pcmk__handle_rsc_config_changes()`` processes resource history entries in the CIB status section. This is used to decide whether certain actions need to be done, such as deleting orphan resources, forcing a restart when a resource definition changes, etc. -* ``allocate_resources()`` assigns resources to nodes. +* ``assign_resources()`` assigns resources to nodes. * ``schedule_resource_actions()`` schedules resource-specific actions (which might or might not end up in the final graph). * ``pcmk__apply_orderings()`` processes ordering constraints in order to modify action attributes such as optional or required. * ``pcmk__create_graph()`` creates the transition graph. Challenges __________ Working with the scheduler is difficult. Challenges include: * It is far too much code to keep more than a small portion in your head at one time. * Small changes can have large (and unexpected) effects. This is why we have a large number of regression tests (``cts/cts-scheduler``), which should be run after making code changes. * It produces an insane amount of log messages at debug and trace levels. You can put resource ID(s) in the ``PCMK_trace_tags`` environment variable to enable trace-level messages only when related to specific resources. * Different parts of the main ``pe_working_set_t`` structure are finalized at different points in the scheduling process, so you have to keep in mind whether information you're using at one point of the code can possibly change later. For example, data unpacked from the CIB can safely be used anytime after ``unpack_cib(),`` but actions may become optional or required anytime before ``pcmk__create_graph()``. There's no easy way to deal with this. * Many names of struct members, functions, etc., are suboptimal, but are part of the public API and cannot be changed until an API backward compatibility break. .. index:: single: pe_working_set_t Cluster Working Set ___________________ The main data object for the scheduler is ``pe_working_set_t``, which contains all information needed about nodes, resources, constraints, etc., both as the raw CIB XML and parsed into more usable data structures, plus the resulting transition graph XML. The variable name is usually ``data_set``. .. index:: single: pe_resource_t Resources _________ ``pe_resource_t`` is the data object representing cluster resources. A resource has a variant: primitive (a.k.a. native), group, clone, or bundle. The resource object has members for two sets of methods, ``resource_object_functions_t`` from the ``libpe_status`` public API, and ``resource_alloc_functions_t`` whose implementation is internal to ``libpacemaker``. The actual functions vary by variant. The object functions have basic capabilities such as unpacking the resource XML, and determining the current or planned location of the resource. -The allocation functions have more obscure capabilities needed for scheduling, +The assignment functions have more obscure capabilities needed for scheduling, such as processing location and ordering constraints. For example, ``pcmk__create_internal_constraints()`` simply calls the ``internal_constraints()`` method for each top-level resource in the cluster. .. index:: single: pe_node_t Nodes _____ -Allocation of resources to nodes is done by choosing the node with the highest +Assignment of resources to nodes is done by choosing the node with the highest score for a given resource. The scheduler does a bunch of processing to -generate the scores, then the actual allocation is straightforward. +generate the scores, then the actual assignment is straightforward. Node lists are frequently used. For example, ``pe_working_set_t`` has a ``nodes`` member which is a list of all nodes in the cluster, and ``pe_resource_t`` has a ``running_on`` member which is a list of all nodes on which the resource is (or might be) active. These are lists of ``pe_node_t`` objects. The ``pe_node_t`` object contains a ``struct pe_node_shared_s *details`` member -with all node information that is independent of resource allocation (the node +with all node information that is independent of resource assignment (the node name, etc.). The working set's ``nodes`` member contains the original of this information. All other node lists contain copies of ``pe_node_t`` where only the ``details`` member points to the originals in the working set's ``nodes`` list. In this way, the other members of ``pe_node_t`` (such as ``weight``, which is the node score) may vary by node list, while the common details are shared. .. index:: single: pe_action_t single: pe_action_flags Actions _______ ``pe_action_t`` is the data object representing actions that might need to be taken. These could be resource actions, cluster-wide actions such as fencing a node, or "pseudo-actions" which are abstractions used as convenient points for ordering other actions against. It has a ``flags`` member which is a bitmask of ``enum pe_action_flags``. The most important of these are ``pe_action_runnable`` (if not set, the action is "blocked" and cannot be added to the transition graph) and ``pe_action_optional`` (actions with this set will not be added to the transition graph; actions often start out as optional, and may become required later). .. index:: single: pe__colocation_t Colocations ___________ ``pcmk__colocation_t`` is the data object representing colocations. Colocation constraints come into play in these parts of the scheduler code: * When sorting resources for assignment, so resources with highest node score are assigned first (see ``cmp_resources()``) * When updating node scores for resource assigment or promotion priority * When assigning resources, so any resources to be colocated with can be assigned first, and so colocations affect where the resource is assigned * When choosing roles for promotable clone instances, so colocations involving a specific role can affect which instances are promoted -The resource allocation functions have several methods related to colocations: +The resource assignment functions have several methods related to colocations: * ``apply_coloc_score():`` This applies a colocation's score to either the dependent's allowed node scores (if called while resources are being assigned) or the dependent's priority (if called while choosing promotable instance roles). It can behave differently depending on whether it is being called as the primary's method or as the dependent's method. * ``add_colocated_node_scores():`` This updates a table of nodes for a given colocation attribute and score. It goes through colocations involving a given resource, and updates the scores of the nodes in the table with the best scores of nodes that match up according to the colocation criteria. * ``colocated_resources():`` This generates a list of all resources involved in mandatory colocations (directly or indirectly via colocation chains) with a given resource. .. index:: single: pe__ordering_t single: pe_ordering Orderings _________ Ordering constraints are simple in concept, but they are one of the most important, powerful, and difficult to follow aspects of the scheduler code. ``pe__ordering_t`` is the data object representing an ordering, better thought of as a relationship between two actions, since the relation can be more complex than just "this one runs after that one". For an ordering "A then B", the code generally refers to A as "first" or "before", and B as "then" or "after". Much of the power comes from ``enum pe_ordering``, which are flags that determine how an ordering behaves. There are many obscure flags with big effects. A few examples: * ``pe_order_none`` means the ordering is disabled and will be ignored. It's 0, meaning no flags set, so it must be compared with equality rather than ``pcmk_is_set()``. * ``pe_order_optional`` means the ordering does not make either action required, so it only applies if they both become required for other reasons. * ``pe_order_implies_first`` means that if action B becomes required for any reason, then action A will become required as well. diff --git a/doc/sphinx/Pacemaker_Explained/advanced-resources.rst b/doc/sphinx/Pacemaker_Explained/advanced-resources.rst index a61b76db2f..07583507a4 100644 --- a/doc/sphinx/Pacemaker_Explained/advanced-resources.rst +++ b/doc/sphinx/Pacemaker_Explained/advanced-resources.rst @@ -1,1629 +1,1629 @@ Advanced Resource Types ----------------------- .. index: single: group resource single: resource; group .. _group-resources: Groups - A Syntactic Shortcut ############################# One of the most common elements of a cluster is a set of resources that need to be located together, start sequentially, and stop in the reverse order. To simplify this configuration, we support the concept of groups. .. topic:: A group of two primitive resources .. code-block:: xml Although the example above contains only two resources, there is no limit to the number of resources a group can contain. The example is also sufficient to explain the fundamental properties of a group: * Resources are started in the order they appear in (**Public-IP** first, then **Email**) * Resources are stopped in the reverse order to which they appear in (**Email** first, then **Public-IP**) If a resource in the group can't run anywhere, then nothing after that is allowed to run, too. * If **Public-IP** can't run anywhere, neither can **Email**; * but if **Email** can't run anywhere, this does not affect **Public-IP** in any way The group above is logically equivalent to writing: .. topic:: How the cluster sees a group resource .. code-block:: xml Obviously as the group grows bigger, the reduced configuration effort can become significant. Another (typical) example of a group is a DRBD volume, the filesystem mount, an IP address, and an application that uses them. .. index:: pair: XML element; group Group Properties ________________ .. table:: **Properties of a Group Resource** :widths: 1 4 +-------------+------------------------------------------------------------------+ | Field | Description | +=============+==================================================================+ | id | .. index:: | | | single: group; property, id | | | single: property; id (group) | | | single: id; group property | | | | | | A unique name for the group | +-------------+------------------------------------------------------------------+ | description | .. index:: | | | single: group; attribute, description | | | single: attribute; description (group) | | | single: description; group attribute | | | | | | An optional description of the group, for the user's own | | | purposes. | | | E.g. ``resources needed for website`` | +-------------+------------------------------------------------------------------+ Group Options _____________ Groups inherit the ``priority``, ``target-role``, and ``is-managed`` properties from primitive resources. See :ref:`resource_options` for information about those properties. Group Instance Attributes _________________________ Groups have no instance attributes. However, any that are set for the group object will be inherited by the group's children. Group Contents ______________ Groups may only contain a collection of cluster resources (see :ref:`primitive-resource`). To refer to a child of a group resource, just use the child's ``id`` instead of the group's. Group Constraints _________________ Although it is possible to reference a group's children in constraints, it is usually preferable to reference the group itself. .. topic:: Some constraints involving groups .. code-block:: xml .. index:: pair: resource-stickiness; group Group Stickiness ________________ Stickiness, the measure of how much a resource wants to stay where it is, is additive in groups. Every active resource of the group will contribute its stickiness value to the group's total. So if the default ``resource-stickiness`` is 100, and a group has seven members, five of which are active, then the group as a whole will prefer its current location with a score of 500. .. index:: single: clone single: resource; clone .. _s-resource-clone: Clones - Resources That Can Have Multiple Active Instances ########################################################## *Clone* resources are resources that can have more than one copy active at the same time. This allows you, for example, to run a copy of a daemon on every node. You can clone any primitive or group resource [#]_. Anonymous versus Unique Clones ______________________________ A clone resource is configured to be either *anonymous* or *globally unique*. Anonymous clones are the simplest. These behave completely identically everywhere they are running. Because of this, there can be only one instance of an anonymous clone active per node. The instances of globally unique clones are distinct entities. All instances are launched identically, but one instance of the clone is not identical to any other instance, whether running on the same node or a different node. As an example, a cloned IP address can use special kernel functionality such that each instance handles a subset of requests for the same IP address. .. index:: single: promotable clone single: resource; promotable .. _s-resource-promotable: Promotable clones _________________ If a clone is *promotable*, its instances can perform a special role that Pacemaker will manage via the ``promote`` and ``demote`` actions of the resource agent. Services that support such a special role have various terms for the special role and the default role: primary and secondary, master and replica, controller and worker, etc. Pacemaker uses the terms *promoted* and *unpromoted* to be agnostic to what the service calls them or what they do. All that Pacemaker cares about is that an instance comes up in the unpromoted role when started, and the resource agent supports the ``promote`` and ``demote`` actions to manage entering and exiting the promoted role. .. index:: pair: XML element; clone Clone Properties ________________ .. table:: **Properties of a Clone Resource** :widths: 1 4 +-------------+------------------------------------------------------------------+ | Field | Description | +=============+==================================================================+ | id | .. index:: | | | single: clone; property, id | | | single: property; id (clone) | | | single: id; clone property | | | | | | A unique name for the clone | +-------------+------------------------------------------------------------------+ | description | .. index:: | | | single: clone; attribute, description | | | single: attribute; description (clone) | | | single: description; clone attribute | | | | | | An optional description of the clone, for the user's own | | | purposes. | | | E.g. ``IP address for website`` | +-------------+------------------------------------------------------------------+ .. index:: pair: options; clone Clone Options _____________ :ref:`Options ` inherited from primitive resources: ``priority, target-role, is-managed`` .. table:: **Clone-specific configuration options** :class: longtable :widths: 1 1 3 +-------------------+-----------------+-------------------------------------------------------+ | Field | Default | Description | +===================+=================+=======================================================+ | globally-unique | false | .. index:: | | | | single: clone; option, globally-unique | | | | single: option; globally-unique (clone) | | | | single: globally-unique; clone option | | | | | | | | If **true**, each clone instance performs a | | | | distinct function | +-------------------+-----------------+-------------------------------------------------------+ | clone-max | 0 | .. index:: | | | | single: clone; option, clone-max | | | | single: option; clone-max (clone) | | | | single: clone-max; clone option | | | | | | | | The maximum number of clone instances that can | | | | be started across the entire cluster. If 0, the | | | | number of nodes in the cluster will be used. | +-------------------+-----------------+-------------------------------------------------------+ | clone-node-max | 1 | .. index:: | | | | single: clone; option, clone-node-max | | | | single: option; clone-node-max (clone) | | | | single: clone-node-max; clone option | | | | | | | | If ``globally-unique`` is **true**, the maximum | | | | number of clone instances that can be started | | | | on a single node | +-------------------+-----------------+-------------------------------------------------------+ | clone-min | 0 | .. index:: | | | | single: clone; option, clone-min | | | | single: option; clone-min (clone) | | | | single: clone-min; clone option | | | | | | | | Require at least this number of clone instances | | | | to be runnable before allowing resources | | | | depending on the clone to be runnable. A value | | | | of 0 means require all clone instances to be | | | | runnable. | +-------------------+-----------------+-------------------------------------------------------+ | notify | false | .. index:: | | | | single: clone; option, notify | | | | single: option; notify (clone) | | | | single: notify; clone option | | | | | | | | Call the resource agent's **notify** action for | | | | all active instances, before and after starting | | | | or stopping any clone instance. The resource | | | | agent must support this action. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | ordered | false | .. index:: | | | | single: clone; option, ordered | | | | single: option; ordered (clone) | | | | single: ordered; clone option | | | | | | | | If **true**, clone instances must be started | | | | sequentially instead of in parallel. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | interleave | false | .. index:: | | | | single: clone; option, interleave | | | | single: option; interleave (clone) | | | | single: interleave; clone option | | | | | | | | When this clone is ordered relative to another | | | | clone, if this option is **false** (the default), | | | | the ordering is relative to *all* instances of | | | | the other clone, whereas if this option is | | | | **true**, the ordering is relative only to | | | | instances on the same node. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | promotable | false | .. index:: | | | | single: clone; option, promotable | | | | single: option; promotable (clone) | | | | single: promotable; clone option | | | | | | | | If **true**, clone instances can perform a | | | | special role that Pacemaker will manage via the | | | | resource agent's **promote** and **demote** | | | | actions. The resource agent must support these | | | | actions. | | | | Allowed values: **false**, **true** | +-------------------+-----------------+-------------------------------------------------------+ | promoted-max | 1 | .. index:: | | | | single: clone; option, promoted-max | | | | single: option; promoted-max (clone) | | | | single: promoted-max; clone option | | | | | | | | If ``promotable`` is **true**, the number of | | | | instances that can be promoted at one time | | | | across the entire cluster | +-------------------+-----------------+-------------------------------------------------------+ | promoted-node-max | 1 | .. index:: | | | | single: clone; option, promoted-node-max | | | | single: option; promoted-node-max (clone) | | | | single: promoted-node-max; clone option | | | | | | | | If ``promotable`` is **true** and ``globally-unique`` | | | | is **false**, the number of clone instances can be | | | | promoted at one time on a single node | +-------------------+-----------------+-------------------------------------------------------+ .. note:: **Deprecated Terminology** In older documentation and online examples, you may see promotable clones referred to as *multi-state*, *stateful*, or *master/slave*; these mean the same thing as *promotable*. Certain syntax is supported for backward compatibility, but is deprecated and will be removed in a future version: * Using a ``master`` tag, instead of a ``clone`` tag with the ``promotable`` meta-attribute set to ``true`` * Using the ``master-max`` meta-attribute instead of ``promoted-max`` * Using the ``master-node-max`` meta-attribute instead of ``promoted-node-max`` * Using ``Master`` as a role name instead of ``Promoted`` * Using ``Slave`` as a role name instead of ``Unpromoted`` Clone Contents ______________ Clones must contain exactly one primitive or group resource. .. topic:: A clone that runs a web server on all nodes .. code-block:: xml .. warning:: You should never reference the name of a clone's child (the primitive or group resource being cloned). If you think you need to do this, you probably need to re-evaluate your design. Clone Instance Attribute ________________________ Clones have no instance attributes; however, any that are set here will be inherited by the clone's child. .. index:: single: clone; constraint Clone Constraints _________________ In most cases, a clone will have a single instance on each active cluster node. If this is not the case, you can indicate which nodes the cluster should preferentially assign copies to with resource location constraints. These constraints are written no differently from those for primitive resources except that the clone's **id** is used. .. topic:: Some constraints involving clones .. code-block:: xml Ordering constraints behave slightly differently for clones. In the example above, ``apache-stats`` will wait until all copies of ``apache-clone`` that need to be started have done so before being started itself. Only if *no* copies can be started will ``apache-stats`` be prevented from being active. Additionally, the clone will wait for ``apache-stats`` to be stopped before stopping itself. Colocation of a primitive or group resource with a clone means that the resource can run on any node with an active instance of the clone. The cluster will choose an instance based on where the clone is running and the resource's own location preferences. Colocation between clones is also possible. If one clone **A** is colocated with another clone **B**, the set of allowed locations for **A** is limited to nodes on which **B** is (or will be) active. Placement is then performed normally. .. index:: single: promotable clone; constraint .. _promotable-clone-constraints: Promotable Clone Constraints ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For promotable clone resources, the ``first-action`` and/or ``then-action`` fields for ordering constraints may be set to ``promote`` or ``demote`` to constrain the promoted role, and colocation constraints may contain ``rsc-role`` and/or ``with-rsc-role`` fields. .. topic:: Constraints involving promotable clone resources .. code-block:: xml In the example above, **myApp** will wait until one of the database copies has been started and promoted before being started itself on the same node. Only if no copies can be promoted will **myApp** be prevented from being active. Additionally, the cluster will wait for **myApp** to be stopped before demoting the database. Colocation of a primitive or group resource with a promotable clone resource means that it can run on any node with an active instance of the promotable clone resource that has the specified role (``Promoted`` or ``Unpromoted``). In the example above, the cluster will choose a location based on where database is running in the promoted role, and if there are multiple promoted instances it will also factor in **myApp**'s own location preferences when deciding which location to choose. Colocation with regular clones and other promotable clone resources is also possible. In such cases, the set of allowed locations for the **rsc** clone is (after role filtering) limited to nodes on which the ``with-rsc`` promotable clone resource is (or will be) in the specified role. Placement is then performed as normal. Using Promotable Clone Resources in Colocation Sets ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ When a promotable clone is used in a :ref:`resource set ` inside a colocation constraint, the resource set may take a ``role`` attribute. In the following example, an instance of **B** may be promoted only on a node where **A** is in the promoted role. Additionally, resources **C** and **D** must be located on a node where both **A** and **B** are promoted. .. topic:: Colocate C and D with A's and B's promoted instances .. code-block:: xml Using Promotable Clone Resources in Ordered Sets ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ When a promotable clone is used in a :ref:`resource set ` inside an ordering constraint, the resource set may take an ``action`` attribute. .. topic:: Start C and D after first promoting A and B .. code-block:: xml In the above example, **B** cannot be promoted until **A** has been promoted. Additionally, resources **C** and **D** must wait until **A** and **B** have been promoted before they can start. .. index:: pair: resource-stickiness; clone .. _s-clone-stickiness: Clone Stickiness ________________ -To achieve a stable allocation pattern, clones are slightly sticky by -default. If no value for ``resource-stickiness`` is provided, the clone -will use a value of 1. Being a small value, it causes minimal -disturbance to the score calculations of other resources but is enough -to prevent Pacemaker from needlessly moving copies around the cluster. +To achieve stable assignments, clones are slightly sticky by default. If no +value for ``resource-stickiness`` is provided, the clone will use a value of 1. +Being a small value, it causes minimal disturbance to the score calculations of +other resources but is enough to prevent Pacemaker from needlessly moving +instances around the cluster. .. note:: For globally unique clones, this may result in multiple instances of the clone staying on a single node, even after another eligible node becomes active (for example, after being put into standby mode then made active again). If you do not want this behavior, specify a ``resource-stickiness`` of 0 for the clone temporarily and let the cluster adjust, then set it back to 1 if you want the default behavior to apply again. .. important:: If ``resource-stickiness`` is set in the ``rsc_defaults`` section, it will apply to clone instances as well. This means an explicit ``resource-stickiness`` of 0 in ``rsc_defaults`` works differently from the implicit default used when ``resource-stickiness`` is not specified. Clone Resource Agent Requirements _________________________________ Any resource can be used as an anonymous clone, as it requires no additional support from the resource agent. Whether it makes sense to do so depends on your resource and its resource agent. Resource Agent Requirements for Globally Unique Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Globally unique clones require additional support in the resource agent. In particular, it must only respond with ``${OCF_SUCCESS}`` if the node has that exact instance active. All other probes for instances of the clone should result in ``${OCF_NOT_RUNNING}`` (or one of the other OCF error codes if they are failed). Individual instances of a clone are identified by appending a colon and a numerical offset, e.g. **apache:2**. Resource agents can find out how many copies there are by examining the ``OCF_RESKEY_CRM_meta_clone_max`` environment variable and which instance it is by examining ``OCF_RESKEY_CRM_meta_clone``. The resource agent must not make any assumptions (based on ``OCF_RESKEY_CRM_meta_clone``) about which numerical instances are active. In particular, the list of active copies will not always be an unbroken sequence, nor always start at 0. Resource Agent Requirements for Promotable Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Promotable clone resources require two extra actions, ``demote`` and ``promote``, which are responsible for changing the state of the resource. Like **start** and **stop**, they should return ``${OCF_SUCCESS}`` if they completed successfully or a relevant error code if they did not. The states can mean whatever you wish, but when the resource is started, it must come up in the unpromoted role. From there, the cluster will decide which instances to promote. In addition to the clone requirements for monitor actions, agents must also *accurately* report which state they are in. The cluster relies on the agent to report its status (including role) accurately and does not indicate to the agent what role it currently believes it to be in. .. table:: **Role implications of OCF return codes** :widths: 1 3 +----------------------+--------------------------------------------------+ | Monitor Return Code | Description | +======================+==================================================+ | OCF_NOT_RUNNING | .. index:: | | | single: OCF_NOT_RUNNING | | | single: OCF return code; OCF_NOT_RUNNING | | | | | | Stopped | +----------------------+--------------------------------------------------+ | OCF_SUCCESS | .. index:: | | | single: OCF_SUCCESS | | | single: OCF return code; OCF_SUCCESS | | | | | | Running (Unpromoted) | +----------------------+--------------------------------------------------+ | OCF_RUNNING_PROMOTED | .. index:: | | | single: OCF_RUNNING_PROMOTED | | | single: OCF return code; OCF_RUNNING_PROMOTED | | | | | | Running (Promoted) | +----------------------+--------------------------------------------------+ | OCF_FAILED_PROMOTED | .. index:: | | | single: OCF_FAILED_PROMOTED | | | single: OCF return code; OCF_FAILED_PROMOTED | | | | | | Failed (Promoted) | +----------------------+--------------------------------------------------+ | Other | .. index:: | | | single: return code | | | | | | Failed (Unpromoted) | +----------------------+--------------------------------------------------+ Clone Notifications ~~~~~~~~~~~~~~~~~~~ If the clone has the ``notify`` meta-attribute set to **true**, and the resource agent supports the ``notify`` action, Pacemaker will call the action when appropriate, passing a number of extra variables which, when combined with additional context, can be used to calculate the current state of the cluster and what is about to happen to it. .. index:: single: clone; environment variables single: notify; environment variables .. table:: **Environment variables supplied with Clone notify actions** :widths: 1 1 +----------------------------------------------+-------------------------------------------------------------------------------+ | Variable | Description | +==============================================+===============================================================================+ | OCF_RESKEY_CRM_meta_notify_type | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_type | | | single: OCF_RESKEY_CRM_meta_notify_type | | | | | | Allowed values: **pre**, **post** | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_operation | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_operation | | | single: OCF_RESKEY_CRM_meta_notify_operation | | | | | | Allowed values: **start**, **stop** | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_start_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_start_resource | | | single: OCF_RESKEY_CRM_meta_notify_start_resource | | | | | | Resources to be started | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_stop_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_stop_resource | | | single: OCF_RESKEY_CRM_meta_notify_stop_resource | | | | | | Resources to be stopped | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_active_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_active_resource | | | single: OCF_RESKEY_CRM_meta_notify_active_resource | | | | | | Resources that are running | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_inactive_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_inactive_resource | | | single: OCF_RESKEY_CRM_meta_notify_inactive_resource | | | | | | Resources that are not running | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_start_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_start_uname | | | single: OCF_RESKEY_CRM_meta_notify_start_uname | | | | | | Nodes on which resources will be started | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_stop_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_stop_uname | | | single: OCF_RESKEY_CRM_meta_notify_stop_uname | | | | | | Nodes on which resources will be stopped | +----------------------------------------------+-------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_active_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_active_uname | | | single: OCF_RESKEY_CRM_meta_notify_active_uname | | | | | | Nodes on which resources are running | +----------------------------------------------+-------------------------------------------------------------------------------+ The variables come in pairs, such as ``OCF_RESKEY_CRM_meta_notify_start_resource`` and ``OCF_RESKEY_CRM_meta_notify_start_uname``, and should be treated as an array of whitespace-separated elements. ``OCF_RESKEY_CRM_meta_notify_inactive_resource`` is an exception, as the matching **uname** variable does not exist since inactive resources are not running on any node. Thus, in order to indicate that **clone:0** will be started on **sles-1**, **clone:2** will be started on **sles-3**, and **clone:3** will be started on **sles-2**, the cluster would set: .. topic:: Notification variables .. code-block:: none OCF_RESKEY_CRM_meta_notify_start_resource="clone:0 clone:2 clone:3" OCF_RESKEY_CRM_meta_notify_start_uname="sles-1 sles-3 sles-2" .. note:: Pacemaker will log but otherwise ignore failures of notify actions. Interpretation of Notification Variables ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ **Pre-notification (stop):** * Active resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (stop) / Pre-notification (start):** * Active resources * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Inactive resources * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (start):** * Active resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` Extra Notifications for Promotable Clones ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. index:: single: clone; environment variables single: promotable; environment variables .. table:: **Extra environment variables supplied for promotable clones** :widths: 1 1 +------------------------------------------------+---------------------------------------------------------------------------------+ | Variable | Description | +================================================+=================================================================================+ | OCF_RESKEY_CRM_meta_notify_promoted_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promoted_resource | | | single: OCF_RESKEY_CRM_meta_notify_promoted_resource | | | | | | Resources that are running in the promoted role | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_unpromoted_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_unpromoted_resource | | | single: OCF_RESKEY_CRM_meta_notify_unpromoted_resource | | | | | | Resources that are running in the unpromoted role | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_promote_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promote_resource | | | single: OCF_RESKEY_CRM_meta_notify_promote_resource | | | | | | Resources to be promoted | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_demote_resource | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_demote_resource | | | single: OCF_RESKEY_CRM_meta_notify_demote_resource | | | | | | Resources to be demoted | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_promote_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promote_uname | | | single: OCF_RESKEY_CRM_meta_notify_promote_uname | | | | | | Nodes on which resources will be promoted | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_demote_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_demote_uname | | | single: OCF_RESKEY_CRM_meta_notify_demote_uname | | | | | | Nodes on which resources will be demoted | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_promoted_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_promoted_uname | | | single: OCF_RESKEY_CRM_meta_notify_promoted_uname | | | | | | Nodes on which resources are running in the promoted role | +------------------------------------------------+---------------------------------------------------------------------------------+ | OCF_RESKEY_CRM_meta_notify_unpromoted_uname | .. index:: | | | single: environment variable; OCF_RESKEY_CRM_meta_notify_unpromoted_uname | | | single: OCF_RESKEY_CRM_meta_notify_unpromoted_uname | | | | | | Nodes on which resources are running in the unpromoted role | +------------------------------------------------+---------------------------------------------------------------------------------+ Interpretation of Promotable Notification Variables ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ **Pre-notification (demote):** * Active resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * Promoted resources: ``$OCF_RESKEY_CRM_meta_notify_promoted_resource`` * Unpromoted resources: ``$OCF_RESKEY_CRM_meta_notify_unpromoted_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (demote) / Pre-notification (stop):** * Active resources: ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * Promoted resources: * ``$OCF_RESKEY_CRM_meta_notify_promoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Unpromoted resources: ``$OCF_RESKEY_CRM_meta_notify_unpromoted_resource`` * Inactive resources: ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` **Post-notification (stop) / Pre-notification (start)** * Active resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Promoted resources: * ``$OCF_RESKEY_CRM_meta_notify_promoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Unpromoted resources: * ``$OCF_RESKEY_CRM_meta_notify_unpromoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (start) / Pre-notification (promote)** * Active resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Promoted resources: * ``$OCF_RESKEY_CRM_meta_notify_promoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Unpromoted resources: * ``$OCF_RESKEY_CRM_meta_notify_unpromoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` **Post-notification (promote)** * Active resources: * ``$OCF_RESKEY_CRM_meta_notify_active_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Promoted resources: * ``$OCF_RESKEY_CRM_meta_notify_promoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Unpromoted resources: * ``$OCF_RESKEY_CRM_meta_notify_unpromoted_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Inactive resources: * ``$OCF_RESKEY_CRM_meta_notify_inactive_resource`` * plus ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * minus ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources to be promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources to be demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources to be stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` * Resources that were started: ``$OCF_RESKEY_CRM_meta_notify_start_resource`` * Resources that were promoted: ``$OCF_RESKEY_CRM_meta_notify_promote_resource`` * Resources that were demoted: ``$OCF_RESKEY_CRM_meta_notify_demote_resource`` * Resources that were stopped: ``$OCF_RESKEY_CRM_meta_notify_stop_resource`` Monitoring Promotable Clone Resources _____________________________________ The usual monitor actions are insufficient to monitor a promotable clone resource, because Pacemaker needs to verify not only that the resource is active, but also that its actual role matches its intended one. Define two monitoring actions: the usual one will cover the unpromoted role, and an additional one with ``role="Promoted"`` will cover the promoted role. .. topic:: Monitoring both states of a promotable clone resource .. code-block:: xml .. important:: It is crucial that *every* monitor operation has a different interval! Pacemaker currently differentiates between operations only by resource and interval; so if (for example) a promotable clone resource had the same monitor interval for both roles, Pacemaker would ignore the role when checking the status -- which would cause unexpected return codes, and therefore unnecessary complications. .. _s-promotion-scores: Determining Which Instance is Promoted ______________________________________ Pacemaker can choose a promotable clone instance to be promoted in one of two ways: * Promotion scores: These are node attributes set via the ``crm_attribute`` command using the ``--promotion`` option, which generally would be called by the resource agent's start action if it supports promotable clones. This tool automatically detects both the resource and host, and should be used to set a preference for being promoted. Based on this, ``promoted-max``, and ``promoted-node-max``, the instance(s) with the highest preference will be promoted. * Constraints: Location constraints can indicate which nodes are most preferred to be promoted. .. topic:: Explicitly preferring node1 to be promoted .. code-block:: xml .. index: single: bundle single: resource; bundle pair: container; Docker pair: container; podman pair: container; rkt .. _s-resource-bundle: Bundles - Containerized Resources ################################# Pacemaker supports a special syntax for launching a service inside a `container `_ with any infrastructure it requires: the *bundle*. Pacemaker bundles support `Docker `_, `podman `_ *(since 2.0.1)*, and `rkt `_ container technologies. [#]_ .. topic:: A bundle for a containerized web server .. code-block:: xml Bundle Prerequisites ____________________ Before configuring a bundle in Pacemaker, the user must install the appropriate container launch technology (Docker, podman, or rkt), and supply a fully configured container image, on every node allowed to run the bundle. Pacemaker will create an implicit resource of type **ocf:heartbeat:docker**, **ocf:heartbeat:podman**, or **ocf:heartbeat:rkt** to manage a bundle's container. The user must ensure that the appropriate resource agent is installed on every node allowed to run the bundle. .. index:: pair: XML element; bundle Bundle Properties _________________ .. table:: **XML Attributes of a bundle Element** :widths: 1 4 +-------------+------------------------------------------------------------------+ | Field | Description | +=============+==================================================================+ | id | .. index:: | | | single: bundle; attribute, id | | | single: attribute; id (bundle) | | | single: id; bundle attribute | | | | | | A unique name for the bundle (required) | +-------------+------------------------------------------------------------------+ | description | .. index:: | | | single: bundle; attribute, description | | | single: attribute; description (bundle) | | | single: description; bundle attribute | | | | | | An optional description of the group, for the user's own | | | purposes. | | | E.g. ``manages the container that runs the service`` | +-------------+------------------------------------------------------------------+ A bundle must contain exactly one ``docker``, ``podman``, or ``rkt`` element. .. index:: pair: XML element; docker pair: XML element; podman pair: XML element; rkt Bundle Container Properties ___________________________ .. table:: **XML attributes of a docker, podman, or rkt Element** :class: longtable :widths: 2 3 4 +-------------------+------------------------------------+---------------------------------------------------+ | Attribute | Default | Description | +===================+====================================+===================================================+ | image | | .. index:: | | | | single: docker; attribute, image | | | | single: attribute; image (docker) | | | | single: image; docker attribute | | | | single: podman; attribute, image | | | | single: attribute; image (podman) | | | | single: image; podman attribute | | | | single: rkt; attribute, image | | | | single: attribute; image (rkt) | | | | single: image; rkt attribute | | | | | | | | Container image tag (required) | +-------------------+------------------------------------+---------------------------------------------------+ | replicas | Value of ``promoted-max`` | .. index:: | | | if that is positive, else 1 | single: docker; attribute, replicas | | | | single: attribute; replicas (docker) | | | | single: replicas; docker attribute | | | | single: podman; attribute, replicas | | | | single: attribute; replicas (podman) | | | | single: replicas; podman attribute | | | | single: rkt; attribute, replicas | | | | single: attribute; replicas (rkt) | | | | single: replicas; rkt attribute | | | | | | | | A positive integer specifying the number of | | | | container instances to launch | +-------------------+------------------------------------+---------------------------------------------------+ | replicas-per-host | 1 | .. index:: | | | | single: docker; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (docker) | | | | single: replicas-per-host; docker attribute | | | | single: podman; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (podman) | | | | single: replicas-per-host; podman attribute | | | | single: rkt; attribute, replicas-per-host | | | | single: attribute; replicas-per-host (rkt) | | | | single: replicas-per-host; rkt attribute | | | | | | | | A positive integer specifying the number of | | | | container instances allowed to run on a | | | | single node | +-------------------+------------------------------------+---------------------------------------------------+ | promoted-max | 0 | .. index:: | | | | single: docker; attribute, promoted-max | | | | single: attribute; promoted-max (docker) | | | | single: promoted-max; docker attribute | | | | single: podman; attribute, promoted-max | | | | single: attribute; promoted-max (podman) | | | | single: promoted-max; podman attribute | | | | single: rkt; attribute, promoted-max | | | | single: attribute; promoted-max (rkt) | | | | single: promoted-max; rkt attribute | | | | | | | | A non-negative integer that, if positive, | | | | indicates that the containerized service | | | | should be treated as a promotable service, | | | | with this many replicas allowed to run the | | | | service in the promoted role | +-------------------+------------------------------------+---------------------------------------------------+ | network | | .. index:: | | | | single: docker; attribute, network | | | | single: attribute; network (docker) | | | | single: network; docker attribute | | | | single: podman; attribute, network | | | | single: attribute; network (podman) | | | | single: network; podman attribute | | | | single: rkt; attribute, network | | | | single: attribute; network (rkt) | | | | single: network; rkt attribute | | | | | | | | If specified, this will be passed to the | | | | ``docker run``, ``podman run``, or | | | | ``rkt run`` command as the network setting | | | | for the container. | +-------------------+------------------------------------+---------------------------------------------------+ | run-command | ``/usr/sbin/pacemaker-remoted`` if | .. index:: | | | bundle contains a **primitive**, | single: docker; attribute, run-command | | | otherwise none | single: attribute; run-command (docker) | | | | single: run-command; docker attribute | | | | single: podman; attribute, run-command | | | | single: attribute; run-command (podman) | | | | single: run-command; podman attribute | | | | single: rkt; attribute, run-command | | | | single: attribute; run-command (rkt) | | | | single: run-command; rkt attribute | | | | | | | | This command will be run inside the container | | | | when launching it ("PID 1"). If the bundle | | | | contains a **primitive**, this command *must* | | | | start ``pacemaker-remoted`` (but could, for | | | | example, be a script that does other stuff, too). | +-------------------+------------------------------------+---------------------------------------------------+ | options | | .. index:: | | | | single: docker; attribute, options | | | | single: attribute; options (docker) | | | | single: options; docker attribute | | | | single: podman; attribute, options | | | | single: attribute; options (podman) | | | | single: options; podman attribute | | | | single: rkt; attribute, options | | | | single: attribute; options (rkt) | | | | single: options; rkt attribute | | | | | | | | Extra command-line options to pass to the | | | | ``docker run``, ``podman run``, or ``rkt run`` | | | | command | +-------------------+------------------------------------+---------------------------------------------------+ .. note:: Considerations when using cluster configurations or container images from Pacemaker 1.1: * If the container image has a pre-2.0.0 version of Pacemaker, set ``run-command`` to ``/usr/sbin/pacemaker_remoted`` (note the underbar instead of dash). * ``masters`` is accepted as an alias for ``promoted-max``, but is deprecated since 2.0.0, and support for it will be removed in a future version. Bundle Network Properties _________________________ A bundle may optionally contain one ```` element. .. index:: pair: XML element; network single: bundle; network .. table:: **XML attributes of a network Element** :widths: 2 1 5 +----------------+---------+------------------------------------------------------------+ | Attribute | Default | Description | +================+=========+============================================================+ | add-host | TRUE | .. index:: | | | | single: network; attribute, add-host | | | | single: attribute; add-host (network) | | | | single: add-host; network attribute | | | | | | | | If TRUE, and ``ip-range-start`` is used, Pacemaker will | | | | automatically ensure that ``/etc/hosts`` inside the | | | | containers has entries for each | | | | :ref:`replica name ` | | | | and its assigned IP. | +----------------+---------+------------------------------------------------------------+ | ip-range-start | | .. index:: | | | | single: network; attribute, ip-range-start | | | | single: attribute; ip-range-start (network) | | | | single: ip-range-start; network attribute | | | | | | | | If specified, Pacemaker will create an implicit | | | | ``ocf:heartbeat:IPaddr2`` resource for each container | | | | instance, starting with this IP address, using up to | | | | ``replicas`` sequential addresses. These addresses can be | | | | used from the host's network to reach the service inside | | | | the container, though it is not visible within the | | | | container itself. Only IPv4 addresses are currently | | | | supported. | +----------------+---------+------------------------------------------------------------+ | host-netmask | 32 | .. index:: | | | | single: network; attribute; host-netmask | | | | single: attribute; host-netmask (network) | | | | single: host-netmask; network attribute | | | | | | | | If ``ip-range-start`` is specified, the IP addresses | | | | are created with this CIDR netmask (as a number of bits). | +----------------+---------+------------------------------------------------------------+ | host-interface | | .. index:: | | | | single: network; attribute; host-interface | | | | single: attribute; host-interface (network) | | | | single: host-interface; network attribute | | | | | | | | If ``ip-range-start`` is specified, the IP addresses are | | | | created on this host interface (by default, it will be | | | | determined from the IP address). | +----------------+---------+------------------------------------------------------------+ | control-port | 3121 | .. index:: | | | | single: network; attribute; control-port | | | | single: attribute; control-port (network) | | | | single: control-port; network attribute | | | | | | | | If the bundle contains a ``primitive``, the cluster will | | | | use this integer TCP port for communication with | | | | Pacemaker Remote inside the container. Changing this is | | | | useful when the container is unable to listen on the | | | | default port, for example, when the container uses the | | | | host's network rather than ``ip-range-start`` (in which | | | | case ``replicas-per-host`` must be 1), or when the bundle | | | | may run on a Pacemaker Remote node that is already | | | | listening on the default port. Any ``PCMK_remote_port`` | | | | environment variable set on the host or in the container | | | | is ignored for bundle connections. | +----------------+---------+------------------------------------------------------------+ .. _s-resource-bundle-note-replica-names: .. note:: Replicas are named by the bundle id plus a dash and an integer counter starting with zero. For example, if a bundle named **httpd-bundle** has **replicas=2**, its containers will be named **httpd-bundle-0** and **httpd-bundle-1**. .. index:: pair: XML element; port-mapping Additionally, a ``network`` element may optionally contain one or more ``port-mapping`` elements. .. table:: **Attributes of a port-mapping Element** :widths: 2 1 5 +---------------+-------------------+------------------------------------------------------+ | Attribute | Default | Description | +===============+===================+======================================================+ | id | | .. index:: | | | | single: port-mapping; attribute, id | | | | single: attribute; id (port-mapping) | | | | single: id; port-mapping attribute | | | | | | | | A unique name for the port mapping (required) | +---------------+-------------------+------------------------------------------------------+ | port | | .. index:: | | | | single: port-mapping; attribute, port | | | | single: attribute; port (port-mapping) | | | | single: port; port-mapping attribute | | | | | | | | If this is specified, connections to this TCP port | | | | number on the host network (on the container's | | | | assigned IP address, if ``ip-range-start`` is | | | | specified) will be forwarded to the container | | | | network. Exactly one of ``port`` or ``range`` | | | | must be specified in a ``port-mapping``. | +---------------+-------------------+------------------------------------------------------+ | internal-port | value of ``port`` | .. index:: | | | | single: port-mapping; attribute, internal-port | | | | single: attribute; internal-port (port-mapping) | | | | single: internal-port; port-mapping attribute | | | | | | | | If ``port`` and this are specified, connections | | | | to ``port`` on the host's network will be | | | | forwarded to this port on the container network. | +---------------+-------------------+------------------------------------------------------+ | range | | .. index:: | | | | single: port-mapping; attribute, range | | | | single: attribute; range (port-mapping) | | | | single: range; port-mapping attribute | | | | | | | | If this is specified, connections to these TCP | | | | port numbers (expressed as *first_port*-*last_port*) | | | | on the host network (on the container's assigned IP | | | | address, if ``ip-range-start`` is specified) will | | | | be forwarded to the same ports in the container | | | | network. Exactly one of ``port`` or ``range`` | | | | must be specified in a ``port-mapping``. | +---------------+-------------------+------------------------------------------------------+ .. note:: If the bundle contains a ``primitive``, Pacemaker will automatically map the ``control-port``, so it is not necessary to specify that port in a ``port-mapping``. .. index: pair: XML element; storage pair: XML element; storage-mapping single: bundle; storage .. _s-bundle-storage: Bundle Storage Properties _________________________ A bundle may optionally contain one ``storage`` element. A ``storage`` element has no properties of its own, but may contain one or more ``storage-mapping`` elements. .. table:: **Attributes of a storage-mapping Element** :widths: 2 1 5 +-----------------+---------+-------------------------------------------------------------+ | Attribute | Default | Description | +=================+=========+=============================================================+ | id | | .. index:: | | | | single: storage-mapping; attribute, id | | | | single: attribute; id (storage-mapping) | | | | single: id; storage-mapping attribute | | | | | | | | A unique name for the storage mapping (required) | +-----------------+---------+-------------------------------------------------------------+ | source-dir | | .. index:: | | | | single: storage-mapping; attribute, source-dir | | | | single: attribute; source-dir (storage-mapping) | | | | single: source-dir; storage-mapping attribute | | | | | | | | The absolute path on the host's filesystem that will be | | | | mapped into the container. Exactly one of ``source-dir`` | | | | and ``source-dir-root`` must be specified in a | | | | ``storage-mapping``. | +-----------------+---------+-------------------------------------------------------------+ | source-dir-root | | .. index:: | | | | single: storage-mapping; attribute, source-dir-root | | | | single: attribute; source-dir-root (storage-mapping) | | | | single: source-dir-root; storage-mapping attribute | | | | | | | | The start of a path on the host's filesystem that will | | | | be mapped into the container, using a different | | | | subdirectory on the host for each container instance. | | | | The subdirectory will be named the same as the | | | | :ref:`replica name `. | | | | Exactly one of ``source-dir`` and ``source-dir-root`` | | | | must be specified in a ``storage-mapping``. | +-----------------+---------+-------------------------------------------------------------+ | target-dir | | .. index:: | | | | single: storage-mapping; attribute, target-dir | | | | single: attribute; target-dir (storage-mapping) | | | | single: target-dir; storage-mapping attribute | | | | | | | | The path name within the container where the host | | | | storage will be mapped (required) | +-----------------+---------+-------------------------------------------------------------+ | options | | .. index:: | | | | single: storage-mapping; attribute, options | | | | single: attribute; options (storage-mapping) | | | | single: options; storage-mapping attribute | | | | | | | | A comma-separated list of file system mount | | | | options to use when mapping the storage | +-----------------+---------+-------------------------------------------------------------+ .. note:: Pacemaker does not define the behavior if the source directory does not already exist on the host. However, it is expected that the container technology and/or its resource agent will create the source directory in that case. .. note:: If the bundle contains a ``primitive``, Pacemaker will automatically map the equivalent of ``source-dir=/etc/pacemaker/authkey target-dir=/etc/pacemaker/authkey`` and ``source-dir-root=/var/log/pacemaker/bundles target-dir=/var/log`` into the container, so it is not necessary to specify those paths in a ``storage-mapping``. .. important:: The ``PCMK_authkey_location`` environment variable must not be set to anything other than the default of ``/etc/pacemaker/authkey`` on any node in the cluster. .. important:: If SELinux is used in enforcing mode on the host, you must ensure the container is allowed to use any storage you mount into it. For Docker and podman bundles, adding "Z" to the mount options will create a container-specific label for the mount that allows the container access. .. index:: single: bundle; primitive Bundle Primitive ________________ A bundle may optionally contain one :ref:`primitive ` resource. The primitive may have operations, instance attributes, and meta-attributes defined, as usual. If a bundle contains a primitive resource, the container image must include the Pacemaker Remote daemon, and at least one of ``ip-range-start`` or ``control-port`` must be configured in the bundle. Pacemaker will create an implicit **ocf:pacemaker:remote** resource for the connection, launch Pacemaker Remote within the container, and monitor and manage the primitive resource via Pacemaker Remote. If the bundle has more than one container instance (replica), the primitive resource will function as an implicit :ref:`clone ` -- a :ref:`promotable clone ` if the bundle has ``promoted-max`` greater than zero. .. note:: If you want to pass environment variables to a bundle's Pacemaker Remote connection or primitive, you have two options: * Environment variables whose value is the same regardless of the underlying host may be set using the container element's ``options`` attribute. * If you want variables to have host-specific values, you can use the :ref:`storage-mapping ` element to map a file on the host as ``/etc/pacemaker/pcmk-init.env`` in the container *(since 2.0.3)*. Pacemaker Remote will parse this file as a shell-like format, with variables set as NAME=VALUE, ignoring blank lines and comments starting with "#". .. important:: When a bundle has a ``primitive``, Pacemaker on all cluster nodes must be able to contact Pacemaker Remote inside the bundle's containers. * The containers must have an accessible network (for example, ``network`` should not be set to "none" with a ``primitive``). * The default, using a distinct network space inside the container, works in combination with ``ip-range-start``. Any firewall must allow access from all cluster nodes to the ``control-port`` on the container IPs. * If the container shares the host's network space (for example, by setting ``network`` to "host"), a unique ``control-port`` should be specified for each bundle. Any firewall must allow access from all cluster nodes to the ``control-port`` on all cluster and remote node IPs. .. index:: single: bundle; node attributes .. _s-bundle-attributes: Bundle Node Attributes ______________________ If the bundle has a ``primitive``, the primitive's resource agent may want to set node attributes such as :ref:`promotion scores `. However, with containers, it is not apparent which node should get the attribute. If the container uses shared storage that is the same no matter which node the container is hosted on, then it is appropriate to use the promotion score on the bundle node itself. On the other hand, if the container uses storage exported from the underlying host, then it may be more appropriate to use the promotion score on the underlying host. Since this depends on the particular situation, the ``container-attribute-target`` resource meta-attribute allows the user to specify which approach to use. If it is set to ``host``, then user-defined node attributes will be checked on the underlying host. If it is anything else, the local node (in this case the bundle node) is used as usual. This only applies to user-defined attributes; the cluster will always check the local node for cluster-defined attributes such as ``#uname``. If ``container-attribute-target`` is ``host``, the cluster will pass additional environment variables to the primitive's resource agent that allow it to set node attributes appropriately: ``CRM_meta_container_attribute_target`` (identical to the meta-attribute value) and ``CRM_meta_physical_host`` (the name of the underlying host). .. note:: When called by a resource agent, the ``attrd_updater`` and ``crm_attribute`` commands will automatically check those environment variables and set attributes appropriately. .. index:: single: bundle; meta-attributes Bundle Meta-Attributes ______________________ Any meta-attribute set on a bundle will be inherited by the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. This includes options such as ``priority``, ``target-role``, and ``is-managed``. See :ref:`resource_options` for more information. Bundles support clone meta-attributes including ``notify``, ``ordered``, and ``interleave``. Limitations of Bundles ______________________ Restarting pacemaker while a bundle is unmanaged or the cluster is in maintenance mode may cause the bundle to fail. Bundles may not be explicitly cloned or included in groups. This includes the bundle's primitive and any resources implicitly created by Pacemaker for the bundle. (If ``replicas`` is greater than 1, the bundle will behave like a clone implicitly.) Bundles do not have instance attributes, utilization attributes, or operations, though a bundle's primitive may have them. A bundle with a primitive can run on a Pacemaker Remote node only if the bundle uses a distinct ``control-port``. .. [#] Of course, the service must support running multiple instances. .. [#] Docker is a trademark of Docker, Inc. No endorsement by or association with Docker, Inc. is implied. diff --git a/doc/sphinx/Pacemaker_Explained/options.rst b/doc/sphinx/Pacemaker_Explained/options.rst index 5d95e4c867..ca7ea2a8a3 100644 --- a/doc/sphinx/Pacemaker_Explained/options.rst +++ b/doc/sphinx/Pacemaker_Explained/options.rst @@ -1,631 +1,631 @@ Cluster-Wide Configuration -------------------------- .. index:: pair: XML element; cib pair: XML element; configuration Configuration Layout #################### The cluster is defined by the Cluster Information Base (CIB), which uses XML notation. The simplest CIB, an empty one, looks like this: .. topic:: An empty configuration .. code-block:: xml The empty configuration above contains the major sections that make up a CIB: * ``cib``: The entire CIB is enclosed with a ``cib`` element. Certain fundamental settings are defined as attributes of this element. * ``configuration``: This section -- the primary focus of this document -- contains traditional configuration information such as what resources the cluster serves and the relationships among them. * ``crm_config``: cluster-wide configuration options * ``nodes``: the machines that host the cluster * ``resources``: the services run by the cluster * ``constraints``: indications of how resources should be placed * ``status``: This section contains the history of each resource on each node. Based on this data, the cluster can construct the complete current state of the cluster. The authoritative source for this section is the local executor (pacemaker-execd process) on each cluster node, and the cluster will occasionally repopulate the entire section. For this reason, it is never written to disk, and administrators are advised against modifying it in any way. In this document, configuration settings will be described as properties or options based on how they are defined in the CIB: * Properties are XML attributes of an XML element. * Options are name-value pairs expressed as ``nvpair`` child elements of an XML element. Normally, you will use command-line tools that abstract the XML, so the distinction will be unimportant; both properties and options are cluster settings you can tweak. CIB Properties ############## Certain settings are defined by CIB properties (that is, attributes of the ``cib`` tag) rather than with the rest of the cluster configuration in the ``configuration`` section. The reason is simply a matter of parsing. These options are used by the configuration database which is, by design, mostly ignorant of the content it holds. So the decision was made to place them in an easy-to-find location. .. table:: **CIB Properties** :class: longtable :widths: 1 3 +------------------+-----------------------------------------------------------+ | Attribute | Description | +==================+===========================================================+ | admin_epoch | .. index:: | | | pair: admin_epoch; cib | | | | | | When a node joins the cluster, the cluster performs a | | | check to see which node has the best configuration. It | | | asks the node with the highest (``admin_epoch``, | | | ``epoch``, ``num_updates``) tuple to replace the | | | configuration on all the nodes -- which makes setting | | | them, and setting them correctly, very important. | | | ``admin_epoch`` is never modified by the cluster; you can | | | use this to make the configurations on any inactive nodes | | | obsolete. | | | | | | **Warning:** Never set this value to zero. In such cases, | | | the cluster cannot tell the difference between your | | | configuration and the "empty" one used when nothing is | | | found on disk. | +------------------+-----------------------------------------------------------+ | epoch | .. index:: | | | pair: epoch; cib | | | | | | The cluster increments this every time the configuration | | | is updated (usually by the administrator). | +------------------+-----------------------------------------------------------+ | num_updates | .. index:: | | | pair: num_updates; cib | | | | | | The cluster increments this every time the configuration | | | or status is updated (usually by the cluster) and resets | | | it to 0 when epoch changes. | +------------------+-----------------------------------------------------------+ | validate-with | .. index:: | | | pair: validate-with; cib | | | | | | Determines the type of XML validation that will be done | | | on the configuration. If set to ``none``, the cluster | | | will not verify that updates conform to the DTD (nor | | | reject ones that don't). | +------------------+-----------------------------------------------------------+ | cib-last-written | .. index:: | | | pair: cib-last-written; cib | | | | | | Indicates when the configuration was last written to | | | disk. Maintained by the cluster; for informational | | | purposes only. | +------------------+-----------------------------------------------------------+ | have-quorum | .. index:: | | | pair: have-quorum; cib | | | | | | Indicates if the cluster has quorum. If false, this may | | | mean that the cluster cannot start resources or fence | | | other nodes (see ``no-quorum-policy`` below). Maintained | | | by the cluster. | +------------------+-----------------------------------------------------------+ | dc-uuid | .. index:: | | | pair: dc-uuid; cib | | | | | | Indicates which cluster node is the current leader. Used | | | by the cluster when placing resources and determining the | | | order of some events. Maintained by the cluster. | +------------------+-----------------------------------------------------------+ .. _cluster_options: Cluster Options ############### Cluster options, as you might expect, control how the cluster behaves when confronted with various situations. They are grouped into sets within the ``crm_config`` section. In advanced configurations, there may be more than one set. (This will be described later in the chapter on :ref:`rules` where we will show how to have the cluster use different sets of options during working hours than during weekends.) For now, we will describe the simple case where each option is present at most once. You can obtain an up-to-date list of cluster options, including their default values, by running the ``man pacemaker-schedulerd`` and ``man pacemaker-controld`` commands. .. table:: **Cluster Options** :class: longtable :widths: 2 1 4 +---------------------------+---------+----------------------------------------------------+ | Option | Default | Description | +===========================+=========+====================================================+ | cluster-name | | .. index:: | | | | pair: cluster option; cluster-name | | | | | | | | An (optional) name for the cluster as a whole. | | | | This is mostly for users' convenience for use | | | | as desired in administration, but this can be | | | | used in the Pacemaker configuration in | | | | :ref:`rules` (as the ``#cluster-name`` | | | | :ref:`node attribute | | | | `. It may | | | | also be used by higher-level tools when | | | | displaying cluster information, and by | | | | certain resource agents (for example, the | | | | ``ocf:heartbeat:GFS2`` agent stores the | | | | cluster name in filesystem meta-data). | +---------------------------+---------+----------------------------------------------------+ | dc-version | | .. index:: | | | | pair: cluster option; dc-version | | | | | | | | Version of Pacemaker on the cluster's DC. | | | | Determined automatically by the cluster. Often | | | | includes the hash which identifies the exact | | | | Git changeset it was built from. Used for | | | | diagnostic purposes. | +---------------------------+---------+----------------------------------------------------+ | cluster-infrastructure | | .. index:: | | | | pair: cluster option; cluster-infrastructure | | | | | | | | The messaging stack on which Pacemaker is | | | | currently running. Determined automatically by | | | | the cluster. Used for informational and | | | | diagnostic purposes. | +---------------------------+---------+----------------------------------------------------+ | no-quorum-policy | stop | .. index:: | | | | pair: cluster option; no-quorum-policy | | | | | | | | What to do when the cluster does not have | | | | quorum. Allowed values: | | | | | | | | * ``ignore:`` continue all resource management | | | | * ``freeze:`` continue resource management, but | | | | don't recover resources from nodes not in the | | | | affected partition | | | | * ``stop:`` stop all resources in the affected | | | | cluster partition | | | | * ``demote:`` demote promotable resources and | | | | stop all other resources in the affected | | | | cluster partition *(since 2.0.5)* | | | | * ``suicide:`` fence all nodes in the affected | | | | cluster partition | +---------------------------+---------+----------------------------------------------------+ | batch-limit | 0 | .. index:: | | | | pair: cluster option; batch-limit | | | | | | | | The maximum number of actions that the cluster | | | | may execute in parallel across all nodes. The | | | | "correct" value will depend on the speed and | | | | load of your network and cluster nodes. If zero, | | | | the cluster will impose a dynamically calculated | | | | limit only when any node has high load. If -1, the | | | | cluster will not impose any limit. | +---------------------------+---------+----------------------------------------------------+ | migration-limit | -1 | .. index:: | | | | pair: cluster option; migration-limit | | | | | | | | The number of | | | | :ref:`live migration ` actions | | | | that the cluster is allowed to execute in | | | | parallel on a node. A value of -1 means | | | | unlimited. | +---------------------------+---------+----------------------------------------------------+ | symmetric-cluster | true | .. index:: | | | | pair: cluster option; symmetric-cluster | | | | | | | | Whether resources can run on any node by default | | | | (if false, a resource is allowed to run on a | | | | node only if a | | | | :ref:`location constraint ` | | | | enables it) | +---------------------------+---------+----------------------------------------------------+ | stop-all-resources | false | .. index:: | | | | pair: cluster option; stop-all-resources | | | | | | | | Whether all resources should be disallowed from | | | | running (can be useful during maintenance) | +---------------------------+---------+----------------------------------------------------+ | stop-orphan-resources | true | .. index:: | | | | pair: cluster option; stop-orphan-resources | | | | | | | | Whether resources that have been deleted from | | | | the configuration should be stopped. This value | | | | takes precedence over ``is-managed`` (that is, | | | | even unmanaged resources will be stopped when | | | | orphaned if this value is ``true`` | +---------------------------+---------+----------------------------------------------------+ | stop-orphan-actions | true | .. index:: | | | | pair: cluster option; stop-orphan-actions | | | | | | | | Whether recurring :ref:`operations ` | | | | that have been deleted from the configuration | | | | should be cancelled | +---------------------------+---------+----------------------------------------------------+ | start-failure-is-fatal | true | .. index:: | | | | pair: cluster option; start-failure-is-fatal | | | | | | | | Whether a failure to start a resource on a | | | | particular node prevents further start attempts | | | | on that node? If ``false``, the cluster will | | | | decide whether the node is still eligible based | | | | on the resource's current failure count and | | | | :ref:`migration-threshold `. | +---------------------------+---------+----------------------------------------------------+ | enable-startup-probes | true | .. index:: | | | | pair: cluster option; enable-startup-probes | | | | | | | | Whether the cluster should check the | | | | pre-existing state of resources when the cluster | | | | starts | +---------------------------+---------+----------------------------------------------------+ | maintenance-mode | false | .. index:: | | | | pair: cluster option; maintenance-mode | | | | | | | | Whether the cluster should refrain from | | | | monitoring, starting and stopping resources | +---------------------------+---------+----------------------------------------------------+ | stonith-enabled | true | .. index:: | | | | pair: cluster option; stonith-enabled | | | | | | | | Whether the cluster is allowed to fence nodes | | | | (for example, failed nodes and nodes with | | | | resources that can't be stopped. | | | | | | | | If true, at least one fence device must be | | | | configured before resources are allowed to run. | | | | | | | | If false, unresponsive nodes are immediately | | | | assumed to be running no resources, and resource | | | | recovery on online nodes starts without any | | | | further protection (which can mean *data loss* | | | | if the unresponsive node still accesses shared | | | | storage, for example). See also the | | | | :ref:`requires ` resource | | | | meta-attribute. | +---------------------------+---------+----------------------------------------------------+ | stonith-action | reboot | .. index:: | | | | pair: cluster option; stonith-action | | | | | | | | Action the cluster should send to the fence agent | | | | when a node must be fenced. Allowed values are | | | | ``reboot``, ``off``, and (for legacy agents only) | | | | ``poweroff``. | +---------------------------+---------+----------------------------------------------------+ | stonith-timeout | 60s | .. index:: | | | | pair: cluster option; stonith-timeout | | | | | | | | How long to wait for ``on``, ``off``, and | | | | ``reboot`` fence actions to complete by default. | +---------------------------+---------+----------------------------------------------------+ | stonith-max-attempts | 10 | .. index:: | | | | pair: cluster option; stonith-max-attempts | | | | | | | | How many times fencing can fail for a target | | | | before the cluster will no longer immediately | | | | re-attempt it. | +---------------------------+---------+----------------------------------------------------+ | stonith-watchdog-timeout | 0 | .. index:: | | | | pair: cluster option; stonith-watchdog-timeout | | | | | | | | If nonzero, and the cluster detects | | | | ``have-watchdog`` as ``true``, then watchdog-based | | | | self-fencing will be performed via SBD when | | | | fencing is required, without requiring a fencing | | | | resource explicitly configured. | | | | | | | | If this is set to a positive value, unseen nodes | | | | are assumed to self-fence within this much time. | | | | | | | | **Warning:** It must be ensured that this value is | | | | larger than the ``SBD_WATCHDOG_TIMEOUT`` | | | | environment variable on all nodes. Pacemaker | | | | verifies the settings individually on all nodes | | | | and prevents startup or shuts down if configured | | | | wrongly on the fly. It is strongly recommended | | | | that ``SBD_WATCHDOG_TIMEOUT`` be set to the same | | | | value on all nodes. | | | | | | | | If this is set to a negative value, and | | | | ``SBD_WATCHDOG_TIMEOUT`` is set, twice that value | | | | will be used. | | | | | | | | **Warning:** In this case, it is essential (and | | | | currently not verified by pacemaker) that | | | | ``SBD_WATCHDOG_TIMEOUT`` is set to the same | | | | value on all nodes. | +---------------------------+---------+----------------------------------------------------+ | concurrent-fencing | false | .. index:: | | | | pair: cluster option; concurrent-fencing | | | | | | | | Whether the cluster is allowed to initiate | | | | multiple fence actions concurrently. Fence actions | | | | initiated externally, such as via the | | | | ``stonith_admin`` tool or an application such as | | | | DLM, or by the fencer itself such as recurring | | | | device monitors and ``status`` and ``list`` | | | | commands, are not limited by this option. | +---------------------------+---------+----------------------------------------------------+ | fence-reaction | stop | .. index:: | | | | pair: cluster option; fence-reaction | | | | | | | | How should a cluster node react if notified of its | | | | own fencing? A cluster node may receive | | | | notification of its own fencing if fencing is | | | | misconfigured, or if fabric fencing is in use that | | | | doesn't cut cluster communication. Allowed values | | | | are ``stop`` to attempt to immediately stop | | | | pacemaker and stay stopped, or ``panic`` to | | | | attempt to immediately reboot the local node, | | | | falling back to stop on failure. The default is | | | | likely to be changed to ``panic`` in a future | | | | release. *(since 2.0.3)* | +---------------------------+---------+----------------------------------------------------+ | priority-fencing-delay | 0 | .. index:: | | | | pair: cluster option; priority-fencing-delay | | | | | | | | Apply this delay to any fencing targeting the lost | | | | nodes with the highest total resource priority in | | | | case we don't have the majority of the nodes in | | | | our cluster partition, so that the more | | | | significant nodes potentially win any fencing | | | | match (especially meaningful in a split-brain of a | | | | 2-node cluster). A promoted resource instance | | | | takes the resource's priority plus 1 if the | | | | resource's priority is not 0. Any static or random | | | | delays introduced by ``pcmk_delay_base`` and | | | | ``pcmk_delay_max`` configured for the | | | | corresponding fencing resources will be added to | | | | this delay. This delay should be significantly | | | | greater than (safely twice) the maximum delay from | | | | those parameters. *(since 2.0.4)* | +---------------------------+---------+----------------------------------------------------+ | node-pending-timeout | 10min | .. index:: | | | | pair: cluster option; node-pending-timeout | | | | | | | | A node that has joined the cluster can be pending | | | | on joining the process group. We wait up to this | | | | much time for it. If it times out, fencing | | | | targeting the node will be issued if enabled. | | | | *(since 2.1.7)* | +---------------------------+---------+----------------------------------------------------+ | cluster-delay | 60s | .. index:: | | | | pair: cluster option; cluster-delay | | | | | | | | Estimated maximum round-trip delay over the | | | | network (excluding action execution). If the DC | | | | requires an action to be executed on another node, | | | | it will consider the action failed if it does not | | | | get a response from the other node in this time | | | | (after considering the action's own timeout). The | | | | "correct" value will depend on the speed and load | | | | of your network and cluster nodes. | +---------------------------+---------+----------------------------------------------------+ | dc-deadtime | 20s | .. index:: | | | | pair: cluster option; dc-deadtime | | | | | | | | How long to wait for a response from other nodes | | | | during startup. The "correct" value will depend on | | | | the speed/load of your network and the type of | | | | switches used. | +---------------------------+---------+----------------------------------------------------+ | cluster-ipc-limit | 500 | .. index:: | | | | pair: cluster option; cluster-ipc-limit | | | | | | | | The maximum IPC message backlog before one cluster | | | | daemon will disconnect another. This is of use in | | | | large clusters, for which a good value is the | | | | number of resources in the cluster multiplied by | | | | the number of nodes. The default of 500 is also | | | | the minimum. Raise this if you see | | | | "Evicting client" messages for cluster daemon PIDs | | | | in the logs. | +---------------------------+---------+----------------------------------------------------+ | pe-error-series-max | -1 | .. index:: | | | | pair: cluster option; pe-error-series-max | | | | | | | | The number of scheduler inputs resulting in errors | | | | to save. Used when reporting problems. A value of | | | | -1 means unlimited (report all), and 0 means none. | +---------------------------+---------+----------------------------------------------------+ | pe-warn-series-max | 5000 | .. index:: | | | | pair: cluster option; pe-warn-series-max | | | | | | | | The number of scheduler inputs resulting in | | | | warnings to save. Used when reporting problems. A | | | | value of -1 means unlimited (report all), and 0 | | | | means none. | +---------------------------+---------+----------------------------------------------------+ | pe-input-series-max | 4000 | .. index:: | | | | pair: cluster option; pe-input-series-max | | | | | | | | The number of "normal" scheduler inputs to save. | | | | Used when reporting problems. A value of -1 means | | | | unlimited (report all), and 0 means none. | +---------------------------+---------+----------------------------------------------------+ | enable-acl | false | .. index:: | | | | pair: cluster option; enable-acl | | | | | | | | Whether :ref:`acl` should be used to authorize | | | | modifications to the CIB | +---------------------------+---------+----------------------------------------------------+ | placement-strategy | default | .. index:: | | | | pair: cluster option; placement-strategy | | | | | - | | | How the cluster should allocate resources to nodes | + | | | How the cluster should assign resources to nodes | | | | (see :ref:`utilization`). Allowed values are | | | | ``default``, ``utilization``, ``balanced``, and | | | | ``minimal``. | +---------------------------+---------+----------------------------------------------------+ | node-health-strategy | none | .. index:: | | | | pair: cluster option; node-health-strategy | | | | | | | | How the cluster should react to node health | | | | attributes (see :ref:`node-health`). Allowed values| | | | are ``none``, ``migrate-on-red``, ``only-green``, | | | | ``progressive``, and ``custom``. | +---------------------------+---------+----------------------------------------------------+ | node-health-base | 0 | .. index:: | | | | pair: cluster option; node-health-base | | | | | | | | The base health score assigned to a node. Only | | | | used when ``node-health-strategy`` is | | | | ``progressive``. | +---------------------------+---------+----------------------------------------------------+ | node-health-green | 0 | .. index:: | | | | pair: cluster option; node-health-green | | | | | | | | The score to use for a node health attribute whose | | | | value is ``green``. Only used when | | | | ``node-health-strategy`` is ``progressive`` or | | | | ``custom``. | +---------------------------+---------+----------------------------------------------------+ | node-health-yellow | 0 | .. index:: | | | | pair: cluster option; node-health-yellow | | | | | | | | The score to use for a node health attribute whose | | | | value is ``yellow``. Only used when | | | | ``node-health-strategy`` is ``progressive`` or | | | | ``custom``. | +---------------------------+---------+----------------------------------------------------+ | node-health-red | 0 | .. index:: | | | | pair: cluster option; node-health-red | | | | | | | | The score to use for a node health attribute whose | | | | value is ``red``. Only used when | | | | ``node-health-strategy`` is ``progressive`` or | | | | ``custom``. | +---------------------------+---------+----------------------------------------------------+ | cluster-recheck-interval | 15min | .. index:: | | | | pair: cluster option; cluster-recheck-interval | | | | | | | | Pacemaker is primarily event-driven, and looks | | | | ahead to know when to recheck the cluster for | | | | failure timeouts and most time-based rules | | | | *(since 2.0.3)*. However, it will also recheck the | | | | cluster after this amount of inactivity. This has | | | | two goals: rules with ``date_spec`` are only | | | | guaranteed to be checked this often, and it also | | | | serves as a fail-safe for some kinds of scheduler | | | | bugs. A value of 0 disables this polling; positive | | | | values are a time interval. | +---------------------------+---------+----------------------------------------------------+ | shutdown-lock | false | .. index:: | | | | pair: cluster option; shutdown-lock | | | | | | | | The default of false allows active resources to be | | | | recovered elsewhere when their node is cleanly | | | | shut down, which is what the vast majority of | | | | users will want. However, some users prefer to | | | | make resources highly available only for failures, | | | | with no recovery for clean shutdowns. If this | | | | option is true, resources active on a node when it | | | | is cleanly shut down are kept "locked" to that | | | | node (not allowed to run elsewhere) until they | | | | start again on that node after it rejoins (or for | | | | at most ``shutdown-lock-limit``, if set). Stonith | | | | resources and Pacemaker Remote connections are | | | | never locked. Clone and bundle instances and the | | | | promoted role of promotable clones are currently | | | | never locked, though support could be added in a | | | | future release. Locks may be manually cleared | | | | using the ``--refresh`` option of ``crm_resource`` | | | | (both the resource and node must be specified; | | | | this works with remote nodes if their connection | | | | resource's ``target-role`` is set to ``Stopped``, | | | | but not if Pacemaker Remote is stopped on the | | | | remote node without disabling the connection | | | | resource). *(since 2.0.4)* | +---------------------------+---------+----------------------------------------------------+ | shutdown-lock-limit | 0 | .. index:: | | | | pair: cluster option; shutdown-lock-limit | | | | | | | | If ``shutdown-lock`` is true, and this is set to a | | | | nonzero time duration, locked resources will be | | | | allowed to start after this much time has passed | | | | since the node shutdown was initiated, even if the | | | | node has not rejoined. (This works with remote | | | | nodes only if their connection resource's | | | | ``target-role`` is set to ``Stopped``.) | | | | *(since 2.0.4)* | +---------------------------+---------+----------------------------------------------------+ | remove-after-stop | false | .. index:: | | | | pair: cluster option; remove-after-stop | | | | | | | | *Deprecated* Should the cluster remove | | | | resources from Pacemaker's executor after they are | | | | stopped? Values other than the default are, at | | | | best, poorly tested and potentially dangerous. | | | | This option is deprecated and will be removed in a | | | | future release. | +---------------------------+---------+----------------------------------------------------+ | startup-fencing | true | .. index:: | | | | pair: cluster option; startup-fencing | | | | | | | | *Advanced Use Only:* Should the cluster fence | | | | unseen nodes at start-up? Setting this to false is | | | | unsafe, because the unseen nodes could be active | | | | and running resources but unreachable. | +---------------------------+---------+----------------------------------------------------+ | election-timeout | 2min | .. index:: | | | | pair: cluster option; election-timeout | | | | | | | | *Advanced Use Only:* If you need to adjust this | | | | value, it probably indicates the presence of a bug.| +---------------------------+---------+----------------------------------------------------+ | shutdown-escalation | 20min | .. index:: | | | | pair: cluster option; shutdown-escalation | | | | | | | | *Advanced Use Only:* If you need to adjust this | | | | value, it probably indicates the presence of a bug.| +---------------------------+---------+----------------------------------------------------+ | join-integration-timeout | 3min | .. index:: | | | | pair: cluster option; join-integration-timeout | | | | | | | | *Advanced Use Only:* If you need to adjust this | | | | value, it probably indicates the presence of a bug.| +---------------------------+---------+----------------------------------------------------+ | join-finalization-timeout | 30min | .. index:: | | | | pair: cluster option; join-finalization-timeout | | | | | | | | *Advanced Use Only:* If you need to adjust this | | | | value, it probably indicates the presence of a bug.| +---------------------------+---------+----------------------------------------------------+ | transition-delay | 0s | .. index:: | | | | pair: cluster option; transition-delay | | | | | | | | *Advanced Use Only:* Delay cluster recovery for | | | | the configured interval to allow for additional or | | | | related events to occur. This can be useful if | | | | your configuration is sensitive to the order in | | | | which ping updates arrive. Enabling this option | | | | will slow down cluster recovery under all | | | | conditions. | +---------------------------+---------+----------------------------------------------------+ diff --git a/doc/sphinx/Pacemaker_Explained/utilization.rst b/doc/sphinx/Pacemaker_Explained/utilization.rst index 93c67cdf31..87eef6021e 100644 --- a/doc/sphinx/Pacemaker_Explained/utilization.rst +++ b/doc/sphinx/Pacemaker_Explained/utilization.rst @@ -1,264 +1,264 @@ .. _utilization: Utilization and Placement Strategy ---------------------------------- Pacemaker decides where to place a resource according to the resource -allocation scores on every node. The resource will be allocated to the +assignment scores on every node. The resource will be assigned to the node where the resource has the highest score. -If the resource allocation scores on all the nodes are equal, by the default +If the resource assignment scores on all the nodes are equal, by the default placement strategy, Pacemaker will choose a node with the least number of -allocated resources for balancing the load. If the number of resources on each +assigned resources for balancing the load. If the number of resources on each node is equal, the first eligible node listed in the CIB will be chosen to run the resource. Often, in real-world situations, different resources use significantly different proportions of a node's capacities (memory, I/O, etc.). We cannot balance the load ideally just according to the number of resources -allocated to a node. Besides, if resources are placed such that their combined +assigned to a node. Besides, if resources are placed such that their combined requirements exceed the provided capacity, they may fail to start completely or run with degraded performance. To take these factors into account, Pacemaker allows you to configure: #. The capacity a certain node provides. #. The capacity a certain resource requires. #. An overall strategy for placement of resources. Utilization attributes ###################### To configure the capacity that a node provides or a resource requires, you can use *utilization attributes* in ``node`` and ``resource`` objects. You can name utilization attributes according to your preferences and define as many name/value pairs as your configuration needs. However, the attributes' values must be integers. .. topic:: Specifying CPU and RAM capacities of two nodes .. code-block:: xml .. topic:: Specifying CPU and RAM consumed by several resources .. code-block:: xml A node is considered eligible for a resource if it has sufficient free capacity to satisfy the resource's requirements. The nature of the required or provided capacities is completely irrelevant to Pacemaker -- it just makes sure that all capacity requirements of a resource are satisfied before placing a resource to a node. Utilization attributes used on a node object can also be *transient* *(since 2.1.6)*. These attributes are added to a ``transient_attributes`` section for the node and are forgotten by the cluster when the node goes offline. The ``attrd_updater`` tool can be used to set these attributes. .. topic:: Transient utilization attribute for node cluster-1 .. code-block:: xml .. note:: Utilization is supported for bundles *(since 2.1.3)*, but only for bundles with an inner primitive. Any resource utilization values should be specified for the inner primitive, but any priority meta-attribute should be specified for the outer bundle. Placement Strategy ################## After you have configured the capacities your nodes provide and the capacities your resources require, you need to set the ``placement-strategy`` in the global cluster options, otherwise the capacity configurations have *no effect*. Four values are available for the ``placement-strategy``: * **default** Utilization values are not taken into account at all. - Resources are allocated according to allocation scores. If scores are equal, + Resources are assigned according to assignment scores. If scores are equal, resources are evenly distributed across nodes. * **utilization** Utilization values are taken into account *only* when deciding whether a node is considered eligible (i.e. whether it has sufficient free capacity to satisfy the resource's requirements). Load-balancing is still done based on the - number of resources allocated to a node. + number of resources assigned to a node. * **balanced** Utilization values are taken into account when deciding whether a node is eligible to serve a resource *and* when load-balancing, so an attempt is made to spread the resources in a way that optimizes resource performance. * **minimal** Utilization values are taken into account *only* when deciding whether a node is eligible to serve a resource. For load-balancing, an attempt is made to concentrate the resources on as few nodes as possible, thereby enabling possible power savings on the remaining nodes. Set ``placement-strategy`` with ``crm_attribute``: .. code-block:: none # crm_attribute --name placement-strategy --update balanced Now Pacemaker will ensure the load from your resources will be distributed evenly throughout the cluster, without the need for convoluted sets of colocation constraints. -Allocation Details +Assignment Details ################## -Which node is preferred to get consumed first when allocating resources? -________________________________________________________________________ +Which node is preferred to get consumed first when assigning resources? +_______________________________________________________________________ * The node with the highest node weight gets consumed first. Node weight is a score maintained by the cluster to represent node health. * If multiple nodes have the same node weight: * If ``placement-strategy`` is ``default`` or ``utilization``, - the node that has the least number of allocated resources gets consumed first. + the node that has the least number of assigned resources gets consumed first. - * If their numbers of allocated resources are equal, + * If their numbers of assigned resources are equal, the first eligible node listed in the CIB gets consumed first. * If ``placement-strategy`` is ``balanced``, the node that has the most free capacity gets consumed first. * If the free capacities of the nodes are equal, - the node that has the least number of allocated resources gets consumed first. + the node that has the least number of assigned resources gets consumed first. - * If their numbers of allocated resources are equal, + * If their numbers of assigned resources are equal, the first eligible node listed in the CIB gets consumed first. * If ``placement-strategy`` is ``minimal``, the first eligible node listed in the CIB gets consumed first. Which node has more free capacity? __________________________________ If only one type of utilization attribute has been defined, free capacity is a simple numeric comparison. If multiple types of utilization attributes have been defined, then the node that is numerically highest in the the most attribute types has the most free capacity. For example: * If ``nodeA`` has more free ``cpus``, and ``nodeB`` has more free ``memory``, then their free capacities are equal. * If ``nodeA`` has more free ``cpus``, while ``nodeB`` has more free ``memory`` and ``storage``, then ``nodeB`` has more free capacity. Which resource is preferred to be assigned first? _________________________________________________ * The resource that has the highest ``priority`` (see :ref:`resource_options`) gets - allocated first. + assigned first. * If their priorities are equal, check whether they are already running. The - resource that has the highest score on the node where it's running gets allocated + resource that has the highest score on the node where it's running gets assigned first, to prevent resource shuffling. * If the scores above are equal or the resources are not running, the resource has - the highest score on the preferred node gets allocated first. + the highest score on the preferred node gets assigned first. * If the scores above are equal, the first runnable resource listed in the CIB - gets allocated first. + gets assigned first. Limitations and Workarounds ########################### The type of problem Pacemaker is dealing with here is known as the `knapsack problem `_ and falls into the `NP-complete `_ category of computer science problems -- a fancy way of saying "it takes a really long time to solve". Clearly in a HA cluster, it's not acceptable to spend minutes, let alone hours or days, finding an optimal solution while services remain unavailable. So instead of trying to solve the problem completely, Pacemaker uses a *best effort* algorithm for determining which node should host a particular service. This means it arrives at a solution much faster than traditional linear programming algorithms, but by doing so at the price of leaving some services stopped. In the contrived example at the start of this chapter: -* ``rsc-small`` would be allocated to ``node1`` +* ``rsc-small`` would be assigned to ``node1`` -* ``rsc-medium`` would be allocated to ``node2`` +* ``rsc-medium`` would be assigned to ``node2`` * ``rsc-large`` would remain inactive Which is not ideal. There are various approaches to dealing with the limitations of pacemaker's placement strategy: * **Ensure you have sufficient physical capacity.** It might sound obvious, but if the physical capacity of your nodes is (close to) maxed out by the cluster under normal conditions, then failover isn't going to go well. Even without the utilization feature, you'll start hitting timeouts and getting secondary failures. * **Build some buffer into the capabilities advertised by the nodes.** Advertise slightly more resources than we physically have, on the (usually valid) assumption that a resource will not use 100% of the configured amount of CPU, memory and so forth *all* the time. This practice is sometimes called *overcommit*. * **Specify resource priorities.** If the cluster is going to sacrifice services, it should be the ones you care about (comparatively) the least. Ensure that resource priorities are properly set so that your most important resources are scheduled first. diff --git a/include/crm/pengine/internal.h b/include/crm/pengine/internal.h index c0c142fff7..f470830553 100644 --- a/include/crm/pengine/internal.h +++ b/include/crm/pengine/internal.h @@ -1,742 +1,742 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PE_INTERNAL__H # define PE_INTERNAL__H # include # include # include # include # include # include # include # include # include const char *pe__resource_description(const pe_resource_t *rsc, uint32_t show_opts); enum pe__clone_flags { // Whether instances should be started sequentially pe__clone_ordered = (1 << 0), // Whether promotion scores have been added pe__clone_promotion_added = (1 << 1), // Whether promotion constraints have been added pe__clone_promotion_constrained = (1 << 2), }; bool pe__clone_is_ordered(const pe_resource_t *clone); int pe__set_clone_flag(pe_resource_t *clone, enum pe__clone_flags flag); enum pe__group_flags { pe__group_ordered = (1 << 0), // Members start sequentially pe__group_colocated = (1 << 1), // Members must be on same node }; bool pe__group_flag_is_set(const pe_resource_t *group, uint32_t flags); pe_resource_t *pe__last_group_member(const pe_resource_t *group); # define pe_rsc_info(rsc, fmt, args...) crm_log_tag(LOG_INFO, rsc ? rsc->id : "", fmt, ##args) # define pe_rsc_debug(rsc, fmt, args...) crm_log_tag(LOG_DEBUG, rsc ? rsc->id : "", fmt, ##args) # define pe_rsc_trace(rsc, fmt, args...) crm_log_tag(LOG_TRACE, rsc ? rsc->id : "", fmt, ##args) # define pe_err(fmt...) do { \ was_processing_error = TRUE; \ pcmk__config_err(fmt); \ } while (0) # define pe_warn(fmt...) do { \ was_processing_warning = TRUE; \ pcmk__config_warn(fmt); \ } while (0) # define pe_proc_err(fmt...) { was_processing_error = TRUE; crm_err(fmt); } # define pe_proc_warn(fmt...) { was_processing_warning = TRUE; crm_warn(fmt); } #define pe__set_working_set_flags(working_set, flags_to_set) do { \ (working_set)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", crm_system_name, \ (working_set)->flags, (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_working_set_flags(working_set, flags_to_clear) do { \ (working_set)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", crm_system_name, \ (working_set)->flags, (flags_to_clear), #flags_to_clear); \ } while (0) #define pe__set_resource_flags(resource, flags_to_set) do { \ (resource)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \ (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_resource_flags(resource, flags_to_clear) do { \ (resource)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Resource", (resource)->id, (resource)->flags, \ (flags_to_clear), #flags_to_clear); \ } while (0) #define pe__set_action_flags(action, flags_to_set) do { \ (action)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_action_flags(action, flags_to_clear) do { \ (action)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_raw_action_flags(action_flags, action_name, flags_to_set) do { \ action_flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Action", action_name, \ (action_flags), \ (flags_to_set), #flags_to_set); \ } while (0) #define pe__clear_raw_action_flags(action_flags, action_name, flags_to_clear) do { \ action_flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", action_name, \ (action_flags), \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_action_flags_as(function, line, action, flags_to_set) do { \ (action)->flags = pcmk__set_flags_as((function), (line), \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_action_flags_as(function, line, action, flags_to_clear) do { \ (action)->flags = pcmk__clear_flags_as((function), (line), \ LOG_TRACE, \ "Action", (action)->uuid, \ (action)->flags, \ (flags_to_clear), \ #flags_to_clear); \ } while (0) #define pe__set_order_flags(order_flags, flags_to_set) do { \ order_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Ordering", "constraint", \ order_flags, (flags_to_set), \ #flags_to_set); \ } while (0) #define pe__clear_order_flags(order_flags, flags_to_clear) do { \ order_flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Ordering", "constraint", \ order_flags, (flags_to_clear), \ #flags_to_clear); \ } while (0) // Some warnings we don't want to print every transition enum pe_warn_once_e { pe_wo_blind = (1 << 0), pe_wo_restart_type = (1 << 1), pe_wo_role_after = (1 << 2), pe_wo_poweroff = (1 << 3), pe_wo_require_all = (1 << 4), pe_wo_order_score = (1 << 5), pe_wo_neg_threshold = (1 << 6), pe_wo_remove_after = (1 << 7), pe_wo_ping_node = (1 << 8), pe_wo_order_inst = (1 << 9), pe_wo_coloc_inst = (1 << 10), pe_wo_group_order = (1 << 11), pe_wo_group_coloc = (1 << 12), pe_wo_upstart = (1 << 13), pe_wo_nagios = (1 << 14), }; extern uint32_t pe_wo; #define pe_warn_once(pe_wo_bit, fmt...) do { \ if (!pcmk_is_set(pe_wo, pe_wo_bit)) { \ if (pe_wo_bit == pe_wo_blind) { \ crm_warn(fmt); \ } else { \ pe_warn(fmt); \ } \ pe_wo = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Warn-once", "logging", pe_wo, \ (pe_wo_bit), #pe_wo_bit); \ } \ } while (0); typedef struct pe__location_constraint_s { char *id; // Constraint XML ID pe_resource_t *rsc_lh; // Resource being located enum rsc_role_e role_filter; // Role to locate enum pe_discover_e discover_mode; // Resource discovery GList *node_list_rh; // List of pe_node_t* } pe__location_t; typedef struct pe__order_constraint_s { int id; uint32_t flags; // Group of enum pe_ordering flags void *lh_opaque; pe_resource_t *lh_rsc; pe_action_t *lh_action; char *lh_action_task; void *rh_opaque; pe_resource_t *rh_rsc; pe_action_t *rh_action; char *rh_action_task; } pe__ordering_t; const pe_resource_t *pe__const_top_resource(const pe_resource_t *rsc, bool include_bundle); int pe__clone_max(const pe_resource_t *clone); int pe__clone_node_max(const pe_resource_t *clone); int pe__clone_promoted_max(const pe_resource_t *clone); int pe__clone_promoted_node_max(const pe_resource_t *clone); void pe__create_clone_notifications(pe_resource_t *clone); void pe__free_clone_notification_data(pe_resource_t *clone); void pe__create_clone_notif_pseudo_ops(pe_resource_t *clone, pe_action_t *start, pe_action_t *started, pe_action_t *stop, pe_action_t *stopped); pe_action_t *pe__new_rsc_pseudo_action(pe_resource_t *rsc, const char *task, bool optional, bool runnable); void pe__create_promotable_pseudo_ops(pe_resource_t *clone, bool any_promoting, bool any_demoting); bool pe_can_fence(const pe_working_set_t *data_set, const pe_node_t *node); void add_hash_param(GHashTable * hash, const char *name, const char *value); char *native_parameter(pe_resource_t * rsc, pe_node_t * node, gboolean create, const char *name, pe_working_set_t * data_set); pe_node_t *native_location(const pe_resource_t *rsc, GList **list, int current); void pe_metadata(pcmk__output_t *out); void verify_pe_options(GHashTable * options); void native_add_running(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set, gboolean failed); gboolean native_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean group_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean clone_unpack(pe_resource_t * rsc, pe_working_set_t * data_set); gboolean pe__unpack_bundle(pe_resource_t *rsc, pe_working_set_t *data_set); pe_resource_t *native_find_rsc(pe_resource_t *rsc, const char *id, const pe_node_t *node, int flags); gboolean native_active(pe_resource_t * rsc, gboolean all); gboolean group_active(pe_resource_t * rsc, gboolean all); gboolean clone_active(pe_resource_t * rsc, gboolean all); gboolean pe__bundle_active(pe_resource_t *rsc, gboolean all); //! \deprecated This function will be removed in a future release void native_print(pe_resource_t *rsc, const char *pre_text, long options, void *print_data); //! \deprecated This function will be removed in a future release void group_print(pe_resource_t *rsc, const char *pre_text, long options, void *print_data); //! \deprecated This function will be removed in a future release void clone_print(pe_resource_t *rsc, const char *pre_text, long options, void *print_data); //! \deprecated This function will be removed in a future release void pe__print_bundle(pe_resource_t *rsc, const char *pre_text, long options, void *print_data); gchar *pcmk__native_output_string(const pe_resource_t *rsc, const char *name, const pe_node_t *node, uint32_t show_opts, const char *target_role, bool show_nodes); int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name , size_t pairs_count, ...); char *pe__node_display_name(pe_node_t *node, bool print_detail); // Clone notifications (pe_notif.c) void pe__order_notifs_after_fencing(const pe_action_t *action, pe_resource_t *rsc, pe_action_t *stonith_op); static inline const char * pe__rsc_bool_str(const pe_resource_t *rsc, uint64_t rsc_flag) { return pcmk__btoa(pcmk_is_set(rsc->flags, rsc_flag)); } int pe__clone_xml(pcmk__output_t *out, va_list args); int pe__clone_default(pcmk__output_t *out, va_list args); int pe__group_xml(pcmk__output_t *out, va_list args); int pe__group_default(pcmk__output_t *out, va_list args); int pe__bundle_xml(pcmk__output_t *out, va_list args); int pe__bundle_html(pcmk__output_t *out, va_list args); int pe__bundle_text(pcmk__output_t *out, va_list args); int pe__node_html(pcmk__output_t *out, va_list args); int pe__node_text(pcmk__output_t *out, va_list args); int pe__node_xml(pcmk__output_t *out, va_list args); int pe__resource_xml(pcmk__output_t *out, va_list args); int pe__resource_html(pcmk__output_t *out, va_list args); int pe__resource_text(pcmk__output_t *out, va_list args); void native_free(pe_resource_t * rsc); void group_free(pe_resource_t * rsc); void clone_free(pe_resource_t * rsc); void pe__free_bundle(pe_resource_t *rsc); enum rsc_role_e native_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e group_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e clone_resource_state(const pe_resource_t * rsc, gboolean current); enum rsc_role_e pe__bundle_resource_state(const pe_resource_t *rsc, gboolean current); void pe__count_common(pe_resource_t *rsc); void pe__count_bundle(pe_resource_t *rsc); void common_free(pe_resource_t * rsc); pe_node_t *pe__copy_node(const pe_node_t *this_node); extern time_t get_effective_time(pe_working_set_t * data_set); /* Failure handling utilities (from failcounts.c) */ // bit flags for fail count handling options enum pe_fc_flags_e { pe_fc_default = (1 << 0), pe_fc_effective = (1 << 1), // don't count expired failures pe_fc_fillers = (1 << 2), // if container, include filler failures in count }; int pe_get_failcount(const pe_node_t *node, pe_resource_t *rsc, time_t *last_failure, uint32_t flags, const xmlNode *xml_op); pe_action_t *pe__clear_failcount(pe_resource_t *rsc, const pe_node_t *node, const char *reason, pe_working_set_t *data_set); /* Functions for finding/counting a resource's active nodes */ bool pe__count_active_node(const pe_resource_t *rsc, pe_node_t *node, pe_node_t **active, unsigned int *count_all, unsigned int *count_clean); pe_node_t *pe__find_active_requires(const pe_resource_t *rsc, unsigned int *count); static inline pe_node_t * pe__current_node(const pe_resource_t *rsc) { return (rsc == NULL)? NULL : rsc->fns->active_node(rsc, NULL, NULL); } /* Binary like operators for lists of nodes */ extern void node_list_exclude(GHashTable * list, GList *list2, gboolean merge_scores); GHashTable *pe__node_list2table(const GList *list); static inline gpointer pe_hash_table_lookup(GHashTable * hash, gconstpointer key) { if (hash) { return g_hash_table_lookup(hash, key); } return NULL; } extern pe_action_t *get_pseudo_op(const char *name, pe_working_set_t * data_set); extern gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order); -void pe__show_node_weights_as(const char *file, const char *function, - int line, bool to_log, const pe_resource_t *rsc, - const char *comment, GHashTable *nodes, - pe_working_set_t *data_set); - -#define pe__show_node_weights(level, rsc, text, nodes, data_set) \ - pe__show_node_weights_as(__FILE__, __func__, __LINE__, \ - (level), (rsc), (text), (nodes), (data_set)) +void pe__show_node_scores_as(const char *file, const char *function, + int line, bool to_log, const pe_resource_t *rsc, + const char *comment, GHashTable *nodes, + pe_working_set_t *data_set); + +#define pe__show_node_scores(level, rsc, text, nodes, data_set) \ + pe__show_node_scores_as(__FILE__, __func__, __LINE__, \ + (level), (rsc), (text), (nodes), (data_set)) xmlNode *find_rsc_op_entry(const pe_resource_t *rsc, const char *key); pe_action_t *custom_action(pe_resource_t *rsc, char *key, const char *task, const pe_node_t *on_node, gboolean optional, gboolean foo, pe_working_set_t *data_set); # define delete_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DELETE, 0) # define delete_action(rsc, node, optional) custom_action( \ rsc, delete_key(rsc), CRMD_ACTION_DELETE, node, \ optional, TRUE, rsc->cluster); # define stopped_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOPPED, 0) # define stopped_action(rsc, node, optional) custom_action( \ rsc, stopped_key(rsc), CRMD_ACTION_STOPPED, node, \ optional, TRUE, rsc->cluster); # define stop_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOP, 0) # define stop_action(rsc, node, optional) custom_action( \ rsc, stop_key(rsc), CRMD_ACTION_STOP, node, \ optional, TRUE, rsc->cluster); # define reload_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_RELOAD_AGENT, 0) # define start_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_START, 0) # define start_action(rsc, node, optional) custom_action( \ rsc, start_key(rsc), CRMD_ACTION_START, node, \ optional, TRUE, rsc->cluster) # define started_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STARTED, 0) # define started_action(rsc, node, optional) custom_action( \ rsc, started_key(rsc), CRMD_ACTION_STARTED, node, \ optional, TRUE, rsc->cluster) # define promote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTE, 0) # define promote_action(rsc, node, optional) custom_action( \ rsc, promote_key(rsc), CRMD_ACTION_PROMOTE, node, \ optional, TRUE, rsc->cluster) # define promoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTED, 0) # define promoted_action(rsc, node, optional) custom_action( \ rsc, promoted_key(rsc), CRMD_ACTION_PROMOTED, node, \ optional, TRUE, rsc->cluster) # define demote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTE, 0) # define demote_action(rsc, node, optional) custom_action( \ rsc, demote_key(rsc), CRMD_ACTION_DEMOTE, node, \ optional, TRUE, rsc->cluster) # define demoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTED, 0) # define demoted_action(rsc, node, optional) custom_action( \ rsc, demoted_key(rsc), CRMD_ACTION_DEMOTED, node, \ optional, TRUE, rsc->cluster) extern int pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set); pe_action_t *find_first_action(const GList *input, const char *uuid, const char *task, const pe_node_t *on_node); enum action_tasks get_complex_task(const pe_resource_t *rsc, const char *name); extern GList *find_actions(GList *input, const char *key, const pe_node_t *on_node); GList *find_actions_exact(GList *input, const char *key, const pe_node_t *on_node); GList *pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node, const char *task, bool require_node); extern void pe_free_action(pe_action_t * action); void resource_location(pe_resource_t *rsc, const pe_node_t *node, int score, const char *tag, pe_working_set_t *data_set); extern int pe__is_newer_op(const xmlNode *xml_a, const xmlNode *xml_b, bool same_node_default); extern gint sort_op_by_callid(gconstpointer a, gconstpointer b); gboolean get_target_role(const pe_resource_t *rsc, enum rsc_role_e *role); void pe__set_next_role(pe_resource_t *rsc, enum rsc_role_e role, const char *why); pe_resource_t *find_clone_instance(const pe_resource_t *rsc, const char *sub_id); extern void destroy_ticket(gpointer data); extern pe_ticket_t *ticket_new(const char *ticket_id, pe_working_set_t * data_set); // Resources for manipulating resource names const char *pe_base_name_end(const char *id); char *clone_strip(const char *last_rsc_id); char *clone_zero(const char *last_rsc_id); static inline bool pe_base_name_eq(const pe_resource_t *rsc, const char *id) { if (id && rsc && rsc->id) { // Number of characters in rsc->id before any clone suffix size_t base_len = pe_base_name_end(rsc->id) - rsc->id + 1; return (strlen(id) == base_len) && !strncmp(id, rsc->id, base_len); } return false; } int pe__target_rc_from_xml(const xmlNode *xml_op); gint pe__cmp_node_name(gconstpointer a, gconstpointer b); bool is_set_recursive(const pe_resource_t *rsc, long long flag, bool any); enum rsc_digest_cmp_val { /*! Digests are the same */ RSC_DIGEST_MATCH = 0, /*! Params that require a restart changed */ RSC_DIGEST_RESTART, /*! Some parameter changed. */ RSC_DIGEST_ALL, /*! rsc op didn't have a digest associated with it, so * it is unknown if parameters changed or not. */ RSC_DIGEST_UNKNOWN, }; typedef struct op_digest_cache_s { enum rsc_digest_cmp_val rc; xmlNode *params_all; xmlNode *params_secure; xmlNode *params_restart; char *digest_all_calc; char *digest_secure_calc; char *digest_restart_calc; } op_digest_cache_t; op_digest_cache_t *pe__calculate_digests(pe_resource_t *rsc, const char *task, guint *interval_ms, const pe_node_t *node, const xmlNode *xml_op, GHashTable *overrides, bool calc_secure, pe_working_set_t *data_set); void pe__free_digests(gpointer ptr); op_digest_cache_t *rsc_action_digest_cmp(pe_resource_t *rsc, const xmlNode *xml_op, pe_node_t *node, pe_working_set_t *data_set); pe_action_t *pe_fence_op(pe_node_t *node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t *data_set); void trigger_unfencing(pe_resource_t *rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t *data_set); char *pe__action2reason(const pe_action_t *action, enum pe_action_flags flag); void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite); void pe__add_action_expected_result(pe_action_t *action, int expected_result); void pe__set_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags); void pe__clear_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags); void pe__clear_resource_flags_on_all(pe_working_set_t *data_set, uint64_t flag); gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref); //! \deprecated This function will be removed in a future release void print_rscs_brief(GList *rsc_list, const char * pre_text, long options, void * print_data, gboolean print_all); int pe__rscs_brief_output(pcmk__output_t *out, GList *rsc_list, unsigned int options); void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay); pe_node_t *pe_create_node(const char *id, const char *uname, const char *type, const char *score, pe_working_set_t * data_set); //! \deprecated This function will be removed in a future release void common_print(pe_resource_t *rsc, const char *pre_text, const char *name, const pe_node_t *node, long options, void *print_data); int pe__common_output_text(pcmk__output_t *out, const pe_resource_t *rsc, const char *name, const pe_node_t *node, unsigned int options); int pe__common_output_html(pcmk__output_t *out, const pe_resource_t *rsc, const char *name, const pe_node_t *node, unsigned int options); //! A single instance of a bundle typedef struct { int offset; //!< 0-origin index of this instance in bundle char *ipaddr; //!< IP address associated with this instance pe_node_t *node; //!< Node created for this instance pe_resource_t *ip; //!< IP address resource for ipaddr pe_resource_t *child; //!< Instance of bundled resource pe_resource_t *container; //!< Container associated with this instance pe_resource_t *remote; //!< Pacemaker Remote connection into container } pe__bundle_replica_t; GList *pe__bundle_containers(const pe_resource_t *bundle); int pe__bundle_max(const pe_resource_t *rsc); bool pe__node_is_bundle_instance(const pe_resource_t *bundle, const pe_node_t *node); pe_resource_t *pe__bundled_resource(const pe_resource_t *rsc); const pe_resource_t *pe__get_rsc_in_container(const pe_resource_t *instance); pe_resource_t *pe__first_container(const pe_resource_t *bundle); void pe__foreach_bundle_replica(pe_resource_t *bundle, bool (*fn)(pe__bundle_replica_t *, void *), void *user_data); pe_resource_t *pe__find_bundle_replica(const pe_resource_t *bundle, const pe_node_t *node); bool pe__bundle_needs_remote_name(pe_resource_t *rsc); const char *pe__add_bundle_remote_name(pe_resource_t *rsc, pe_working_set_t *data_set, xmlNode *xml, const char *field); const char *pe_node_attribute_calculated(const pe_node_t *node, const char *name, const pe_resource_t *rsc); const char *pe_node_attribute_raw(const pe_node_t *node, const char *name); bool pe__is_universal_clone(const pe_resource_t *rsc, const pe_working_set_t *data_set); void pe__add_param_check(const xmlNode *rsc_op, pe_resource_t *rsc, pe_node_t *node, enum pe_check_parameters, pe_working_set_t *data_set); void pe__foreach_param_check(pe_working_set_t *data_set, void (*cb)(pe_resource_t*, pe_node_t*, const xmlNode*, enum pe_check_parameters)); void pe__free_param_checks(pe_working_set_t *data_set); bool pe__shutdown_requested(const pe_node_t *node); void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set); /*! * \internal * \brief Register xml formatting message functions. * * \param[in,out] out Output object to register messages with */ void pe__register_messages(pcmk__output_t *out); void pe__unpack_dataset_nvpairs(const xmlNode *xml_obj, const char *set_name, const pe_rule_eval_data_t *rule_data, GHashTable *hash, const char *always_first, gboolean overwrite, pe_working_set_t *data_set); bool pe__resource_is_disabled(const pe_resource_t *rsc); pe_action_t *pe__clear_resource_history(pe_resource_t *rsc, const pe_node_t *node, pe_working_set_t *data_set); GList *pe__rscs_with_tag(pe_working_set_t *data_set, const char *tag_name); GList *pe__unames_with_tag(pe_working_set_t *data_set, const char *tag_name); bool pe__rsc_has_tag(pe_working_set_t *data_set, const char *rsc, const char *tag); bool pe__uname_has_tag(pe_working_set_t *data_set, const char *node, const char *tag); bool pe__rsc_running_on_only(const pe_resource_t *rsc, const pe_node_t *node); bool pe__rsc_running_on_any(pe_resource_t *rsc, GList *node_list); GList *pe__filter_rsc_list(GList *rscs, GList *filter); GList * pe__build_node_name_list(pe_working_set_t *data_set, const char *s); GList * pe__build_rsc_list(pe_working_set_t *data_set, const char *s); bool pcmk__rsc_filtered_by_node(pe_resource_t *rsc, GList *only_node); gboolean pe__bundle_is_filtered(const pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__clone_is_filtered(const pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__group_is_filtered(const pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); gboolean pe__native_is_filtered(const pe_resource_t *rsc, GList *only_rsc, gboolean check_parent); xmlNode *pe__failed_probe_for_rsc(const pe_resource_t *rsc, const char *name); const char *pe__clone_child_id(const pe_resource_t *rsc); int pe__sum_node_health_scores(const pe_node_t *node, int base_health); int pe__node_health(pe_node_t *node); static inline enum pcmk__health_strategy pe__health_strategy(pe_working_set_t *data_set) { return pcmk__parse_health_strategy(pe_pref(data_set->config_hash, PCMK__OPT_NODE_HEALTH_STRATEGY)); } static inline int pe__health_score(const char *option, pe_working_set_t *data_set) { return char2score(pe_pref(data_set->config_hash, option)); } /*! * \internal * \brief Return a string suitable for logging as a node name * * \param[in] node Node to return a node name string for * * \return Node name if available, otherwise node ID if available, * otherwise "unspecified node" if node is NULL or "unidentified node" * if node has neither a name nor ID. */ static inline const char * pe__node_name(const pe_node_t *node) { if (node == NULL) { return "unspecified node"; } else if (node->details->uname != NULL) { return node->details->uname; } else if (node->details->id != NULL) { return node->details->id; } else { return "unidentified node"; } } /*! * \internal * \brief Check whether two node objects refer to the same node * * \param[in] node1 First node object to compare * \param[in] node2 Second node object to compare * * \return true if \p node1 and \p node2 refer to the same node */ static inline bool pe__same_node(const pe_node_t *node1, const pe_node_t *node2) { return (node1 != NULL) && (node2 != NULL) && (node1->details == node2->details); } /*! * \internal * \brief Get the operation key from an action history entry * * \param[in] xml Action history entry * * \return Entry's operation key */ static inline const char * pe__xe_history_key(const xmlNode *xml) { if (xml == NULL) { return NULL; } else { /* @COMPAT Pacemaker <= 1.1.5 did not add the key, and used the ID * instead. Checking for that allows us to process old saved CIBs, * including some regression tests. */ const char *key = crm_element_value(xml, XML_LRM_ATTR_TASK_KEY); return pcmk__str_empty(key)? ID(xml) : key; } } #endif diff --git a/include/pacemaker-internal.h b/include/pacemaker-internal.h index 8610d1ed4d..5cc904b845 100644 --- a/include/pacemaker-internal.h +++ b/include/pacemaker-internal.h @@ -1,27 +1,25 @@ /* - * Copyright 2019-2022 the Pacemaker project contributors + * Copyright 2019-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PACEMAKER_INTERNAL__H # define PACEMAKER_INTERNAL__H # include # include # include # include # include # include # include -# include -# include # include # include # include # include #endif diff --git a/include/pcmki/Makefile.am b/include/pcmki/Makefile.am index b379fdb0c0..824ef6c838 100644 --- a/include/pcmki/Makefile.am +++ b/include/pcmki/Makefile.am @@ -1,26 +1,24 @@ # -# Copyright 2019-2022 the Pacemaker project contributors +# Copyright 2019-2023 the Pacemaker project contributors # # The version control history for this file may have further details. # # This source code is licensed under the GNU General Public License version 2 # or later (GPLv2+) WITHOUT ANY WARRANTY. # MAINTAINERCLEANFILES = Makefile.in noinst_HEADERS = pcmki_acl.h \ pcmki_cluster_queries.h \ pcmki_fence.h \ pcmki_output.h \ pcmki_resource.h \ pcmki_result_code.h \ pcmki_rule.h \ - pcmki_sched_allocate.h \ - pcmki_sched_utils.h \ pcmki_scheduler.h \ pcmki_simulate.h \ pcmki_status.h \ pcmki_transition.h .PHONY: $(ARCHIVE_VERSION) diff --git a/include/pcmki/pcmki_cluster_queries.h b/include/pcmki/pcmki_cluster_queries.h index 776aa27937..3fa4c23f26 100644 --- a/include/pcmki/pcmki_cluster_queries.h +++ b/include/pcmki/pcmki_cluster_queries.h @@ -1,69 +1,70 @@ /* - * Copyright 2020-2022 the Pacemaker project contributors + * Copyright 2020-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__PCMKI_PCMKI_CLUSTER_QUERIES__H # define PCMK__PCMKI_PCMKI_CLUSTER_QUERIES__H -#include // gboolean, GMainLoop, etc. +#include +#include #include #include #include #include // CIB queries int pcmk__list_nodes(pcmk__output_t *out, const char *node_types, - gboolean bash_export); + bool bash_export); // Controller queries int pcmk__controller_status(pcmk__output_t *out, const char *node_name, unsigned int message_timeout_ms); int pcmk__designated_controller(pcmk__output_t *out, unsigned int message_timeout_ms); int pcmk__pacemakerd_status(pcmk__output_t *out, const char *ipc_name, unsigned int message_timeout_ms, bool show_output, enum pcmk_pacemakerd_state *state); int pcmk__query_node_info(pcmk__output_t *out, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms); /*! * \internal * \brief Get the node name corresponding to a node ID from the controller * * \param[in,out] out Output object * \param[in] node_id ID of node whose name to get (or 0 for * the local node) * \param[out] node_name If not \p NULL, where to store the node * name * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code * * \note The caller is responsible for freeing \p *node_name using \p free(). */ static inline int pcmk__query_node_name(pcmk__output_t *out, uint32_t nodeid, char **node_name, unsigned int message_timeout_ms) { return pcmk__query_node_info(out, &nodeid, node_name, NULL, NULL, NULL, NULL, false, message_timeout_ms); } // pacemakerd queries int pcmk__pacemakerd_status(pcmk__output_t *out, const char *ipc_name, unsigned int message_timeout_ms, bool show_output, enum pcmk_pacemakerd_state *state); #endif diff --git a/include/pcmki/pcmki_sched_allocate.h b/include/pcmki/pcmki_sched_allocate.h deleted file mode 100644 index 83440fe305..0000000000 --- a/include/pcmki/pcmki_sched_allocate.h +++ /dev/null @@ -1,24 +0,0 @@ -/* - * Copyright 2004-2023 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU Lesser General Public License - * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. - */ - -#ifndef PCMK__PCMKI_PCMKI_SCHED_ALLOCATE__H -# define PCMK__PCMKI_PCMKI_SCHED_ALLOCATE__H - -# include -# include -# include -# include -# include -# include -# include -# include - -void pcmk__log_transition_summary(const char *filename); - -#endif diff --git a/include/pcmki/pcmki_sched_utils.h b/include/pcmki/pcmki_sched_utils.h deleted file mode 100644 index 154856aea1..0000000000 --- a/include/pcmki/pcmki_sched_utils.h +++ /dev/null @@ -1,31 +0,0 @@ -/* - * Copyright 2004-2023 the Pacemaker project contributors - * - * The version control history for this file may have further details. - * - * This source code is licensed under the GNU Lesser General Public License - * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. - */ - -#ifndef PCMK__PCMKI_PCMKI_SCHED_UTILS__H -# define PCMK__PCMKI_PCMKI_SCHED_UTILS__H - -#include // bool -#include // GList, GHashTable, gboolean, guint -#include // lrmd_event_data_t -#include // cib_t -#include -#include -#include -#include -#include -#include - -/* Constraint helper functions */ -GList *pcmk__copy_node_list(const GList *list, bool reset); - -xmlNode *pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *event, - const char *caller_version, int target_rc, - const char *node, const char *origin); - -#endif diff --git a/include/pcmki/pcmki_scheduler.h b/include/pcmki/pcmki_scheduler.h index dde50a57e3..3f66cea0a8 100644 --- a/include/pcmki/pcmki_scheduler.h +++ b/include/pcmki/pcmki_scheduler.h @@ -1,43 +1,47 @@ /* * Copyright 2014-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__PCMKI_PCMKI_SCHEDULER__H -# define PCMK__PCMKI_PCMKI_SCHEDULER__H +#define PCMK__PCMKI_PCMKI_SCHEDULER__H -# include -# include -# include -# include -# include -# include +#include // GList +#include // bool +#include // xmlNode -# include +#include // lrmd_event_data_t +#include // pe_resource_t, pe_working_set_t typedef struct { const char *id; const char *node_attribute; pe_resource_t *dependent; // The resource being colocated pe_resource_t *primary; // The resource the dependent is colocated with int dependent_role; // Colocation applies only if dependent has this role int primary_role; // Colocation applies only if primary has this role int score; bool influence; // Whether dependent influences active primary placement } pcmk__colocation_t; void pcmk__unpack_constraints(pe_working_set_t *data_set); void pcmk__schedule_actions(xmlNode *cib, unsigned long long flags, pe_working_set_t *data_set); GList *pcmk__with_this_colocations(const pe_resource_t *rsc); GList *pcmk__this_with_colocations(const pe_resource_t *rsc); +GList *pcmk__copy_node_list(const GList *list, bool reset); + +xmlNode *pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *event, + const char *caller_version, int target_rc, + const char *node, const char *origin); + #endif diff --git a/include/pcmki/pcmki_transition.h b/include/pcmki/pcmki_transition.h index 5dc310179d..85f2d61be9 100644 --- a/include/pcmki/pcmki_transition.h +++ b/include/pcmki/pcmki_transition.h @@ -1,176 +1,177 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__PCMKI_PCMKI_TRANSITION__H # define PCMK__PCMKI_PCMKI_TRANSITION__H # include # include # include # include #ifdef __cplusplus extern "C" { #endif enum pcmk__graph_action_type { pcmk__pseudo_graph_action, pcmk__rsc_graph_action, pcmk__cluster_graph_action, }; enum pcmk__synapse_flags { pcmk__synapse_ready = (1 << 0), pcmk__synapse_failed = (1 << 1), pcmk__synapse_executed = (1 << 2), pcmk__synapse_confirmed = (1 << 3), }; typedef struct { int id; int priority; uint32_t flags; // Group of pcmk__synapse_flags GList *actions; /* pcmk__graph_action_t* */ GList *inputs; /* pcmk__graph_action_t* */ } pcmk__graph_synapse_t; #define pcmk__set_synapse_flags(synapse, flags_to_set) do { \ (synapse)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Synapse", "synapse", \ (synapse)->flags, (flags_to_set), #flags_to_set); \ } while (0) #define pcmk__clear_synapse_flags(synapse, flags_to_clear) do { \ (synapse)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Synapse", "synapse", \ (synapse)->flags, (flags_to_clear), #flags_to_clear); \ } while (0) enum pcmk__graph_action_flags { pcmk__graph_action_sent_update = (1 << 0), /* sent to the CIB */ pcmk__graph_action_executed = (1 << 1), /* sent to the CRM */ pcmk__graph_action_confirmed = (1 << 2), pcmk__graph_action_failed = (1 << 3), pcmk__graph_action_can_fail = (1 << 4), //! \deprecated Will be removed in a future release }; typedef struct { int id; int timeout; int timer; guint interval_ms; GHashTable *params; enum pcmk__graph_action_type type; pcmk__graph_synapse_t *synapse; uint32_t flags; // Group of pcmk__graph_action_flags xmlNode *xml; } pcmk__graph_action_t; #define pcmk__set_graph_action_flags(action, flags_to_set) do { \ (action)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", "action", \ (action)->flags, (flags_to_set), #flags_to_set); \ } while (0) #define pcmk__clear_graph_action_flags(action, flags_to_clear) do { \ (action)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Action", "action", \ (action)->flags, (flags_to_clear), #flags_to_clear); \ } while (0) // What to do after finished processing a transition graph enum pcmk__graph_next { // Order matters: lowest priority to highest pcmk__graph_done, // Transition complete, nothing further needed pcmk__graph_wait, // Transition interrupted, wait for further changes pcmk__graph_restart, // Transition interrupted, start a new one pcmk__graph_shutdown, // Transition interrupted, local shutdown needed }; typedef struct { int id; char *source; int abort_priority; bool complete; const char *abort_reason; enum pcmk__graph_next completion_action; int num_actions; int num_synapses; int batch_limit; guint network_delay; guint stonith_timeout; int fired; int pending; int skipped; int completed; int incomplete; GList *synapses; /* pcmk__graph_synapse_t* */ int migration_limit; //! Failcount after one failed stop action char *failed_stop_offset; //! Failcount after one failed start action char *failed_start_offset; //! Time (from epoch) by which the controller should re-run the scheduler time_t recheck_by; } pcmk__graph_t; typedef struct { int (*pseudo) (pcmk__graph_t *graph, pcmk__graph_action_t *action); int (*rsc) (pcmk__graph_t *graph, pcmk__graph_action_t *action); int (*cluster) (pcmk__graph_t *graph, pcmk__graph_action_t *action); int (*fence) (pcmk__graph_t *graph, pcmk__graph_action_t *action); bool (*allowed) (pcmk__graph_t *graph, pcmk__graph_action_t *action); } pcmk__graph_functions_t; enum pcmk__graph_status { pcmk__graph_active, // Some actions have been performed pcmk__graph_pending, // No actions performed yet pcmk__graph_complete, pcmk__graph_terminated, }; void pcmk__set_graph_functions(pcmk__graph_functions_t *fns); pcmk__graph_t *pcmk__unpack_graph(const xmlNode *xml_graph, const char *reference); enum pcmk__graph_status pcmk__execute_graph(pcmk__graph_t *graph); void pcmk__update_graph(pcmk__graph_t *graph, const pcmk__graph_action_t *action); void pcmk__free_graph(pcmk__graph_t *graph); const char *pcmk__graph_status2text(enum pcmk__graph_status state); void pcmk__log_graph(unsigned int log_level, pcmk__graph_t *graph); void pcmk__log_graph_action(int log_level, pcmk__graph_action_t *action); +void pcmk__log_transition_summary(const char *filename); lrmd_event_data_t *pcmk__event_from_graph_action(const xmlNode *resource, const pcmk__graph_action_t *action, int status, int rc, const char *exit_reason); #ifdef __cplusplus } #endif #endif diff --git a/lib/common/output_xml.c b/lib/common/output_xml.c index e34223fd53..e22b601c9d 100644 --- a/lib/common/output_xml.c +++ b/lib/common/output_xml.c @@ -1,542 +1,542 @@ /* - * Copyright 2019-2022 the Pacemaker project contributors + * Copyright 2019-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include /* pcmk__xml2fd */ #include #include #include static gboolean legacy_xml = FALSE; static gboolean simple_list = FALSE; static gboolean substitute = FALSE; GOptionEntry pcmk__xml_output_entries[] = { { "xml-legacy", 0, G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &legacy_xml, NULL, NULL }, { "xml-simple-list", 0, G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &simple_list, NULL, NULL }, { "xml-substitute", 0, G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &substitute, NULL, NULL }, { NULL } }; typedef struct subst_s { const char *from; const char *to; } subst_t; static subst_t substitutions[] = { { "Active Resources", "resources" }, - { "Allocation Scores", "allocations" }, - { "Allocation Scores and Utilization Information", "allocations_utilizations" }, + { "Assignment Scores", "allocations" }, + { "Assignment Scores and Utilization Information", "allocations_utilizations" }, { "Cluster Summary", "summary" }, { "Current cluster status", "cluster_status" }, { "Executing Cluster Transition", "transition" }, { "Failed Resource Actions", "failures" }, { "Fencing History", "fence_history" }, { "Full List of Resources", "resources" }, { "Inactive Resources", "resources" }, { "Migration Summary", "node_history" }, { "Negative Location Constraints", "bans" }, { "Node Attributes", "node_attributes" }, { "Operations", "node_history" }, { "Resource Config", "resource_config" }, { "Resource Operations", "operations" }, { "Revised Cluster Status", "revised_cluster_status" }, { "Transition Summary", "actions" }, { "Utilization Information", "utilizations" }, { NULL, NULL } }; /* The first several elements of this struct must be the same as the first * several elements of private_data_s in lib/common/output_html.c. That * struct gets passed to a bunch of the pcmk__output_xml_* functions which * assume an XML private_data_s. Keeping them laid out the same means this * still works. */ typedef struct private_data_s { /* Begin members that must match the HTML version */ xmlNode *root; GQueue *parent_q; GSList *errors; /* End members that must match the HTML version */ bool legacy_xml; } private_data_t; static void xml_free_priv(pcmk__output_t *out) { private_data_t *priv = NULL; if (out == NULL || out->priv == NULL) { return; } priv = out->priv; free_xml(priv->root); g_queue_free(priv->parent_q); g_slist_free(priv->errors); free(priv); out->priv = NULL; } static bool xml_init(pcmk__output_t *out) { private_data_t *priv = NULL; CRM_ASSERT(out != NULL); /* If xml_init was previously called on this output struct, just return. */ if (out->priv != NULL) { return true; } else { out->priv = calloc(1, sizeof(private_data_t)); if (out->priv == NULL) { return false; } priv = out->priv; } if (legacy_xml) { priv->root = create_xml_node(NULL, "crm_mon"); crm_xml_add(priv->root, "version", PACEMAKER_VERSION); } else { priv->root = create_xml_node(NULL, "pacemaker-result"); crm_xml_add(priv->root, "api-version", PCMK__API_VERSION); if (out->request != NULL) { crm_xml_add(priv->root, "request", out->request); } } priv->parent_q = g_queue_new(); priv->errors = NULL; g_queue_push_tail(priv->parent_q, priv->root); /* Copy this from the file-level variable. This means that it is only settable * as a command line option, and that pcmk__output_new must be called after all * command line processing is completed. */ priv->legacy_xml = legacy_xml; return true; } static void add_error_node(gpointer data, gpointer user_data) { char *str = (char *) data; xmlNodePtr node = (xmlNodePtr) user_data; pcmk_create_xml_text_node(node, "error", str); } static void xml_finish(pcmk__output_t *out, crm_exit_t exit_status, bool print, void **copy_dest) { private_data_t *priv = NULL; xmlNodePtr node; CRM_ASSERT(out != NULL); priv = out->priv; /* If root is NULL, xml_init failed and we are being called from pcmk__output_free * in the pcmk__output_new path. */ if (priv == NULL || priv->root == NULL) { return; } if (legacy_xml) { GSList *node = priv->errors; if (exit_status != CRM_EX_OK) { fprintf(stderr, "%s\n", crm_exit_str(exit_status)); } while (node != NULL) { fprintf(stderr, "%s\n", (char *) node->data); node = node->next; } } else { char *rc_as_str = pcmk__itoa(exit_status); node = create_xml_node(priv->root, "status"); pcmk__xe_set_props(node, "code", rc_as_str, "message", crm_exit_str(exit_status), NULL); if (g_slist_length(priv->errors) > 0) { xmlNodePtr errors_node = create_xml_node(node, "errors"); g_slist_foreach(priv->errors, add_error_node, (gpointer) errors_node); } free(rc_as_str); } if (print) { pcmk__xml2fd(fileno(out->dest), priv->root); } if (copy_dest != NULL) { *copy_dest = copy_xml(priv->root); } } static void xml_reset(pcmk__output_t *out) { CRM_ASSERT(out != NULL); out->dest = freopen(NULL, "w", out->dest); CRM_ASSERT(out->dest != NULL); xml_free_priv(out); xml_init(out); } static void xml_subprocess_output(pcmk__output_t *out, int exit_status, const char *proc_stdout, const char *proc_stderr) { xmlNodePtr node, child_node; char *rc_as_str = NULL; CRM_ASSERT(out != NULL); rc_as_str = pcmk__itoa(exit_status); node = pcmk__output_xml_create_parent(out, "command", "code", rc_as_str, NULL); if (proc_stdout != NULL) { child_node = pcmk_create_xml_text_node(node, "output", proc_stdout); crm_xml_add(child_node, "source", "stdout"); } if (proc_stderr != NULL) { child_node = pcmk_create_xml_text_node(node, "output", proc_stderr); crm_xml_add(child_node, "source", "stderr"); } free(rc_as_str); } static void xml_version(pcmk__output_t *out, bool extended) { CRM_ASSERT(out != NULL); pcmk__output_create_xml_node(out, "version", "program", "Pacemaker", "version", PACEMAKER_VERSION, "author", "Andrew Beekhof and the " "Pacemaker project contributors", "build", BUILD_VERSION, "features", CRM_FEATURES, NULL); } G_GNUC_PRINTF(2, 3) static void xml_err(pcmk__output_t *out, const char *format, ...) { private_data_t *priv = NULL; int len = 0; char *buf = NULL; va_list ap; CRM_ASSERT(out != NULL && out->priv != NULL); priv = out->priv; va_start(ap, format); len = vasprintf(&buf, format, ap); CRM_ASSERT(len > 0); va_end(ap); priv->errors = g_slist_append(priv->errors, buf); } G_GNUC_PRINTF(2, 3) static int xml_info(pcmk__output_t *out, const char *format, ...) { return pcmk_rc_no_output; } static void xml_output_xml(pcmk__output_t *out, const char *name, const char *buf) { xmlNodePtr parent = NULL; xmlNodePtr cdata_node = NULL; CRM_ASSERT(out != NULL); parent = pcmk__output_create_xml_node(out, name, NULL); cdata_node = xmlNewCDataBlock(getDocPtr(parent), (pcmkXmlStr) buf, strlen(buf)); xmlAddChild(parent, cdata_node); } G_GNUC_PRINTF(4, 5) static void xml_begin_list(pcmk__output_t *out, const char *singular_noun, const char *plural_noun, const char *format, ...) { va_list ap; char *name = NULL; char *buf = NULL; int len; CRM_ASSERT(out != NULL); va_start(ap, format); len = vasprintf(&buf, format, ap); CRM_ASSERT(len >= 0); va_end(ap); if (substitute) { for (subst_t *s = substitutions; s->from != NULL; s++) { if (!strcmp(s->from, buf)) { name = g_strdup(s->to); break; } } } if (name == NULL) { name = g_ascii_strdown(buf, -1); } if (legacy_xml || simple_list) { pcmk__output_xml_create_parent(out, name, NULL); } else { pcmk__output_xml_create_parent(out, "list", "name", name, NULL); } g_free(name); free(buf); } G_GNUC_PRINTF(3, 4) static void xml_list_item(pcmk__output_t *out, const char *name, const char *format, ...) { xmlNodePtr item_node = NULL; va_list ap; char *buf = NULL; int len; CRM_ASSERT(out != NULL); va_start(ap, format); len = vasprintf(&buf, format, ap); CRM_ASSERT(len >= 0); va_end(ap); item_node = pcmk__output_create_xml_text_node(out, "item", buf); if (name != NULL) { crm_xml_add(item_node, "name", name); } free(buf); } static void xml_increment_list(pcmk__output_t *out) { /* This function intentially left blank */ } static void xml_end_list(pcmk__output_t *out) { private_data_t *priv = NULL; CRM_ASSERT(out != NULL && out->priv != NULL); priv = out->priv; if (priv->legacy_xml || simple_list) { g_queue_pop_tail(priv->parent_q); } else { char *buf = NULL; xmlNodePtr node; node = g_queue_pop_tail(priv->parent_q); buf = crm_strdup_printf("%lu", xmlChildElementCount(node)); crm_xml_add(node, "count", buf); free(buf); } } static bool xml_is_quiet(pcmk__output_t *out) { return false; } static void xml_spacer(pcmk__output_t *out) { /* This function intentionally left blank */ } static void xml_progress(pcmk__output_t *out, bool end) { /* This function intentionally left blank */ } pcmk__output_t * pcmk__mk_xml_output(char **argv) { pcmk__output_t *retval = calloc(1, sizeof(pcmk__output_t)); if (retval == NULL) { return NULL; } retval->fmt_name = "xml"; retval->request = pcmk__quote_cmdline(argv); retval->init = xml_init; retval->free_priv = xml_free_priv; retval->finish = xml_finish; retval->reset = xml_reset; retval->register_message = pcmk__register_message; retval->message = pcmk__call_message; retval->subprocess_output = xml_subprocess_output; retval->version = xml_version; retval->info = xml_info; retval->transient = xml_info; retval->err = xml_err; retval->output_xml = xml_output_xml; retval->begin_list = xml_begin_list; retval->list_item = xml_list_item; retval->increment_list = xml_increment_list; retval->end_list = xml_end_list; retval->is_quiet = xml_is_quiet; retval->spacer = xml_spacer; retval->progress = xml_progress; retval->prompt = pcmk__text_prompt; return retval; } xmlNodePtr pcmk__output_xml_create_parent(pcmk__output_t *out, const char *name, ...) { va_list args; xmlNodePtr node = NULL; CRM_ASSERT(out != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return NULL); node = pcmk__output_create_xml_node(out, name, NULL); va_start(args, name); pcmk__xe_set_propv(node, args); va_end(args); pcmk__output_xml_push_parent(out, node); return node; } void pcmk__output_xml_add_node_copy(pcmk__output_t *out, xmlNodePtr node) { private_data_t *priv = NULL; xmlNodePtr parent = NULL; CRM_ASSERT(out != NULL && out->priv != NULL); CRM_ASSERT(node != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return); priv = out->priv; parent = g_queue_peek_tail(priv->parent_q); // Shouldn't happen unless the caller popped priv->root CRM_CHECK(parent != NULL, return); add_node_copy(parent, node); } xmlNodePtr pcmk__output_create_xml_node(pcmk__output_t *out, const char *name, ...) { xmlNodePtr node = NULL; private_data_t *priv = NULL; va_list args; CRM_ASSERT(out != NULL && out->priv != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return NULL); priv = out->priv; node = create_xml_node(g_queue_peek_tail(priv->parent_q), name); va_start(args, name); pcmk__xe_set_propv(node, args); va_end(args); return node; } xmlNodePtr pcmk__output_create_xml_text_node(pcmk__output_t *out, const char *name, const char *content) { xmlNodePtr node = NULL; CRM_ASSERT(out != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return NULL); node = pcmk__output_create_xml_node(out, name, NULL); xmlNodeSetContent(node, (pcmkXmlStr) content); return node; } void pcmk__output_xml_push_parent(pcmk__output_t *out, xmlNodePtr parent) { private_data_t *priv = NULL; CRM_ASSERT(out != NULL && out->priv != NULL); CRM_ASSERT(parent != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return); priv = out->priv; g_queue_push_tail(priv->parent_q, parent); } void pcmk__output_xml_pop_parent(pcmk__output_t *out) { private_data_t *priv = NULL; CRM_ASSERT(out != NULL && out->priv != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return); priv = out->priv; CRM_ASSERT(g_queue_get_length(priv->parent_q) > 0); g_queue_pop_tail(priv->parent_q); } xmlNodePtr pcmk__output_xml_peek_parent(pcmk__output_t *out) { private_data_t *priv = NULL; CRM_ASSERT(out != NULL && out->priv != NULL); CRM_CHECK(pcmk__str_any_of(out->fmt_name, "xml", "html", NULL), return NULL); priv = out->priv; /* If queue is empty NULL will be returned */ return g_queue_peek_tail(priv->parent_q); } diff --git a/lib/pacemaker/libpacemaker_private.h b/lib/pacemaker/libpacemaker_private.h index 4adaffdd99..57ec35e139 100644 --- a/lib/pacemaker/libpacemaker_private.h +++ b/lib/pacemaker/libpacemaker_private.h @@ -1,1055 +1,1056 @@ /* * Copyright 2021-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__LIBPACEMAKER_PRIVATE__H # define PCMK__LIBPACEMAKER_PRIVATE__H /* This header is for the sole use of libpacemaker, so that functions can be * declared with G_GNUC_INTERNAL for efficiency. */ #include // pe_action_t, pe_node_t, pe_working_set_t +#include // pe__location_t // Flags to modify the behavior of add_colocated_node_scores() enum pcmk__coloc_select { // With no other flags, apply all "with this" colocations pcmk__coloc_select_default = 0, // Apply "this with" colocations instead of "with this" colocations pcmk__coloc_select_this_with = (1 << 0), // Apply only colocations with non-negative scores pcmk__coloc_select_nonnegative = (1 << 1), // Apply only colocations with at least one matching node pcmk__coloc_select_active = (1 << 2), }; // Flags the update_ordered_actions() method can return enum pcmk__updated { pcmk__updated_none = 0, // Nothing changed pcmk__updated_first = (1 << 0), // First action was updated pcmk__updated_then = (1 << 1), // Then action was updated }; #define pcmk__set_updated_flags(au_flags, action, flags_to_set) do { \ au_flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Action update", \ (action)->uuid, au_flags, \ (flags_to_set), #flags_to_set); \ } while (0) #define pcmk__clear_updated_flags(au_flags, action, flags_to_clear) do { \ au_flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Action update", \ (action)->uuid, au_flags, \ (flags_to_clear), #flags_to_clear); \ } while (0) -// Resource allocation methods +// Resource assignment methods struct resource_alloc_functions_s { /*! * \internal * \brief Assign a resource to a node * * \param[in,out] rsc Resource to assign to a node * \param[in] prefer Node to prefer, if all else is equal * * \return Node that \p rsc is assigned to, if assigned entirely to one node */ pe_node_t *(*assign)(pe_resource_t *rsc, const pe_node_t *prefer); /*! * \internal * \brief Create all actions needed for a given resource * * \param[in,out] rsc Resource to create actions for */ void (*create_actions)(pe_resource_t *rsc); /*! * \internal * \brief Schedule any probes needed for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return true if any probe was created, otherwise false */ bool (*create_probe)(pe_resource_t *rsc, pe_node_t *node); /*! * \internal * \brief Create implicit constraints needed for a resource * * \param[in,out] rsc Resource to create implicit constraints for */ void (*internal_constraints)(pe_resource_t *rsc); /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint, apply its score to the dependent's - * allowed node weights (if we are still placing resources) or priority (if + * allowed node scores (if we are still placing resources) or priority (if * we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply * \param[in] for_dependent true if called on behalf of dependent */ void (*apply_coloc_score)(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent); /*! * \internal * \brief Create list of all resources in colocations with a given resource * * Given a resource, create a list of all resources involved in mandatory * colocations with it, whether directly or indirectly via chained colocations. * * \param[in] rsc Resource to add to colocated list * \param[in] orig_rsc Resource originally requested * \param[in,out] colocated_rscs Existing list * * \return List of given resource and all resources involved in colocations * * \note This function is recursive; top-level callers should pass NULL as * \p colocated_rscs and \p orig_rsc, and the desired resource as * \p rsc. The recursive calls will use other values. */ GList *(*colocated_resources)(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *colocated_rscs); /*! * \internal * \brief Add colocations affecting a resource as primary to a list * * Given a resource being assigned (\p orig_rsc) and a resource somewhere in * its chain of ancestors (\p rsc, which may be \p orig_rsc), get * colocations that affect the ancestor as primary and should affect the * resource, and add them to a given list. * * \param[in] rsc Resource whose colocations should be added * \param[in] orig_rsc Affected resource (\p rsc or a descendant) * \param[in,out] list List of colocations to add to * * \note All arguments should be non-NULL. * \note The pcmk__with_this_colocations() wrapper should usually be used * instead of using this method directly. */ void (*with_this_colocations)(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); /*! * \internal * \brief Add colocations affecting a resource as dependent to a list * * Given a resource being assigned (\p orig_rsc) and a resource somewhere in * its chain of ancestors (\p rsc, which may be \p orig_rsc), get * colocations that affect the ancestor as dependent and should affect the * resource, and add them to a given list. * * * \param[in] rsc Resource whose colocations should be added * \param[in] orig_rsc Affected resource (\p rsc or a descendant) * \param[in,out] list List of colocations to add to * * \note All arguments should be non-NULL. * \note The pcmk__this_with_colocations() wrapper should usually be used * instead of using this method directly. */ void (*this_with_colocations)(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); /*! * \internal * \brief Update nodes with scores of colocated resources' nodes * * Given a table of nodes and a resource, update the nodes' scores with the * scores of the best nodes matching the attribute used for each of the * resource's relevant colocations. * * \param[in,out] rsc Resource to check colocations for * \param[in] log_id Resource ID for logs (if NULL, use \p rsc ID) * \param[in,out] nodes Nodes to update (set initial contents to NULL * to copy \p rsc's allowed nodes) * \param[in] colocation Original colocation constraint (used to get * configured primary resource's stickiness, and * to get colocation node attribute; if NULL, * \p rsc's own matching node scores will not be * added, and *nodes must be NULL as well) * \param[in] factor Incorporate scores multiplied by this factor * \param[in] flags Bitmask of enum pcmk__coloc_select values * * \note NULL *nodes, NULL colocation, and the pcmk__coloc_select_this_with * flag are used together (and only by cmp_resources()). * \note The caller remains responsible for freeing \p *nodes. */ void (*add_colocated_node_scores)(pe_resource_t *rsc, const char *log_id, GHashTable **nodes, pcmk__colocation_t *colocation, float factor, uint32_t flags); /*! * \internal * \brief Apply a location constraint to a resource's allowed node scores * * \param[in,out] rsc Resource to apply constraint to * \param[in,out] location Location constraint to apply */ void (*apply_location)(pe_resource_t *rsc, pe__location_t *location); /*! * \internal * \brief Return action flags for a given resource action * * \param[in,out] action Action to get flags for * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node * \note For primitives, this will be the same as action->flags regardless * of node. For collective resources, the flags can differ due to * multiple instances possibly being involved. */ uint32_t (*action_flags)(pe_action_t *action, const pe_node_t *node); /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two actions, update the actions' * flags (and runnable_before members if appropriate) as appropriate for the * ordering. Effects may cascade to other orderings involving the actions as * well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this * node (only used when interleaving instances) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates * (may include pe_action_optional to affect only * mandatory actions, and pe_action_runnable to * affect only runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * \param[in,out] data_set Cluster working set * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t (*update_ordered_actions)(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set); /*! * \internal * \brief Output a summary of scheduled actions for a resource * * \param[in,out] rsc Resource to output actions for */ void (*output_actions)(pe_resource_t *rsc); /*! * \internal * \brief Add a resource's actions to the transition graph * * \param[in,out] rsc Resource whose actions should be added */ void (*add_actions_to_graph)(pe_resource_t *rsc); /*! * \internal * \brief Add meta-attributes relevant to transition graph actions to XML * * If a given resource supports variant-specific meta-attributes that are * needed for transition graph actions, add them to a given XML element. * * \param[in] rsc Resource whose meta-attributes should be added * \param[in,out] xml Transition graph action attributes XML to add to */ void (*add_graph_meta)(const pe_resource_t *rsc, xmlNode *xml); /*! * \internal * \brief Add a resource's utilization to a table of utilization values * * This function is used when summing the utilization of a resource and all * resources colocated with it, to determine whether a node has sufficient * capacity. Given a resource and a table of utilization values, it will add * the resource's utilization to the existing values, if the resource has - * not yet been allocated to a node. + * not yet been assigned to a node. * * \param[in] rsc Resource with utilization to add - * \param[in] orig_rsc Resource being allocated (for logging only) + * \param[in] orig_rsc Resource being assigned (for logging only) * \param[in] all_rscs List of all resources that will be summed * \param[in,out] utilization Table of utilization values to add to */ void (*add_utilization)(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization); /*! * \internal * \brief Apply a shutdown lock for a resource, if appropriate * * \param[in,out] rsc Resource to check for shutdown lock */ void (*shutdown_lock)(pe_resource_t *rsc); }; // Actions (pcmk_sched_actions.c) G_GNUC_INTERNAL void pcmk__update_action_for_orderings(pe_action_t *action, pe_working_set_t *data_set); G_GNUC_INTERNAL uint32_t pcmk__update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__log_action(const char *pre_text, const pe_action_t *action, bool details); G_GNUC_INTERNAL pe_action_t *pcmk__new_cancel_action(pe_resource_t *rsc, const char *name, guint interval_ms, const pe_node_t *node); G_GNUC_INTERNAL pe_action_t *pcmk__new_shutdown_action(pe_node_t *node); G_GNUC_INTERNAL bool pcmk__action_locks_rsc_to_node(const pe_action_t *action); G_GNUC_INTERNAL void pcmk__deduplicate_action_inputs(pe_action_t *action); G_GNUC_INTERNAL void pcmk__output_actions(pe_working_set_t *data_set); G_GNUC_INTERNAL bool pcmk__check_action_config(pe_resource_t *rsc, pe_node_t *node, const xmlNode *xml_op); G_GNUC_INTERNAL void pcmk__handle_rsc_config_changes(pe_working_set_t *data_set); // Recurring actions (pcmk_sched_recurring.c) G_GNUC_INTERNAL void pcmk__create_recurring_actions(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__schedule_cancel(pe_resource_t *rsc, const char *call_id, const char *task, guint interval_ms, const pe_node_t *node, const char *reason); G_GNUC_INTERNAL void pcmk__reschedule_recurring(pe_resource_t *rsc, const char *task, guint interval_ms, pe_node_t *node); G_GNUC_INTERNAL bool pcmk__action_is_recurring(const pe_action_t *action); // Producing transition graphs (pcmk_graph_producer.c) G_GNUC_INTERNAL bool pcmk__graph_has_loop(const pe_action_t *init_action, const pe_action_t *action, pe_action_wrapper_t *input); G_GNUC_INTERNAL void pcmk__add_rsc_actions_to_graph(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__create_graph(pe_working_set_t *data_set); // Fencing (pcmk_sched_fencing.c) G_GNUC_INTERNAL void pcmk__order_vs_fence(pe_action_t *stonith_op, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__order_vs_unfence(const pe_resource_t *rsc, pe_node_t *node, pe_action_t *action, enum pe_ordering order); G_GNUC_INTERNAL void pcmk__fence_guest(pe_node_t *node); G_GNUC_INTERNAL bool pcmk__node_unfenced(const pe_node_t *node); G_GNUC_INTERNAL void pcmk__order_restart_vs_unfence(gpointer data, gpointer user_data); // Injected scheduler inputs (pcmk_sched_injections.c) void pcmk__inject_scheduler_input(pe_working_set_t *data_set, cib_t *cib, const pcmk_injections_t *injections); // Constraints of any type (pcmk_sched_constraints.c) G_GNUC_INTERNAL pe_resource_t *pcmk__find_constraint_resource(GList *rsc_list, const char *id); G_GNUC_INTERNAL xmlNode *pcmk__expand_tags_in_sets(xmlNode *xml_obj, const pe_working_set_t *data_set); G_GNUC_INTERNAL bool pcmk__valid_resource_or_tag(const pe_working_set_t *data_set, const char *id, pe_resource_t **rsc, pe_tag_t **tag); G_GNUC_INTERNAL bool pcmk__tag_to_set(xmlNode *xml_obj, xmlNode **rsc_set, const char *attr, bool convert_rsc, const pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__create_internal_constraints(pe_working_set_t *data_set); // Location constraints G_GNUC_INTERNAL void pcmk__unpack_location(xmlNode *xml_obj, pe_working_set_t *data_set); G_GNUC_INTERNAL pe__location_t *pcmk__new_location(const char *id, pe_resource_t *rsc, - int node_weight, const char *discover_mode, + int node_score, const char *discover_mode, pe_node_t *foo_node, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__apply_locations(pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__apply_location(pe_resource_t *rsc, pe__location_t *constraint); // Colocation constraints (pcmk_sched_colocation.c) enum pcmk__coloc_affects { pcmk__coloc_affects_nothing = 0, pcmk__coloc_affects_location, pcmk__coloc_affects_role, }; G_GNUC_INTERNAL enum pcmk__coloc_affects pcmk__colocation_affects(const pe_resource_t *dependent, const pe_resource_t *primary, const pcmk__colocation_t *colocation, bool preview); G_GNUC_INTERNAL -void pcmk__apply_coloc_to_weights(pe_resource_t *dependent, - const pe_resource_t *primary, - const pcmk__colocation_t *colocation); +void pcmk__apply_coloc_to_scores(pe_resource_t *dependent, + const pe_resource_t *primary, + const pcmk__colocation_t *colocation); G_GNUC_INTERNAL void pcmk__apply_coloc_to_priority(pe_resource_t *dependent, const pe_resource_t *primary, const pcmk__colocation_t *colocation); G_GNUC_INTERNAL void pcmk__add_colocated_node_scores(pe_resource_t *rsc, const char *log_id, GHashTable **nodes, pcmk__colocation_t *colocation, float factor, uint32_t flags); G_GNUC_INTERNAL void pcmk__add_dependent_scores(gpointer data, gpointer user_data); G_GNUC_INTERNAL void pcmk__unpack_colocation(xmlNode *xml_obj, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__add_this_with(GList **list, const pcmk__colocation_t *colocation); G_GNUC_INTERNAL void pcmk__add_this_with_list(GList **list, GList *addition); G_GNUC_INTERNAL void pcmk__add_with_this(GList **list, const pcmk__colocation_t *colocation); G_GNUC_INTERNAL void pcmk__add_with_this_list(GList **list, GList *addition); G_GNUC_INTERNAL void pcmk__new_colocation(const char *id, const char *node_attr, int score, pe_resource_t *dependent, pe_resource_t *primary, const char *dependent_role, const char *primary_role, bool influence, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__block_colocation_dependents(pe_action_t *action, pe_working_set_t *data_set); /*! * \internal * \brief Check whether colocation's dependent preferences should be considered * * \param[in] colocation Colocation constraint * \param[in] rsc Primary instance (normally this will be * colocation->primary, which NULL will be treated as, * but for clones or bundles with multiple instances * this can be a particular instance) * * \return true if colocation influence should be effective, otherwise false */ static inline bool pcmk__colocation_has_influence(const pcmk__colocation_t *colocation, const pe_resource_t *rsc) { if (rsc == NULL) { rsc = colocation->primary; } /* A bundle replica colocates its remote connection with its container, * using a finite score so that the container can run on Pacemaker Remote * nodes. * * Moving a connection is lightweight and does not interrupt the service, * while moving a container is heavyweight and does interrupt the service, * so don't move a clean, active container based solely on the preferences * of its connection. * * This also avoids problematic scenarios where two containers want to * perpetually swap places. */ if (pcmk_is_set(colocation->dependent->flags, pe_rsc_allow_remote_remotes) && !pcmk_is_set(rsc->flags, pe_rsc_failed) && pcmk__list_of_1(rsc->running_on)) { return false; } /* The dependent in a colocation influences the primary's location * if the influence option is true or the primary is not yet active. */ return colocation->influence || (rsc->running_on == NULL); } // Ordering constraints (pcmk_sched_ordering.c) G_GNUC_INTERNAL void pcmk__new_ordering(pe_resource_t *first_rsc, char *first_task, pe_action_t *first_action, pe_resource_t *then_rsc, char *then_task, pe_action_t *then_action, uint32_t flags, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__unpack_ordering(xmlNode *xml_obj, pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__disable_invalid_orderings(pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__order_stops_before_shutdown(pe_node_t *node, pe_action_t *shutdown_op); G_GNUC_INTERNAL void pcmk__apply_orderings(pe_working_set_t *data_set); G_GNUC_INTERNAL void pcmk__order_after_each(pe_action_t *after, GList *list); /*! * \internal * \brief Create a new ordering between two resource actions * * \param[in,out] first_rsc Resource for 'first' action * \param[in,out] first_task Action key for 'first' action * \param[in] then_rsc Resource for 'then' action * \param[in,out] then_task Action key for 'then' action * \param[in] flags Bitmask of enum pe_ordering flags */ #define pcmk__order_resource_actions(first_rsc, first_task, \ then_rsc, then_task, flags) \ pcmk__new_ordering((first_rsc), \ pcmk__op_key((first_rsc)->id, (first_task), 0), \ NULL, \ (then_rsc), \ pcmk__op_key((then_rsc)->id, (then_task), 0), \ NULL, (flags), (first_rsc)->cluster) #define pcmk__order_starts(rsc1, rsc2, flags) \ pcmk__order_resource_actions((rsc1), CRMD_ACTION_START, \ (rsc2), CRMD_ACTION_START, (flags)) #define pcmk__order_stops(rsc1, rsc2, flags) \ pcmk__order_resource_actions((rsc1), CRMD_ACTION_STOP, \ (rsc2), CRMD_ACTION_STOP, (flags)) // Ticket constraints (pcmk_sched_tickets.c) G_GNUC_INTERNAL void pcmk__unpack_rsc_ticket(xmlNode *xml_obj, pe_working_set_t *data_set); // Promotable clone resources (pcmk_sched_promotable.c) G_GNUC_INTERNAL void pcmk__add_promotion_scores(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__require_promotion_tickets(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__set_instance_roles(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__create_promotable_actions(pe_resource_t *clone); G_GNUC_INTERNAL void pcmk__promotable_restart_ordering(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__order_promotable_instances(pe_resource_t *clone); G_GNUC_INTERNAL void pcmk__update_dependent_with_promotable(const pe_resource_t *primary, pe_resource_t *dependent, const pcmk__colocation_t *colocation); G_GNUC_INTERNAL void pcmk__update_promotable_dependent_priority(const pe_resource_t *primary, pe_resource_t *dependent, const pcmk__colocation_t *colocation); // Pacemaker Remote nodes (pcmk_sched_remote.c) G_GNUC_INTERNAL bool pcmk__is_failed_remote_node(const pe_node_t *node); G_GNUC_INTERNAL void pcmk__order_remote_connection_actions(pe_working_set_t *data_set); G_GNUC_INTERNAL bool pcmk__rsc_corresponds_to_guest(const pe_resource_t *rsc, const pe_node_t *node); G_GNUC_INTERNAL pe_node_t *pcmk__connection_host_for_action(const pe_action_t *action); G_GNUC_INTERNAL void pcmk__substitute_remote_addr(pe_resource_t *rsc, GHashTable *params); G_GNUC_INTERNAL void pcmk__add_bundle_meta_to_xml(xmlNode *args_xml, const pe_action_t *action); // Primitives (pcmk_sched_primitive.c) G_GNUC_INTERNAL pe_node_t *pcmk__primitive_assign(pe_resource_t *rsc, const pe_node_t *prefer); G_GNUC_INTERNAL void pcmk__primitive_create_actions(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__primitive_internal_constraints(pe_resource_t *rsc); G_GNUC_INTERNAL uint32_t pcmk__primitive_action_flags(pe_action_t *action, const pe_node_t *node); G_GNUC_INTERNAL void pcmk__primitive_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent); G_GNUC_INTERNAL void pcmk__with_primitive_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__primitive_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__schedule_cleanup(pe_resource_t *rsc, const pe_node_t *node, bool optional); G_GNUC_INTERNAL void pcmk__primitive_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml); G_GNUC_INTERNAL void pcmk__primitive_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization); G_GNUC_INTERNAL void pcmk__primitive_shutdown_lock(pe_resource_t *rsc); // Groups (pcmk_sched_group.c) G_GNUC_INTERNAL pe_node_t *pcmk__group_assign(pe_resource_t *rsc, const pe_node_t *prefer); G_GNUC_INTERNAL void pcmk__group_create_actions(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__group_internal_constraints(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__group_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent); G_GNUC_INTERNAL void pcmk__with_group_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__group_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__group_add_colocated_node_scores(pe_resource_t *rsc, const char *log_id, GHashTable **nodes, pcmk__colocation_t *colocation, float factor, uint32_t flags); G_GNUC_INTERNAL void pcmk__group_apply_location(pe_resource_t *rsc, pe__location_t *location); G_GNUC_INTERNAL uint32_t pcmk__group_action_flags(pe_action_t *action, const pe_node_t *node); G_GNUC_INTERNAL uint32_t pcmk__group_update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set); G_GNUC_INTERNAL GList *pcmk__group_colocated_resources(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *colocated_rscs); G_GNUC_INTERNAL void pcmk__group_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization); G_GNUC_INTERNAL void pcmk__group_shutdown_lock(pe_resource_t *rsc); // Clones (pcmk_sched_clone.c) G_GNUC_INTERNAL pe_node_t *pcmk__clone_assign(pe_resource_t *rsc, const pe_node_t *prefer); G_GNUC_INTERNAL void pcmk__clone_create_actions(pe_resource_t *rsc); G_GNUC_INTERNAL bool pcmk__clone_create_probe(pe_resource_t *rsc, pe_node_t *node); G_GNUC_INTERNAL void pcmk__clone_internal_constraints(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__clone_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent); G_GNUC_INTERNAL void pcmk__with_clone_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__clone_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__clone_apply_location(pe_resource_t *rsc, pe__location_t *constraint); G_GNUC_INTERNAL uint32_t pcmk__clone_action_flags(pe_action_t *action, const pe_node_t *node); G_GNUC_INTERNAL void pcmk__clone_add_actions_to_graph(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__clone_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml); G_GNUC_INTERNAL void pcmk__clone_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization); G_GNUC_INTERNAL void pcmk__clone_shutdown_lock(pe_resource_t *rsc); // Bundles (pcmk_sched_bundle.c) G_GNUC_INTERNAL pe_node_t *pcmk__bundle_assign(pe_resource_t *rsc, const pe_node_t *prefer); G_GNUC_INTERNAL void pcmk__bundle_create_actions(pe_resource_t *rsc); G_GNUC_INTERNAL bool pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node); G_GNUC_INTERNAL void pcmk__bundle_internal_constraints(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__bundle_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent); G_GNUC_INTERNAL void pcmk__with_bundle_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__bundle_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list); G_GNUC_INTERNAL void pcmk__bundle_apply_location(pe_resource_t *rsc, pe__location_t *constraint); G_GNUC_INTERNAL uint32_t pcmk__bundle_action_flags(pe_action_t *action, const pe_node_t *node); G_GNUC_INTERNAL void pcmk__output_bundle_actions(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__bundle_add_actions_to_graph(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__bundle_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization); G_GNUC_INTERNAL void pcmk__bundle_shutdown_lock(pe_resource_t *rsc); // Clone instances or bundle replica containers (pcmk_sched_instances.c) G_GNUC_INTERNAL void pcmk__assign_instances(pe_resource_t *collective, GList *instances, int max_total, int max_per_node); G_GNUC_INTERNAL void pcmk__create_instance_actions(pe_resource_t *rsc, GList *instances); G_GNUC_INTERNAL bool pcmk__instance_matches(const pe_resource_t *instance, const pe_node_t *node, enum rsc_role_e role, bool current); G_GNUC_INTERNAL pe_resource_t *pcmk__find_compatible_instance(const pe_resource_t *match_rsc, const pe_resource_t *rsc, enum rsc_role_e role, bool current); G_GNUC_INTERNAL uint32_t pcmk__instance_update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set); G_GNUC_INTERNAL uint32_t pcmk__collective_action_flags(pe_action_t *action, const GList *instances, const pe_node_t *node); G_GNUC_INTERNAL void pcmk__add_collective_constraints(GList **list, const pe_resource_t *instance, const pe_resource_t *collective, bool with_this); // Injections (pcmk_injections.c) G_GNUC_INTERNAL xmlNode *pcmk__inject_node(cib_t *cib_conn, const char *node, const char *uuid); G_GNUC_INTERNAL xmlNode *pcmk__inject_node_state_change(cib_t *cib_conn, const char *node, bool up); G_GNUC_INTERNAL xmlNode *pcmk__inject_resource_history(pcmk__output_t *out, xmlNode *cib_node, const char *resource, const char *lrm_name, const char *rclass, const char *rtype, const char *rprovider); G_GNUC_INTERNAL void pcmk__inject_failcount(pcmk__output_t *out, xmlNode *cib_node, const char *resource, const char *task, guint interval_ms, int rc); G_GNUC_INTERNAL xmlNode *pcmk__inject_action_result(xmlNode *cib_resource, lrmd_event_data_t *op, int target_rc); // Nodes (pcmk_sched_nodes.c) G_GNUC_INTERNAL bool pcmk__node_available(const pe_node_t *node, bool consider_score, bool consider_guest); G_GNUC_INTERNAL bool pcmk__any_node_available(GHashTable *nodes); G_GNUC_INTERNAL GHashTable *pcmk__copy_node_table(GHashTable *nodes); G_GNUC_INTERNAL GList *pcmk__sort_nodes(GList *nodes, pe_node_t *active_node); G_GNUC_INTERNAL void pcmk__apply_node_health(pe_working_set_t *data_set); G_GNUC_INTERNAL pe_node_t *pcmk__top_allowed_node(const pe_resource_t *rsc, const pe_node_t *node); // Functions applying to more than one variant (pcmk_sched_resource.c) G_GNUC_INTERNAL -void pcmk__set_allocation_methods(pe_working_set_t *data_set); +void pcmk__set_assignment_methods(pe_working_set_t *data_set); G_GNUC_INTERNAL bool pcmk__rsc_agent_changed(pe_resource_t *rsc, pe_node_t *node, const xmlNode *rsc_entry, bool active_on_node); G_GNUC_INTERNAL GList *pcmk__rscs_matching_id(const char *id, const pe_working_set_t *data_set); G_GNUC_INTERNAL GList *pcmk__colocated_resources(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *colocated_rscs); G_GNUC_INTERNAL void pcmk__noop_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml); G_GNUC_INTERNAL void pcmk__output_resource_actions(pe_resource_t *rsc); G_GNUC_INTERNAL bool pcmk__finalize_assignment(pe_resource_t *rsc, pe_node_t *chosen, bool force); G_GNUC_INTERNAL bool pcmk__assign_resource(pe_resource_t *rsc, pe_node_t *node, bool force); G_GNUC_INTERNAL void pcmk__unassign_resource(pe_resource_t *rsc); G_GNUC_INTERNAL bool pcmk__threshold_reached(pe_resource_t *rsc, const pe_node_t *node, pe_resource_t **failed); G_GNUC_INTERNAL void pcmk__sort_resources(pe_working_set_t *data_set); G_GNUC_INTERNAL gint pcmk__cmp_instance(gconstpointer a, gconstpointer b); G_GNUC_INTERNAL gint pcmk__cmp_instance_number(gconstpointer a, gconstpointer b); // Functions related to probes (pcmk_sched_probes.c) G_GNUC_INTERNAL bool pcmk__probe_rsc_on_node(pe_resource_t *rsc, pe_node_t *node); G_GNUC_INTERNAL void pcmk__order_probes(pe_working_set_t *data_set); G_GNUC_INTERNAL bool pcmk__probe_resource_list(GList *rscs, pe_node_t *node); G_GNUC_INTERNAL void pcmk__schedule_probes(pe_working_set_t *data_set); // Functions related to live migration (pcmk_sched_migration.c) void pcmk__create_migration_actions(pe_resource_t *rsc, const pe_node_t *current); void pcmk__abort_dangling_migration(void *data, void *user_data); bool pcmk__rsc_can_migrate(const pe_resource_t *rsc, const pe_node_t *current); void pcmk__order_migration_equivalents(pe__ordering_t *order); // Functions related to node utilization (pcmk_sched_utilization.c) G_GNUC_INTERNAL int pcmk__compare_node_capacities(const pe_node_t *node1, const pe_node_t *node2); G_GNUC_INTERNAL void pcmk__consume_node_capacity(GHashTable *current_utilization, const pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__release_node_capacity(GHashTable *current_utilization, const pe_resource_t *rsc); G_GNUC_INTERNAL const pe_node_t *pcmk__ban_insufficient_capacity(pe_resource_t *rsc); G_GNUC_INTERNAL void pcmk__create_utilization_constraints(pe_resource_t *rsc, const GList *allowed_nodes); G_GNUC_INTERNAL void pcmk__show_node_capacities(const char *desc, pe_working_set_t *data_set); #endif // PCMK__LIBPACEMAKER_PRIVATE__H diff --git a/lib/pacemaker/pcmk_cluster_queries.c b/lib/pacemaker/pcmk_cluster_queries.c index 6002cd495f..6b8289a638 100644 --- a/lib/pacemaker/pcmk_cluster_queries.c +++ b/lib/pacemaker/pcmk_cluster_queries.c @@ -1,900 +1,896 @@ /* - * Copyright 2020-2022 the Pacemaker project contributors + * Copyright 2020-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include -#include // gboolean, GMainLoop, etc. #include // xmlNode #include #include #include #include #include #include #include #include #include #include #include #include //! Object to store node info from the controller API typedef struct { /* Adapted from pcmk_controld_api_reply_t:data:node_info. * (char **) are convenient here for use within callbacks: we can skip * copying strings unless the caller passes a non-NULL value. */ uint32_t id; char **node_name; char **uuid; char **state; bool have_quorum; bool is_remote; } node_info_t; //! Object to store API results, a timeout, and an output object typedef struct { pcmk__output_t *out; bool show_output; int rc; unsigned int message_timeout_ms; enum pcmk_pacemakerd_state pcmkd_state; node_info_t node_info; } data_t; /*! * \internal * \brief Validate that an IPC API event is a good reply * * \param[in,out] data API results and options * \param[in] api IPC API connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * * \return Standard Pacemaker return code */ static int validate_reply_event(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status) { pcmk__output_t *out = data->out; switch (event_type) { case pcmk_ipc_event_reply: break; case pcmk_ipc_event_disconnect: if (data->rc == ECONNRESET) { // Unexpected out->err(out, "error: Lost connection to %s", pcmk_ipc_name(api, true)); } // Nothing bad but not the reply we're looking for return ENOTSUP; default: // Ditto return ENOTSUP; } if (status != CRM_EX_OK) { out->err(out, "error: Bad reply from %s: %s", pcmk_ipc_name(api, true), crm_exit_str(status)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a controller API event is a good reply of expected type * * \param[in,out] data API results and options * \param[in] api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_controld_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_controld_api_reply expected_type) { pcmk__output_t *out = data->out; int rc = pcmk_rc_ok; const pcmk_controld_api_reply_t *reply = NULL; rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from controller", pcmk__controld_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a \p pacemakerd API event is a good reply of expected * type * * \param[in,out] data API results and options * \param[in] api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_pcmkd_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_pacemakerd_api_reply expected_type) { pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_pacemakerd_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from pacemakerd", pcmk__pcmkd_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Process a controller status IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void controller_status_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "health", reply->data.ping.sys_from, reply->host_from, reply->data.ping.fsa_state, reply->data.ping.result); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a designated controller IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void designated_controller_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "dc", reply->host_from); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a node info IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void node_info_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_info); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->data.node_info.uname == NULL) { out->err(out, "Node is not known to cluster"); data->rc = pcmk_rc_node_unknown; return; } data->node_info.have_quorum = reply->data.node_info.have_quorum; data->node_info.is_remote = reply->data.node_info.is_remote; data->node_info.id = (uint32_t) reply->data.node_info.id; pcmk__str_update(data->node_info.node_name, reply->data.node_info.uname); pcmk__str_update(data->node_info.uuid, reply->data.node_info.uuid); pcmk__str_update(data->node_info.state, reply->data.node_info.state); if (data->show_output) { out->message(out, "node-info", reply->data.node_info.id, reply->data.node_info.uname, reply->data.node_info.uuid, reply->data.node_info.state, reply->data.node_info.have_quorum, reply->data.node_info.is_remote); } data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a \p pacemakerd status IPC event * * \param[in,out] pacemakerd_api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_pacemakerd_api_reply_t object * containing event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void pacemakerd_event_cb(pcmk_ipc_api_t *pacemakerd_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = user_data; pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_pcmkd_reply(data, pacemakerd_api, event_type, status, event_data, pcmk_pacemakerd_reply_ping); if (rc != pcmk_rc_ok) { return; } // Parse desired information from reply reply = (const pcmk_pacemakerd_api_reply_t *) event_data; data->pcmkd_state = reply->data.ping.state; data->rc = pcmk_rc_ok; if (!data->show_output) { return; } if (reply->data.ping.status == pcmk_rc_ok) { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, NULL, reply->data.ping.last_good); } else { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, "query failed", time(NULL)); } } static pcmk_ipc_api_t * ipc_connect(data_t *data, enum pcmk_ipc_server server, pcmk_ipc_callback_t cb, enum pcmk_ipc_dispatch dispatch_type, bool eremoteio_ok) { int rc; pcmk__output_t *out = data->out; pcmk_ipc_api_t *api = NULL; rc = pcmk_new_ipc_api(&api, server); if (api == NULL) { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); data->rc = rc; return NULL; } if (cb != NULL) { pcmk_register_ipc_callback(api, cb, data); } rc = pcmk_connect_ipc(api, dispatch_type); if (rc != pcmk_rc_ok) { if (rc == EREMOTEIO) { data->pcmkd_state = pcmk_pacemakerd_state_remote; if (eremoteio_ok) { /* EREMOTEIO may be expected and acceptable for some callers * on a Pacemaker Remote node */ rc = pcmk_rc_ok; } else { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); } } data->rc = rc; pcmk_free_ipc_api(api); return NULL; } return api; } /*! * \internal * \brief Poll an IPC API connection until timeout or a reply is received * * \param[in,out] data API results and options * \param[in,out] api IPC API connection * \param[in] on_node If not \p NULL, name of the node to poll (used only * for logging) * * \note Sets the \p rc member of \p data on error */ static void poll_until_reply(data_t *data, pcmk_ipc_api_t *api, const char *on_node) { pcmk__output_t *out = data->out; uint64_t start_nsec = qb_util_nano_current_get(); uint64_t end_nsec = start_nsec; uint64_t elapsed_ms = 0; uint64_t remaining_ms = data->message_timeout_ms; while (remaining_ms > 0) { int rc = pcmk_poll_ipc(api, remaining_ms); if (rc == EAGAIN) { // Poll timed out break; } if (rc != pcmk_rc_ok) { out->err(out, "error: Failed to poll %s API%s%s: %s", pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, ""), pcmk_rc_str(rc)); data->rc = rc; return; } pcmk_dispatch_ipc(api); if (data->rc != EAGAIN) { // Received a reply return; } end_nsec = qb_util_nano_current_get(); elapsed_ms = (end_nsec - start_nsec) / QB_TIME_NS_IN_MSEC; remaining_ms = data->message_timeout_ms - elapsed_ms; } out->err(out, "error: Timed out after %ums waiting for reply from %s API%s%s", data->message_timeout_ms, pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, "")); data->rc = EAGAIN; } /*! * \internal * \brief Get and output controller status * * \param[in,out] out Output object * \param[in] node_name Name of node whose status is desired * (\p NULL for DC) * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__controller_status(pcmk__output_t *out, const char *node_name, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, controller_status_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, node_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on %s: %s", pcmk__s(node_name, "DC"), pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, pcmk__s(node_name, "DC")); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_controller_status(xmlNodePtr *xml, const char *node_name, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__controller_status(out, node_name, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and output designated controller node name * * \param[in,out] out Output object * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__designated_controller(pcmk__output_t *out, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, designated_controller_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, NULL); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on DC: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "DC"); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_designated_controller(xmlNodePtr *xml, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__designated_controller(out, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and optionally output node info corresponding to a node ID from * the controller * * \param[in,out] out Output object * \param[in,out] node_id ID of node whose name to get. If \p NULL * or 0, get the local node name. If not * \p NULL, store the true node ID here on * success. * \param[out] node_name If not \p NULL, where to store the node * name * \param[out] uuid If not \p NULL, where to store the node * UUID * \param[out] state If not \p NULL, where to store the * membership state * \param[out] is_remote If not \p NULL, where to store whether the * node is a Pacemaker Remote node * \param[out] have_quorum If not \p NULL, where to store whether the * node has quorum * \param[in] show_output Whether to show the node info * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code * * \note The caller is responsible for freeing \p *node_name, \p *uuid, and * \p *state using \p free(). */ int pcmk__query_node_info(pcmk__output_t *out, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .node_info = { .id = (node_id == NULL)? 0 : *node_id, .node_name = node_name, .uuid = uuid, .state = state, }, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (node_name != NULL) { *node_name = NULL; } if (uuid != NULL) { *uuid = NULL; } if (state != NULL) { *state = NULL; } if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, node_info_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_node_info(controld_api, (node_id != NULL)? *node_id : 0); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not send request to controller API on local " "node: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "local node"); } pcmk_free_ipc_api(controld_api); } if (data.rc != pcmk_rc_ok) { return data.rc; } // String outputs are set in callback if (node_id != NULL) { *node_id = data.node_info.id; } if (have_quorum != NULL) { *have_quorum = data.node_info.have_quorum; } if (is_remote != NULL) { *is_remote = data.node_info.is_remote; } return data.rc; } // Documented in header int pcmk_query_node_info(xmlNodePtr *xml, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; CRM_ASSERT(node_name != NULL); rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__query_node_info(out, node_id, node_name, uuid, state, have_quorum, is_remote, show_output, message_timeout_ms); pcmk__xml_output_finish(out, xml); return rc; } /*! * \internal * \brief Get and optionally output \p pacemakerd status * * \param[in,out] out Output object * \param[in] ipc_name IPC name for request * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemakerd API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * \param[in] show_output Whether to output the \p pacemakerd state * \param[out] state Where to store the \p pacemakerd state, if * not \p NULL * * \return Standard Pacemaker return code * * \note This function sets \p state to \p pcmk_pacemakerd_state_remote and * returns \p pcmk_rc_ok if the IPC connection attempt returns * \p EREMOTEIO. That code indicates that this is a Pacemaker Remote node * with \p pacemaker-remoted running. The node may be connected to the * cluster. */ int pcmk__pacemakerd_status(pcmk__output_t *out, const char *ipc_name, unsigned int message_timeout_ms, bool show_output, enum pcmk_pacemakerd_state *state) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .pcmkd_state = pcmk_pacemakerd_state_invalid, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *pacemakerd_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } pacemakerd_api = ipc_connect(&data, pcmk_ipc_pacemakerd, pacemakerd_event_cb, dispatch_type, true); if (pacemakerd_api != NULL) { int rc = pcmk_pacemakerd_api_ping(pacemakerd_api, ipc_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping launcher API: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, pacemakerd_api, NULL); } pcmk_free_ipc_api(pacemakerd_api); } else if ((data.pcmkd_state == pcmk_pacemakerd_state_remote) && show_output) { // No API connection so the callback wasn't run out->message(out, "pacemakerd-health", NULL, data.pcmkd_state, NULL, time(NULL)); } if (state != NULL) { *state = data.pcmkd_state; } return data.rc; } // Documented in header int pcmk_pacemakerd_status(xmlNodePtr *xml, const char *ipc_name, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__pacemakerd_status(out, ipc_name, message_timeout_ms, true, NULL); pcmk__xml_output_finish(out, xml); return rc; } /* user data for looping through remote node xpath searches */ struct node_data { pcmk__output_t *out; int found; const char *field; /* XML attribute to check for node name */ const char *type; - gboolean bash_export; + bool bash_export; }; static void remote_node_print_helper(xmlNode *result, void *user_data) { struct node_data *data = user_data; pcmk__output_t *out = data->out; const char *name = crm_element_value(result, XML_ATTR_UNAME); const char *id = crm_element_value(result, data->field); // node name and node id are the same for remote/guest nodes out->message(out, "crmadmin-node", data->type, - name ? name : id, - id, - data->bash_export); + pcmk__s(name, id), id, data->bash_export); data->found++; } // \return Standard Pacemaker return code int -pcmk__list_nodes(pcmk__output_t *out, const char *node_types, - gboolean bash_export) +pcmk__list_nodes(pcmk__output_t *out, const char *node_types, bool bash_export) { xmlNode *xml_node = NULL; int rc; rc = cib__signon_query(out, NULL, &xml_node); if (rc == pcmk_rc_ok) { struct node_data data = { .out = out, .found = 0, .bash_export = bash_export }; out->begin_list(out, NULL, NULL, "nodes"); if (!pcmk__str_empty(node_types) && strstr(node_types, "all")) { node_types = NULL; } if (pcmk__str_empty(node_types) || strstr(node_types, "cluster")) { data.field = "id"; data.type = "cluster"; crm_foreach_xpath_result(xml_node, PCMK__XP_MEMBER_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || strstr(node_types, "guest")) { data.field = "value"; data.type = "guest"; crm_foreach_xpath_result(xml_node, PCMK__XP_GUEST_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || !pcmk__strcmp(node_types, ",|^remote", pcmk__str_regex)) { data.field = "id"; data.type = "remote"; crm_foreach_xpath_result(xml_node, PCMK__XP_REMOTE_NODE_CONFIG, remote_node_print_helper, &data); } out->end_list(out); if (data.found == 0) { out->info(out, "No nodes configured"); } free_xml(xml_node); } return rc; } int pcmk_list_nodes(xmlNodePtr *xml, const char *node_types) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__list_nodes(out, node_types, FALSE); pcmk__xml_output_finish(out, xml); return rc; } diff --git a/lib/pacemaker/pcmk_graph_consumer.c b/lib/pacemaker/pcmk_graph_consumer.c index f2f172efe4..44bec8e62a 100644 --- a/lib/pacemaker/pcmk_graph_consumer.c +++ b/lib/pacemaker/pcmk_graph_consumer.c @@ -1,874 +1,875 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include /* * Functions for updating graph */ /*! * \internal * \brief Update synapse after completed prerequisite * * A synapse is ready to be executed once all its prerequisite actions (inputs) * complete. Given a completed action, check whether it is an input for a given * synapse, and if so, mark the input as confirmed, and mark the synapse as * ready if appropriate. * * \param[in,out] synapse Transition graph synapse to update * \param[in] action_id ID of an action that completed * * \note The only substantial effect here is confirming synapse inputs. * should_fire_synapse() will recalculate pcmk__synapse_ready, so the only * thing that uses the pcmk__synapse_ready from here is * synapse_state_str(). */ static void update_synapse_ready(pcmk__graph_synapse_t *synapse, int action_id) { if (pcmk_is_set(synapse->flags, pcmk__synapse_ready)) { return; // All inputs have already been confirmed } pcmk__set_synapse_flags(synapse, pcmk__synapse_ready); // Presume ready until proven otherwise for (GList *lpc = synapse->inputs; lpc != NULL; lpc = lpc->next) { pcmk__graph_action_t *prereq = (pcmk__graph_action_t *) lpc->data; if (prereq->id == action_id) { crm_trace("Confirming input %d of synapse %d", action_id, synapse->id); pcmk__set_graph_action_flags(prereq, pcmk__graph_action_confirmed); } else if (!(pcmk_is_set(prereq->flags, pcmk__graph_action_confirmed))) { pcmk__clear_synapse_flags(synapse, pcmk__synapse_ready); crm_trace("Synapse %d still not ready after action %d", synapse->id, action_id); } } if (pcmk_is_set(synapse->flags, pcmk__synapse_ready)) { crm_trace("Synapse %d is now ready to execute", synapse->id); } } /*! * \internal * \brief Update action and synapse confirmation after action completion * * \param[in,out] synapse Transition graph synapse that action belongs to * \param[in] action_id ID of action that completed */ static void update_synapse_confirmed(pcmk__graph_synapse_t *synapse, int action_id) { bool all_confirmed = true; for (GList *lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) lpc->data; if (action->id == action_id) { crm_trace("Confirmed action %d of synapse %d", action_id, synapse->id); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); } else if (all_confirmed && !(pcmk_is_set(action->flags, pcmk__graph_action_confirmed))) { all_confirmed = false; crm_trace("Synapse %d still not confirmed after action %d", synapse->id, action_id); } } if (all_confirmed && !(pcmk_is_set(synapse->flags, pcmk__synapse_confirmed))) { crm_trace("Confirmed synapse %d", synapse->id); pcmk__set_synapse_flags(synapse, pcmk__synapse_confirmed); } } /*! * \internal * \brief Update the transition graph with a completed action result * * \param[in,out] graph Transition graph to update * \param[in] action Action that completed */ void pcmk__update_graph(pcmk__graph_t *graph, const pcmk__graph_action_t *action) { for (GList *lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) lpc->data; if (pcmk_any_flags_set(synapse->flags, pcmk__synapse_confirmed|pcmk__synapse_failed)) { continue; // This synapse already completed } else if (pcmk_is_set(synapse->flags, pcmk__synapse_executed)) { update_synapse_confirmed(synapse, action->id); } else if (!(pcmk_is_set(action->flags, pcmk__graph_action_failed)) || (synapse->priority == INFINITY)) { update_synapse_ready(synapse, action->id); } } } /* * Functions for executing graph */ /* A transition graph consists of various types of actions. The library caller * registers execution functions for each action type, which will be stored * here. */ static pcmk__graph_functions_t *graph_fns = NULL; /*! * \internal * \brief Set transition graph execution functions * * \param[in] Execution functions to use */ void pcmk__set_graph_functions(pcmk__graph_functions_t *fns) { crm_debug("Setting custom functions for executing transition graphs"); graph_fns = fns; CRM_ASSERT(graph_fns != NULL); CRM_ASSERT(graph_fns->rsc != NULL); CRM_ASSERT(graph_fns->cluster != NULL); CRM_ASSERT(graph_fns->pseudo != NULL); CRM_ASSERT(graph_fns->fence != NULL); } /*! * \internal * \brief Check whether a graph synapse is ready to be executed * * \param[in,out] graph Transition graph that synapse is part of * \param[in,out] synapse Synapse to check * * \return true if synapse is ready, false otherwise */ static bool should_fire_synapse(pcmk__graph_t *graph, pcmk__graph_synapse_t *synapse) { GList *lpc = NULL; pcmk__set_synapse_flags(synapse, pcmk__synapse_ready); for (lpc = synapse->inputs; lpc != NULL; lpc = lpc->next) { pcmk__graph_action_t *prereq = (pcmk__graph_action_t *) lpc->data; if (!(pcmk_is_set(prereq->flags, pcmk__graph_action_confirmed))) { crm_trace("Input %d for synapse %d not yet confirmed", prereq->id, synapse->id); pcmk__clear_synapse_flags(synapse, pcmk__synapse_ready); break; } else if (pcmk_is_set(prereq->flags, pcmk__graph_action_failed) && !(pcmk_is_set(prereq->flags, pcmk__graph_action_can_fail))) { crm_trace("Input %d for synapse %d confirmed but failed", prereq->id, synapse->id); pcmk__clear_synapse_flags(synapse, pcmk__synapse_ready); break; } } if (pcmk_is_set(synapse->flags, pcmk__synapse_ready)) { crm_trace("Synapse %d is ready to execute", synapse->id); } else { return false; } for (lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { pcmk__graph_action_t *a = (pcmk__graph_action_t *) lpc->data; if (a->type == pcmk__pseudo_graph_action) { /* None of the below applies to pseudo ops */ } else if (synapse->priority < graph->abort_priority) { crm_trace("Skipping synapse %d: priority %d is less than " "abort priority %d", synapse->id, synapse->priority, graph->abort_priority); graph->skipped++; return false; } else if (graph_fns->allowed && !(graph_fns->allowed(graph, a))) { crm_trace("Deferring synapse %d: not allowed", synapse->id); return false; } } return true; } /*! * \internal * \brief Initiate an action from a transition graph * * \param[in,out] graph Transition graph containing action * \param[in,out] action Action to execute * * \return Standard Pacemaker return code */ static int initiate_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *id = ID(action->xml); CRM_CHECK(id != NULL, return EINVAL); CRM_CHECK(!pcmk_is_set(action->flags, pcmk__graph_action_executed), return pcmk_rc_already); pcmk__set_graph_action_flags(action, pcmk__graph_action_executed); switch (action->type) { case pcmk__pseudo_graph_action: crm_trace("Executing pseudo-action %d (%s)", action->id, id); return graph_fns->pseudo(graph, action); case pcmk__rsc_graph_action: crm_trace("Executing resource action %d (%s)", action->id, id); return graph_fns->rsc(graph, action); case pcmk__cluster_graph_action: if (pcmk__str_eq(crm_element_value(action->xml, XML_LRM_ATTR_TASK), CRM_OP_FENCE, pcmk__str_casei)) { crm_trace("Executing fencing action %d (%s)", action->id, id); return graph_fns->fence(graph, action); } crm_trace("Executing cluster action %d (%s)", action->id, id); return graph_fns->cluster(graph, action); default: crm_err("Unsupported graph action type <%s " XML_ATTR_ID "='%s'> " "(bug?)", crm_element_name(action->xml), id); return EINVAL; } } /*! * \internal * \brief Execute a graph synapse * * \param[in,out] graph Transition graph with synapse to execute * \param[in,out] synapse Synapse to execute * * \return Standard Pacemaker return value */ static int fire_synapse(pcmk__graph_t *graph, pcmk__graph_synapse_t *synapse) { pcmk__set_synapse_flags(synapse, pcmk__synapse_executed); for (GList *lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) lpc->data; int rc = initiate_action(graph, action); if (rc != pcmk_rc_ok) { crm_err("Failed initiating <%s " XML_ATTR_ID "=%d> in synapse %d: " "%s", crm_element_name(action->xml), action->id, synapse->id, pcmk_rc_str(rc)); pcmk__set_synapse_flags(synapse, pcmk__synapse_confirmed); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed |pcmk__graph_action_failed); return pcmk_rc_error; } } return pcmk_rc_ok; } /*! * \internal * \brief Dummy graph method that can be used with simulations * * \param[in,out] graph Transition graph containing action * \param[in,out] action Graph action to be initiated * * \return Standard Pacemaker return code * \note If the PE_fail environment variable is set to the action ID, * then the graph action will be marked as failed. */ static int pseudo_action_dummy(pcmk__graph_t *graph, pcmk__graph_action_t *action) { static int fail = -1; if (fail < 0) { long long fail_ll; if ((pcmk__scan_ll(getenv("PE_fail"), &fail_ll, 0LL) == pcmk_rc_ok) && (fail_ll > 0LL) && (fail_ll <= INT_MAX)) { fail = (int) fail_ll; } else { fail = 0; } } if (action->id == fail) { crm_err("Dummy event handler: pretending action %d failed", action->id); pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); graph->abort_priority = INFINITY; } else { crm_trace("Dummy event handler: action %d initiated", action->id); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); return pcmk_rc_ok; } static pcmk__graph_functions_t default_fns = { pseudo_action_dummy, pseudo_action_dummy, pseudo_action_dummy, pseudo_action_dummy }; /*! * \internal * \brief Execute all actions in a transition graph * * \param[in,out] graph Transition graph to execute * * \return Status of transition after execution */ enum pcmk__graph_status pcmk__execute_graph(pcmk__graph_t *graph) { GList *lpc = NULL; int log_level = LOG_DEBUG; enum pcmk__graph_status pass_result = pcmk__graph_active; const char *status = "In progress"; if (graph_fns == NULL) { graph_fns = &default_fns; } if (graph == NULL) { return pcmk__graph_complete; } graph->fired = 0; graph->pending = 0; graph->skipped = 0; graph->completed = 0; graph->incomplete = 0; // Count completed and in-flight synapses for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) lpc->data; if (pcmk_is_set(synapse->flags, pcmk__synapse_confirmed)) { graph->completed++; } else if (!(pcmk_is_set(synapse->flags, pcmk__synapse_failed)) && pcmk_is_set(synapse->flags, pcmk__synapse_executed)) { graph->pending++; } } crm_trace("Executing graph %d (%d synapses already completed, %d pending)", graph->id, graph->completed, graph->pending); // Execute any synapses that are ready for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) lpc->data; if ((graph->batch_limit > 0) && (graph->pending >= graph->batch_limit)) { crm_debug("Throttling graph execution: batch limit (%d) reached", graph->batch_limit); break; } else if (pcmk_is_set(synapse->flags, pcmk__synapse_failed)) { graph->skipped++; continue; } else if (pcmk_any_flags_set(synapse->flags, pcmk__synapse_confirmed|pcmk__synapse_executed)) { continue; // Already handled } else if (should_fire_synapse(graph, synapse)) { graph->fired++; if (fire_synapse(graph, synapse) != pcmk_rc_ok) { crm_err("Synapse %d failed to fire", synapse->id); log_level = LOG_ERR; graph->abort_priority = INFINITY; graph->incomplete++; graph->fired--; } if (!(pcmk_is_set(synapse->flags, pcmk__synapse_confirmed))) { graph->pending++; } } else { crm_trace("Synapse %d cannot fire", synapse->id); graph->incomplete++; } } if ((graph->pending == 0) && (graph->fired == 0)) { graph->complete = true; if ((graph->incomplete != 0) && (graph->abort_priority <= 0)) { log_level = LOG_WARNING; pass_result = pcmk__graph_terminated; status = "Terminated"; } else if (graph->skipped != 0) { log_level = LOG_NOTICE; pass_result = pcmk__graph_complete; status = "Stopped"; } else { log_level = LOG_NOTICE; pass_result = pcmk__graph_complete; status = "Complete"; } } else if (graph->fired == 0) { pass_result = pcmk__graph_pending; } do_crm_log(log_level, "Transition %d (Complete=%d, Pending=%d," " Fired=%d, Skipped=%d, Incomplete=%d, Source=%s): %s", graph->id, graph->completed, graph->pending, graph->fired, graph->skipped, graph->incomplete, graph->source, status); return pass_result; } /* * Functions for unpacking transition graph XML into structs */ /*! * \internal * \brief Unpack a transition graph action from XML * * \param[in] parent Synapse that action is part of * \param[in] xml_action Action XML to unparse * * \return Newly allocated action on success, or NULL otherwise */ static pcmk__graph_action_t * unpack_action(pcmk__graph_synapse_t *parent, xmlNode *xml_action) { enum pcmk__graph_action_type action_type; pcmk__graph_action_t *action = NULL; const char *element = TYPE(xml_action); const char *value = ID(xml_action); if (value == NULL) { crm_err("Ignoring transition graph action without id (bug?)"); crm_log_xml_trace(xml_action, "invalid"); return NULL; } if (pcmk__str_eq(element, XML_GRAPH_TAG_RSC_OP, pcmk__str_casei)) { action_type = pcmk__rsc_graph_action; } else if (pcmk__str_eq(element, XML_GRAPH_TAG_PSEUDO_EVENT, pcmk__str_casei)) { action_type = pcmk__pseudo_graph_action; } else if (pcmk__str_eq(element, XML_GRAPH_TAG_CRM_EVENT, pcmk__str_casei)) { action_type = pcmk__cluster_graph_action; } else { crm_err("Ignoring transition graph action of unknown type '%s' (bug?)", element); crm_log_xml_trace(xml_action, "invalid"); return NULL; } action = calloc(1, sizeof(pcmk__graph_action_t)); if (action == NULL) { crm_perror(LOG_CRIT, "Cannot unpack transition graph action"); crm_log_xml_trace(xml_action, "lost"); return NULL; } pcmk__scan_min_int(value, &(action->id), -1); action->type = pcmk__rsc_graph_action; action->xml = copy_xml(xml_action); action->synapse = parent; action->type = action_type; action->params = xml2list(action->xml); value = g_hash_table_lookup(action->params, "CRM_meta_timeout"); pcmk__scan_min_int(value, &(action->timeout), 0); /* Take start-delay into account for the timeout of the action timer */ value = g_hash_table_lookup(action->params, "CRM_meta_start_delay"); { int start_delay; pcmk__scan_min_int(value, &start_delay, 0); action->timeout += start_delay; } if (pcmk__guint_from_hash(action->params, CRM_META "_" XML_LRM_ATTR_INTERVAL, 0, &(action->interval_ms)) != pcmk_rc_ok) { action->interval_ms = 0; } value = g_hash_table_lookup(action->params, "CRM_meta_can_fail"); if (value != NULL) { - gboolean can_fail = FALSE; + int can_fail = 0; + crm_str_to_boolean(value, &can_fail); if (can_fail) { pcmk__set_graph_action_flags(action, pcmk__graph_action_can_fail); } else { pcmk__clear_graph_action_flags(action, pcmk__graph_action_can_fail); } #ifndef PCMK__COMPAT_2_0 if (pcmk_is_set(action->flags, pcmk__graph_action_can_fail)) { crm_warn("Support for the can_fail meta-attribute is deprecated" " and will be removed in a future release"); } #endif } crm_trace("Action %d has timer set to %dms", action->id, action->timeout); return action; } /*! * \internal * \brief Unpack transition graph synapse from XML * * \param[in,out] new_graph Transition graph that synapse is part of * \param[in] xml_synapse Synapse XML * * \return Newly allocated synapse on success, or NULL otherwise */ static pcmk__graph_synapse_t * unpack_synapse(pcmk__graph_t *new_graph, const xmlNode *xml_synapse) { const char *value = NULL; xmlNode *action_set = NULL; pcmk__graph_synapse_t *new_synapse = NULL; crm_trace("Unpacking synapse %s", ID(xml_synapse)); new_synapse = calloc(1, sizeof(pcmk__graph_synapse_t)); if (new_synapse == NULL) { return NULL; } pcmk__scan_min_int(ID(xml_synapse), &(new_synapse->id), 0); value = crm_element_value(xml_synapse, XML_CIB_ATTR_PRIORITY); pcmk__scan_min_int(value, &(new_synapse->priority), 0); CRM_CHECK(new_synapse->id >= 0, free(new_synapse); return NULL); new_graph->num_synapses++; crm_trace("Unpacking synapse %s action sets", crm_element_value(xml_synapse, XML_ATTR_ID)); for (action_set = first_named_child(xml_synapse, "action_set"); action_set != NULL; action_set = crm_next_same_xml(action_set)) { for (xmlNode *action = pcmk__xml_first_child(action_set); action != NULL; action = pcmk__xml_next(action)) { pcmk__graph_action_t *new_action = unpack_action(new_synapse, action); if (new_action == NULL) { continue; } crm_trace("Adding action %d to synapse %d", new_action->id, new_synapse->id); new_graph->num_actions++; new_synapse->actions = g_list_append(new_synapse->actions, new_action); } } crm_trace("Unpacking synapse %s inputs", ID(xml_synapse)); for (xmlNode *inputs = first_named_child(xml_synapse, "inputs"); inputs != NULL; inputs = crm_next_same_xml(inputs)) { for (xmlNode *trigger = first_named_child(inputs, "trigger"); trigger != NULL; trigger = crm_next_same_xml(trigger)) { for (xmlNode *input = pcmk__xml_first_child(trigger); input != NULL; input = pcmk__xml_next(input)) { pcmk__graph_action_t *new_input = unpack_action(new_synapse, input); if (new_input == NULL) { continue; } crm_trace("Adding input %d to synapse %d", new_input->id, new_synapse->id); new_synapse->inputs = g_list_append(new_synapse->inputs, new_input); } } } return new_synapse; } /*! * \internal * \brief Unpack transition graph XML * * \param[in] xml_graph Transition graph XML to unpack * \param[in] reference Where the XML came from (for logging) * * \return Newly allocated transition graph on success, NULL otherwise * \note The caller is responsible for freeing the return value using * pcmk__free_graph(). * \note The XML is expected to be structured like: ... ... */ pcmk__graph_t * pcmk__unpack_graph(const xmlNode *xml_graph, const char *reference) { pcmk__graph_t *new_graph = NULL; new_graph = calloc(1, sizeof(pcmk__graph_t)); if (new_graph == NULL) { return NULL; } new_graph->source = strdup((reference == NULL)? "unknown" : reference); if (new_graph->source == NULL) { free(new_graph); return NULL; } new_graph->id = -1; new_graph->abort_priority = 0; new_graph->network_delay = 0; new_graph->stonith_timeout = 0; new_graph->completion_action = pcmk__graph_done; // Parse top-level attributes from if (xml_graph != NULL) { const char *buf = crm_element_value(xml_graph, "transition_id"); CRM_CHECK(buf != NULL, free(new_graph); return NULL); pcmk__scan_min_int(buf, &(new_graph->id), -1); buf = crm_element_value(xml_graph, "cluster-delay"); CRM_CHECK(buf != NULL, free(new_graph); return NULL); new_graph->network_delay = crm_parse_interval_spec(buf); buf = crm_element_value(xml_graph, "stonith-timeout"); if (buf == NULL) { new_graph->stonith_timeout = new_graph->network_delay; } else { new_graph->stonith_timeout = crm_parse_interval_spec(buf); } // Use 0 (dynamic limit) as default/invalid, -1 (no limit) as minimum buf = crm_element_value(xml_graph, "batch-limit"); if ((buf == NULL) || (pcmk__scan_min_int(buf, &(new_graph->batch_limit), -1) != pcmk_rc_ok)) { new_graph->batch_limit = 0; } buf = crm_element_value(xml_graph, "migration-limit"); pcmk__scan_min_int(buf, &(new_graph->migration_limit), -1); pcmk__str_update(&(new_graph->failed_stop_offset), crm_element_value(xml_graph, "failed-stop-offset")); pcmk__str_update(&(new_graph->failed_start_offset), crm_element_value(xml_graph, "failed-start-offset")); if (crm_element_value_epoch(xml_graph, "recheck-by", &(new_graph->recheck_by)) != pcmk_ok) { new_graph->recheck_by = 0; } } // Unpack each child element for (const xmlNode *synapse_xml = first_named_child(xml_graph, "synapse"); synapse_xml != NULL; synapse_xml = crm_next_same_xml(synapse_xml)) { pcmk__graph_synapse_t *new_synapse = unpack_synapse(new_graph, synapse_xml); if (new_synapse != NULL) { new_graph->synapses = g_list_append(new_graph->synapses, new_synapse); } } crm_debug("Unpacked transition %d from %s: %d actions in %d synapses", new_graph->id, new_graph->source, new_graph->num_actions, new_graph->num_synapses); return new_graph; } /* * Functions for freeing transition graph objects */ /*! * \internal * \brief Free a transition graph action object * * \param[in,out] user_data Action to free */ static void free_graph_action(gpointer user_data) { pcmk__graph_action_t *action = user_data; if (action->timer != 0) { crm_warn("Cancelling timer for graph action %d", action->id); g_source_remove(action->timer); } if (action->params != NULL) { g_hash_table_destroy(action->params); } free_xml(action->xml); free(action); } /*! * \internal * \brief Free a transition graph synapse object * * \param[in,out] user_data Synapse to free */ static void free_graph_synapse(gpointer user_data) { pcmk__graph_synapse_t *synapse = user_data; g_list_free_full(synapse->actions, free_graph_action); g_list_free_full(synapse->inputs, free_graph_action); free(synapse); } /*! * \internal * \brief Free a transition graph object * * \param[in,out] graph Transition graph to free */ void pcmk__free_graph(pcmk__graph_t *graph) { if (graph != NULL) { g_list_free_full(graph->synapses, free_graph_synapse); free(graph->source); free(graph->failed_stop_offset); free(graph->failed_start_offset); free(graph); } } /* * Other transition graph utilities */ /*! * \internal * \brief Synthesize an executor event from a graph action * * \param[in] resource If not NULL, use greater call ID than in this XML * \param[in] action Graph action * \param[in] status What to use as event execution status * \param[in] rc What to use as event exit status * \param[in] exit_reason What to use as event exit reason * * \return Newly allocated executor event on success, or NULL otherwise */ lrmd_event_data_t * pcmk__event_from_graph_action(const xmlNode *resource, const pcmk__graph_action_t *action, int status, int rc, const char *exit_reason) { lrmd_event_data_t *op = NULL; GHashTableIter iter; const char *name = NULL; const char *value = NULL; xmlNode *action_resource = NULL; CRM_CHECK(action != NULL, return NULL); CRM_CHECK(action->type == pcmk__rsc_graph_action, return NULL); action_resource = first_named_child(action->xml, XML_CIB_TAG_RESOURCE); CRM_CHECK(action_resource != NULL, crm_log_xml_warn(action->xml, "invalid"); return NULL); op = lrmd_new_event(ID(action_resource), crm_element_value(action->xml, XML_LRM_ATTR_TASK), action->interval_ms); lrmd__set_result(op, rc, status, exit_reason); op->t_run = time(NULL); op->t_rcchange = op->t_run; op->params = pcmk__strkey_table(free, free); g_hash_table_iter_init(&iter, action->params); while (g_hash_table_iter_next(&iter, (void **)&name, (void **)&value)) { g_hash_table_insert(op->params, strdup(name), strdup(value)); } for (xmlNode *xop = pcmk__xml_first_child(resource); xop != NULL; xop = pcmk__xml_next(xop)) { int tmp = 0; crm_element_value_int(xop, XML_LRM_ATTR_CALLID, &tmp); crm_debug("Got call_id=%d for %s", tmp, ID(resource)); if (tmp > op->call_id) { op->call_id = tmp; } } op->call_id++; return op; } diff --git a/lib/pacemaker/pcmk_graph_producer.c b/lib/pacemaker/pcmk_graph_producer.c index 5484e8b88b..9d430d24cb 100644 --- a/lib/pacemaker/pcmk_graph_producer.c +++ b/lib/pacemaker/pcmk_graph_producer.c @@ -1,1078 +1,1076 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include "libpacemaker_private.h" // Convenience macros for logging action properties #define action_type_str(flags) \ (pcmk_is_set((flags), pe_action_pseudo)? "pseudo-action" : "action") #define action_optional_str(flags) \ (pcmk_is_set((flags), pe_action_optional)? "optional" : "required") #define action_runnable_str(flags) \ (pcmk_is_set((flags), pe_action_runnable)? "runnable" : "unrunnable") #define action_node_str(a) \ (((a)->node == NULL)? "no node" : (a)->node->details->uname) /*! * \internal * \brief Add an XML node tag for a specified ID * * \param[in] id Node UUID to add * \param[in,out] xml Parent XML tag to add to */ static xmlNode* add_node_to_xml_by_id(const char *id, xmlNode *xml) { xmlNode *node_xml; node_xml = create_xml_node(xml, XML_CIB_TAG_NODE); crm_xml_add(node_xml, XML_ATTR_ID, id); return node_xml; } /*! * \internal * \brief Add an XML node tag for a specified node * * \param[in] node Node to add * \param[in,out] xml XML to add node to */ static void add_node_to_xml(const pe_node_t *node, void *xml) { add_node_to_xml_by_id(node->details->id, (xmlNode *) xml); } /*! * \internal * \brief Add XML with nodes that need an update of their maintenance state * * \param[in,out] xml Parent XML tag to add to * \param[in] data_set Working set for cluster */ static int add_maintenance_nodes(xmlNode *xml, const pe_working_set_t *data_set) { GList *gIter = NULL; xmlNode *maintenance = xml?create_xml_node(xml, XML_GRAPH_TAG_MAINTENANCE):NULL; int count = 0; for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; struct pe_node_shared_s *details = node->details; if (!pe__is_guest_or_remote_node(node)) { continue; /* just remote nodes need to know atm */ } if (details->maintenance != details->remote_maintenance) { if (maintenance) { crm_xml_add( add_node_to_xml_by_id(node->details->id, maintenance), XML_NODE_IS_MAINTENANCE, details->maintenance?"1":"0"); } count++; } } crm_trace("%s %d nodes to adjust maintenance-mode " "to transition", maintenance?"Added":"Counted", count); return count; } /*! * \internal * \brief Add pseudo action with nodes needing maintenance state update * * \param[in,out] data_set Working set for cluster */ static void add_maintenance_update(pe_working_set_t *data_set) { pe_action_t *action = NULL; if (add_maintenance_nodes(NULL, data_set)) { crm_trace("adding maintenance state update pseudo action"); action = get_pseudo_op(CRM_OP_MAINTENANCE_NODES, data_set); pe__set_action_flags(action, pe_action_print_always); } } /*! * \internal * \brief Add XML with nodes that an action is expected to bring down * * If a specified action is expected to bring any nodes down, add an XML block * with their UUIDs. When a node is lost, this allows the controller to * determine whether it was expected. * * \param[in,out] xml Parent XML tag to add to * \param[in] action Action to check for downed nodes * \param[in] data_set Working set for cluster */ static void add_downed_nodes(xmlNode *xml, const pe_action_t *action, const pe_working_set_t *data_set) { CRM_CHECK(xml && action && action->node && data_set, return); if (pcmk__str_eq(action->task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { /* Shutdown makes the action's node down */ xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* Fencing makes the action's node and any hosted guest nodes down */ const char *fence = g_hash_table_lookup(action->meta, "stonith_action"); if (pcmk__is_fencing_action(fence)) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); pe_foreach_guest_node(data_set, action->node, add_node_to_xml, downed); } } else if (action->rsc && action->rsc->is_remote_node && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { /* Stopping a remote connection resource makes connected node down, * unless it's part of a migration */ GList *iter; pe_action_t *input; - gboolean migrating = FALSE; + bool migrating = false; for (iter = action->actions_before; iter != NULL; iter = iter->next) { input = ((pe_action_wrapper_t *) iter->data)->action; if (input->rsc && pcmk__str_eq(action->rsc->id, input->rsc->id, pcmk__str_casei) && pcmk__str_eq(input->task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { - migrating = TRUE; + migrating = true; break; } } if (!migrating) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->rsc->id, downed); } } } /*! * \internal * \brief Create a transition graph operation key for a clone action * * \param[in] action Clone action * \param[in] interval_ms Action interval in milliseconds * * \return Newly allocated string with transition graph operation key */ static char * clone_op_key(const pe_action_t *action, guint interval_ms) { if (pcmk__str_eq(action->task, RSC_NOTIFY, pcmk__str_none)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_operation"); CRM_LOG_ASSERT((n_type != NULL) && (n_task != NULL)); return pcmk__notify_key(action->rsc->clone_name, n_type, n_task); } else if (action->cancel_task != NULL) { return pcmk__op_key(action->rsc->clone_name, action->cancel_task, interval_ms); } else { return pcmk__op_key(action->rsc->clone_name, action->task, interval_ms); } } /*! * \internal * \brief Add node details to transition graph action XML * * \param[in] action Scheduled action * \param[in,out] xml Transition graph action XML for \p action */ static void add_node_details(const pe_action_t *action, xmlNode *xml) { pe_node_t *router_node = pcmk__connection_host_for_action(action); crm_xml_add(xml, XML_LRM_ATTR_TARGET, action->node->details->uname); crm_xml_add(xml, XML_LRM_ATTR_TARGET_UUID, action->node->details->id); if (router_node != NULL) { crm_xml_add(xml, XML_LRM_ATTR_ROUTER_NODE, router_node->details->uname); } } /*! * \internal * \brief Add resource details to transition graph action XML * * \param[in] action Scheduled action * \param[in,out] action_xml Transition graph action XML for \p action */ static void add_resource_details(const pe_action_t *action, xmlNode *action_xml) { xmlNode *rsc_xml = NULL; const char *attr_list[] = { XML_AGENT_ATTR_CLASS, XML_AGENT_ATTR_PROVIDER, XML_ATTR_TYPE }; /* If a resource is locked to a node via shutdown-lock, mark its actions * so the controller can preserve the lock when the action completes. */ if (pcmk__action_locks_rsc_to_node(action)) { crm_xml_add_ll(action_xml, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) action->rsc->lock_time); } // List affected resource rsc_xml = create_xml_node(action_xml, crm_element_name(action->rsc->xml)); if (pcmk_is_set(action->rsc->flags, pe_rsc_orphan) && (action->rsc->clone_name != NULL)) { /* Use the numbered instance name here, because if there is more * than one instance on a node, we need to make sure the command * goes to the right one. * * This is important even for anonymous clones, because the clone's * unique meta-attribute might have just been toggled from on to * off. */ crm_debug("Using orphan clone name %s instead of %s", action->rsc->id, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } else if (!pcmk_is_set(action->rsc->flags, pe_rsc_unique)) { const char *xml_id = ID(action->rsc->xml); crm_debug("Using anonymous clone name %s for %s (aka %s)", xml_id, action->rsc->id, action->rsc->clone_name); /* ID is what we'd like client to use * ID_LONG is what they might know it as instead * * ID_LONG is only strictly needed /here/ during the * transition period until all nodes in the cluster * are running the new software /and/ have rebooted * once (meaning that they've only ever spoken to a DC * supporting this feature). * * If anyone toggles the unique flag to 'on', the * 'instance free' name will correspond to an orphan * and fall into the clause above instead */ crm_xml_add(rsc_xml, XML_ATTR_ID, xml_id); if ((action->rsc->clone_name != NULL) && !pcmk__str_eq(xml_id, action->rsc->clone_name, pcmk__str_none)) { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->clone_name); } else { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } } else { CRM_ASSERT(action->rsc->clone_name == NULL); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->id); } for (int lpc = 0; lpc < PCMK__NELEM(attr_list); lpc++) { crm_xml_add(rsc_xml, attr_list[lpc], g_hash_table_lookup(action->rsc->meta, attr_list[lpc])); } } /*! * \internal * \brief Add action attributes to transition graph action XML * * \param[in,out] action Scheduled action * \param[in,out] action_xml Transition graph action XML for \p action */ static void add_action_attributes(pe_action_t *action, xmlNode *action_xml) { xmlNode *args_xml = NULL; /* We create free-standing XML to start, so we can sort the attributes * before adding it to action_xml, which keeps the scheduler regression * test graphs comparable. */ args_xml = create_xml_node(NULL, XML_TAG_ATTRS); crm_xml_add(args_xml, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); g_hash_table_foreach(action->extra, hash2field, args_xml); if ((action->rsc != NULL) && (action->node != NULL)) { // Get the resource instance attributes, evaluated properly for node GHashTable *params = pe_rsc_params(action->rsc, action->node, action->rsc->cluster); pcmk__substitute_remote_addr(action->rsc, params); g_hash_table_foreach(params, hash2smartfield, args_xml); } else if ((action->rsc != NULL) && (action->rsc->variant <= pe_native)) { GHashTable *params = pe_rsc_params(action->rsc, NULL, action->rsc->cluster); g_hash_table_foreach(params, hash2smartfield, args_xml); } g_hash_table_foreach(action->meta, hash2metafield, args_xml); if (action->rsc != NULL) { pe_resource_t *parent = action->rsc; while (parent != NULL) { parent->cmds->add_graph_meta(parent, args_xml); parent = parent->parent; } pcmk__add_bundle_meta_to_xml(args_xml, action); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_none) && (action->node != NULL)) { /* Pass the node's attributes as meta-attributes. * * @TODO: Determine whether it is still necessary to do this. It was * added in 33d99707, probably for the libfence-based implementation in * c9a90bd, which is no longer used. */ g_hash_table_foreach(action->node->details->attrs, hash2metafield, args_xml); } sorted_xml(args_xml, action_xml, FALSE); free_xml(args_xml); } /*! * \internal * \brief Create the transition graph XML for a scheduled action * * \param[in,out] parent Parent XML element to add action to * \param[in,out] action Scheduled action * \param[in] skip_details If false, add action details as sub-elements * \param[in] data_set Cluster working set */ static void create_graph_action(xmlNode *parent, pe_action_t *action, bool skip_details, const pe_working_set_t *data_set) { bool needs_node_info = true; bool needs_maintenance_info = false; xmlNode *action_xml = NULL; if ((action == NULL) || (data_set == NULL)) { return; } // Create the top-level element based on task if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* All fences need node info; guest node fences are pseudo-events */ action_xml = create_xml_node(parent, pcmk_is_set(action->flags, pe_action_pseudo)? XML_GRAPH_TAG_PSEUDO_EVENT : XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_any_of(action->task, CRM_OP_SHUTDOWN, CRM_OP_CLEAR_FAILCOUNT, NULL)) { action_xml = create_xml_node(parent, XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_LRM_DELETE, pcmk__str_none)) { // CIB-only clean-up for shutdown locks action_xml = create_xml_node(parent, XML_GRAPH_TAG_CRM_EVENT); crm_xml_add(action_xml, PCMK__XA_MODE, XML_TAG_CIB); } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { if (pcmk__str_eq(action->task, CRM_OP_MAINTENANCE_NODES, pcmk__str_none)) { needs_maintenance_info = true; } action_xml = create_xml_node(parent, XML_GRAPH_TAG_PSEUDO_EVENT); needs_node_info = false; } else { action_xml = create_xml_node(parent, XML_GRAPH_TAG_RSC_OP); } crm_xml_add_int(action_xml, XML_ATTR_ID, action->id); crm_xml_add(action_xml, XML_LRM_ATTR_TASK, action->task); if ((action->rsc != NULL) && (action->rsc->clone_name != NULL)) { char *clone_key = NULL; guint interval_ms; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } clone_key = clone_op_key(action, interval_ms); crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, clone_key); crm_xml_add(action_xml, "internal_" XML_LRM_ATTR_TASK_KEY, action->uuid); free(clone_key); } else { crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, action->uuid); } if (needs_node_info && (action->node != NULL)) { add_node_details(action, action_xml); g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET), strdup(action->node->details->uname)); g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET_UUID), strdup(action->node->details->id)); } if (skip_details) { return; } if ((action->rsc != NULL) && !pcmk_is_set(action->flags, pe_action_pseudo)) { // This is a real resource action, so add resource details add_resource_details(action, action_xml); } /* List any attributes in effect */ add_action_attributes(action, action_xml); /* List any nodes this action is expected to make down */ if (needs_node_info && (action->node != NULL)) { add_downed_nodes(action_xml, action, data_set); } if (needs_maintenance_info) { add_maintenance_nodes(action_xml, data_set); } } /*! * \internal * \brief Check whether an action should be added to the transition graph * * \param[in] action Action to check * * \return true if action should be added to graph, otherwise false */ static bool should_add_action_to_graph(const pe_action_t *action) { if (!pcmk_is_set(action->flags, pe_action_runnable)) { crm_trace("Ignoring action %s (%d): unrunnable", action->uuid, action->id); return false; } if (pcmk_is_set(action->flags, pe_action_optional) && !pcmk_is_set(action->flags, pe_action_print_always)) { crm_trace("Ignoring action %s (%d): optional", action->uuid, action->id); return false; } /* Actions for unmanaged resources should be excluded from the graph, * with the exception of monitors and cancellation of recurring monitors. */ if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pe_rsc_managed) && !pcmk__str_eq(action->task, RSC_STATUS, pcmk__str_none)) { const char *interval_ms_s; /* A cancellation of a recurring monitor will get here because the task * is cancel rather than monitor, but the interval can still be used to * recognize it. The interval has been normalized to milliseconds by * this point, so a string comparison is sufficient. */ interval_ms_s = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); if (pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches)) { crm_trace("Ignoring action %s (%d): for unmanaged resource (%s)", action->uuid, action->id, action->rsc->id); return false; } } /* Always add pseudo-actions, fence actions, and shutdown actions (already * determined to be required and runnable by this point) */ if (pcmk_is_set(action->flags, pe_action_pseudo) || pcmk__strcase_any_of(action->task, CRM_OP_FENCE, CRM_OP_SHUTDOWN, NULL)) { return true; } if (action->node == NULL) { pe_err("Skipping action %s (%d) " - "because it was not allocated to a node (bug?)", + "because it was not assigned to a node (bug?)", action->uuid, action->id); - pcmk__log_action("Unallocated", action, false); + pcmk__log_action("Unassigned", action, false); return false; } if (pcmk_is_set(action->flags, pe_action_dc)) { crm_trace("Action %s (%d) should be dumped: " "can run on DC instead of %s", action->uuid, action->id, pe__node_name(action->node)); } else if (pe__is_guest_node(action->node) && !action->node->details->remote_requires_reset) { crm_trace("Action %s (%d) should be dumped: " "assuming will be runnable on guest %s", action->uuid, action->id, pe__node_name(action->node)); } else if (!action->node->details->online) { pe_err("Skipping action %s (%d) " "because it was scheduled for offline node (bug?)", action->uuid, action->id); pcmk__log_action("Offline node", action, false); return false; } else if (action->node->details->unclean) { pe_err("Skipping action %s (%d) " "because it was scheduled for unclean node (bug?)", action->uuid, action->id); pcmk__log_action("Unclean node", action, false); return false; } return true; } /*! * \internal * \brief Check whether an ordering's flags can change an action * * \param[in] ordering Ordering to check * * \return true if ordering has flags that can change an action, false otherwise */ static bool ordering_can_change_actions(const pe_action_wrapper_t *ordering) { return pcmk_any_flags_set(ordering->type, ~(pe_order_implies_first_printed |pe_order_implies_then_printed |pe_order_optional)); } /*! * \internal * \brief Check whether an action input should be in the transition graph * * \param[in] action Action to check * \param[in,out] input Action input to check * * \return true if input should be in graph, false otherwise * \note This function may not only check an input, but disable it under certian * circumstances (load or anti-colocation orderings that are not needed). */ static bool should_add_input_to_graph(const pe_action_t *action, pe_action_wrapper_t *input) { if (input->state == pe_link_dumped) { return true; } if (input->type == pe_order_none) { crm_trace("Ignoring %s (%d) input %s (%d): " "ordering disabled", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && !ordering_can_change_actions(input)) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && pcmk_is_set(input->type, pe_order_one_or_more)) { crm_trace("Ignoring %s (%d) input %s (%d): " "one-or-more and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_implies_first_migratable) && !pcmk_is_set(input->action->flags, pe_action_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "implies input migratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_apply_first_non_migratable) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "only if input unmigratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if ((input->type == pe_order_optional) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable) && pcmk__ends_with(input->action->uuid, "_stop_0")) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional but stop in migration", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (input->type == pe_order_load) { pe_node_t *input_node = input->action->node; // load orderings are relevant only if actions are for same node if (action->rsc && pcmk__str_eq(action->task, RSC_MIGRATE, pcmk__str_casei)) { - pe_node_t *allocated = action->rsc->allocated_to; + pe_node_t *assigned = action->rsc->allocated_to; /* For load_stopped -> migrate_to orderings, we care about where it - * has been allocated to, not where it will be executed. + * has been assigned to, not where it will be executed. */ - if ((input_node == NULL) || (allocated == NULL) - || (input_node->details != allocated->details)) { + if (!pe__same_node(input_node, assigned)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, - (allocated? allocated->details->uname : ""), + (assigned? assigned->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } - } else if ((input_node == NULL) || (action->node == NULL) - || (input_node->details != action->node->details)) { + } else if (!pe__same_node(input_node, action->node)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, (action->node? action->node->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->type == pe_order_anti_colocation) { if (input->action->node && action->node - && (input->action->node->details != action->node->details)) { + && !pe__same_node(input->action->node, action->node)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, pe__node_name(action->node), pe__node_name(input->action->node)); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->action->rsc && input->action->rsc != action->rsc && pcmk_is_set(input->action->rsc->flags, pe_rsc_failed) && !pcmk_is_set(input->action->rsc->flags, pe_rsc_managed) && pcmk__ends_with(input->action->uuid, "_stop_0") && action->rsc && pe_rsc_is_clone(action->rsc)) { crm_warn("Ignoring requirement that %s complete before %s:" " unmanaged failed resources cannot prevent clone shutdown", input->action->uuid, action->uuid); return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional) && !pcmk_any_flags_set(input->action->flags, pe_action_print_always|pe_action_dumped) && !should_add_action_to_graph(input->action)) { crm_trace("Ignoring %s (%d) input %s (%d): " "input optional", action->uuid, action->id, input->action->uuid, input->action->id); return false; } crm_trace("%s (%d) input %s %s (%d) on %s should be dumped: %s %s %#.6x", action->uuid, action->id, action_type_str(input->action->flags), input->action->uuid, input->action->id, action_node_str(input->action), action_runnable_str(input->action->flags), action_optional_str(input->action->flags), input->type); return true; } /*! * \internal * \brief Check whether an ordering creates an ordering loop * * \param[in] init_action "First" action in ordering * \param[in] action Callers should always set this the same as * \p init_action (this function may use a different * value for recursive calls) * \param[in,out] input Action wrapper for "then" action in ordering * * \return true if the ordering creates a loop, otherwise false */ bool pcmk__graph_has_loop(const pe_action_t *init_action, const pe_action_t *action, pe_action_wrapper_t *input) { bool has_loop = false; if (pcmk_is_set(input->action->flags, pe_action_tracking)) { crm_trace("Breaking tracking loop: %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); return false; } // Don't need to check inputs that won't be used if (!should_add_input_to_graph(action, input)) { return false; } if (input->action == init_action) { crm_debug("Input loop found in %s@%s ->...-> %s@%s", action->uuid, action->node? action->node->details->uname : "", init_action->uuid, init_action->node? init_action->node->details->uname : ""); return true; } pe__set_action_flags(input->action, pe_action_tracking); crm_trace("Checking inputs of action %s@%s input %s@%s (%#.6x)" "for graph loop with %s@%s ", action->uuid, action->node? action->node->details->uname : "", input->action->uuid, input->action->node? input->action->node->details->uname : "", input->type, init_action->uuid, init_action->node? init_action->node->details->uname : ""); // Recursively check input itself for loops for (GList *iter = input->action->actions_before; iter != NULL; iter = iter->next) { if (pcmk__graph_has_loop(init_action, input->action, (pe_action_wrapper_t *) iter->data)) { // Recursive call already logged a debug message has_loop = true; break; } } pe__clear_action_flags(input->action, pe_action_tracking); if (!has_loop) { crm_trace("No input loop found in %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); } return has_loop; } /*! * \internal * \brief Create a synapse XML element for a transition graph * * \param[in] action Action that synapse is for * \param[in,out] data_set Cluster working set containing graph * * \return Newly added XML element for new graph synapse */ static xmlNode * create_graph_synapse(const pe_action_t *action, pe_working_set_t *data_set) { int synapse_priority = 0; xmlNode *syn = create_xml_node(data_set->graph, "synapse"); crm_xml_add_int(syn, XML_ATTR_ID, data_set->num_synapse); data_set->num_synapse++; if (action->rsc != NULL) { synapse_priority = action->rsc->priority; } if (action->priority > synapse_priority) { synapse_priority = action->priority; } if (synapse_priority > 0) { crm_xml_add_int(syn, XML_CIB_ATTR_PRIORITY, synapse_priority); } return syn; } /*! * \internal * \brief Add an action to the transition graph XML if appropriate * * \param[in,out] data Action to possibly add * \param[in,out] user_data Cluster working set * * \note This will de-duplicate the action inputs, meaning that the * pe_action_wrapper_t:type flags can no longer be relied on to retain * their original settings. That means this MUST be called after * pcmk__apply_orderings() is complete, and nothing after this should rely * on those type flags. (For example, some code looks for type equal to * some flag rather than whether the flag is set, and some code looks for * particular combinations of flags -- such code must be done before * pcmk__create_graph().) */ static void add_action_to_graph(gpointer data, gpointer user_data) { pe_action_t *action = (pe_action_t *) data; pe_working_set_t *data_set = (pe_working_set_t *) user_data; xmlNode *syn = NULL; xmlNode *set = NULL; xmlNode *in = NULL; /* If we haven't already, de-duplicate inputs (even if we won't be adding * the action to the graph, so that crm_simulate's dot graphs don't have * duplicates). */ if (!pcmk_is_set(action->flags, pe_action_dedup)) { pcmk__deduplicate_action_inputs(action); pe__set_action_flags(action, pe_action_dedup); } if (pcmk_is_set(action->flags, pe_action_dumped) // Already added, or || !should_add_action_to_graph(action)) { // shouldn't be added return; } pe__set_action_flags(action, pe_action_dumped); crm_trace("Adding action %d (%s%s%s) to graph", action->id, action->uuid, ((action->node == NULL)? "" : " on "), ((action->node == NULL)? "" : action->node->details->uname)); syn = create_graph_synapse(action, data_set); set = create_xml_node(syn, "action_set"); in = create_xml_node(syn, "inputs"); create_graph_action(set, action, false, data_set); for (GList *lpc = action->actions_before; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *input = (pe_action_wrapper_t *) lpc->data; if (should_add_input_to_graph(action, input)) { xmlNode *input_xml = create_xml_node(in, "trigger"); input->state = pe_link_dumped; create_graph_action(input_xml, input->action, true, data_set); } } } static int transition_id = -1; /*! * \internal * \brief Log a message after calculating a transition * * \param[in] filename Where transition input is stored */ void pcmk__log_transition_summary(const char *filename) { if (was_processing_error) { crm_err("Calculated transition %d (with errors)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else if (was_processing_warning) { crm_warn("Calculated transition %d (with warnings)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else { crm_notice("Calculated transition %d%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } if (crm_config_error) { crm_notice("Configuration errors found during scheduler processing," " please run \"crm_verify -L\" to identify issues"); } } /*! * \internal * \brief Add a resource's actions to the transition graph * * \param[in,out] rsc Resource whose actions should be added */ void pcmk__add_rsc_actions_to_graph(pe_resource_t *rsc) { GList *iter = NULL; CRM_ASSERT(rsc != NULL); pe_rsc_trace(rsc, "Adding actions for %s to graph", rsc->id); // First add the resource's own actions g_list_foreach(rsc->actions, add_action_to_graph, rsc->cluster); // Then recursively add its children's actions (appropriate to variant) for (iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child_rsc = (pe_resource_t *) iter->data; child_rsc->cmds->add_actions_to_graph(child_rsc); } } /*! * \internal * \brief Create a transition graph with all cluster actions needed * * \param[in,out] data_set Cluster working set */ void pcmk__create_graph(pe_working_set_t *data_set) { GList *iter = NULL; const char *value = NULL; long long limit = 0LL; transition_id++; crm_trace("Creating transition graph %d", transition_id); data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH); value = pe_pref(data_set->config_hash, "cluster-delay"); crm_xml_add(data_set->graph, "cluster-delay", value); value = pe_pref(data_set->config_hash, "stonith-timeout"); crm_xml_add(data_set->graph, "stonith-timeout", value); crm_xml_add(data_set->graph, "failed-stop-offset", "INFINITY"); if (pcmk_is_set(data_set->flags, pe_flag_start_failure_fatal)) { crm_xml_add(data_set->graph, "failed-start-offset", "INFINITY"); } else { crm_xml_add(data_set->graph, "failed-start-offset", "1"); } value = pe_pref(data_set->config_hash, "batch-limit"); crm_xml_add(data_set->graph, "batch-limit", value); crm_xml_add_int(data_set->graph, "transition_id", transition_id); value = pe_pref(data_set->config_hash, "migration-limit"); if ((pcmk__scan_ll(value, &limit, 0LL) == pcmk_rc_ok) && (limit > 0)) { crm_xml_add(data_set->graph, "migration-limit", value); } if (data_set->recheck_by > 0) { char *recheck_epoch = NULL; recheck_epoch = crm_strdup_printf("%llu", (long long) data_set->recheck_by); crm_xml_add(data_set->graph, "recheck-by", recheck_epoch); free(recheck_epoch); } /* The following code will de-duplicate action inputs, so nothing past this * should rely on the action input type flags retaining their original * values. */ // Add resource actions to graph for (iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; pe_rsc_trace(rsc, "Processing actions for %s", rsc->id); rsc->cmds->add_actions_to_graph(rsc); } // Add pseudo-action for list of nodes with maintenance state update add_maintenance_update(data_set); // Add non-resource (node) actions for (iter = data_set->actions; iter != NULL; iter = iter->next) { pe_action_t *action = (pe_action_t *) iter->data; if ((action->rsc != NULL) && (action->node != NULL) && action->node->details->shutdown && !pcmk_is_set(action->rsc->flags, pe_rsc_maintenance) && !pcmk_any_flags_set(action->flags, pe_action_optional|pe_action_runnable) && pcmk__str_eq(action->task, RSC_STOP, pcmk__str_none)) { /* Eventually we should just ignore the 'fence' case, but for now * it's the best way to detect (in CTS) when CIB resource updates * are being lost. */ if (pcmk_is_set(data_set->flags, pe_flag_have_quorum) || (data_set->no_quorum_policy == no_quorum_ignore)) { crm_crit("Cannot %s %s because of %s:%s%s (%s)", action->node->details->unclean? "fence" : "shut down", pe__node_name(action->node), action->rsc->id, pcmk_is_set(action->rsc->flags, pe_rsc_managed)? " blocked" : " unmanaged", pcmk_is_set(action->rsc->flags, pe_rsc_failed)? " failed" : "", action->uuid); } } add_action_to_graph((gpointer) action, (gpointer) data_set); } crm_log_xml_trace(data_set->graph, "graph"); } diff --git a/lib/pacemaker/pcmk_output.c b/lib/pacemaker/pcmk_output.c index 73795169e7..2ea3b9d892 100644 --- a/lib/pacemaker/pcmk_output.c +++ b/lib/pacemaker/pcmk_output.c @@ -1,2331 +1,2331 @@ /* * Copyright 2019-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include static char * colocations_header(pe_resource_t *rsc, pcmk__colocation_t *cons, bool dependents) { char *retval = NULL; if (cons->primary_role > RSC_ROLE_STARTED) { retval = crm_strdup_printf("%s (score=%s, %s role=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), (dependents? "needs" : "with"), role2text(cons->primary_role), cons->id); } else { retval = crm_strdup_printf("%s (score=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), cons->id); } return retval; } static void colocations_xml_node(pcmk__output_t *out, pe_resource_t *rsc, pcmk__colocation_t *cons) { xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_DEPEND, "id", cons->id, "rsc", cons->dependent->id, "with-rsc", cons->primary->id, "score", pcmk_readable_score(cons->score), NULL); if (cons->node_attribute) { xmlSetProp(node, (pcmkXmlStr) "node-attribute", (pcmkXmlStr) cons->node_attribute); } if (cons->dependent_role != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "rsc-role", (pcmkXmlStr) role2text(cons->dependent_role)); } if (cons->primary_role != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "with-rsc-role", (pcmkXmlStr) role2text(cons->primary_role)); } } static int do_locations_list_xml(pcmk__output_t *out, pe_resource_t *rsc, bool add_header) { GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; if (add_header) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "locations"); } pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_LOCATION, "node", node->details->uname, "rsc", rsc->id, "id", cons->id, "score", pcmk_readable_score(node->weight), NULL); } } if (add_header) { PCMK__OUTPUT_LIST_FOOTER(out, rc); } return rc; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pe_resource_t *", "pe_node_t *", "pe_node_t *", "pe_action_t *", "pe_action_t *") static int rsc_action_item(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *origin = va_arg(args, pe_node_t *); pe_node_t *destination = va_arg(args, pe_node_t *); pe_action_t *action = va_arg(args, pe_action_t *); pe_action_t *source = va_arg(args, pe_action_t *); int len = 0; char *reason = NULL; char *details = NULL; bool same_host = false; bool same_role = false; bool need_role = false; static int rsc_width = 5; static int detail_width = 5; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if(source == NULL) { source = action; } len = strlen(rsc->id); if(len > rsc_width) { rsc_width = len + 2; } if ((rsc->role > RSC_ROLE_STARTED) || (rsc->next_role > RSC_ROLE_UNPROMOTED)) { need_role = true; } - if(origin != NULL && destination != NULL && origin->details == destination->details) { + if (pe__same_node(origin, destination)) { same_host = true; } if(rsc->role == rsc->next_role) { same_role = true; } if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(destination)); } else if (origin == NULL) { /* Starting a resource */ details = crm_strdup_printf("%s", pe__node_name(destination)); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (destination == NULL) { /* Stopping a resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role) { /* Moving a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", pe__node_name(origin), pe__node_name(destination), role2text(rsc->role)); } else if (same_role) { /* Moving a normal resource */ details = crm_strdup_printf("%s -> %s", pe__node_name(origin), pe__node_name(destination)); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(origin)); } else { /* Moving and promoting/demoting */ details = crm_strdup_printf("%s %s -> %s %s", role2text(rsc->role), pe__node_name(origin), role2text(rsc->next_role), pe__node_name(destination)); } len = strlen(details); if(len > detail_width) { detail_width = len; } if(source->reason && !pcmk_is_set(action->flags, pe_action_runnable)) { reason = crm_strdup_printf("due to %s (blocked)", source->reason); } else if(source->reason) { reason = crm_strdup_printf("due to %s", source->reason); } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { reason = strdup("blocked"); } out->list_item(out, NULL, "%-8s %-*s ( %*s )%s%s", change, rsc_width, rsc->id, detail_width, details, reason ? " " : "", reason ? reason : ""); free(details); free(reason); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pe_resource_t *", "pe_node_t *", "pe_node_t *", "pe_action_t *", "pe_action_t *") static int rsc_action_item_xml(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *origin = va_arg(args, pe_node_t *); pe_node_t *destination = va_arg(args, pe_node_t *); pe_action_t *action = va_arg(args, pe_action_t *); pe_action_t *source = va_arg(args, pe_action_t *); char *change_str = NULL; bool same_host = false; bool same_role = false; bool need_role = false; xmlNode *xml = NULL; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if (source == NULL) { source = action; } if ((rsc->role > RSC_ROLE_STARTED) || (rsc->next_role > RSC_ROLE_UNPROMOTED)) { need_role = true; } - if(origin != NULL && destination != NULL && origin->details == destination->details) { + if (pe__same_node(origin, destination)) { same_host = true; } if(rsc->role == rsc->next_role) { same_role = true; } change_str = g_ascii_strdown(change, -1); xml = pcmk__output_create_xml_node(out, "rsc_action", "action", change_str, "resource", rsc->id, NULL); g_free(change_str); if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } else if (origin == NULL) { /* Starting a resource */ crm_xml_add(xml, "node", destination->details->uname); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "node", origin->details->uname, NULL); } else if (destination == NULL) { /* Stopping a resource */ crm_xml_add(xml, "node", origin->details->uname); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, NULL); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ crm_xml_add(xml, "source", origin->details->uname); } else if (need_role && same_role) { /* Moving a promotable clone instance */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, "role", role2text(rsc->role), NULL); } else if (same_role) { /* Moving a normal resource */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, NULL); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "source", origin->details->uname, NULL); } else { /* Moving and promoting/demoting */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } if (source->reason && !pcmk_is_set(action->flags, pe_action_runnable)) { pcmk__xe_set_props(xml, "reason", source->reason, "blocked", "true", NULL); } else if(source->reason) { crm_xml_add(xml, "reason", source->reason); } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { pcmk__xe_set_bool_attr(xml, "blocked", true); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "bool") static int rsc_is_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons * directly rather than rsc->cmds->this_with_colocations(). */ pe__set_resource_flags(rsc, pe_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources %s is colocated with", rsc->id); if (pcmk_is_set(cons->primary->flags, pe_rsc_detect_loop)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->primary->id, cons->id); continue; } hdr = colocations_header(cons->primary, cons, false); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->primary); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "bool") static int rsc_is_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons * directly rather than rsc->cmds->this_with_colocations(). */ pe__set_resource_flags(rsc, pe_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->primary->flags, pe_rsc_detect_loop)) { colocations_xml_node(out, cons->primary, cons); continue; } colocations_xml_node(out, cons->primary, cons); do_locations_list_xml(out, cons->primary, false); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } } return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "bool") static int rscs_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use * rsc->rsc_cons_lhs directly rather than * rsc->cmds->with_this_colocations(). */ pe__set_resource_flags(rsc, pe_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources colocated with %s", rsc->id); if (pcmk_is_set(cons->dependent->flags, pe_rsc_detect_loop)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->dependent->id, cons->id); continue; } hdr = colocations_header(cons->dependent, cons, true); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->dependent); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "bool") static int rscs_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use * rsc->rsc_cons_lhs directly rather than * rsc->cmds->with_this_colocations(). */ pe__set_resource_flags(rsc, pe_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->dependent->flags, pe_rsc_detect_loop)) { colocations_xml_node(out, cons->dependent, cons); continue; } colocations_xml_node(out, cons->dependent, cons); do_locations_list_xml(out, cons->dependent, false); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } } return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Locations"); out->list_item(out, NULL, "Node %s (score=%s, id=%s, rsc=%s)", pe__node_name(node), pcmk_readable_score(node->weight), cons->id, rsc->id); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); return do_locations_list_xml(out, rsc, true); } PCMK__OUTPUT_ARGS("locations-and-colocations", "pe_resource_t *", "pe_working_set_t *", "bool", "bool") static int locations_and_colocations(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(data_set); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } out->message(out, "locations-list", rsc); pe__clear_resource_flags_on_all(data_set, pe_rsc_detect_loop); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_detect_loop); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("locations-and-colocations", "pe_resource_t *", "pe_working_set_t *", "bool", "bool") static int locations_and_colocations_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(data_set); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } pcmk__output_xml_create_parent(out, "constraints", NULL); do_locations_list_xml(out, rsc, false); pe__clear_resource_flags_on_all(data_set, pe_rsc_detect_loop); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_detect_loop); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); pcmk__output_xml_pop_parent(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health(pcmk__output_t *out, va_list args) { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); return out->info(out, "Controller on %s in state %s: %s", pcmk__s(host_from, "unknown node"), pcmk__s(fsa_state, "unknown"), pcmk__s(result, "unknown result")); } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from G_GNUC_UNUSED = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result G_GNUC_UNUSED = va_arg(args, const char *); if (fsa_state != NULL) { pcmk__formatted_printf(out, "%s\n", fsa_state); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); pcmk__output_create_xml_node(out, pcmk__s(sys_from, ""), "node_name", pcmk__s(host_from, ""), "state", pcmk__s(fsa_state, ""), "result", pcmk__s(result, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; int rc = pcmk_rc_ok; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } rc = out->info(out, "Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); free(last_updated_s); return rc; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_html(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; char *msg = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } msg = crm_strdup_printf("Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); pcmk__output_create_html_node(out, "li", NULL, NULL, msg); free(msg); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return pacemakerd_health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated G_GNUC_UNUSED = va_arg(args, time_t); if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } pcmk__formatted_printf(out, "%s\n", state_s); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } pcmk__output_create_xml_node(out, "pacemakerd", "sys_from", sys_from, "state", state_s, "last_updated", last_updated_s, NULL); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_default(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); out->list_item(out, NULL, "Testing %s ... %.2f secs", xml_file, (end - start) / (float) CLOCKS_PER_SEC); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_xml(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); char *duration = pcmk__ftoa((end - start) / (float) CLOCKS_PER_SEC); pcmk__output_create_xml_node(out, "timing", "file", xml_file, "duration", duration, NULL); free(duration); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); return out->info(out, "Designated Controller is: %s", pcmk__s(dc, "not yet elected")); } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return dc(out, args); } else { const char *dc = va_arg(args, const char *); if (dc != NULL) { pcmk__formatted_printf(out, "%s\n", pcmk__s(dc, "")); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_xml(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); pcmk__output_create_xml_node(out, "dc", "node_name", pcmk__s(dc, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export = va_arg(args, int); if (bash_export) { return out->info(out, "export %s=%s", pcmk__s(name, ""), pcmk__s(id, "")); } else { return out->info(out, "%s node: %s (%s)", type ? type : "cluster", pcmk__s(name, ""), pcmk__s(id, "")); } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return crmadmin_node(out, args); } else { const char *type G_GNUC_UNUSED = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id G_GNUC_UNUSED = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__formatted_printf(out, "%s\n", pcmk__s(name, "")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_xml(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__output_create_xml_node(out, "node", "type", type ? type : "cluster", "name", pcmk__s(name, ""), "id", pcmk__s(id, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("digests", "const pe_resource_t *", "const pe_node_t *", "const char *", "guint", "const op_digest_cache_t *") static int digests_text(pcmk__output_t *out, va_list args) { const pe_resource_t *rsc = va_arg(args, const pe_resource_t *); const pe_node_t *node = va_arg(args, const pe_node_t *); const char *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); const op_digest_cache_t *digests = va_arg(args, const op_digest_cache_t *); char *action_desc = NULL; const char *rsc_desc = "unknown resource"; const char *node_desc = "unknown node"; if (interval_ms != 0) { action_desc = crm_strdup_printf("%ums-interval %s action", interval_ms, ((task == NULL)? "unknown" : task)); } else if (pcmk__str_eq(task, "monitor", pcmk__str_none)) { action_desc = strdup("probe action"); } else { action_desc = crm_strdup_printf("%s action", ((task == NULL)? "unknown" : task)); } if ((rsc != NULL) && (rsc->id != NULL)) { rsc_desc = rsc->id; } if ((node != NULL) && (node->details->uname != NULL)) { node_desc = node->details->uname; } out->begin_list(out, NULL, NULL, "Digests for %s %s on %s", rsc_desc, action_desc, node_desc); free(action_desc); if (digests == NULL) { out->list_item(out, NULL, "none"); out->end_list(out); return pcmk_rc_ok; } if (digests->digest_all_calc != NULL) { out->list_item(out, NULL, "%s (all parameters)", digests->digest_all_calc); } if (digests->digest_secure_calc != NULL) { out->list_item(out, NULL, "%s (non-private parameters)", digests->digest_secure_calc); } if (digests->digest_restart_calc != NULL) { out->list_item(out, NULL, "%s (non-reloadable parameters)", digests->digest_restart_calc); } out->end_list(out); return pcmk_rc_ok; } static void add_digest_xml(xmlNode *parent, const char *type, const char *digest, xmlNode *digest_source) { if (digest != NULL) { xmlNodePtr digest_xml = create_xml_node(parent, "digest"); crm_xml_add(digest_xml, "type", ((type == NULL)? "unspecified" : type)); crm_xml_add(digest_xml, "hash", digest); if (digest_source != NULL) { add_node_copy(digest_xml, digest_source); } } } PCMK__OUTPUT_ARGS("digests", "const pe_resource_t *", "const pe_node_t *", "const char *", "guint", "const op_digest_cache_t *") static int digests_xml(pcmk__output_t *out, va_list args) { const pe_resource_t *rsc = va_arg(args, const pe_resource_t *); const pe_node_t *node = va_arg(args, const pe_node_t *); const char *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); const op_digest_cache_t *digests = va_arg(args, const op_digest_cache_t *); char *interval_s = crm_strdup_printf("%ums", interval_ms); xmlNode *xml = NULL; xml = pcmk__output_create_xml_node(out, "digests", "resource", pcmk__s(rsc->id, ""), "node", pcmk__s(node->details->uname, ""), "task", pcmk__s(task, ""), "interval", interval_s, NULL); free(interval_s); if (digests != NULL) { add_digest_xml(xml, "all", digests->digest_all_calc, digests->params_all); add_digest_xml(xml, "nonprivate", digests->digest_secure_calc, digests->params_secure); add_digest_xml(xml, "nonreloadable", digests->digest_restart_calc, digests->params_restart); } return pcmk_rc_ok; } #define STOP_SANITY_ASSERT(lineno) do { \ if(current && current->details->unclean) { \ /* It will be a pseudo op */ \ } else if(stop == NULL) { \ crm_err("%s:%d: No stop action exists for %s", \ __func__, lineno, rsc->id); \ CRM_ASSERT(stop != NULL); \ } else if (pcmk_is_set(stop->flags, pe_action_optional)) { \ crm_err("%s:%d: Action %s is still optional", \ __func__, lineno, stop->uuid); \ CRM_ASSERT(!pcmk_is_set(stop->flags, pe_action_optional)); \ } \ } while(0) PCMK__OUTPUT_ARGS("rsc-action", "pe_resource_t *", "pe_node_t *", "pe_node_t *") static int rsc_action_default(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *current = va_arg(args, pe_node_t *); pe_node_t *next = va_arg(args, pe_node_t *); GList *possible_matches = NULL; char *key = NULL; int rc = pcmk_rc_no_output; bool moving = false; pe_node_t *start_node = NULL; pe_action_t *start = NULL; pe_action_t *stop = NULL; pe_action_t *promote = NULL; pe_action_t *demote = NULL; if (!pcmk_is_set(rsc->flags, pe_rsc_managed) || (current == NULL && next == NULL)) { pe_rsc_info(rsc, "Leave %s\t(%s%s)", rsc->id, role2text(rsc->role), !pcmk_is_set(rsc->flags, pe_rsc_managed)? " unmanaged" : ""); return rc; } moving = (current != NULL) && (next != NULL) - && (current->details != next->details); + && !pe__same_node(current, next); possible_matches = pe__resource_actions(rsc, next, RSC_START, false); if (possible_matches) { start = possible_matches->data; g_list_free(possible_matches); } if ((start == NULL) || !pcmk_is_set(start->flags, pe_action_runnable)) { start_node = NULL; } else { start_node = current; } possible_matches = pe__resource_actions(rsc, start_node, RSC_STOP, false); if (possible_matches) { stop = possible_matches->data; g_list_free(possible_matches); } else if (pcmk_is_set(rsc->flags, pe_rsc_stop_unexpected)) { /* The resource is multiply active with multiple-active set to * stop_unexpected, and not stopping on its current node, but it should * be stopping elsewhere. */ possible_matches = pe__resource_actions(rsc, NULL, RSC_STOP, false); if (possible_matches != NULL) { stop = possible_matches->data; g_list_free(possible_matches); } } possible_matches = pe__resource_actions(rsc, next, RSC_PROMOTE, false); if (possible_matches) { promote = possible_matches->data; g_list_free(possible_matches); } possible_matches = pe__resource_actions(rsc, next, RSC_DEMOTE, false); if (possible_matches) { demote = possible_matches->data; g_list_free(possible_matches); } if (rsc->role == rsc->next_role) { pe_action_t *migrate_op = NULL; CRM_CHECK(next != NULL, return rc); possible_matches = pe__resource_actions(rsc, next, RSC_MIGRATED, false); if (possible_matches) { migrate_op = possible_matches->data; } if ((migrate_op != NULL) && (current != NULL) && pcmk_is_set(migrate_op->flags, pe_action_runnable)) { rc = out->message(out, "rsc-action-item", "Migrate", rsc, current, next, start, NULL); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if (start == NULL || pcmk_is_set(start->flags, pe_action_optional)) { if ((demote != NULL) && (promote != NULL) && !pcmk_is_set(demote->flags, pe_action_optional) && !pcmk_is_set(promote->flags, pe_action_optional)) { rc = out->message(out, "rsc-action-item", "Re-promote", rsc, current, next, promote, demote); } else { pe_rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id, role2text(rsc->role), pe__node_name(next)); } } else if (!pcmk_is_set(start->flags, pe_action_runnable)) { rc = out->message(out, "rsc-action-item", "Stop", rsc, current, NULL, stop, (stop && stop->reason)? stop : start); STOP_SANITY_ASSERT(__LINE__); } else if (moving && current) { rc = out->message(out, "rsc-action-item", pcmk_is_set(rsc->flags, pe_rsc_failed)? "Recover" : "Move", rsc, current, next, stop, NULL); } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { rc = out->message(out, "rsc-action-item", "Recover", rsc, current, NULL, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); /* STOP_SANITY_ASSERT(__LINE__); False positive for migrate-fail-7 */ } g_list_free(possible_matches); return rc; } if(stop && (rsc->next_role == RSC_ROLE_STOPPED || (start && !pcmk_is_set(start->flags, pe_action_runnable)))) { GList *gIter = NULL; key = stop_key(rsc); for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pe_action_t *stop_op = NULL; possible_matches = find_actions(rsc->actions, key, node); if (possible_matches) { stop_op = possible_matches->data; g_list_free(possible_matches); } if (stop_op && (stop_op->flags & pe_action_runnable)) { STOP_SANITY_ASSERT(__LINE__); } if (out->message(out, "rsc-action-item", "Stop", rsc, node, NULL, stop_op, (stop_op && stop_op->reason)? stop_op : start) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } free(key); } else if ((stop != NULL) && pcmk_all_flags_set(rsc->flags, pe_rsc_failed|pe_rsc_stop)) { /* 'stop' may be NULL if the failure was ignored */ rc = out->message(out, "rsc-action-item", "Recover", rsc, current, next, stop, start); STOP_SANITY_ASSERT(__LINE__); } else if (moving) { rc = out->message(out, "rsc-action-item", "Move", rsc, current, next, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if (stop != NULL && !pcmk_is_set(stop->flags, pe_action_optional)) { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (rsc->role == RSC_ROLE_PROMOTED) { CRM_LOG_ASSERT(current != NULL); rc = out->message(out, "rsc-action-item", "Demote", rsc, current, next, demote, NULL); } else if (rsc->next_role == RSC_ROLE_PROMOTED) { CRM_LOG_ASSERT(next); rc = out->message(out, "rsc-action-item", "Promote", rsc, current, next, promote, NULL); } else if (rsc->role == RSC_ROLE_STOPPED && rsc->next_role > RSC_ROLE_STOPPED) { rc = out->message(out, "rsc-action-item", "Start", rsc, current, next, start, NULL); } return rc; } PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *") static int node_action(pcmk__output_t *out, va_list args) { const char *task = va_arg(args, const char *); const char *node_name = va_arg(args, const char *); const char *reason = va_arg(args, const char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { out->list_item(out, NULL, "%s %s '%s'", task, node_name, reason); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *") static int node_action_xml(pcmk__output_t *out, va_list args) { const char *task = va_arg(args, const char *); const char *node_name = va_arg(args, const char *); const char *reason = va_arg(args, const char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { pcmk__output_create_xml_node(out, "node_action", "task", task, "node", node_name, "reason", reason, NULL); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-info", "int", "const char *", "const char *", "const char *", "bool", "bool") static int node_info_default(pcmk__output_t *out, va_list args) { int node_id = va_arg(args, int); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); return out->info(out, "Node %d: %s " "(uuid=%s, state=%s, have_quorum=%s, is_remote=%s)", node_id, pcmk__s(node_name, "unknown"), pcmk__s(uuid, "unknown"), pcmk__s(state, "unknown"), pcmk__btoa(have_quorum), pcmk__btoa(is_remote)); } PCMK__OUTPUT_ARGS("node-info", "int", "const char *", "const char *", "const char *", "bool", "bool") static int node_info_xml(pcmk__output_t *out, va_list args) { int node_id = va_arg(args, int); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); char *id_s = crm_strdup_printf("%d", node_id); pcmk__output_create_xml_node(out, "node-info", "nodeid", id_s, XML_ATTR_UNAME, node_name, XML_ATTR_ID, uuid, XML_NODE_IS_PEER, state, XML_ATTR_HAVE_QUORUM, pcmk__btoa(have_quorum), XML_NODE_IS_REMOTE, pcmk__btoa(is_remote), NULL); free(id_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if(rsc) { out->list_item(out, NULL, "Cluster action: %s for %s on %s", task, ID(rsc), node); } else { out->list_item(out, NULL, "Cluster action: %s on %s", task, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "cluster_action", "task", task, "node", node, NULL); if (rsc) { crm_xml_add(xml_node, "id", ID(rsc)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *") static int inject_fencing_action(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Fencing %s (%s)", target, op); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *") static int inject_fencing_action_xml(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "fencing_action", "target", target, "op", op, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); out->list_item(out, NULL, "Injecting attribute %s=%s into %s '%s'", name, value, node_path, ID(cib_node)); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); pcmk__output_create_xml_node(out, "inject_attr", "name", name, "value", value, "node_path", node_path, "cib_node", ID(cib_node), NULL); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Injecting %s into the configuration", spec); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec_xml(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "inject_spec", "spec", spec, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *") static int inject_modify_config(pcmk__output_t *out, va_list args) { const char *quorum = va_arg(args, const char *); const char *watchdog = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->begin_list(out, NULL, NULL, "Performing Requested Modifications"); if (quorum) { out->list_item(out, NULL, "Setting quorum: %s", quorum); } if (watchdog) { out->list_item(out, NULL, "Setting watchdog: %s", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *") static int inject_modify_config_xml(pcmk__output_t *out, va_list args) { const char *quorum = va_arg(args, const char *); const char *watchdog = va_arg(args, const char *); xmlNodePtr node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node = pcmk__output_xml_create_parent(out, "modifications", NULL); if (quorum) { crm_xml_add(node, "quorum", quorum); } if (watchdog) { crm_xml_add(node, "watchdog", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *") static int inject_modify_node(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *node = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Online", pcmk__str_none)) { out->list_item(out, NULL, "Bringing node %s online", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Offline", pcmk__str_none)) { out->list_item(out, NULL, "Taking node %s offline", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Failing", pcmk__str_none)) { out->list_item(out, NULL, "Failing node %s", node); return pcmk_rc_ok; } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *") static int inject_modify_node_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *node = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_node", "action", action, "node", node, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *") static int inject_modify_ticket(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *ticket = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Standby", pcmk__str_none)) { out->list_item(out, NULL, "Making ticket %s standby", ticket); } else { out->list_item(out, NULL, "%s ticket %s", action, ticket); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *") static int inject_modify_ticket_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *ticket = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_ticket", "action", action, "ticket", ticket, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Pseudo action: %s%s%s", task, node ? " on " : "", node ? node : ""); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "pseudo_action", "task", task, NULL); if (node) { crm_xml_add(xml_node, "node", node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "const char *", "guint") static int inject_rsc_action(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); const char *node = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (interval_ms) { out->list_item(out, NULL, "Resource action: %-15s %s=%u on %s", rsc, operation, interval_ms, node); } else { out->list_item(out, NULL, "Resource action: %-15s %s on %s", rsc, operation, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "const char *", "guint") static int inject_rsc_action_xml(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); const char *node = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "rsc_action", "resource", rsc, "op", operation, "node", node, NULL); if (interval_ms) { char *interval_s = pcmk__itoa(interval_ms); crm_xml_add(xml_node, "interval", interval_s); free(interval_s); } return pcmk_rc_ok; } #define CHECK_RC(retcode, retval) \ if (retval == pcmk_rc_ok) { \ retcode = pcmk_rc_ok; \ } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") int pcmk__cluster_status_text(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); int rc = pcmk_rc_no_output; bool already_printed_failure = false; CHECK_RC(rc, out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts)); if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { CHECK_RC(rc, out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { CHECK_RC(rc, out->message(out, "resource-list", data_set, show_opts, true, unames, resources, rc == pcmk_rc_ok)); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { CHECK_RC(rc, out->message(out, "node-attribute-list", data_set, show_opts, rc == pcmk_rc_ok, unames, resources)); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { CHECK_RC(rc, out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, rc == pcmk_rc_ok)); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { CHECK_RC(rc, out->message(out, "failed-action-list", data_set, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); already_printed_failure = true; } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { CHECK_RC(rc, out->message(out, "ticket-list", data_set, rc == pcmk_rc_ok)); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { CHECK_RC(rc, out->message(out, "ban-list", data_set, prefix, resources, show_opts, rc == pcmk_rc_ok)); } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { CHECK_RC(rc, out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } } return rc; } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_xml(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts); /*** NODES ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes)) { out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { /* XML output always displays full details. */ uint32_t full_show_opts = show_opts & ~pcmk_show_brief; out->message(out, "resource-list", data_set, full_show_opts, false, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", data_set, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { out->message(out, "failed-action-list", data_set, unames, resources, show_opts, false); } /* Print stonith history */ if (pcmk_is_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { out->message(out, "full-fencing-list", history_rc, stonith_history, unames, section_opts, show_opts, false); } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", data_set, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", data_set, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_html(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); bool already_printed_failure = false; out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts); /*** NODE LIST ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { out->message(out, "resource-list", data_set, show_opts, true, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", data_set, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { out->message(out, "failed-action-list", data_set, unames, resources, show_opts, false); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, false); } } else { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "fencing-list", hp, unames, section_opts, show_opts, false); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, false); } } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", data_set, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", data_set, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_default(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); GString *s = g_string_sized_new(50); if (!pcmk__str_empty(scope)) { pcmk__g_strcat(s, "scope=\"", scope, "\" ", NULL); } if (!pcmk__str_empty(instance)) { pcmk__g_strcat(s, "id=\"", instance, "\" ", NULL); } pcmk__g_strcat(s, "name=\"", pcmk__s(name, ""), "\" ", NULL); if (!pcmk__str_empty(host)) { pcmk__g_strcat(s, "host=\"", host, "\" ", NULL); } pcmk__g_strcat(s, "value=\"", pcmk__s(value, ""), "\"", NULL); out->info(out, "%s", s->str); g_string_free(s, TRUE); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_xml(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, "attribute", "name", name, "value", value ? value : "", NULL); if (!pcmk__str_empty(scope)) { crm_xml_add(node, "scope", scope); } if (!pcmk__str_empty(instance)) { crm_xml_add(node, "id", instance); } if (!pcmk__str_empty(host)) { crm_xml_add(node, "host", host); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_default(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); switch (result) { case pcmk_rc_within_range: return out->info(out, "Rule %s is still in effect", rule_id); case pcmk_rc_ok: return out->info(out, "Rule %s satisfies conditions", rule_id); case pcmk_rc_after_range: return out->info(out, "Rule %s is expired", rule_id); case pcmk_rc_before_range: return out->info(out, "Rule %s has not yet taken effect", rule_id); case pcmk_rc_op_unsatisfied: return out->info(out, "Rule %s does not satisfy conditions", rule_id); default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_xml(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); char *rc_str = pcmk__itoa(pcmk_rc2exitc(result)); pcmk__output_create_xml_node(out, "rule-check", "rule-id", rule_id, "rc", rc_str, NULL); free(rc_str); switch (result) { case pcmk_rc_within_range: case pcmk_rc_ok: case pcmk_rc_after_range: case pcmk_rc_before_range: case pcmk_rc_op_unsatisfied: return pcmk_rc_ok; default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_none(pcmk__output_t *out, va_list args) { return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_text(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); static int code_width = 0; if (out->is_quiet(out)) { /* If out->is_quiet(), don't print the code. Print name and/or desc in a * compact format for text output, or print nothing at all for none-type * output. */ if ((name != NULL) && (desc != NULL)) { pcmk__formatted_printf(out, "%s - %s\n", name, desc); } else if ((name != NULL) || (desc != NULL)) { pcmk__formatted_printf(out, "%s\n", ((name != NULL)? name : desc)); } return pcmk_rc_ok; } /* Get length of longest (most negative) standard Pacemaker return code * This should be longer than all the values of any other type of return * code. */ if (code_width == 0) { long long most_negative = pcmk_rc_error - (long long) pcmk__n_rc + 1; code_width = (int) snprintf(NULL, 0, "%lld", most_negative); } if ((name != NULL) && (desc != NULL)) { static int name_width = 0; if (name_width == 0) { // Get length of longest standard Pacemaker return code name for (int lpc = 0; lpc < pcmk__n_rc; lpc++) { int len = (int) strlen(pcmk_rc_name(pcmk_rc_error - lpc)); name_width = QB_MAX(name_width, len); } } return out->info(out, "% *d: %-*s %s", code_width, code, name_width, name, desc); } if ((name != NULL) || (desc != NULL)) { return out->info(out, "% *d: %s", code_width, code, ((name != NULL)? name : desc)); } return out->info(out, "% *d", code_width, code); } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_xml(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); char *code_str = pcmk__itoa(code); pcmk__output_create_xml_node(out, "result-code", "code", code_str, XML_ATTR_NAME, name, XML_ATTR_DESC, desc, NULL); free(code_str); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "attribute", "default", attribute_default }, { "attribute", "xml", attribute_xml }, { "cluster-status", "default", pcmk__cluster_status_text }, { "cluster-status", "html", cluster_status_html }, { "cluster-status", "xml", cluster_status_xml }, { "crmadmin-node", "default", crmadmin_node }, { "crmadmin-node", "text", crmadmin_node_text }, { "crmadmin-node", "xml", crmadmin_node_xml }, { "dc", "default", dc }, { "dc", "text", dc_text }, { "dc", "xml", dc_xml }, { "digests", "default", digests_text }, { "digests", "xml", digests_xml }, { "health", "default", health }, { "health", "text", health_text }, { "health", "xml", health_xml }, { "inject-attr", "default", inject_attr }, { "inject-attr", "xml", inject_attr_xml }, { "inject-cluster-action", "default", inject_cluster_action }, { "inject-cluster-action", "xml", inject_cluster_action_xml }, { "inject-fencing-action", "default", inject_fencing_action }, { "inject-fencing-action", "xml", inject_fencing_action_xml }, { "inject-modify-config", "default", inject_modify_config }, { "inject-modify-config", "xml", inject_modify_config_xml }, { "inject-modify-node", "default", inject_modify_node }, { "inject-modify-node", "xml", inject_modify_node_xml }, { "inject-modify-ticket", "default", inject_modify_ticket }, { "inject-modify-ticket", "xml", inject_modify_ticket_xml }, { "inject-pseudo-action", "default", inject_pseudo_action }, { "inject-pseudo-action", "xml", inject_pseudo_action_xml }, { "inject-rsc-action", "default", inject_rsc_action }, { "inject-rsc-action", "xml", inject_rsc_action_xml }, { "inject-spec", "default", inject_spec }, { "inject-spec", "xml", inject_spec_xml }, { "locations-list", "default", locations_list }, { "locations-list", "xml", locations_list_xml }, { "node-action", "default", node_action }, { "node-action", "xml", node_action_xml }, { "node-info", "default", node_info_default }, { "node-info", "xml", node_info_xml }, { "pacemakerd-health", "default", pacemakerd_health }, { "pacemakerd-health", "html", pacemakerd_health_html }, { "pacemakerd-health", "text", pacemakerd_health_text }, { "pacemakerd-health", "xml", pacemakerd_health_xml }, { "profile", "default", profile_default, }, { "profile", "xml", profile_xml }, { "result-code", "none", result_code_none }, { "result-code", "text", result_code_text }, { "result-code", "xml", result_code_xml }, { "rsc-action", "default", rsc_action_default }, { "rsc-action-item", "default", rsc_action_item }, { "rsc-action-item", "xml", rsc_action_item_xml }, { "rsc-is-colocated-with-list", "default", rsc_is_colocated_with_list }, { "rsc-is-colocated-with-list", "xml", rsc_is_colocated_with_list_xml }, { "rscs-colocated-with-list", "default", rscs_colocated_with_list }, { "rscs-colocated-with-list", "xml", rscs_colocated_with_list_xml }, { "rule-check", "default", rule_check_default }, { "rule-check", "xml", rule_check_xml }, { "locations-and-colocations", "default", locations_and_colocations }, { "locations-and-colocations", "xml", locations_and_colocations_xml }, { NULL, NULL, NULL } }; void pcmk__register_lib_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } diff --git a/lib/pacemaker/pcmk_rule.c b/lib/pacemaker/pcmk_rule.c index b8ca453a52..3ad8a62be4 100644 --- a/lib/pacemaker/pcmk_rule.c +++ b/lib/pacemaker/pcmk_rule.c @@ -1,295 +1,296 @@ /* * Copyright 2022-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include +#include #include #include /*! * \internal * \brief Evaluate a date expression for a specific time * * \param[in] expr date_expression XML * \param[in] now Time for which to evaluate expression * * \return Standard Pacemaker return code */ static int eval_date_expression(const xmlNode *expr, crm_time_t *now) { pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; return pe__eval_date_expr(expr, &rule_data, NULL); } /*! * \internal * \brief Initialize the cluster working set for checking rules * * Make our own copies of the CIB XML and date/time object, if they're not * \c NULL. This way we don't have to take ownership of the objects passed via * the API. * * \param[in,out] out Output object * \param[in] input The CIB XML to check (if \c NULL, use current CIB) * \param[in] date Check whether the rule is in effect at this date * and time (if \c NULL, use current date and time) * \param[out] data_set Where to store the cluster working set * * \return Standard Pacemaker return code */ static int init_rule_check(pcmk__output_t *out, xmlNodePtr input, const crm_time_t *date, pe_working_set_t **data_set) { // Allows for cleaner syntax than dereferencing the data_set argument pe_working_set_t *new_data_set = NULL; new_data_set = pe_new_working_set(); if (new_data_set == NULL) { return ENOMEM; } pe__set_working_set_flags(new_data_set, pe_flag_no_counts|pe_flag_no_compat); // Populate the working set instance // Make our own copy of the given input or fetch the CIB and use that if (input != NULL) { new_data_set->input = copy_xml(input); if (new_data_set->input == NULL) { out->err(out, "Failed to copy input XML"); pe_free_working_set(new_data_set); return ENOMEM; } } else { int rc = cib__signon_query(out, NULL, &(new_data_set->input)); if (rc != pcmk_rc_ok) { pe_free_working_set(new_data_set); return rc; } } // Make our own copy of the given crm_time_t object; otherwise // cluster_status() populates with the current time if (date != NULL) { // pcmk_copy_time() guarantees non-NULL new_data_set->now = pcmk_copy_time(date); } // Unpack everything cluster_status(new_data_set); *data_set = new_data_set; return pcmk_rc_ok; } #define XPATH_NODE_RULE "//" XML_TAG_RULE "[@" XML_ATTR_ID "='%s']" /*! * \internal * \brief Check whether a given rule is in effect * * \param[in] data_set Cluster working set * \param[in] rule_id The ID of the rule to check * \param[out] error Where to store a rule evaluation error message * * \return Standard Pacemaker return code */ static int eval_rule(pe_working_set_t *data_set, const char *rule_id, const char **error) { xmlNodePtr cib_constraints = NULL; xmlNodePtr match = NULL; xmlXPathObjectPtr xpath_obj = NULL; char *xpath = NULL; int rc = pcmk_rc_ok; int num_results = 0; *error = NULL; /* Rules are under the constraints node in the XML, so first find that. */ cib_constraints = pcmk_find_cib_element(data_set->input, XML_CIB_TAG_CONSTRAINTS); /* Get all rules matching the given ID that are also simple enough for us * to check. For the moment, these rules must only have a single * date_expression child and: * - Do not have a date_spec operation, or * - Have a date_spec operation that contains years= but does not contain * moon=. * * We do this in steps to provide better error messages. First, check that * there's any rule with the given ID. */ xpath = crm_strdup_printf(XPATH_NODE_RULE, rule_id); xpath_obj = xpath_search(cib_constraints, xpath); num_results = numXpathResults(xpath_obj); free(xpath); freeXpathObject(xpath_obj); if (num_results == 0) { *error = "Rule not found"; return ENXIO; } if (num_results > 1) { // Should not be possible; schema prevents this *error = "Found more than one rule with matching ID"; return pcmk_rc_duplicate_id; } /* Next, make sure it has exactly one date_expression. */ xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression", rule_id); xpath_obj = xpath_search(cib_constraints, xpath); num_results = numXpathResults(xpath_obj); free(xpath); freeXpathObject(xpath_obj); if (num_results != 1) { if (num_results == 0) { *error = "Rule does not have a date expression"; } else { *error = "Rule has more than one date expression"; } return EOPNOTSUPP; } /* Then, check that it's something we actually support. */ xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression[" "@" XML_EXPR_ATTR_OPERATION "!='date_spec']", rule_id); xpath_obj = xpath_search(cib_constraints, xpath); num_results = numXpathResults(xpath_obj); free(xpath); if (num_results == 0) { freeXpathObject(xpath_obj); xpath = crm_strdup_printf(XPATH_NODE_RULE "//date_expression[" "@" XML_EXPR_ATTR_OPERATION "='date_spec' " "and date_spec/@years " "and not(date_spec/@moon)]", rule_id); xpath_obj = xpath_search(cib_constraints, xpath); num_results = numXpathResults(xpath_obj); free(xpath); if (num_results == 0) { freeXpathObject(xpath_obj); *error = "Rule must either not use date_spec, or use date_spec " "with years= but not moon="; return EOPNOTSUPP; } } match = getXpathResult(xpath_obj, 0); /* We should have ensured this with the xpath query above, but double- * checking can't hurt. */ CRM_ASSERT(match != NULL); CRM_ASSERT(find_expression_type(match) == time_expr); rc = eval_date_expression(match, data_set->now); if (rc == pcmk_rc_undetermined) { /* pe__eval_date_expr() should return this only if something is * malformed or missing */ *error = "Error parsing rule"; } freeXpathObject(xpath_obj); return rc; } /*! * \internal * \brief Check whether each rule in a list is in effect * * \param[in,out] out Output object * \param[in] input The CIB XML to check (if \c NULL, use current CIB) * \param[in] date Check whether the rule is in effect at this date and * time (if \c NULL, use current date and time) * \param[in] rule_ids The IDs of the rules to check, as a NULL- * terminated list. * * \return Standard Pacemaker return code */ int pcmk__check_rules(pcmk__output_t *out, xmlNodePtr input, const crm_time_t *date, const char **rule_ids) { pe_working_set_t *data_set = NULL; int rc = pcmk_rc_ok; CRM_ASSERT(out != NULL); if (rule_ids == NULL) { // Trivial case; every rule specified is in effect return pcmk_rc_ok; } rc = init_rule_check(out, input, date, &data_set); if (rc != pcmk_rc_ok) { return rc; } for (const char **rule_id = rule_ids; *rule_id != NULL; rule_id++) { const char *error = NULL; int last_rc = eval_rule(data_set, *rule_id, &error); out->message(out, "rule-check", *rule_id, last_rc, error); if (last_rc != pcmk_rc_ok) { rc = last_rc; } } pe_free_working_set(data_set); return rc; } // Documented in pacemaker.h int pcmk_check_rules(xmlNodePtr *xml, xmlNodePtr input, const crm_time_t *date, const char **rule_ids) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__check_rules(out, input, date, rule_ids); pcmk__xml_output_finish(out, xml); return rc; } diff --git a/lib/pacemaker/pcmk_sched_actions.c b/lib/pacemaker/pcmk_sched_actions.c index 13d0b4b4e6..2fa66ce383 100644 --- a/lib/pacemaker/pcmk_sched_actions.c +++ b/lib/pacemaker/pcmk_sched_actions.c @@ -1,1918 +1,1917 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Get the action flags relevant to ordering constraints * * \param[in,out] action Action to check * \param[in] node Node that *other* action in the ordering is on * (used only for clone resource actions) * * \return Action flags that should be used for orderings */ static uint32_t action_flags_for_ordering(pe_action_t *action, const pe_node_t *node) { bool runnable = false; uint32_t flags; // For non-resource actions, return the action flags if (action->rsc == NULL) { return action->flags; } /* For non-clone resources, or a clone action not assigned to a node, * return the flags as determined by the resource method without a node * specified. */ flags = action->rsc->cmds->action_flags(action, NULL); if ((node == NULL) || !pe_rsc_is_clone(action->rsc)) { return flags; } /* Otherwise (i.e., for clone resource actions on a specific node), first * remember whether the non-node-specific action is runnable. */ runnable = pcmk_is_set(flags, pe_action_runnable); // Then recheck the resource method with the node flags = action->rsc->cmds->action_flags(action, node); /* For clones in ordering constraints, the node-specific "runnable" doesn't * matter, just the non-node-specific setting (i.e., is the action runnable * anywhere). * * This applies only to runnable, and only for ordering constraints. This * function shouldn't be used for other types of constraints without * changes. Not very satisfying, but it's logical and appears to work well. */ if (runnable && !pcmk_is_set(flags, pe_action_runnable)) { pe__set_raw_action_flags(flags, action->rsc->id, pe_action_runnable); } return flags; } /*! * \internal * \brief Get action UUID that should be used with a resource ordering * * When an action is ordered relative to an action for a collective resource * (clone, group, or bundle), it actually needs to be ordered after all * instances of the collective have completed the relevant action (for example, * given "start CLONE then start RSC", RSC must wait until all instances of * CLONE have started). Given the UUID and resource of the first action in an * ordering, this returns the UUID of the action that should actually be used * for ordering (for example, "CLONE_started_0" instead of "CLONE_start_0"). * * \param[in] first_uuid UUID of first action in ordering * \param[in] first_rsc Resource of first action in ordering * * \return Newly allocated copy of UUID to use with ordering * \note It is the caller's responsibility to free the return value. */ static char * action_uuid_for_ordering(const char *first_uuid, const pe_resource_t *first_rsc) { guint interval_ms = 0; char *uuid = NULL; char *rid = NULL; char *first_task_str = NULL; enum action_tasks first_task = no_action; enum action_tasks remapped_task = no_action; // Only non-notify actions for collective resources need remapping if ((strstr(first_uuid, "notify") != NULL) || (first_rsc->variant < pe_group)) { goto done; } // Only non-recurring actions need remapping CRM_ASSERT(parse_op_key(first_uuid, &rid, &first_task_str, &interval_ms)); if (interval_ms > 0) { goto done; } first_task = text2task(first_task_str); switch (first_task) { case stop_rsc: case start_rsc: case action_notify: case action_promote: case action_demote: remapped_task = first_task + 1; break; case stopped_rsc: case started_rsc: case action_notified: case action_promoted: case action_demoted: remapped_task = first_task; break; case monitor_rsc: case shutdown_crm: case stonith_node: break; default: crm_err("Unknown action '%s' in ordering", first_task_str); break; } if (remapped_task != no_action) { /* If a (clone) resource has notifications enabled, we want to order * relative to when all notifications have been sent for the remapped * task. Only outermost resources or those in bundles have * notifications. */ if (pcmk_is_set(first_rsc->flags, pe_rsc_notify) && ((first_rsc->parent == NULL) || (pe_rsc_is_clone(first_rsc) && (first_rsc->parent->variant == pe_container)))) { uuid = pcmk__notify_key(rid, "confirmed-post", task2text(remapped_task)); } else { uuid = pcmk__op_key(rid, task2text(remapped_task), 0); } pe_rsc_trace(first_rsc, "Remapped action UUID %s to %s for ordering purposes", first_uuid, uuid); } done: if (uuid == NULL) { uuid = strdup(first_uuid); CRM_ASSERT(uuid != NULL); } free(first_task_str); free(rid); return uuid; } /*! * \internal * \brief Get actual action that should be used with an ordering * * When an action is ordered relative to an action for a collective resource * (clone, group, or bundle), it actually needs to be ordered after all * instances of the collective have completed the relevant action (for example, * given "start CLONE then start RSC", RSC must wait until all instances of * CLONE have started). Given the first action in an ordering, this returns the * the action that should actually be used for ordering (for example, the * started action instead of the start action). * * \param[in] action First action in an ordering * * \return Actual action that should be used for the ordering */ static pe_action_t * action_for_ordering(pe_action_t *action) { pe_action_t *result = action; pe_resource_t *rsc = action->rsc; if ((rsc != NULL) && (rsc->variant >= pe_group) && (action->uuid != NULL)) { char *uuid = action_uuid_for_ordering(action->uuid, rsc); result = find_first_action(rsc->actions, uuid, NULL, NULL); if (result == NULL) { crm_warn("Not remapping %s to %s because %s does not have " "remapped action", action->uuid, uuid, rsc->id); result = action; } free(uuid); } return result; } /*! * \internal * \brief Update flags for ordering's actions appropriately for ordering's flags * * \param[in,out] first First action in an ordering * \param[in,out] then Then action in an ordering * \param[in] first_flags Action flags for \p first for ordering purposes * \param[in] then_flags Action flags for \p then for ordering purposes * \param[in,out] order Action wrapper for \p first in ordering * \param[in,out] data_set Cluster working set * * \return Group of enum pcmk__updated flags */ static uint32_t update_action_for_ordering_flags(pe_action_t *first, pe_action_t *then, uint32_t first_flags, uint32_t then_flags, pe_action_wrapper_t *order, pe_working_set_t *data_set) { uint32_t changed = pcmk__updated_none; /* The node will only be used for clones. If interleaved, node will be NULL, * otherwise the ordering scope will be limited to the node. Normally, the * whole 'then' clone should restart if 'first' is restarted, so then->node * is needed. */ pe_node_t *node = then->node; if (pcmk_is_set(order->type, pe_order_implies_then_on_node)) { /* For unfencing, only instances of 'then' on the same node as 'first' * (the unfencing operation) should restart, so reset node to * first->node, at which point this case is handled like a normal * pe_order_implies_then. */ pe__clear_order_flags(order->type, pe_order_implies_then_on_node); pe__set_order_flags(order->type, pe_order_implies_then); node = first->node; pe_rsc_trace(then->rsc, "%s then %s: mapped pe_order_implies_then_on_node to " "pe_order_implies_then on %s", first->uuid, then->uuid, pe__node_name(node)); } if (pcmk_is_set(order->type, pe_order_implies_then)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags & pe_action_optional, pe_action_optional, pe_order_implies_then, data_set); } else if (!pcmk_is_set(first_flags, pe_action_optional) && pcmk_is_set(then->flags, pe_action_optional)) { pe__clear_action_flags(then, pe_action_optional); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_implies_then", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_restart) && (then->rsc != NULL)) { enum pe_action_flags restart = pe_action_optional|pe_action_runnable; changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, restart, pe_order_restart, data_set); pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_restart", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_implies_first)) { if (first->rsc != NULL) { changed |= first->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_optional, pe_order_implies_first, data_set); } else if (!pcmk_is_set(first_flags, pe_action_optional) && pcmk_is_set(first->flags, pe_action_runnable)) { pe__clear_action_flags(first, pe_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_first); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_implies_first", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_promoted_implies_first)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags & pe_action_optional, pe_action_optional, pe_order_promoted_implies_first, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_promoted_implies_first", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_one_or_more)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_runnable, pe_order_one_or_more, data_set); } else if (pcmk_is_set(first_flags, pe_action_runnable)) { // We have another runnable instance of "first" then->runnable_before++; /* Mark "then" as runnable if it requires a certain number of * "before" instances to be runnable, and they now are. */ if ((then->runnable_before >= then->required_runnable_before) && !pcmk_is_set(then->flags, pe_action_runnable)) { pe__set_action_flags(then, pe_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_one_or_more", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_probe) && (then->rsc != NULL)) { if (!pcmk_is_set(first_flags, pe_action_runnable) && (first->rsc->running_on != NULL)) { pe_rsc_trace(then->rsc, "%s then %s: ignoring because first is stopping", first->uuid, then->uuid); order->type = pe_order_none; } else { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_runnable, pe_order_runnable_left, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_probe", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_runnable_left)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_runnable, pe_order_runnable_left, data_set); } else if (!pcmk_is_set(first_flags, pe_action_runnable) && pcmk_is_set(then->flags, pe_action_runnable)) { pe__clear_action_flags(then, pe_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_runnable_left", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_implies_first_migratable)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_optional, pe_order_implies_first_migratable, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after " "pe_order_implies_first_migratable", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_pseudo_left)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_optional, pe_order_pseudo_left, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_pseudo_left", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_optional)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_runnable, pe_order_optional, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_optional", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pe_order_asymmetrical)) { if (then->rsc != NULL) { changed |= then->rsc->cmds->update_ordered_actions(first, then, node, first_flags, pe_action_runnable, pe_order_asymmetrical, data_set); } pe_rsc_trace(then->rsc, "%s then %s: %s after pe_order_asymmetrical", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(first->flags, pe_action_runnable) && pcmk_is_set(order->type, pe_order_implies_then_printed) && !pcmk_is_set(first_flags, pe_action_optional)) { pe_rsc_trace(then->rsc, "%s will be in graph because %s is required", then->uuid, first->uuid); pe__set_action_flags(then, pe_action_print_always); // Don't bother marking 'then' as changed just for this } if (pcmk_is_set(order->type, pe_order_implies_first_printed) && !pcmk_is_set(then_flags, pe_action_optional)) { pe_rsc_trace(then->rsc, "%s will be in graph because %s is required", first->uuid, then->uuid); pe__set_action_flags(first, pe_action_print_always); // Don't bother marking 'first' as changed just for this } if (pcmk_any_flags_set(order->type, pe_order_implies_then |pe_order_implies_first |pe_order_restart) && (first->rsc != NULL) && !pcmk_is_set(first->rsc->flags, pe_rsc_managed) && pcmk_is_set(first->rsc->flags, pe_rsc_block) && !pcmk_is_set(first->flags, pe_action_runnable) && pcmk__str_eq(first->task, RSC_STOP, pcmk__str_casei)) { if (pcmk_is_set(then->flags, pe_action_runnable)) { pe__clear_action_flags(then, pe_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pe_rsc_trace(then->rsc, "%s then %s: %s after checking whether first " "is blocked, unmanaged, unrunnable stop", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } return changed; } // Convenience macros for logging action properties #define action_type_str(flags) \ (pcmk_is_set((flags), pe_action_pseudo)? "pseudo-action" : "action") #define action_optional_str(flags) \ (pcmk_is_set((flags), pe_action_optional)? "optional" : "required") #define action_runnable_str(flags) \ (pcmk_is_set((flags), pe_action_runnable)? "runnable" : "unrunnable") #define action_node_str(a) \ (((a)->node == NULL)? "no node" : (a)->node->details->uname) /*! * \internal * \brief Update an action's flags for all orderings where it is "then" * * \param[in,out] then Action to update * \param[in,out] data_set Cluster working set */ void pcmk__update_action_for_orderings(pe_action_t *then, pe_working_set_t *data_set) { GList *lpc = NULL; uint32_t changed = pcmk__updated_none; int last_flags = then->flags; pe_rsc_trace(then->rsc, "Updating %s %s (%s %s) on %s", action_type_str(then->flags), then->uuid, action_optional_str(then->flags), action_runnable_str(then->flags), action_node_str(then)); if (pcmk_is_set(then->flags, pe_action_requires_any)) { /* Initialize current known "runnable before" actions. As * update_action_for_ordering_flags() is called for each of then's * before actions, this number will increment as runnable 'first' * actions are encountered. */ then->runnable_before = 0; if (then->required_runnable_before == 0) { /* @COMPAT This ordering constraint uses the deprecated * "require-all=false" attribute. Treat it like "clone-min=1". */ then->required_runnable_before = 1; } /* The pe_order_one_or_more clause of update_action_for_ordering_flags() * (called below) will reset runnable if appropriate. */ pe__clear_action_flags(then, pe_action_runnable); } for (lpc = then->actions_before; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc->data; pe_action_t *first = other->action; pe_node_t *then_node = then->node; pe_node_t *first_node = first->node; if ((first->rsc != NULL) && (first->rsc->variant == pe_group) && pcmk__str_eq(first->task, RSC_START, pcmk__str_casei)) { first_node = first->rsc->fns->location(first->rsc, NULL, FALSE); if (first_node != NULL) { pe_rsc_trace(first->rsc, "Found %s for 'first' %s", pe__node_name(first_node), first->uuid); } } if ((then->rsc != NULL) && (then->rsc->variant == pe_group) && pcmk__str_eq(then->task, RSC_START, pcmk__str_casei)) { then_node = then->rsc->fns->location(then->rsc, NULL, FALSE); if (then_node != NULL) { pe_rsc_trace(then->rsc, "Found %s for 'then' %s", pe__node_name(then_node), then->uuid); } } // Disable constraint if it only applies when on same node, but isn't if (pcmk_is_set(other->type, pe_order_same_node) && (first_node != NULL) && (then_node != NULL) - && (first_node->details != then_node->details)) { + && !pe__same_node(first_node, then_node)) { pe_rsc_trace(then->rsc, "Disabled ordering %s on %s then %s on %s: not same node", other->action->uuid, pe__node_name(first_node), then->uuid, pe__node_name(then_node)); other->type = pe_order_none; continue; } pcmk__clear_updated_flags(changed, then, pcmk__updated_first); if ((first->rsc != NULL) && pcmk_is_set(other->type, pe_order_then_cancels_first) && !pcmk_is_set(then->flags, pe_action_optional)) { /* 'then' is required, so we must abandon 'first' * (e.g. a required stop cancels any agent reload). */ pe__set_action_flags(other->action, pe_action_optional); if (!strcmp(first->task, CRMD_ACTION_RELOAD_AGENT)) { pe__clear_resource_flags(first->rsc, pe_rsc_reload); } } if ((first->rsc != NULL) && (then->rsc != NULL) && (first->rsc != then->rsc) && !is_parent(then->rsc, first->rsc)) { first = action_for_ordering(first); } if (first != other->action) { pe_rsc_trace(then->rsc, "Ordering %s after %s instead of %s", then->uuid, first->uuid, other->action->uuid); } pe_rsc_trace(then->rsc, "%s (%#.6x) then %s (%#.6x): type=%#.6x node=%s", first->uuid, first->flags, then->uuid, then->flags, other->type, action_node_str(first)); if (first == other->action) { /* 'first' was not remapped (e.g. from 'start' to 'running'), which * could mean it is a non-resource action, a primitive resource * action, or already expanded. */ uint32_t first_flags, then_flags; first_flags = action_flags_for_ordering(first, then_node); then_flags = action_flags_for_ordering(then, first_node); changed |= update_action_for_ordering_flags(first, then, first_flags, then_flags, other, data_set); /* 'first' was for a complex resource (clone, group, etc), * create a new dependency if necessary */ } else if (order_actions(first, then, other->type)) { /* This was the first time 'first' and 'then' were associated, * start again to get the new actions_before list */ pcmk__set_updated_flags(changed, then, pcmk__updated_then); pe_rsc_trace(then->rsc, "Disabled ordering %s then %s in favor of %s then %s", other->action->uuid, then->uuid, first->uuid, then->uuid); other->type = pe_order_none; } if (pcmk_is_set(changed, pcmk__updated_first)) { crm_trace("Re-processing %s and its 'after' actions " "because it changed", first->uuid); for (GList *lpc2 = first->actions_after; lpc2 != NULL; lpc2 = lpc2->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc2->data; pcmk__update_action_for_orderings(other->action, data_set); } pcmk__update_action_for_orderings(first, data_set); } } if (pcmk_is_set(then->flags, pe_action_requires_any)) { if (last_flags == then->flags) { pcmk__clear_updated_flags(changed, then, pcmk__updated_then); } else { pcmk__set_updated_flags(changed, then, pcmk__updated_then); } } if (pcmk_is_set(changed, pcmk__updated_then)) { crm_trace("Re-processing %s and its 'after' actions because it changed", then->uuid); if (pcmk_is_set(last_flags, pe_action_runnable) && !pcmk_is_set(then->flags, pe_action_runnable)) { pcmk__block_colocation_dependents(then, data_set); } pcmk__update_action_for_orderings(then, data_set); for (lpc = then->actions_after; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) lpc->data; pcmk__update_action_for_orderings(other->action, data_set); } } } static inline bool is_primitive_action(const pe_action_t *action) { return action && action->rsc && (action->rsc->variant == pe_native); } /*! * \internal * \brief Clear a single action flag and set reason text * * \param[in,out] action Action whose flag should be cleared * \param[in] flag Action flag that should be cleared * \param[in] reason Action that is the reason why flag is being cleared */ #define clear_action_flag_because(action, flag, reason) do { \ if (pcmk_is_set((action)->flags, (flag))) { \ pe__clear_action_flags(action, flag); \ if ((action)->rsc != (reason)->rsc) { \ char *reason_text = pe__action2reason((reason), (flag)); \ pe_action_set_reason((action), reason_text, \ ((flag) == pe_action_migrate_runnable)); \ free(reason_text); \ } \ } \ } while (0) /*! * \internal * \brief Update actions in an asymmetric ordering * * If the "first" action in an asymmetric ordering is unrunnable, make the * "second" action unrunnable as well, if appropriate. * * \param[in] first 'First' action in an asymmetric ordering * \param[in,out] then 'Then' action in an asymmetric ordering */ static void handle_asymmetric_ordering(const pe_action_t *first, pe_action_t *then) { /* Only resource actions after an unrunnable 'first' action need updates for * asymmetric ordering. */ if ((then->rsc == NULL) || pcmk_is_set(first->flags, pe_action_runnable)) { return; } // Certain optional 'then' actions are unaffected by unrunnable 'first' if (pcmk_is_set(then->flags, pe_action_optional)) { enum rsc_role_e then_rsc_role = then->rsc->fns->state(then->rsc, TRUE); if ((then_rsc_role == RSC_ROLE_STOPPED) && pcmk__str_eq(then->task, RSC_STOP, pcmk__str_none)) { /* If 'then' should stop after 'first' but is already stopped, the * ordering is irrelevant. */ return; } else if ((then_rsc_role >= RSC_ROLE_STARTED) && pcmk__str_eq(then->task, RSC_START, pcmk__str_none) && pe__rsc_running_on_only(then->rsc, then->node)) { /* Similarly if 'then' should start after 'first' but is already * started on a single node. */ return; } } // 'First' can't run, so 'then' can't either clear_action_flag_because(then, pe_action_optional, first); clear_action_flag_because(then, pe_action_runnable, first); } /*! * \internal * \brief Set action bits appropriately when pe_restart_order is used * * \param[in,out] first 'First' action in an ordering with pe_restart_order * \param[in,out] then 'Then' action in an ordering with pe_restart_order * \param[in] filter What action flags to care about * * \note pe_restart_order is set for "stop resource before starting it" and * "stop later group member before stopping earlier group member" */ static void handle_restart_ordering(pe_action_t *first, pe_action_t *then, uint32_t filter) { const char *reason = NULL; CRM_ASSERT(is_primitive_action(first)); CRM_ASSERT(is_primitive_action(then)); // We need to update the action in two cases: // ... if 'then' is required if (pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(then->flags, pe_action_optional)) { reason = "restart"; } /* ... if 'then' is unrunnable action on same resource (if a resource * should restart but can't start, we still want to stop) */ if (pcmk_is_set(filter, pe_action_runnable) && !pcmk_is_set(then->flags, pe_action_runnable) && pcmk_is_set(then->rsc->flags, pe_rsc_managed) && (first->rsc == then->rsc)) { reason = "stop"; } if (reason == NULL) { return; } pe_rsc_trace(first->rsc, "Handling %s -> %s for %s", first->uuid, then->uuid, reason); // Make 'first' required if it is runnable if (pcmk_is_set(first->flags, pe_action_runnable)) { clear_action_flag_because(first, pe_action_optional, then); } // Make 'first' required if 'then' is required if (!pcmk_is_set(then->flags, pe_action_optional)) { clear_action_flag_because(first, pe_action_optional, then); } // Make 'first' unmigratable if 'then' is unmigratable if (!pcmk_is_set(then->flags, pe_action_migrate_runnable)) { clear_action_flag_because(first, pe_action_migrate_runnable, then); } // Make 'then' unrunnable if 'first' is required but unrunnable if (!pcmk_is_set(first->flags, pe_action_optional) && !pcmk_is_set(first->flags, pe_action_runnable)) { clear_action_flag_because(then, pe_action_runnable, first); } } /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two actions, update the actions' flags * (and runnable_before members if appropriate) as appropriate for the ordering. * Effects may cascade to other orderings involving the actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * (ignored) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pe_action_optional to affect only mandatory * actions, and pe_action_runnable to affect only * runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * \param[in,out] data_set Cluster working set * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t pcmk__update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set) { uint32_t changed = pcmk__updated_none; uint32_t then_flags = then->flags; uint32_t first_flags = first->flags; if (pcmk_is_set(type, pe_order_asymmetrical)) { handle_asymmetric_ordering(first, then); } if (pcmk_is_set(type, pe_order_implies_first) && !pcmk_is_set(then_flags, pe_action_optional)) { // Then is required, and implies first should be, too if (pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(flags, pe_action_optional) && pcmk_is_set(first_flags, pe_action_optional)) { clear_action_flag_because(first, pe_action_optional, then); } if (pcmk_is_set(flags, pe_action_migrate_runnable) && !pcmk_is_set(then->flags, pe_action_migrate_runnable)) { clear_action_flag_because(first, pe_action_migrate_runnable, then); } } if (pcmk_is_set(type, pe_order_promoted_implies_first) && (then->rsc != NULL) && (then->rsc->role == RSC_ROLE_PROMOTED) && pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(then->flags, pe_action_optional)) { clear_action_flag_because(first, pe_action_optional, then); if (pcmk_is_set(first->flags, pe_action_migrate_runnable) && !pcmk_is_set(then->flags, pe_action_migrate_runnable)) { clear_action_flag_because(first, pe_action_migrate_runnable, then); } } if (pcmk_is_set(type, pe_order_implies_first_migratable) && pcmk_is_set(filter, pe_action_optional)) { if (!pcmk_all_flags_set(then->flags, pe_action_migrate_runnable|pe_action_runnable)) { clear_action_flag_because(first, pe_action_runnable, then); } if (!pcmk_is_set(then->flags, pe_action_optional)) { clear_action_flag_because(first, pe_action_optional, then); } } if (pcmk_is_set(type, pe_order_pseudo_left) && pcmk_is_set(filter, pe_action_optional) && !pcmk_is_set(first->flags, pe_action_runnable)) { clear_action_flag_because(then, pe_action_migrate_runnable, first); pe__clear_action_flags(then, pe_action_pseudo); } if (pcmk_is_set(type, pe_order_runnable_left) && pcmk_is_set(filter, pe_action_runnable) && pcmk_is_set(then->flags, pe_action_runnable) && !pcmk_is_set(flags, pe_action_runnable)) { clear_action_flag_because(then, pe_action_runnable, first); clear_action_flag_because(then, pe_action_migrate_runnable, first); } if (pcmk_is_set(type, pe_order_implies_then) && pcmk_is_set(filter, pe_action_optional) && pcmk_is_set(then->flags, pe_action_optional) && !pcmk_is_set(flags, pe_action_optional) && !pcmk_is_set(first->flags, pe_action_migrate_runnable)) { clear_action_flag_because(then, pe_action_optional, first); } if (pcmk_is_set(type, pe_order_restart)) { handle_restart_ordering(first, then, filter); } if (then_flags != then->flags) { pcmk__set_updated_flags(changed, first, pcmk__updated_then); pe_rsc_trace(then->rsc, "%s on %s: flags are now %#.6x (was %#.6x) " "because of 'first' %s (%#.6x)", then->uuid, pe__node_name(then->node), then->flags, then_flags, first->uuid, first->flags); if ((then->rsc != NULL) && (then->rsc->parent != NULL)) { // Required to handle "X_stop then X_start" for cloned groups pcmk__update_action_for_orderings(then, data_set); } } if (first_flags != first->flags) { pcmk__set_updated_flags(changed, first, pcmk__updated_first); pe_rsc_trace(first->rsc, "%s on %s: flags are now %#.6x (was %#.6x) " "because of 'then' %s (%#.6x)", first->uuid, pe__node_name(first->node), first->flags, first_flags, then->uuid, then->flags); } return changed; } /*! * \internal * \brief Trace-log an action (optionally with its dependent actions) * * \param[in] pre_text If not NULL, prefix the log with this plus ": " * \param[in] action Action to log * \param[in] details If true, recursively log dependent actions */ void pcmk__log_action(const char *pre_text, const pe_action_t *action, bool details) { const char *node_uname = NULL; const char *node_uuid = NULL; const char *desc = NULL; CRM_CHECK(action != NULL, return); if (!pcmk_is_set(action->flags, pe_action_pseudo)) { if (action->node != NULL) { node_uname = action->node->details->uname; node_uuid = action->node->details->id; } else { node_uname = ""; } } switch (text2task(action->task)) { case stonith_node: case shutdown_crm: if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (pcmk_is_set(action->flags, pe_action_optional)) { desc = "Optional "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; default: if (pcmk_is_set(action->flags, pe_action_optional)) { desc = "Optional "; } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (action->rsc? action->rsc->id : ""), (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; } if (details) { const GList *iter = NULL; const pe_action_wrapper_t *other = NULL; crm_trace("\t\t====== Preceding Actions"); for (iter = action->actions_before; iter != NULL; iter = iter->next) { other = (const pe_action_wrapper_t *) iter->data; pcmk__log_action("\t\t", other->action, false); } crm_trace("\t\t====== Subsequent Actions"); for (iter = action->actions_after; iter != NULL; iter = iter->next) { other = (const pe_action_wrapper_t *) iter->data; pcmk__log_action("\t\t", other->action, false); } crm_trace("\t\t====== End"); } else { crm_trace("\t\t(before=%d, after=%d)", g_list_length(action->actions_before), g_list_length(action->actions_after)); } } /*! * \internal * \brief Create a new shutdown action for a node * * \param[in,out] node Node being shut down * * \return Newly created shutdown action for \p node */ pe_action_t * pcmk__new_shutdown_action(pe_node_t *node) { char *shutdown_id = NULL; pe_action_t *shutdown_op = NULL; CRM_ASSERT(node != NULL); shutdown_id = crm_strdup_printf("%s-%s", CRM_OP_SHUTDOWN, node->details->uname); shutdown_op = custom_action(NULL, shutdown_id, CRM_OP_SHUTDOWN, node, FALSE, TRUE, node->details->data_set); pcmk__order_stops_before_shutdown(node, shutdown_op); add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); return shutdown_op; } /*! * \internal * \brief Calculate and add an operation digest to XML * * Calculate an operation digest, which enables us to later determine when a * restart is needed due to the resource's parameters being changed, and add it * to given XML. * * \param[in] op Operation result from executor * \param[in,out] update XML to add digest to */ static void add_op_digest_to_xml(const lrmd_event_data_t *op, xmlNode *update) { char *digest = NULL; xmlNode *args_xml = NULL; if (op->params == NULL) { return; } args_xml = create_xml_node(NULL, XML_TAG_PARAMS); g_hash_table_foreach(op->params, hash2field, args_xml); pcmk__filter_op_for_digest(args_xml); digest = calculate_operation_digest(args_xml, NULL); crm_xml_add(update, XML_LRM_ATTR_OP_DIGEST, digest); free_xml(args_xml); free(digest); } #define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx" /*! * \internal * \brief Create XML for resource operation history update * * \param[in,out] parent Parent XML node to add to * \param[in,out] op Operation event data * \param[in] caller_version DC feature set * \param[in] target_rc Expected result of operation * \param[in] node Name of node on which operation was performed * \param[in] origin Arbitrary description of update source * * \return Newly created XML node for history update */ xmlNode * pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *op, const char *caller_version, int target_rc, const char *node, const char *origin) { char *key = NULL; char *magic = NULL; char *op_id = NULL; char *op_id_additional = NULL; char *local_user_data = NULL; const char *exit_reason = NULL; xmlNode *xml_op = NULL; const char *task = NULL; CRM_CHECK(op != NULL, return NULL); crm_trace("Creating history XML for %s-interval %s action for %s on %s " "(DC version: %s, origin: %s)", pcmk__readable_interval(op->interval_ms), op->op_type, op->rsc_id, ((node == NULL)? "no node" : node), caller_version, origin); task = op->op_type; /* Record a successful agent reload as a start, and a failed one as a * monitor, to make life easier for the scheduler when determining the * current state. * * @COMPAT We should check "reload" here only if the operation was for a * pre-OCF-1.1 resource agent, but we don't know that here, and we should * only ever get results for actions scheduled by us, so we can reasonably * assume any "reload" is actually a pre-1.1 agent reload. */ if (pcmk__str_any_of(task, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { if (op->op_status == PCMK_EXEC_DONE) { task = CRMD_ACTION_START; } else { task = CRMD_ACTION_STATUS; } } key = pcmk__op_key(op->rsc_id, task, op->interval_ms); if (pcmk__str_eq(task, CRMD_ACTION_NOTIFY, pcmk__str_none)) { const char *n_type = crm_meta_value(op->params, "notify_type"); const char *n_task = crm_meta_value(op->params, "notify_operation"); CRM_LOG_ASSERT(n_type != NULL); CRM_LOG_ASSERT(n_task != NULL); op_id = pcmk__notify_key(op->rsc_id, n_type, n_task); if (op->op_status != PCMK_EXEC_PENDING) { /* Ignore notify errors. * * @TODO It might be better to keep the correct result here, and * ignore it in process_graph_event(). */ lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } /* Migration history is preserved separately, which usually matters for * multiple nodes and is important for future cluster transitions. */ } else if (pcmk__str_any_of(op->op_type, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { op_id = strdup(key); } else if (did_rsc_op_fail(op, target_rc)) { op_id = pcmk__op_key(op->rsc_id, "last_failure", 0); if (op->interval_ms == 0) { // Ensure 'last' gets updated, in case record-pending is true op_id_additional = pcmk__op_key(op->rsc_id, "last", 0); } exit_reason = op->exit_reason; } else if (op->interval_ms > 0) { op_id = strdup(key); } else { op_id = pcmk__op_key(op->rsc_id, "last", 0); } again: xml_op = pcmk__xe_match(parent, XML_LRM_TAG_RSC_OP, XML_ATTR_ID, op_id); if (xml_op == NULL) { xml_op = create_xml_node(parent, XML_LRM_TAG_RSC_OP); } if (op->user_data == NULL) { crm_debug("Generating fake transition key for: " PCMK__OP_FMT " %d from %s", op->rsc_id, op->op_type, op->interval_ms, op->call_id, origin); local_user_data = pcmk__transition_key(-1, op->call_id, target_rc, FAKE_TE_ID); op->user_data = local_user_data; } if (magic == NULL) { magic = crm_strdup_printf("%d:%d;%s", op->op_status, op->rc, (const char *) op->user_data); } crm_xml_add(xml_op, XML_ATTR_ID, op_id); crm_xml_add(xml_op, XML_LRM_ATTR_TASK_KEY, key); crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task); crm_xml_add(xml_op, XML_ATTR_ORIGIN, origin); crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, caller_version); crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, op->user_data); crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, magic); crm_xml_add(xml_op, XML_LRM_ATTR_EXIT_REASON, exit_reason == NULL ? "" : exit_reason); crm_xml_add(xml_op, XML_LRM_ATTR_TARGET, node); /* For context during triage */ crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op->rc); crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, op->op_status); crm_xml_add_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, op->interval_ms); if (compare_version("2.1", caller_version) <= 0) { if (op->t_run || op->t_rcchange || op->exec_time || op->queue_time) { crm_trace("Timing data (" PCMK__OP_FMT "): last=%u change=%u exec=%u queue=%u", op->rsc_id, op->op_type, op->interval_ms, op->t_run, op->t_rcchange, op->exec_time, op->queue_time); if ((op->interval_ms != 0) && (op->t_rcchange != 0)) { // Recurring ops may have changed rc after initial run crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_rcchange); } else { crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_run); } crm_xml_add_int(xml_op, XML_RSC_OP_T_EXEC, op->exec_time); crm_xml_add_int(xml_op, XML_RSC_OP_T_QUEUE, op->queue_time); } } if (pcmk__str_any_of(op->op_type, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { /* * Record migrate_source and migrate_target always for migrate ops. */ const char *name = XML_LRM_ATTR_MIGRATE_SOURCE; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); name = XML_LRM_ATTR_MIGRATE_TARGET; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); } add_op_digest_to_xml(op, xml_op); if (op_id_additional) { free(op_id); op_id = op_id_additional; op_id_additional = NULL; goto again; } if (local_user_data) { free(local_user_data); op->user_data = NULL; } free(magic); free(op_id); free(key); return xml_op; } /*! * \internal * \brief Check whether an action shutdown-locks a resource to a node * * If the shutdown-lock cluster property is set, resources will not be recovered * on a different node if cleanly stopped, and may start only on that same node. * This function checks whether that applies to a given action, so that the * transition graph can be marked appropriately. * * \param[in] action Action to check * * \return true if \p action locks its resource to the action's node, * otherwise false */ bool pcmk__action_locks_rsc_to_node(const pe_action_t *action) { // Only resource actions taking place on resource's lock node are locked if ((action == NULL) || (action->rsc == NULL) - || (action->rsc->lock_node == NULL) || (action->node == NULL) - || (action->node->details != action->rsc->lock_node->details)) { + || !pe__same_node(action->node, action->rsc->lock_node)) { return false; } /* During shutdown, only stops are locked (otherwise, another action such as * a demote would cause the controller to clear the lock) */ if (action->node->details->shutdown && (action->task != NULL) && (strcmp(action->task, RSC_STOP) != 0)) { return false; } return true; } /* lowest to highest */ static gint sort_action_id(gconstpointer a, gconstpointer b) { const pe_action_wrapper_t *action_wrapper2 = (const pe_action_wrapper_t *)a; const pe_action_wrapper_t *action_wrapper1 = (const pe_action_wrapper_t *)b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (action_wrapper1->action->id < action_wrapper2->action->id) { return 1; } if (action_wrapper1->action->id > action_wrapper2->action->id) { return -1; } return 0; } /*! * \internal * \brief Remove any duplicate action inputs, merging action flags * * \param[in,out] action Action whose inputs should be checked */ void pcmk__deduplicate_action_inputs(pe_action_t *action) { GList *item = NULL; GList *next = NULL; pe_action_wrapper_t *last_input = NULL; action->actions_before = g_list_sort(action->actions_before, sort_action_id); for (item = action->actions_before; item != NULL; item = next) { pe_action_wrapper_t *input = (pe_action_wrapper_t *) item->data; next = item->next; if ((last_input != NULL) && (input->action->id == last_input->action->id)) { crm_trace("Input %s (%d) duplicate skipped for action %s (%d)", input->action->uuid, input->action->id, action->uuid, action->id); /* For the purposes of scheduling, the ordering flags no longer * matter, but crm_simulate looks at certain ones when creating a * dot graph. Combining the flags is sufficient for that purpose. */ last_input->type |= input->type; if (input->state == pe_link_dumped) { last_input->state = pe_link_dumped; } free(item->data); action->actions_before = g_list_delete_link(action->actions_before, item); } else { last_input = input; input->state = pe_link_not_dumped; } } } /*! * \internal * \brief Output all scheduled actions * * \param[in,out] data_set Cluster working set */ void pcmk__output_actions(pe_working_set_t *data_set) { pcmk__output_t *out = data_set->priv; // Output node (non-resource) actions for (GList *iter = data_set->actions; iter != NULL; iter = iter->next) { char *node_name = NULL; char *task = NULL; pe_action_t *action = (pe_action_t *) iter->data; if (action->rsc != NULL) { continue; // Resource actions will be output later } else if (pcmk_is_set(action->flags, pe_action_optional)) { continue; // This action was not scheduled } if (pcmk__str_eq(action->task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { task = strdup("Shutdown"); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); task = crm_strdup_printf("Fence (%s)", op); } else { continue; // Don't display other node action types } if (pe__is_guest_node(action->node)) { node_name = crm_strdup_printf("%s (resource: %s)", pe__node_name(action->node), action->node->details->remote_rsc->container->id); } else if (action->node != NULL) { node_name = crm_strdup_printf("%s", pe__node_name(action->node)); } out->message(out, "node-action", task, node_name, action->reason); free(node_name); free(task); } // Output resource actions for (GList *iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; rsc->cmds->output_actions(rsc); } } /*! * \internal * \brief Check whether action from resource history is still in configuration * * \param[in] rsc Resource that action is for * \param[in] task Action's name * \param[in] interval_ms Action's interval (in milliseconds) * * \return true if action is still in resource configuration, otherwise false */ static bool action_in_config(const pe_resource_t *rsc, const char *task, guint interval_ms) { char *key = pcmk__op_key(rsc->id, task, interval_ms); bool config = (find_rsc_op_entry(rsc, key) != NULL); free(key); return config; } /*! * \internal * \brief Get action name needed to compare digest for configuration changes * * \param[in] task Action name from history * \param[in] interval_ms Action interval (in milliseconds) * * \return Action name whose digest should be compared */ static const char * task_for_digest(const char *task, guint interval_ms) { /* Certain actions need to be compared against the parameters used to start * the resource. */ if ((interval_ms == 0) && pcmk__str_any_of(task, RSC_STATUS, RSC_MIGRATED, RSC_PROMOTE, NULL)) { task = RSC_START; } return task; } /*! * \internal * \brief Check whether only sanitized parameters to an action changed * * When collecting CIB files for troubleshooting, crm_report will mask * sensitive resource parameters. If simulations were run using that, affected * resources would appear to need a restart, which would complicate * troubleshooting. To avoid that, we save a "secure digest" of non-sensitive * parameters. This function used that digest to check whether only masked * parameters are different. * * \param[in] xml_op Resource history entry with secure digest * \param[in] digest_data Operation digest information being compared * \param[in] data_set Cluster working set * * \return true if only sanitized parameters changed, otherwise false */ static bool only_sanitized_changed(const xmlNode *xml_op, const op_digest_cache_t *digest_data, const pe_working_set_t *data_set) { const char *digest_secure = NULL; if (!pcmk_is_set(data_set->flags, pe_flag_sanitized)) { // The scheduler is not being run as a simulation return false; } digest_secure = crm_element_value(xml_op, XML_LRM_ATTR_SECURE_DIGEST); return (digest_data->rc != RSC_DIGEST_MATCH) && (digest_secure != NULL) && (digest_data->digest_secure_calc != NULL) && (strcmp(digest_data->digest_secure_calc, digest_secure) == 0); } /*! * \internal * \brief Force a restart due to a configuration change * * \param[in,out] rsc Resource that action is for * \param[in] task Name of action whose configuration changed * \param[in] interval_ms Action interval (in milliseconds) * \param[in,out] node Node where resource should be restarted */ static void force_restart(pe_resource_t *rsc, const char *task, guint interval_ms, pe_node_t *node) { char *key = pcmk__op_key(rsc->id, task, interval_ms); pe_action_t *required = custom_action(rsc, key, task, NULL, FALSE, TRUE, rsc->cluster); pe_action_set_reason(required, "resource definition change", true); trigger_unfencing(rsc, node, "Device parameters changed", NULL, rsc->cluster); } /*! * \internal * \brief Schedule a reload of a resource on a node * * \param[in,out] rsc Resource to reload * \param[in] node Where resource should be reloaded */ static void schedule_reload(pe_resource_t *rsc, const pe_node_t *node) { pe_action_t *reload = NULL; // For collective resources, just call recursively for children if (rsc->variant > pe_native) { g_list_foreach(rsc->children, (GFunc) schedule_reload, (gpointer) node); return; } // Skip the reload in certain situations if ((node == NULL) || !pcmk_is_set(rsc->flags, pe_rsc_managed) || pcmk_is_set(rsc->flags, pe_rsc_failed)) { pe_rsc_trace(rsc, "Skip reload of %s:%s%s %s", rsc->id, pcmk_is_set(rsc->flags, pe_rsc_managed)? "" : " unmanaged", pcmk_is_set(rsc->flags, pe_rsc_failed)? " failed" : "", (node == NULL)? "inactive" : node->details->uname); return; } /* If a resource's configuration changed while a start was pending, * force a full restart instead of a reload. */ if (pcmk_is_set(rsc->flags, pe_rsc_start_pending)) { pe_rsc_trace(rsc, "%s: preventing agent reload because start pending", rsc->id); custom_action(rsc, stop_key(rsc), CRMD_ACTION_STOP, node, FALSE, TRUE, rsc->cluster); return; } // Schedule the reload pe__set_resource_flags(rsc, pe_rsc_reload); reload = custom_action(rsc, reload_key(rsc), CRMD_ACTION_RELOAD_AGENT, node, FALSE, TRUE, rsc->cluster); pe_action_set_reason(reload, "resource definition change", FALSE); // Set orderings so that a required stop or demote cancels the reload pcmk__new_ordering(NULL, NULL, reload, rsc, stop_key(rsc), NULL, pe_order_optional|pe_order_then_cancels_first, rsc->cluster); pcmk__new_ordering(NULL, NULL, reload, rsc, demote_key(rsc), NULL, pe_order_optional|pe_order_then_cancels_first, rsc->cluster); } /*! * \internal * \brief Handle any configuration change for an action * * Given an action from resource history, if the resource's configuration * changed since the action was done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, etc.). * * \param[in,out] rsc Resource that action is for * \param[in,out] node Node that action was on * \param[in] xml_op Action XML from resource history * * \return true if action configuration changed, otherwise false */ bool pcmk__check_action_config(pe_resource_t *rsc, pe_node_t *node, const xmlNode *xml_op) { guint interval_ms = 0; const char *task = NULL; const op_digest_cache_t *digest_data = NULL; CRM_CHECK((rsc != NULL) && (node != NULL) && (xml_op != NULL), return false); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); CRM_CHECK(task != NULL, return false); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); // If this is a recurring action, check whether it has been orphaned if (interval_ms > 0) { if (action_in_config(rsc, task, interval_ms)) { pe_rsc_trace(rsc, "%s-interval %s for %s on %s is in configuration", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); } else if (pcmk_is_set(rsc->cluster->flags, pe_flag_stop_action_orphans)) { pcmk__schedule_cancel(rsc, crm_element_value(xml_op, XML_LRM_ATTR_CALLID), task, interval_ms, node, "orphan"); return true; } else { pe_rsc_debug(rsc, "%s-interval %s for %s on %s is orphaned", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); return true; } } crm_trace("Checking %s-interval %s for %s on %s for configuration changes", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); task = task_for_digest(task, interval_ms); digest_data = rsc_action_digest_cmp(rsc, xml_op, node, rsc->cluster); if (only_sanitized_changed(xml_op, digest_data, rsc->cluster)) { if (!pcmk__is_daemon && (rsc->cluster->priv != NULL)) { pcmk__output_t *out = rsc->cluster->priv; out->info(out, "Only 'private' parameters to %s-interval %s for %s " "on %s changed: %s", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node), crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); } return false; } switch (digest_data->rc) { case RSC_DIGEST_RESTART: crm_log_xml_debug(digest_data->params_restart, "params:restart"); force_restart(rsc, task, interval_ms, node); return true; case RSC_DIGEST_ALL: case RSC_DIGEST_UNKNOWN: // Changes that can potentially be handled by an agent reload if (interval_ms > 0) { /* Recurring actions aren't reloaded per se, they are just * re-scheduled so the next run uses the new parameters. * The old instance will be cancelled automatically. */ crm_log_xml_debug(digest_data->params_all, "params:reschedule"); pcmk__reschedule_recurring(rsc, task, interval_ms, node); } else if (crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) { // Agent supports reload, so use it trigger_unfencing(rsc, node, "Device parameters changed (reload)", NULL, rsc->cluster); crm_log_xml_debug(digest_data->params_all, "params:reload"); schedule_reload(rsc, node); } else { pe_rsc_trace(rsc, "Restarting %s because agent doesn't support reload", rsc->id); crm_log_xml_debug(digest_data->params_restart, "params:restart"); force_restart(rsc, task, interval_ms, node); } return true; default: break; } return false; } /*! * \internal * \brief Create a list of resource's action history entries, sorted by call ID * * \param[in] rsc_entry Resource's status XML * \param[out] start_index Where to store index of start-like action, if any * \param[out] stop_index Where to store index of stop action, if any */ static GList * rsc_history_as_list(const xmlNode *rsc_entry, int *start_index, int *stop_index) { GList *ops = NULL; for (xmlNode *rsc_op = first_named_child(rsc_entry, XML_LRM_TAG_RSC_OP); rsc_op != NULL; rsc_op = crm_next_same_xml(rsc_op)) { ops = g_list_prepend(ops, rsc_op); } ops = g_list_sort(ops, sort_op_by_callid); calculate_active_ops(ops, start_index, stop_index); return ops; } /*! * \internal * \brief Process a resource's action history from the CIB status * * Given a resource's action history, if the resource's configuration * changed since the actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in] rsc_entry Resource's status XML * \param[in,out] rsc Resource whose history is being processed * \param[in,out] node Node whose history is being processed */ static void process_rsc_history(const xmlNode *rsc_entry, pe_resource_t *rsc, pe_node_t *node) { int offset = -1; int stop_index = 0; int start_index = 0; GList *sorted_op_list = NULL; if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { if (pe_rsc_is_anon_clone(pe__const_top_resource(rsc, false))) { pe_rsc_trace(rsc, "Skipping configuration check " "for orphaned clone instance %s", rsc->id); } else { pe_rsc_trace(rsc, "Skipping configuration check and scheduling clean-up " "for orphaned resource %s", rsc->id); pcmk__schedule_cleanup(rsc, node, false); } return; } if (pe_find_node_id(rsc->running_on, node->details->id) == NULL) { if (pcmk__rsc_agent_changed(rsc, node, rsc_entry, false)) { pcmk__schedule_cleanup(rsc, node, false); } pe_rsc_trace(rsc, "Skipping configuration check for %s " "because no longer active on %s", rsc->id, pe__node_name(node)); return; } pe_rsc_trace(rsc, "Checking for configuration changes for %s on %s", rsc->id, pe__node_name(node)); if (pcmk__rsc_agent_changed(rsc, node, rsc_entry, true)) { pcmk__schedule_cleanup(rsc, node, false); } sorted_op_list = rsc_history_as_list(rsc_entry, &start_index, &stop_index); if (start_index < stop_index) { return; // Resource is stopped } for (GList *iter = sorted_op_list; iter != NULL; iter = iter->next) { xmlNode *rsc_op = (xmlNode *) iter->data; const char *task = NULL; guint interval_ms = 0; if (++offset < start_index) { // Skip actions that happened before a start continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((interval_ms > 0) && (pcmk_is_set(rsc->flags, pe_rsc_maintenance) || node->details->maintenance)) { // Maintenance mode cancels recurring operations pcmk__schedule_cancel(rsc, crm_element_value(rsc_op, XML_LRM_ATTR_CALLID), task, interval_ms, node, "maintenance mode"); } else if ((interval_ms > 0) || pcmk__strcase_any_of(task, RSC_STATUS, RSC_START, RSC_PROMOTE, RSC_MIGRATED, NULL)) { /* If a resource operation failed, and the operation's definition * has changed, clear any fail count so they can be retried fresh. */ if (pe__bundle_needs_remote_name(rsc)) { - /* We haven't allocated resources to nodes yet, so if the + /* We haven't assigned resources to nodes yet, so if the * REMOTE_CONTAINER_HACK is used, we may calculate the digest * based on the literal "#uname" value rather than the properly * substituted value. That would mistakenly make the action * definition appear to have been changed. Defer the check until * later in this case. */ pe__add_param_check(rsc_op, rsc, node, pe_check_active, rsc->cluster); } else if (pcmk__check_action_config(rsc, node, rsc_op) && (pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL) != 0)) { pe__clear_failcount(rsc, node, "action definition changed", rsc->cluster); } } } g_list_free(sorted_op_list); } /*! * \internal * \brief Process a node's action history from the CIB status * * Given a node's resource history, if the resource's configuration changed * since the actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in,out] node Node whose history is being processed * \param[in] lrm_rscs Node's from CIB status XML */ static void process_node_history(pe_node_t *node, const xmlNode *lrm_rscs) { crm_trace("Processing node history for %s", pe__node_name(node)); for (const xmlNode *rsc_entry = first_named_child(lrm_rscs, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { if (xml_has_children(rsc_entry)) { GList *result = pcmk__rscs_matching_id(ID(rsc_entry), node->details->data_set); for (GList *iter = result; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; if (rsc->variant == pe_native) { process_rsc_history(rsc_entry, rsc, node); } } g_list_free(result); } } } // XPath to find a node's resource history #define XPATH_NODE_HISTORY "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES /*! * \internal * \brief Process any resource configuration changes in the CIB status * * Go through all nodes' resource history, and if a resource's configuration * changed since its actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in,out] data_set Cluster working set */ void pcmk__handle_rsc_config_changes(pe_working_set_t *data_set) { crm_trace("Check resource and action configuration for changes"); /* Rather than iterate through the status section, iterate through the nodes * and search for the appropriate status subsection for each. This skips * orphaned nodes and lets us eliminate some cases before searching the XML. */ for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; /* Don't bother checking actions for a node that can't run actions ... * unless it's in maintenance mode, in which case we still need to * cancel any existing recurring monitors. */ if (node->details->maintenance || pcmk__node_available(node, false, false)) { char *xpath = NULL; xmlNode *history = NULL; xpath = crm_strdup_printf(XPATH_NODE_HISTORY, node->details->uname); history = get_xpath_object(xpath, data_set->input, LOG_NEVER); free(xpath); process_node_history(node, history); } } } diff --git a/lib/pacemaker/pcmk_sched_bundle.c b/lib/pacemaker/pcmk_sched_bundle.c index 63d02c31ae..241f967b4f 100644 --- a/lib/pacemaker/pcmk_sched_bundle.c +++ b/lib/pacemaker/pcmk_sched_bundle.c @@ -1,939 +1,939 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Assign a single bundle replica's resources (other than container) * * \param[in,out] replica Replica to assign * \param[in] user_data Preferred node, if any * * \return true (to indicate that any further replicas should be processed) */ static bool assign_replica(pe__bundle_replica_t *replica, void *user_data) { pe_node_t *container_host = NULL; const pe_node_t *prefer = user_data; const pe_resource_t *bundle = pe__const_top_resource(replica->container, true); if (replica->ip != NULL) { pe_rsc_trace(bundle, "Assigning bundle %s IP %s", bundle->id, replica->ip->id); replica->ip->cmds->assign(replica->ip, prefer); } container_host = replica->container->allocated_to; if (replica->remote != NULL) { if (pe__is_guest_or_remote_node(container_host)) { /* REMOTE_CONTAINER_HACK: "Nested" connection resources must be on * the same host because Pacemaker Remote only supports a single * active connection. */ pcmk__new_colocation("replica-remote-with-host-remote", NULL, INFINITY, replica->remote, container_host->details->remote_rsc, NULL, NULL, true, bundle->cluster); } pe_rsc_trace(bundle, "Assigning bundle %s connection %s", bundle->id, replica->remote->id); replica->remote->cmds->assign(replica->remote, prefer); } if (replica->child != NULL) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, replica->child->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (!pe__same_node(node, replica->node)) { node->weight = -INFINITY; } else if (!pcmk__threshold_reached(replica->child, node, NULL)) { node->weight = INFINITY; } } pe__set_resource_flags(replica->child->parent, pe_rsc_allocating); pe_rsc_trace(bundle, "Assigning bundle %s replica child %s", bundle->id, replica->child->id); replica->child->cmds->assign(replica->child, replica->node); pe__clear_resource_flags(replica->child->parent, pe_rsc_allocating); } return true; } /*! * \internal * \brief Assign a bundle resource to a node * * \param[in,out] rsc Resource to assign to a node * \param[in] prefer Node to prefer, if all else is equal * * \return Node that \p rsc is assigned to, if assigned entirely to one node */ pe_node_t * pcmk__bundle_assign(pe_resource_t *rsc, const pe_node_t *prefer) { GList *containers = NULL; pe_resource_t *bundled_resource = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); pe_rsc_trace(rsc, "Assigning bundle %s", rsc->id); pe__set_resource_flags(rsc, pe_rsc_allocating); - pe__show_node_weights(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), - rsc, __func__, rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), + rsc, __func__, rsc->allowed_nodes, rsc->cluster); // Assign all containers first, so we know what nodes the bundle will be on containers = g_list_sort(pe__bundle_containers(rsc), pcmk__cmp_instance); pcmk__assign_instances(rsc, containers, pe__bundle_max(rsc), rsc->fns->max_per_node(rsc)); g_list_free(containers); // Then assign remaining replica resources pe__foreach_bundle_replica(rsc, assign_replica, (void *) prefer); // Finally, assign the bundled resources to each bundle node bundled_resource = pe__bundled_resource(rsc); if (bundled_resource != NULL) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, bundled_resource->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { if (pe__node_is_bundle_instance(rsc, node)) { node->weight = 0; } else { node->weight = -INFINITY; } } bundled_resource->cmds->assign(bundled_resource, prefer); } pe__clear_resource_flags(rsc, pe_rsc_allocating|pe_rsc_provisional); return NULL; } /*! * \internal * \brief Create actions for a bundle replica's resources (other than child) * * \param[in,out] replica Replica to create actions for * \param[in] user_data Unused * * \return true (to indicate that any further replicas should be processed) */ static bool create_replica_actions(pe__bundle_replica_t *replica, void *user_data) { if (replica->ip != NULL) { replica->ip->cmds->create_actions(replica->ip); } if (replica->container != NULL) { replica->container->cmds->create_actions(replica->container); } if (replica->remote != NULL) { replica->remote->cmds->create_actions(replica->remote); } return true; } /*! * \internal * \brief Create all actions needed for a given bundle resource * * \param[in,out] rsc Bundle resource to create actions for */ void pcmk__bundle_create_actions(pe_resource_t *rsc) { pe_action_t *action = NULL; GList *containers = NULL; pe_resource_t *bundled_resource = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); pe__foreach_bundle_replica(rsc, create_replica_actions, NULL); containers = pe__bundle_containers(rsc); pcmk__create_instance_actions(rsc, containers); g_list_free(containers); bundled_resource = pe__bundled_resource(rsc); if (bundled_resource != NULL) { bundled_resource->cmds->create_actions(bundled_resource); if (pcmk_is_set(bundled_resource->flags, pe_rsc_promotable)) { pe__new_rsc_pseudo_action(rsc, RSC_PROMOTE, true, true); action = pe__new_rsc_pseudo_action(rsc, RSC_PROMOTED, true, true); action->priority = INFINITY; pe__new_rsc_pseudo_action(rsc, RSC_DEMOTE, true, true); action = pe__new_rsc_pseudo_action(rsc, RSC_DEMOTED, true, true); action->priority = INFINITY; } } } /*! * \internal * \brief Create internal constraints for a bundle replica's resources * * \param[in,out] replica Replica to create internal constraints for * \param[in,out] user_data Replica's parent bundle * * \return true (to indicate that any further replicas should be processed) */ static bool replica_internal_constraints(pe__bundle_replica_t *replica, void *user_data) { pe_resource_t *bundle = user_data; replica->container->cmds->internal_constraints(replica->container); // Start bundle -> start replica container pcmk__order_starts(bundle, replica->container, pe_order_runnable_left|pe_order_implies_first_printed); // Stop bundle -> stop replica child and container if (replica->child != NULL) { pcmk__order_stops(bundle, replica->child, pe_order_implies_first_printed); } pcmk__order_stops(bundle, replica->container, pe_order_implies_first_printed); // Start replica container -> bundle is started pcmk__order_resource_actions(replica->container, RSC_START, bundle, RSC_STARTED, pe_order_implies_then_printed); // Stop replica container -> bundle is stopped pcmk__order_resource_actions(replica->container, RSC_STOP, bundle, RSC_STOPPED, pe_order_implies_then_printed); if (replica->ip != NULL) { replica->ip->cmds->internal_constraints(replica->ip); // Replica IP address -> replica container (symmetric) pcmk__order_starts(replica->ip, replica->container, pe_order_runnable_left|pe_order_preserve); pcmk__order_stops(replica->container, replica->ip, pe_order_implies_first|pe_order_preserve); pcmk__new_colocation("ip-with-container", NULL, INFINITY, replica->ip, replica->container, NULL, NULL, true, bundle->cluster); } if (replica->remote != NULL) { /* This handles ordering and colocating remote relative to container * (via "resource-with-container"). Since IP is also ordered and * colocated relative to the container, we don't need to do anything * explicit here with IP. */ replica->remote->cmds->internal_constraints(replica->remote); } if (replica->child != NULL) { CRM_ASSERT(replica->remote != NULL); // "Start remote then child" is implicit in scheduler's remote logic } return true; } /*! * \internal * \brief Create implicit constraints needed for a bundle resource * * \param[in,out] rsc Bundle resource to create implicit constraints for */ void pcmk__bundle_internal_constraints(pe_resource_t *rsc) { pe_resource_t *bundled_resource = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); pe__foreach_bundle_replica(rsc, replica_internal_constraints, rsc); bundled_resource = pe__bundled_resource(rsc); if (bundled_resource == NULL) { return; } // Start bundle -> start bundled clone pcmk__order_resource_actions(rsc, RSC_START, bundled_resource, RSC_START, pe_order_implies_first_printed); // Bundled clone is started -> bundle is started pcmk__order_resource_actions(bundled_resource, RSC_STARTED, rsc, RSC_STARTED, pe_order_implies_then_printed); // Stop bundle -> stop bundled clone pcmk__order_resource_actions(rsc, RSC_STOP, bundled_resource, RSC_STOP, pe_order_implies_first_printed); // Bundled clone is stopped -> bundle is stopped pcmk__order_resource_actions(bundled_resource, RSC_STOPPED, rsc, RSC_STOPPED, pe_order_implies_then_printed); bundled_resource->cmds->internal_constraints(bundled_resource); if (!pcmk_is_set(bundled_resource->flags, pe_rsc_promotable)) { return; } pcmk__promotable_restart_ordering(rsc); // Demote bundle -> demote bundled clone pcmk__order_resource_actions(rsc, RSC_DEMOTE, bundled_resource, RSC_DEMOTE, pe_order_implies_first_printed); // Bundled clone is demoted -> bundle is demoted pcmk__order_resource_actions(bundled_resource, RSC_DEMOTED, rsc, RSC_DEMOTED, pe_order_implies_then_printed); // Promote bundle -> promote bundled clone pcmk__order_resource_actions(rsc, RSC_PROMOTE, bundled_resource, RSC_PROMOTE, pe_order_implies_first_printed); // Bundled clone is promoted -> bundle is promoted pcmk__order_resource_actions(bundled_resource, RSC_PROMOTED, rsc, RSC_PROMOTED, pe_order_implies_then_printed); } struct match_data { const pe_node_t *node; // Node to compare against replica pe_resource_t *container; // Replica container corresponding to node }; /*! * \internal * \brief Check whether a replica container is assigned to a given node * * \param[in,out] replica Replica to check * \param[in,out] user_data struct match_data with node to compare against * * \return true if the replica does not match (to indicate further replicas * should be processed), otherwise false */ static bool match_replica_container(pe__bundle_replica_t *replica, void *user_data) { struct match_data *match_data = user_data; if (pcmk__instance_matches(replica->container, match_data->node, RSC_ROLE_UNKNOWN, false)) { match_data->container = replica->container; return false; // Match found, don't bother searching further replicas } return true; // No match, keep searching } /*! * \internal * \brief Find a bundle container compatible with a dependent resource * * \param[in] dependent Dependent resource in colocation with bundle * \param[in] bundle Bundle that \p dependent is colocated with * * \return A container from \p bundle assigned to the same node as \p dependent * if assigned, otherwise assigned to any of dependent's allowed nodes, * otherwise NULL. */ static pe_resource_t * compatible_container(const pe_resource_t *dependent, pe_resource_t *bundle) { GList *scratch = NULL; struct match_data match_data = { NULL, NULL }; // If dependent is assigned, only check there match_data.node = dependent->fns->location(dependent, NULL, 0); if (match_data.node != NULL) { pe__foreach_bundle_replica(bundle, match_replica_container, &match_data); return match_data.container; } // Otherwise, check for any of the dependent's allowed nodes scratch = g_hash_table_get_values(dependent->allowed_nodes); scratch = pcmk__sort_nodes(scratch, NULL); for (const GList *iter = scratch; iter != NULL; iter = iter->next) { match_data.node = (const pe_node_t *) iter->data; pe__foreach_bundle_replica(bundle, match_replica_container, &match_data); if (match_data.container != NULL) { break; } } g_list_free(scratch); return match_data.container; } struct coloc_data { const pcmk__colocation_t *colocation; pe_resource_t *dependent; GList *container_hosts; }; /*! * \internal - * \brief Apply a colocation score to replica node weights or resource priority + * \brief Apply a colocation score to replica node scores or resource priority * * \param[in,out] replica Replica to apply colocation score to * \param[in,out] user_data struct coloc_data for colocation being applied * * \return true (to indicate that any further replicas should be processed) */ static bool replica_apply_coloc_score(pe__bundle_replica_t *replica, void *user_data) { struct coloc_data *coloc_data = user_data; pe_node_t *chosen = NULL; if (coloc_data->colocation->score < INFINITY) { replica->container->cmds->apply_coloc_score(coloc_data->dependent, replica->container, coloc_data->colocation, false); return true; } chosen = replica->container->fns->location(replica->container, NULL, 0); if ((chosen == NULL) || is_set_recursive(replica->container, pe_rsc_block, true)) { return true; } if ((coloc_data->colocation->primary_role >= RSC_ROLE_PROMOTED) && ((replica->child == NULL) || (replica->child->next_role < RSC_ROLE_PROMOTED))) { return true; } pe_rsc_trace(pe__const_top_resource(replica->container, true), "Allowing mandatory colocation %s using %s @%d", coloc_data->colocation->id, pe__node_name(chosen), chosen->weight); coloc_data->container_hosts = g_list_prepend(coloc_data->container_hosts, chosen); return true; } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint, apply its score to the dependent's - * allowed node weights (if we are still placing resources) or priority (if + * allowed node scores (if we are still placing resources) or priority (if * we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply * \param[in] for_dependent true if called on behalf of dependent */ void pcmk__bundle_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent) { struct coloc_data coloc_data = { colocation, dependent, NULL }; /* This should never be called for the bundle itself as a dependent. * Instead, we add its colocation constraints to its containers and call the * apply_coloc_score() method for the containers as dependents. */ CRM_ASSERT((primary != NULL) && (primary->variant == pe_container) && (dependent != NULL) && (dependent->variant == pe_native) && (colocation != NULL) && !for_dependent); if (pcmk_is_set(primary->flags, pe_rsc_provisional)) { pe_rsc_trace(primary, "Skipping applying colocation %s " "because %s is still provisional", colocation->id, primary->id); return; } pe_rsc_trace(primary, "Applying colocation %s (%s with %s at %s)", colocation->id, dependent->id, primary->id, pcmk_readable_score(colocation->score)); /* If the constraint dependent is a clone or bundle, "dependent" here is one * of its instances. Look for a compatible instance of this bundle. */ if (colocation->dependent->variant > pe_group) { pe_resource_t *primary_container = NULL; primary_container = compatible_container(dependent, primary); if (primary_container != NULL) { // Success, we found one pe_rsc_debug(primary, "Pairing %s with %s", dependent->id, primary_container->id); dependent->cmds->apply_coloc_score(dependent, primary_container, colocation, true); } else if (colocation->score >= INFINITY) { // Failure, and it's fatal crm_notice("%s cannot run because there is no compatible " "instance of %s to colocate with", dependent->id, primary->id); pcmk__assign_resource(dependent, NULL, true); } else { // Failure, but we can ignore it pe_rsc_debug(primary, "%s cannot be colocated with any instance of %s", dependent->id, primary->id); } return; } pe__foreach_bundle_replica(primary, replica_apply_coloc_score, &coloc_data); if (colocation->score >= INFINITY) { node_list_exclude(dependent->allowed_nodes, coloc_data.container_hosts, FALSE); } g_list_free(coloc_data.container_hosts); } // Bundle implementation of resource_alloc_functions_t:with_this_colocations() void pcmk__with_bundle_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (rsc->variant == pe_container) && (orig_rsc != NULL) && (list != NULL), return); if (rsc == orig_rsc) { // Colocations are wanted for bundle itself pcmk__add_with_this_list(list, rsc->rsc_cons_lhs); // Only the bundle replicas' containers get the bundle's constraints } else if (pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) { pcmk__add_collective_constraints(list, orig_rsc, rsc, true); } } // Bundle implementation of resource_alloc_functions_t:this_with_colocations() void pcmk__bundle_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (rsc->variant == pe_container) && (orig_rsc != NULL) && (list != NULL), return); if (rsc == orig_rsc) { // Colocations are wanted for bundle itself pcmk__add_this_with_list(list, rsc->rsc_cons); // Only the bundle replicas' containers get the bundle's constraints } else if (pcmk_is_set(orig_rsc->flags, pe_rsc_replica_container)) { pcmk__add_collective_constraints(list, orig_rsc, rsc, false); } } /*! * \internal * \brief Return action flags for a given bundle resource action * * \param[in,out] action Bundle resource action to get flags for * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__bundle_action_flags(pe_action_t *action, const pe_node_t *node) { GList *containers = NULL; uint32_t flags = 0; pe_resource_t *bundled_resource = NULL; CRM_ASSERT((action != NULL) && (action->rsc != NULL) && (action->rsc->variant == pe_container)); bundled_resource = pe__bundled_resource(action->rsc); if (bundled_resource != NULL) { // Clone actions are done on the bundled clone resource, not container switch (get_complex_task(bundled_resource, action->task)) { case no_action: case action_notify: case action_notified: case action_promote: case action_promoted: case action_demote: case action_demoted: return pcmk__collective_action_flags(action, bundled_resource->children, node); default: break; } } containers = pe__bundle_containers(action->rsc); flags = pcmk__collective_action_flags(action, containers, node); g_list_free(containers); return flags; } /*! * \internal * \brief Apply a location constraint to a bundle replica * * \param[in,out] replica Replica to apply constraint to * \param[in,out] user_data Location constraint to apply * * \return true (to indicate that any further replicas should be processed) */ static bool apply_location_to_replica(pe__bundle_replica_t *replica, void *user_data) { pe__location_t *location = user_data; if (replica->container != NULL) { replica->container->cmds->apply_location(replica->container, location); } if (replica->ip != NULL) { replica->ip->cmds->apply_location(replica->ip, location); } return true; } /*! * \internal * \brief Apply a location constraint to a bundle resource's allowed node scores * * \param[in,out] rsc Bundle resource to apply constraint to * \param[in,out] location Location constraint to apply */ void pcmk__bundle_apply_location(pe_resource_t *rsc, pe__location_t *location) { pe_resource_t *bundled_resource = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container) && (location != NULL)); pcmk__apply_location(rsc, location); pe__foreach_bundle_replica(rsc, apply_location_to_replica, location); bundled_resource = pe__bundled_resource(rsc); if ((bundled_resource != NULL) && ((location->role_filter == RSC_ROLE_UNPROMOTED) || (location->role_filter == RSC_ROLE_PROMOTED))) { bundled_resource->cmds->apply_location(bundled_resource, location); bundled_resource->rsc_location = g_list_prepend( bundled_resource->rsc_location, location); } } #define XPATH_REMOTE "//nvpair[@name='" XML_RSC_ATTR_REMOTE_RA_ADDR "']" /*! * \internal * \brief Add a bundle replica's actions to transition graph * * \param[in,out] replica Replica to add to graph * \param[in] user_data Bundle that replica belongs to (for logging only) * * \return true (to indicate that any further replicas should be processed) */ static bool add_replica_actions_to_graph(pe__bundle_replica_t *replica, void *user_data) { if ((replica->remote != NULL) && (replica->container != NULL) && pe__bundle_needs_remote_name(replica->remote)) { /* REMOTE_CONTAINER_HACK: Allow remote nodes to run containers that * run pacemaker-remoted inside, without needing a separate IP for * the container. This is done by configuring the inner remote's * connection host as the magic string "#uname", then * replacing it with the underlying host when needed. */ xmlNode *nvpair = get_xpath_object(XPATH_REMOTE, replica->remote->xml, LOG_ERR); const char *calculated_addr = NULL; // Replace the value in replica->remote->xml (if appropriate) calculated_addr = pe__add_bundle_remote_name(replica->remote, replica->remote->cluster, nvpair, "value"); if (calculated_addr != NULL) { /* Since this is for the bundle as a resource, and not any * particular action, replace the value in the default * parameters (not evaluated for node). create_graph_action() * will grab it from there to replace it in node-evaluated * parameters. */ GHashTable *params = pe_rsc_params(replica->remote, NULL, replica->remote->cluster); g_hash_table_replace(params, strdup(XML_RSC_ATTR_REMOTE_RA_ADDR), strdup(calculated_addr)); } else { pe_resource_t *bundle = user_data; /* The only way to get here is if the remote connection is * neither currently running nor scheduled to run. That means we * won't be doing any operations that require addr (only start * requires it; we additionally use it to compare digests when * unpacking status, promote, and migrate_from history, but * that's already happened by this point). */ pe_rsc_info(bundle, "Unable to determine address for bundle %s " "remote connection", bundle->id); } } if (replica->ip != NULL) { replica->ip->cmds->add_actions_to_graph(replica->ip); } if (replica->container != NULL) { replica->container->cmds->add_actions_to_graph(replica->container); } if (replica->remote != NULL) { replica->remote->cmds->add_actions_to_graph(replica->remote); } return true; } /*! * \internal * \brief Add a bundle resource's actions to the transition graph * * \param[in,out] rsc Bundle resource whose actions should be added */ void pcmk__bundle_add_actions_to_graph(pe_resource_t *rsc) { pe_resource_t *bundled_resource = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); bundled_resource = pe__bundled_resource(rsc); if (bundled_resource != NULL) { bundled_resource->cmds->add_actions_to_graph(bundled_resource); } pe__foreach_bundle_replica(rsc, add_replica_actions_to_graph, rsc); } struct probe_data { pe_resource_t *bundle; // Bundle being probed pe_node_t *node; // Node to create probes on bool any_created; // Whether any probes have been created }; /*! * \internal * \brief Order a bundle replica's start after another replica's probe * * \param[in,out] replica Replica to order start for * \param[in,out] user_data Replica with probe to order after * * \return true (to indicate that any further replicas should be processed) */ static bool order_replica_start_after(pe__bundle_replica_t *replica, void *user_data) { pe__bundle_replica_t *probed_replica = user_data; if ((replica == probed_replica) || (replica->container == NULL)) { return true; } pcmk__new_ordering(probed_replica->container, pcmk__op_key(probed_replica->container->id, RSC_STATUS, 0), NULL, replica->container, pcmk__op_key(replica->container->id, RSC_START, 0), NULL, pe_order_optional|pe_order_same_node, replica->container->cluster); return true; } /*! * \internal * \brief Create probes for a bundle replica's resources * * \param[in,out] replica Replica to create probes for * \param[in,out] user_data struct probe_data * * \return true (to indicate that any further replicas should be processed) */ static bool create_replica_probes(pe__bundle_replica_t *replica, void *user_data) { struct probe_data *probe_data = user_data; if ((replica->ip != NULL) && replica->ip->cmds->create_probe(replica->ip, probe_data->node)) { probe_data->any_created = true; } if ((replica->child != NULL) && pe__same_node(probe_data->node, replica->node) && replica->child->cmds->create_probe(replica->child, probe_data->node)) { probe_data->any_created = true; } if ((replica->container != NULL) && replica->container->cmds->create_probe(replica->container, probe_data->node)) { probe_data->any_created = true; /* If we're limited to one replica per host (due to * the lack of an IP range probably), then we don't * want any of our peer containers starting until * we've established that no other copies are already * running. * * Partly this is to ensure that the maximum replicas per host is * observed, but also to ensure that the containers * don't fail to start because the necessary port * mappings (which won't include an IP for uniqueness) * are already taken */ if (probe_data->bundle->fns->max_per_node(probe_data->bundle) == 1) { pe__foreach_bundle_replica(probe_data->bundle, order_replica_start_after, replica); } } if ((replica->container != NULL) && (replica->remote != NULL) && replica->remote->cmds->create_probe(replica->remote, probe_data->node)) { /* Do not probe the remote resource until we know where the container is * running. This is required for REMOTE_CONTAINER_HACK to correctly * probe remote resources. */ char *probe_uuid = pcmk__op_key(replica->remote->id, RSC_STATUS, 0); pe_action_t *probe = find_first_action(replica->remote->actions, probe_uuid, NULL, probe_data->node); free(probe_uuid); if (probe != NULL) { probe_data->any_created = true; pe_rsc_trace(probe_data->bundle, "Ordering %s probe on %s", replica->remote->id, pe__node_name(probe_data->node)); pcmk__new_ordering(replica->container, pcmk__op_key(replica->container->id, RSC_START, 0), NULL, replica->remote, NULL, probe, pe_order_probe, probe_data->bundle->cluster); } } return true; } /*! * \internal * * \brief Schedule any probes needed for a bundle resource on a node * * \param[in,out] rsc Bundle resource to create probes for * \param[in,out] node Node to create probe on * * \return true if any probe was created, otherwise false */ bool pcmk__bundle_create_probe(pe_resource_t *rsc, pe_node_t *node) { struct probe_data probe_data = { rsc, node, false }; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); pe__foreach_bundle_replica(rsc, create_replica_probes, &probe_data); return probe_data.any_created; } /*! * \internal * \brief Output actions for one bundle replica * * \param[in,out] replica Replica to output actions for * \param[in] user_data Unused * * \return true (to indicate that any further replicas should be processed) */ static bool output_replica_actions(pe__bundle_replica_t *replica, void *user_data) { if (replica->ip != NULL) { replica->ip->cmds->output_actions(replica->ip); } if (replica->container != NULL) { replica->container->cmds->output_actions(replica->container); } if (replica->remote != NULL) { replica->remote->cmds->output_actions(replica->remote); } if (replica->child != NULL) { replica->child->cmds->output_actions(replica->child); } return true; } /*! * \internal * \brief Output a summary of scheduled actions for a bundle resource * * \param[in,out] rsc Bundle resource to output actions for */ void pcmk__output_bundle_actions(pe_resource_t *rsc) { CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); pe__foreach_bundle_replica(rsc, output_replica_actions, NULL); } // Bundle implementation of resource_alloc_functions_t:add_utilization() void pcmk__bundle_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization) { pe_resource_t *container = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } /* All bundle replicas are identical, so using the utilization of the first * is sufficient for any. Only the implicit container resource can have * utilization values. */ container = pe__first_container(rsc); if (container != NULL) { container->cmds->add_utilization(container, orig_rsc, all_rscs, utilization); } } // Bundle implementation of resource_alloc_functions_t:shutdown_lock() void pcmk__bundle_shutdown_lock(pe_resource_t *rsc) { CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_container)); // Bundles currently don't support shutdown locks } diff --git a/lib/pacemaker/pcmk_sched_clone.c b/lib/pacemaker/pcmk_sched_clone.c index e0ff0b0a23..8df2a1d9f1 100644 --- a/lib/pacemaker/pcmk_sched_clone.c +++ b/lib/pacemaker/pcmk_sched_clone.c @@ -1,672 +1,672 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Assign a clone resource's instances to nodes * * \param[in,out] rsc Clone resource to assign * \param[in] prefer Node to prefer, if all else is equal * * \return NULL (clones are not assigned to a single node) */ pe_node_t * pcmk__clone_assign(pe_resource_t *rsc, const pe_node_t *prefer) { CRM_ASSERT(pe_rsc_is_clone(rsc)); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return NULL; // Assignment has already been done } // Detect assignment loops if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Breaking assignment loop involving %s", rsc->id); return NULL; } pe__set_resource_flags(rsc, pe_rsc_allocating); // If this clone is promotable, consider nodes' promotion scores if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__add_promotion_scores(rsc); } /* If this clone is colocated with any other resources, assign those first. * Since the this_with_colocations() method boils down to a copy of rsc_cons * for clones, we can use that here directly for efficiency. */ for (GList *iter = rsc->rsc_cons; iter != NULL; iter = iter->next) { pcmk__colocation_t *constraint = (pcmk__colocation_t *) iter->data; pe_rsc_trace(rsc, "%s: Assigning colocation %s primary %s first", rsc->id, constraint->id, constraint->primary->id); constraint->primary->cmds->assign(constraint->primary, prefer); } /* If any resources are colocated with this one, consider their preferences. * Because the with_this_colocations() method boils down to a copy of * rsc_cons_lhs for clones, we can use that here directly for efficiency. */ g_list_foreach(rsc->rsc_cons_lhs, pcmk__add_dependent_scores, rsc); - pe__show_node_weights(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), - rsc, __func__, rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), + rsc, __func__, rsc->allowed_nodes, rsc->cluster); rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance); pcmk__assign_instances(rsc, rsc->children, pe__clone_max(rsc), pe__clone_node_max(rsc)); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__set_instance_roles(rsc); } pe__clear_resource_flags(rsc, pe_rsc_provisional|pe_rsc_allocating); pe_rsc_trace(rsc, "Assigned clone %s", rsc->id); return NULL; } /*! * \internal * \brief Create all actions needed for a given clone resource * * \param[in,out] rsc Clone resource to create actions for */ void pcmk__clone_create_actions(pe_resource_t *rsc) { CRM_ASSERT(pe_rsc_is_clone(rsc)); pe_rsc_trace(rsc, "Creating actions for clone %s", rsc->id); pcmk__create_instance_actions(rsc, rsc->children); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__create_promotable_actions(rsc); } } /*! * \internal * \brief Create implicit constraints needed for a clone resource * * \param[in,out] rsc Clone resource to create implicit constraints for */ void pcmk__clone_internal_constraints(pe_resource_t *rsc) { bool ordered = pe__clone_is_ordered(rsc); CRM_ASSERT(pe_rsc_is_clone(rsc)); pe_rsc_trace(rsc, "Creating internal constraints for clone %s", rsc->id); // Restart ordering: Stop -> stopped -> start -> started pcmk__order_resource_actions(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_START, rsc, RSC_STARTED, pe_order_runnable_left); pcmk__order_resource_actions(rsc, RSC_STOP, rsc, RSC_STOPPED, pe_order_runnable_left); // Demoted -> stop and started -> promote if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__order_resource_actions(rsc, RSC_DEMOTED, rsc, RSC_STOP, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_STARTED, rsc, RSC_PROMOTE, pe_order_runnable_left); } if (ordered) { /* Ordered clone instances must start and stop by instance number. The * instances might have been previously shuffled for assignment or * promotion purposes, so re-sort them. */ rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance_number); } for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->cmds->internal_constraints(instance); // Start clone -> start instance -> clone started pcmk__order_starts(rsc, instance, pe_order_runnable_left |pe_order_implies_first_printed); pcmk__order_resource_actions(instance, RSC_START, rsc, RSC_STARTED, pe_order_implies_then_printed); // Stop clone -> stop instance -> clone stopped pcmk__order_stops(rsc, instance, pe_order_implies_first_printed); pcmk__order_resource_actions(instance, RSC_STOP, rsc, RSC_STOPPED, pe_order_implies_then_printed); /* Instances of ordered clones must be started and stopped by instance * number. Since only some instances may be starting or stopping, order * each instance relative to every later instance. */ if (ordered) { for (GList *later = iter->next; later != NULL; later = later->next) { pcmk__order_starts(instance, (pe_resource_t *) later->data, pe_order_optional); pcmk__order_stops((pe_resource_t *) later->data, instance, pe_order_optional); } } } if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { pcmk__order_promotable_instances(rsc); } } /*! * \internal * \brief Check whether colocated resources can be interleaved * * \param[in] colocation Colocation constraint with clone as primary * * \return true if colocated resources can be interleaved, otherwise false */ static bool can_interleave(const pcmk__colocation_t *colocation) { const pe_resource_t *dependent = colocation->dependent; // Only colocations between clone or bundle resources use interleaving if (dependent->variant <= pe_group) { return false; } // Only the dependent needs to be marked for interleaving if (!crm_is_true(g_hash_table_lookup(dependent->meta, XML_RSC_ATTR_INTERLEAVE))) { return false; } /* @TODO Do we actually care about multiple primary instances sharing a * dependent instance? */ if (dependent->fns->max_per_node(dependent) != colocation->primary->fns->max_per_node(colocation->primary)) { pcmk__config_err("Cannot interleave %s and %s because they do not " "support the same number of instances per node", dependent->id, colocation->primary->id); return false; } return true; } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint, apply its score to the dependent's - * allowed node weights (if we are still placing resources) or priority (if + * allowed node scores (if we are still placing resources) or priority (if * we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply * \param[in] for_dependent true if called on behalf of dependent */ void pcmk__clone_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent) { GList *iter = NULL; /* This should never be called for the clone itself as a dependent. Instead, * we add its colocation constraints to its instances and call the * apply_coloc_score() method for the instances as dependents. */ CRM_ASSERT(!for_dependent); CRM_ASSERT((colocation != NULL) && pe_rsc_is_clone(primary) && (dependent != NULL) && (dependent->variant == pe_native)); if (pcmk_is_set(primary->flags, pe_rsc_provisional)) { pe_rsc_trace(primary, "Delaying processing colocation %s " "because cloned primary %s is still provisional", colocation->id, primary->id); return; } pe_rsc_trace(primary, "Processing colocation %s (%s with clone %s @%s)", colocation->id, dependent->id, primary->id, pcmk_readable_score(colocation->score)); // Apply role-specific colocations if (pcmk_is_set(primary->flags, pe_rsc_promotable) && (colocation->primary_role != RSC_ROLE_UNKNOWN)) { if (pcmk_is_set(dependent->flags, pe_rsc_provisional)) { // We're assigning the dependent to a node pcmk__update_dependent_with_promotable(primary, dependent, colocation); return; } if (colocation->dependent_role == RSC_ROLE_PROMOTED) { // We're choosing a role for the dependent pcmk__update_promotable_dependent_priority(primary, dependent, colocation); return; } } // Apply interleaved colocations if (can_interleave(colocation)) { pe_resource_t *primary_instance = NULL; primary_instance = pcmk__find_compatible_instance(dependent, primary, RSC_ROLE_UNKNOWN, false); if (primary_instance != NULL) { pe_rsc_debug(primary, "Interleaving %s with %s", dependent->id, primary_instance->id); dependent->cmds->apply_coloc_score(dependent, primary_instance, colocation, true); } else if (colocation->score >= INFINITY) { crm_notice("%s cannot run because it cannot interleave with " "any instance of %s", dependent->id, primary->id); pcmk__assign_resource(dependent, NULL, true); } else { pe_rsc_debug(primary, "%s will not colocate with %s " "because no instance can interleave with it", dependent->id, primary->id); } return; } // Apply mandatory colocations if (colocation->score >= INFINITY) { GList *primary_nodes = NULL; // Dependent can run only where primary will have unblocked instances for (iter = primary->children; iter != NULL; iter = iter->next) { const pe_resource_t *instance = iter->data; pe_node_t *chosen = instance->fns->location(instance, NULL, 0); if ((chosen != NULL) && !is_set_recursive(instance, pe_rsc_block, TRUE)) { pe_rsc_trace(primary, "Allowing %s: %s %d", colocation->id, pe__node_name(chosen), chosen->weight); primary_nodes = g_list_prepend(primary_nodes, chosen); } } node_list_exclude(dependent->allowed_nodes, primary_nodes, FALSE); g_list_free(primary_nodes); return; } // Apply optional colocations for (iter = primary->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->cmds->apply_coloc_score(dependent, instance, colocation, false); } } // Clone implementation of resource_alloc_functions_t:with_this_colocations() void pcmk__with_clone_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (orig_rsc != NULL) && (list != NULL), return); if (rsc == orig_rsc) { // Colocations are wanted for clone itself pcmk__add_with_this_list(list, rsc->rsc_cons_lhs); } else { pcmk__add_collective_constraints(list, orig_rsc, rsc, true); } } // Clone implementation of resource_alloc_functions_t:this_with_colocations() void pcmk__clone_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (orig_rsc != NULL) && (list != NULL), return); if (rsc == orig_rsc) { // Colocations are wanted for clone itself pcmk__add_this_with_list(list, rsc->rsc_cons); } else { pcmk__add_collective_constraints(list, orig_rsc, rsc, false); } } /*! * \internal * \brief Return action flags for a given clone resource action * * \param[in,out] action Action to get flags for * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__clone_action_flags(pe_action_t *action, const pe_node_t *node) { CRM_ASSERT((action != NULL) && pe_rsc_is_clone(action->rsc)); return pcmk__collective_action_flags(action, action->rsc->children, node); } /*! * \internal * \brief Apply a location constraint to a clone resource's allowed node scores * * \param[in,out] rsc Clone resource to apply constraint to * \param[in,out] location Location constraint to apply */ void pcmk__clone_apply_location(pe_resource_t *rsc, pe__location_t *location) { CRM_CHECK((location != NULL) && pe_rsc_is_clone(rsc), return); pcmk__apply_location(rsc, location); for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->cmds->apply_location(instance, location); } } /*! * \internal * \brief Add a clone resource's actions to the transition graph * * \param[in,out] rsc Resource whose actions should be added */ void pcmk__clone_add_actions_to_graph(pe_resource_t *rsc) { CRM_CHECK(pe_rsc_is_clone(rsc), return); g_list_foreach(rsc->actions, (GFunc) rsc->cmds->action_flags, NULL); pe__create_clone_notifications(rsc); for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child_rsc = (pe_resource_t *) iter->data; child_rsc->cmds->add_actions_to_graph(child_rsc); } pcmk__add_rsc_actions_to_graph(rsc); pe__free_clone_notification_data(rsc); } /*! * \internal * \brief Check whether a resource or any children have been probed on a node * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return true if \p node is in the known_on table of \p rsc or any of its * children, otherwise false */ static bool rsc_probed_on(const pe_resource_t *rsc, const pe_node_t *node) { if (rsc->children != NULL) { for (GList *child_iter = rsc->children; child_iter != NULL; child_iter = child_iter->next) { pe_resource_t *child = (pe_resource_t *) child_iter->data; if (rsc_probed_on(child, node)) { return true; } } return false; } if (rsc->known_on != NULL) { GHashTableIter iter; pe_node_t *known_node = NULL; g_hash_table_iter_init(&iter, rsc->known_on); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &known_node)) { if (pe__same_node(node, known_node)) { return true; } } } return false; } /*! * \internal * \brief Find clone instance that has been probed on given node * * \param[in] clone Clone resource to check * \param[in] node Node to check * * \return Instance of \p clone that has been probed on \p node if any, * otherwise NULL */ static pe_resource_t * find_probed_instance_on(const pe_resource_t *clone, const pe_node_t *node) { for (GList *iter = clone->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; if (rsc_probed_on(instance, node)) { return instance; } } return NULL; } /*! * \internal * \brief Probe an anonymous clone on a node * * \param[in,out] clone Anonymous clone to probe * \param[in,out] node Node to probe \p clone on */ static bool probe_anonymous_clone(pe_resource_t *clone, pe_node_t *node) { // Check whether we already probed an instance on this node pe_resource_t *child = find_probed_instance_on(clone, node); // Otherwise, check if we plan to start an instance on this node for (GList *iter = clone->children; (iter != NULL) && (child == NULL); iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; const pe_node_t *instance_node = NULL; instance_node = instance->fns->location(instance, NULL, 0); if (pe__same_node(instance_node, node)) { child = instance; } } // Otherwise, use the first clone instance if (child == NULL) { child = clone->children->data; } // Anonymous clones only need to probe a single instance return child->cmds->create_probe(child, node); } /*! * \internal * \brief Schedule any probes needed for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return true if any probe was created, otherwise false */ bool pcmk__clone_create_probe(pe_resource_t *rsc, pe_node_t *node) { CRM_CHECK((node != NULL) && pe_rsc_is_clone(rsc), return false); if (rsc->exclusive_discover) { /* The clone is configured to be probed only where a location constraint * exists with resource-discovery set to exclusive. * * This check is not strictly necessary here since the instance's * create_probe() method would also check, but doing it here is more * efficient (especially for unique clones with a large number of * instances), and affects the CRM_meta_notify_available_uname variable * passed with notify actions. */ pe_node_t *allowed = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if ((allowed == NULL) || (allowed->rsc_discover_mode != pe_discover_exclusive)) { /* This node is not marked for resource discovery. Remove it from * allowed_nodes so that notifications contain only nodes that the * clone can possibly run on. */ pe_rsc_trace(rsc, "Skipping probe for %s on %s because resource has " "exclusive discovery but is not allowed on node", rsc->id, pe__node_name(node)); g_hash_table_remove(rsc->allowed_nodes, node->details->id); return false; } } rsc->children = g_list_sort(rsc->children, pcmk__cmp_instance_number); if (pcmk_is_set(rsc->flags, pe_rsc_unique)) { return pcmk__probe_resource_list(rsc->children, node); } else { return probe_anonymous_clone(rsc, node); } } /*! * \internal * \brief Add meta-attributes relevant to transition graph actions to XML * * Add clone-specific meta-attributes needed for transition graph actions. * * \param[in] rsc Clone resource whose meta-attributes should be added * \param[in,out] xml Transition graph action attributes XML to add to */ void pcmk__clone_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml) { char *name = NULL; CRM_ASSERT(pe_rsc_is_clone(rsc) && (xml != NULL)); name = crm_meta_name(XML_RSC_ATTR_UNIQUE); crm_xml_add(xml, name, pe__rsc_bool_str(rsc, pe_rsc_unique)); free(name); name = crm_meta_name(XML_RSC_ATTR_NOTIFY); crm_xml_add(xml, name, pe__rsc_bool_str(rsc, pe_rsc_notify)); free(name); name = crm_meta_name(XML_RSC_ATTR_INCARNATION_MAX); crm_xml_add_int(xml, name, pe__clone_max(rsc)); free(name); name = crm_meta_name(XML_RSC_ATTR_INCARNATION_NODEMAX); crm_xml_add_int(xml, name, pe__clone_node_max(rsc)); free(name); if (pcmk_is_set(rsc->flags, pe_rsc_promotable)) { int promoted_max = pe__clone_promoted_max(rsc); int promoted_node_max = pe__clone_promoted_node_max(rsc); name = crm_meta_name(XML_RSC_ATTR_PROMOTED_MAX); crm_xml_add_int(xml, name, promoted_max); free(name); name = crm_meta_name(XML_RSC_ATTR_PROMOTED_NODEMAX); crm_xml_add_int(xml, name, promoted_node_max); free(name); /* @COMPAT Maintain backward compatibility with resource agents that * expect the old names (deprecated since 2.0.0). */ name = crm_meta_name(PCMK_XA_PROMOTED_MAX_LEGACY); crm_xml_add_int(xml, name, promoted_max); free(name); name = crm_meta_name(PCMK_XA_PROMOTED_NODE_MAX_LEGACY); crm_xml_add_int(xml, name, promoted_node_max); free(name); } } // Clone implementation of resource_alloc_functions_t:add_utilization() void pcmk__clone_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization) { bool existing = false; pe_resource_t *child = NULL; if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } // Look for any child already existing in the list for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { child = (pe_resource_t *) iter->data; if (g_list_find(all_rscs, child)) { existing = true; // Keep checking remaining children } else { // If this is a clone of a group, look for group's members for (GList *member_iter = child->children; member_iter != NULL; member_iter = member_iter->next) { pe_resource_t *member = (pe_resource_t *) member_iter->data; if (g_list_find(all_rscs, member) != NULL) { // Add *child's* utilization, not group member's child->cmds->add_utilization(child, orig_rsc, all_rscs, utilization); existing = true; break; } } } } if (!existing && (rsc->children != NULL)) { // If nothing was found, still add first child's utilization child = (pe_resource_t *) rsc->children->data; child->cmds->add_utilization(child, orig_rsc, all_rscs, utilization); } } // Clone implementation of resource_alloc_functions_t:shutdown_lock() void pcmk__clone_shutdown_lock(pe_resource_t *rsc) { return; // Clones currently don't support shutdown locks } diff --git a/lib/pacemaker/pcmk_sched_colocation.c b/lib/pacemaker/pcmk_sched_colocation.c index a262633a92..d1be935978 100644 --- a/lib/pacemaker/pcmk_sched_colocation.c +++ b/lib/pacemaker/pcmk_sched_colocation.c @@ -1,1665 +1,1663 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include "crm/common/util.h" #include "crm/common/xml_internal.h" #include "crm/msg_xml.h" #include "libpacemaker_private.h" #define EXPAND_CONSTRAINT_IDREF(__set, __rsc, __name) do { \ __rsc = pcmk__find_constraint_resource(data_set->resources, __name); \ if (__rsc == NULL) { \ pcmk__config_err("%s: No resource found for %s", __set, __name); \ return; \ } \ } while(0) // Used to temporarily mark a node as unusable #define INFINITY_HACK (INFINITY * -100) static gint cmp_dependent_priority(gconstpointer a, gconstpointer b) { const pcmk__colocation_t *rsc_constraint1 = (const pcmk__colocation_t *) a; const pcmk__colocation_t *rsc_constraint2 = (const pcmk__colocation_t *) b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } CRM_ASSERT(rsc_constraint1->dependent != NULL); CRM_ASSERT(rsc_constraint1->primary != NULL); if (rsc_constraint1->dependent->priority > rsc_constraint2->dependent->priority) { return -1; } if (rsc_constraint1->dependent->priority < rsc_constraint2->dependent->priority) { return 1; } /* Process clones before primitives and groups */ if (rsc_constraint1->dependent->variant > rsc_constraint2->dependent->variant) { return -1; } if (rsc_constraint1->dependent->variant < rsc_constraint2->dependent->variant) { return 1; } /* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable * clones (probably unnecessary, but avoids having to update regression * tests) */ if (rsc_constraint1->dependent->variant == pe_clone) { if (pcmk_is_set(rsc_constraint1->dependent->flags, pe_rsc_promotable) && !pcmk_is_set(rsc_constraint2->dependent->flags, pe_rsc_promotable)) { return -1; } else if (!pcmk_is_set(rsc_constraint1->dependent->flags, pe_rsc_promotable) && pcmk_is_set(rsc_constraint2->dependent->flags, pe_rsc_promotable)) { return 1; } } return strcmp(rsc_constraint1->dependent->id, rsc_constraint2->dependent->id); } static gint cmp_primary_priority(gconstpointer a, gconstpointer b) { const pcmk__colocation_t *rsc_constraint1 = (const pcmk__colocation_t *) a; const pcmk__colocation_t *rsc_constraint2 = (const pcmk__colocation_t *) b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } CRM_ASSERT(rsc_constraint1->dependent != NULL); CRM_ASSERT(rsc_constraint1->primary != NULL); if (rsc_constraint1->primary->priority > rsc_constraint2->primary->priority) { return -1; } if (rsc_constraint1->primary->priority < rsc_constraint2->primary->priority) { return 1; } /* Process clones before primitives and groups */ if (rsc_constraint1->primary->variant > rsc_constraint2->primary->variant) { return -1; } else if (rsc_constraint1->primary->variant < rsc_constraint2->primary->variant) { return 1; } /* @COMPAT scheduler <2.0.0: Process promotable clones before nonpromotable * clones (probably unnecessary, but avoids having to update regression * tests) */ if (rsc_constraint1->primary->variant == pe_clone) { if (pcmk_is_set(rsc_constraint1->primary->flags, pe_rsc_promotable) && !pcmk_is_set(rsc_constraint2->primary->flags, pe_rsc_promotable)) { return -1; } else if (!pcmk_is_set(rsc_constraint1->primary->flags, pe_rsc_promotable) && pcmk_is_set(rsc_constraint2->primary->flags, pe_rsc_promotable)) { return 1; } } return strcmp(rsc_constraint1->primary->id, rsc_constraint2->primary->id); } /*! * \internal * \brief Add a "this with" colocation constraint to a sorted list * * \param[in,out] list List of constraints to add \p colocation to * \param[in] colocation Colocation constraint to add to \p list * * \note The list will be sorted using cmp_primary_priority(). */ void pcmk__add_this_with(GList **list, const pcmk__colocation_t *colocation) { CRM_ASSERT((list != NULL) && (colocation != NULL)); crm_trace("Adding colocation %s (%s with %s%s%s @%d) " "to 'this with' list", colocation->id, colocation->dependent->id, colocation->primary->id, (colocation->node_attribute == NULL)? "" : " using ", pcmk__s(colocation->node_attribute, ""), colocation->score); *list = g_list_insert_sorted(*list, (gpointer) colocation, cmp_primary_priority); } /*! * \internal * \brief Add a list of "this with" colocation constraints to a list * * \param[in,out] list List of constraints to add \p addition to * \param[in] addition List of colocation constraints to add to \p list * * \note The lists must be pre-sorted by cmp_primary_priority(). */ void pcmk__add_this_with_list(GList **list, GList *addition) { CRM_CHECK((list != NULL), return); if (*list == NULL) { // Trivial case for efficiency crm_trace("Copying %u 'this with' colocations to new list", g_list_length(addition)); *list = g_list_copy(addition); } else { while (addition != NULL) { pcmk__add_this_with(list, addition->data); addition = addition->next; } } } /*! * \internal * \brief Add a "with this" colocation constraint to a sorted list * * \param[in,out] list List of constraints to add \p colocation to * \param[in] colocation Colocation constraint to add to \p list * * \note The list will be sorted using cmp_dependent_priority(). */ void pcmk__add_with_this(GList **list, const pcmk__colocation_t *colocation) { CRM_ASSERT((list != NULL) && (colocation != NULL)); crm_trace("Adding colocation %s (%s with %s%s%s @%d) " "to 'with this' list", colocation->id, colocation->dependent->id, colocation->primary->id, (colocation->node_attribute == NULL)? "" : " using ", pcmk__s(colocation->node_attribute, ""), colocation->score); *list = g_list_insert_sorted(*list, (gpointer) colocation, cmp_dependent_priority); } /*! * \internal * \brief Add a list of "with this" colocation constraints to a list * * \param[in,out] list List of constraints to add \p addition to * \param[in] addition List of colocation constraints to add to \p list * * \note The lists must be pre-sorted by cmp_dependent_priority(). */ void pcmk__add_with_this_list(GList **list, GList *addition) { CRM_CHECK((list != NULL), return); if (*list == NULL) { // Trivial case for efficiency crm_trace("Copying %u 'with this' colocations to new list", g_list_length(addition)); *list = g_list_copy(addition); } else { while (addition != NULL) { pcmk__add_with_this(list, addition->data); addition = addition->next; } } } /*! * \internal * \brief Add orderings necessary for an anti-colocation constraint * * \param[in,out] first_rsc One resource in an anti-colocation * \param[in] first_role Anti-colocation role of \p first_rsc * \param[in] then_rsc Other resource in the anti-colocation * \param[in] then_role Anti-colocation role of \p then_rsc */ static void anti_colocation_order(pe_resource_t *first_rsc, int first_role, pe_resource_t *then_rsc, int then_role) { const char *first_tasks[] = { NULL, NULL }; const char *then_tasks[] = { NULL, NULL }; /* Actions to make first_rsc lose first_role */ if (first_role == RSC_ROLE_PROMOTED) { first_tasks[0] = CRMD_ACTION_DEMOTE; } else { first_tasks[0] = CRMD_ACTION_STOP; if (first_role == RSC_ROLE_UNPROMOTED) { first_tasks[1] = CRMD_ACTION_PROMOTE; } } /* Actions to make then_rsc gain then_role */ if (then_role == RSC_ROLE_PROMOTED) { then_tasks[0] = CRMD_ACTION_PROMOTE; } else { then_tasks[0] = CRMD_ACTION_START; if (then_role == RSC_ROLE_UNPROMOTED) { then_tasks[1] = CRMD_ACTION_DEMOTE; } } for (int first_lpc = 0; (first_lpc <= 1) && (first_tasks[first_lpc] != NULL); first_lpc++) { for (int then_lpc = 0; (then_lpc <= 1) && (then_tasks[then_lpc] != NULL); then_lpc++) { pcmk__order_resource_actions(first_rsc, first_tasks[first_lpc], then_rsc, then_tasks[then_lpc], pe_order_anti_colocation); } } } /*! * \internal * \brief Add a new colocation constraint to a cluster working set * * \param[in] id XML ID for this constraint * \param[in] node_attr Colocate by this attribute (NULL for #uname) * \param[in] score Constraint score * \param[in,out] dependent Resource to be colocated * \param[in,out] primary Resource to colocate \p dependent with * \param[in] dependent_role Current role of \p dependent * \param[in] primary_role Current role of \p primary * \param[in] influence Whether colocation constraint has influence * \param[in,out] data_set Cluster working set to add constraint to */ void pcmk__new_colocation(const char *id, const char *node_attr, int score, pe_resource_t *dependent, pe_resource_t *primary, const char *dependent_role, const char *primary_role, bool influence, pe_working_set_t *data_set) { pcmk__colocation_t *new_con = NULL; if (score == 0) { crm_trace("Ignoring colocation '%s' because score is 0", id); return; } if ((dependent == NULL) || (primary == NULL)) { pcmk__config_err("Ignoring colocation '%s' because resource " "does not exist", id); return; } new_con = calloc(1, sizeof(pcmk__colocation_t)); if (new_con == NULL) { return; } if (pcmk__str_eq(dependent_role, RSC_ROLE_STARTED_S, pcmk__str_null_matches|pcmk__str_casei)) { dependent_role = RSC_ROLE_UNKNOWN_S; } if (pcmk__str_eq(primary_role, RSC_ROLE_STARTED_S, pcmk__str_null_matches|pcmk__str_casei)) { primary_role = RSC_ROLE_UNKNOWN_S; } new_con->id = id; new_con->dependent = dependent; new_con->primary = primary; new_con->score = score; new_con->dependent_role = text2role(dependent_role); new_con->primary_role = text2role(primary_role); new_con->node_attribute = node_attr; new_con->influence = influence; if (node_attr == NULL) { node_attr = CRM_ATTR_UNAME; } pe_rsc_trace(dependent, "%s ==> %s (%s %d)", dependent->id, primary->id, node_attr, score); pcmk__add_this_with(&(dependent->rsc_cons), new_con); pcmk__add_with_this(&(primary->rsc_cons_lhs), new_con); data_set->colocation_constraints = g_list_append(data_set->colocation_constraints, new_con); if (score <= -INFINITY) { anti_colocation_order(dependent, new_con->dependent_role, primary, new_con->primary_role); anti_colocation_order(primary, new_con->primary_role, dependent, new_con->dependent_role); } } /*! * \internal * \brief Return the boolean influence corresponding to configuration * * \param[in] coloc_id Colocation XML ID (for error logging) * \param[in] rsc Resource involved in constraint (for default) * \param[in] influence_s String value of influence option * * \return true if string evaluates true, false if string evaluates false, * or value of resource's critical option if string is NULL or invalid */ static bool unpack_influence(const char *coloc_id, const pe_resource_t *rsc, const char *influence_s) { if (influence_s != NULL) { int influence_i = 0; if (crm_str_to_boolean(influence_s, &influence_i) < 0) { pcmk__config_err("Constraint '%s' has invalid value for " XML_COLOC_ATTR_INFLUENCE " (using default)", coloc_id); } else { return (influence_i != 0); } } return pcmk_is_set(rsc->flags, pe_rsc_critical); } static void unpack_colocation_set(xmlNode *set, int score, const char *coloc_id, const char *influence_s, pe_working_set_t *data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *with = NULL; pe_resource_t *resource = NULL; const char *set_id = ID(set); const char *role = crm_element_value(set, "role"); const char *ordering = crm_element_value(set, "ordering"); int local_score = score; bool sequential = false; const char *score_s = crm_element_value(set, XML_RULE_ATTR_SCORE); if (score_s) { local_score = char2score(score_s); } if (local_score == 0) { crm_trace("Ignoring colocation '%s' for set '%s' because score is 0", coloc_id, set_id); return; } if (ordering == NULL) { ordering = "group"; } if (pcmk__xe_get_bool_attr(set, "sequential", &sequential) == pcmk_rc_ok && !sequential) { return; } else if ((local_score > 0) && pcmk__str_eq(ordering, "group", pcmk__str_casei)) { for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); if (with != NULL) { pe_rsc_trace(resource, "Colocating %s with %s", resource->id, with->id); pcmk__new_colocation(set_id, NULL, local_score, resource, with, role, role, unpack_influence(coloc_id, resource, influence_s), data_set); } with = resource; } } else if (local_score > 0) { pe_resource_t *last = NULL; for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); if (last != NULL) { pe_rsc_trace(resource, "Colocating %s with %s", last->id, resource->id); pcmk__new_colocation(set_id, NULL, local_score, last, resource, role, role, unpack_influence(coloc_id, last, influence_s), data_set); } last = resource; } } else { /* Anti-colocating with every prior resource is * the only way to ensure the intuitive result * (i.e. that no one in the set can run with anyone else in the set) */ for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { xmlNode *xml_rsc_with = NULL; bool influence = true; EXPAND_CONSTRAINT_IDREF(set_id, resource, ID(xml_rsc)); influence = unpack_influence(coloc_id, resource, influence_s); for (xml_rsc_with = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc_with != NULL; xml_rsc_with = crm_next_same_xml(xml_rsc_with)) { if (pcmk__str_eq(resource->id, ID(xml_rsc_with), pcmk__str_casei)) { break; } EXPAND_CONSTRAINT_IDREF(set_id, with, ID(xml_rsc_with)); pe_rsc_trace(resource, "Anti-Colocating %s with %s", resource->id, with->id); pcmk__new_colocation(set_id, NULL, local_score, resource, with, role, role, influence, data_set); } } } } static void colocate_rsc_sets(const char *id, xmlNode *set1, xmlNode *set2, int score, const char *influence_s, pe_working_set_t *data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *rsc_1 = NULL; pe_resource_t *rsc_2 = NULL; const char *role_1 = crm_element_value(set1, "role"); const char *role_2 = crm_element_value(set2, "role"); int rc = pcmk_rc_ok; bool sequential = false; if (score == 0) { crm_trace("Ignoring colocation '%s' between sets because score is 0", id); return; } rc = pcmk__xe_get_bool_attr(set1, "sequential", &sequential); if (rc != pcmk_rc_ok || sequential) { // Get the first one xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); if (xml_rsc != NULL) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); } } rc = pcmk__xe_get_bool_attr(set2, "sequential", &sequential); if (rc != pcmk_rc_ok || sequential) { // Get the last one const char *rid = NULL; for (xml_rsc = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { rid = ID(xml_rsc); } EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid); } if ((rsc_1 != NULL) && (rsc_2 != NULL)) { pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, role_2, unpack_influence(id, rsc_1, influence_s), data_set); } else if (rsc_1 != NULL) { bool influence = unpack_influence(id, rsc_1, influence_s); for (xml_rsc = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, role_2, influence, data_set); } } else if (rsc_2 != NULL) { for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, role_2, unpack_influence(id, rsc_1, influence_s), data_set); } } else { for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { xmlNode *xml_rsc_2 = NULL; bool influence = true; EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); influence = unpack_influence(id, rsc_1, influence_s); for (xml_rsc_2 = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc_2 != NULL; xml_rsc_2 = crm_next_same_xml(xml_rsc_2)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); pcmk__new_colocation(id, NULL, score, rsc_1, rsc_2, role_1, role_2, influence, data_set); } } } } static void unpack_simple_colocation(xmlNode *xml_obj, const char *id, const char *influence_s, pe_working_set_t *data_set) { int score_i = 0; const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); const char *dependent_id = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); const char *primary_id = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET); const char *dependent_role = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); const char *primary_role = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE); const char *attr = crm_element_value(xml_obj, XML_COLOC_ATTR_NODE_ATTR); // @COMPAT: Deprecated since 2.1.5 const char *dependent_instance = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_INSTANCE); // @COMPAT: Deprecated since 2.1.5 const char *primary_instance = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_INSTANCE); pe_resource_t *dependent = pcmk__find_constraint_resource(data_set->resources, dependent_id); pe_resource_t *primary = pcmk__find_constraint_resource(data_set->resources, primary_id); if (dependent_instance != NULL) { pe_warn_once(pe_wo_coloc_inst, "Support for " XML_COLOC_ATTR_SOURCE_INSTANCE " is " "deprecated and will be removed in a future release."); } if (primary_instance != NULL) { pe_warn_once(pe_wo_coloc_inst, "Support for " XML_COLOC_ATTR_TARGET_INSTANCE " is " "deprecated and will be removed in a future release."); } if (dependent == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, dependent_id); return; } else if (primary == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", id, primary_id); return; } else if ((dependent_instance != NULL) && !pe_rsc_is_clone(dependent)) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, dependent_id, dependent_instance); return; } else if ((primary_instance != NULL) && !pe_rsc_is_clone(primary)) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", id, primary_id, primary_instance); return; } if (dependent_instance != NULL) { dependent = find_clone_instance(dependent, dependent_instance); if (dependent == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", id, dependent_id, dependent_instance); return; } } if (primary_instance != NULL) { primary = find_clone_instance(primary, primary_instance); if (primary == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", "'%s'", id, primary_id, primary_instance); return; } } if (pcmk__xe_attr_is_true(xml_obj, XML_CONS_ATTR_SYMMETRICAL)) { pcmk__config_warn("The colocation constraint '" XML_CONS_ATTR_SYMMETRICAL "' attribute has been removed"); } if (score) { score_i = char2score(score); } pcmk__new_colocation(id, attr, score_i, dependent, primary, dependent_role, primary_role, unpack_influence(id, dependent, influence_s), data_set); } // \return Standard Pacemaker return code static int unpack_colocation_tags(xmlNode *xml_obj, xmlNode **expanded_xml, pe_working_set_t *data_set) { const char *id = NULL; const char *dependent_id = NULL; const char *primary_id = NULL; const char *dependent_role = NULL; const char *primary_role = NULL; pe_resource_t *dependent = NULL; pe_resource_t *primary = NULL; pe_tag_t *dependent_tag = NULL; pe_tag_t *primary_tag = NULL; xmlNode *dependent_set = NULL; xmlNode *primary_set = NULL; bool any_sets = false; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return EINVAL); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return pcmk_rc_unpack_error; } // Check whether there are any resource sets with template or tag references *expanded_xml = pcmk__expand_tags_in_sets(xml_obj, data_set); if (*expanded_xml != NULL) { crm_log_xml_trace(*expanded_xml, "Expanded rsc_colocation"); return pcmk_rc_ok; } dependent_id = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE); primary_id = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET); if ((dependent_id == NULL) || (primary_id == NULL)) { return pcmk_rc_ok; } if (!pcmk__valid_resource_or_tag(data_set, dependent_id, &dependent, &dependent_tag)) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, dependent_id); return pcmk_rc_unpack_error; } if (!pcmk__valid_resource_or_tag(data_set, primary_id, &primary, &primary_tag)) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, primary_id); return pcmk_rc_unpack_error; } if ((dependent != NULL) && (primary != NULL)) { /* Neither side references any template/tag. */ return pcmk_rc_ok; } if ((dependent_tag != NULL) && (primary_tag != NULL)) { // A colocation constraint between two templates/tags makes no sense pcmk__config_err("Ignoring constraint '%s' because two templates or " "tags cannot be colocated", id); return pcmk_rc_unpack_error; } dependent_role = crm_element_value(xml_obj, XML_COLOC_ATTR_SOURCE_ROLE); primary_role = crm_element_value(xml_obj, XML_COLOC_ATTR_TARGET_ROLE); *expanded_xml = copy_xml(xml_obj); // Convert template/tag reference in "rsc" into resource_set under constraint if (!pcmk__tag_to_set(*expanded_xml, &dependent_set, XML_COLOC_ATTR_SOURCE, true, data_set)) { free_xml(*expanded_xml); *expanded_xml = NULL; return pcmk_rc_unpack_error; } if (dependent_set != NULL) { if (dependent_role != NULL) { // Move "rsc-role" into converted resource_set as "role" crm_xml_add(dependent_set, "role", dependent_role); xml_remove_prop(*expanded_xml, XML_COLOC_ATTR_SOURCE_ROLE); } any_sets = true; } // Convert template/tag reference in "with-rsc" into resource_set under constraint if (!pcmk__tag_to_set(*expanded_xml, &primary_set, XML_COLOC_ATTR_TARGET, true, data_set)) { free_xml(*expanded_xml); *expanded_xml = NULL; return pcmk_rc_unpack_error; } if (primary_set != NULL) { if (primary_role != NULL) { // Move "with-rsc-role" into converted resource_set as "role" crm_xml_add(primary_set, "role", primary_role); xml_remove_prop(*expanded_xml, XML_COLOC_ATTR_TARGET_ROLE); } any_sets = true; } if (any_sets) { crm_log_xml_trace(*expanded_xml, "Expanded rsc_colocation"); } else { free_xml(*expanded_xml); *expanded_xml = NULL; } return pcmk_rc_ok; } /*! * \internal * \brief Parse a colocation constraint from XML into a cluster working set * * \param[in,out] xml_obj Colocation constraint XML to unpack * \param[in,out] data_set Cluster working set to add constraint to */ void pcmk__unpack_colocation(xmlNode *xml_obj, pe_working_set_t *data_set) { int score_i = 0; xmlNode *set = NULL; xmlNode *last = NULL; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); const char *influence_s = crm_element_value(xml_obj, XML_COLOC_ATTR_INFLUENCE); if (score) { score_i = char2score(score); } if (unpack_colocation_tags(xml_obj, &expanded_xml, data_set) != pcmk_rc_ok) { return; } if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; } for (set = first_named_child(xml_obj, XML_CONS_TAG_RSC_SET); set != NULL; set = crm_next_same_xml(set)) { set = expand_idref(set, data_set->input); if (set == NULL) { // Configuration error, message already logged if (expanded_xml != NULL) { free_xml(expanded_xml); } return; } unpack_colocation_set(set, score_i, id, influence_s, data_set); if (last != NULL) { colocate_rsc_sets(id, last, set, score_i, influence_s, data_set); } last = set; } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } if (last == NULL) { unpack_simple_colocation(xml_obj, id, influence_s, data_set); } } /*! * \internal * \brief Make actions of a given type unrunnable for a given resource * * \param[in,out] rsc Resource whose actions should be blocked * \param[in] task Name of action to block * \param[in] reason Unrunnable start action causing the block */ static void mark_action_blocked(pe_resource_t *rsc, const char *task, const pe_resource_t *reason) { char *reason_text = crm_strdup_printf("colocation with %s", reason->id); for (GList *gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (pcmk_is_set(action->flags, pe_action_runnable) && pcmk__str_eq(action->task, task, pcmk__str_casei)) { pe__clear_action_flags(action, pe_action_runnable); pe_action_set_reason(action, reason_text, false); pcmk__block_colocation_dependents(action, rsc->cluster); pcmk__update_action_for_orderings(action, rsc->cluster); } } // If parent resource can't perform an action, neither can any children for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { mark_action_blocked((pe_resource_t *) (iter->data), task, reason); } free(reason_text); } /*! * \internal * \brief If an action is unrunnable, block any relevant dependent actions * * If a given action is an unrunnable start or promote, block the start or * promote actions of resources colocated with it, as appropriate to the * colocations' configured roles. * * \param[in,out] action Action to check * \param[in] data_set Cluster working set (ignored) */ void pcmk__block_colocation_dependents(pe_action_t *action, pe_working_set_t *data_set) { GList *gIter = NULL; GList *colocations = NULL; pe_resource_t *rsc = NULL; bool is_start = false; if (pcmk_is_set(action->flags, pe_action_runnable)) { return; // Only unrunnable actions block dependents } is_start = pcmk__str_eq(action->task, RSC_START, pcmk__str_none); if (!is_start && !pcmk__str_eq(action->task, RSC_PROMOTE, pcmk__str_none)) { return; // Only unrunnable starts and promotes block dependents } CRM_ASSERT(action->rsc != NULL); // Start and promote are resource actions /* If this resource is part of a collective resource, dependents are blocked * only if all instances of the collective are unrunnable, so check the * collective resource. */ rsc = uber_parent(action->rsc); if (rsc->parent != NULL) { rsc = rsc->parent; // Bundle } // Colocation fails only if entire primary can't reach desired role for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; pe_action_t *child_action = find_first_action(child->actions, NULL, action->task, NULL); if ((child_action == NULL) || pcmk_is_set(child_action->flags, pe_action_runnable)) { crm_trace("Not blocking %s colocation dependents because " "at least %s has runnable %s", rsc->id, child->id, action->task); return; // At least one child can reach desired role } } crm_trace("Blocking %s colocation dependents due to unrunnable %s %s", rsc->id, action->rsc->id, action->task); // Check each colocation where this resource is primary colocations = pcmk__with_this_colocations(rsc); for (gIter = colocations; gIter != NULL; gIter = gIter->next) { pcmk__colocation_t *colocation = (pcmk__colocation_t *) gIter->data; if (colocation->score < INFINITY) { continue; // Only mandatory colocations block dependent } /* If the primary can't start, the dependent can't reach its colocated * role, regardless of what the primary or dependent colocation role is. * * If the primary can't be promoted, the dependent can't reach its * colocated role if the primary's colocation role is promoted. */ if (!is_start && (colocation->primary_role != RSC_ROLE_PROMOTED)) { continue; } // Block the dependent from reaching its colocated role if (colocation->dependent_role == RSC_ROLE_PROMOTED) { mark_action_blocked(colocation->dependent, RSC_PROMOTE, action->rsc); } else { mark_action_blocked(colocation->dependent, RSC_START, action->rsc); } } g_list_free(colocations); } /*! * \internal * \brief Determine how a colocation constraint should affect a resource * * Colocation constraints have different effects at different points in the * scheduler sequence. Initially, they affect a resource's location; once that * is determined, then for promotable clones they can affect a resource * instance's role; after both are determined, the constraints no longer matter. * Given a specific colocation constraint, check what has been done so far to * determine what should be affected at the current point in the scheduler. * * \param[in] dependent Dependent resource in colocation * \param[in] primary Primary resource in colocation * \param[in] colocation Colocation constraint - * \param[in] preview If true, pretend resources have already been allocated + * \param[in] preview If true, pretend resources have already been assigned * * \return How colocation constraint should be applied at this point */ enum pcmk__coloc_affects pcmk__colocation_affects(const pe_resource_t *dependent, const pe_resource_t *primary, const pcmk__colocation_t *colocation, bool preview) { if (!preview && pcmk_is_set(primary->flags, pe_rsc_provisional)) { - // Primary resource has not been allocated yet, so we can't do anything + // Primary resource has not been assigned yet, so we can't do anything return pcmk__coloc_affects_nothing; } if ((colocation->dependent_role >= RSC_ROLE_UNPROMOTED) && (dependent->parent != NULL) && pcmk_is_set(dependent->parent->flags, pe_rsc_promotable) && !pcmk_is_set(dependent->flags, pe_rsc_provisional)) { /* This is a colocation by role, and the dependent is a promotable clone - * that has already been allocated, so the colocation should now affect + * that has already been assigned, so the colocation should now affect * the role. */ return pcmk__coloc_affects_role; } if (!preview && !pcmk_is_set(dependent->flags, pe_rsc_provisional)) { - /* The dependent resource has already been through allocation, so the + /* The dependent resource has already been through assignment, so the * constraint no longer has any effect. Log an error if a mandatory * colocation constraint has been violated. */ const pe_node_t *primary_node = primary->allocated_to; if (dependent->allocated_to == NULL) { crm_trace("Skipping colocation '%s': %s will not run anywhere", colocation->id, dependent->id); } else if (colocation->score >= INFINITY) { // Dependent resource must colocate with primary resource - if ((primary_node == NULL) || - (primary_node->details != dependent->allocated_to->details)) { + if (!pe__same_node(primary_node, dependent->allocated_to)) { crm_err("%s must be colocated with %s but is not (%s vs. %s)", dependent->id, primary->id, pe__node_name(dependent->allocated_to), pe__node_name(primary_node)); } } else if (colocation->score <= -CRM_SCORE_INFINITY) { // Dependent resource must anti-colocate with primary resource - if ((primary_node != NULL) && - (dependent->allocated_to->details == primary_node->details)) { - crm_err("%s and %s must be anti-colocated but are allocated " + if (pe__same_node(dependent->allocated_to, primary_node)) { + crm_err("%s and %s must be anti-colocated but are assigned " "to the same node (%s)", dependent->id, primary->id, pe__node_name(primary_node)); } } return pcmk__coloc_affects_nothing; } if ((colocation->score > 0) && (colocation->dependent_role != RSC_ROLE_UNKNOWN) && (colocation->dependent_role != dependent->next_role)) { crm_trace("Skipping colocation '%s': dependent limited to %s role " "but %s next role is %s", colocation->id, role2text(colocation->dependent_role), dependent->id, role2text(dependent->next_role)); return pcmk__coloc_affects_nothing; } if ((colocation->score > 0) && (colocation->primary_role != RSC_ROLE_UNKNOWN) && (colocation->primary_role != primary->next_role)) { crm_trace("Skipping colocation '%s': primary limited to %s role " "but %s next role is %s", colocation->id, role2text(colocation->primary_role), primary->id, role2text(primary->next_role)); return pcmk__coloc_affects_nothing; } if ((colocation->score < 0) && (colocation->dependent_role != RSC_ROLE_UNKNOWN) && (colocation->dependent_role == dependent->next_role)) { crm_trace("Skipping anti-colocation '%s': dependent role %s matches", colocation->id, role2text(colocation->dependent_role)); return pcmk__coloc_affects_nothing; } if ((colocation->score < 0) && (colocation->primary_role != RSC_ROLE_UNKNOWN) && (colocation->primary_role == primary->next_role)) { crm_trace("Skipping anti-colocation '%s': primary role %s matches", colocation->id, role2text(colocation->primary_role)); return pcmk__coloc_affects_nothing; } return pcmk__coloc_affects_location; } /*! * \internal - * \brief Apply colocation to dependent for allocation purposes + * \brief Apply colocation to dependent for assignment purposes * - * Update the allowed node weights of the dependent resource in a colocation, - * for the purposes of allocating it to a node + * Update the allowed node scores of the dependent resource in a colocation, + * for the purposes of assigning it to a node. * * \param[in,out] dependent Dependent resource in colocation * \param[in] primary Primary resource in colocation * \param[in] colocation Colocation constraint */ void -pcmk__apply_coloc_to_weights(pe_resource_t *dependent, - const pe_resource_t *primary, - const pcmk__colocation_t *colocation) +pcmk__apply_coloc_to_scores(pe_resource_t *dependent, + const pe_resource_t *primary, + const pcmk__colocation_t *colocation) { const char *attribute = CRM_ATTR_ID; const char *value = NULL; GHashTable *work = NULL; GHashTableIter iter; pe_node_t *node = NULL; if (colocation->node_attribute != NULL) { attribute = colocation->node_attribute; } if (primary->allocated_to != NULL) { value = pe_node_attribute_raw(primary->allocated_to, attribute); } else if (colocation->score < 0) { // Nothing to do (anti-colocation with something that is not running) return; } work = pcmk__copy_node_table(dependent->allowed_nodes); g_hash_table_iter_init(&iter, work); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (primary->allocated_to == NULL) { node->weight = pcmk__add_scores(-colocation->score, node->weight); pe_rsc_trace(dependent, "Applied %s to %s score on %s (now %s after " "subtracting %s because primary %s inactive)", colocation->id, dependent->id, pe__node_name(node), pcmk_readable_score(node->weight), pcmk_readable_score(colocation->score), primary->id); } else if (pcmk__str_eq(pe_node_attribute_raw(node, attribute), value, pcmk__str_casei)) { /* Add colocation score only if optional (or minus infinity). A * mandatory colocation is a requirement rather than a preference, * so we don't need to consider it for relative assignment purposes. * The resource will simply be forbidden from running on the node if * the primary isn't active there (via the condition above). */ if (colocation->score < CRM_SCORE_INFINITY) { node->weight = pcmk__add_scores(colocation->score, node->weight); pe_rsc_trace(dependent, "Applied %s to %s score on %s (now %s after " "adding %s)", colocation->id, dependent->id, pe__node_name(node), pcmk_readable_score(node->weight), pcmk_readable_score(colocation->score)); } } else if (colocation->score >= CRM_SCORE_INFINITY) { /* Only mandatory colocations are relevant when the colocation * attribute doesn't match, because an attribute not matching is not * a negative preference -- the colocation is simply relevant only * where it matches. */ node->weight = -CRM_SCORE_INFINITY; pe_rsc_trace(dependent, "Banned %s from %s because colocation %s attribute %s " "does not match", dependent->id, pe__node_name(node), colocation->id, attribute); } } if ((colocation->score <= -INFINITY) || (colocation->score >= INFINITY) || pcmk__any_node_available(work)) { g_hash_table_destroy(dependent->allowed_nodes); dependent->allowed_nodes = work; work = NULL; } else { pe_rsc_info(dependent, "%s: Rolling back scores from %s (no available nodes)", dependent->id, primary->id); } if (work != NULL) { g_hash_table_destroy(work); } } /*! * \internal * \brief Apply colocation to dependent for role purposes * * Update the priority of the dependent resource in a colocation, for the * purposes of selecting its role * * \param[in,out] dependent Dependent resource in colocation * \param[in] primary Primary resource in colocation * \param[in] colocation Colocation constraint */ void pcmk__apply_coloc_to_priority(pe_resource_t *dependent, const pe_resource_t *primary, const pcmk__colocation_t *colocation) { const char *dependent_value = NULL; const char *primary_value = NULL; const char *attribute = CRM_ATTR_ID; int score_multiplier = 1; if ((primary->allocated_to == NULL) || (dependent->allocated_to == NULL)) { return; } if (colocation->node_attribute != NULL) { attribute = colocation->node_attribute; } dependent_value = pe_node_attribute_raw(dependent->allocated_to, attribute); primary_value = pe_node_attribute_raw(primary->allocated_to, attribute); if (!pcmk__str_eq(dependent_value, primary_value, pcmk__str_casei)) { if ((colocation->score == INFINITY) && (colocation->dependent_role == RSC_ROLE_PROMOTED)) { dependent->priority = -INFINITY; } return; } if ((colocation->primary_role != RSC_ROLE_UNKNOWN) && (colocation->primary_role != primary->next_role)) { return; } if (colocation->dependent_role == RSC_ROLE_UNPROMOTED) { score_multiplier = -1; } dependent->priority = pcmk__add_scores(score_multiplier * colocation->score, dependent->priority); pe_rsc_trace(dependent, "Applied %s to %s promotion priority (now %s after %s %s)", colocation->id, dependent->id, pcmk_readable_score(dependent->priority), ((score_multiplier == 1)? "adding" : "subtracting"), pcmk_readable_score(colocation->score)); } /*! * \internal * \brief Find score of highest-scored node that matches colocation attribute * * \param[in] rsc Resource whose allowed nodes should be searched * \param[in] attr Colocation attribute name (must not be NULL) * \param[in] value Colocation attribute value to require */ static int best_node_score_matching_attr(const pe_resource_t *rsc, const char *attr, const char *value) { GHashTableIter iter; pe_node_t *node = NULL; int best_score = -INFINITY; const char *best_node = NULL; // Find best allowed node with matching attribute g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if ((node->weight > best_score) && pcmk__node_available(node, false, false) && pcmk__str_eq(value, pe_node_attribute_raw(node, attr), pcmk__str_casei)) { best_score = node->weight; best_node = node->details->uname; } } if (!pcmk__str_eq(attr, CRM_ATTR_UNAME, pcmk__str_casei)) { if (best_node == NULL) { crm_info("No allowed node for %s matches node attribute %s=%s", rsc->id, attr, value); } else { crm_info("Allowed node %s for %s had best score (%d) " "of those matching node attribute %s=%s", best_node, rsc->id, best_score, attr, value); } } return best_score; } /*! * \internal * \brief Check whether a resource is allowed only on a single node * * \param[in] rsc Resource to check * * \return \c true if \p rsc is allowed only on one node, otherwise \c false */ static bool allowed_on_one(const pe_resource_t *rsc) { GHashTableIter iter; pe_node_t *allowed_node = NULL; int allowed_nodes = 0; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &allowed_node)) { if ((allowed_node->weight >= 0) && (++allowed_nodes > 1)) { pe_rsc_trace(rsc, "%s is allowed on multiple nodes", rsc->id); return false; } } pe_rsc_trace(rsc, "%s is allowed %s", rsc->id, ((allowed_nodes == 1)? "on a single node" : "nowhere")); return (allowed_nodes == 1); } /*! * \internal - * \brief Add resource's colocation matches to current node allocation scores + * \brief Add resource's colocation matches to current node assignment scores * * For each node in a given table, if any of a given resource's allowed nodes * have a matching value for the colocation attribute, add the highest of those * nodes' scores to the node's score. * * \param[in,out] nodes Table of nodes with assignment scores so far * \param[in] rsc Resource whose allowed nodes should be compared * \param[in] colocation Original colocation constraint (used to get * configured primary resource's stickiness, and * to get colocation node attribute; pass NULL to * ignore stickiness and use default attribute) * \param[in] factor Factor by which to multiply scores being added * \param[in] only_positive Whether to add only positive scores */ static void add_node_scores_matching_attr(GHashTable *nodes, const pe_resource_t *rsc, pcmk__colocation_t *colocation, float factor, bool only_positive) { GHashTableIter iter; pe_node_t *node = NULL; const char *attr = CRM_ATTR_UNAME; if ((colocation != NULL) && (colocation->node_attribute != NULL)) { attr = colocation->node_attribute; } // Iterate through each node g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { - float weight_f = 0; - int weight = 0; + float delta_f = 0; + int delta = 0; int score = 0; int new_score = 0; const char *value = pe_node_attribute_raw(node, attr); score = best_node_score_matching_attr(rsc, attr, value); if ((factor < 0) && (score < 0)) { /* If the dependent is anti-colocated, we generally don't want the * primary to prefer nodes that the dependent avoids. That could * lead to unnecessary shuffling of the primary when the dependent * hits its migration threshold somewhere, for example. * * However, there are cases when it is desirable. If the dependent * can't run anywhere but where the primary is, it would be * worthwhile to move the primary for the sake of keeping the * dependent active. * * We can't know that exactly at this point since we don't know * where the primary will be assigned, but we can limit considering * the preference to when the dependent is allowed only on one node. * This is less than ideal for multiple reasons: * * - the dependent could be allowed on more than one node but have * anti-colocation primaries on each; * - the dependent could be a clone or bundle with multiple * instances, and the dependent as a whole is allowed on multiple * nodes but some instance still can't run * - the dependent has considered node-specific criteria such as * location constraints and stickiness by this point, but might * have other factors that end up disallowing a node * * but the alternative is making the primary move when it doesn't * need to. * * We also consider the primary's stickiness and influence, so the * user has some say in the matter. (This is the configured primary, * not a particular instance of the primary, but that doesn't matter * unless stickiness uses a rule to vary by node, and that seems * acceptable to ignore.) */ if ((colocation == NULL) || (colocation->primary->stickiness >= -score) || !pcmk__colocation_has_influence(colocation, NULL) || !allowed_on_one(colocation->dependent)) { crm_trace("%s: Filtering %d + %f * %d " "(double negative disallowed)", pe__node_name(node), node->weight, factor, score); continue; } } if (node->weight == INFINITY_HACK) { crm_trace("%s: Filtering %d + %f * %d (node was marked unusable)", pe__node_name(node), node->weight, factor, score); continue; } - weight_f = factor * score; + delta_f = factor * score; // Round the number; see http://c-faq.com/fp/round.html - weight = (int) ((weight_f < 0)? (weight_f - 0.5) : (weight_f + 0.5)); + delta = (int) ((delta_f < 0)? (delta_f - 0.5) : (delta_f + 0.5)); /* Small factors can obliterate the small scores that are often actually * used in configurations. If the score and factor are nonzero, ensure * that the result is nonzero as well. */ - if ((weight == 0) && (score != 0)) { + if ((delta == 0) && (score != 0)) { if (factor > 0.0) { - weight = 1; + delta = 1; } else if (factor < 0.0) { - weight = -1; + delta = -1; } } - new_score = pcmk__add_scores(weight, node->weight); + new_score = pcmk__add_scores(delta, node->weight); if (only_positive && (new_score < 0) && (node->weight > 0)) { crm_trace("%s: Filtering %d + %f * %d = %d " "(negative disallowed, marking node unusable)", pe__node_name(node), node->weight, factor, score, new_score); node->weight = INFINITY_HACK; continue; } if (only_positive && (new_score < 0) && (node->weight == 0)) { crm_trace("%s: Filtering %d + %f * %d = %d (negative disallowed)", pe__node_name(node), node->weight, factor, score, new_score); continue; } crm_trace("%s: %d + %f * %d = %d", pe__node_name(node), node->weight, factor, score, new_score); node->weight = new_score; } } /*! * \internal * \brief Update nodes with scores of colocated resources' nodes * * Given a table of nodes and a resource, update the nodes' scores with the * scores of the best nodes matching the attribute used for each of the * resource's relevant colocations. * * \param[in,out] rsc Resource to check colocations for * \param[in] log_id Resource ID for logs (if NULL, use \p rsc ID) * \param[in,out] nodes Nodes to update (set initial contents to NULL * to copy \p rsc's allowed nodes) * \param[in] colocation Original colocation constraint (used to get * configured primary resource's stickiness, and * to get colocation node attribute; if NULL, * \p rsc's own matching node scores will not be * added, and *nodes must be NULL as well) * \param[in] factor Incorporate scores multiplied by this factor * \param[in] flags Bitmask of enum pcmk__coloc_select values * * \note NULL *nodes, NULL colocation, and the pcmk__coloc_select_this_with * flag are used together (and only by cmp_resources()). * \note The caller remains responsible for freeing \p *nodes. */ void pcmk__add_colocated_node_scores(pe_resource_t *rsc, const char *log_id, GHashTable **nodes, pcmk__colocation_t *colocation, float factor, uint32_t flags) { GHashTable *work = NULL; CRM_ASSERT((rsc != NULL) && (nodes != NULL) && ((colocation != NULL) || (*nodes == NULL))); if (log_id == NULL) { log_id = rsc->id; } // Avoid infinite recursion if (pcmk_is_set(rsc->flags, pe_rsc_merging)) { pe_rsc_info(rsc, "%s: Breaking dependency loop at %s", log_id, rsc->id); return; } pe__set_resource_flags(rsc, pe_rsc_merging); if (*nodes == NULL) { work = pcmk__copy_node_table(rsc->allowed_nodes); } else { pe_rsc_trace(rsc, "%s: Merging scores from %s (at %.6f)", log_id, rsc->id, factor); work = pcmk__copy_node_table(*nodes); add_node_scores_matching_attr(work, rsc, colocation, factor, pcmk_is_set(flags, pcmk__coloc_select_nonnegative)); } if (work == NULL) { pe__clear_resource_flags(rsc, pe_rsc_merging); return; } if (pcmk__any_node_available(work)) { GList *colocations = NULL; if (pcmk_is_set(flags, pcmk__coloc_select_this_with)) { colocations = pcmk__this_with_colocations(rsc); pe_rsc_trace(rsc, "Checking additional %d optional '%s with' constraints", g_list_length(colocations), rsc->id); } else { colocations = pcmk__with_this_colocations(rsc); pe_rsc_trace(rsc, "Checking additional %d optional 'with %s' constraints", g_list_length(colocations), rsc->id); } flags |= pcmk__coloc_select_active; for (GList *iter = colocations; iter != NULL; iter = iter->next) { pcmk__colocation_t *constraint = (pcmk__colocation_t *) iter->data; pe_resource_t *other = NULL; float other_factor = factor * constraint->score / (float) INFINITY; if (pcmk_is_set(flags, pcmk__coloc_select_this_with)) { other = constraint->primary; } else if (!pcmk__colocation_has_influence(constraint, NULL)) { continue; } else { other = constraint->dependent; } pe_rsc_trace(rsc, "Optionally merging score of '%s' constraint (%s with %s)", constraint->id, constraint->dependent->id, constraint->primary->id); other->cmds->add_colocated_node_scores(other, log_id, &work, constraint, other_factor, flags); - pe__show_node_weights(true, NULL, log_id, work, rsc->cluster); + pe__show_node_scores(true, NULL, log_id, work, rsc->cluster); } g_list_free(colocations); } else if (pcmk_is_set(flags, pcmk__coloc_select_active)) { pe_rsc_info(rsc, "%s: Rolling back optional scores from %s", log_id, rsc->id); g_hash_table_destroy(work); pe__clear_resource_flags(rsc, pe_rsc_merging); return; } if (pcmk_is_set(flags, pcmk__coloc_select_nonnegative)) { pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, work); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (node->weight == INFINITY_HACK) { node->weight = 1; } } } if (*nodes != NULL) { g_hash_table_destroy(*nodes); } *nodes = work; pe__clear_resource_flags(rsc, pe_rsc_merging); } /*! * \internal * \brief Apply a "with this" colocation to a resource's allowed node scores * * \param[in,out] data Colocation to apply * \param[in,out] user_data Resource being assigned */ void pcmk__add_dependent_scores(gpointer data, gpointer user_data) { pcmk__colocation_t *colocation = (pcmk__colocation_t *) data; pe_resource_t *rsc = (pe_resource_t *) user_data; pe_resource_t *other = colocation->dependent; const float factor = colocation->score / (float) INFINITY; uint32_t flags = pcmk__coloc_select_active; if (!pcmk__colocation_has_influence(colocation, NULL)) { return; } if (rsc->variant == pe_clone) { flags |= pcmk__coloc_select_nonnegative; } pe_rsc_trace(rsc, "%s: Incorporating attenuated %s assignment scores due " "to colocation %s", rsc->id, other->id, colocation->id); other->cmds->add_colocated_node_scores(other, rsc->id, &rsc->allowed_nodes, colocation, factor, flags); } /*! * \internal * \brief Get all colocations affecting a resource as the primary * * \param[in] rsc Resource to get colocations for * * \return Newly allocated list of colocations affecting \p rsc as primary * * \note This is a convenience wrapper for the with_this_colocations() method. */ GList * pcmk__with_this_colocations(const pe_resource_t *rsc) { GList *list = NULL; rsc->cmds->with_this_colocations(rsc, rsc, &list); return list; } /*! * \internal * \brief Get all colocations affecting a resource as the dependent * * \param[in] rsc Resource to get colocations for * * \return Newly allocated list of colocations affecting \p rsc as dependent * * \note This is a convenience wrapper for the this_with_colocations() method. */ GList * pcmk__this_with_colocations(const pe_resource_t *rsc) { GList *list = NULL; rsc->cmds->this_with_colocations(rsc, rsc, &list); return list; } diff --git a/lib/pacemaker/pcmk_sched_group.c b/lib/pacemaker/pcmk_sched_group.c index f284f34d86..7de33e4a4e 100644 --- a/lib/pacemaker/pcmk_sched_group.c +++ b/lib/pacemaker/pcmk_sched_group.c @@ -1,870 +1,870 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Assign a group resource to a node * * \param[in,out] rsc Group resource to assign to a node * \param[in] prefer Node to prefer, if all else is equal * * \return Node that \p rsc is assigned to, if assigned entirely to one node */ pe_node_t * pcmk__group_assign(pe_resource_t *rsc, const pe_node_t *prefer) { pe_node_t *first_assigned_node = NULL; pe_resource_t *first_member = NULL; CRM_ASSERT(rsc != NULL); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->allocated_to; // Assignment already done } if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Assignment dependency loop detected involving %s", rsc->id); return NULL; } if (rsc->children == NULL) { // No members to assign pe__clear_resource_flags(rsc, pe_rsc_provisional); return NULL; } pe__set_resource_flags(rsc, pe_rsc_allocating); first_member = (pe_resource_t *) rsc->children->data; rsc->role = first_member->role; - pe__show_node_weights(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), - rsc, __func__, rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), + rsc, __func__, rsc->allowed_nodes, rsc->cluster); for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; pe_node_t *node = NULL; pe_rsc_trace(rsc, "Assigning group %s member %s", rsc->id, member->id); node = member->cmds->assign(member, prefer); if (first_assigned_node == NULL) { first_assigned_node = node; } } pe__set_next_role(rsc, first_member->next_role, "first group member"); pe__clear_resource_flags(rsc, pe_rsc_allocating|pe_rsc_provisional); if (!pe__group_flag_is_set(rsc, pe__group_colocated)) { return NULL; } return first_assigned_node; } /*! * \internal * \brief Create a pseudo-operation for a group as an ordering point * * \param[in,out] group Group resource to create action for * \param[in] action Action name * * \return Newly created pseudo-operation */ static pe_action_t * create_group_pseudo_op(pe_resource_t *group, const char *action) { pe_action_t *op = custom_action(group, pcmk__op_key(group->id, action, 0), action, NULL, TRUE, TRUE, group->cluster); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); return op; } /*! * \internal * \brief Create all actions needed for a given group resource * * \param[in,out] rsc Group resource to create actions for */ void pcmk__group_create_actions(pe_resource_t *rsc) { CRM_ASSERT(rsc != NULL); pe_rsc_trace(rsc, "Creating actions for group %s", rsc->id); // Create actions for individual group members for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; member->cmds->create_actions(member); } // Create pseudo-actions for group itself to serve as ordering points create_group_pseudo_op(rsc, RSC_START); create_group_pseudo_op(rsc, RSC_STARTED); create_group_pseudo_op(rsc, RSC_STOP); create_group_pseudo_op(rsc, RSC_STOPPED); if (crm_is_true(g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_PROMOTABLE))) { create_group_pseudo_op(rsc, RSC_DEMOTE); create_group_pseudo_op(rsc, RSC_DEMOTED); create_group_pseudo_op(rsc, RSC_PROMOTE); create_group_pseudo_op(rsc, RSC_PROMOTED); } } // User data for member_internal_constraints() struct member_data { // These could be derived from member but this avoids some function calls bool ordered; bool colocated; bool promotable; pe_resource_t *last_active; pe_resource_t *previous_member; }; /*! * \internal * \brief Create implicit constraints needed for a group member * * \param[in,out] data Group member to create implicit constraints for * \param[in,out] user_data Member data (struct member_data *) */ static void member_internal_constraints(gpointer data, gpointer user_data) { pe_resource_t *member = (pe_resource_t *) data; struct member_data *member_data = (struct member_data *) user_data; // For ordering demote vs demote or stop vs stop uint32_t down_flags = pe_order_implies_first_printed; // For ordering demote vs demoted or stop vs stopped uint32_t post_down_flags = pe_order_implies_then_printed; // Create the individual member's implicit constraints member->cmds->internal_constraints(member); if (member_data->previous_member == NULL) { // This is first member if (member_data->ordered) { pe__set_order_flags(down_flags, pe_order_optional); post_down_flags = pe_order_implies_then; } } else if (member_data->colocated) { // Colocate this member with the previous one pcmk__new_colocation("group:internal_colocation", NULL, INFINITY, member, member_data->previous_member, NULL, NULL, pcmk_is_set(member->flags, pe_rsc_critical), member->cluster); } if (member_data->promotable) { // Demote group -> demote member -> group is demoted pcmk__order_resource_actions(member->parent, RSC_DEMOTE, member, RSC_DEMOTE, down_flags); pcmk__order_resource_actions(member, RSC_DEMOTE, member->parent, RSC_DEMOTED, post_down_flags); // Promote group -> promote member -> group is promoted pcmk__order_resource_actions(member, RSC_PROMOTE, member->parent, RSC_PROMOTED, pe_order_runnable_left |pe_order_implies_then |pe_order_implies_then_printed); pcmk__order_resource_actions(member->parent, RSC_PROMOTE, member, RSC_PROMOTE, pe_order_implies_first_printed); } // Stop group -> stop member -> group is stopped pcmk__order_stops(member->parent, member, down_flags); pcmk__order_resource_actions(member, RSC_STOP, member->parent, RSC_STOPPED, post_down_flags); // Start group -> start member -> group is started pcmk__order_starts(member->parent, member, pe_order_implies_first_printed); pcmk__order_resource_actions(member, RSC_START, member->parent, RSC_STARTED, pe_order_runnable_left |pe_order_implies_then |pe_order_implies_then_printed); if (!member_data->ordered) { pcmk__order_starts(member->parent, member, pe_order_implies_then |pe_order_runnable_left |pe_order_implies_first_printed); if (member_data->promotable) { pcmk__order_resource_actions(member->parent, RSC_PROMOTE, member, RSC_PROMOTE, pe_order_implies_then |pe_order_runnable_left |pe_order_implies_first_printed); } } else if (member_data->previous_member == NULL) { pcmk__order_starts(member->parent, member, pe_order_none); if (member_data->promotable) { pcmk__order_resource_actions(member->parent, RSC_PROMOTE, member, RSC_PROMOTE, pe_order_none); } } else { // Order this member relative to the previous one pcmk__order_starts(member_data->previous_member, member, pe_order_implies_then|pe_order_runnable_left); pcmk__order_stops(member, member_data->previous_member, pe_order_optional|pe_order_restart); /* In unusual circumstances (such as adding a new member to the middle * of a group with unmanaged later members), this member may be active * while the previous (new) member is inactive. In this situation, the * usual restart orderings will be irrelevant, so we need to order this * member's stop before the previous member's start. */ if ((member->running_on != NULL) && (member_data->previous_member->running_on == NULL)) { pcmk__order_resource_actions(member, RSC_STOP, member_data->previous_member, RSC_START, pe_order_implies_first |pe_order_runnable_left); } if (member_data->promotable) { pcmk__order_resource_actions(member_data->previous_member, RSC_PROMOTE, member, RSC_PROMOTE, pe_order_implies_then |pe_order_runnable_left); pcmk__order_resource_actions(member, RSC_DEMOTE, member_data->previous_member, RSC_DEMOTE, pe_order_optional); } } // Make sure partially active groups shut down in sequence if (member->running_on != NULL) { if (member_data->ordered && (member_data->previous_member != NULL) && (member_data->previous_member->running_on == NULL) && (member_data->last_active != NULL) && (member_data->last_active->running_on != NULL)) { pcmk__order_stops(member, member_data->last_active, pe_order_optional); } member_data->last_active = member; } member_data->previous_member = member; } /*! * \internal * \brief Create implicit constraints needed for a group resource * * \param[in,out] rsc Group resource to create implicit constraints for */ void pcmk__group_internal_constraints(pe_resource_t *rsc) { struct member_data member_data = { false, }; CRM_ASSERT(rsc != NULL); /* Order group pseudo-actions relative to each other for restarting: * stop group -> group is stopped -> start group -> group is started */ pcmk__order_resource_actions(rsc, RSC_STOP, rsc, RSC_STOPPED, pe_order_runnable_left); pcmk__order_resource_actions(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_START, rsc, RSC_STARTED, pe_order_runnable_left); member_data.ordered = pe__group_flag_is_set(rsc, pe__group_ordered); member_data.colocated = pe__group_flag_is_set(rsc, pe__group_colocated); member_data.promotable = pcmk_is_set(pe__const_top_resource(rsc, false)->flags, pe_rsc_promotable); g_list_foreach(rsc->children, member_internal_constraints, &member_data); } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint for a group with some other resource, apply the - * score to the dependent's allowed node weights (if we are still placing + * score to the dependent's allowed node scores (if we are still placing * resources) or priority (if we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent group resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply */ static void colocate_group_with(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation) { pe_resource_t *member = NULL; if (dependent->children == NULL) { return; } pe_rsc_trace(primary, "Processing %s (group %s with %s) for dependent", colocation->id, dependent->id, primary->id); if (pe__group_flag_is_set(dependent, pe__group_colocated)) { // Colocate first member (internal colocations will handle the rest) member = (pe_resource_t *) dependent->children->data; member->cmds->apply_coloc_score(member, primary, colocation, true); return; } if (colocation->score >= INFINITY) { pcmk__config_err("%s: Cannot perform mandatory colocation between " "non-colocated group and %s", dependent->id, primary->id); return; } // Colocate each member individually for (GList *iter = dependent->children; iter != NULL; iter = iter->next) { member = (pe_resource_t *) iter->data; member->cmds->apply_coloc_score(member, primary, colocation, true); } } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint for some other resource with a group, apply the - * score to the dependent's allowed node weights (if we are still placing + * score to the dependent's allowed node scores (if we are still placing * resources) or priority (if we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in] primary Primary group resource in colocation * \param[in] colocation Colocation constraint to apply */ static void colocate_with_group(pe_resource_t *dependent, const pe_resource_t *primary, const pcmk__colocation_t *colocation) { pe_resource_t *member = NULL; pe_rsc_trace(primary, "Processing colocation %s (%s with group %s) for primary", colocation->id, dependent->id, primary->id); if (pcmk_is_set(primary->flags, pe_rsc_provisional)) { return; } if (pe__group_flag_is_set(primary, pe__group_colocated)) { if (colocation->score >= INFINITY) { /* For mandatory colocations, the entire group must be assignable * (and in the specified role if any), so apply the colocation based * on the last member. */ member = pe__last_group_member(primary); } else if (primary->children != NULL) { /* For optional colocations, whether the group is partially or fully * up doesn't matter, so apply the colocation based on the first * member. */ member = (pe_resource_t *) primary->children->data; } if (member == NULL) { return; // Nothing to colocate with } member->cmds->apply_coloc_score(dependent, member, colocation, false); return; } if (colocation->score >= INFINITY) { pcmk__config_err("%s: Cannot perform mandatory colocation with" " non-colocated group %s", dependent->id, primary->id); return; } // Colocate dependent with each member individually for (GList *iter = primary->children; iter != NULL; iter = iter->next) { member = (pe_resource_t *) iter->data; member->cmds->apply_coloc_score(dependent, member, colocation, false); } } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint, apply its score to the dependent's - * allowed node weights (if we are still placing resources) or priority (if + * allowed node scores (if we are still placing resources) or priority (if * we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply * \param[in] for_dependent true if called on behalf of dependent */ void pcmk__group_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent) { CRM_ASSERT((dependent != NULL) && (primary != NULL) && (colocation != NULL)); if (for_dependent) { colocate_group_with(dependent, primary, colocation); } else { // Method should only be called for primitive dependents CRM_ASSERT(dependent->variant == pe_native); colocate_with_group(dependent, primary, colocation); } } /*! * \internal * \brief Return action flags for a given group resource action * * \param[in,out] action Group action to get flags for * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__group_action_flags(pe_action_t *action, const pe_node_t *node) { // Default flags for a group action uint32_t flags = pe_action_optional|pe_action_runnable|pe_action_pseudo; CRM_ASSERT(action != NULL); // Update flags considering each member's own flags for same action for (GList *iter = action->rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; // Check whether member has the same action enum action_tasks task = get_complex_task(member, action->task); const char *task_s = task2text(task); pe_action_t *member_action = find_first_action(member->actions, NULL, task_s, node); if (member_action != NULL) { uint32_t member_flags = member->cmds->action_flags(member_action, node); // Group action is mandatory if any member action is if (pcmk_is_set(flags, pe_action_optional) && !pcmk_is_set(member_flags, pe_action_optional)) { pe_rsc_trace(action->rsc, "%s is mandatory because %s is", action->uuid, member_action->uuid); pe__clear_raw_action_flags(flags, "group action", pe_action_optional); pe__clear_action_flags(action, pe_action_optional); } // Group action is unrunnable if any member action is if (!pcmk__str_eq(task_s, action->task, pcmk__str_none) && pcmk_is_set(flags, pe_action_runnable) && !pcmk_is_set(member_flags, pe_action_runnable)) { pe_rsc_trace(action->rsc, "%s is unrunnable because %s is", action->uuid, member_action->uuid); pe__clear_raw_action_flags(flags, "group action", pe_action_runnable); pe__clear_action_flags(action, pe_action_runnable); } /* Group (pseudo-)actions other than stop or demote are unrunnable * unless every member will do it. */ } else if ((task != stop_rsc) && (task != action_demote)) { pe_rsc_trace(action->rsc, "%s is not runnable because %s will not %s", action->uuid, member->id, task_s); pe__clear_raw_action_flags(flags, "group action", pe_action_runnable); } } return flags; } /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two actions, update the actions' flags * (and runnable_before members if appropriate) as appropriate for the ordering. * Effects may cascade to other orderings involving the actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * (only used when interleaving instances) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pe_action_optional to affect only mandatory * actions, and pe_action_runnable to affect only * runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * \param[in,out] data_set Cluster working set * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t pcmk__group_update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set) { uint32_t changed = pcmk__updated_none; CRM_ASSERT((first != NULL) && (then != NULL) && (data_set != NULL)); // Group method can be called only for group action as "then" action CRM_ASSERT(then->rsc != NULL); // Update the actions for the group itself changed |= pcmk__update_ordered_actions(first, then, node, flags, filter, type, data_set); // Update the actions for each group member for (GList *iter = then->rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; pe_action_t *member_action = find_first_action(member->actions, NULL, then->task, node); if (member_action != NULL) { changed |= member->cmds->update_ordered_actions(first, member_action, node, flags, filter, type, data_set); } } return changed; } /*! * \internal * \brief Apply a location constraint to a group's allowed node scores * * \param[in,out] rsc Group resource to apply constraint to * \param[in,out] location Location constraint to apply */ void pcmk__group_apply_location(pe_resource_t *rsc, pe__location_t *location) { GList *node_list_orig = NULL; GList *node_list_copy = NULL; bool reset_scores = true; CRM_ASSERT((rsc != NULL) && (location != NULL)); node_list_orig = location->node_list_rh; node_list_copy = pcmk__copy_node_list(node_list_orig, true); reset_scores = pe__group_flag_is_set(rsc, pe__group_colocated); // Apply the constraint for the group itself (updates node scores) pcmk__apply_location(rsc, location); // Apply the constraint for each member for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; member->cmds->apply_location(member, location); if (reset_scores) { /* The first member of colocated groups needs to use the original * node scores, but subsequent members should work on a copy, since * the first member's scores already incorporate theirs. */ reset_scores = false; location->node_list_rh = node_list_copy; } } location->node_list_rh = node_list_orig; g_list_free_full(node_list_copy, free); } // Group implementation of resource_alloc_functions_t:colocated_resources() GList * pcmk__group_colocated_resources(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *colocated_rscs) { const pe_resource_t *member = NULL; CRM_ASSERT(rsc != NULL); if (orig_rsc == NULL) { orig_rsc = rsc; } if (pe__group_flag_is_set(rsc, pe__group_colocated) || pe_rsc_is_clone(rsc->parent)) { /* This group has colocated members and/or is cloned -- either way, * add every child's colocated resources to the list. The first and last * members will include the group's own colocations. */ colocated_rscs = g_list_prepend(colocated_rscs, (gpointer) rsc); for (const GList *iter = rsc->children; iter != NULL; iter = iter->next) { member = (const pe_resource_t *) iter->data; colocated_rscs = member->cmds->colocated_resources(member, orig_rsc, colocated_rscs); } } else if (rsc->children != NULL) { /* This group's members are not colocated, and the group is not cloned, * so just add the group's own colocations to the list. */ colocated_rscs = pcmk__colocated_resources(rsc, orig_rsc, colocated_rscs); } return colocated_rscs; } // Group implementation of resource_alloc_functions_t:with_this_colocations() void pcmk__with_group_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (rsc->variant == pe_group) && (orig_rsc != NULL) && (list != NULL), return); // Ignore empty groups if (rsc->children == NULL) { return; } /* "With this" colocations are needed only for the group itself and for its * last member. Add the group's colocations plus any relevant * parent colocations if cloned. */ if ((rsc == orig_rsc) || (orig_rsc == pe__last_group_member(rsc))) { crm_trace("Adding 'with %s' colocations to list for %s", rsc->id, orig_rsc->id); pcmk__add_with_this_list(list, rsc->rsc_cons_lhs); if (rsc->parent != NULL) { // Cloned group rsc->parent->cmds->with_this_colocations(rsc->parent, orig_rsc, list); } } } // Group implementation of resource_alloc_functions_t:this_with_colocations() void pcmk__group_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { CRM_CHECK((rsc != NULL) && (rsc->variant == pe_group) && (orig_rsc != NULL) && (list != NULL), return); // Ignore empty groups if (rsc->children == NULL) { return; } /* Colocations for the group itself, or for its first member, consist of the * group's colocations plus any relevant parent colocations if cloned. */ if ((rsc == orig_rsc) || (orig_rsc == (const pe_resource_t *) rsc->children->data)) { crm_trace("Adding '%s with' colocations to list for %s", rsc->id, orig_rsc->id); pcmk__add_this_with_list(list, rsc->rsc_cons); if (rsc->parent != NULL) { // Cloned group rsc->parent->cmds->this_with_colocations(rsc->parent, orig_rsc, list); } return; } /* Later group members honor the group's colocations indirectly, due to the * internal group colocations that chain everything from the first member. * However, if an earlier group member is unmanaged, this chaining will not * happen, so the group's mandatory colocations must be explicitly added. */ for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { const pe_resource_t *member = (const pe_resource_t *) iter->data; if (orig_rsc == member) { break; // We've seen all earlier members, and none are unmanaged } if (!pcmk_is_set(member->flags, pe_rsc_managed)) { crm_trace("Adding mandatory '%s with' colocations to list for " "member %s because earlier member %s is unmanaged", rsc->id, orig_rsc->id, member->id); for (const GList *cons_iter = rsc->rsc_cons; cons_iter != NULL; cons_iter = cons_iter->next) { const pcmk__colocation_t *colocation = NULL; colocation = (const pcmk__colocation_t *) cons_iter->data; if (colocation->score == INFINITY) { pcmk__add_this_with(list, colocation); } } // @TODO Add mandatory (or all?) clone constraints if cloned break; } } } /*! * \internal * \brief Update nodes with scores of colocated resources' nodes * * Given a table of nodes and a resource, update the nodes' scores with the * scores of the best nodes matching the attribute used for each of the * resource's relevant colocations. * * \param[in,out] rsc Group resource to check colocations for * \param[in] log_id Resource ID for logs (if NULL, use \p rsc ID) * \param[in,out] nodes Nodes to update (set initial contents to NULL * to copy \p rsc's allowed nodes) * \param[in] colocation Original colocation constraint (used to get * configured primary resource's stickiness, and * to get colocation node attribute; if NULL, * \p rsc's own matching node scores will not be * added, and *nodes must be NULL as well) * \param[in] factor Incorporate scores multiplied by this factor * \param[in] flags Bitmask of enum pcmk__coloc_select values * * \note NULL *nodes, NULL colocation, and the pcmk__coloc_select_this_with * flag are used together (and only by cmp_resources()). * \note The caller remains responsible for freeing \p *nodes. */ void pcmk__group_add_colocated_node_scores(pe_resource_t *rsc, const char *log_id, GHashTable **nodes, pcmk__colocation_t *colocation, float factor, uint32_t flags) { pe_resource_t *member = NULL; CRM_ASSERT((rsc != NULL) && (rsc->variant == pe_group) && (nodes != NULL) && ((colocation != NULL) || (*nodes == NULL))); if (log_id == NULL) { log_id = rsc->id; } // Avoid infinite recursion if (pcmk_is_set(rsc->flags, pe_rsc_merging)) { pe_rsc_info(rsc, "%s: Breaking dependency loop at %s", log_id, rsc->id); return; } pe__set_resource_flags(rsc, pe_rsc_merging); // Ignore empty groups (only possible with schema validation disabled) if (rsc->children == NULL) { return; } /* Refer the operation to the first or last member as appropriate. * * cmp_resources() is the only caller that passes a NULL nodes table, * and is also the only caller using pcmk__coloc_select_this_with. * For "this with" colocations, the last member will recursively incorporate * all the other members' "this with" colocations via the internal group * colocations (and via the first member, the group's own colocations). * * For "with this" colocations, the first member works similarly. */ if (*nodes == NULL) { member = pe__last_group_member(rsc); } else { member = rsc->children->data; } pe_rsc_trace(rsc, "%s: Merging scores from group %s using member %s " "(at %.6f)", log_id, rsc->id, member->id, factor); member->cmds->add_colocated_node_scores(member, log_id, nodes, colocation, factor, flags); pe__clear_resource_flags(rsc, pe_rsc_merging); } // Group implementation of resource_alloc_functions_t:add_utilization() void pcmk__group_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization) { pe_resource_t *member = NULL; CRM_ASSERT((rsc != NULL) && (orig_rsc != NULL) && (utilization != NULL)); if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } pe_rsc_trace(orig_rsc, "%s: Adding group %s as colocated utilization", orig_rsc->id, rsc->id); if (pe__group_flag_is_set(rsc, pe__group_colocated) || pe_rsc_is_clone(rsc->parent)) { // Every group member will be on same node, so sum all members for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { member = (pe_resource_t *) iter->data; if (pcmk_is_set(member->flags, pe_rsc_provisional) && (g_list_find(all_rscs, member) == NULL)) { member->cmds->add_utilization(member, orig_rsc, all_rscs, utilization); } } } else if (rsc->children != NULL) { // Just add first member's utilization member = (pe_resource_t *) rsc->children->data; if ((member != NULL) && pcmk_is_set(member->flags, pe_rsc_provisional) && (g_list_find(all_rscs, member) == NULL)) { member->cmds->add_utilization(member, orig_rsc, all_rscs, utilization); } } } // Group implementation of resource_alloc_functions_t:shutdown_lock() void pcmk__group_shutdown_lock(pe_resource_t *rsc) { CRM_ASSERT(rsc != NULL); for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *member = (pe_resource_t *) iter->data; member->cmds->shutdown_lock(member); } } diff --git a/lib/pacemaker/pcmk_sched_instances.c b/lib/pacemaker/pcmk_sched_instances.c index 8fe59e6def..a06172b3e3 100644 --- a/lib/pacemaker/pcmk_sched_instances.c +++ b/lib/pacemaker/pcmk_sched_instances.c @@ -1,1656 +1,1656 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ /* This file is intended for code usable with both clone instances and bundle * replica containers. */ #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Check whether a clone or bundle has instances for all available nodes * * \param[in] collective Clone or bundle to check * * \return true if \p collective has enough instances for all of its available * allowed nodes, otherwise false */ static bool can_run_everywhere(const pe_resource_t *collective) { GHashTableIter iter; pe_node_t *node = NULL; int available_nodes = 0; int max_instances = 0; switch (collective->variant) { case pe_clone: max_instances = pe__clone_max(collective); break; case pe_container: max_instances = pe__bundle_max(collective); break; default: return false; // Not actually possible } g_hash_table_iter_init(&iter, collective->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (pcmk__node_available(node, false, false) && (max_instances < ++available_nodes)) { return false; } } return true; } /*! * \internal * \brief Check whether a node is allowed to run an instance * * \param[in] instance Clone instance or bundle container to check * \param[in] node Node to check * \param[in] max_per_node Maximum number of instances allowed to run on a node * * \return true if \p node is allowed to run \p instance, otherwise false */ static bool can_run_instance(const pe_resource_t *instance, const pe_node_t *node, int max_per_node) { pe_node_t *allowed_node = NULL; if (pcmk_is_set(instance->flags, pe_rsc_orphan)) { pe_rsc_trace(instance, "%s cannot run on %s: orphaned", instance->id, pe__node_name(node)); return false; } if (!pcmk__node_available(node, false, false)) { pe_rsc_trace(instance, "%s cannot run on %s: node cannot run resources", instance->id, pe__node_name(node)); return false; } allowed_node = pcmk__top_allowed_node(instance, node); if (allowed_node == NULL) { crm_warn("%s cannot run on %s: node not allowed", instance->id, pe__node_name(node)); return false; } if (allowed_node->weight < 0) { pe_rsc_trace(instance, "%s cannot run on %s: parent score is %s there", instance->id, pe__node_name(node), pcmk_readable_score(allowed_node->weight)); return false; } if (allowed_node->count >= max_per_node) { pe_rsc_trace(instance, "%s cannot run on %s: node already has %d instance%s", instance->id, pe__node_name(node), max_per_node, pcmk__plural_s(max_per_node)); return false; } pe_rsc_trace(instance, "%s can run on %s (%d already running)", instance->id, pe__node_name(node), allowed_node->count); return true; } /*! * \internal * \brief Ban a clone instance or bundle replica from unavailable allowed nodes * * \param[in,out] instance Clone instance or bundle replica to ban * \param[in] max_per_node Maximum instances allowed to run on a node */ static void ban_unavailable_allowed_nodes(pe_resource_t *instance, int max_per_node) { if (instance->allowed_nodes != NULL) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, instance->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (!can_run_instance(instance, node, max_per_node)) { pe_rsc_trace(instance, "Banning %s from unavailable node %s", instance->id, pe__node_name(node)); node->weight = -INFINITY; for (GList *child_iter = instance->children; child_iter != NULL; child_iter = child_iter->next) { pe_resource_t *child = (pe_resource_t *) child_iter->data; pe_node_t *child_node = NULL; child_node = pe_hash_table_lookup(child->allowed_nodes, node->details->id); if (child_node != NULL) { pe_rsc_trace(instance, "Banning %s child %s " "from unavailable node %s", instance->id, child->id, pe__node_name(node)); child_node->weight = -INFINITY; } } } } } } /*! * \internal * \brief Create a hash table with a single node in it * * \param[in] node Node to copy into new table * * \return Newly created hash table containing a copy of \p node * \note The caller is responsible for freeing the result with * g_hash_table_destroy(). */ static GHashTable * new_node_table(pe_node_t *node) { GHashTable *table = pcmk__strkey_table(NULL, free); node = pe__copy_node(node); g_hash_table_insert(table, (gpointer) node->details->id, node); return table; } /*! * \internal * \brief Apply a resource's parent's colocation scores to a node table * * \param[in] rsc Resource whose colocations should be applied * \param[in,out] nodes Node table to apply colocations to */ static void apply_parent_colocations(const pe_resource_t *rsc, GHashTable **nodes) { GList *iter = NULL; pcmk__colocation_t *colocation = NULL; pe_resource_t *other = NULL; float factor = 0.0; /* Because the this_with_colocations() and with_this_colocations() methods * boil down to copies of rsc_cons and rsc_cons_lhs for clones and bundles, * we can use those here directly for efficiency. */ for (iter = rsc->parent->rsc_cons; iter != NULL; iter = iter->next) { colocation = (pcmk__colocation_t *) iter->data; other = colocation->primary; factor = colocation->score / (float) INFINITY, other->cmds->add_colocated_node_scores(other, rsc->id, nodes, colocation, factor, pcmk__coloc_select_default); } for (iter = rsc->parent->rsc_cons_lhs; iter != NULL; iter = iter->next) { colocation = (pcmk__colocation_t *) iter->data; if (!pcmk__colocation_has_influence(colocation, rsc)) { continue; } other = colocation->dependent; factor = colocation->score / (float) INFINITY, other->cmds->add_colocated_node_scores(other, rsc->id, nodes, colocation, factor, pcmk__coloc_select_nonnegative); } } /*! * \internal * \brief Compare clone or bundle instances based on colocation scores * * Determine the relative order in which two clone or bundle instances should be * assigned to nodes, considering the scores of colocation constraints directly * or indirectly involving them. * * \param[in] instance1 First instance to compare * \param[in] instance2 Second instance to compare * * \return A negative number if \p instance1 should be assigned first, * a positive number if \p instance2 should be assigned first, * or 0 if assignment order doesn't matter */ static int cmp_instance_by_colocation(const pe_resource_t *instance1, const pe_resource_t *instance2) { int rc = 0; pe_node_t *node1 = NULL; pe_node_t *node2 = NULL; pe_node_t *current_node1 = pe__current_node(instance1); pe_node_t *current_node2 = pe__current_node(instance2); GHashTable *colocated_scores1 = NULL; GHashTable *colocated_scores2 = NULL; CRM_ASSERT((instance1 != NULL) && (instance1->parent != NULL) && (instance2 != NULL) && (instance2->parent != NULL) && (current_node1 != NULL) && (current_node2 != NULL)); // Create node tables initialized with each node colocated_scores1 = new_node_table(current_node1); colocated_scores2 = new_node_table(current_node2); // Apply parental colocations apply_parent_colocations(instance1, &colocated_scores1); apply_parent_colocations(instance2, &colocated_scores2); // Find original nodes again, with scores updated for colocations node1 = g_hash_table_lookup(colocated_scores1, current_node1->details->id); node2 = g_hash_table_lookup(colocated_scores2, current_node2->details->id); // Compare nodes by updated scores if (node1->weight < node2->weight) { crm_trace("Assign %s (%d on %s) after %s (%d on %s)", instance1->id, node1->weight, pe__node_name(node1), instance2->id, node2->weight, pe__node_name(node2)); rc = 1; } else if (node1->weight > node2->weight) { crm_trace("Assign %s (%d on %s) before %s (%d on %s)", instance1->id, node1->weight, pe__node_name(node1), instance2->id, node2->weight, pe__node_name(node2)); rc = -1; } g_hash_table_destroy(colocated_scores1); g_hash_table_destroy(colocated_scores2); return rc; } /*! * \internal * \brief Check whether a resource or any of its children are failed * * \param[in] rsc Resource to check * * \return true if \p rsc or any of its children are failed, otherwise false */ static bool did_fail(const pe_resource_t *rsc) { if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { return true; } for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { if (did_fail((const pe_resource_t *) iter->data)) { return true; } } return false; } /*! * \internal * \brief Check whether a node is allowed to run a resource * * \param[in] rsc Resource to check * \param[in,out] node Node to check (will be set NULL if not allowed) * * \return true if *node is either NULL or allowed for \p rsc, otherwise false */ static bool node_is_allowed(const pe_resource_t *rsc, pe_node_t **node) { if (*node != NULL) { pe_node_t *allowed = pe_hash_table_lookup(rsc->allowed_nodes, (*node)->details->id); if ((allowed == NULL) || (allowed->weight < 0)) { pe_rsc_trace(rsc, "%s: current location (%s) is unavailable", rsc->id, pe__node_name(*node)); *node = NULL; return false; } } return true; } /*! * \internal * \brief Compare two clone or bundle instances' instance numbers * * \param[in] a First instance to compare * \param[in] b Second instance to compare * * \return A negative number if \p a's instance number is lower, * a positive number if \p b's instance number is lower, * or 0 if their instance numbers are the same */ gint pcmk__cmp_instance_number(gconstpointer a, gconstpointer b) { const pe_resource_t *instance1 = (const pe_resource_t *) a; const pe_resource_t *instance2 = (const pe_resource_t *) b; char *div1 = NULL; char *div2 = NULL; CRM_ASSERT((instance1 != NULL) && (instance2 != NULL)); // Clone numbers are after a colon, bundle numbers after a dash div1 = strrchr(instance1->id, ':'); if (div1 == NULL) { div1 = strrchr(instance1->id, '-'); } div2 = strrchr(instance2->id, ':'); if (div2 == NULL) { div2 = strrchr(instance2->id, '-'); } CRM_ASSERT((div1 != NULL) && (div2 != NULL)); return (gint) (strtol(div1 + 1, NULL, 10) - strtol(div2 + 1, NULL, 10)); } /*! * \internal * \brief Compare clone or bundle instances according to assignment order * * Compare two clone or bundle instances according to the order they should be * assigned to nodes, preferring (in order): * * - Active instance that is less multiply active * - Instance that is not active on a disallowed node * - Instance with higher configured priority * - Active instance whose current node can run resources * - Active instance whose parent is allowed on current node * - Active instance whose current node has fewer other instances * - Active instance * - Instance that isn't failed * - Instance whose colocations result in higher score on current node * - Instance with lower ID in lexicographic order * * \param[in] a First instance to compare * \param[in] b Second instance to compare * * \return A negative number if \p a should be assigned first, * a positive number if \p b should be assigned first, * or 0 if assignment order doesn't matter */ gint pcmk__cmp_instance(gconstpointer a, gconstpointer b) { int rc = 0; pe_node_t *node1 = NULL; pe_node_t *node2 = NULL; unsigned int nnodes1 = 0; unsigned int nnodes2 = 0; bool can1 = true; bool can2 = true; const pe_resource_t *instance1 = (const pe_resource_t *) a; const pe_resource_t *instance2 = (const pe_resource_t *) b; CRM_ASSERT((instance1 != NULL) && (instance2 != NULL)); node1 = instance1->fns->active_node(instance1, &nnodes1, NULL); node2 = instance2->fns->active_node(instance2, &nnodes2, NULL); /* If both instances are running and at least one is multiply * active, prefer instance that's running on fewer nodes. */ if ((nnodes1 > 0) && (nnodes2 > 0)) { if (nnodes1 < nnodes2) { crm_trace("Assign %s (active on %d) before %s (active on %d): " "less multiply active", instance1->id, nnodes1, instance2->id, nnodes2); return -1; } else if (nnodes1 > nnodes2) { crm_trace("Assign %s (active on %d) after %s (active on %d): " "more multiply active", instance1->id, nnodes1, instance2->id, nnodes2); return 1; } } /* An instance that is either inactive or active on an allowed node is * preferred over an instance that is active on a no-longer-allowed node. */ can1 = node_is_allowed(instance1, &node1); can2 = node_is_allowed(instance2, &node2); if (can1 && !can2) { crm_trace("Assign %s before %s: not active on a disallowed node", instance1->id, instance2->id); return -1; } else if (!can1 && can2) { crm_trace("Assign %s after %s: active on a disallowed node", instance1->id, instance2->id); return 1; } // Prefer instance with higher configured priority if (instance1->priority > instance2->priority) { crm_trace("Assign %s before %s: priority (%d > %d)", instance1->id, instance2->id, instance1->priority, instance2->priority); return -1; } else if (instance1->priority < instance2->priority) { crm_trace("Assign %s after %s: priority (%d < %d)", instance1->id, instance2->id, instance1->priority, instance2->priority); return 1; } // Prefer active instance if ((node1 == NULL) && (node2 == NULL)) { crm_trace("No assignment preference for %s vs. %s: inactive", instance1->id, instance2->id); return 0; } else if (node1 == NULL) { crm_trace("Assign %s after %s: active", instance1->id, instance2->id); return 1; } else if (node2 == NULL) { crm_trace("Assign %s before %s: active", instance1->id, instance2->id); return -1; } // Prefer instance whose current node can run resources can1 = pcmk__node_available(node1, false, false); can2 = pcmk__node_available(node2, false, false); if (can1 && !can2) { crm_trace("Assign %s before %s: current node can run resources", instance1->id, instance2->id); return -1; } else if (!can1 && can2) { crm_trace("Assign %s after %s: current node can't run resources", instance1->id, instance2->id); return 1; } // Prefer instance whose parent is allowed to run on instance's current node node1 = pcmk__top_allowed_node(instance1, node1); node2 = pcmk__top_allowed_node(instance2, node2); if ((node1 == NULL) && (node2 == NULL)) { crm_trace("No assignment preference for %s vs. %s: " "parent not allowed on either instance's current node", instance1->id, instance2->id); return 0; } else if (node1 == NULL) { crm_trace("Assign %s after %s: parent not allowed on current node", instance1->id, instance2->id); return 1; } else if (node2 == NULL) { crm_trace("Assign %s before %s: parent allowed on current node", instance1->id, instance2->id); return -1; } // Prefer instance whose current node is running fewer other instances if (node1->count < node2->count) { crm_trace("Assign %s before %s: fewer active instances on current node", instance1->id, instance2->id); return -1; } else if (node1->count > node2->count) { crm_trace("Assign %s after %s: more active instances on current node", instance1->id, instance2->id); return 1; } // Prefer instance that isn't failed can1 = did_fail(instance1); can2 = did_fail(instance2); if (!can1 && can2) { crm_trace("Assign %s before %s: not failed", instance1->id, instance2->id); return -1; } else if (can1 && !can2) { crm_trace("Assign %s after %s: failed", instance1->id, instance2->id); return 1; } // Prefer instance with higher cumulative colocation score on current node rc = cmp_instance_by_colocation(instance1, instance2); if (rc != 0) { return rc; } // Prefer instance with lower instance number rc = pcmk__cmp_instance_number(instance1, instance2); if (rc < 0) { crm_trace("Assign %s before %s: instance number", instance1->id, instance2->id); } else if (rc > 0) { crm_trace("Assign %s after %s: instance number", instance1->id, instance2->id); } else { crm_trace("No assignment preference for %s vs. %s", instance1->id, instance2->id); } return rc; } /*! * \internal * \brief Choose a node for an instance * * \param[in,out] instance Clone instance or bundle replica container * \param[in] prefer If not NULL, attempt early assignment to this * node, if still the best choice; otherwise, * perform final assignment * \param[in] max_per_node Assign at most this many instances to one node * * \return true if \p instance could be assigned to a node, otherwise false */ static bool assign_instance(pe_resource_t *instance, const pe_node_t *prefer, int max_per_node) { pe_node_t *chosen = NULL; pe_node_t *allowed = NULL; CRM_ASSERT(instance != NULL); pe_rsc_trace(instance, "Assigning %s (preferring %s)", instance->id, ((prefer == NULL)? "no node" : prefer->details->uname)); if (!pcmk_is_set(instance->flags, pe_rsc_provisional)) { // Instance is already assigned return instance->fns->location(instance, NULL, FALSE) != NULL; } if (pcmk_is_set(instance->flags, pe_rsc_allocating)) { pe_rsc_debug(instance, "Assignment loop detected involving %s colocations", instance->id); return false; } if (prefer != NULL) { // Possible early assignment to preferred node // Get preferred node with instance's scores allowed = g_hash_table_lookup(instance->allowed_nodes, prefer->details->id); if ((allowed == NULL) || (allowed->weight < 0)) { pe_rsc_trace(instance, "Not assigning %s to preferred node %s: unavailable", instance->id, pe__node_name(prefer)); return false; } } ban_unavailable_allowed_nodes(instance, max_per_node); if (prefer == NULL) { // Final assignment chosen = instance->cmds->assign(instance, NULL); } else { // Possible early assignment to preferred node GHashTable *backup = pcmk__copy_node_table(instance->allowed_nodes); chosen = instance->cmds->assign(instance, prefer); // Revert nodes if preferred node won't be assigned - if ((chosen != NULL) && (chosen->details != prefer->details)) { + if ((chosen != NULL) && !pe__same_node(chosen, prefer)) { crm_info("Not assigning %s to preferred node %s: %s is better", instance->id, pe__node_name(prefer), pe__node_name(chosen)); g_hash_table_destroy(instance->allowed_nodes); instance->allowed_nodes = backup; pcmk__unassign_resource(instance); chosen = NULL; } else if (backup != NULL) { g_hash_table_destroy(backup); } } // The parent tracks how many instances have been assigned to each node if (chosen != NULL) { allowed = pcmk__top_allowed_node(instance, chosen); if (allowed == NULL) { /* The instance is allowed on the node, but its parent isn't. This * shouldn't be possible if the resource is managed, and we won't be * able to limit the number of instances assigned to the node. */ CRM_LOG_ASSERT(!pcmk_is_set(instance->flags, pe_rsc_managed)); } else { allowed->count++; } } return chosen != NULL; } /*! * \internal * \brief Reset the node counts of a resource's allowed nodes to zero * * \param[in,out] rsc Resource to reset * * \return Number of nodes that are available to run resources */ static unsigned int reset_allowed_node_counts(pe_resource_t *rsc) { unsigned int available_nodes = 0; pe_node_t *node = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { node->count = 0; if (pcmk__node_available(node, false, false)) { available_nodes++; } } return available_nodes; } /*! * \internal * \brief Check whether an instance has a preferred node * * \param[in] rsc Clone or bundle being assigned (for logs only) * \param[in] instance Clone instance or bundle replica container * \param[in] optimal_per_node Optimal number of instances per node * * \return Instance's current node if still available, otherwise NULL */ static const pe_node_t * preferred_node(const pe_resource_t *rsc, const pe_resource_t *instance, int optimal_per_node) { const pe_node_t *node = NULL; const pe_node_t *parent_node = NULL; // Check whether instance is active, healthy, and not yet assigned if ((instance->running_on == NULL) || !pcmk_is_set(instance->flags, pe_rsc_provisional) || pcmk_is_set(instance->flags, pe_rsc_failed)) { return NULL; } // Check whether instance's current node can run resources node = pe__current_node(instance); if (!pcmk__node_available(node, true, false)) { pe_rsc_trace(rsc, "Not assigning %s to %s early (unavailable)", instance->id, pe__node_name(node)); return NULL; } // Check whether node already has optimal number of instances assigned parent_node = pcmk__top_allowed_node(instance, node); if ((parent_node != NULL) && (parent_node->count >= optimal_per_node)) { pe_rsc_trace(rsc, "Not assigning %s to %s early " "(optimal instances already assigned)", instance->id, pe__node_name(node)); return NULL; } return node; } /*! * \internal * \brief Assign collective instances to nodes * * \param[in,out] collective Clone or bundle resource being assigned * \param[in,out] instances List of clone instances or bundle containers * \param[in] max_total Maximum instances to assign in total * \param[in] max_per_node Maximum instances to assign to any one node */ void pcmk__assign_instances(pe_resource_t *collective, GList *instances, int max_total, int max_per_node) { // Reuse node count to track number of assigned instances unsigned int available_nodes = reset_allowed_node_counts(collective); int optimal_per_node = 0; int assigned = 0; GList *iter = NULL; pe_resource_t *instance = NULL; const pe_node_t *current = NULL; if (available_nodes > 0) { optimal_per_node = max_total / available_nodes; } if (optimal_per_node < 1) { optimal_per_node = 1; } pe_rsc_debug(collective, "Assigning up to %d %s instance%s to up to %u node%s " "(at most %d per host, %d optimal)", max_total, collective->id, pcmk__plural_s(max_total), available_nodes, pcmk__plural_s(available_nodes), max_per_node, optimal_per_node); // Assign as many instances as possible to their current location for (iter = instances; (iter != NULL) && (assigned < max_total); iter = iter->next) { instance = (pe_resource_t *) iter->data; current = preferred_node(collective, instance, optimal_per_node); if ((current != NULL) && assign_instance(instance, current, max_per_node)) { pe_rsc_trace(collective, "Assigned %s to current node %s", instance->id, pe__node_name(current)); assigned++; } } pe_rsc_trace(collective, "Assigned %d of %d instance%s to current node", assigned, max_total, pcmk__plural_s(max_total)); for (iter = instances; iter != NULL; iter = iter->next) { instance = (pe_resource_t *) iter->data; if (!pcmk_is_set(instance->flags, pe_rsc_provisional)) { continue; // Already assigned } if (instance->running_on != NULL) { current = pe__current_node(instance); if (pcmk__top_allowed_node(instance, current) == NULL) { const char *unmanaged = ""; if (!pcmk_is_set(instance->flags, pe_rsc_managed)) { unmanaged = "Unmanaged resource "; } crm_notice("%s%s is running on %s which is no longer allowed", unmanaged, instance->id, pe__node_name(current)); } } if (assigned >= max_total) { pe_rsc_debug(collective, "Not assigning %s because maximum %d instances " "already assigned", instance->id, max_total); resource_location(instance, NULL, -INFINITY, "collective_limit_reached", collective->cluster); } else if (assign_instance(instance, NULL, max_per_node)) { assigned++; } } pe_rsc_debug(collective, "Assigned %d of %d possible instance%s of %s", assigned, max_total, pcmk__plural_s(max_total), collective->id); } enum instance_state { instance_starting = (1 << 0), instance_stopping = (1 << 1), /* This indicates that some instance is restarting. It's not the same as * instance_starting|instance_stopping, which would indicate that some * instance is starting, and some instance (not necessarily the same one) is * stopping. */ instance_restarting = (1 << 2), instance_active = (1 << 3), instance_all = instance_starting|instance_stopping |instance_restarting|instance_active, }; /*! * \internal * \brief Check whether an instance is active, starting, and/or stopping * * \param[in] instance Clone instance or bundle replica container * \param[in,out] state Whether any instance is starting, stopping, etc. */ static void check_instance_state(const pe_resource_t *instance, uint32_t *state) { const GList *iter = NULL; uint32_t instance_state = 0; // State of just this instance // No need to check further if all conditions have already been detected if (pcmk_all_flags_set(*state, instance_all)) { return; } // If instance is a collective (a cloned group), check its children instead if (instance->variant > pe_native) { for (iter = instance->children; (iter != NULL) && !pcmk_all_flags_set(*state, instance_all); iter = iter->next) { check_instance_state((const pe_resource_t *) iter->data, state); } return; } // If we get here, instance is a primitive if (instance->running_on != NULL) { instance_state |= instance_active; } // Check each of the instance's actions for runnable start or stop for (iter = instance->actions; (iter != NULL) && !pcmk_all_flags_set(instance_state, instance_starting |instance_stopping); iter = iter->next) { const pe_action_t *action = (const pe_action_t *) iter->data; const bool optional = pcmk_is_set(action->flags, pe_action_optional); if (pcmk__str_eq(RSC_START, action->task, pcmk__str_none)) { if (!optional && pcmk_is_set(action->flags, pe_action_runnable)) { pe_rsc_trace(instance, "Instance is starting due to %s", action->uuid); instance_state |= instance_starting; } else { pe_rsc_trace(instance, "%s doesn't affect %s state (%s)", action->uuid, instance->id, (optional? "optional" : "unrunnable")); } } else if (pcmk__str_eq(RSC_STOP, action->task, pcmk__str_none)) { /* Only stop actions can be pseudo-actions for primitives. That * indicates that the node they are on is being fenced, so the stop * is implied rather than actually executed. */ if (!optional && pcmk_any_flags_set(action->flags, pe_action_pseudo|pe_action_runnable)) { pe_rsc_trace(instance, "Instance is stopping due to %s", action->uuid); instance_state |= instance_stopping; } else { pe_rsc_trace(instance, "%s doesn't affect %s state (%s)", action->uuid, instance->id, (optional? "optional" : "unrunnable")); } } } if (pcmk_all_flags_set(instance_state, instance_starting|instance_stopping)) { instance_state |= instance_restarting; } *state |= instance_state; } /*! * \internal * \brief Create actions for collective resource instances * * \param[in,out] collective Clone or bundle resource to create actions for * \param[in,out] instances List of clone instances or bundle containers */ void pcmk__create_instance_actions(pe_resource_t *collective, GList *instances) { uint32_t state = 0; pe_action_t *stop = NULL; pe_action_t *stopped = NULL; pe_action_t *start = NULL; pe_action_t *started = NULL; pe_rsc_trace(collective, "Creating collective instance actions for %s", collective->id); // Create actions for each instance appropriate to its variant for (GList *iter = instances; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->cmds->create_actions(instance); check_instance_state(instance, &state); } // Create pseudo-actions for rsc start and started start = pe__new_rsc_pseudo_action(collective, RSC_START, !pcmk_is_set(state, instance_starting), true); started = pe__new_rsc_pseudo_action(collective, RSC_STARTED, !pcmk_is_set(state, instance_starting), false); started->priority = INFINITY; if (pcmk_any_flags_set(state, instance_active|instance_starting)) { pe__set_action_flags(started, pe_action_runnable); } // Create pseudo-actions for rsc stop and stopped stop = pe__new_rsc_pseudo_action(collective, RSC_STOP, !pcmk_is_set(state, instance_stopping), true); stopped = pe__new_rsc_pseudo_action(collective, RSC_STOPPED, !pcmk_is_set(state, instance_stopping), true); stopped->priority = INFINITY; if (!pcmk_is_set(state, instance_restarting)) { pe__set_action_flags(stop, pe_action_migrate_runnable); } if (collective->variant == pe_clone) { pe__create_clone_notif_pseudo_ops(collective, start, started, stop, stopped); } } /*! * \internal * \brief Get a list of clone instances or bundle replica containers * * \param[in] rsc Clone or bundle resource * * \return Clone instances if \p rsc is a clone, or a newly created list of * \p rsc's replica containers if \p rsc is a bundle * \note The caller must call free_instance_list() on the result when the list * is no longer needed. */ static inline GList * get_instance_list(const pe_resource_t *rsc) { if (rsc->variant == pe_container) { return pe__bundle_containers(rsc); } else { return rsc->children; } } /*! * \internal * \brief Free any memory created by get_instance_list() * * \param[in] rsc Clone or bundle resource passed to get_instance_list() * \param[in,out] list Return value of get_instance_list() for \p rsc */ static inline void free_instance_list(const pe_resource_t *rsc, GList *list) { if (list != rsc->children) { g_list_free(list); } } /*! * \internal * \brief Check whether an instance is compatible with a role and node * * \param[in] instance Clone instance or bundle replica container * \param[in] node Instance must match this node * \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return true if \p instance is compatible with \p node and \p role, * otherwise false */ bool pcmk__instance_matches(const pe_resource_t *instance, const pe_node_t *node, enum rsc_role_e role, bool current) { pe_node_t *instance_node = NULL; CRM_CHECK((instance != NULL) && (node != NULL), return false); if ((role != RSC_ROLE_UNKNOWN) && (role != instance->fns->state(instance, current))) { pe_rsc_trace(instance, "%s is not a compatible instance (role is not %s)", instance->id, role2text(role)); return false; } if (!is_set_recursive(instance, pe_rsc_block, true)) { // We only want instances that haven't failed instance_node = instance->fns->location(instance, NULL, current); } if (instance_node == NULL) { pe_rsc_trace(instance, "%s is not a compatible instance (not assigned to a node)", instance->id); return false; } - if (instance_node->details != node->details) { + if (!pe__same_node(instance_node, node)) { pe_rsc_trace(instance, "%s is not a compatible instance (assigned to %s not %s)", instance->id, pe__node_name(instance_node), pe__node_name(node)); return false; } return true; } /*! * \internal * \brief Find an instance that matches a given resource by node and role * * \param[in] match_rsc Resource that instance must match (for logging only) * \param[in] rsc Clone or bundle resource to check for matching instance * \param[in] node Instance must match this node * \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return \p rsc instance matching \p node and \p role if any, otherwise NULL */ static pe_resource_t * find_compatible_instance_on_node(const pe_resource_t *match_rsc, const pe_resource_t *rsc, const pe_node_t *node, enum rsc_role_e role, bool current) { GList *instances = NULL; instances = get_instance_list(rsc); for (GList *iter = instances; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; if (pcmk__instance_matches(instance, node, role, current)) { pe_rsc_trace(match_rsc, "Found %s %s instance %s compatible with %s on %s", role == RSC_ROLE_UNKNOWN? "matching" : role2text(role), rsc->id, instance->id, match_rsc->id, pe__node_name(node)); free_instance_list(rsc, instances); // Only frees list, not contents return instance; } } free_instance_list(rsc, instances); pe_rsc_trace(match_rsc, "No %s %s instance found compatible with %s on %s", ((role == RSC_ROLE_UNKNOWN)? "matching" : role2text(role)), rsc->id, match_rsc->id, pe__node_name(node)); return NULL; } /*! * \internal * \brief Find a clone instance or bundle container compatible with a resource * * \param[in] match_rsc Resource that instance must match * \param[in] rsc Clone or bundle resource to check for matching instance * \param[in] role If not RSC_ROLE_UNKNOWN, instance must match this role * \param[in] current If true, compare instance's original node and role, * otherwise compare assigned next node and role * * \return Compatible (by \p role and \p match_rsc location) instance of \p rsc * if any, otherwise NULL */ pe_resource_t * pcmk__find_compatible_instance(const pe_resource_t *match_rsc, const pe_resource_t *rsc, enum rsc_role_e role, bool current) { pe_resource_t *instance = NULL; GList *nodes = NULL; const pe_node_t *node = match_rsc->fns->location(match_rsc, NULL, current); // If match_rsc has a node, check only that node if (node != NULL) { return find_compatible_instance_on_node(match_rsc, rsc, node, role, current); } // Otherwise check for an instance matching any of match_rsc's allowed nodes nodes = pcmk__sort_nodes(g_hash_table_get_values(match_rsc->allowed_nodes), NULL); for (GList *iter = nodes; (iter != NULL) && (instance == NULL); iter = iter->next) { instance = find_compatible_instance_on_node(match_rsc, rsc, (pe_node_t *) iter->data, role, current); } if (instance == NULL) { pe_rsc_debug(rsc, "No %s instance found compatible with %s", rsc->id, match_rsc->id); } g_list_free(nodes); return instance; } /*! * \internal * \brief Unassign an instance if mandatory ordering has no interleave match * * \param[in] first 'First' action in an ordering * \param[in] then 'Then' action in an ordering * \param[in,out] then_instance 'Then' instance that has no interleave match * \param[in] type Group of enum pe_ordering flags to apply * \param[in] current If true, "then" action is stopped or demoted * * \return true if \p then_instance was unassigned, otherwise false */ static bool unassign_if_mandatory(const pe_action_t *first, const pe_action_t *then, pe_resource_t *then_instance, uint32_t type, bool current) { // Allow "then" instance to go down even without an interleave match if (current) { pe_rsc_trace(then->rsc, "%s has no instance to order before stopping " "or demoting %s", first->rsc->id, then_instance->id); /* If the "first" action must be runnable, but there is no "first" * instance, the "then" instance must not be allowed to come up. */ } else if (pcmk_any_flags_set(type, pe_order_runnable_left |pe_order_implies_then)) { pe_rsc_info(then->rsc, "Inhibiting %s from being active " "because there is no %s instance to interleave", then_instance->id, first->rsc->id); return pcmk__assign_resource(then_instance, NULL, true); } return false; } /*! * \internal * \brief Find first matching action for a clone instance or bundle container * * \param[in] action Action in an interleaved ordering * \param[in] instance Clone instance or bundle container being interleaved * \param[in] action_name Action to look for * \param[in] node If not NULL, require action to be on this node * \param[in] for_first If true, \p instance is the 'first' resource in the * ordering, otherwise it is the 'then' resource * * \return First action for \p instance (or in some cases if \p instance is a * bundle container, its containerized resource) that matches * \p action_name and \p node if any, otherwise NULL */ static pe_action_t * find_instance_action(const pe_action_t *action, const pe_resource_t *instance, const char *action_name, const pe_node_t *node, bool for_first) { const pe_resource_t *rsc = NULL; pe_action_t *matching_action = NULL; /* If instance is a bundle container, sometimes we should interleave the * action for the container itself, and sometimes for the containerized * resource. * * For example, given "start bundle A then bundle B", B likely requires the * service inside A's container to be active, rather than just the * container, so we should interleave the action for A's containerized * resource. On the other hand, it's possible B's container itself requires * something from A, so we should interleave the action for B's container. * * Essentially, for 'first', we should use the containerized resource for * everything except stop, and for 'then', we should use the container for * everything except promote and demote (which can only be performed on the * containerized resource). */ if ((for_first && !pcmk__str_any_of(action->task, CRMD_ACTION_STOP, CRMD_ACTION_STOPPED, NULL)) || (!for_first && pcmk__str_any_of(action->task, CRMD_ACTION_PROMOTE, CRMD_ACTION_PROMOTED, CRMD_ACTION_DEMOTE, CRMD_ACTION_DEMOTED, NULL))) { rsc = pe__get_rsc_in_container(instance); } if (rsc == NULL) { rsc = instance; // No containerized resource, use instance itself } else { node = NULL; // Containerized actions are on bundle-created guest } matching_action = find_first_action(rsc->actions, NULL, action_name, node); if (matching_action != NULL) { return matching_action; } if (pcmk_is_set(instance->flags, pe_rsc_orphan) || pcmk__str_any_of(action_name, RSC_STOP, RSC_DEMOTE, NULL)) { crm_trace("No %s action found for %s%s", action_name, pcmk_is_set(instance->flags, pe_rsc_orphan)? "orphan " : "", instance->id); } else { crm_err("No %s action found for %s to interleave (bug?)", action_name, instance->id); } return NULL; } /*! * \internal * \brief Get the original action name of a bundle or clone action * * Given an action for a bundle or clone, get the original action name, * mapping notify to the action being notified, and if the instances are * primitives, mapping completion actions to the action that was completed * (for example, stopped to stop). * * \param[in] action Clone or bundle action to check * * \return Original action name for \p action */ static const char * orig_action_name(const pe_action_t *action) { const pe_resource_t *instance = action->rsc->children->data; // Any instance char *action_type = NULL; const char *action_name = action->task; enum action_tasks orig_task = no_action; if (pcmk__strcase_any_of(action->task, CRMD_ACTION_NOTIFY, CRMD_ACTION_NOTIFIED, NULL)) { // action->uuid is RSC_(confirmed-){pre,post}_notify_ACTION_INTERVAL CRM_CHECK(parse_op_key(action->uuid, NULL, &action_type, NULL), return task2text(no_action)); action_name = strstr(action_type, "_notify_"); CRM_CHECK(action_name != NULL, return task2text(no_action)); action_name += strlen("_notify_"); } orig_task = get_complex_task(instance, action_name); free(action_type); return task2text(orig_task); } /*! * \internal * \brief Update two interleaved actions according to an ordering between them * * Given information about an ordering of two interleaved actions, update the * actions' flags (and runnable_before members if appropriate) as appropriate * for the ordering. Effects may cascade to other orderings involving the * actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * \param[in] filter Action flags to limit scope of certain updates (may * include pe_action_optional to affect only mandatory * actions, and pe_action_runnable to affect only * runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * * \return Group of enum pcmk__updated flags indicating what was updated */ static uint32_t update_interleaved_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t filter, uint32_t type) { GList *instances = NULL; uint32_t changed = pcmk__updated_none; const char *orig_first_task = orig_action_name(first); // Stops and demotes must be interleaved with instance on current node bool current = pcmk__ends_with(first->uuid, "_" CRMD_ACTION_STOPPED "_0") || pcmk__ends_with(first->uuid, "_" CRMD_ACTION_DEMOTED "_0"); // Update the specified actions for each "then" instance individually instances = get_instance_list(then->rsc); for (GList *iter = instances; iter != NULL; iter = iter->next) { pe_resource_t *first_instance = NULL; pe_resource_t *then_instance = iter->data; pe_action_t *first_action = NULL; pe_action_t *then_action = NULL; // Find a "first" instance to interleave with this "then" instance first_instance = pcmk__find_compatible_instance(then_instance, first->rsc, RSC_ROLE_UNKNOWN, current); if (first_instance == NULL) { // No instance can be interleaved if (unassign_if_mandatory(first, then, then_instance, type, current)) { pcmk__set_updated_flags(changed, first, pcmk__updated_then); } continue; } first_action = find_instance_action(first, first_instance, orig_first_task, node, true); if (first_action == NULL) { continue; } then_action = find_instance_action(then, then_instance, then->task, node, false); if (then_action == NULL) { continue; } if (order_actions(first_action, then_action, type)) { pcmk__set_updated_flags(changed, first, pcmk__updated_first|pcmk__updated_then); } changed |= then_instance->cmds->update_ordered_actions( first_action, then_action, node, first_instance->cmds->action_flags(first_action, node), filter, type, then->rsc->cluster); } free_instance_list(then->rsc, instances); return changed; } /*! * \internal * \brief Check whether two actions in an ordering can be interleaved * * \param[in] first 'First' action in the ordering * \param[in] then 'Then' action in the ordering * * \return true if \p first and \p then can be interleaved, otherwise false */ static bool can_interleave_actions(const pe_action_t *first, const pe_action_t *then) { bool interleave = false; pe_resource_t *rsc = NULL; if ((first->rsc == NULL) || (then->rsc == NULL)) { crm_trace("Not interleaving %s with %s: not resource actions", first->uuid, then->uuid); return false; } if (first->rsc == then->rsc) { crm_trace("Not interleaving %s with %s: same resource", first->uuid, then->uuid); return false; } if ((first->rsc->variant < pe_clone) || (then->rsc->variant < pe_clone)) { crm_trace("Not interleaving %s with %s: not clones or bundles", first->uuid, then->uuid); return false; } if (pcmk__ends_with(then->uuid, "_stop_0") || pcmk__ends_with(then->uuid, "_demote_0")) { rsc = first->rsc; } else { rsc = then->rsc; } interleave = crm_is_true(g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INTERLEAVE)); pe_rsc_trace(rsc, "'%s then %s' will %sbe interleaved (based on %s)", first->uuid, then->uuid, (interleave? "" : "not "), rsc->id); return interleave; } /*! * \internal * \brief Update non-interleaved instance actions according to an ordering * * Given information about an ordering of two non-interleaved actions, update * the actions' flags (and runnable_before members if appropriate) as * appropriate for the ordering. Effects may cascade to other orderings * involving the actions as well. * * \param[in,out] instance Clone instance or bundle container * \param[in,out] first "First" action in ordering * \param[in] then "Then" action in ordering (for \p instance's parent) * \param[in] node If not NULL, limit scope of ordering to this node * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pe_action_optional to affect only mandatory * actions, and pe_action_runnable to affect only * runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * * \return Group of enum pcmk__updated flags indicating what was updated */ static uint32_t update_noninterleaved_actions(pe_resource_t *instance, pe_action_t *first, const pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type) { pe_action_t *instance_action = NULL; uint32_t instance_flags = 0; uint32_t changed = pcmk__updated_none; // Check whether instance has an equivalent of "then" action instance_action = find_first_action(instance->actions, NULL, then->task, node); if (instance_action == NULL) { return changed; } // Check whether action is runnable instance_flags = instance->cmds->action_flags(instance_action, node); if (!pcmk_is_set(instance_flags, pe_action_runnable)) { return changed; } // If so, update actions for the instance changed = instance->cmds->update_ordered_actions(first, instance_action, node, flags, filter, type, instance->cluster); // Propagate any changes to later actions if (pcmk_is_set(changed, pcmk__updated_then)) { for (GList *after_iter = instance_action->actions_after; after_iter != NULL; after_iter = after_iter->next) { pe_action_wrapper_t *after = after_iter->data; pcmk__update_action_for_orderings(after->action, instance->cluster); } } return changed; } /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two clone or bundle actions, update * the actions' flags (and runnable_before members if appropriate) as * appropriate for the ordering. Effects may cascade to other orderings * involving the actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * (only used when interleaving instances) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pe_action_optional to affect only mandatory * actions, and pe_action_runnable to affect only * runnable actions) * \param[in] type Group of enum pe_ordering flags to apply * \param[in,out] data_set Cluster working set * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t pcmk__instance_update_ordered_actions(pe_action_t *first, pe_action_t *then, const pe_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pe_working_set_t *data_set) { if (then->rsc == NULL) { return pcmk__updated_none; } else if (can_interleave_actions(first, then)) { return update_interleaved_actions(first, then, node, filter, type); } else { uint32_t changed = pcmk__updated_none; GList *instances = get_instance_list(then->rsc); // Update actions for the clone or bundle resource itself changed |= pcmk__update_ordered_actions(first, then, node, flags, filter, type, data_set); // Update the 'then' clone instances or bundle containers individually for (GList *iter = instances; iter != NULL; iter = iter->next) { pe_resource_t *instance = iter->data; changed |= update_noninterleaved_actions(instance, first, then, node, flags, filter, type); } free_instance_list(then->rsc, instances); return changed; } } #define pe__clear_action_summary_flags(flags, action, flag) do { \ flags = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Action summary", action->rsc->id, \ flags, flag, #flag); \ } while (0) /*! * \internal * \brief Return action flags for a given clone or bundle action * * \param[in,out] action Action for a clone or bundle * \param[in] instances Clone instances or bundle containers * \param[in] node If not NULL, limit effects to this node * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__collective_action_flags(pe_action_t *action, const GList *instances, const pe_node_t *node) { bool any_runnable = false; const char *action_name = orig_action_name(action); // Set original assumptions (optional and runnable may be cleared below) uint32_t flags = pe_action_optional|pe_action_runnable|pe_action_pseudo; for (const GList *iter = instances; iter != NULL; iter = iter->next) { const pe_resource_t *instance = iter->data; const pe_node_t *instance_node = NULL; pe_action_t *instance_action = NULL; uint32_t instance_flags; // Node is relevant only to primitive instances if (instance->variant == pe_native) { instance_node = node; } instance_action = find_first_action(instance->actions, NULL, action_name, instance_node); if (instance_action == NULL) { pe_rsc_trace(action->rsc, "%s has no %s action on %s", instance->id, action_name, pe__node_name(node)); continue; } pe_rsc_trace(action->rsc, "%s has %s for %s on %s", instance->id, instance_action->uuid, action_name, pe__node_name(node)); instance_flags = instance->cmds->action_flags(instance_action, node); // If any instance action is mandatory, so is the collective action if (pcmk_is_set(flags, pe_action_optional) && !pcmk_is_set(instance_flags, pe_action_optional)) { pe_rsc_trace(instance, "%s is mandatory because %s is", action->uuid, instance_action->uuid); pe__clear_action_summary_flags(flags, action, pe_action_optional); pe__clear_action_flags(action, pe_action_optional); } // If any instance action is runnable, so is the collective action if (pcmk_is_set(instance_flags, pe_action_runnable)) { any_runnable = true; } } if (!any_runnable) { pe_rsc_trace(action->rsc, "%s is not runnable because no instance can run %s", action->uuid, action_name); pe__clear_action_summary_flags(flags, action, pe_action_runnable); if (node == NULL) { pe__clear_action_flags(action, pe_action_runnable); } } return flags; } /*! * \internal * \brief Add a collective resource's colocations to a list for an instance * * \param[in,out] list Colocation list to add to * \param[in] instance Clone or bundle instance or instance group member * \param[in] collective Clone or bundle resource with colocations to add * \param[in] with_this If true, add collective's "with this" colocations, * otherwise add its "this with" colocations */ void pcmk__add_collective_constraints(GList **list, const pe_resource_t *instance, const pe_resource_t *collective, bool with_this) { const GList *colocations = NULL; bool everywhere = false; CRM_CHECK((list != NULL) && (instance != NULL), return); if (collective == NULL) { return; } switch (collective->variant) { case pe_clone: case pe_container: break; default: return; } everywhere = can_run_everywhere(collective); if (with_this) { colocations = collective->rsc_cons_lhs; } else { colocations = collective->rsc_cons; } for (const GList *iter = colocations; iter != NULL; iter = iter->next) { const pcmk__colocation_t *colocation = iter->data; if (with_this && !pcmk__colocation_has_influence(colocation, instance)) { continue; } if (!everywhere || (colocation->score < 0) || (!with_this && (colocation->score == INFINITY))) { if (with_this) { pcmk__add_with_this(list, colocation); } else { pcmk__add_this_with(list, colocation); } } } } diff --git a/lib/pacemaker/pcmk_sched_location.c b/lib/pacemaker/pcmk_sched_location.c index b4ce4ff07d..9aecf4291c 100644 --- a/lib/pacemaker/pcmk_sched_location.c +++ b/lib/pacemaker/pcmk_sched_location.c @@ -1,678 +1,679 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include +#include #include #include "libpacemaker_private.h" static int get_node_score(const char *rule, const char *score, bool raw, pe_node_t *node, pe_resource_t *rsc) { int score_f = 0; if (score == NULL) { pe_err("Rule %s: no score specified. Assuming 0.", rule); } else if (raw) { score_f = char2score(score); } else { const char *attr_score = pe_node_attribute_calculated(node, score, rsc); if (attr_score == NULL) { crm_debug("Rule %s: %s did not have a value for %s", rule, pe__node_name(node), score); score_f = -INFINITY; } else { crm_debug("Rule %s: %s had value %s for %s", rule, pe__node_name(node), attr_score, score); score_f = char2score(attr_score); } } return score_f; } static pe__location_t * generate_location_rule(pe_resource_t *rsc, xmlNode *rule_xml, const char *discovery, crm_time_t *next_change, pe_working_set_t *data_set, pe_re_match_data_t *re_match_data) { const char *rule_id = NULL; const char *score = NULL; const char *boolean = NULL; const char *role = NULL; GList *gIter = NULL; GList *match_L = NULL; bool do_and = true; bool accept = true; bool raw_score = true; bool score_allocated = false; pe__location_t *location_rule = NULL; rule_xml = expand_idref(rule_xml, data_set->input); if (rule_xml == NULL) { return NULL; } rule_id = crm_element_value(rule_xml, XML_ATTR_ID); boolean = crm_element_value(rule_xml, XML_RULE_ATTR_BOOLEAN_OP); role = crm_element_value(rule_xml, XML_RULE_ATTR_ROLE); crm_trace("Processing rule: %s", rule_id); if ((role != NULL) && (text2role(role) == RSC_ROLE_UNKNOWN)) { pe_err("Bad role specified for %s: %s", rule_id, role); return NULL; } score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE); if (score == NULL) { score = crm_element_value(rule_xml, XML_RULE_ATTR_SCORE_ATTRIBUTE); if (score != NULL) { raw_score = false; } } if (pcmk__str_eq(boolean, "or", pcmk__str_casei)) { do_and = false; } location_rule = pcmk__new_location(rule_id, rsc, 0, discovery, NULL, data_set); if (location_rule == NULL) { return NULL; } if ((re_match_data != NULL) && (re_match_data->nregs > 0) && (re_match_data->pmatch[0].rm_so != -1) && !raw_score) { char *result = pe_expand_re_matches(score, re_match_data); if (result != NULL) { score = result; score_allocated = true; } } if (role != NULL) { crm_trace("Setting role filter: %s", role); location_rule->role_filter = text2role(role); if (location_rule->role_filter == RSC_ROLE_UNPROMOTED) { /* Any promotable clone cannot be promoted without being in the * unpromoted role first. Ergo, any constraint for the unpromoted * role applies to every role. */ location_rule->role_filter = RSC_ROLE_UNKNOWN; } } if (do_and) { GList *gIter = NULL; match_L = pcmk__copy_node_list(data_set->nodes, true); for (gIter = match_L; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; node->weight = get_node_score(rule_id, score, raw_score, node, rsc); } } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { int score_f = 0; pe_node_t *node = (pe_node_t *) gIter->data; pe_match_data_t match_data = { .re = re_match_data, .params = pe_rsc_params(rsc, node, data_set), .meta = rsc->meta, }; accept = pe_test_rule(rule_xml, node->details->attrs, RSC_ROLE_UNKNOWN, data_set->now, next_change, &match_data); crm_trace("Rule %s %s on %s", ID(rule_xml), accept? "passed" : "failed", pe__node_name(node)); score_f = get_node_score(rule_id, score, raw_score, node, rsc); if (accept) { pe_node_t *local = pe_find_node_id(match_L, node->details->id); if ((local == NULL) && do_and) { continue; } else if (local == NULL) { local = pe__copy_node(node); match_L = g_list_append(match_L, local); } if (!do_and) { local->weight = pcmk__add_scores(local->weight, score_f); } crm_trace("%s has score %s after %s", pe__node_name(node), pcmk_readable_score(local->weight), rule_id); } else if (do_and && !accept) { // Remove it pe_node_t *delete = pe_find_node_id(match_L, node->details->id); if (delete != NULL) { match_L = g_list_remove(match_L, delete); crm_trace("%s did not match", pe__node_name(node)); } free(delete); } } if (score_allocated) { free((char *)score); } location_rule->node_list_rh = match_L; if (location_rule->node_list_rh == NULL) { crm_trace("No matching nodes for rule %s", rule_id); return NULL; } crm_trace("%s: %d nodes matched", rule_id, g_list_length(location_rule->node_list_rh)); return location_rule; } static void unpack_rsc_location(xmlNode *xml_obj, pe_resource_t *rsc, const char *role, const char *score, pe_working_set_t *data_set, pe_re_match_data_t *re_match_data) { pe__location_t *location = NULL; const char *rsc_id = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *node = crm_element_value(xml_obj, XML_CIB_TAG_NODE); const char *discovery = crm_element_value(xml_obj, XML_LOCATION_ATTR_DISCOVERY); if (rsc == NULL) { pcmk__config_warn("Ignoring constraint '%s' because resource '%s' " "does not exist", id, rsc_id); return; } if (score == NULL) { score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); } if ((node != NULL) && (score != NULL)) { int score_i = char2score(score); pe_node_t *match = pe_find_node(data_set->nodes, node); if (!match) { return; } location = pcmk__new_location(id, rsc, score_i, discovery, match, data_set); } else { bool empty = true; crm_time_t *next_change = crm_time_new_undefined(); /* This loop is logically parallel to pe_evaluate_rules(), except * instead of checking whether any rule is active, we set up location * constraints for each active rule. */ for (xmlNode *rule_xml = first_named_child(xml_obj, XML_TAG_RULE); rule_xml != NULL; rule_xml = crm_next_same_xml(rule_xml)) { empty = false; crm_trace("Unpacking %s/%s", id, ID(rule_xml)); generate_location_rule(rsc, rule_xml, discovery, next_change, data_set, re_match_data); } if (empty) { pcmk__config_err("Ignoring constraint '%s' because it contains " "no rules", id); } /* If there is a point in the future when the evaluation of a rule will * change, make sure the scheduler is re-run by that time. */ if (crm_time_is_defined(next_change)) { time_t t = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(t, data_set); } crm_time_free(next_change); return; } if (role == NULL) { role = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE); } if ((location != NULL) && (role != NULL)) { if (text2role(role) == RSC_ROLE_UNKNOWN) { pe_err("Invalid constraint %s: Bad role %s", id, role); return; } else { enum rsc_role_e r = text2role(role); switch(r) { case RSC_ROLE_UNKNOWN: case RSC_ROLE_STARTED: case RSC_ROLE_UNPROMOTED: /* Applies to all */ location->role_filter = RSC_ROLE_UNKNOWN; break; default: location->role_filter = r; break; } } } } static void unpack_simple_location(xmlNode *xml_obj, pe_working_set_t *data_set) { const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); if (value) { pe_resource_t *rsc; rsc = pcmk__find_constraint_resource(data_set->resources, value); unpack_rsc_location(xml_obj, rsc, NULL, NULL, data_set, NULL); } value = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE_PATTERN); if (value) { regex_t *r_patt = calloc(1, sizeof(regex_t)); bool invert = false; GList *rIter = NULL; if (value[0] == '!') { value++; invert = true; } if (regcomp(r_patt, value, REG_EXTENDED) != 0) { pcmk__config_err("Ignoring constraint '%s' because " XML_LOC_ATTR_SOURCE_PATTERN " has invalid value '%s'", id, value); free(r_patt); return; } for (rIter = data_set->resources; rIter; rIter = rIter->next) { pe_resource_t *r = rIter->data; int nregs = 0; regmatch_t *pmatch = NULL; int status; if(r_patt->re_nsub > 0) { nregs = r_patt->re_nsub + 1; } else { nregs = 1; } pmatch = calloc(nregs, sizeof(regmatch_t)); status = regexec(r_patt, r->id, nregs, pmatch, 0); if (!invert && (status == 0)) { pe_re_match_data_t re_match_data = { .string = r->id, .nregs = nregs, .pmatch = pmatch }; crm_debug("'%s' matched '%s' for %s", r->id, value, id); unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, &re_match_data); } else if (invert && (status != 0)) { crm_debug("'%s' is an inverted match of '%s' for %s", r->id, value, id); unpack_rsc_location(xml_obj, r, NULL, NULL, data_set, NULL); } else { crm_trace("'%s' does not match '%s' for %s", r->id, value, id); } free(pmatch); } regfree(r_patt); free(r_patt); } } // \return Standard Pacemaker return code static int unpack_location_tags(xmlNode *xml_obj, xmlNode **expanded_xml, pe_working_set_t *data_set) { const char *id = NULL; const char *rsc_id = NULL; const char *state = NULL; pe_resource_t *rsc = NULL; pe_tag_t *tag = NULL; xmlNode *rsc_set = NULL; *expanded_xml = NULL; CRM_CHECK(xml_obj != NULL, return EINVAL); id = ID(xml_obj); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return pcmk_rc_unpack_error; } // Check whether there are any resource sets with template or tag references *expanded_xml = pcmk__expand_tags_in_sets(xml_obj, data_set); if (*expanded_xml != NULL) { crm_log_xml_trace(*expanded_xml, "Expanded rsc_location"); return pcmk_rc_ok; } rsc_id = crm_element_value(xml_obj, XML_LOC_ATTR_SOURCE); if (rsc_id == NULL) { return pcmk_rc_ok; } if (!pcmk__valid_resource_or_tag(data_set, rsc_id, &rsc, &tag)) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", id, rsc_id); return pcmk_rc_unpack_error; } else if (rsc != NULL) { // No template is referenced return pcmk_rc_ok; } state = crm_element_value(xml_obj, XML_RULE_ATTR_ROLE); *expanded_xml = copy_xml(xml_obj); // Convert template/tag reference in "rsc" into resource_set under constraint if (!pcmk__tag_to_set(*expanded_xml, &rsc_set, XML_LOC_ATTR_SOURCE, false, data_set)) { free_xml(*expanded_xml); *expanded_xml = NULL; return pcmk_rc_unpack_error; } if (rsc_set != NULL) { if (state != NULL) { // Move "rsc-role" into converted resource_set as "role" attribute crm_xml_add(rsc_set, "role", state); xml_remove_prop(*expanded_xml, XML_RULE_ATTR_ROLE); } crm_log_xml_trace(*expanded_xml, "Expanded rsc_location"); } else { // No sets free_xml(*expanded_xml); *expanded_xml = NULL; } return pcmk_rc_ok; } // \return Standard Pacemaker return code static int unpack_location_set(xmlNode *location, xmlNode *set, pe_working_set_t *data_set) { xmlNode *xml_rsc = NULL; pe_resource_t *resource = NULL; const char *set_id; const char *role; const char *local_score; CRM_CHECK(set != NULL, return EINVAL); set_id = ID(set); if (set_id == NULL) { pcmk__config_err("Ignoring " XML_CONS_TAG_RSC_SET " without " XML_ATTR_ID " in constraint '%s'", pcmk__s(ID(location), "(missing ID)")); return pcmk_rc_unpack_error; } role = crm_element_value(set, "role"); local_score = crm_element_value(set, XML_RULE_ATTR_SCORE); for (xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { resource = pcmk__find_constraint_resource(data_set->resources, ID(xml_rsc)); if (resource == NULL) { pcmk__config_err("%s: No resource found for %s", set_id, ID(xml_rsc)); return pcmk_rc_unpack_error; } unpack_rsc_location(location, resource, role, local_score, data_set, NULL); } return pcmk_rc_ok; } void pcmk__unpack_location(xmlNode *xml_obj, pe_working_set_t *data_set) { xmlNode *set = NULL; bool any_sets = false; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; if (unpack_location_tags(xml_obj, &expanded_xml, data_set) != pcmk_rc_ok) { return; } if (expanded_xml) { orig_xml = xml_obj; xml_obj = expanded_xml; } for (set = first_named_child(xml_obj, XML_CONS_TAG_RSC_SET); set != NULL; set = crm_next_same_xml(set)) { any_sets = true; set = expand_idref(set, data_set->input); if ((set == NULL) // Configuration error, message already logged || (unpack_location_set(xml_obj, set, data_set) != pcmk_rc_ok)) { if (expanded_xml) { free_xml(expanded_xml); } return; } } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } if (!any_sets) { unpack_simple_location(xml_obj, data_set); } } /*! * \internal * \brief Add a new location constraint to a cluster working set * * \param[in] id XML ID of location constraint * \param[in,out] rsc Resource in location constraint - * \param[in] node_weight Constraint score + * \param[in] node_score Constraint score * \param[in] discover_mode Resource discovery option for constraint * \param[in] node Node in constraint (or NULL if rule-based) * \param[in,out] data_set Cluster working set to add constraint to * * \return Newly allocated location constraint * \note The result will be added to \p data_set and should not be freed * separately. */ pe__location_t * pcmk__new_location(const char *id, pe_resource_t *rsc, - int node_weight, const char *discover_mode, + int node_score, const char *discover_mode, pe_node_t *node, pe_working_set_t *data_set) { pe__location_t *new_con = NULL; if (id == NULL) { pe_err("Invalid constraint: no ID specified"); return NULL; } else if (rsc == NULL) { pe_err("Invalid constraint %s: no resource specified", id); return NULL; } else if (node == NULL) { - CRM_CHECK(node_weight == 0, return NULL); + CRM_CHECK(node_score == 0, return NULL); } new_con = calloc(1, sizeof(pe__location_t)); if (new_con != NULL) { new_con->id = strdup(id); new_con->rsc_lh = rsc; new_con->node_list_rh = NULL; new_con->role_filter = RSC_ROLE_UNKNOWN; if (pcmk__str_eq(discover_mode, "always", pcmk__str_null_matches|pcmk__str_casei)) { new_con->discover_mode = pe_discover_always; } else if (pcmk__str_eq(discover_mode, "never", pcmk__str_casei)) { new_con->discover_mode = pe_discover_never; } else if (pcmk__str_eq(discover_mode, "exclusive", pcmk__str_casei)) { new_con->discover_mode = pe_discover_exclusive; rsc->exclusive_discover = TRUE; } else { pe_err("Invalid " XML_LOCATION_ATTR_DISCOVERY " value %s " "in location constraint", discover_mode); } if (node != NULL) { pe_node_t *copy = pe__copy_node(node); - copy->weight = node_weight; + copy->weight = node_score; new_con->node_list_rh = g_list_prepend(NULL, copy); } data_set->placement_constraints = g_list_prepend(data_set->placement_constraints, new_con); rsc->rsc_location = g_list_prepend(rsc->rsc_location, new_con); } return new_con; } /*! * \internal * \brief Apply all location constraints * * \param[in,out] data_set Cluster working set */ void pcmk__apply_locations(pe_working_set_t *data_set) { for (GList *iter = data_set->placement_constraints; iter != NULL; iter = iter->next) { pe__location_t *location = iter->data; location->rsc_lh->cmds->apply_location(location->rsc_lh, location); } } /*! * \internal * \brief Apply a location constraint to a resource's allowed node scores * * \param[in,out] rsc Resource to apply constraint to * \param[in,out] location Location constraint to apply * * \note This does not consider the resource's children, so the resource's * apply_location() method should be used instead in most cases. */ void pcmk__apply_location(pe_resource_t *rsc, pe__location_t *location) { bool need_role = false; CRM_CHECK((rsc != NULL) && (location != NULL), return); // If a role was specified, ensure constraint is applicable need_role = (location->role_filter > RSC_ROLE_UNKNOWN); if (need_role && (location->role_filter != rsc->next_role)) { pe_rsc_trace(rsc, "Not applying %s to %s because role will be %s not %s", location->id, rsc->id, role2text(rsc->next_role), role2text(location->role_filter)); return; } if (location->node_list_rh == NULL) { pe_rsc_trace(rsc, "Not applying %s to %s because no nodes match", location->id, rsc->id); return; } pe_rsc_trace(rsc, "Applying %s%s%s to %s", location->id, (need_role? " for role " : ""), (need_role? role2text(location->role_filter) : ""), rsc->id); for (GList *gIter = location->node_list_rh; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; - pe_node_t *weighted_node = NULL; + pe_node_t *allowed_node = NULL; - weighted_node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, - node->details->id); - if (weighted_node == NULL) { + allowed_node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, + node->details->id); + if (allowed_node == NULL) { pe_rsc_trace(rsc, "* = %d on %s", node->weight, pe__node_name(node)); - weighted_node = pe__copy_node(node); + allowed_node = pe__copy_node(node); g_hash_table_insert(rsc->allowed_nodes, - (gpointer) weighted_node->details->id, - weighted_node); + (gpointer) allowed_node->details->id, + allowed_node); } else { pe_rsc_trace(rsc, "* + %d on %s", node->weight, pe__node_name(node)); - weighted_node->weight = pcmk__add_scores(weighted_node->weight, - node->weight); + allowed_node->weight = pcmk__add_scores(allowed_node->weight, + node->weight); } - if (weighted_node->rsc_discover_mode < location->discover_mode) { + if (allowed_node->rsc_discover_mode < location->discover_mode) { if (location->discover_mode == pe_discover_exclusive) { rsc->exclusive_discover = TRUE; } /* exclusive > never > always... always is default */ - weighted_node->rsc_discover_mode = location->discover_mode; + allowed_node->rsc_discover_mode = location->discover_mode; } } } diff --git a/lib/pacemaker/pcmk_sched_nodes.c b/lib/pacemaker/pcmk_sched_nodes.c index d7d5ba4616..fea91c513e 100644 --- a/lib/pacemaker/pcmk_sched_nodes.c +++ b/lib/pacemaker/pcmk_sched_nodes.c @@ -1,351 +1,351 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include // lrmd_event_data_t #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Check whether a node is available to run resources * * \param[in] node Node to check * \param[in] consider_score If true, consider a negative score unavailable * \param[in] consider_guest If true, consider a guest node unavailable whose * resource will not be active * * \return true if node is online and not shutting down, unclean, or in standby * or maintenance mode, otherwise false */ bool pcmk__node_available(const pe_node_t *node, bool consider_score, bool consider_guest) { if ((node == NULL) || (node->details == NULL) || !node->details->online || node->details->shutdown || node->details->unclean || node->details->standby || node->details->maintenance) { return false; } if (consider_score && (node->weight < 0)) { return false; } // @TODO Go through all callers to see which should set consider_guest if (consider_guest && pe__is_guest_node(node)) { pe_resource_t *guest = node->details->remote_rsc->container; if (guest->fns->location(guest, NULL, FALSE) == NULL) { return false; } } return true; } /*! * \internal * \brief Copy a hash table of node objects * * \param[in] nodes Hash table to copy * * \return New copy of nodes (or NULL if nodes is NULL) */ GHashTable * pcmk__copy_node_table(GHashTable *nodes) { GHashTable *new_table = NULL; GHashTableIter iter; pe_node_t *node = NULL; if (nodes == NULL) { return NULL; } new_table = pcmk__strkey_table(NULL, free); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { pe_node_t *new_node = pe__copy_node(node); g_hash_table_insert(new_table, (gpointer) new_node->details->id, new_node); } return new_table; } /*! * \internal * \brief Copy a list of node objects * * \param[in] list List to copy * \param[in] reset Set copies' scores to 0 * * \return New list of shallow copies of nodes in original list */ GList * pcmk__copy_node_list(const GList *list, bool reset) { GList *result = NULL; for (const GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = NULL; pe_node_t *this_node = (pe_node_t *) gIter->data; new_node = pe__copy_node(this_node); if (reset) { new_node->weight = 0; } result = g_list_prepend(result, new_node); } return result; } /*! * \internal - * \brief Compare two nodes for allocation desirability + * \brief Compare two nodes for assignment preference * - * Given two nodes, check which one is more preferred by allocation criteria - * such as node weight and utilization. + * Given two nodes, check which one is more preferred by assignment criteria + * such as node score and utilization. * * \param[in] a First node to compare * \param[in] b Second node to compare * \param[in] data Node that resource being assigned is active on, if any * * \return -1 if \p a is preferred, +1 if \p b is preferred, or 0 if they are * equally preferred */ static gint compare_nodes(gconstpointer a, gconstpointer b, gpointer data) { const pe_node_t *node1 = (const pe_node_t *) a; const pe_node_t *node2 = (const pe_node_t *) b; const pe_node_t *active = (const pe_node_t *) data; - int node1_weight = 0; - int node2_weight = 0; + int node1_score = 0; + int node2_score = 0; int result = 0; if (a == NULL) { return 1; } if (b == NULL) { return -1; } - // Compare node weights + // Compare node scores - node1_weight = pcmk__node_available(node1, false, false)? node1->weight : -INFINITY; - node2_weight = pcmk__node_available(node2, false, false)? node2->weight : -INFINITY; + node1_score = pcmk__node_available(node1, false, false)? node1->weight : -INFINITY; + node2_score = pcmk__node_available(node2, false, false)? node2->weight : -INFINITY; - if (node1_weight > node2_weight) { - crm_trace("%s (%d) > %s (%d) : weight", - pe__node_name(node1), node1_weight, pe__node_name(node2), - node2_weight); + if (node1_score > node2_score) { + crm_trace("%s (%d) > %s (%d) : score", + pe__node_name(node1), node1_score, pe__node_name(node2), + node2_score); return -1; } - if (node1_weight < node2_weight) { - crm_trace("%s (%d) < %s (%d) : weight", - pe__node_name(node1), node1_weight, pe__node_name(node2), - node2_weight); + if (node1_score < node2_score) { + crm_trace("%s (%d) < %s (%d) : score", + pe__node_name(node1), node1_score, pe__node_name(node2), + node2_score); return 1; } - crm_trace("%s (%d) == %s (%d) : weight", - pe__node_name(node1), node1_weight, pe__node_name(node2), - node2_weight); + crm_trace("%s (%d) == %s (%d) : score", + pe__node_name(node1), node1_score, pe__node_name(node2), + node2_score); // If appropriate, compare node utilization if (pcmk__str_eq(node1->details->data_set->placement_strategy, "minimal", pcmk__str_casei)) { goto equal; } if (pcmk__str_eq(node1->details->data_set->placement_strategy, "balanced", pcmk__str_casei)) { result = pcmk__compare_node_capacities(node1, node2); if (result < 0) { crm_trace("%s > %s : capacity (%d)", pe__node_name(node1), pe__node_name(node2), result); return -1; } else if (result > 0) { crm_trace("%s < %s : capacity (%d)", pe__node_name(node1), pe__node_name(node2), result); return 1; } } - // Compare number of allocated resources + // Compare number of resources already assigned to node if (node1->details->num_resources < node2->details->num_resources) { crm_trace("%s (%d) > %s (%d) : resources", pe__node_name(node1), node1->details->num_resources, pe__node_name(node2), node2->details->num_resources); return -1; } else if (node1->details->num_resources > node2->details->num_resources) { crm_trace("%s (%d) < %s (%d) : resources", pe__node_name(node1), node1->details->num_resources, pe__node_name(node2), node2->details->num_resources); return 1; } // Check whether one node is already running desired resource if (active != NULL) { - if (active->details == node1->details) { + if (pe__same_node(active, node1)) { crm_trace("%s (%d) > %s (%d) : active", pe__node_name(node1), node1->details->num_resources, pe__node_name(node2), node2->details->num_resources); return -1; - } else if (active->details == node2->details) { + } else if (pe__same_node(active, node2)) { crm_trace("%s (%d) < %s (%d) : active", pe__node_name(node1), node1->details->num_resources, pe__node_name(node2), node2->details->num_resources); return 1; } } // If all else is equal, prefer node with lowest-sorting name equal: crm_trace("%s = %s", pe__node_name(node1), pe__node_name(node2)); return strcmp(node1->details->uname, node2->details->uname); } /*! * \internal - * \brief Sort a list of nodes by allocation desirability + * \brief Sort a list of nodes by assigment preference * * \param[in,out] nodes Node list to sort * \param[in] active_node Node where resource being assigned is active * * \return New head of sorted list */ GList * pcmk__sort_nodes(GList *nodes, pe_node_t *active_node) { return g_list_sort_with_data(nodes, compare_nodes, active_node); } /*! * \internal * \brief Check whether any node is available to run resources * * \param[in] nodes Nodes to check * * \return true if any node in \p nodes is available to run resources, * otherwise false */ bool pcmk__any_node_available(GHashTable *nodes) { GHashTableIter iter; const pe_node_t *node = NULL; if (nodes == NULL) { return false; } g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (pcmk__node_available(node, true, false)) { return true; } } return false; } /*! * \internal * \brief Apply node health values for all nodes in cluster * * \param[in,out] data_set Cluster working set */ void pcmk__apply_node_health(pe_working_set_t *data_set) { int base_health = 0; enum pcmk__health_strategy strategy; const char *strategy_str = pe_pref(data_set->config_hash, PCMK__OPT_NODE_HEALTH_STRATEGY); strategy = pcmk__parse_health_strategy(strategy_str); if (strategy == pcmk__health_strategy_none) { return; } crm_info("Applying node health strategy '%s'", strategy_str); // The progressive strategy can use a base health score if (strategy == pcmk__health_strategy_progressive) { base_health = pe__health_score(PCMK__OPT_NODE_HEALTH_BASE, data_set); } for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; int health = pe__sum_node_health_scores(node, base_health); // An overall health score of 0 has no effect if (health == 0) { continue; } crm_info("Overall system health of %s is %d", pe__node_name(node), health); // Use node health as a location score for each resource on the node for (GList *r = data_set->resources; r != NULL; r = r->next) { pe_resource_t *rsc = (pe_resource_t *) r->data; bool constrain = true; if (health < 0) { /* Negative health scores do not apply to resources with * allow-unhealthy-nodes=true. */ constrain = !crm_is_true(g_hash_table_lookup(rsc->meta, PCMK__META_ALLOW_UNHEALTHY_NODES)); } if (constrain) { pcmk__new_location(strategy_str, rsc, health, NULL, node, data_set); } else { pe_rsc_trace(rsc, "%s is immune from health ban on %s", rsc->id, pe__node_name(node)); } } } } /*! * \internal * \brief Check for a node in a resource's parent's allowed nodes * * \param[in] rsc Resource whose parent should be checked * \param[in] node Node to check for * * \return Equivalent of \p node from \p rsc's parent's allowed nodes if any, * otherwise NULL */ pe_node_t * pcmk__top_allowed_node(const pe_resource_t *rsc, const pe_node_t *node) { GHashTable *allowed_nodes = NULL; if ((rsc == NULL) || (node == NULL)) { return NULL; } else if (rsc->parent == NULL) { allowed_nodes = rsc->allowed_nodes; } else { allowed_nodes = rsc->parent->allowed_nodes; } return pe_hash_table_lookup(allowed_nodes, node->details->id); } diff --git a/lib/pacemaker/pcmk_sched_ordering.c b/lib/pacemaker/pcmk_sched_ordering.c index 6629999014..1b884dca0b 100644 --- a/lib/pacemaker/pcmk_sched_ordering.c +++ b/lib/pacemaker/pcmk_sched_ordering.c @@ -1,1463 +1,1462 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include // PRIx32 #include #include #include #include #include "libpacemaker_private.h" enum pe_order_kind { pe_order_kind_optional, pe_order_kind_mandatory, pe_order_kind_serialize, }; enum ordering_symmetry { ordering_asymmetric, // the only relation in an asymmetric ordering ordering_symmetric, // the normal relation in a symmetric ordering ordering_symmetric_inverse, // the inverse relation in a symmetric ordering }; #define EXPAND_CONSTRAINT_IDREF(__set, __rsc, __name) do { \ __rsc = pcmk__find_constraint_resource(data_set->resources, __name); \ if (__rsc == NULL) { \ pcmk__config_err("%s: No resource found for %s", __set, __name); \ return pcmk_rc_unpack_error; \ } \ } while (0) static const char * invert_action(const char *action) { if (pcmk__str_eq(action, RSC_START, pcmk__str_casei)) { return RSC_STOP; } else if (pcmk__str_eq(action, RSC_STOP, pcmk__str_casei)) { return RSC_START; } else if (pcmk__str_eq(action, RSC_PROMOTE, pcmk__str_casei)) { return RSC_DEMOTE; } else if (pcmk__str_eq(action, RSC_DEMOTE, pcmk__str_casei)) { return RSC_PROMOTE; } else if (pcmk__str_eq(action, RSC_PROMOTED, pcmk__str_casei)) { return RSC_DEMOTED; } else if (pcmk__str_eq(action, RSC_DEMOTED, pcmk__str_casei)) { return RSC_PROMOTED; } else if (pcmk__str_eq(action, RSC_STARTED, pcmk__str_casei)) { return RSC_STOPPED; } else if (pcmk__str_eq(action, RSC_STOPPED, pcmk__str_casei)) { return RSC_STARTED; } crm_warn("Unknown action '%s' specified in order constraint", action); return NULL; } static enum pe_order_kind get_ordering_type(const xmlNode *xml_obj) { enum pe_order_kind kind_e = pe_order_kind_mandatory; const char *kind = crm_element_value(xml_obj, XML_ORDER_ATTR_KIND); if (kind == NULL) { const char *score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); kind_e = pe_order_kind_mandatory; if (score) { // @COMPAT deprecated informally since 1.0.7, formally since 2.0.1 int score_i = char2score(score); if (score_i == 0) { kind_e = pe_order_kind_optional; } pe_warn_once(pe_wo_order_score, "Support for 'score' in rsc_order is deprecated " "and will be removed in a future release " "(use 'kind' instead)"); } } else if (pcmk__str_eq(kind, "Mandatory", pcmk__str_casei)) { kind_e = pe_order_kind_mandatory; } else if (pcmk__str_eq(kind, "Optional", pcmk__str_casei)) { kind_e = pe_order_kind_optional; } else if (pcmk__str_eq(kind, "Serialize", pcmk__str_casei)) { kind_e = pe_order_kind_serialize; } else { pcmk__config_err("Resetting '" XML_ORDER_ATTR_KIND "' for constraint " "%s to 'Mandatory' because '%s' is not valid", pcmk__s(ID(xml_obj), "missing ID"), kind); } return kind_e; } /*! * \internal * \brief Get ordering symmetry from XML * * \param[in] xml_obj Ordering XML * \param[in] parent_kind Default ordering kind * \param[in] parent_symmetrical_s Parent element's symmetrical setting, if any * * \retval ordering_symmetric Ordering is symmetric * \retval ordering_asymmetric Ordering is asymmetric */ static enum ordering_symmetry get_ordering_symmetry(const xmlNode *xml_obj, enum pe_order_kind parent_kind, const char *parent_symmetrical_s) { int rc = pcmk_rc_ok; bool symmetric = false; enum pe_order_kind kind = parent_kind; // Default to parent's kind // Check ordering XML for explicit kind if ((crm_element_value(xml_obj, XML_ORDER_ATTR_KIND) != NULL) || (crm_element_value(xml_obj, XML_RULE_ATTR_SCORE) != NULL)) { kind = get_ordering_type(xml_obj); } // Check ordering XML (and parent) for explicit symmetrical setting rc = pcmk__xe_get_bool_attr(xml_obj, XML_CONS_ATTR_SYMMETRICAL, &symmetric); if (rc != pcmk_rc_ok && parent_symmetrical_s != NULL) { symmetric = crm_is_true(parent_symmetrical_s); rc = pcmk_rc_ok; } if (rc == pcmk_rc_ok) { if (symmetric) { if (kind == pe_order_kind_serialize) { pcmk__config_warn("Ignoring " XML_CONS_ATTR_SYMMETRICAL " for '%s' because not valid with " XML_ORDER_ATTR_KIND " of 'Serialize'", ID(xml_obj)); } else { return ordering_symmetric; } } return ordering_asymmetric; } // Use default symmetry if (kind == pe_order_kind_serialize) { return ordering_asymmetric; } return ordering_symmetric; } /*! * \internal * \brief Get ordering flags appropriate to ordering kind * * \param[in] kind Ordering kind * \param[in] first Action name for 'first' action * \param[in] symmetry This ordering's symmetry role * * \return Minimal ordering flags appropriate to \p kind */ static uint32_t ordering_flags_for_kind(enum pe_order_kind kind, const char *first, enum ordering_symmetry symmetry) { uint32_t flags = pe_order_none; // so we trace-log all flags set pe__set_order_flags(flags, pe_order_optional); switch (kind) { case pe_order_kind_optional: break; case pe_order_kind_serialize: pe__set_order_flags(flags, pe_order_serialize_only); break; case pe_order_kind_mandatory: switch (symmetry) { case ordering_asymmetric: pe__set_order_flags(flags, pe_order_asymmetrical); break; case ordering_symmetric: pe__set_order_flags(flags, pe_order_implies_then); if (pcmk__strcase_any_of(first, RSC_START, RSC_PROMOTE, NULL)) { pe__set_order_flags(flags, pe_order_runnable_left); } break; case ordering_symmetric_inverse: pe__set_order_flags(flags, pe_order_implies_first); break; } break; } return flags; } /*! * \internal * \brief Find resource corresponding to ID specified in ordering * * \param[in] xml Ordering XML * \param[in] resource_attr XML attribute name for resource ID * \param[in] instance_attr XML attribute name for instance number. * This option is deprecated and will be removed in a * future release. * \param[in] data_set Cluster working set * * \return Resource corresponding to \p id, or NULL if none */ static pe_resource_t * get_ordering_resource(const xmlNode *xml, const char *resource_attr, const char *instance_attr, const pe_working_set_t *data_set) { // @COMPAT: instance_attr and instance_id variables deprecated since 2.1.5 pe_resource_t *rsc = NULL; const char *rsc_id = crm_element_value(xml, resource_attr); const char *instance_id = crm_element_value(xml, instance_attr); if (rsc_id == NULL) { pcmk__config_err("Ignoring constraint '%s' without %s", ID(xml), resource_attr); return NULL; } rsc = pcmk__find_constraint_resource(data_set->resources, rsc_id); if (rsc == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not exist", ID(xml), rsc_id); return NULL; } if (instance_id != NULL) { pe_warn_once(pe_wo_order_inst, "Support for " XML_ORDER_ATTR_FIRST_INSTANCE " and " XML_ORDER_ATTR_THEN_INSTANCE " is deprecated and will be " "removed in a future release."); if (!pe_rsc_is_clone(rsc)) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "is not a clone but instance '%s' was requested", ID(xml), rsc_id, instance_id); return NULL; } rsc = find_clone_instance(rsc, instance_id); if (rsc == NULL) { pcmk__config_err("Ignoring constraint '%s' because resource '%s' " "does not have an instance '%s'", "'%s'", ID(xml), rsc_id, instance_id); return NULL; } } return rsc; } /*! * \internal * \brief Determine minimum number of 'first' instances required in ordering * * \param[in] rsc 'First' resource in ordering * \param[in] xml Ordering XML * * \return Minimum 'first' instances required (or 0 if not applicable) */ static int get_minimum_first_instances(const pe_resource_t *rsc, const xmlNode *xml) { const char *clone_min = NULL; bool require_all = false; if (!pe_rsc_is_clone(rsc)) { return 0; } clone_min = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INCARNATION_MIN); if (clone_min != NULL) { int clone_min_int = 0; pcmk__scan_min_int(clone_min, &clone_min_int, 0); return clone_min_int; } /* @COMPAT 1.1.13: * require-all=false is deprecated equivalent of clone-min=1 */ if (pcmk__xe_get_bool_attr(xml, "require-all", &require_all) != ENODATA) { pe_warn_once(pe_wo_require_all, "Support for require-all in ordering constraints " "is deprecated and will be removed in a future release" " (use clone-min clone meta-attribute instead)"); if (!require_all) { return 1; } } return 0; } /*! * \internal * \brief Create orderings for a constraint with clone-min > 0 * * \param[in] id Ordering ID * \param[in,out] rsc_first 'First' resource in ordering (a clone) * \param[in] action_first 'First' action in ordering * \param[in] rsc_then 'Then' resource in ordering * \param[in] action_then 'Then' action in ordering * \param[in] flags Ordering flags * \param[in] clone_min Minimum required instances of 'first' * \param[in,out] data_set Cluster working set */ static void clone_min_ordering(const char *id, pe_resource_t *rsc_first, const char *action_first, pe_resource_t *rsc_then, const char *action_then, uint32_t flags, int clone_min, pe_working_set_t *data_set) { // Create a pseudo-action for when the minimum instances are active char *task = crm_strdup_printf(CRM_OP_RELAXED_CLONE ":%s", id); pe_action_t *clone_min_met = get_pseudo_op(task, data_set); free(task); /* Require the pseudo-action to have the required number of actions to be * considered runnable before allowing the pseudo-action to be runnable. */ clone_min_met->required_runnable_before = clone_min; pe__set_action_flags(clone_min_met, pe_action_requires_any); // Order the actions for each clone instance before the pseudo-action for (GList *rIter = rsc_first->children; rIter != NULL; rIter = rIter->next) { pe_resource_t *child = rIter->data; pcmk__new_ordering(child, pcmk__op_key(child->id, action_first, 0), NULL, NULL, NULL, clone_min_met, pe_order_one_or_more|pe_order_implies_then_printed, data_set); } // Order "then" action after the pseudo-action (if runnable) pcmk__new_ordering(NULL, NULL, clone_min_met, rsc_then, pcmk__op_key(rsc_then->id, action_then, 0), NULL, flags|pe_order_runnable_left, data_set); } /*! * \internal * \brief Update ordering flags for restart-type=restart * * \param[in] rsc 'Then' resource in ordering * \param[in] kind Ordering kind * \param[in] flag Ordering flag to set (when applicable) * \param[in,out] flags Ordering flag set to update * * \compat The restart-type resource meta-attribute is deprecated. Eventually, * it will be removed, and pe_restart_ignore will be the only behavior, * at which time this can just be removed entirely. */ #define handle_restart_type(rsc, kind, flag, flags) do { \ if (((kind) == pe_order_kind_optional) \ && ((rsc)->restart_type == pe_restart_restart)) { \ pe__set_order_flags((flags), (flag)); \ } \ } while (0) /*! * \internal * \brief Create new ordering for inverse of symmetric constraint * * \param[in] id Ordering ID (for logging only) * \param[in] kind Ordering kind * \param[in] rsc_first 'First' resource in ordering (a clone) * \param[in] action_first 'First' action in ordering * \param[in,out] rsc_then 'Then' resource in ordering * \param[in] action_then 'Then' action in ordering */ static void inverse_ordering(const char *id, enum pe_order_kind kind, pe_resource_t *rsc_first, const char *action_first, pe_resource_t *rsc_then, const char *action_then) { action_then = invert_action(action_then); action_first = invert_action(action_first); if ((action_then == NULL) || (action_first == NULL)) { pcmk__config_warn("Cannot invert constraint '%s' " "(please specify inverse manually)", id); } else { uint32_t flags = ordering_flags_for_kind(kind, action_first, ordering_symmetric_inverse); handle_restart_type(rsc_then, kind, pe_order_implies_first, flags); pcmk__order_resource_actions(rsc_then, action_then, rsc_first, action_first, flags); } } static void unpack_simple_rsc_order(xmlNode *xml_obj, pe_working_set_t *data_set) { pe_resource_t *rsc_then = NULL; pe_resource_t *rsc_first = NULL; int min_required_before = 0; enum pe_order_kind kind = pe_order_kind_mandatory; - uint32_t cons_weight = pe_order_none; + uint32_t flags = pe_order_none; enum ordering_symmetry symmetry; const char *action_then = NULL; const char *action_first = NULL; const char *id = NULL; CRM_CHECK(xml_obj != NULL, return); id = crm_element_value(xml_obj, XML_ATTR_ID); if (id == NULL) { pcmk__config_err("Ignoring <%s> constraint without " XML_ATTR_ID, crm_element_name(xml_obj)); return; } rsc_first = get_ordering_resource(xml_obj, XML_ORDER_ATTR_FIRST, XML_ORDER_ATTR_FIRST_INSTANCE, data_set); if (rsc_first == NULL) { return; } rsc_then = get_ordering_resource(xml_obj, XML_ORDER_ATTR_THEN, XML_ORDER_ATTR_THEN_INSTANCE, data_set); if (rsc_then == NULL) { return; } action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION); if (action_first == NULL) { action_first = RSC_START; } action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION); if (action_then == NULL) { action_then = action_first; } kind = get_ordering_type(xml_obj); symmetry = get_ordering_symmetry(xml_obj, kind, NULL); - cons_weight = ordering_flags_for_kind(kind, action_first, symmetry); + flags = ordering_flags_for_kind(kind, action_first, symmetry); - handle_restart_type(rsc_then, kind, pe_order_implies_then, cons_weight); + handle_restart_type(rsc_then, kind, pe_order_implies_then, flags); /* If there is a minimum number of instances that must be runnable before * the 'then' action is runnable, we use a pseudo-action for convenience: * minimum number of clone instances have runnable actions -> * pseudo-action is runnable -> dependency is runnable. */ min_required_before = get_minimum_first_instances(rsc_first, xml_obj); if (min_required_before > 0) { clone_min_ordering(id, rsc_first, action_first, rsc_then, action_then, - cons_weight, min_required_before, data_set); + flags, min_required_before, data_set); } else { pcmk__order_resource_actions(rsc_first, action_first, rsc_then, - action_then, cons_weight); + action_then, flags); } if (symmetry == ordering_symmetric) { inverse_ordering(id, kind, rsc_first, action_first, rsc_then, action_then); } } /*! * \internal * \brief Create a new ordering between two actions * * \param[in,out] first_rsc Resource for 'first' action (if NULL and * \p first_action is a resource action, that * resource will be used) * \param[in,out] first_action_task Action key for 'first' action (if NULL and * \p first_action is not NULL, its UUID will * be used) * \param[in,out] first_action 'first' action (if NULL, \p first_rsc and * \p first_action_task must be set) * * \param[in] then_rsc Resource for 'then' action (if NULL and * \p then_action is a resource action, that * resource will be used) * \param[in,out] then_action_task Action key for 'then' action (if NULL and * \p then_action is not NULL, its UUID will * be used) * \param[in] then_action 'then' action (if NULL, \p then_rsc and * \p then_action_task must be set) * * \param[in] flags Flag set of enum pe_ordering * \param[in,out] data_set Cluster working set to add ordering to * * \note This function takes ownership of first_action_task and * then_action_task, which do not need to be freed by the caller. */ void pcmk__new_ordering(pe_resource_t *first_rsc, char *first_action_task, pe_action_t *first_action, pe_resource_t *then_rsc, char *then_action_task, pe_action_t *then_action, uint32_t flags, pe_working_set_t *data_set) { pe__ordering_t *order = NULL; // One of action or resource must be specified for each side CRM_CHECK(((first_action != NULL) || (first_rsc != NULL)) && ((then_action != NULL) || (then_rsc != NULL)), free(first_action_task); free(then_action_task); return); if ((first_rsc == NULL) && (first_action != NULL)) { first_rsc = first_action->rsc; } if ((then_rsc == NULL) && (then_action != NULL)) { then_rsc = then_action->rsc; } order = calloc(1, sizeof(pe__ordering_t)); CRM_ASSERT(order != NULL); order->id = data_set->order_id++; order->flags = flags; order->lh_rsc = first_rsc; order->rh_rsc = then_rsc; order->lh_action = first_action; order->rh_action = then_action; order->lh_action_task = first_action_task; order->rh_action_task = then_action_task; if ((order->lh_action_task == NULL) && (first_action != NULL)) { order->lh_action_task = strdup(first_action->uuid); } if ((order->rh_action_task == NULL) && (then_action != NULL)) { order->rh_action_task = strdup(then_action->uuid); } if ((order->lh_rsc == NULL) && (first_action != NULL)) { order->lh_rsc = first_action->rsc; } if ((order->rh_rsc == NULL) && (then_action != NULL)) { order->rh_rsc = then_action->rsc; } pe_rsc_trace(first_rsc, "Created ordering %d for %s then %s", (data_set->order_id - 1), pcmk__s(order->lh_action_task, "an underspecified action"), pcmk__s(order->rh_action_task, "an underspecified action")); data_set->ordering_constraints = g_list_prepend(data_set->ordering_constraints, order); pcmk__order_migration_equivalents(order); } /*! * \brief Unpack a set in an ordering constraint * * \param[in] set Set XML to unpack * \param[in] parent_kind rsc_order XML "kind" attribute * \param[in] parent_symmetrical_s rsc_order XML "symmetrical" attribute * \param[in,out] data_set Cluster working set * * \return Standard Pacemaker return code */ static int unpack_order_set(const xmlNode *set, enum pe_order_kind parent_kind, const char *parent_symmetrical_s, pe_working_set_t *data_set) { GList *set_iter = NULL; GList *resources = NULL; pe_resource_t *last = NULL; pe_resource_t *resource = NULL; int local_kind = parent_kind; bool sequential = false; uint32_t flags = pe_order_optional; enum ordering_symmetry symmetry; char *key = NULL; const char *id = ID(set); const char *action = crm_element_value(set, "action"); const char *sequential_s = crm_element_value(set, "sequential"); const char *kind_s = crm_element_value(set, XML_ORDER_ATTR_KIND); if (action == NULL) { action = RSC_START; } if (kind_s) { local_kind = get_ordering_type(set); } if (sequential_s == NULL) { sequential_s = "1"; } sequential = crm_is_true(sequential_s); symmetry = get_ordering_symmetry(set, parent_kind, parent_symmetrical_s); flags = ordering_flags_for_kind(local_kind, action, symmetry); for (const xmlNode *xml_rsc = first_named_child(set, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, resource, ID(xml_rsc)); resources = g_list_append(resources, resource); } if (pcmk__list_of_1(resources)) { crm_trace("Single set: %s", id); goto done; } set_iter = resources; while (set_iter != NULL) { resource = (pe_resource_t *) set_iter->data; set_iter = set_iter->next; key = pcmk__op_key(resource->id, action, 0); if (local_kind == pe_order_kind_serialize) { /* Serialize before everything that comes after */ for (GList *gIter = set_iter; gIter != NULL; gIter = gIter->next) { pe_resource_t *then_rsc = (pe_resource_t *) gIter->data; char *then_key = pcmk__op_key(then_rsc->id, action, 0); pcmk__new_ordering(resource, strdup(key), NULL, then_rsc, then_key, NULL, flags, data_set); } } else if (sequential) { if (last != NULL) { pcmk__order_resource_actions(last, action, resource, action, flags); } last = resource; } free(key); } if (symmetry == ordering_asymmetric) { goto done; } last = NULL; action = invert_action(action); flags = ordering_flags_for_kind(local_kind, action, ordering_symmetric_inverse); set_iter = resources; while (set_iter != NULL) { resource = (pe_resource_t *) set_iter->data; set_iter = set_iter->next; if (sequential) { if (last != NULL) { pcmk__order_resource_actions(resource, action, last, action, flags); } last = resource; } } done: g_list_free(resources); return pcmk_rc_ok; } /*! * \brief Order two resource sets relative to each other * * \param[in] id Ordering ID (for logging) * \param[in] set1 First listed set * \param[in] set2 Second listed set * \param[in] kind Ordering kind * \param[in,out] data_set Cluster working set * \param[in] symmetry Which ordering symmetry applies to this relation * * \return Standard Pacemaker return code */ static int order_rsc_sets(const char *id, const xmlNode *set1, const xmlNode *set2, enum pe_order_kind kind, pe_working_set_t *data_set, enum ordering_symmetry symmetry) { const xmlNode *xml_rsc = NULL; const xmlNode *xml_rsc_2 = NULL; pe_resource_t *rsc_1 = NULL; pe_resource_t *rsc_2 = NULL; const char *action_1 = crm_element_value(set1, "action"); const char *action_2 = crm_element_value(set2, "action"); uint32_t flags = pe_order_none; bool require_all = true; (void) pcmk__xe_get_bool_attr(set1, "require-all", &require_all); if (action_1 == NULL) { action_1 = RSC_START; } if (action_2 == NULL) { action_2 = RSC_START; } if (symmetry == ordering_symmetric_inverse) { action_1 = invert_action(action_1); action_2 = invert_action(action_2); } if (pcmk__str_eq(RSC_STOP, action_1, pcmk__str_casei) || pcmk__str_eq(RSC_DEMOTE, action_1, pcmk__str_casei)) { /* Assuming: A -> ( B || C) -> D * The one-or-more logic only applies during the start/promote phase. * During shutdown neither B nor can shutdown until D is down, so simply * turn require_all back on. */ require_all = true; } flags = ordering_flags_for_kind(kind, action_1, symmetry); /* If we have an unordered set1, whether it is sequential or not is * irrelevant in regards to set2. */ if (!require_all) { char *task = crm_strdup_printf(CRM_OP_RELAXED_SET ":%s", ID(set1)); pe_action_t *unordered_action = get_pseudo_op(task, data_set); free(task); pe__set_action_flags(unordered_action, pe_action_requires_any); for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); /* Add an ordering constraint between every element in set1 and the * pseudo action. If any action in set1 is runnable the pseudo * action will be runnable. */ pcmk__new_ordering(rsc_1, pcmk__op_key(rsc_1->id, action_1, 0), NULL, NULL, NULL, unordered_action, pe_order_one_or_more|pe_order_implies_then_printed, data_set); } for (xml_rsc_2 = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc_2 != NULL; xml_rsc_2 = crm_next_same_xml(xml_rsc_2)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); /* Add an ordering constraint between the pseudo-action and every * element in set2. If the pseudo-action is runnable, every action * in set2 will be runnable. */ pcmk__new_ordering(NULL, NULL, unordered_action, rsc_2, pcmk__op_key(rsc_2->id, action_2, 0), NULL, flags|pe_order_runnable_left, data_set); } return pcmk_rc_ok; } if (pcmk__xe_attr_is_true(set1, "sequential")) { if (symmetry == ordering_symmetric_inverse) { // Get the first one xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); if (xml_rsc != NULL) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); } } else { // Get the last one const char *rid = NULL; for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { rid = ID(xml_rsc); } EXPAND_CONSTRAINT_IDREF(id, rsc_1, rid); } } if (pcmk__xe_attr_is_true(set2, "sequential")) { if (symmetry == ordering_symmetric_inverse) { // Get the last one const char *rid = NULL; for (xml_rsc = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { rid = ID(xml_rsc); } EXPAND_CONSTRAINT_IDREF(id, rsc_2, rid); } else { // Get the first one xml_rsc = first_named_child(set2, XML_TAG_RESOURCE_REF); if (xml_rsc != NULL) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); } } } if ((rsc_1 != NULL) && (rsc_2 != NULL)) { pcmk__order_resource_actions(rsc_1, action_1, rsc_2, action_2, flags); } else if (rsc_1 != NULL) { for (xml_rsc = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc)); pcmk__order_resource_actions(rsc_1, action_1, rsc_2, action_2, flags); } } else if (rsc_2 != NULL) { for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); pcmk__order_resource_actions(rsc_1, action_1, rsc_2, action_2, flags); } } else { for (xml_rsc = first_named_child(set1, XML_TAG_RESOURCE_REF); xml_rsc != NULL; xml_rsc = crm_next_same_xml(xml_rsc)) { EXPAND_CONSTRAINT_IDREF(id, rsc_1, ID(xml_rsc)); for (xmlNode *xml_rsc_2 = first_named_child(set2, XML_TAG_RESOURCE_REF); xml_rsc_2 != NULL; xml_rsc_2 = crm_next_same_xml(xml_rsc_2)) { EXPAND_CONSTRAINT_IDREF(id, rsc_2, ID(xml_rsc_2)); pcmk__order_resource_actions(rsc_1, action_1, rsc_2, action_2, flags); } } } return pcmk_rc_ok; } /*! * \internal * \brief If an ordering constraint uses resource tags, expand them * * \param[in,out] xml_obj Ordering constraint XML * \param[out] expanded_xml Equivalent XML with tags expanded * \param[in] data_set Cluster working set * * \return Standard Pacemaker return code (specifically, pcmk_rc_ok on success, * and pcmk_rc_unpack_error on invalid configuration) */ static int unpack_order_tags(xmlNode *xml_obj, xmlNode **expanded_xml, const pe_working_set_t *data_set) { const char *id_first = NULL; const char *id_then = NULL; const char *action_first = NULL; const char *action_then = NULL; pe_resource_t *rsc_first = NULL; pe_resource_t *rsc_then = NULL; pe_tag_t *tag_first = NULL; pe_tag_t *tag_then = NULL; xmlNode *rsc_set_first = NULL; xmlNode *rsc_set_then = NULL; bool any_sets = false; // Check whether there are any resource sets with template or tag references *expanded_xml = pcmk__expand_tags_in_sets(xml_obj, data_set); if (*expanded_xml != NULL) { crm_log_xml_trace(*expanded_xml, "Expanded rsc_order"); return pcmk_rc_ok; } id_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST); id_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN); if ((id_first == NULL) || (id_then == NULL)) { return pcmk_rc_ok; } if (!pcmk__valid_resource_or_tag(data_set, id_first, &rsc_first, &tag_first)) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", ID(xml_obj), id_first); return pcmk_rc_unpack_error; } if (!pcmk__valid_resource_or_tag(data_set, id_then, &rsc_then, &tag_then)) { pcmk__config_err("Ignoring constraint '%s' because '%s' is not a " "valid resource or tag", ID(xml_obj), id_then); return pcmk_rc_unpack_error; } if ((rsc_first != NULL) && (rsc_then != NULL)) { // Neither side references a template or tag return pcmk_rc_ok; } action_first = crm_element_value(xml_obj, XML_ORDER_ATTR_FIRST_ACTION); action_then = crm_element_value(xml_obj, XML_ORDER_ATTR_THEN_ACTION); *expanded_xml = copy_xml(xml_obj); // Convert template/tag reference in "first" into resource_set under constraint if (!pcmk__tag_to_set(*expanded_xml, &rsc_set_first, XML_ORDER_ATTR_FIRST, true, data_set)) { free_xml(*expanded_xml); *expanded_xml = NULL; return pcmk_rc_unpack_error; } if (rsc_set_first != NULL) { if (action_first != NULL) { // Move "first-action" into converted resource_set as "action" crm_xml_add(rsc_set_first, "action", action_first); xml_remove_prop(*expanded_xml, XML_ORDER_ATTR_FIRST_ACTION); } any_sets = true; } // Convert template/tag reference in "then" into resource_set under constraint if (!pcmk__tag_to_set(*expanded_xml, &rsc_set_then, XML_ORDER_ATTR_THEN, true, data_set)) { free_xml(*expanded_xml); *expanded_xml = NULL; return pcmk_rc_unpack_error; } if (rsc_set_then != NULL) { if (action_then != NULL) { // Move "then-action" into converted resource_set as "action" crm_xml_add(rsc_set_then, "action", action_then); xml_remove_prop(*expanded_xml, XML_ORDER_ATTR_THEN_ACTION); } any_sets = true; } if (any_sets) { crm_log_xml_trace(*expanded_xml, "Expanded rsc_order"); } else { free_xml(*expanded_xml); *expanded_xml = NULL; } return pcmk_rc_ok; } /*! * \internal * \brief Unpack ordering constraint XML * * \param[in,out] xml_obj Ordering constraint XML to unpack * \param[in,out] data_set Cluster working set */ void pcmk__unpack_ordering(xmlNode *xml_obj, pe_working_set_t *data_set) { xmlNode *set = NULL; xmlNode *last = NULL; xmlNode *orig_xml = NULL; xmlNode *expanded_xml = NULL; const char *id = crm_element_value(xml_obj, XML_ATTR_ID); const char *invert = crm_element_value(xml_obj, XML_CONS_ATTR_SYMMETRICAL); enum pe_order_kind kind = get_ordering_type(xml_obj); enum ordering_symmetry symmetry = get_ordering_symmetry(xml_obj, kind, NULL); // Expand any resource tags in the constraint XML if (unpack_order_tags(xml_obj, &expanded_xml, data_set) != pcmk_rc_ok) { return; } if (expanded_xml != NULL) { orig_xml = xml_obj; xml_obj = expanded_xml; } // If the constraint has resource sets, unpack them for (set = first_named_child(xml_obj, XML_CONS_TAG_RSC_SET); set != NULL; set = crm_next_same_xml(set)) { set = expand_idref(set, data_set->input); if ((set == NULL) // Configuration error, message already logged || (unpack_order_set(set, kind, invert, data_set) != pcmk_rc_ok)) { if (expanded_xml != NULL) { free_xml(expanded_xml); } return; } if (last != NULL) { if (order_rsc_sets(id, last, set, kind, data_set, symmetry) != pcmk_rc_ok) { if (expanded_xml != NULL) { free_xml(expanded_xml); } return; } if ((symmetry == ordering_symmetric) && (order_rsc_sets(id, set, last, kind, data_set, ordering_symmetric_inverse) != pcmk_rc_ok)) { if (expanded_xml != NULL) { free_xml(expanded_xml); } return; } } last = set; } if (expanded_xml) { free_xml(expanded_xml); xml_obj = orig_xml; } // If the constraint has no resource sets, unpack it as a simple ordering if (last == NULL) { return unpack_simple_rsc_order(xml_obj, data_set); } } static bool ordering_is_invalid(pe_action_t *action, pe_action_wrapper_t *input) { /* Prevent user-defined ordering constraints between resources * running in a guest node and the resource that defines that node. */ if (!pcmk_is_set(input->type, pe_order_preserve) && (input->action->rsc != NULL) && pcmk__rsc_corresponds_to_guest(action->rsc, input->action->node)) { crm_warn("Invalid ordering constraint between %s and %s", input->action->rsc->id, action->rsc->id); return true; } /* If there's an order like * "rscB_stop node2"-> "load_stopped_node2" -> "rscA_migrate_to node1" * * then rscA is being migrated from node1 to node2, while rscB is being * migrated from node2 to node1. If there would be a graph loop, * break the order "load_stopped_node2" -> "rscA_migrate_to node1". */ if ((input->type == pe_order_load) && action->rsc && pcmk__str_eq(action->task, RSC_MIGRATE, pcmk__str_casei) && pcmk__graph_has_loop(action, action, input)) { return true; } return false; } void pcmk__disable_invalid_orderings(pe_working_set_t *data_set) { for (GList *iter = data_set->actions; iter != NULL; iter = iter->next) { pe_action_t *action = (pe_action_t *) iter->data; pe_action_wrapper_t *input = NULL; for (GList *input_iter = action->actions_before; input_iter != NULL; input_iter = input_iter->next) { input = (pe_action_wrapper_t *) input_iter->data; if (ordering_is_invalid(action, input)) { input->type = pe_order_none; } } } } /*! * \internal * \brief Order stops on a node before the node's shutdown * * \param[in,out] node Node being shut down * \param[in] shutdown_op Shutdown action for node */ void pcmk__order_stops_before_shutdown(pe_node_t *node, pe_action_t *shutdown_op) { for (GList *iter = node->details->data_set->actions; iter != NULL; iter = iter->next) { pe_action_t *action = (pe_action_t *) iter->data; // Only stops on the node shutting down are relevant - if ((action->rsc == NULL) || (action->node == NULL) - || (action->node->details != node->details) + if (!pe__same_node(action->node, node) || !pcmk__str_eq(action->task, RSC_STOP, pcmk__str_casei)) { continue; } // Resources and nodes in maintenance mode won't be touched if (pcmk_is_set(action->rsc->flags, pe_rsc_maintenance)) { pe_rsc_trace(action->rsc, "Not ordering %s before shutdown of %s because " "resource in maintenance mode", action->uuid, pe__node_name(node)); continue; } else if (node->details->maintenance) { pe_rsc_trace(action->rsc, "Not ordering %s before shutdown of %s because " "node in maintenance mode", action->uuid, pe__node_name(node)); continue; } /* Don't touch a resource that is unmanaged or blocked, to avoid * blocking the shutdown (though if another action depends on this one, * we may still end up blocking) */ if (!pcmk_any_flags_set(action->rsc->flags, pe_rsc_managed|pe_rsc_block)) { pe_rsc_trace(action->rsc, "Not ordering %s before shutdown of %s because " "resource is unmanaged or blocked", action->uuid, pe__node_name(node)); continue; } pe_rsc_trace(action->rsc, "Ordering %s before shutdown of %s", action->uuid, pe__node_name(node)); pe__clear_action_flags(action, pe_action_optional); pcmk__new_ordering(action->rsc, NULL, action, NULL, strdup(CRM_OP_SHUTDOWN), shutdown_op, pe_order_optional|pe_order_runnable_left, node->details->data_set); } } /*! * \brief Find resource actions matching directly or as child * * \param[in] rsc Resource to check * \param[in] original_key Action key to search for (possibly referencing * parent of \rsc) * * \return Newly allocated list of matching actions * \note It is the caller's responsibility to free the result with g_list_free() */ static GList * find_actions_by_task(const pe_resource_t *rsc, const char *original_key) { // Search under given task key directly GList *list = find_actions(rsc->actions, original_key, NULL); if (list == NULL) { // Search again using this resource's ID char *key = NULL; char *task = NULL; guint interval_ms = 0; if (parse_op_key(original_key, NULL, &task, &interval_ms)) { key = pcmk__op_key(rsc->id, task, interval_ms); list = find_actions(rsc->actions, key, NULL); free(key); free(task); } else { crm_err("Invalid operation key (bug?): %s", original_key); } } return list; } /*! * \internal * \brief Order relevant resource actions after a given action * * \param[in,out] first_action Action to order after (or NULL if none runnable) * \param[in] rsc Resource whose actions should be ordered * \param[in,out] order Ordering constraint being applied */ static void order_resource_actions_after(pe_action_t *first_action, const pe_resource_t *rsc, pe__ordering_t *order) { GList *then_actions = NULL; uint32_t flags = pe_order_none; CRM_CHECK((rsc != NULL) && (order != NULL), return); flags = order->flags; pe_rsc_trace(rsc, "Applying ordering %d for 'then' resource %s", order->id, rsc->id); if (order->rh_action != NULL) { then_actions = g_list_prepend(NULL, order->rh_action); } else { then_actions = find_actions_by_task(rsc, order->rh_action_task); } if (then_actions == NULL) { pe_rsc_trace(rsc, "Ignoring ordering %d: no %s actions found for %s", order->id, order->rh_action_task, rsc->id); return; } if ((first_action != NULL) && (first_action->rsc == rsc) && pcmk_is_set(first_action->flags, pe_action_dangle)) { pe_rsc_trace(rsc, "Detected dangling migration ordering (%s then %s %s)", first_action->uuid, order->rh_action_task, rsc->id); pe__clear_order_flags(flags, pe_order_implies_then); } if ((first_action == NULL) && !pcmk_is_set(flags, pe_order_implies_then)) { pe_rsc_debug(rsc, "Ignoring ordering %d for %s: No first action found", order->id, rsc->id); g_list_free(then_actions); return; } for (GList *iter = then_actions; iter != NULL; iter = iter->next) { pe_action_t *then_action_iter = (pe_action_t *) iter->data; if (first_action != NULL) { order_actions(first_action, then_action_iter, flags); } else { pe__clear_action_flags(then_action_iter, pe_action_runnable); crm_warn("%s of %s is unrunnable because there is no %s of %s " "to order it after", then_action_iter->task, rsc->id, order->lh_action_task, order->lh_rsc->id); } } g_list_free(then_actions); } static void rsc_order_first(pe_resource_t *first_rsc, pe__ordering_t *order, pe_working_set_t *data_set) { GList *first_actions = NULL; pe_action_t *first_action = order->lh_action; pe_resource_t *then_rsc = order->rh_rsc; CRM_ASSERT(first_rsc != NULL); pe_rsc_trace(first_rsc, "Applying ordering constraint %d (first: %s)", order->id, first_rsc->id); if (first_action != NULL) { first_actions = g_list_prepend(NULL, first_action); } else { first_actions = find_actions_by_task(first_rsc, order->lh_action_task); } if ((first_actions == NULL) && (first_rsc == then_rsc)) { pe_rsc_trace(first_rsc, "Ignoring constraint %d: first (%s for %s) not found", order->id, order->lh_action_task, first_rsc->id); } else if (first_actions == NULL) { char *key = NULL; char *op_type = NULL; guint interval_ms = 0; parse_op_key(order->lh_action_task, NULL, &op_type, &interval_ms); key = pcmk__op_key(first_rsc->id, op_type, interval_ms); if ((first_rsc->fns->state(first_rsc, TRUE) == RSC_ROLE_STOPPED) && pcmk__str_eq(op_type, RSC_STOP, pcmk__str_casei)) { free(key); pe_rsc_trace(first_rsc, "Ignoring constraint %d: first (%s for %s) not found", order->id, order->lh_action_task, first_rsc->id); } else if ((first_rsc->fns->state(first_rsc, TRUE) == RSC_ROLE_UNPROMOTED) && pcmk__str_eq(op_type, RSC_DEMOTE, pcmk__str_casei)) { free(key); pe_rsc_trace(first_rsc, "Ignoring constraint %d: first (%s for %s) not found", order->id, order->lh_action_task, first_rsc->id); } else { pe_rsc_trace(first_rsc, "Creating first (%s for %s) for constraint %d ", order->lh_action_task, first_rsc->id, order->id); first_action = custom_action(first_rsc, key, op_type, NULL, TRUE, TRUE, data_set); first_actions = g_list_prepend(NULL, first_action); } free(op_type); } if (then_rsc == NULL) { if (order->rh_action == NULL) { pe_rsc_trace(first_rsc, "Ignoring constraint %d: then not found", order->id); return; } then_rsc = order->rh_action->rsc; } for (GList *gIter = first_actions; gIter != NULL; gIter = gIter->next) { first_action = (pe_action_t *) gIter->data; if (then_rsc == NULL) { order_actions(first_action, order->rh_action, order->flags); } else { order_resource_actions_after(first_action, then_rsc, order); } } g_list_free(first_actions); } void pcmk__apply_orderings(pe_working_set_t *data_set) { crm_trace("Applying ordering constraints"); /* Ordering constraints need to be processed in the order they were created. * rsc_order_first() and order_resource_actions_after() require the relevant * actions to already exist in some cases, but rsc_order_first() will create * the 'first' action in certain cases. Thus calling rsc_order_first() can * change the behavior of later-created orderings. * * Also, g_list_append() should be avoided for performance reasons, so we * prepend orderings when creating them and reverse the list here. * * @TODO This is brittle and should be carefully redesigned so that the * order of creation doesn't matter, and the reverse becomes unneeded. */ data_set->ordering_constraints = g_list_reverse(data_set->ordering_constraints); for (GList *gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) { pe__ordering_t *order = gIter->data; pe_resource_t *rsc = order->lh_rsc; if (rsc != NULL) { rsc_order_first(rsc, order, data_set); continue; } rsc = order->rh_rsc; if (rsc != NULL) { order_resource_actions_after(order->lh_action, rsc, order); } else { crm_trace("Applying ordering constraint %d (non-resource actions)", order->id); order_actions(order->lh_action, order->rh_action, order->flags); } } g_list_foreach(data_set->actions, (GFunc) pcmk__block_colocation_dependents, data_set); crm_trace("Ordering probes"); pcmk__order_probes(data_set); crm_trace("Updating %d actions", g_list_length(data_set->actions)); g_list_foreach(data_set->actions, (GFunc) pcmk__update_action_for_orderings, data_set); pcmk__disable_invalid_orderings(data_set); } /*! * \internal * \brief Order a given action after each action in a given list * * \param[in,out] after "After" action * \param[in,out] list List of "before" actions */ void pcmk__order_after_each(pe_action_t *after, GList *list) { const char *after_desc = (after->task == NULL)? after->uuid : after->task; for (GList *iter = list; iter != NULL; iter = iter->next) { pe_action_t *before = (pe_action_t *) iter->data; const char *before_desc = before->task? before->task : before->uuid; crm_debug("Ordering %s on %s before %s on %s", before_desc, pe__node_name(before->node), after_desc, pe__node_name(after->node)); order_actions(before, after, pe_order_optional); } } /*! * \internal * \brief Order promotions and demotions for restarts of a clone or bundle * * \param[in,out] rsc Clone or bundle to order */ void pcmk__promotable_restart_ordering(pe_resource_t *rsc) { // Order start and promote after all instances are stopped pcmk__order_resource_actions(rsc, RSC_STOPPED, rsc, RSC_START, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_STOPPED, rsc, RSC_PROMOTE, pe_order_optional); // Order stop, start, and promote after all instances are demoted pcmk__order_resource_actions(rsc, RSC_DEMOTED, rsc, RSC_STOP, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_DEMOTED, rsc, RSC_START, pe_order_optional); pcmk__order_resource_actions(rsc, RSC_DEMOTED, rsc, RSC_PROMOTE, pe_order_optional); // Order promote after all instances are started pcmk__order_resource_actions(rsc, RSC_STARTED, rsc, RSC_PROMOTE, pe_order_optional); // Order demote after all instances are demoted pcmk__order_resource_actions(rsc, RSC_DEMOTE, rsc, RSC_DEMOTED, pe_order_optional); } diff --git a/lib/pacemaker/pcmk_sched_primitive.c b/lib/pacemaker/pcmk_sched_primitive.c index e34133a5ef..b78e5c1f68 100644 --- a/lib/pacemaker/pcmk_sched_primitive.c +++ b/lib/pacemaker/pcmk_sched_primitive.c @@ -1,1573 +1,1573 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include "libpacemaker_private.h" static void stop_resource(pe_resource_t *rsc, pe_node_t *node, bool optional); static void start_resource(pe_resource_t *rsc, pe_node_t *node, bool optional); static void demote_resource(pe_resource_t *rsc, pe_node_t *node, bool optional); static void promote_resource(pe_resource_t *rsc, pe_node_t *node, bool optional); static void assert_role_error(pe_resource_t *rsc, pe_node_t *node, bool optional); static enum rsc_role_e rsc_state_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX] = { /* This array lists the immediate next role when transitioning from one role * to a target role. For example, when going from Stopped to Promoted, the * next role is Unpromoted, because the resource must be started before it * can be promoted. The current state then becomes Started, which is fed * into this array again, giving a next role of Promoted. * * Current role Immediate next role Final target role * ------------ ------------------- ----------------- */ /* Unknown */ { RSC_ROLE_UNKNOWN, /* Unknown */ RSC_ROLE_STOPPED, /* Stopped */ RSC_ROLE_STOPPED, /* Started */ RSC_ROLE_STOPPED, /* Unpromoted */ RSC_ROLE_STOPPED, /* Promoted */ }, /* Stopped */ { RSC_ROLE_STOPPED, /* Unknown */ RSC_ROLE_STOPPED, /* Stopped */ RSC_ROLE_STARTED, /* Started */ RSC_ROLE_UNPROMOTED, /* Unpromoted */ RSC_ROLE_UNPROMOTED, /* Promoted */ }, /* Started */ { RSC_ROLE_STOPPED, /* Unknown */ RSC_ROLE_STOPPED, /* Stopped */ RSC_ROLE_STARTED, /* Started */ RSC_ROLE_UNPROMOTED, /* Unpromoted */ RSC_ROLE_PROMOTED, /* Promoted */ }, /* Unpromoted */ { RSC_ROLE_STOPPED, /* Unknown */ RSC_ROLE_STOPPED, /* Stopped */ RSC_ROLE_STOPPED, /* Started */ RSC_ROLE_UNPROMOTED, /* Unpromoted */ RSC_ROLE_PROMOTED, /* Promoted */ }, /* Promoted */ { RSC_ROLE_STOPPED, /* Unknown */ RSC_ROLE_UNPROMOTED, /* Stopped */ RSC_ROLE_UNPROMOTED, /* Started */ RSC_ROLE_UNPROMOTED, /* Unpromoted */ RSC_ROLE_PROMOTED, /* Promoted */ }, }; /*! * \internal * \brief Function to schedule actions needed for a role change * * \param[in,out] rsc Resource whose role is changing * \param[in,out] node Node where resource will be in its next role * \param[in] optional Whether scheduled actions should be optional */ typedef void (*rsc_transition_fn)(pe_resource_t *rsc, pe_node_t *node, bool optional); static rsc_transition_fn rsc_action_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX] = { /* This array lists the function needed to transition directly from one role * to another. NULL indicates that nothing is needed. * * Current role Transition function Next role * ------------ ------------------- ---------- */ /* Unknown */ { assert_role_error, /* Unknown */ stop_resource, /* Stopped */ assert_role_error, /* Started */ assert_role_error, /* Unpromoted */ assert_role_error, /* Promoted */ }, /* Stopped */ { assert_role_error, /* Unknown */ NULL, /* Stopped */ start_resource, /* Started */ start_resource, /* Unpromoted */ assert_role_error, /* Promoted */ }, /* Started */ { assert_role_error, /* Unknown */ stop_resource, /* Stopped */ NULL, /* Started */ NULL, /* Unpromoted */ promote_resource, /* Promoted */ }, /* Unpromoted */ { assert_role_error, /* Unknown */ stop_resource, /* Stopped */ stop_resource, /* Started */ NULL, /* Unpromoted */ promote_resource, /* Promoted */ }, /* Promoted */ { assert_role_error, /* Unknown */ demote_resource, /* Stopped */ demote_resource, /* Started */ demote_resource, /* Unpromoted */ NULL, /* Promoted */ }, }; /*! * \internal - * \brief Get a list of a resource's allowed nodes sorted by node weight + * \brief Get a list of a resource's allowed nodes sorted by node score * * \param[in] rsc Resource to check * - * \return List of allowed nodes sorted by node weight + * \return List of allowed nodes sorted by node score */ static GList * sorted_allowed_nodes(const pe_resource_t *rsc) { if (rsc->allowed_nodes != NULL) { GList *nodes = g_hash_table_get_values(rsc->allowed_nodes); if (nodes != NULL) { return pcmk__sort_nodes(nodes, pe__current_node(rsc)); } } return NULL; } /*! * \internal * \brief Assign a resource to its best allowed node, if possible * * \param[in,out] rsc Resource to choose a node for * \param[in] prefer If not NULL, prefer this node when all else equal * * \return true if \p rsc could be assigned to a node, otherwise false */ static bool assign_best_node(pe_resource_t *rsc, const pe_node_t *prefer) { GList *nodes = NULL; pe_node_t *chosen = NULL; pe_node_t *best = NULL; bool result = false; const pe_node_t *most_free_node = pcmk__ban_insufficient_capacity(rsc); if (prefer == NULL) { prefer = most_free_node; } if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { // We've already finished assignment of resources to nodes return rsc->allocated_to != NULL; } - // Sort allowed nodes by weight + // Sort allowed nodes by score nodes = sorted_allowed_nodes(rsc); if (nodes != NULL) { best = (pe_node_t *) nodes->data; // First node has best score } if ((prefer != NULL) && (nodes != NULL)) { // Get the allowed node version of prefer chosen = g_hash_table_lookup(rsc->allowed_nodes, prefer->details->id); if (chosen == NULL) { pe_rsc_trace(rsc, "Preferred node %s for %s was unknown", pe__node_name(prefer), rsc->id); - /* Favor the preferred node as long as its weight is at least as good as + /* Favor the preferred node as long as its score is at least as good as * the best allowed node's. * * An alternative would be to favor the preferred node even if the best - * node is better, when the best node's weight is less than INFINITY. + * node is better, when the best node's score is less than INFINITY. */ } else if (chosen->weight < best->weight) { pe_rsc_trace(rsc, "Preferred node %s for %s was unsuitable", pe__node_name(chosen), rsc->id); chosen = NULL; } else if (!pcmk__node_available(chosen, true, false)) { pe_rsc_trace(rsc, "Preferred node %s for %s was unavailable", pe__node_name(chosen), rsc->id); chosen = NULL; } else { pe_rsc_trace(rsc, "Chose preferred node %s for %s (ignoring %d candidates)", pe__node_name(chosen), rsc->id, g_list_length(nodes)); } } if ((chosen == NULL) && (best != NULL)) { /* Either there is no preferred node, or the preferred node is not * suitable, but another node is allowed to run the resource. */ chosen = best; if (!pe_rsc_is_unique_clone(rsc->parent) && (chosen->weight > 0) // Zero not acceptable && pcmk__node_available(chosen, false, false)) { /* If the resource is already running on a node, prefer that node if * it is just as good as the chosen node. * * We don't do this for unique clone instances, because * pcmk__assign_instances() has already assigned instances to their * running nodes when appropriate, and if we get here, we don't want * remaining unassigned instances to prefer a node that's already * running another instance. */ pe_node_t *running = pe__current_node(rsc); if (running == NULL) { // Nothing to do } else if (!pcmk__node_available(running, true, false)) { pe_rsc_trace(rsc, "Current node for %s (%s) can't run resources", rsc->id, pe__node_name(running)); } else { int nodes_with_best_score = 1; for (GList *iter = nodes->next; iter; iter = iter->next) { pe_node_t *allowed = (pe_node_t *) iter->data; if (allowed->weight != chosen->weight) { - // The nodes are sorted by weight, so no more are equal + // The nodes are sorted by score, so no more are equal break; } if (pe__same_node(allowed, running)) { // Scores are equal, so prefer the current node chosen = allowed; } nodes_with_best_score++; } if (nodes_with_best_score > 1) { do_crm_log(((chosen->weight >= INFINITY)? LOG_WARNING : LOG_INFO), "Chose %s for %s from %d nodes with score %s", pe__node_name(chosen), rsc->id, nodes_with_best_score, pcmk_readable_score(chosen->weight)); } } } pe_rsc_trace(rsc, "Chose %s for %s from %d candidates", pe__node_name(chosen), rsc->id, g_list_length(nodes)); } result = pcmk__finalize_assignment(rsc, chosen, false); g_list_free(nodes); return result; } /*! * \internal * \brief Apply a "this with" colocation to a node's allowed node scores * * \param[in,out] data Colocation to apply * \param[in,out] user_data Resource being assigned */ static void apply_this_with(gpointer data, gpointer user_data) { pcmk__colocation_t *colocation = (pcmk__colocation_t *) data; pe_resource_t *rsc = (pe_resource_t *) user_data; GHashTable *archive = NULL; pe_resource_t *other = colocation->primary; // In certain cases, we will need to revert the node scores if ((colocation->dependent_role >= RSC_ROLE_PROMOTED) || ((colocation->score < 0) && (colocation->score > -INFINITY))) { archive = pcmk__copy_node_table(rsc->allowed_nodes); } if (pcmk_is_set(other->flags, pe_rsc_provisional)) { pe_rsc_trace(rsc, "%s: Assigning colocation %s primary %s first" "(score=%d role=%s)", rsc->id, colocation->id, other->id, colocation->score, role2text(colocation->dependent_role)); other->cmds->assign(other, NULL); } // Apply the colocation score to this resource's allowed node scores rsc->cmds->apply_coloc_score(rsc, other, colocation, true); if ((archive != NULL) && !pcmk__any_node_available(rsc->allowed_nodes)) { pe_rsc_info(rsc, "%s: Reverting scores from colocation with %s " "because no nodes allowed", rsc->id, other->id); g_hash_table_destroy(rsc->allowed_nodes); rsc->allowed_nodes = archive; archive = NULL; } if (archive != NULL) { g_hash_table_destroy(archive); } } /*! * \internal * \brief Update a Pacemaker Remote node once its connection has been assigned * * \param[in] connection Connection resource that has been assigned */ static void remote_connection_assigned(const pe_resource_t *connection) { pe_node_t *remote_node = pe_find_node(connection->cluster->nodes, connection->id); CRM_CHECK(remote_node != NULL, return); if ((connection->allocated_to != NULL) && (connection->next_role != RSC_ROLE_STOPPED)) { crm_trace("Pacemaker Remote node %s will be online", remote_node->details->id); remote_node->details->online = TRUE; if (remote_node->details->unseen) { // Avoid unnecessary fence, since we will attempt connection remote_node->details->unclean = FALSE; } } else { crm_trace("Pacemaker Remote node %s will be shut down " "(%sassigned connection's next role is %s)", remote_node->details->id, ((connection->allocated_to == NULL)? "un" : ""), role2text(connection->next_role)); remote_node->details->shutdown = TRUE; } } /*! * \internal * \brief Assign a primitive resource to a node * * \param[in,out] rsc Resource to assign to a node * \param[in] prefer Node to prefer, if all else is equal * * \return Node that \p rsc is assigned to, if assigned entirely to one node */ pe_node_t * pcmk__primitive_assign(pe_resource_t *rsc, const pe_node_t *prefer) { GList *this_with_colocations = NULL; GList *with_this_colocations = NULL; GList *iter = NULL; pcmk__colocation_t *colocation = NULL; CRM_ASSERT(rsc != NULL); // Never assign a child without parent being assigned first if ((rsc->parent != NULL) && !pcmk_is_set(rsc->parent->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "%s: Assigning parent %s first", rsc->id, rsc->parent->id); rsc->parent->cmds->assign(rsc->parent, prefer); } if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return rsc->allocated_to; // Assignment has already been done } // Ensure we detect assignment loops if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { pe_rsc_debug(rsc, "Breaking assignment loop involving %s", rsc->id); return NULL; } pe__set_resource_flags(rsc, pe_rsc_allocating); - pe__show_node_weights(true, rsc, "Pre-assignment", rsc->allowed_nodes, - rsc->cluster); + pe__show_node_scores(true, rsc, "Pre-assignment", rsc->allowed_nodes, + rsc->cluster); this_with_colocations = pcmk__this_with_colocations(rsc); with_this_colocations = pcmk__with_this_colocations(rsc); // Apply mandatory colocations first, to satisfy as many as possible for (iter = this_with_colocations; iter != NULL; iter = iter->next) { colocation = iter->data; if ((colocation->score <= -CRM_SCORE_INFINITY) || (colocation->score >= CRM_SCORE_INFINITY)) { apply_this_with(iter->data, rsc); } } for (iter = with_this_colocations; iter != NULL; iter = iter->next) { colocation = iter->data; if ((colocation->score <= -CRM_SCORE_INFINITY) || (colocation->score >= CRM_SCORE_INFINITY)) { pcmk__add_dependent_scores(iter->data, rsc); } } - pe__show_node_weights(true, rsc, "Mandatory-colocations", - rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(true, rsc, "Mandatory-colocations", + rsc->allowed_nodes, rsc->cluster); // Then apply optional colocations for (iter = this_with_colocations; iter != NULL; iter = iter->next) { colocation = iter->data; if ((colocation->score > -CRM_SCORE_INFINITY) && (colocation->score < CRM_SCORE_INFINITY)) { apply_this_with(iter->data, rsc); } } for (iter = with_this_colocations; iter != NULL; iter = iter->next) { colocation = iter->data; if ((colocation->score > -CRM_SCORE_INFINITY) && (colocation->score < CRM_SCORE_INFINITY)) { pcmk__add_dependent_scores(iter->data, rsc); } } g_list_free(this_with_colocations); g_list_free(with_this_colocations); if (rsc->next_role == RSC_ROLE_STOPPED) { pe_rsc_trace(rsc, "Banning %s from all nodes because it will be stopped", rsc->id); resource_location(rsc, NULL, -INFINITY, XML_RSC_ATTR_TARGET_ROLE, rsc->cluster); } else if ((rsc->next_role > rsc->role) && !pcmk_is_set(rsc->cluster->flags, pe_flag_have_quorum) && (rsc->cluster->no_quorum_policy == no_quorum_freeze)) { crm_notice("Resource %s cannot be elevated from %s to %s due to " "no-quorum-policy=freeze", rsc->id, role2text(rsc->role), role2text(rsc->next_role)); pe__set_next_role(rsc, rsc->role, "no-quorum-policy=freeze"); } - pe__show_node_weights(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), - rsc, __func__, rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(!pcmk_is_set(rsc->cluster->flags, pe_flag_show_scores), + rsc, __func__, rsc->allowed_nodes, rsc->cluster); // Unmanage resource if fencing is enabled but no device is configured if (pcmk_is_set(rsc->cluster->flags, pe_flag_stonith_enabled) && !pcmk_is_set(rsc->cluster->flags, pe_flag_have_stonith_resource)) { pe__clear_resource_flags(rsc, pe_rsc_managed); } if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { // Unmanaged resources stay on their current node const char *reason = NULL; pe_node_t *assign_to = NULL; pe__set_next_role(rsc, rsc->role, "unmanaged"); assign_to = pe__current_node(rsc); if (assign_to == NULL) { reason = "inactive"; } else if (rsc->role == RSC_ROLE_PROMOTED) { reason = "promoted"; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { reason = "failed"; } else { reason = "active"; } pe_rsc_info(rsc, "Unmanaged resource %s assigned to %s: %s", rsc->id, (assign_to? assign_to->details->uname : "no node"), reason); pcmk__finalize_assignment(rsc, assign_to, true); } else if (pcmk_is_set(rsc->cluster->flags, pe_flag_stop_everything)) { pe_rsc_debug(rsc, "Forcing %s to stop: stop-all-resources", rsc->id); pcmk__finalize_assignment(rsc, NULL, true); } else if (pcmk_is_set(rsc->flags, pe_rsc_provisional) && assign_best_node(rsc, prefer)) { // Assignment successful } else if (rsc->allocated_to == NULL) { if (!pcmk_is_set(rsc->flags, pe_rsc_orphan)) { pe_rsc_info(rsc, "Resource %s cannot run anywhere", rsc->id); } else if (rsc->running_on != NULL) { pe_rsc_info(rsc, "Stopping orphan resource %s", rsc->id); } } else { pe_rsc_debug(rsc, "%s: pre-assigned to %s", rsc->id, pe__node_name(rsc->allocated_to)); } pe__clear_resource_flags(rsc, pe_rsc_allocating); if (rsc->is_remote_node) { remote_connection_assigned(rsc); } return rsc->allocated_to; } /*! * \internal * \brief Schedule actions to bring resource down and back to current role * * \param[in,out] rsc Resource to restart * \param[in,out] current Node that resource should be brought down on * \param[in] need_stop Whether the resource must be stopped * \param[in] need_promote Whether the resource must be promoted * * \return Role that resource would have after scheduled actions are taken */ static void schedule_restart_actions(pe_resource_t *rsc, pe_node_t *current, bool need_stop, bool need_promote) { enum rsc_role_e role = rsc->role; enum rsc_role_e next_role; rsc_transition_fn fn = NULL; pe__set_resource_flags(rsc, pe_rsc_restarting); // Bring resource down to a stop on its current node while (role != RSC_ROLE_STOPPED) { next_role = rsc_state_matrix[role][RSC_ROLE_STOPPED]; pe_rsc_trace(rsc, "Creating %s action to take %s down from %s to %s", (need_stop? "required" : "optional"), rsc->id, role2text(role), role2text(next_role)); fn = rsc_action_matrix[role][next_role]; if (fn == NULL) { break; } fn(rsc, current, !need_stop); role = next_role; } // Bring resource up to its next role on its next node while ((rsc->role <= rsc->next_role) && (role != rsc->role) && !pcmk_is_set(rsc->flags, pe_rsc_block)) { bool required = need_stop; next_role = rsc_state_matrix[role][rsc->role]; if ((next_role == RSC_ROLE_PROMOTED) && need_promote) { required = true; } pe_rsc_trace(rsc, "Creating %s action to take %s up from %s to %s", (required? "required" : "optional"), rsc->id, role2text(role), role2text(next_role)); fn = rsc_action_matrix[role][next_role]; if (fn == NULL) { break; } fn(rsc, rsc->allocated_to, !required); role = next_role; } pe__clear_resource_flags(rsc, pe_rsc_restarting); } /*! * \internal * \brief If a resource's next role is not explicitly specified, set a default * * \param[in,out] rsc Resource to set next role for * * \return "explicit" if next role was explicitly set, otherwise "implicit" */ static const char * set_default_next_role(pe_resource_t *rsc) { if (rsc->next_role != RSC_ROLE_UNKNOWN) { return "explicit"; } if (rsc->allocated_to == NULL) { pe__set_next_role(rsc, RSC_ROLE_STOPPED, "assignment"); } else { pe__set_next_role(rsc, RSC_ROLE_STARTED, "assignment"); } return "implicit"; } /*! * \internal * \brief Create an action to represent an already pending start * * \param[in,out] rsc Resource to create start action for */ static void create_pending_start(pe_resource_t *rsc) { pe_action_t *start = NULL; pe_rsc_trace(rsc, "Creating action for %s to represent already pending start", rsc->id); start = start_action(rsc, rsc->allocated_to, TRUE); pe__set_action_flags(start, pe_action_print_always); } /*! * \internal * \brief Schedule actions needed to take a resource to its next role * * \param[in,out] rsc Resource to schedule actions for */ static void schedule_role_transition_actions(pe_resource_t *rsc) { enum rsc_role_e role = rsc->role; while (role != rsc->next_role) { enum rsc_role_e next_role = rsc_state_matrix[role][rsc->next_role]; rsc_transition_fn fn = NULL; pe_rsc_trace(rsc, "Creating action to take %s from %s to %s (ending at %s)", rsc->id, role2text(role), role2text(next_role), role2text(rsc->next_role)); fn = rsc_action_matrix[role][next_role]; if (fn == NULL) { break; } fn(rsc, rsc->allocated_to, false); role = next_role; } } /*! * \internal * \brief Create all actions needed for a given primitive resource * * \param[in,out] rsc Primitive resource to create actions for */ void pcmk__primitive_create_actions(pe_resource_t *rsc) { bool need_stop = false; bool need_promote = false; bool is_moving = false; bool allow_migrate = false; bool multiply_active = false; pe_node_t *current = NULL; unsigned int num_all_active = 0; unsigned int num_clean_active = 0; const char *next_role_source = NULL; CRM_ASSERT(rsc != NULL); next_role_source = set_default_next_role(rsc); pe_rsc_trace(rsc, "Creating all actions for %s transition from %s to %s " "(%s) on %s", rsc->id, role2text(rsc->role), role2text(rsc->next_role), next_role_source, pe__node_name(rsc->allocated_to)); current = rsc->fns->active_node(rsc, &num_all_active, &num_clean_active); g_list_foreach(rsc->dangling_migrations, pcmk__abort_dangling_migration, rsc); if ((current != NULL) && (rsc->allocated_to != NULL) - && (current->details != rsc->allocated_to->details) + && !pe__same_node(current, rsc->allocated_to) && (rsc->next_role >= RSC_ROLE_STARTED)) { pe_rsc_trace(rsc, "Moving %s from %s to %s", rsc->id, pe__node_name(current), pe__node_name(rsc->allocated_to)); is_moving = true; allow_migrate = pcmk__rsc_can_migrate(rsc, current); // This is needed even if migrating (though I'm not sure why ...) need_stop = true; } // Check whether resource is partially migrated and/or multiply active if ((rsc->partial_migration_source != NULL) && (rsc->partial_migration_target != NULL) && allow_migrate && (num_all_active == 2) && pe__same_node(current, rsc->partial_migration_source) && pe__same_node(rsc->allocated_to, rsc->partial_migration_target)) { /* A partial migration is in progress, and the migration target remains * the same as when the migration began. */ pe_rsc_trace(rsc, "Partial migration of %s from %s to %s will continue", rsc->id, pe__node_name(rsc->partial_migration_source), pe__node_name(rsc->partial_migration_target)); } else if ((rsc->partial_migration_source != NULL) || (rsc->partial_migration_target != NULL)) { // A partial migration is in progress but can't be continued if (num_all_active > 2) { // The resource is migrating *and* multiply active! crm_notice("Forcing recovery of %s because it is migrating " "from %s to %s and possibly active elsewhere", rsc->id, pe__node_name(rsc->partial_migration_source), pe__node_name(rsc->partial_migration_target)); } else { // The migration source or target isn't available crm_notice("Forcing recovery of %s because it can no longer " "migrate from %s to %s", rsc->id, pe__node_name(rsc->partial_migration_source), pe__node_name(rsc->partial_migration_target)); } need_stop = true; rsc->partial_migration_source = rsc->partial_migration_target = NULL; allow_migrate = false; } else if (pcmk_is_set(rsc->flags, pe_rsc_needs_fencing)) { multiply_active = (num_all_active > 1); } else { /* If a resource has "requires" set to nothing or quorum, don't consider * it active on unclean nodes (similar to how all resources behave when * stonith-enabled is false). We can start such resources elsewhere * before fencing completes, and if we considered the resource active on * the failed node, we would attempt recovery for being active on * multiple nodes. */ multiply_active = (num_clean_active > 1); } if (multiply_active) { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); // Resource was (possibly) incorrectly multiply active pe_proc_err("%s resource %s might be active on %u nodes (%s)", pcmk__s(class, "Untyped"), rsc->id, num_all_active, recovery2text(rsc->recovery_type)); crm_notice("See https://wiki.clusterlabs.org/wiki/FAQ" "#Resource_is_Too_Active for more information"); switch (rsc->recovery_type) { case recovery_stop_start: need_stop = true; break; case recovery_stop_unexpected: need_stop = true; // stop_resource() will skip expected node pe__set_resource_flags(rsc, pe_rsc_stop_unexpected); break; default: break; } } else { pe__clear_resource_flags(rsc, pe_rsc_stop_unexpected); } if (pcmk_is_set(rsc->flags, pe_rsc_start_pending)) { create_pending_start(rsc); } if (is_moving) { // Remaining tests are only for resources staying where they are } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { if (pcmk_is_set(rsc->flags, pe_rsc_stop)) { need_stop = true; pe_rsc_trace(rsc, "Recovering %s", rsc->id); } else { pe_rsc_trace(rsc, "Recovering %s by demotion", rsc->id); if (rsc->next_role == RSC_ROLE_PROMOTED) { need_promote = true; } } } else if (pcmk_is_set(rsc->flags, pe_rsc_block)) { pe_rsc_trace(rsc, "Blocking further actions on %s", rsc->id); need_stop = true; } else if ((rsc->role > RSC_ROLE_STARTED) && (current != NULL) && (rsc->allocated_to != NULL)) { pe_action_t *start = NULL; pe_rsc_trace(rsc, "Creating start action for promoted resource %s", rsc->id); start = start_action(rsc, rsc->allocated_to, TRUE); if (!pcmk_is_set(start->flags, pe_action_optional)) { // Recovery of a promoted resource pe_rsc_trace(rsc, "%s restart is required for recovery", rsc->id); need_stop = true; } } // Create any actions needed to bring resource down and back up to same role schedule_restart_actions(rsc, current, need_stop, need_promote); // Create any actions needed to take resource from this role to the next schedule_role_transition_actions(rsc); pcmk__create_recurring_actions(rsc); if (allow_migrate) { pcmk__create_migration_actions(rsc, current); } } /*! * \internal * \brief Ban a resource from any allowed nodes that are Pacemaker Remote nodes * * \param[in] rsc Resource to check */ static void rsc_avoids_remote_nodes(const pe_resource_t *rsc) { GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (node->details->remote_rsc != NULL) { node->weight = -INFINITY; } } } /*! * \internal * \brief Return allowed nodes as (possibly sorted) list * * Convert a resource's hash table of allowed nodes to a list. If printing to * stdout, sort the list, to keep action ID numbers consistent for regression * test output (while avoiding the performance hit on a live cluster). * * \param[in] rsc Resource to check for allowed nodes * * \return List of resource's allowed nodes * \note Callers should take care not to rely on the list being sorted. */ static GList * allowed_nodes_as_list(const pe_resource_t *rsc) { GList *allowed_nodes = NULL; if (rsc->allowed_nodes) { allowed_nodes = g_hash_table_get_values(rsc->allowed_nodes); } if (!pcmk__is_daemon) { allowed_nodes = g_list_sort(allowed_nodes, pe__cmp_node_name); } return allowed_nodes; } /*! * \internal * \brief Create implicit constraints needed for a primitive resource * * \param[in,out] rsc Primitive resource to create implicit constraints for */ void pcmk__primitive_internal_constraints(pe_resource_t *rsc) { GList *allowed_nodes = NULL; bool check_unfencing = false; bool check_utilization = false; CRM_ASSERT(rsc != NULL); if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "Skipping implicit constraints for unmanaged resource %s", rsc->id); return; } // Whether resource requires unfencing check_unfencing = !pcmk_is_set(rsc->flags, pe_rsc_fence_device) && pcmk_is_set(rsc->cluster->flags, pe_flag_enable_unfencing) && pcmk_is_set(rsc->flags, pe_rsc_needs_unfencing); // Whether a non-default placement strategy is used check_utilization = (g_hash_table_size(rsc->utilization) > 0) && !pcmk__str_eq(rsc->cluster->placement_strategy, "default", pcmk__str_casei); // Order stops before starts (i.e. restart) pcmk__new_ordering(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_optional|pe_order_implies_then|pe_order_restart, rsc->cluster); // Promotable ordering: demote before stop, start before promote if (pcmk_is_set(pe__const_top_resource(rsc, false)->flags, pe_rsc_promotable) || (rsc->role > RSC_ROLE_UNPROMOTED)) { pcmk__new_ordering(rsc, pcmk__op_key(rsc->id, RSC_DEMOTE, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, pe_order_promoted_implies_first, rsc->cluster); pcmk__new_ordering(rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_PROMOTE, 0), NULL, pe_order_runnable_left, rsc->cluster); } // Don't clear resource history if probing on same node pcmk__new_ordering(rsc, pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL, pe_order_same_node|pe_order_then_cancels_first, rsc->cluster); // Certain checks need allowed nodes if (check_unfencing || check_utilization || (rsc->container != NULL)) { allowed_nodes = allowed_nodes_as_list(rsc); } if (check_unfencing) { g_list_foreach(allowed_nodes, pcmk__order_restart_vs_unfence, rsc); } if (check_utilization) { pcmk__create_utilization_constraints(rsc, allowed_nodes); } if (rsc->container != NULL) { pe_resource_t *remote_rsc = NULL; if (rsc->is_remote_node) { // rsc is the implicit remote connection for a guest or bundle node /* Guest resources are not allowed to run on Pacemaker Remote nodes, * to avoid nesting remotes. However, bundles are allowed. */ if (!pcmk_is_set(rsc->flags, pe_rsc_allow_remote_remotes)) { rsc_avoids_remote_nodes(rsc->container); } /* If someone cleans up a guest or bundle node's container, we will * likely schedule a (re-)probe of the container and recovery of the * connection. Order the connection stop after the container probe, * so that if we detect the container running, we will trigger a new * transition and avoid the unnecessary recovery. */ pcmk__order_resource_actions(rsc->container, RSC_STATUS, rsc, RSC_STOP, pe_order_optional); /* A user can specify that a resource must start on a Pacemaker Remote * node by explicitly configuring it with the container=NODENAME * meta-attribute. This is of questionable merit, since location * constraints can accomplish the same thing. But we support it, so here * we check whether a resource (that is not itself a remote connection) * has container set to a remote node or guest node resource. */ } else if (rsc->container->is_remote_node) { remote_rsc = rsc->container; } else { remote_rsc = pe__resource_contains_guest_node(rsc->cluster, rsc->container); } if (remote_rsc != NULL) { /* Force the resource on the Pacemaker Remote node instead of * colocating the resource with the container resource. */ for (GList *item = allowed_nodes; item; item = item->next) { pe_node_t *node = item->data; if (node->details->remote_rsc != remote_rsc) { node->weight = -INFINITY; } } } else { /* This resource is either a filler for a container that does NOT * represent a Pacemaker Remote node, or a Pacemaker Remote * connection resource for a guest node or bundle. */ int score; crm_trace("Order and colocate %s relative to its container %s", rsc->id, rsc->container->id); pcmk__new_ordering(rsc->container, pcmk__op_key(rsc->container->id, RSC_START, 0), NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL, pe_order_implies_then|pe_order_runnable_left, rsc->cluster); pcmk__new_ordering(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL, rsc->container, pcmk__op_key(rsc->container->id, RSC_STOP, 0), NULL, pe_order_implies_first, rsc->cluster); if (pcmk_is_set(rsc->flags, pe_rsc_allow_remote_remotes)) { score = 10000; /* Highly preferred but not essential */ } else { score = INFINITY; /* Force them to run on the same host */ } pcmk__new_colocation("resource-with-container", NULL, score, rsc, rsc->container, NULL, NULL, true, rsc->cluster); } } if (rsc->is_remote_node || pcmk_is_set(rsc->flags, pe_rsc_fence_device)) { /* Remote connections and fencing devices are not allowed to run on * Pacemaker Remote nodes */ rsc_avoids_remote_nodes(rsc); } g_list_free(allowed_nodes); } /*! * \internal - * \brief Apply a colocation's score to node weights or resource priority + * \brief Apply a colocation's score to node scores or resource priority * * Given a colocation constraint, apply its score to the dependent's - * allowed node weights (if we are still placing resources) or priority (if + * allowed node scores (if we are still placing resources) or priority (if * we are choosing promotable clone instance roles). * * \param[in,out] dependent Dependent resource in colocation * \param[in,out] primary Primary resource in colocation * \param[in] colocation Colocation constraint to apply * \param[in] for_dependent true if called on behalf of dependent */ void pcmk__primitive_apply_coloc_score(pe_resource_t *dependent, pe_resource_t *primary, const pcmk__colocation_t *colocation, bool for_dependent) { enum pcmk__coloc_affects filter_results; CRM_CHECK((colocation != NULL) && (dependent != NULL) && (primary != NULL), return); if (for_dependent) { // Always process on behalf of primary resource primary->cmds->apply_coloc_score(dependent, primary, colocation, false); return; } filter_results = pcmk__colocation_affects(dependent, primary, colocation, false); pe_rsc_trace(dependent, "%s %s with %s (%s, score=%d, filter=%d)", ((colocation->score > 0)? "Colocating" : "Anti-colocating"), dependent->id, primary->id, colocation->id, colocation->score, filter_results); switch (filter_results) { case pcmk__coloc_affects_role: pcmk__apply_coloc_to_priority(dependent, primary, colocation); break; case pcmk__coloc_affects_location: - pcmk__apply_coloc_to_weights(dependent, primary, colocation); + pcmk__apply_coloc_to_scores(dependent, primary, colocation); break; default: // pcmk__coloc_affects_nothing return; } } /* Primitive implementation of * resource_alloc_functions_t:with_this_colocations() */ void pcmk__with_primitive_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { // Primitives don't have children, so rsc should also be orig_rsc CRM_CHECK((rsc != NULL) && (rsc->variant == pe_native) && (rsc == orig_rsc) && (list != NULL), return); // Add primitive's own colocations plus any relevant ones from parent pcmk__add_with_this_list(list, rsc->rsc_cons_lhs); if (rsc->parent != NULL) { rsc->parent->cmds->with_this_colocations(rsc->parent, rsc, list); } } /* Primitive implementation of * resource_alloc_functions_t:this_with_colocations() */ void pcmk__primitive_with_colocations(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList **list) { // Primitives don't have children, so rsc should also be orig_rsc CRM_CHECK((rsc != NULL) && (rsc->variant == pe_native) && (rsc == orig_rsc) && (list != NULL), return); // Add primitive's own colocations plus any relevant ones from parent pcmk__add_this_with_list(list, rsc->rsc_cons); if (rsc->parent != NULL) { rsc->parent->cmds->this_with_colocations(rsc->parent, rsc, list); } } /*! * \internal * \brief Return action flags for a given primitive resource action * * \param[in,out] action Action to get flags for * \param[in] node If not NULL, limit effects to this node (ignored) * * \return Flags appropriate to \p action on \p node */ uint32_t pcmk__primitive_action_flags(pe_action_t *action, const pe_node_t *node) { CRM_ASSERT(action != NULL); return (uint32_t) action->flags; } /*! * \internal * \brief Check whether a node is a multiply active resource's expected node * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return true if \p rsc is multiply active with multiple-active set to * stop_unexpected, and \p node is the node where it will remain active * \note This assumes that the resource's next role cannot be changed to stopped * after this is called, which should be reasonable if status has already * been unpacked and resources have been assigned to nodes. */ static bool is_expected_node(const pe_resource_t *rsc, const pe_node_t *node) { return pcmk_all_flags_set(rsc->flags, pe_rsc_stop_unexpected|pe_rsc_restarting) && (rsc->next_role > RSC_ROLE_STOPPED) && pe__same_node(rsc->allocated_to, node); } /*! * \internal * \brief Schedule actions needed to stop a resource wherever it is active * * \param[in,out] rsc Resource being stopped * \param[in] node Node where resource is being stopped (ignored) * \param[in] optional Whether actions should be optional */ static void stop_resource(pe_resource_t *rsc, pe_node_t *node, bool optional) { for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) { pe_node_t *current = (pe_node_t *) iter->data; pe_action_t *stop = NULL; if (is_expected_node(rsc, current)) { /* We are scheduling restart actions for a multiply active resource * with multiple-active=stop_unexpected, and this is where it should * not be stopped. */ pe_rsc_trace(rsc, "Skipping stop of multiply active resource %s " "on expected node %s", rsc->id, pe__node_name(current)); continue; } if (rsc->partial_migration_target != NULL) { // Continue migration if node originally was and remains target if (pe__same_node(current, rsc->partial_migration_target) && pe__same_node(current, rsc->allocated_to)) { pe_rsc_trace(rsc, "Skipping stop of %s on %s " "because partial migration there will continue", rsc->id, pe__node_name(current)); continue; } else { pe_rsc_trace(rsc, "Forcing stop of %s on %s " "because migration target changed", rsc->id, pe__node_name(current)); optional = false; } } pe_rsc_trace(rsc, "Scheduling stop of %s on %s", rsc->id, pe__node_name(current)); stop = stop_action(rsc, current, optional); if (rsc->allocated_to == NULL) { pe_action_set_reason(stop, "node availability", true); } else if (pcmk_all_flags_set(rsc->flags, pe_rsc_restarting |pe_rsc_stop_unexpected)) { /* We are stopping a multiply active resource on a node that is * not its expected node, and we are still scheduling restart * actions, so the stop is for being multiply active. */ pe_action_set_reason(stop, "being multiply active", true); } if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe__clear_action_flags(stop, pe_action_runnable); } if (pcmk_is_set(rsc->cluster->flags, pe_flag_remove_after_stop)) { pcmk__schedule_cleanup(rsc, current, optional); } if (pcmk_is_set(rsc->flags, pe_rsc_needs_unfencing)) { pe_action_t *unfence = pe_fence_op(current, "on", true, NULL, false, rsc->cluster); order_actions(stop, unfence, pe_order_implies_first); if (!pcmk__node_unfenced(current)) { pe_proc_err("Stopping %s until %s can be unfenced", rsc->id, pe__node_name(current)); } } } } /*! * \internal * \brief Schedule actions needed to start a resource on a node * * \param[in,out] rsc Resource being started * \param[in,out] node Node where resource should be started * \param[in] optional Whether actions should be optional */ static void start_resource(pe_resource_t *rsc, pe_node_t *node, bool optional) { pe_action_t *start = NULL; CRM_ASSERT(node != NULL); pe_rsc_trace(rsc, "Scheduling %s start of %s on %s (score %d)", (optional? "optional" : "required"), rsc->id, pe__node_name(node), node->weight); start = start_action(rsc, node, TRUE); pcmk__order_vs_unfence(rsc, node, start, pe_order_implies_then); if (pcmk_is_set(start->flags, pe_action_runnable) && !optional) { pe__clear_action_flags(start, pe_action_optional); } if (is_expected_node(rsc, node)) { /* This could be a problem if the start becomes necessary for other * reasons later. */ pe_rsc_trace(rsc, "Start of multiply active resouce %s " "on expected node %s will be a pseudo-action", rsc->id, pe__node_name(node)); pe__set_action_flags(start, pe_action_pseudo); } } /*! * \internal * \brief Schedule actions needed to promote a resource on a node * * \param[in,out] rsc Resource being promoted * \param[in] node Node where resource should be promoted * \param[in] optional Whether actions should be optional */ static void promote_resource(pe_resource_t *rsc, pe_node_t *node, bool optional) { GList *iter = NULL; GList *action_list = NULL; bool runnable = true; CRM_ASSERT(node != NULL); // Any start must be runnable for promotion to be runnable action_list = pe__resource_actions(rsc, node, RSC_START, true); for (iter = action_list; iter != NULL; iter = iter->next) { pe_action_t *start = (pe_action_t *) iter->data; if (!pcmk_is_set(start->flags, pe_action_runnable)) { runnable = false; } } g_list_free(action_list); if (runnable) { pe_action_t *promote = promote_action(rsc, node, optional); pe_rsc_trace(rsc, "Scheduling %s promotion of %s on %s", (optional? "optional" : "required"), rsc->id, pe__node_name(node)); if (is_expected_node(rsc, node)) { /* This could be a problem if the promote becomes necessary for * other reasons later. */ pe_rsc_trace(rsc, "Promotion of multiply active resouce %s " "on expected node %s will be a pseudo-action", rsc->id, pe__node_name(node)); pe__set_action_flags(promote, pe_action_pseudo); } } else { pe_rsc_trace(rsc, "Not promoting %s on %s: start unrunnable", rsc->id, pe__node_name(node)); action_list = pe__resource_actions(rsc, node, RSC_PROMOTE, true); for (iter = action_list; iter != NULL; iter = iter->next) { pe_action_t *promote = (pe_action_t *) iter->data; pe__clear_action_flags(promote, pe_action_runnable); } g_list_free(action_list); } } /*! * \internal * \brief Schedule actions needed to demote a resource wherever it is active * * \param[in,out] rsc Resource being demoted * \param[in] node Node where resource should be demoted (ignored) * \param[in] optional Whether actions should be optional */ static void demote_resource(pe_resource_t *rsc, pe_node_t *node, bool optional) { /* Since this will only be called for a primitive (possibly as an instance * of a collective resource), the resource is multiply active if it is * running on more than one node, so we want to demote on all of them as * part of recovery, regardless of which one is the desired node. */ for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) { pe_node_t *current = (pe_node_t *) iter->data; if (is_expected_node(rsc, current)) { pe_rsc_trace(rsc, "Skipping demote of multiply active resource %s " "on expected node %s", rsc->id, pe__node_name(current)); } else { pe_rsc_trace(rsc, "Scheduling %s demotion of %s on %s", (optional? "optional" : "required"), rsc->id, pe__node_name(current)); demote_action(rsc, current, optional); } } } static void assert_role_error(pe_resource_t *rsc, pe_node_t *node, bool optional) { CRM_ASSERT(false); } /*! * \internal * \brief Schedule cleanup of a resource * * \param[in,out] rsc Resource to clean up * \param[in] node Node to clean up on * \param[in] optional Whether clean-up should be optional */ void pcmk__schedule_cleanup(pe_resource_t *rsc, const pe_node_t *node, bool optional) { /* If the cleanup is required, its orderings are optional, because they're * relevant only if both actions are required. Conversely, if the cleanup is * optional, the orderings make the then action required if the first action * becomes required. */ uint32_t flag = optional? pe_order_implies_then : pe_order_optional; CRM_CHECK((rsc != NULL) && (node != NULL), return); if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { pe_rsc_trace(rsc, "Skipping clean-up of %s on %s: resource failed", rsc->id, pe__node_name(node)); return; } if (node->details->unclean || !node->details->online) { pe_rsc_trace(rsc, "Skipping clean-up of %s on %s: node unavailable", rsc->id, pe__node_name(node)); return; } crm_notice("Scheduling clean-up of %s on %s", rsc->id, pe__node_name(node)); delete_action(rsc, node, optional); // stop -> clean-up -> start pcmk__order_resource_actions(rsc, RSC_STOP, rsc, RSC_DELETE, flag); pcmk__order_resource_actions(rsc, RSC_DELETE, rsc, RSC_START, flag); } /*! * \internal * \brief Add primitive meta-attributes relevant to graph actions to XML * * \param[in] rsc Primitive resource whose meta-attributes should be added * \param[in,out] xml Transition graph action attributes XML to add to */ void pcmk__primitive_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml) { char *name = NULL; char *value = NULL; const pe_resource_t *parent = NULL; CRM_ASSERT((rsc != NULL) && (xml != NULL)); /* Clone instance numbers get set internally as meta-attributes, and are * needed in the transition graph (for example, to tell unique clone * instances apart). */ value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INCARNATION); if (value != NULL) { name = crm_meta_name(XML_RSC_ATTR_INCARNATION); crm_xml_add(xml, name, value); free(name); } // Not sure if this one is really needed ... value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_REMOTE_NODE); if (value != NULL) { name = crm_meta_name(XML_RSC_ATTR_REMOTE_NODE); crm_xml_add(xml, name, value); free(name); } /* The container meta-attribute can be set on the primitive itself or one of * its parents (for example, a group inside a container resource), so check * them all, and keep the highest one found. */ for (parent = rsc; parent != NULL; parent = parent->parent) { if (parent->container != NULL) { crm_xml_add(xml, CRM_META "_" XML_RSC_ATTR_CONTAINER, parent->container->id); } } /* Bundle replica children will get their external-ip set internally as a * meta-attribute. The graph action needs it, but under a different naming * convention than other meta-attributes. */ value = g_hash_table_lookup(rsc->meta, "external-ip"); if (value != NULL) { crm_xml_add(xml, "pcmk_external_ip", value); } } // Primitive implementation of resource_alloc_functions_t:add_utilization() void pcmk__primitive_add_utilization(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *all_rscs, GHashTable *utilization) { if (!pcmk_is_set(rsc->flags, pe_rsc_provisional)) { return; } pe_rsc_trace(orig_rsc, "%s: Adding primitive %s as colocated utilization", orig_rsc->id, rsc->id); pcmk__release_node_capacity(utilization, rsc); } /*! * \internal * \brief Get epoch time of node's shutdown attribute (or now if none) * * \param[in,out] node Node to check * * \return Epoch time corresponding to shutdown attribute if set or now if not */ static time_t shutdown_time(pe_node_t *node) { const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN); time_t result = 0; if (shutdown != NULL) { long long result_ll; if (pcmk__scan_ll(shutdown, &result_ll, 0LL) == pcmk_rc_ok) { result = (time_t) result_ll; } } return (result == 0)? get_effective_time(node->details->data_set) : result; } /*! * \internal * \brief Ban a resource from a node if it's not locked to the node * * \param[in] data Node to check * \param[in,out] user_data Resource to check */ static void ban_if_not_locked(gpointer data, gpointer user_data) { const pe_node_t *node = (const pe_node_t *) data; pe_resource_t *rsc = (pe_resource_t *) user_data; if (strcmp(node->details->uname, rsc->lock_node->details->uname) != 0) { resource_location(rsc, node, -CRM_SCORE_INFINITY, XML_CONFIG_ATTR_SHUTDOWN_LOCK, rsc->cluster); } } // Primitive implementation of resource_alloc_functions_t:shutdown_lock() void pcmk__primitive_shutdown_lock(pe_resource_t *rsc) { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); // Fence devices and remote connections can't be locked if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_null_matches) || pe__resource_is_remote_conn(rsc, rsc->cluster)) { return; } if (rsc->lock_node != NULL) { // The lock was obtained from resource history if (rsc->running_on != NULL) { /* The resource was started elsewhere even though it is now * considered locked. This shouldn't be possible, but as a * failsafe, we don't want to disturb the resource now. */ pe_rsc_info(rsc, "Cancelling shutdown lock because %s is already active", rsc->id); pe__clear_resource_history(rsc, rsc->lock_node, rsc->cluster); rsc->lock_node = NULL; rsc->lock_time = 0; } // Only a resource active on exactly one node can be locked } else if (pcmk__list_of_1(rsc->running_on)) { pe_node_t *node = rsc->running_on->data; if (node->details->shutdown) { if (node->details->unclean) { pe_rsc_debug(rsc, "Not locking %s to unclean %s for shutdown", rsc->id, pe__node_name(node)); } else { rsc->lock_node = node; rsc->lock_time = shutdown_time(node); } } } if (rsc->lock_node == NULL) { // No lock needed return; } if (rsc->cluster->shutdown_lock > 0) { time_t lock_expiration = rsc->lock_time + rsc->cluster->shutdown_lock; pe_rsc_info(rsc, "Locking %s to %s due to shutdown (expires @%lld)", rsc->id, pe__node_name(rsc->lock_node), (long long) lock_expiration); pe__update_recheck_time(++lock_expiration, rsc->cluster); } else { pe_rsc_info(rsc, "Locking %s to %s due to shutdown", rsc->id, pe__node_name(rsc->lock_node)); } // If resource is locked to one node, ban it from all other nodes g_list_foreach(rsc->cluster->nodes, ban_if_not_locked, rsc); } diff --git a/lib/pacemaker/pcmk_sched_probes.c b/lib/pacemaker/pcmk_sched_probes.c index 919e5233da..cdc8a581b7 100644 --- a/lib/pacemaker/pcmk_sched_probes.c +++ b/lib/pacemaker/pcmk_sched_probes.c @@ -1,896 +1,895 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Add the expected result to a newly created probe * * \param[in,out] probe Probe action to add expected result to * \param[in] rsc Resource that probe is for * \param[in] node Node that probe will run on */ static void add_expected_result(pe_action_t *probe, const pe_resource_t *rsc, const pe_node_t *node) { // Check whether resource is currently active on node pe_node_t *running = pe_find_node_id(rsc->running_on, node->details->id); // The expected result is what we think the resource's current state is if (running == NULL) { pe__add_action_expected_result(probe, CRM_EX_NOT_RUNNING); } else if (rsc->role == RSC_ROLE_PROMOTED) { pe__add_action_expected_result(probe, CRM_EX_PROMOTED); } } /*! * \internal * \brief Create any needed robes on a node for a list of resources * * \param[in,out] rscs List of resources to create probes for * \param[in,out] node Node to create probes on * * \return true if any probe was created, otherwise false */ bool pcmk__probe_resource_list(GList *rscs, pe_node_t *node) { bool any_created = false; for (GList *iter = rscs; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; if (rsc->cmds->create_probe(rsc, node)) { any_created = true; } } return any_created; } /*! * \internal * \brief Order one resource's start after another's start-up probe * * \param[in,out] rsc1 Resource that might get start-up probe * \param[in] rsc2 Resource that might be started */ static void probe_then_start(pe_resource_t *rsc1, pe_resource_t *rsc2) { if ((rsc1->allocated_to != NULL) && (g_hash_table_lookup(rsc1->known_on, rsc1->allocated_to->details->id) == NULL)) { pcmk__new_ordering(rsc1, pcmk__op_key(rsc1->id, RSC_STATUS, 0), NULL, rsc2, pcmk__op_key(rsc2->id, RSC_START, 0), NULL, pe_order_optional, rsc1->cluster); } } /*! * \internal * \brief Check whether a guest resource will stop * * \param[in] node Guest node to check * * \return true if guest resource will likely stop, otherwise false */ static bool guest_resource_will_stop(const pe_node_t *node) { const pe_resource_t *guest_rsc = node->details->remote_rsc->container; /* Ideally, we'd check whether the guest has a required stop, but that * information doesn't exist yet, so approximate it ... */ return node->details->remote_requires_reset || node->details->unclean || pcmk_is_set(guest_rsc->flags, pe_rsc_failed) || (guest_rsc->next_role == RSC_ROLE_STOPPED) // Guest is moving || ((guest_rsc->role > RSC_ROLE_STOPPED) && (guest_rsc->allocated_to != NULL) && (pe_find_node(guest_rsc->running_on, guest_rsc->allocated_to->details->uname) == NULL)); } /*! * \internal * \brief Create a probe action for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return Newly created probe action */ static pe_action_t * probe_action(pe_resource_t *rsc, pe_node_t *node) { pe_action_t *probe = NULL; char *key = pcmk__op_key(rsc->id, RSC_STATUS, 0); crm_debug("Scheduling probe of %s %s on %s", role2text(rsc->role), rsc->id, pe__node_name(node)); probe = custom_action(rsc, key, RSC_STATUS, node, FALSE, TRUE, rsc->cluster); pe__clear_action_flags(probe, pe_action_optional); pcmk__order_vs_unfence(rsc, node, probe, pe_order_optional); add_expected_result(probe, rsc, node); return probe; } /*! * \internal * \brief Create probes for a resource on a node, if needed * * \brief Schedule any probes needed for a resource on a node * * \param[in,out] rsc Resource to create probe for * \param[in,out] node Node to create probe on * * \return true if any probe was created, otherwise false */ bool pcmk__probe_rsc_on_node(pe_resource_t *rsc, pe_node_t *node) { uint32_t flags = pe_order_optional; pe_action_t *probe = NULL; pe_node_t *allowed = NULL; pe_resource_t *top = uber_parent(rsc); const char *reason = NULL; CRM_CHECK((rsc != NULL) && (node != NULL), return false); if (!pcmk_is_set(rsc->cluster->flags, pe_flag_startup_probes)) { reason = "start-up probes are disabled"; goto no_probe; } if (pe__is_guest_or_remote_node(node)) { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_none)) { reason = "Pacemaker Remote nodes cannot run stonith agents"; goto no_probe; } else if (pe__is_guest_node(node) && pe__resource_contains_guest_node(rsc->cluster, rsc)) { reason = "guest nodes cannot run resources containing guest nodes"; goto no_probe; } else if (rsc->is_remote_node) { reason = "Pacemaker Remote nodes cannot host remote connections"; goto no_probe; } } // If this is a collective resource, probes are created for its children if (rsc->children != NULL) { return pcmk__probe_resource_list(rsc->children, node); } if ((rsc->container != NULL) && !rsc->is_remote_node) { reason = "resource is inside a container"; goto no_probe; } else if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { reason = "resource is orphaned"; goto no_probe; } else if (g_hash_table_lookup(rsc->known_on, node->details->id) != NULL) { reason = "resource state is already known"; goto no_probe; } allowed = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (rsc->exclusive_discover || top->exclusive_discover) { // Exclusive discovery is enabled ... if (allowed == NULL) { // ... but this node is not allowed to run the resource reason = "resource has exclusive discovery but is not allowed " "on node"; goto no_probe; } else if (allowed->rsc_discover_mode != pe_discover_exclusive) { // ... but no constraint marks this node for discovery of resource reason = "resource has exclusive discovery but is not enabled " "on node"; goto no_probe; } } if (allowed == NULL) { allowed = node; } if (allowed->rsc_discover_mode == pe_discover_never) { reason = "node has discovery disabled"; goto no_probe; } if (pe__is_guest_node(node)) { pe_resource_t *guest = node->details->remote_rsc->container; if (guest->role == RSC_ROLE_STOPPED) { // The guest is stopped, so we know no resource is active there reason = "node's guest is stopped"; probe_then_start(guest, top); goto no_probe; } else if (guest_resource_will_stop(node)) { reason = "node's guest will stop"; // Order resource start after guest stop (in case it's restarting) pcmk__new_ordering(guest, pcmk__op_key(guest->id, RSC_STOP, 0), NULL, top, pcmk__op_key(top->id, RSC_START, 0), NULL, pe_order_optional, rsc->cluster); goto no_probe; } } // We've eliminated all cases where a probe is not needed, so now it is probe = probe_action(rsc, node); /* Below, we will order the probe relative to start or reload. If this is a * clone instance, the start or reload is for the entire clone rather than * just the instance. Otherwise, the start or reload is for the resource * itself. */ if (!pe_rsc_is_clone(top)) { top = rsc; } /* Prevent a start if the resource can't be probed, but don't cause the * resource or entire clone to stop if already active. */ if (!pcmk_is_set(probe->flags, pe_action_runnable) && (top->running_on == NULL)) { pe__set_order_flags(flags, pe_order_runnable_left); } // Start or reload after probing the resource pcmk__new_ordering(rsc, NULL, probe, top, pcmk__op_key(top->id, RSC_START, 0), NULL, flags, rsc->cluster); pcmk__new_ordering(rsc, NULL, probe, top, reload_key(rsc), NULL, pe_order_optional, rsc->cluster); return true; no_probe: pe_rsc_trace(rsc, "Skipping probe for %s on %s because %s", rsc->id, node->details->id, reason); return false; } /*! * \internal * \brief Check whether a probe should be ordered before another action * * \param[in] probe Probe action to check * \param[in] then Other action to check * * \return true if \p probe should be ordered before \p then, otherwise false */ static bool probe_needed_before_action(const pe_action_t *probe, const pe_action_t *then) { // Probes on a node are performed after unfencing it, not before if (pcmk__str_eq(then->task, CRM_OP_FENCE, pcmk__str_casei) - && (probe->node != NULL) && (then->node != NULL) - && (probe->node->details == then->node->details)) { + && pe__same_node(probe->node, then->node)) { const char *op = g_hash_table_lookup(then->meta, "stonith_action"); if (pcmk__str_eq(op, "on", pcmk__str_casei)) { return false; } } // Probes should be done on a node before shutting it down if (pcmk__str_eq(then->task, CRM_OP_SHUTDOWN, pcmk__str_none) && (probe->node != NULL) && (then->node != NULL) - && (probe->node->details != then->node->details)) { + && !pe__same_node(probe->node, then->node)) { return false; } // Otherwise probes should always be done before any other action return true; } /*! * \internal * \brief Add implicit "probe then X" orderings for "stop then X" orderings * * If the state of a resource is not known yet, a probe will be scheduled, * expecting a "not running" result. If the probe fails, a stop will not be * scheduled until the next transition. Thus, if there are ordering constraints * like "stop this resource then do something else that's not for the same * resource", add implicit "probe this resource then do something" equivalents * so the relation is upheld until we know whether a stop is needed. * * \param[in,out] data_set Cluster working set */ static void add_probe_orderings_for_stops(pe_working_set_t *data_set) { for (GList *iter = data_set->ordering_constraints; iter != NULL; iter = iter->next) { pe__ordering_t *order = iter->data; uint32_t order_flags = pe_order_optional; GList *probes = NULL; GList *then_actions = NULL; // Skip disabled orderings if (order->flags == pe_order_none) { continue; } // Skip non-resource orderings, and orderings for the same resource if ((order->lh_rsc == NULL) || (order->lh_rsc == order->rh_rsc)) { continue; } // Skip invalid orderings (shouldn't be possible) if (((order->lh_action == NULL) && (order->lh_action_task == NULL)) || ((order->rh_action == NULL) && (order->rh_action_task == NULL))) { continue; } // Skip orderings for first actions other than stop if ((order->lh_action != NULL) && !pcmk__str_eq(order->lh_action->task, RSC_STOP, pcmk__str_none)) { continue; } else if ((order->lh_action == NULL) && !pcmk__ends_with(order->lh_action_task, "_" RSC_STOP "_0")) { continue; } /* Do not imply a probe ordering for a resource inside of a stopping * container. Otherwise, it might introduce a transition loop, since a * probe could be scheduled after the container starts again. */ if ((order->rh_rsc != NULL) && (order->lh_rsc->container == order->rh_rsc)) { if ((order->rh_action != NULL) && pcmk__str_eq(order->rh_action->task, RSC_STOP, pcmk__str_none)) { continue; } else if ((order->rh_action == NULL) && pcmk__ends_with(order->rh_action_task, "_" RSC_STOP "_0")) { continue; } } // Preserve certain order options for future filtering if (pcmk_is_set(order->flags, pe_order_apply_first_non_migratable)) { pe__set_order_flags(order_flags, pe_order_apply_first_non_migratable); } if (pcmk_is_set(order->flags, pe_order_same_node)) { pe__set_order_flags(order_flags, pe_order_same_node); } // Preserve certain order types for future filtering if ((order->flags == pe_order_anti_colocation) || (order->flags == pe_order_load)) { order_flags = order->flags; } // List all scheduled probes for the first resource probes = pe__resource_actions(order->lh_rsc, NULL, RSC_STATUS, FALSE); if (probes == NULL) { // There aren't any continue; } // List all relevant "then" actions if (order->rh_action != NULL) { then_actions = g_list_prepend(NULL, order->rh_action); } else if (order->rh_rsc != NULL) { then_actions = find_actions(order->rh_rsc->actions, order->rh_action_task, NULL); if (then_actions == NULL) { // There aren't any g_list_free(probes); continue; } } crm_trace("Implying 'probe then' orderings for '%s then %s' " "(id=%d, type=%.6x)", order->lh_action? order->lh_action->uuid : order->lh_action_task, order->rh_action? order->rh_action->uuid : order->rh_action_task, order->id, order->flags); for (GList *probe_iter = probes; probe_iter != NULL; probe_iter = probe_iter->next) { pe_action_t *probe = (pe_action_t *) probe_iter->data; for (GList *then_iter = then_actions; then_iter != NULL; then_iter = then_iter->next) { pe_action_t *then = (pe_action_t *) then_iter->data; if (probe_needed_before_action(probe, then)) { order_actions(probe, then, order_flags); } } } g_list_free(then_actions); g_list_free(probes); } } /*! * \internal * \brief Add necessary orderings between probe and starts of clone instances * * , in additon to the ordering with the parent resource added upon creating * the probe. * * \param[in,out] probe Probe as 'first' action in an ordering * \param[in,out] after 'then' action wrapper in the ordering */ static void add_start_orderings_for_probe(pe_action_t *probe, pe_action_wrapper_t *after) { uint32_t flags = pe_order_optional|pe_order_runnable_left; /* Although the ordering between the probe of the clone instance and the * start of its parent has been added in pcmk__probe_rsc_on_node(), we * avoided enforcing `pe_order_runnable_left` order type for that as long as * any of the clone instances are running to prevent them from being * unexpectedly stopped. * * On the other hand, we still need to prevent any inactive instances from * starting unless the probe is runnable so that we don't risk starting too * many instances before we know the state on all nodes. */ if (after->action->rsc->variant <= pe_group || pcmk_is_set(probe->flags, pe_action_runnable) // The order type is already enforced for its parent. || pcmk_is_set(after->type, pe_order_runnable_left) || (pe__const_top_resource(probe->rsc, false) != after->action->rsc) || !pcmk__str_eq(after->action->task, RSC_START, pcmk__str_none)) { return; } crm_trace("Adding probe start orderings for '%s@%s (%s) " "then instances of %s@%s'", probe->uuid, pe__node_name(probe->node), pcmk_is_set(probe->flags, pe_action_runnable)? "runnable" : "unrunnable", after->action->uuid, pe__node_name(after->action->node)); for (GList *then_iter = after->action->actions_after; then_iter != NULL; then_iter = then_iter->next) { pe_action_wrapper_t *then = (pe_action_wrapper_t *) then_iter->data; if (then->action->rsc->running_on || (pe__const_top_resource(then->action->rsc, false) != after->action->rsc) || !pcmk__str_eq(then->action->task, RSC_START, pcmk__str_none)) { continue; } crm_trace("Adding probe start ordering for '%s@%s (%s) " "then %s@%s' (type=%#.6x)", probe->uuid, pe__node_name(probe->node), pcmk_is_set(probe->flags, pe_action_runnable)? "runnable" : "unrunnable", then->action->uuid, pe__node_name(then->action->node), flags); /* Prevent the instance from starting if the instance can't, but don't * cause any other intances to stop if already active. */ order_actions(probe, then->action, flags); } return; } /*! * \internal * \brief Order probes before restarts and re-promotes * * If a given ordering is a "probe then start" or "probe then promote" ordering, * add an implicit "probe then stop/demote" ordering in case the action is part * of a restart/re-promote, and do the same recursively for all actions ordered * after the "then" action. * * \param[in,out] probe Probe as 'first' action in an ordering * \param[in,out] after 'then' action in the ordering * \param[in,out] data_set Cluster working set */ static void add_restart_orderings_for_probe(pe_action_t *probe, pe_action_t *after, pe_working_set_t *data_set) { GList *iter = NULL; bool interleave = false; pe_resource_t *compatible_rsc = NULL; // Validate that this is a resource probe followed by some action if ((after == NULL) || (probe == NULL) || (probe->rsc == NULL) || (probe->rsc->variant != pe_native) || !pcmk__str_eq(probe->task, RSC_STATUS, pcmk__str_casei)) { return; } // Avoid running into any possible loop if (pcmk_is_set(after->flags, pe_action_tracking)) { return; } pe__set_action_flags(after, pe_action_tracking); crm_trace("Adding probe restart orderings for '%s@%s then %s@%s'", probe->uuid, pe__node_name(probe->node), after->uuid, pe__node_name(after->node)); /* Add restart orderings if "then" is for a different primitive. * Orderings for collective resources will be added later. */ if ((after->rsc != NULL) && (after->rsc->variant == pe_native) && (probe->rsc != after->rsc)) { GList *then_actions = NULL; if (pcmk__str_eq(after->task, RSC_START, pcmk__str_casei)) { then_actions = pe__resource_actions(after->rsc, NULL, RSC_STOP, FALSE); } else if (pcmk__str_eq(after->task, RSC_PROMOTE, pcmk__str_casei)) { then_actions = pe__resource_actions(after->rsc, NULL, RSC_DEMOTE, FALSE); } for (iter = then_actions; iter != NULL; iter = iter->next) { pe_action_t *then = (pe_action_t *) iter->data; // Skip pseudo-actions (for example, those implied by fencing) if (!pcmk_is_set(then->flags, pe_action_pseudo)) { order_actions(probe, then, pe_order_optional); } } g_list_free(then_actions); } /* Detect whether "then" is an interleaved clone action. For these, we want * to add orderings only for the relevant instance. */ if ((after->rsc != NULL) && (after->rsc->variant > pe_group)) { const char *interleave_s = g_hash_table_lookup(after->rsc->meta, XML_RSC_ATTR_INTERLEAVE); interleave = crm_is_true(interleave_s); if (interleave) { compatible_rsc = pcmk__find_compatible_instance(probe->rsc, after->rsc, RSC_ROLE_UNKNOWN, false); } } /* Now recursively do the same for all actions ordered after "then". This * also handles collective resources since the collective action will be * ordered before its individual instances' actions. */ for (iter = after->actions_after; iter != NULL; iter = iter->next) { pe_action_wrapper_t *after_wrapper = (pe_action_wrapper_t *) iter->data; /* pe_order_implies_then is the reason why a required A.start * implies/enforces B.start to be required too, which is the cause of * B.restart/re-promote. * * Not sure about pe_order_implies_then_on_node though. It's now only * used for unfencing case, which tends to introduce transition * loops... */ if (!pcmk_is_set(after_wrapper->type, pe_order_implies_then)) { /* The order type between a group/clone and its child such as * B.start-> B_child.start is: * pe_order_implies_first_printed | pe_order_runnable_left * * Proceed through the ordering chain and build dependencies with * its children. */ if ((after->rsc == NULL) || (after->rsc->variant < pe_group) || (probe->rsc->parent == after->rsc) || (after_wrapper->action->rsc == NULL) || (after_wrapper->action->rsc->variant > pe_group) || (after->rsc != after_wrapper->action->rsc->parent)) { continue; } /* Proceed to the children of a group or a non-interleaved clone. * For an interleaved clone, proceed only to the relevant child. */ if ((after->rsc->variant > pe_group) && interleave && ((compatible_rsc == NULL) || (compatible_rsc != after_wrapper->action->rsc))) { continue; } } crm_trace("Recursively adding probe restart orderings for " "'%s@%s then %s@%s' (type=%#.6x)", after->uuid, pe__node_name(after->node), after_wrapper->action->uuid, pe__node_name(after_wrapper->action->node), after_wrapper->type); add_restart_orderings_for_probe(probe, after_wrapper->action, data_set); } } /*! * \internal * \brief Clear the tracking flag on all scheduled actions * * \param[in,out] data_set Cluster working set */ static void clear_actions_tracking_flag(pe_working_set_t *data_set) { GList *gIter = NULL; for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; pe__clear_action_flags(action, pe_action_tracking); } } /*! * \internal * \brief Add start and restart orderings for probes scheduled for a resource * * \param[in,out] rsc Resource whose probes should be ordered * \param[in,out] data_set Cluster working set */ static void add_start_restart_orderings_for_rsc(pe_resource_t *rsc, pe_working_set_t *data_set) { GList *probes = NULL; // For collective resources, order each instance recursively if (rsc->variant != pe_native) { g_list_foreach(rsc->children, (GFunc) add_start_restart_orderings_for_rsc, data_set); return; } // Find all probes for given resource probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE); // Add probe restart orderings for each probe found for (GList *iter = probes; iter != NULL; iter = iter->next) { pe_action_t *probe = (pe_action_t *) iter->data; for (GList *then_iter = probe->actions_after; then_iter != NULL; then_iter = then_iter->next) { pe_action_wrapper_t *then = (pe_action_wrapper_t *) then_iter->data; add_start_orderings_for_probe(probe, then); add_restart_orderings_for_probe(probe, then->action, data_set); clear_actions_tracking_flag(data_set); } } g_list_free(probes); } /*! * \internal * \brief Add "A then probe B" orderings for "A then B" orderings * * \param[in,out] data_set Cluster working set * * \note This function is currently disabled (see next comment). */ static void order_then_probes(pe_working_set_t *data_set) { #if 0 /* Given an ordering "A then B", we would prefer to wait for A to be started * before probing B. * * For example, if A is a filesystem which B can't even run without, it * would be helpful if the author of B's agent could assume that A is * running before B.monitor will be called. * * However, we can't _only_ probe after A is running, otherwise we wouldn't * detect the state of B if A could not be started. We can't even do an * opportunistic version of this, because B may be moving: * * A.stop -> A.start -> B.probe -> B.stop -> B.start * * and if we add B.stop -> A.stop here, we get a loop: * * A.stop -> A.start -> B.probe -> B.stop -> A.stop * * We could kill the "B.probe -> B.stop" dependency, but that could mean * stopping B "too" soon, because B.start must wait for the probe, and * we don't want to stop B if we can't start it. * * We could add the ordering only if A is an anonymous clone with * clone-max == node-max (since we'll never be moving it). However, we could * still be stopping one instance at the same time as starting another. * * The complexity of checking for allowed conditions combined with the ever * narrowing use case suggests that this code should remain disabled until * someone gets smarter. */ for (GList *iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; pe_action_t *start = NULL; GList *actions = NULL; GList *probes = NULL; actions = pe__resource_actions(rsc, NULL, RSC_START, FALSE); if (actions) { start = actions->data; g_list_free(actions); } if (start == NULL) { crm_err("No start action for %s", rsc->id); continue; } probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE); for (actions = start->actions_before; actions != NULL; actions = actions->next) { pe_action_wrapper_t *before = (pe_action_wrapper_t *) actions->data; pe_action_t *first = before->action; pe_resource_t *first_rsc = first->rsc; if (first->required_runnable_before) { for (GList *clone_actions = first->actions_before; clone_actions != NULL; clone_actions = clone_actions->next) { before = (pe_action_wrapper_t *) clone_actions->data; crm_trace("Testing '%s then %s' for %s", first->uuid, before->action->uuid, start->uuid); CRM_ASSERT(before->action->rsc != NULL); first_rsc = before->action->rsc; break; } } else if (!pcmk__str_eq(first->task, RSC_START, pcmk__str_none)) { crm_trace("Not a start op %s for %s", first->uuid, start->uuid); } if (first_rsc == NULL) { continue; } else if (pe__const_top_resource(first_rsc, false) == pe__const_top_resource(start->rsc, false)) { crm_trace("Same parent %s for %s", first_rsc->id, start->uuid); continue; } else if (!pe_rsc_is_clone(pe__const_top_resource(first_rsc, false))) { crm_trace("Not a clone %s for %s", first_rsc->id, start->uuid); continue; } crm_err("Applying %s before %s %d", first->uuid, start->uuid, pe__const_top_resource(first_rsc, false)->variant); for (GList *probe_iter = probes; probe_iter != NULL; probe_iter = probe_iter->next) { pe_action_t *probe = (pe_action_t *) probe_iter->data; crm_err("Ordering %s before %s", first->uuid, probe->uuid); order_actions(first, probe, pe_order_optional); } } } #endif } void pcmk__order_probes(pe_working_set_t *data_set) { // Add orderings for "probe then X" g_list_foreach(data_set->resources, (GFunc) add_start_restart_orderings_for_rsc, data_set); add_probe_orderings_for_stops(data_set); order_then_probes(data_set); } /*! * \internal * \brief Schedule any probes needed * * \param[in,out] data_set Cluster working set * * \note This may also schedule fencing of failed remote nodes. */ void pcmk__schedule_probes(pe_working_set_t *data_set) { // Schedule probes on each node in the cluster as needed for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; const char *probed = NULL; if (!node->details->online) { // Don't probe offline nodes if (pcmk__is_failed_remote_node(node)) { pe_fence_node(data_set, node, "the connection is unrecoverable", FALSE); } continue; } else if (node->details->unclean) { // ... or nodes that need fencing continue; } else if (!node->details->rsc_discovery_enabled) { // The user requested that probes not be done on this node continue; } /* This is no longer needed for live clusters, since the probe_complete * node attribute will never be in the CIB. However this is still useful * for processing old saved CIBs (< 1.1.14), including the * reprobe-target_rc regression test. */ probed = pe_node_attribute_raw(node, CRM_OP_PROBED); if (probed != NULL && crm_is_true(probed) == FALSE) { pe_action_t *probe_op = NULL; probe_op = custom_action(NULL, crm_strdup_printf("%s-%s", CRM_OP_REPROBE, node->details->uname), CRM_OP_REPROBE, node, FALSE, TRUE, data_set); add_hash_param(probe_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); continue; } // Probe each resource in the cluster on this node, as needed pcmk__probe_resource_list(data_set->resources, node); } } diff --git a/lib/pacemaker/pcmk_sched_promotable.c b/lib/pacemaker/pcmk_sched_promotable.c index d08823e1b4..c67010de6d 100644 --- a/lib/pacemaker/pcmk_sched_promotable.c +++ b/lib/pacemaker/pcmk_sched_promotable.c @@ -1,1284 +1,1285 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include "libpacemaker_private.h" /*! * \internal * \brief Add implicit promotion ordering for a promotable instance * * \param[in,out] clone Clone resource * \param[in,out] child Instance of \p clone being ordered * \param[in,out] last Previous instance ordered (NULL if \p child is first) */ static void order_instance_promotion(pe_resource_t *clone, pe_resource_t *child, pe_resource_t *last) { // "Promote clone" -> promote instance -> "clone promoted" pcmk__order_resource_actions(clone, RSC_PROMOTE, child, RSC_PROMOTE, pe_order_optional); pcmk__order_resource_actions(child, RSC_PROMOTE, clone, RSC_PROMOTED, pe_order_optional); // If clone is ordered, order this instance relative to last if ((last != NULL) && pe__clone_is_ordered(clone)) { pcmk__order_resource_actions(last, RSC_PROMOTE, child, RSC_PROMOTE, pe_order_optional); } } /*! * \internal * \brief Add implicit demotion ordering for a promotable instance * * \param[in,out] clone Clone resource * \param[in,out] child Instance of \p clone being ordered * \param[in] last Previous instance ordered (NULL if \p child is first) */ static void order_instance_demotion(pe_resource_t *clone, pe_resource_t *child, pe_resource_t *last) { // "Demote clone" -> demote instance -> "clone demoted" pcmk__order_resource_actions(clone, RSC_DEMOTE, child, RSC_DEMOTE, pe_order_implies_first_printed); pcmk__order_resource_actions(child, RSC_DEMOTE, clone, RSC_DEMOTED, pe_order_implies_then_printed); // If clone is ordered, order this instance relative to last if ((last != NULL) && pe__clone_is_ordered(clone)) { pcmk__order_resource_actions(child, RSC_DEMOTE, last, RSC_DEMOTE, pe_order_optional); } } /*! * \internal * \brief Check whether an instance will be promoted or demoted * * \param[in] rsc Instance to check * \param[out] demoting If \p rsc will be demoted, this will be set to true * \param[out] promoting If \p rsc will be promoted, this will be set to true */ static void check_for_role_change(const pe_resource_t *rsc, bool *demoting, bool *promoting) { const GList *iter = NULL; // If this is a cloned group, check group members recursively if (rsc->children != NULL) { for (iter = rsc->children; iter != NULL; iter = iter->next) { check_for_role_change((const pe_resource_t *) iter->data, demoting, promoting); } return; } for (iter = rsc->actions; iter != NULL; iter = iter->next) { const pe_action_t *action = (const pe_action_t *) iter->data; if (*promoting && *demoting) { return; } else if (pcmk_is_set(action->flags, pe_action_optional)) { continue; } else if (pcmk__str_eq(RSC_DEMOTE, action->task, pcmk__str_none)) { *demoting = true; } else if (pcmk__str_eq(RSC_PROMOTE, action->task, pcmk__str_none)) { *promoting = true; } } } /*! * \internal * \brief Add promoted-role location constraint scores to an instance's priority * * Adjust a promotable clone instance's promotion priority by the scores of any * location constraints in a list that are both limited to the promoted role and * for the node where the instance will be placed. * * \param[in,out] child Promotable clone instance * \param[in] location_constraints List of location constraints to apply * \param[in] chosen Node where \p child will be placed */ static void apply_promoted_locations(pe_resource_t *child, const GList *location_constraints, const pe_node_t *chosen) { for (const GList *iter = location_constraints; iter; iter = iter->next) { const pe__location_t *location = iter->data; - pe_node_t *weighted_node = NULL; + const pe_node_t *constraint_node = NULL; if (location->role_filter == RSC_ROLE_PROMOTED) { - weighted_node = pe_find_node_id(location->node_list_rh, - chosen->details->id); + constraint_node = pe_find_node_id(location->node_list_rh, + chosen->details->id); } - if (weighted_node != NULL) { + if (constraint_node != NULL) { int new_priority = pcmk__add_scores(child->priority, - weighted_node->weight); + constraint_node->weight); pe_rsc_trace(child, "Applying location %s to %s promotion priority on %s: " "%s + %s = %s", - location->id, child->id, pe__node_name(weighted_node), + location->id, child->id, + pe__node_name(constraint_node), pcmk_readable_score(child->priority), - pcmk_readable_score(weighted_node->weight), + pcmk_readable_score(constraint_node->weight), pcmk_readable_score(new_priority)); child->priority = new_priority; } } } /*! * \internal * \brief Get the node that an instance will be promoted on * * \param[in] rsc Promotable clone instance to check * * \return Node that \p rsc will be promoted on, or NULL if none */ static pe_node_t * node_to_be_promoted_on(const pe_resource_t *rsc) { pe_node_t *node = NULL; pe_node_t *local_node = NULL; const pe_resource_t *parent = NULL; // If this is a cloned group, bail if any group member can't be promoted for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child = (pe_resource_t *) iter->data; if (node_to_be_promoted_on(child) == NULL) { pe_rsc_trace(rsc, "%s can't be promoted because member %s can't", rsc->id, child->id); return NULL; } } node = rsc->fns->location(rsc, NULL, FALSE); if (node == NULL) { pe_rsc_trace(rsc, "%s can't be promoted because it won't be active", rsc->id); return NULL; } else if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { if (rsc->fns->state(rsc, TRUE) == RSC_ROLE_PROMOTED) { crm_notice("Unmanaged instance %s will be left promoted on %s", rsc->id, pe__node_name(node)); } else { pe_rsc_trace(rsc, "%s can't be promoted because it is unmanaged", rsc->id); return NULL; } } else if (rsc->priority < 0) { pe_rsc_trace(rsc, "%s can't be promoted because its promotion priority %d " "is negative", rsc->id, rsc->priority); return NULL; } else if (!pcmk__node_available(node, false, true)) { pe_rsc_trace(rsc, "%s can't be promoted because %s can't run resources", rsc->id, pe__node_name(node)); return NULL; } parent = pe__const_top_resource(rsc, false); local_node = pe_hash_table_lookup(parent->allowed_nodes, node->details->id); if (local_node == NULL) { - /* It should not be possible for the scheduler to have allocated the + /* It should not be possible for the scheduler to have assigned the * instance to a node where its parent is not allowed, but it's good to * have a fail-safe. */ if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { crm_warn("%s can't be promoted because %s is not allowed on %s " "(scheduler bug?)", rsc->id, parent->id, pe__node_name(node)); } // else the instance is unmanaged and already promoted return NULL; } else if ((local_node->count >= pe__clone_promoted_node_max(parent)) && pcmk_is_set(rsc->flags, pe_rsc_managed)) { pe_rsc_trace(rsc, "%s can't be promoted because %s has " "maximum promoted instances already", rsc->id, pe__node_name(node)); return NULL; } return local_node; } /*! * \internal * \brief Compare two promotable clone instances by promotion priority * * \param[in] a First instance to compare * \param[in] b Second instance to compare * * \return A negative number if \p a has higher promotion priority, * a positive number if \p b has higher promotion priority, * or 0 if promotion priorities are equal */ static gint cmp_promotable_instance(gconstpointer a, gconstpointer b) { const pe_resource_t *rsc1 = (const pe_resource_t *) a; const pe_resource_t *rsc2 = (const pe_resource_t *) b; enum rsc_role_e role1 = RSC_ROLE_UNKNOWN; enum rsc_role_e role2 = RSC_ROLE_UNKNOWN; CRM_ASSERT((rsc1 != NULL) && (rsc2 != NULL)); // Check sort index set by pcmk__set_instance_roles() if (rsc1->sort_index > rsc2->sort_index) { pe_rsc_trace(rsc1, "%s has higher promotion priority than %s " "(sort index %d > %d)", rsc1->id, rsc2->id, rsc1->sort_index, rsc2->sort_index); return -1; } else if (rsc1->sort_index < rsc2->sort_index) { pe_rsc_trace(rsc1, "%s has lower promotion priority than %s " "(sort index %d < %d)", rsc1->id, rsc2->id, rsc1->sort_index, rsc2->sort_index); return 1; } // If those are the same, prefer instance whose current role is higher role1 = rsc1->fns->state(rsc1, TRUE); role2 = rsc2->fns->state(rsc2, TRUE); if (role1 > role2) { pe_rsc_trace(rsc1, "%s has higher promotion priority than %s " "(higher current role)", rsc1->id, rsc2->id); return -1; } else if (role1 < role2) { pe_rsc_trace(rsc1, "%s has lower promotion priority than %s " "(lower current role)", rsc1->id, rsc2->id); return 1; } // Finally, do normal clone instance sorting return pcmk__cmp_instance(a, b); } /*! * \internal - * \brief Add a promotable clone instance's sort index to its node's weight + * \brief Add a promotable clone instance's sort index to its node's score * * Add a promotable clone instance's sort index (which sums its promotion * preferences and scores of relevant location constraints for the promoted - * role) to the node weight of the instance's allocated node. + * role) to the node score of the instance's assigned node. * * \param[in] data Promotable clone instance * \param[in,out] user_data Clone parent of \p data */ static void -add_sort_index_to_node_weight(gpointer data, gpointer user_data) +add_sort_index_to_node_score(gpointer data, gpointer user_data) { const pe_resource_t *child = (const pe_resource_t *) data; pe_resource_t *clone = (pe_resource_t *) user_data; pe_node_t *node = NULL; const pe_node_t *chosen = NULL; if (child->sort_index < 0) { pe_rsc_trace(clone, "Not adding sort index of %s: negative", child->id); return; } chosen = child->fns->location(child, NULL, FALSE); if (chosen == NULL) { pe_rsc_trace(clone, "Not adding sort index of %s: inactive", child->id); return; } node = (pe_node_t *) pe_hash_table_lookup(clone->allowed_nodes, chosen->details->id); CRM_ASSERT(node != NULL); node->weight = pcmk__add_scores(child->sort_index, node->weight); pe_rsc_trace(clone, "Added cumulative priority of %s (%s) to score on %s (now %s)", child->id, pcmk_readable_score(child->sort_index), pe__node_name(node), pcmk_readable_score(node->weight)); } /*! * \internal - * \brief Apply colocation to dependent's node weights if for promoted role + * \brief Apply colocation to dependent's node scores if for promoted role * * \param[in,out] data Colocation constraint to apply * \param[in,out] user_data Promotable clone that is constraint's dependent */ static void apply_coloc_to_dependent(gpointer data, gpointer user_data) { pcmk__colocation_t *constraint = (pcmk__colocation_t *) data; pe_resource_t *clone = (pe_resource_t *) user_data; pe_resource_t *primary = constraint->primary; uint32_t flags = pcmk__coloc_select_default; float factor = constraint->score / (float) INFINITY; if (constraint->dependent_role != RSC_ROLE_PROMOTED) { return; } if (constraint->score < INFINITY) { flags = pcmk__coloc_select_active; } pe_rsc_trace(clone, "Applying colocation %s (promoted %s with %s) @%s", constraint->id, constraint->dependent->id, constraint->primary->id, pcmk_readable_score(constraint->score)); primary->cmds->add_colocated_node_scores(primary, clone->id, &clone->allowed_nodes, constraint, factor, flags); } /*! * \internal - * \brief Apply colocation to primary's node weights if for promoted role + * \brief Apply colocation to primary's node scores if for promoted role * * \param[in,out] data Colocation constraint to apply * \param[in,out] user_data Promotable clone that is constraint's primary */ static void apply_coloc_to_primary(gpointer data, gpointer user_data) { pcmk__colocation_t *constraint = (pcmk__colocation_t *) data; pe_resource_t *clone = (pe_resource_t *) user_data; pe_resource_t *dependent = constraint->dependent; const float factor = constraint->score / (float) INFINITY; const uint32_t flags = pcmk__coloc_select_active |pcmk__coloc_select_nonnegative; if ((constraint->primary_role != RSC_ROLE_PROMOTED) || !pcmk__colocation_has_influence(constraint, NULL)) { return; } pe_rsc_trace(clone, "Applying colocation %s (%s with promoted %s) @%s", constraint->id, constraint->dependent->id, constraint->primary->id, pcmk_readable_score(constraint->score)); dependent->cmds->add_colocated_node_scores(dependent, clone->id, &clone->allowed_nodes, constraint, factor, flags); } /*! * \internal - * \brief Set clone instance's sort index to its node's weight + * \brief Set clone instance's sort index to its node's score * * \param[in,out] data Promotable clone instance * \param[in] user_data Parent clone of \p data */ static void -set_sort_index_to_node_weight(gpointer data, gpointer user_data) +set_sort_index_to_node_score(gpointer data, gpointer user_data) { pe_resource_t *child = (pe_resource_t *) data; const pe_resource_t *clone = (const pe_resource_t *) user_data; pe_node_t *chosen = child->fns->location(child, NULL, FALSE); if (!pcmk_is_set(child->flags, pe_rsc_managed) && (child->next_role == RSC_ROLE_PROMOTED)) { child->sort_index = INFINITY; pe_rsc_trace(clone, "Final sort index for %s is INFINITY (unmanaged promoted)", child->id); } else if ((chosen == NULL) || (child->sort_index < 0)) { pe_rsc_trace(clone, - "Final sort index for %s is %d (ignoring node weight)", + "Final sort index for %s is %d (ignoring node score)", child->id, child->sort_index); } else { const pe_node_t *node = NULL; node = pe_hash_table_lookup(clone->allowed_nodes, chosen->details->id); CRM_ASSERT(node != NULL); child->sort_index = node->weight; pe_rsc_trace(clone, - "Merging weights for %s: final sort index for %s is %d", + "Adding scores for %s: final sort index for %s is %d", clone->id, child->id, child->sort_index); } } /*! * \internal * \brief Sort a promotable clone's instances by descending promotion priority * * \param[in,out] clone Promotable clone to sort */ static void sort_promotable_instances(pe_resource_t *clone) { if (pe__set_clone_flag(clone, pe__clone_promotion_constrained) == pcmk_rc_already) { return; } pe__set_resource_flags(clone, pe_rsc_merging); for (GList *iter = clone->children; iter != NULL; iter = iter->next) { pe_resource_t *child = (pe_resource_t *) iter->data; pe_rsc_trace(clone, - "Merging weights for %s: initial sort index for %s is %d", + "Adding scores for %s: initial sort index for %s is %d", clone->id, child->id, child->sort_index); } - pe__show_node_weights(true, clone, "Before", clone->allowed_nodes, - clone->cluster); + pe__show_node_scores(true, clone, "Before", clone->allowed_nodes, + clone->cluster); /* Because the this_with_colocations() and with_this_colocations() methods * boil down to copies of rsc_cons and rsc_cons_lhs for clones, we can use * those here directly for efficiency. */ - g_list_foreach(clone->children, add_sort_index_to_node_weight, clone); + g_list_foreach(clone->children, add_sort_index_to_node_score, clone); g_list_foreach(clone->rsc_cons, apply_coloc_to_dependent, clone); g_list_foreach(clone->rsc_cons_lhs, apply_coloc_to_primary, clone); // Ban resource from all nodes if it needs a ticket but doesn't have it pcmk__require_promotion_tickets(clone); - pe__show_node_weights(true, clone, "After", clone->allowed_nodes, - clone->cluster); + pe__show_node_scores(true, clone, "After", clone->allowed_nodes, + clone->cluster); - // Reset sort indexes to final node weights - g_list_foreach(clone->children, set_sort_index_to_node_weight, clone); + // Reset sort indexes to final node scores + g_list_foreach(clone->children, set_sort_index_to_node_score, clone); // Finally, sort instances in descending order of promotion priority clone->children = g_list_sort(clone->children, cmp_promotable_instance); pe__clear_resource_flags(clone, pe_rsc_merging); } /*! * \internal * \brief Find the active instance (if any) of an anonymous clone on a node * * \param[in] clone Anonymous clone to check * \param[in] id Instance ID (without instance number) to check * \param[in] node Node to check * * \return */ static pe_resource_t * find_active_anon_instance(const pe_resource_t *clone, const char *id, const pe_node_t *node) { for (GList *iter = clone->children; iter; iter = iter->next) { pe_resource_t *child = iter->data; pe_resource_t *active = NULL; // Use ->find_rsc() in case this is a cloned group active = clone->fns->find_rsc(child, id, node, pe_find_clone|pe_find_current); if (active != NULL) { return active; } } return NULL; } /* * \brief Check whether an anonymous clone instance is known on a node * * \param[in] clone Anonymous clone to check * \param[in] id Instance ID (without instance number) to check * \param[in] node Node to check * * \return true if \p id instance of \p clone is known on \p node, * otherwise false */ static bool anonymous_known_on(const pe_resource_t *clone, const char *id, const pe_node_t *node) { for (GList *iter = clone->children; iter; iter = iter->next) { pe_resource_t *child = iter->data; /* Use ->find_rsc() because this might be a cloned group, and knowing * that other members of the group are known here implies nothing. */ child = clone->fns->find_rsc(child, id, NULL, pe_find_clone); CRM_LOG_ASSERT(child != NULL); if (child != NULL) { if (g_hash_table_lookup(child->known_on, node->details->id)) { return true; } } } return false; } /*! * \internal * \brief Check whether a node is allowed to run a resource * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return true if \p node is allowed to run \p rsc, otherwise false */ static bool is_allowed(const pe_resource_t *rsc, const pe_node_t *node) { pe_node_t *allowed = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); return (allowed != NULL) && (allowed->weight >= 0); } /*! * \brief Check whether a clone instance's promotion score should be considered * * \param[in] rsc Promotable clone instance to check * \param[in] node Node where score would be applied * * \return true if \p rsc's promotion score should be considered on \p node, * otherwise false */ static bool promotion_score_applies(const pe_resource_t *rsc, const pe_node_t *node) { char *id = clone_strip(rsc->id); const pe_resource_t *parent = pe__const_top_resource(rsc, false); pe_resource_t *active = NULL; const char *reason = "allowed"; // Some checks apply only to anonymous clone instances if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { // If instance is active on the node, its score definitely applies active = find_active_anon_instance(parent, id, node); if (active == rsc) { reason = "active"; goto check_allowed; } /* If *no* instance is active on this node, this instance's score will * count if it has been probed on this node. */ if ((active == NULL) && anonymous_known_on(parent, id, node)) { reason = "probed"; goto check_allowed; } } /* If this clone's status is unknown on *all* nodes (e.g. cluster startup), * take all instances' scores into account, to make sure we use any * permanent promotion scores. */ if ((rsc->running_on == NULL) && (g_hash_table_size(rsc->known_on) == 0)) { reason = "none probed"; goto check_allowed; } /* Otherwise, we've probed and/or started the resource *somewhere*, so * consider promotion scores on nodes where we know the status. */ if ((pe_hash_table_lookup(rsc->known_on, node->details->id) != NULL) || (pe_find_node_id(rsc->running_on, node->details->id) != NULL)) { reason = "known"; } else { pe_rsc_trace(rsc, "Ignoring %s promotion score (for %s) on %s: not probed", rsc->id, id, pe__node_name(node)); free(id); return false; } check_allowed: if (is_allowed(rsc, node)) { pe_rsc_trace(rsc, "Counting %s promotion score (for %s) on %s: %s", rsc->id, id, pe__node_name(node), reason); free(id); return true; } pe_rsc_trace(rsc, "Ignoring %s promotion score (for %s) on %s: not allowed", rsc->id, id, pe__node_name(node)); free(id); return false; } /*! * \internal * \brief Get the value of a promotion score node attribute * * \param[in] rsc Promotable clone instance to get promotion score for * \param[in] node Node to get promotion score for * \param[in] name Resource name to use in promotion score attribute name * * \return Value of promotion score node attribute for \p rsc on \p node */ static const char * promotion_attr_value(const pe_resource_t *rsc, const pe_node_t *node, const char *name) { char *attr_name = NULL; const char *attr_value = NULL; CRM_CHECK((rsc != NULL) && (node != NULL) && (name != NULL), return NULL); attr_name = pcmk_promotion_score_name(name); attr_value = pe_node_attribute_calculated(node, attr_name, rsc); free(attr_name); return attr_value; } /*! * \internal * \brief Get the promotion score for a clone instance on a node * * \param[in] rsc Promotable clone instance to get score for * \param[in] node Node to get score for * \param[out] is_default If non-NULL, will be set true if no score available * * \return Promotion score for \p rsc on \p node (or 0 if none) */ static int promotion_score(const pe_resource_t *rsc, const pe_node_t *node, bool *is_default) { char *name = NULL; const char *attr_value = NULL; if (is_default != NULL) { *is_default = true; } CRM_CHECK((rsc != NULL) && (node != NULL), return 0); /* If this is an instance of a cloned group, the promotion score is the sum * of all members' promotion scores. */ if (rsc->children != NULL) { int score = 0; for (const GList *iter = rsc->children; iter != NULL; iter = iter->next) { const pe_resource_t *child = (const pe_resource_t *) iter->data; bool child_default = false; int child_score = promotion_score(child, node, &child_default); if (!child_default && (is_default != NULL)) { *is_default = false; } score += child_score; } return score; } if (!promotion_score_applies(rsc, node)) { return 0; } /* For the promotion score attribute name, use the name the resource is * known as in resource history, since that's what crm_attribute --promotion * would have used. */ name = (rsc->clone_name == NULL)? rsc->id : rsc->clone_name; attr_value = promotion_attr_value(rsc, node, name); if (attr_value != NULL) { pe_rsc_trace(rsc, "Promotion score for %s on %s = %s", name, pe__node_name(node), pcmk__s(attr_value, "(unset)")); } else if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { /* If we don't have any resource history yet, we won't have clone_name. * In that case, for anonymous clones, try the resource name without * any instance number. */ name = clone_strip(rsc->id); if (strcmp(rsc->id, name) != 0) { attr_value = promotion_attr_value(rsc, node, name); pe_rsc_trace(rsc, "Promotion score for %s on %s (for %s) = %s", name, pe__node_name(node), rsc->id, pcmk__s(attr_value, "(unset)")); } free(name); } if (attr_value == NULL) { return 0; } if (is_default != NULL) { *is_default = false; } return char2score(attr_value); } /*! * \internal - * \brief Include promotion scores in instances' node weights and priorities + * \brief Include promotion scores in instances' node scores and priorities * * \param[in,out] rsc Promotable clone resource to update */ void pcmk__add_promotion_scores(pe_resource_t *rsc) { if (pe__set_clone_flag(rsc, pe__clone_promotion_added) == pcmk_rc_already) { return; } for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child_rsc = (pe_resource_t *) iter->data; GHashTableIter iter; pe_node_t *node = NULL; int score, new_score; g_hash_table_iter_init(&iter, child_rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (!pcmk__node_available(node, false, false)) { /* This node will never be promoted, so don't apply the * promotion score, as that may lead to clone shuffling. */ continue; } score = promotion_score(child_rsc, node, NULL); if (score > 0) { new_score = pcmk__add_scores(node->weight, score); if (new_score != node->weight) { // Could remain INFINITY node->weight = new_score; pe_rsc_trace(rsc, "Added %s promotion priority (%s) to score " "on %s (now %s)", child_rsc->id, pcmk_readable_score(score), pe__node_name(node), pcmk_readable_score(new_score)); } } if (score > child_rsc->priority) { pe_rsc_trace(rsc, "Updating %s priority to promotion score (%d->%d)", child_rsc->id, child_rsc->priority, score); child_rsc->priority = score; } } } } /*! * \internal * \brief If a resource's current role is started, change it to unpromoted * * \param[in,out] data Resource to update * \param[in] user_data Ignored */ static void set_current_role_unpromoted(void *data, void *user_data) { pe_resource_t *rsc = (pe_resource_t *) data; if (rsc->role == RSC_ROLE_STARTED) { // Promotable clones should use unpromoted role instead of started rsc->role = RSC_ROLE_UNPROMOTED; } g_list_foreach(rsc->children, set_current_role_unpromoted, NULL); } /*! * \internal * \brief Set a resource's next role to unpromoted (or stopped if unassigned) * * \param[in,out] data Resource to update * \param[in] user_data Ignored */ static void set_next_role_unpromoted(void *data, void *user_data) { pe_resource_t *rsc = (pe_resource_t *) data; GList *assigned = NULL; rsc->fns->location(rsc, &assigned, FALSE); if (assigned == NULL) { pe__set_next_role(rsc, RSC_ROLE_STOPPED, "stopped instance"); } else { pe__set_next_role(rsc, RSC_ROLE_UNPROMOTED, "unpromoted instance"); g_list_free(assigned); } g_list_foreach(rsc->children, set_next_role_unpromoted, NULL); } /*! * \internal * \brief Set a resource's next role to promoted if not already set * * \param[in,out] data Resource to update * \param[in] user_data Ignored */ static void set_next_role_promoted(void *data, gpointer user_data) { pe_resource_t *rsc = (pe_resource_t *) data; if (rsc->next_role == RSC_ROLE_UNKNOWN) { pe__set_next_role(rsc, RSC_ROLE_PROMOTED, "promoted instance"); } g_list_foreach(rsc->children, set_next_role_promoted, NULL); } /*! * \internal * \brief Show instance's promotion score on node where it will be active * * \param[in,out] instance Promotable clone instance to show */ static void show_promotion_score(pe_resource_t *instance) { pe_node_t *chosen = instance->fns->location(instance, NULL, FALSE); if (pcmk_is_set(instance->cluster->flags, pe_flag_show_scores) && !pcmk__is_daemon && (instance->cluster->priv != NULL)) { pcmk__output_t *out = instance->cluster->priv; out->message(out, "promotion-score", instance, chosen, pcmk_readable_score(instance->sort_index)); } else { pe_rsc_debug(pe__const_top_resource(instance, false), "%s promotion score on %s: sort=%s priority=%s", instance->id, ((chosen == NULL)? "none" : pe__node_name(chosen)), pcmk_readable_score(instance->sort_index), pcmk_readable_score(instance->priority)); } } /*! * \internal * \brief Set a clone instance's promotion priority * * \param[in,out] data Promotable clone instance to update * \param[in] user_data Instance's parent clone */ static void set_instance_priority(gpointer data, gpointer user_data) { pe_resource_t *instance = (pe_resource_t *) data; const pe_resource_t *clone = (const pe_resource_t *) user_data; const pe_node_t *chosen = NULL; enum rsc_role_e next_role = RSC_ROLE_UNKNOWN; GList *list = NULL; pe_rsc_trace(clone, "Assigning priority for %s: %s", instance->id, role2text(instance->next_role)); if (instance->fns->state(instance, TRUE) == RSC_ROLE_STARTED) { set_current_role_unpromoted(instance, NULL); } // Only an instance that will be active can be promoted chosen = instance->fns->location(instance, &list, FALSE); if (pcmk__list_of_multiple(list)) { pcmk__config_err("Cannot promote non-colocated child %s", instance->id); } g_list_free(list); if (chosen == NULL) { return; } next_role = instance->fns->state(instance, FALSE); switch (next_role) { case RSC_ROLE_STARTED: case RSC_ROLE_UNKNOWN: // Set instance priority to its promotion score (or -1 if none) { bool is_default = false; instance->priority = promotion_score(instance, chosen, &is_default); if (is_default) { /* * Default to -1 if no value is set. This allows * instances eligible for promotion to be specified * based solely on rsc_location constraints, but * prevents any instance from being promoted if neither * a constraint nor a promotion score is present */ instance->priority = -1; } } break; case RSC_ROLE_UNPROMOTED: case RSC_ROLE_STOPPED: // Instance can't be promoted instance->priority = -INFINITY; break; case RSC_ROLE_PROMOTED: // Nothing needed (re-creating actions after scheduling fencing) break; default: CRM_CHECK(FALSE, crm_err("Unknown resource role %d for %s", next_role, instance->id)); } // Add relevant location constraint scores for promoted role apply_promoted_locations(instance, instance->rsc_location, chosen); apply_promoted_locations(instance, clone->rsc_location, chosen); // Consider instance's role-based colocations with other resources list = pcmk__this_with_colocations(instance); for (GList *iter = list; iter != NULL; iter = iter->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) iter->data; instance->cmds->apply_coloc_score(instance, cons->primary, cons, true); } g_list_free(list); instance->sort_index = instance->priority; if (next_role == RSC_ROLE_PROMOTED) { instance->sort_index = INFINITY; } pe_rsc_trace(clone, "Assigning %s priority = %d", instance->id, instance->priority); } /*! * \internal * \brief Set a promotable clone instance's role * * \param[in,out] data Promotable clone instance to update * \param[in,out] user_data Pointer to count of instances chosen for promotion */ static void set_instance_role(gpointer data, gpointer user_data) { pe_resource_t *instance = (pe_resource_t *) data; int *count = (int *) user_data; const pe_resource_t *clone = pe__const_top_resource(instance, false); pe_node_t *chosen = NULL; show_promotion_score(instance); if (instance->sort_index < 0) { pe_rsc_trace(clone, "Not supposed to promote instance %s", instance->id); } else if ((*count < pe__clone_promoted_max(instance)) || !pcmk_is_set(clone->flags, pe_rsc_managed)) { chosen = node_to_be_promoted_on(instance); } if (chosen == NULL) { set_next_role_unpromoted(instance, NULL); return; } if ((instance->role < RSC_ROLE_PROMOTED) && !pcmk_is_set(instance->cluster->flags, pe_flag_have_quorum) && (instance->cluster->no_quorum_policy == no_quorum_freeze)) { crm_notice("Clone instance %s cannot be promoted without quorum", instance->id); set_next_role_unpromoted(instance, NULL); return; } chosen->count++; pe_rsc_info(clone, "Choosing %s (%s) on %s for promotion", instance->id, role2text(instance->role), pe__node_name(chosen)); set_next_role_promoted(instance, NULL); (*count)++; } /*! * \internal * \brief Set roles for all instances of a promotable clone * * \param[in,out] rsc Promotable clone resource to update */ void pcmk__set_instance_roles(pe_resource_t *rsc) { int promoted = 0; GHashTableIter iter; pe_node_t *node = NULL; - // Repurpose count to track the number of promoted instances allocated + // Repurpose count to track the number of promoted instances assigned g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { node->count = 0; } // Set instances' promotion priorities and sort by highest priority first g_list_foreach(rsc->children, set_instance_priority, rsc); sort_promotable_instances(rsc); // Choose the first N eligible instances to be promoted g_list_foreach(rsc->children, set_instance_role, &promoted); pe_rsc_info(rsc, "%s: Promoted %d instances of a possible %d", rsc->id, promoted, pe__clone_promoted_max(rsc)); } /*! * * \internal * \brief Create actions for promotable clone instances * * \param[in,out] clone Promotable clone to create actions for * \param[out] any_promoting Will be set true if any instance is promoting * \param[out] any_demoting Will be set true if any instance is demoting */ static void create_promotable_instance_actions(pe_resource_t *clone, bool *any_promoting, bool *any_demoting) { for (GList *iter = clone->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->cmds->create_actions(instance); check_for_role_change(instance, any_demoting, any_promoting); } } /*! * \internal * \brief Reset each promotable instance's resource priority * * Reset the priority of each instance of a promotable clone to the clone's * priority (after promotion actions are scheduled, when instance priorities * were repurposed as promotion scores). * * \param[in,out] clone Promotable clone to reset */ static void reset_instance_priorities(pe_resource_t *clone) { for (GList *iter = clone->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; instance->priority = clone->priority; } } /*! * \internal * \brief Create actions specific to promotable clones * * \param[in,out] clone Promotable clone to create actions for */ void pcmk__create_promotable_actions(pe_resource_t *clone) { bool any_promoting = false; bool any_demoting = false; // Create actions for each clone instance individually create_promotable_instance_actions(clone, &any_promoting, &any_demoting); // Create pseudo-actions for clone as a whole pe__create_promotable_pseudo_ops(clone, any_promoting, any_demoting); // Undo our temporary repurposing of resource priority for instances reset_instance_priorities(clone); } /*! * \internal * \brief Create internal orderings for a promotable clone's instances * * \param[in,out] clone Promotable clone instance to order */ void pcmk__order_promotable_instances(pe_resource_t *clone) { pe_resource_t *previous = NULL; // Needed for ordered clones pcmk__promotable_restart_ordering(clone); for (GList *iter = clone->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; // Demote before promote pcmk__order_resource_actions(instance, RSC_DEMOTE, instance, RSC_PROMOTE, pe_order_optional); order_instance_promotion(clone, instance, previous); order_instance_demotion(clone, instance, previous); previous = instance; } } /*! * \internal * \brief Update dependent's allowed nodes for colocation with promotable * * \param[in,out] dependent Dependent resource to update * \param[in] primary_node Node where an instance of the primary will be * \param[in] colocation Colocation constraint to apply */ static void update_dependent_allowed_nodes(pe_resource_t *dependent, const pe_node_t *primary_node, const pcmk__colocation_t *colocation) { GHashTableIter iter; pe_node_t *node = NULL; const char *primary_value = NULL; const char *attr = NULL; if (colocation->score >= INFINITY) { return; // Colocation is mandatory, so allowed node scores don't matter } // Get value of primary's colocation node attribute attr = colocation->node_attribute; if (attr == NULL) { attr = CRM_ATTR_UNAME; } primary_value = pe_node_attribute_raw(primary_node, attr); pe_rsc_trace(colocation->primary, "Applying %s (%s with %s on %s by %s @%d) to %s", colocation->id, colocation->dependent->id, colocation->primary->id, pe__node_name(primary_node), attr, colocation->score, dependent->id); g_hash_table_iter_init(&iter, dependent->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { const char *dependent_value = pe_node_attribute_raw(node, attr); if (pcmk__str_eq(primary_value, dependent_value, pcmk__str_casei)) { node->weight = pcmk__add_scores(node->weight, colocation->score); pe_rsc_trace(colocation->primary, "Added %s score (%s) to %s (now %s)", colocation->id, pcmk_readable_score(colocation->score), pe__node_name(node), pcmk_readable_score(node->weight)); } } } /*! * \brief Update dependent for a colocation with a promotable clone * * \param[in] primary Primary resource in the colocation * \param[in,out] dependent Dependent resource in the colocation * \param[in] colocation Colocation constraint to apply */ void pcmk__update_dependent_with_promotable(const pe_resource_t *primary, pe_resource_t *dependent, const pcmk__colocation_t *colocation) { GList *affected_nodes = NULL; /* Build a list of all nodes where an instance of the primary will be, and * (for optional colocations) update the dependent's allowed node scores for * each one. */ for (GList *iter = primary->children; iter != NULL; iter = iter->next) { pe_resource_t *instance = (pe_resource_t *) iter->data; pe_node_t *node = instance->fns->location(instance, NULL, FALSE); if (node == NULL) { continue; } if (instance->fns->state(instance, FALSE) == colocation->primary_role) { update_dependent_allowed_nodes(dependent, node, colocation); affected_nodes = g_list_prepend(affected_nodes, node); } } - /* For mandatory colocations, add the primary's node weight to the - * dependent's node weight for each affected node, and ban the dependent + /* For mandatory colocations, add the primary's node score to the + * dependent's node score for each affected node, and ban the dependent * from all other nodes. * * However, skip this for promoted-with-promoted colocations, otherwise * inactive dependent instances can't start (in the unpromoted role). */ if ((colocation->score >= INFINITY) && ((colocation->dependent_role != RSC_ROLE_PROMOTED) || (colocation->primary_role != RSC_ROLE_PROMOTED))) { pe_rsc_trace(colocation->primary, "Applying %s (mandatory %s with %s) to %s", colocation->id, colocation->dependent->id, colocation->primary->id, dependent->id); node_list_exclude(dependent->allowed_nodes, affected_nodes, TRUE); } g_list_free(affected_nodes); } /*! * \internal * \brief Update dependent priority for colocation with promotable * * \param[in] primary Primary resource in the colocation * \param[in,out] dependent Dependent resource in the colocation * \param[in] colocation Colocation constraint to apply */ void pcmk__update_promotable_dependent_priority(const pe_resource_t *primary, pe_resource_t *dependent, const pcmk__colocation_t *colocation) { pe_resource_t *primary_instance = NULL; // Look for a primary instance where dependent will be primary_instance = pcmk__find_compatible_instance(dependent, primary, colocation->primary_role, false); if (primary_instance != NULL) { // Add primary instance's priority to dependent's int new_priority = pcmk__add_scores(dependent->priority, colocation->score); pe_rsc_trace(colocation->primary, "Applying %s (%s with %s) to %s priority (%s + %s = %s)", colocation->id, colocation->dependent->id, colocation->primary->id, dependent->id, pcmk_readable_score(dependent->priority), pcmk_readable_score(colocation->score), pcmk_readable_score(new_priority)); dependent->priority = new_priority; } else if (colocation->score >= INFINITY) { // Mandatory colocation, but primary won't be here pe_rsc_trace(colocation->primary, "Applying %s (%s with %s) to %s: can't be promoted", colocation->id, colocation->dependent->id, colocation->primary->id, dependent->id); dependent->priority = -INFINITY; } } diff --git a/lib/pacemaker/pcmk_sched_remote.c b/lib/pacemaker/pcmk_sched_remote.c index 6adb5d4d51..dabb78fafc 100644 --- a/lib/pacemaker/pcmk_sched_remote.c +++ b/lib/pacemaker/pcmk_sched_remote.c @@ -1,729 +1,729 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include "libpacemaker_private.h" enum remote_connection_state { remote_state_unknown = 0, remote_state_alive = 1, remote_state_resting = 2, remote_state_failed = 3, remote_state_stopped = 4 }; static const char * state2text(enum remote_connection_state state) { switch (state) { case remote_state_unknown: return "unknown"; case remote_state_alive: return "alive"; case remote_state_resting: return "resting"; case remote_state_failed: return "failed"; case remote_state_stopped: return "stopped"; } return "impossible"; } /* We always use pe_order_preserve with these convenience functions to exempt * internally generated constraints from the prohibition of user constraints * involving remote connection resources. * * The start ordering additionally uses pe_order_runnable_left so that the * specified action is not runnable if the start is not runnable. */ static inline void order_start_then_action(pe_resource_t *first_rsc, pe_action_t *then_action, uint32_t extra, pe_working_set_t *data_set) { if ((first_rsc != NULL) && (then_action != NULL) && (data_set != NULL)) { pcmk__new_ordering(first_rsc, start_key(first_rsc), NULL, then_action->rsc, NULL, then_action, pe_order_preserve|pe_order_runnable_left|extra, data_set); } } static inline void order_action_then_stop(pe_action_t *first_action, pe_resource_t *then_rsc, uint32_t extra, pe_working_set_t *data_set) { if ((first_action != NULL) && (then_rsc != NULL) && (data_set != NULL)) { pcmk__new_ordering(first_action->rsc, NULL, first_action, then_rsc, stop_key(then_rsc), NULL, pe_order_preserve|extra, data_set); } } static enum remote_connection_state get_remote_node_state(const pe_node_t *node) { const pe_resource_t *remote_rsc = NULL; const pe_node_t *cluster_node = NULL; CRM_ASSERT(node != NULL); remote_rsc = node->details->remote_rsc; CRM_ASSERT(remote_rsc != NULL); cluster_node = pe__current_node(remote_rsc); /* If the cluster node the remote connection resource resides on * is unclean or went offline, we can't process any operations * on that remote node until after it starts elsewhere. */ if ((remote_rsc->next_role == RSC_ROLE_STOPPED) || (remote_rsc->allocated_to == NULL)) { // The connection resource is not going to run anywhere if ((cluster_node != NULL) && cluster_node->details->unclean) { /* The remote connection is failed because its resource is on a * failed node and can't be recovered elsewhere, so we must fence. */ return remote_state_failed; } if (!pcmk_is_set(remote_rsc->flags, pe_rsc_failed)) { /* Connection resource is cleanly stopped */ return remote_state_stopped; } /* Connection resource is failed */ if ((remote_rsc->next_role == RSC_ROLE_STOPPED) && remote_rsc->remote_reconnect_ms && node->details->remote_was_fenced && !pe__shutdown_requested(node)) { /* We won't know whether the connection is recoverable until the * reconnect interval expires and we reattempt connection. */ return remote_state_unknown; } /* The remote connection is in a failed state. If there are any * resources known to be active on it (stop) or in an unknown state * (probe), we must assume the worst and fence it. */ return remote_state_failed; } else if (cluster_node == NULL) { /* Connection is recoverable but not currently running anywhere, so see * if we can recover it first */ return remote_state_unknown; } else if (cluster_node->details->unclean || !(cluster_node->details->online)) { // Connection is running on a dead node, see if we can recover it first return remote_state_resting; } else if (pcmk__list_of_multiple(remote_rsc->running_on) && (remote_rsc->partial_migration_source != NULL) && (remote_rsc->partial_migration_target != NULL)) { /* We're in the middle of migrating a connection resource, so wait until * after the migration completes before performing any actions. */ return remote_state_resting; } return remote_state_alive; } /*! * \internal * \brief Order actions on remote node relative to actions for the connection * * \param[in,out] action An action scheduled on a Pacemaker Remote node */ static void apply_remote_ordering(pe_action_t *action) { pe_resource_t *remote_rsc = NULL; enum action_tasks task = text2task(action->task); enum remote_connection_state state = get_remote_node_state(action->node); uint32_t order_opts = pe_order_none; if (action->rsc == NULL) { return; } CRM_ASSERT(pe__is_guest_or_remote_node(action->node)); remote_rsc = action->node->details->remote_rsc; CRM_ASSERT(remote_rsc != NULL); crm_trace("Order %s action %s relative to %s%s (state: %s)", action->task, action->uuid, pcmk_is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "", remote_rsc->id, state2text(state)); if (pcmk__strcase_any_of(action->task, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { /* Migration ops map to "no_action", but we need to apply the same * ordering as for stop or demote (see get_router_node()). */ task = stop_rsc; } switch (task) { case start_rsc: case action_promote: order_opts = pe_order_none; if (state == remote_state_failed) { /* Force recovery, by making this action required */ pe__set_order_flags(order_opts, pe_order_implies_then); } /* Ensure connection is up before running this action */ order_start_then_action(remote_rsc, action, order_opts, remote_rsc->cluster); break; case stop_rsc: if (state == remote_state_alive) { order_action_then_stop(action, remote_rsc, pe_order_implies_first, remote_rsc->cluster); } else if (state == remote_state_failed) { /* The resource is active on the node, but since we don't have a * valid connection, the only way to stop the resource is by * fencing the node. There is no need to order the stop relative * to the remote connection, since the stop will become implied * by the fencing. */ pe_fence_node(remote_rsc->cluster, action->node, "resources are active but connection is unrecoverable", FALSE); } else if (remote_rsc->next_role == RSC_ROLE_STOPPED) { /* State must be remote_state_unknown or remote_state_stopped. * Since the connection is not coming back up in this * transition, stop this resource first. */ order_action_then_stop(action, remote_rsc, pe_order_implies_first, remote_rsc->cluster); } else { /* The connection is going to be started somewhere else, so * stop this resource after that completes. */ order_start_then_action(remote_rsc, action, pe_order_none, remote_rsc->cluster); } break; case action_demote: /* Only order this demote relative to the connection start if the * connection isn't being torn down. Otherwise, the demote would be * blocked because the connection start would not be allowed. */ if ((state == remote_state_resting) || (state == remote_state_unknown)) { order_start_then_action(remote_rsc, action, pe_order_none, remote_rsc->cluster); } /* Otherwise we can rely on the stop ordering */ break; default: /* Wait for the connection resource to be up */ if (pcmk__action_is_recurring(action)) { /* In case we ever get the recovery logic wrong, force * recurring monitors to be restarted, even if just * the connection was re-established */ order_start_then_action(remote_rsc, action, pe_order_implies_then, remote_rsc->cluster); } else { pe_node_t *cluster_node = pe__current_node(remote_rsc); if ((task == monitor_rsc) && (state == remote_state_failed)) { /* We would only be here if we do not know the state of the * resource on the remote node. Since we have no way to find * out, it is necessary to fence the node. */ pe_fence_node(remote_rsc->cluster, action->node, "resources are in unknown state " "and connection is unrecoverable", FALSE); } if ((cluster_node != NULL) && (state == remote_state_stopped)) { /* The connection is currently up, but is going down * permanently. Make sure we check services are actually * stopped _before_ we let the connection get closed. */ order_action_then_stop(action, remote_rsc, pe_order_runnable_left, remote_rsc->cluster); } else { order_start_then_action(remote_rsc, action, pe_order_none, remote_rsc->cluster); } } break; } } static void apply_container_ordering(pe_action_t *action, pe_working_set_t *data_set) { /* VMs are also classified as containers for these purposes... in * that they both involve a 'thing' running on a real or remote * cluster node. * * This allows us to be smarter about the type and extent of * recovery actions required in various scenarios */ pe_resource_t *remote_rsc = NULL; pe_resource_t *container = NULL; enum action_tasks task = text2task(action->task); CRM_ASSERT(action->rsc != NULL); CRM_ASSERT(action->node != NULL); CRM_ASSERT(pe__is_guest_or_remote_node(action->node)); remote_rsc = action->node->details->remote_rsc; CRM_ASSERT(remote_rsc != NULL); container = remote_rsc->container; CRM_ASSERT(container != NULL); if (pcmk_is_set(container->flags, pe_rsc_failed)) { pe_fence_node(data_set, action->node, "container failed", FALSE); } crm_trace("Order %s action %s relative to %s%s for %s%s", action->task, action->uuid, pcmk_is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "", remote_rsc->id, pcmk_is_set(container->flags, pe_rsc_failed)? "failed " : "", container->id); if (pcmk__strcase_any_of(action->task, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { /* Migration ops map to "no_action", but we need to apply the same * ordering as for stop or demote (see get_router_node()). */ task = stop_rsc; } switch (task) { case start_rsc: case action_promote: // Force resource recovery if the container is recovered order_start_then_action(container, action, pe_order_implies_then, data_set); // Wait for the connection resource to be up, too order_start_then_action(remote_rsc, action, pe_order_none, data_set); break; case stop_rsc: case action_demote: if (pcmk_is_set(container->flags, pe_rsc_failed)) { /* When the container representing a guest node fails, any stop * or demote actions for resources running on the guest node * are implied by the container stopping. This is similar to * how fencing operations work for cluster nodes and remote * nodes. */ } else { /* Ensure the operation happens before the connection is brought * down. * * If we really wanted to, we could order these after the * connection start, IFF the container's current role was * stopped (otherwise we re-introduce an ordering loop when the * connection is restarting). */ order_action_then_stop(action, remote_rsc, pe_order_none, data_set); } break; default: /* Wait for the connection resource to be up */ if (pcmk__action_is_recurring(action)) { /* In case we ever get the recovery logic wrong, force * recurring monitors to be restarted, even if just * the connection was re-established */ if(task != no_action) { order_start_then_action(remote_rsc, action, pe_order_implies_then, data_set); } } else { order_start_then_action(remote_rsc, action, pe_order_none, data_set); } break; } } /*! * \internal * \brief Order all relevant actions relative to remote connection actions * * \param[in,out] data_set Cluster working set */ void pcmk__order_remote_connection_actions(pe_working_set_t *data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_have_remote_nodes)) { return; } crm_trace("Creating remote connection orderings"); for (GList *gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; pe_resource_t *remote = NULL; // We are only interested in resource actions if (action->rsc == NULL) { continue; } /* Special case: If we are clearing the failcount of an actual * remote connection resource, then make sure this happens before * any start of the resource in this transition. */ if (action->rsc->is_remote_node && pcmk__str_eq(action->task, CRM_OP_CLEAR_FAILCOUNT, pcmk__str_casei)) { pcmk__new_ordering(action->rsc, NULL, action, action->rsc, pcmk__op_key(action->rsc->id, RSC_START, 0), NULL, pe_order_optional, data_set); continue; } - // We are only interested in actions allocated to a node + // We are only interested in actions assigned to a node if (action->node == NULL) { continue; } if (!pe__is_guest_or_remote_node(action->node)) { continue; } /* We are only interested in real actions. * * @TODO This is probably wrong; pseudo-actions might be converted to * real actions and vice versa later in update_actions() at the end of * pcmk__apply_orderings(). */ if (pcmk_is_set(action->flags, pe_action_pseudo)) { continue; } remote = action->node->details->remote_rsc; if (remote == NULL) { // Orphaned continue; } /* Another special case: if a resource is moving to a Pacemaker Remote * node, order the stop on the original node after any start of the * remote connection. This ensures that if the connection fails to * start, we leave the resource running on the original node. */ if (pcmk__str_eq(action->task, RSC_START, pcmk__str_casei)) { for (GList *item = action->rsc->actions; item != NULL; item = item->next) { pe_action_t *rsc_action = item->data; - if ((rsc_action->node->details != action->node->details) + if (!pe__same_node(rsc_action->node, action->node) && pcmk__str_eq(rsc_action->task, RSC_STOP, pcmk__str_casei)) { pcmk__new_ordering(remote, start_key(remote), NULL, action->rsc, NULL, rsc_action, pe_order_optional, data_set); } } } /* The action occurs across a remote connection, so create * ordering constraints that guarantee the action occurs while the node * is active (after start, before stop ... things like that). * * This is somewhat brittle in that we need to make sure the results of * this ordering are compatible with the result of get_router_node(). * It would probably be better to add XML_LRM_ATTR_ROUTER_NODE as part * of this logic rather than create_graph_action(). */ if (remote->container) { crm_trace("Container ordering for %s", action->uuid); apply_container_ordering(action, data_set); } else { crm_trace("Remote ordering for %s", action->uuid); apply_remote_ordering(action); } } } /*! * \internal * \brief Check whether a node is a failed remote node * * \param[in] node Node to check * * \return true if \p node is a failed remote node, false otherwise */ bool pcmk__is_failed_remote_node(const pe_node_t *node) { return pe__is_remote_node(node) && (node->details->remote_rsc != NULL) && (get_remote_node_state(node) == remote_state_failed); } /*! * \internal * \brief Check whether a given resource corresponds to a given node as guest * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return true if \p node is a guest node and \p rsc is its containing * resource, otherwise false */ bool pcmk__rsc_corresponds_to_guest(const pe_resource_t *rsc, const pe_node_t *node) { return (rsc != NULL) && (rsc->fillers != NULL) && (node != NULL) && (node->details->remote_rsc != NULL) && (node->details->remote_rsc->container == rsc); } /*! * \internal * \brief Get proper connection host that a remote action must be routed through * * A remote connection resource might be starting, stopping, or migrating in the * same transition that an action needs to be executed on its Pacemaker Remote * node. Determine the proper node that the remote action should be routed * through. * * \param[in] action (Potentially remote) action to route * * \return Connection host that action should be routed through if remote, * otherwise NULL */ pe_node_t * pcmk__connection_host_for_action(const pe_action_t *action) { pe_node_t *began_on = NULL; pe_node_t *ended_on = NULL; bool partial_migration = false; const char *task = action->task; if (pcmk__str_eq(task, CRM_OP_FENCE, pcmk__str_casei) || !pe__is_guest_or_remote_node(action->node)) { return NULL; } CRM_ASSERT(action->node->details->remote_rsc != NULL); began_on = pe__current_node(action->node->details->remote_rsc); ended_on = action->node->details->remote_rsc->allocated_to; if (action->node->details->remote_rsc && (action->node->details->remote_rsc->container == NULL) && action->node->details->remote_rsc->partial_migration_target) { partial_migration = true; } if (began_on == NULL) { crm_trace("Routing %s for %s through remote connection's " "next node %s (starting)%s", action->task, (action->rsc? action->rsc->id : "no resource"), (ended_on? ended_on->details->uname : "none"), partial_migration? " (partial migration)" : ""); return ended_on; } if (ended_on == NULL) { crm_trace("Routing %s for %s through remote connection's " "current node %s (stopping)%s", action->task, (action->rsc? action->rsc->id : "no resource"), (began_on? began_on->details->uname : "none"), partial_migration? " (partial migration)" : ""); return began_on; } - if (began_on->details == ended_on->details) { + if (pe__same_node(began_on, ended_on)) { crm_trace("Routing %s for %s through remote connection's " "current node %s (not moving)%s", action->task, (action->rsc? action->rsc->id : "no resource"), (began_on? began_on->details->uname : "none"), partial_migration? " (partial migration)" : ""); return began_on; } /* If we get here, the remote connection is moving during this transition. * This means some actions for resources behind the connection will get * routed through the cluster node the connection resource is currently on, * and others are routed through the cluster node the connection will end up * on. */ if (pcmk__str_eq(task, "notify", pcmk__str_casei)) { task = g_hash_table_lookup(action->meta, "notify_operation"); } /* * Stop, demote, and migration actions must occur before the connection can * move (these actions are required before the remote resource can stop). In * this case, we know these actions have to be routed through the initial * cluster node the connection resource lived on before the move takes * place. * * The exception is a partial migration of a (non-guest) remote connection * resource; in that case, all actions (even these) will be ordered after * the connection's pseudo-start on the migration target, so the target is * the router node. */ if (pcmk__strcase_any_of(task, "cancel", "stop", "demote", "migrate_from", "migrate_to", NULL) && !partial_migration) { crm_trace("Routing %s for %s through remote connection's " "current node %s (moving)%s", action->task, (action->rsc? action->rsc->id : "no resource"), (began_on? began_on->details->uname : "none"), partial_migration? " (partial migration)" : ""); return began_on; } /* Everything else (start, promote, monitor, probe, refresh, * clear failcount, delete, ...) must occur after the connection starts on * the node it is moving to. */ crm_trace("Routing %s for %s through remote connection's " "next node %s (moving)%s", action->task, (action->rsc? action->rsc->id : "no resource"), (ended_on? ended_on->details->uname : "none"), partial_migration? " (partial migration)" : ""); return ended_on; } /*! * \internal * \brief Replace remote connection's addr="#uname" with actual address * * REMOTE_CONTAINER_HACK: If a given resource is a remote connection resource * with its "addr" parameter set to "#uname", pull the actual value from the * parameters evaluated without a node (which was put there earlier in * pcmk__create_graph() when the bundle's expand() method was called). * * \param[in,out] rsc Resource to check * \param[in,out] params Resource parameters evaluated per node */ void pcmk__substitute_remote_addr(pe_resource_t *rsc, GHashTable *params) { const char *remote_addr = g_hash_table_lookup(params, XML_RSC_ATTR_REMOTE_RA_ADDR); if (pcmk__str_eq(remote_addr, "#uname", pcmk__str_none)) { GHashTable *base = pe_rsc_params(rsc, NULL, rsc->cluster); remote_addr = g_hash_table_lookup(base, XML_RSC_ATTR_REMOTE_RA_ADDR); if (remote_addr != NULL) { g_hash_table_insert(params, strdup(XML_RSC_ATTR_REMOTE_RA_ADDR), strdup(remote_addr)); } } } /*! * \brief Add special bundle meta-attributes to XML * * If a given action will be executed on a guest node (including a bundle), * add the special bundle meta-attribute "container-attribute-target" and * environment variable "physical_host" as XML attributes (using meta-attribute * naming). * * \param[in,out] args_xml XML to add attributes to * \param[in] action Action to check */ void pcmk__add_bundle_meta_to_xml(xmlNode *args_xml, const pe_action_t *action) { const pe_node_t *host = NULL; enum action_tasks task; if (!pe__is_guest_node(action->node)) { return; } task = text2task(action->task); if ((task == action_notify) || (task == action_notified)) { task = text2task(g_hash_table_lookup(action->meta, "notify_operation")); } switch (task) { case stop_rsc: case stopped_rsc: case action_demote: case action_demoted: // "Down" actions take place on guest's current host host = pe__current_node(action->node->details->remote_rsc->container); break; case start_rsc: case started_rsc: case monitor_rsc: case action_promote: case action_promoted: // "Up" actions take place on guest's next host host = action->node->details->remote_rsc->container->allocated_to; break; default: break; } if (host != NULL) { hash2metafield((gpointer) XML_RSC_ATTR_TARGET, (gpointer) g_hash_table_lookup(action->rsc->meta, XML_RSC_ATTR_TARGET), (gpointer) args_xml); hash2metafield((gpointer) PCMK__ENV_PHYSICAL_HOST, (gpointer) host->details->uname, (gpointer) args_xml); } } diff --git a/lib/pacemaker/pcmk_sched_resource.c b/lib/pacemaker/pcmk_sched_resource.c index 11efbd1b19..173948684e 100644 --- a/lib/pacemaker/pcmk_sched_resource.c +++ b/lib/pacemaker/pcmk_sched_resource.c @@ -1,722 +1,723 @@ /* * Copyright 2014-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include "libpacemaker_private.h" -// Resource allocation methods that vary by resource variant -static resource_alloc_functions_t allocation_methods[] = { +// Resource assignment methods by resource variant +static resource_alloc_functions_t assignment_methods[] = { { pcmk__primitive_assign, pcmk__primitive_create_actions, pcmk__probe_rsc_on_node, pcmk__primitive_internal_constraints, pcmk__primitive_apply_coloc_score, pcmk__colocated_resources, pcmk__with_primitive_colocations, pcmk__primitive_with_colocations, pcmk__add_colocated_node_scores, pcmk__apply_location, pcmk__primitive_action_flags, pcmk__update_ordered_actions, pcmk__output_resource_actions, pcmk__add_rsc_actions_to_graph, pcmk__primitive_add_graph_meta, pcmk__primitive_add_utilization, pcmk__primitive_shutdown_lock, }, { pcmk__group_assign, pcmk__group_create_actions, pcmk__probe_rsc_on_node, pcmk__group_internal_constraints, pcmk__group_apply_coloc_score, pcmk__group_colocated_resources, pcmk__with_group_colocations, pcmk__group_with_colocations, pcmk__group_add_colocated_node_scores, pcmk__group_apply_location, pcmk__group_action_flags, pcmk__group_update_ordered_actions, pcmk__output_resource_actions, pcmk__add_rsc_actions_to_graph, pcmk__noop_add_graph_meta, pcmk__group_add_utilization, pcmk__group_shutdown_lock, }, { pcmk__clone_assign, pcmk__clone_create_actions, pcmk__clone_create_probe, pcmk__clone_internal_constraints, pcmk__clone_apply_coloc_score, pcmk__colocated_resources, pcmk__with_clone_colocations, pcmk__clone_with_colocations, pcmk__add_colocated_node_scores, pcmk__clone_apply_location, pcmk__clone_action_flags, pcmk__instance_update_ordered_actions, pcmk__output_resource_actions, pcmk__clone_add_actions_to_graph, pcmk__clone_add_graph_meta, pcmk__clone_add_utilization, pcmk__clone_shutdown_lock, }, { pcmk__bundle_assign, pcmk__bundle_create_actions, pcmk__bundle_create_probe, pcmk__bundle_internal_constraints, pcmk__bundle_apply_coloc_score, pcmk__colocated_resources, pcmk__with_bundle_colocations, pcmk__bundle_with_colocations, pcmk__add_colocated_node_scores, pcmk__bundle_apply_location, pcmk__bundle_action_flags, pcmk__instance_update_ordered_actions, pcmk__output_bundle_actions, pcmk__bundle_add_actions_to_graph, pcmk__noop_add_graph_meta, pcmk__bundle_add_utilization, pcmk__bundle_shutdown_lock, } }; /*! * \internal * \brief Check whether a resource's agent standard, provider, or type changed * * \param[in,out] rsc Resource to check * \param[in,out] node Node needing unfencing if agent changed * \param[in] rsc_entry XML with previously known agent information * \param[in] active_on_node Whether \p rsc is active on \p node * * \return true if agent for \p rsc changed, otherwise false */ bool pcmk__rsc_agent_changed(pe_resource_t *rsc, pe_node_t *node, const xmlNode *rsc_entry, bool active_on_node) { bool changed = false; const char *attr_list[] = { XML_ATTR_TYPE, XML_AGENT_ATTR_CLASS, XML_AGENT_ATTR_PROVIDER }; for (int i = 0; i < PCMK__NELEM(attr_list); i++) { const char *value = crm_element_value(rsc->xml, attr_list[i]); const char *old_value = crm_element_value(rsc_entry, attr_list[i]); if (!pcmk__str_eq(value, old_value, pcmk__str_none)) { changed = true; trigger_unfencing(rsc, node, "Device definition changed", NULL, rsc->cluster); if (active_on_node) { crm_notice("Forcing restart of %s on %s " "because %s changed from '%s' to '%s'", rsc->id, pe__node_name(node), attr_list[i], pcmk__s(old_value, ""), pcmk__s(value, "")); } } } if (changed && active_on_node) { // Make sure the resource is restarted custom_action(rsc, stop_key(rsc), CRMD_ACTION_STOP, node, FALSE, TRUE, rsc->cluster); pe__set_resource_flags(rsc, pe_rsc_start_pending); } return changed; } /*! * \internal * \brief Add resource (and any matching children) to list if it matches ID * * \param[in] result List to add resource to * \param[in] rsc Resource to check * \param[in] id ID to match * * \return (Possibly new) head of list */ static GList * add_rsc_if_matching(GList *result, pe_resource_t *rsc, const char *id) { if ((strcmp(rsc->id, id) == 0) || ((rsc->clone_name != NULL) && (strcmp(rsc->clone_name, id) == 0))) { result = g_list_prepend(result, rsc); } for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child = (pe_resource_t *) iter->data; result = add_rsc_if_matching(result, child, id); } return result; } /*! * \internal * \brief Find all resources matching a given ID by either ID or clone name * * \param[in] id Resource ID to check * \param[in] data_set Cluster working set * * \return List of all resources that match \p id * \note The caller is responsible for freeing the return value with * g_list_free(). */ GList * pcmk__rscs_matching_id(const char *id, const pe_working_set_t *data_set) { GList *result = NULL; CRM_CHECK((id != NULL) && (data_set != NULL), return NULL); for (GList *iter = data_set->resources; iter != NULL; iter = iter->next) { result = add_rsc_if_matching(result, (pe_resource_t *) iter->data, id); } return result; } /*! * \internal - * \brief Set the variant-appropriate allocation methods for a resource + * \brief Set the variant-appropriate assignment methods for a resource * - * \param[in,out] rsc Resource to set allocation methods for + * \param[in,out] rsc Resource to set assignment methods for * \param[in] ignored Here so function can be used with g_list_foreach() */ static void -set_allocation_methods_for_rsc(pe_resource_t *rsc, void *ignored) +set_assignment_methods_for_rsc(pe_resource_t *rsc, void *ignored) { - rsc->cmds = &allocation_methods[rsc->variant]; - g_list_foreach(rsc->children, (GFunc) set_allocation_methods_for_rsc, NULL); + rsc->cmds = &assignment_methods[rsc->variant]; + g_list_foreach(rsc->children, (GFunc) set_assignment_methods_for_rsc, NULL); } /*! * \internal - * \brief Set the variant-appropriate allocation methods for all resources + * \brief Set the variant-appropriate assignment methods for all resources * * \param[in,out] data_set Cluster working set */ void -pcmk__set_allocation_methods(pe_working_set_t *data_set) +pcmk__set_assignment_methods(pe_working_set_t *data_set) { - g_list_foreach(data_set->resources, (GFunc) set_allocation_methods_for_rsc, + g_list_foreach(data_set->resources, (GFunc) set_assignment_methods_for_rsc, NULL); } // Shared implementation of resource_alloc_functions_t:colocated_resources() GList * pcmk__colocated_resources(const pe_resource_t *rsc, const pe_resource_t *orig_rsc, GList *colocated_rscs) { const GList *iter = NULL; GList *colocations = NULL; if (orig_rsc == NULL) { orig_rsc = rsc; } if ((rsc == NULL) || (g_list_find(colocated_rscs, rsc) != NULL)) { return colocated_rscs; } pe_rsc_trace(orig_rsc, "%s is in colocation chain with %s", rsc->id, orig_rsc->id); colocated_rscs = g_list_prepend(colocated_rscs, (gpointer) rsc); // Follow colocations where this resource is the dependent resource colocations = pcmk__this_with_colocations(rsc); for (iter = colocations; iter != NULL; iter = iter->next) { const pcmk__colocation_t *constraint = iter->data; const pe_resource_t *primary = constraint->primary; if (primary == orig_rsc) { continue; // Break colocation loop } if ((constraint->score == INFINITY) && (pcmk__colocation_affects(rsc, primary, constraint, true) == pcmk__coloc_affects_location)) { colocated_rscs = primary->cmds->colocated_resources(primary, orig_rsc, colocated_rscs); } } g_list_free(colocations); // Follow colocations where this resource is the primary resource colocations = pcmk__with_this_colocations(rsc); for (iter = colocations; iter != NULL; iter = iter->next) { const pcmk__colocation_t *constraint = iter->data; const pe_resource_t *dependent = constraint->dependent; if (dependent == orig_rsc) { continue; // Break colocation loop } if (pe_rsc_is_clone(rsc) && !pe_rsc_is_clone(dependent)) { continue; // We can't be sure whether dependent will be colocated } if ((constraint->score == INFINITY) && (pcmk__colocation_affects(dependent, rsc, constraint, true) == pcmk__coloc_affects_location)) { colocated_rscs = dependent->cmds->colocated_resources(dependent, orig_rsc, colocated_rscs); } } g_list_free(colocations); return colocated_rscs; } // No-op function for variants that don't need to implement add_graph_meta() void pcmk__noop_add_graph_meta(const pe_resource_t *rsc, xmlNode *xml) { } void pcmk__output_resource_actions(pe_resource_t *rsc) { pcmk__output_t *out = rsc->cluster->priv; pe_node_t *next = NULL; pe_node_t *current = NULL; if (rsc->children != NULL) { for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child = (pe_resource_t *) iter->data; child->cmds->output_actions(child); } return; } next = rsc->allocated_to; if (rsc->running_on) { current = pe__current_node(rsc); if (rsc->role == RSC_ROLE_STOPPED) { /* This can occur when resources are being recovered because * the current role can change in pcmk__primitive_create_actions() */ rsc->role = RSC_ROLE_STARTED; } } if ((current == NULL) && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { /* Don't log stopped orphans */ return; } out->message(out, "rsc-action", rsc, current, next); } /*! * \internal * \brief Assign a specified primitive resource to a node * * Assign a specified primitive resource to a specified node, if the node can * run the resource (or unconditionally, if \p force is true). Mark the resource * as no longer provisional. If the primitive can't be assigned (or \p chosen is * NULL), unassign any previous assignment for it, set its next role to stopped, * and update any existing actions scheduled for it. This is not done * recursively for children, so it should be called only for primitives. * * \param[in,out] rsc Resource to assign * \param[in,out] chosen Node to assign \p rsc to * \param[in] force If true, assign to \p chosen even if unavailable * * \return true if \p rsc could be assigned, otherwise false * * \note Assigning a resource to the NULL node using this function is different * from calling pcmk__unassign_resource(), in that it will also update any * actions created for the resource. */ bool pcmk__finalize_assignment(pe_resource_t *rsc, pe_node_t *chosen, bool force) { pcmk__output_t *out = rsc->cluster->priv; CRM_ASSERT(rsc->variant == pe_native); if (!force && (chosen != NULL)) { if ((chosen->weight < 0) // Allow the graph to assume that guest node connections will come up || (!pcmk__node_available(chosen, true, false) && !pe__is_guest_node(chosen))) { crm_debug("All nodes for resource %s are unavailable, unclean or " - "shutting down (%s can%s run resources, with weight %d)", + "shutting down (%s can%s run resources, with score %s)", rsc->id, pe__node_name(chosen), (pcmk__node_available(chosen, true, false)? "" : "not"), - chosen->weight); + pcmk_readable_score(chosen->weight)); pe__set_next_role(rsc, RSC_ROLE_STOPPED, "node availability"); chosen = NULL; } } pcmk__unassign_resource(rsc); pe__clear_resource_flags(rsc, pe_rsc_provisional); if (chosen == NULL) { - crm_debug("Could not allocate a node for %s", rsc->id); - pe__set_next_role(rsc, RSC_ROLE_STOPPED, "unable to allocate"); + crm_debug("Could not assign %s to a node", rsc->id); + pe__set_next_role(rsc, RSC_ROLE_STOPPED, "unable to assign"); for (GList *iter = rsc->actions; iter != NULL; iter = iter->next) { pe_action_t *op = (pe_action_t *) iter->data; - crm_debug("Updating %s for allocation failure", op->uuid); + pe_rsc_debug(rsc, "Updating %s for %s assignment failure", + op->uuid, rsc->id); if (pcmk__str_eq(op->task, RSC_STOP, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_optional); } else if (pcmk__str_eq(op->task, RSC_START, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_runnable); //pe__set_resource_flags(rsc, pe_rsc_block); } else { // Cancel recurring actions, unless for stopped state const char *interval_ms_s = NULL; const char *target_rc_s = NULL; char *rc_stopped = pcmk__itoa(PCMK_OCF_NOT_RUNNING); interval_ms_s = g_hash_table_lookup(op->meta, XML_LRM_ATTR_INTERVAL_MS); target_rc_s = g_hash_table_lookup(op->meta, XML_ATTR_TE_TARGET_RC); if ((interval_ms_s != NULL) && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_none) && !pcmk__str_eq(rc_stopped, target_rc_s, pcmk__str_none)) { pe__clear_action_flags(op, pe_action_runnable); } free(rc_stopped); } } return false; } crm_debug("Assigning %s to %s", rsc->id, pe__node_name(chosen)); rsc->allocated_to = pe__copy_node(chosen); chosen->details->allocated_rsc = g_list_prepend(chosen->details->allocated_rsc, rsc); chosen->details->num_resources++; chosen->count++; pcmk__consume_node_capacity(chosen->details->utilization, rsc); if (pcmk_is_set(rsc->cluster->flags, pe_flag_show_utilization)) { out->message(out, "resource-util", rsc, chosen, __func__); } return true; } /*! * \internal * \brief Assign a specified resource (of any variant) to a node * * Assign a specified resource and its children (if any) to a specified node, if * the node can run the resource (or unconditionally, if \p force is true). Mark * the resources as no longer provisional. If the resources can't be assigned * (or \p chosen is NULL), unassign any previous assignments, set next role to * stopped, and update any existing actions scheduled for them. * * \param[in,out] rsc Resource to assign * \param[in,out] chosen Node to assign \p rsc to * \param[in] force If true, assign to \p chosen even if unavailable * * \return true if \p rsc could be assigned, otherwise false * * \note Assigning a resource to the NULL node using this function is different * from calling pcmk__unassign_resource(), in that it will also update any * actions created for the resource. */ bool pcmk__assign_resource(pe_resource_t *rsc, pe_node_t *node, bool force) { bool changed = false; if (rsc->children == NULL) { if (rsc->allocated_to != NULL) { changed = true; } pcmk__finalize_assignment(rsc, node, force); } else { for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child_rsc = (pe_resource_t *) iter->data; changed |= pcmk__assign_resource(child_rsc, node, force); } } return changed; } /*! * \internal * \brief Remove any assignment of a specified resource to a node * * If a specified resource has been assigned to a node, remove that assignment * and mark the resource as provisional again. This is not done recursively for * children, so it should be called only for primitives. * * \param[in,out] rsc Resource to unassign */ void pcmk__unassign_resource(pe_resource_t *rsc) { pe_node_t *old = rsc->allocated_to; if (old == NULL) { return; } crm_info("Unassigning %s from %s", rsc->id, pe__node_name(old)); pe__set_resource_flags(rsc, pe_rsc_provisional); rsc->allocated_to = NULL; /* We're going to free the pe_node_t, but its details member is shared and * will remain, so update that appropriately first. */ old->details->allocated_rsc = g_list_remove(old->details->allocated_rsc, rsc); old->details->num_resources--; pcmk__release_node_capacity(old->details->utilization, rsc); free(old); } /*! * \internal * \brief Check whether a resource has reached its migration threshold on a node * * \param[in,out] rsc Resource to check * \param[in] node Node to check * \param[out] failed If threshold has been reached, this will be set to * resource that failed (possibly a parent of \p rsc) * * \return true if the migration threshold has been reached, false otherwise */ bool pcmk__threshold_reached(pe_resource_t *rsc, const pe_node_t *node, pe_resource_t **failed) { int fail_count, remaining_tries; pe_resource_t *rsc_to_ban = rsc; // Migration threshold of 0 means never force away if (rsc->migration_threshold == 0) { return false; } // If we're ignoring failures, also ignore the migration threshold if (pcmk_is_set(rsc->flags, pe_rsc_failure_ignored)) { return false; } // If there are no failures, there's no need to force away fail_count = pe_get_failcount(node, rsc, NULL, pe_fc_effective|pe_fc_fillers, NULL); if (fail_count <= 0) { return false; } // If failed resource is anonymous clone instance, we'll force clone away if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { rsc_to_ban = uber_parent(rsc); } // How many more times recovery will be tried on this node remaining_tries = rsc->migration_threshold - fail_count; if (remaining_tries <= 0) { crm_warn("%s cannot run on %s due to reaching migration threshold " "(clean up resource to allow again)" CRM_XS " failures=%d migration-threshold=%d", rsc_to_ban->id, pe__node_name(node), fail_count, rsc->migration_threshold); if (failed != NULL) { *failed = rsc_to_ban; } return true; } crm_info("%s can fail %d more time%s on " "%s before reaching migration threshold (%d)", rsc_to_ban->id, remaining_tries, pcmk__plural_s(remaining_tries), pe__node_name(node), rsc->migration_threshold); return false; } static void * convert_const_pointer(const void *ptr) { /* Worst function ever */ return (void *)ptr; } /*! * \internal - * \brief Get a node's weight + * \brief Get a node's score * - * \param[in] node Unweighted node to check (for node ID) - * \param[in] nodes List of weighted nodes to look for \p node in + * \param[in] node Node with ID to check + * \param[in] nodes List of nodes to look for \p node score in * - * \return Node's weight, or -INFINITY if not found + * \return Node's score, or -INFINITY if not found */ static int -get_node_weight(const pe_node_t *node, GHashTable *nodes) +get_node_score(const pe_node_t *node, GHashTable *nodes) { - pe_node_t *weighted_node = NULL; + pe_node_t *found_node = NULL; if ((node != NULL) && (nodes != NULL)) { - weighted_node = g_hash_table_lookup(nodes, node->details->id); + found_node = g_hash_table_lookup(nodes, node->details->id); } - return (weighted_node == NULL)? -INFINITY : weighted_node->weight; + return (found_node == NULL)? -INFINITY : found_node->weight; } /*! * \internal - * \brief Compare two resources according to which should be allocated first + * \brief Compare two resources according to which should be assigned first * * \param[in] a First resource to compare * \param[in] b Second resource to compare * \param[in] data Sorted list of all nodes in cluster * - * \return -1 if \p a should be allocated before \b, 0 if they are equal, - * or +1 if \p a should be allocated after \b + * \return -1 if \p a should be assigned before \b, 0 if they are equal, + * or +1 if \p a should be assigned after \b */ static gint cmp_resources(gconstpointer a, gconstpointer b, gpointer data) { const pe_resource_t *resource1 = a; const pe_resource_t *resource2 = b; const GList *nodes = (const GList *) data; int rc = 0; - int r1_weight = -INFINITY; - int r2_weight = -INFINITY; + int r1_score = -INFINITY; + int r2_score = -INFINITY; pe_node_t *r1_node = NULL; pe_node_t *r2_node = NULL; GHashTable *r1_nodes = NULL; GHashTable *r2_nodes = NULL; const char *reason = NULL; - // Resources with highest priority should be allocated first + // Resources with highest priority should be assigned first reason = "priority"; - r1_weight = resource1->priority; - r2_weight = resource2->priority; - if (r1_weight > r2_weight) { + r1_score = resource1->priority; + r2_score = resource2->priority; + if (r1_score > r2_score) { rc = -1; goto done; } - if (r1_weight < r2_weight) { + if (r1_score < r2_score) { rc = 1; goto done; } // We need nodes to make any other useful comparisons reason = "no node list"; if (nodes == NULL) { goto done; } - // Calculate and log node weights + // Calculate and log node scores resource1->cmds->add_colocated_node_scores(convert_const_pointer(resource1), resource1->id, &r1_nodes, NULL, 1, pcmk__coloc_select_this_with); resource2->cmds->add_colocated_node_scores(convert_const_pointer(resource2), resource2->id, &r2_nodes, NULL, 1, pcmk__coloc_select_this_with); - pe__show_node_weights(true, NULL, resource1->id, r1_nodes, - resource1->cluster); - pe__show_node_weights(true, NULL, resource2->id, r2_nodes, - resource2->cluster); + pe__show_node_scores(true, NULL, resource1->id, r1_nodes, + resource1->cluster); + pe__show_node_scores(true, NULL, resource2->id, r2_nodes, + resource2->cluster); // The resource with highest score on its current node goes first reason = "current location"; if (resource1->running_on != NULL) { r1_node = pe__current_node(resource1); } if (resource2->running_on != NULL) { r2_node = pe__current_node(resource2); } - r1_weight = get_node_weight(r1_node, r1_nodes); - r2_weight = get_node_weight(r2_node, r2_nodes); - if (r1_weight > r2_weight) { + r1_score = get_node_score(r1_node, r1_nodes); + r2_score = get_node_score(r2_node, r2_nodes); + if (r1_score > r2_score) { rc = -1; goto done; } - if (r1_weight < r2_weight) { + if (r1_score < r2_score) { rc = 1; goto done; } - // Otherwise a higher weight on any node will do + // Otherwise a higher score on any node will do reason = "score"; for (const GList *iter = nodes; iter != NULL; iter = iter->next) { const pe_node_t *node = (const pe_node_t *) iter->data; - r1_weight = get_node_weight(node, r1_nodes); - r2_weight = get_node_weight(node, r2_nodes); - if (r1_weight > r2_weight) { + r1_score = get_node_score(node, r1_nodes); + r2_score = get_node_score(node, r2_nodes); + if (r1_score > r2_score) { rc = -1; goto done; } - if (r1_weight < r2_weight) { + if (r1_score < r2_score) { rc = 1; goto done; } } done: crm_trace("%s (%d)%s%s %c %s (%d)%s%s: %s", - resource1->id, r1_weight, + resource1->id, r1_score, ((r1_node == NULL)? "" : " on "), ((r1_node == NULL)? "" : r1_node->details->id), ((rc < 0)? '>' : ((rc > 0)? '<' : '=')), - resource2->id, r2_weight, + resource2->id, r2_score, ((r2_node == NULL)? "" : " on "), ((r2_node == NULL)? "" : r2_node->details->id), reason); if (r1_nodes != NULL) { g_hash_table_destroy(r1_nodes); } if (r2_nodes != NULL) { g_hash_table_destroy(r2_nodes); } return rc; } /*! * \internal - * \brief Sort resources in the order they should be allocated to nodes + * \brief Sort resources in the order they should be assigned to nodes * * \param[in,out] data_set Cluster working set */ void pcmk__sort_resources(pe_working_set_t *data_set) { GList *nodes = g_list_copy(data_set->nodes); nodes = pcmk__sort_nodes(nodes, NULL); data_set->resources = g_list_sort_with_data(data_set->resources, cmp_resources, nodes); g_list_free(nodes); } diff --git a/lib/pacemaker/pcmk_sched_utilization.c b/lib/pacemaker/pcmk_sched_utilization.c index 0a4bec373b..7df0ad821e 100644 --- a/lib/pacemaker/pcmk_sched_utilization.c +++ b/lib/pacemaker/pcmk_sched_utilization.c @@ -1,469 +1,469 @@ /* * Copyright 2014-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include "libpacemaker_private.h" // Name for a pseudo-op to use in ordering constraints for utilization #define LOAD_STOPPED "load_stopped" /*! * \internal * \brief Get integer utilization from a string * * \param[in] s String representation of a node utilization value * * \return Integer equivalent of \p s * \todo It would make sense to restrict utilization values to nonnegative * integers, but the documentation just says "integers" and we didn't * restrict them initially, so for backward compatibility, allow any * integer. */ static int utilization_value(const char *s) { int value = 0; if ((s != NULL) && (pcmk__scan_min_int(s, &value, INT_MIN) == EINVAL)) { pe_warn("Using 0 for utilization instead of invalid value '%s'", value); value = 0; } return value; } /* * Functions for comparing node capacities */ struct compare_data { const pe_node_t *node1; const pe_node_t *node2; bool node2_only; int result; }; /*! * \internal * \brief Compare a single utilization attribute for two nodes * * Compare one utilization attribute for two nodes, incrementing the result if * the first node has greater capacity, and decrementing it if the second node * has greater capacity. * * \param[in] key Utilization attribute name to compare * \param[in] value Utilization attribute value to compare * \param[in,out] user_data Comparison data (as struct compare_data*) */ static void compare_utilization_value(gpointer key, gpointer value, gpointer user_data) { int node1_capacity = 0; int node2_capacity = 0; struct compare_data *data = user_data; const char *node2_value = NULL; if (data->node2_only) { if (g_hash_table_lookup(data->node1->details->utilization, key)) { return; // We've already compared this attribute } } else { node1_capacity = utilization_value((const char *) value); } node2_value = g_hash_table_lookup(data->node2->details->utilization, key); node2_capacity = utilization_value(node2_value); if (node1_capacity > node2_capacity) { data->result--; } else if (node1_capacity < node2_capacity) { data->result++; } } /*! * \internal * \brief Compare utilization capacities of two nodes * * \param[in] node1 First node to compare * \param[in] node2 Second node to compare * * \return Negative integer if node1 has more free capacity, * 0 if the capacities are equal, or a positive integer * if node2 has more free capacity */ int pcmk__compare_node_capacities(const pe_node_t *node1, const pe_node_t *node2) { struct compare_data data = { .node1 = node1, .node2 = node2, .node2_only = false, .result = 0, }; // Compare utilization values that node1 and maybe node2 have g_hash_table_foreach(node1->details->utilization, compare_utilization_value, &data); // Compare utilization values that only node2 has data.node2_only = true; g_hash_table_foreach(node2->details->utilization, compare_utilization_value, &data); return data.result; } /* * Functions for updating node capacities */ struct calculate_data { GHashTable *current_utilization; bool plus; }; /*! * \internal * \brief Update a single utilization attribute with a new value * * \param[in] key Name of utilization attribute to update * \param[in] value Value to add or substract * \param[in,out] user_data Calculation data (as struct calculate_data *) */ static void update_utilization_value(gpointer key, gpointer value, gpointer user_data) { int result = 0; const char *current = NULL; struct calculate_data *data = user_data; current = g_hash_table_lookup(data->current_utilization, key); if (data->plus) { result = utilization_value(current) + utilization_value(value); } else if (current) { result = utilization_value(current) - utilization_value(value); } g_hash_table_replace(data->current_utilization, strdup(key), pcmk__itoa(result)); } /*! * \internal * \brief Subtract a resource's utilization from node capacity * * \param[in,out] current_utilization Current node utilization attributes * \param[in] rsc Resource with utilization to subtract */ void pcmk__consume_node_capacity(GHashTable *current_utilization, const pe_resource_t *rsc) { struct calculate_data data = { .current_utilization = current_utilization, .plus = false, }; g_hash_table_foreach(rsc->utilization, update_utilization_value, &data); } /*! * \internal * \brief Add a resource's utilization to node capacity * * \param[in,out] current_utilization Current node utilization attributes * \param[in] rsc Resource with utilization to add */ void pcmk__release_node_capacity(GHashTable *current_utilization, const pe_resource_t *rsc) { struct calculate_data data = { .current_utilization = current_utilization, .plus = true, }; g_hash_table_foreach(rsc->utilization, update_utilization_value, &data); } /* * Functions for checking for sufficient node capacity */ struct capacity_data { const pe_node_t *node; const char *rsc_id; bool is_enough; }; /*! * \internal * \brief Check whether a single utilization attribute has sufficient capacity * * \param[in] key Name of utilization attribute to check * \param[in] value Amount of utilization required * \param[in,out] user_data Capacity data (as struct capacity_data *) */ static void check_capacity(gpointer key, gpointer value, gpointer user_data) { int required = 0; int remaining = 0; const char *node_value_s = NULL; struct capacity_data *data = user_data; node_value_s = g_hash_table_lookup(data->node->details->utilization, key); required = utilization_value(value); remaining = utilization_value(node_value_s); if (required > remaining) { crm_debug("Remaining capacity for %s on %s (%d) is insufficient " "for resource %s usage (%d)", (const char *) key, pe__node_name(data->node), remaining, data->rsc_id, required); data->is_enough = false; } } /*! * \internal * \brief Check whether a node has sufficient capacity for a resource * * \param[in] node Node to check * \param[in] rsc_id ID of resource to check (for debug logs only) * \param[in] utilization Required utilization amounts * * \return true if node has sufficient capacity for resource, otherwise false */ static bool have_enough_capacity(const pe_node_t *node, const char *rsc_id, GHashTable *utilization) { struct capacity_data data = { .node = node, .rsc_id = rsc_id, .is_enough = true, }; g_hash_table_foreach(utilization, check_capacity, &data); return data.is_enough; } /*! * \internal * \brief Sum the utilization requirements of a list of resources * - * \param[in] orig_rsc Resource being allocated (for logging purposes) + * \param[in] orig_rsc Resource being assigned (for logging purposes) * \param[in] rscs Resources whose utilization should be summed * * \return Newly allocated hash table with sum of all utilization values * \note It is the caller's responsibility to free the return value using * g_hash_table_destroy(). */ static GHashTable * sum_resource_utilization(const pe_resource_t *orig_rsc, GList *rscs) { GHashTable *utilization = pcmk__strkey_table(free, free); for (GList *iter = rscs; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; rsc->cmds->add_utilization(rsc, orig_rsc, rscs, utilization); } return utilization; } /*! * \internal * \brief Ban resource from nodes with insufficient utilization capacity * * \param[in,out] rsc Resource to check * * \return Allowed node for \p rsc with most spare capacity, if there are no * nodes with enough capacity for \p rsc and all its colocated resources */ const pe_node_t * pcmk__ban_insufficient_capacity(pe_resource_t *rsc) { bool any_capable = false; char *rscs_id = NULL; pe_node_t *node = NULL; const pe_node_t *most_capable_node = NULL; GList *colocated_rscs = NULL; - GHashTable *unallocated_utilization = NULL; + GHashTable *unassigned_utilization = NULL; GHashTableIter iter; CRM_CHECK(rsc != NULL, return NULL); // The default placement strategy ignores utilization if (pcmk__str_eq(rsc->cluster->placement_strategy, "default", pcmk__str_casei)) { return NULL; } // Check whether any resources are colocated with this one colocated_rscs = rsc->cmds->colocated_resources(rsc, NULL, NULL); if (colocated_rscs == NULL) { return NULL; } rscs_id = crm_strdup_printf("%s and its colocated resources", rsc->id); // If rsc isn't in the list, add it so we include its utilization if (g_list_find(colocated_rscs, rsc) == NULL) { colocated_rscs = g_list_append(colocated_rscs, rsc); } - // Sum utilization of colocated resources that haven't been allocated yet - unallocated_utilization = sum_resource_utilization(rsc, colocated_rscs); + // Sum utilization of colocated resources that haven't been assigned yet + unassigned_utilization = sum_resource_utilization(rsc, colocated_rscs); // Check whether any node has enough capacity for all the resources g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (!pcmk__node_available(node, true, false)) { continue; } - if (have_enough_capacity(node, rscs_id, unallocated_utilization)) { + if (have_enough_capacity(node, rscs_id, unassigned_utilization)) { any_capable = true; } // Keep track of node with most free capacity if ((most_capable_node == NULL) || (pcmk__compare_node_capacities(node, most_capable_node) < 0)) { most_capable_node = node; } } if (any_capable) { // If so, ban resource from any node with insufficient capacity g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (pcmk__node_available(node, true, false) && !have_enough_capacity(node, rscs_id, - unallocated_utilization)) { + unassigned_utilization)) { pe_rsc_debug(rsc, "%s does not have enough capacity for %s", pe__node_name(node), rscs_id); resource_location(rsc, node, -INFINITY, "__limit_utilization__", rsc->cluster); } } most_capable_node = NULL; } else { // Otherwise, ban from nodes with insufficient capacity for rsc alone g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (pcmk__node_available(node, true, false) && !have_enough_capacity(node, rsc->id, rsc->utilization)) { pe_rsc_debug(rsc, "%s does not have enough capacity for %s", pe__node_name(node), rsc->id); resource_location(rsc, node, -INFINITY, "__limit_utilization__", rsc->cluster); } } } - g_hash_table_destroy(unallocated_utilization); + g_hash_table_destroy(unassigned_utilization); g_list_free(colocated_rscs); free(rscs_id); - pe__show_node_weights(true, rsc, "Post-utilization", - rsc->allowed_nodes, rsc->cluster); + pe__show_node_scores(true, rsc, "Post-utilization", rsc->allowed_nodes, + rsc->cluster); return most_capable_node; } /*! * \internal * \brief Create a new load_stopped pseudo-op for a node * * \param[in] node Node to create op for * \param[in,out] data_set Cluster working set * * \return Newly created load_stopped op */ static pe_action_t * new_load_stopped_op(const pe_node_t *node, pe_working_set_t *data_set) { char *load_stopped_task = crm_strdup_printf(LOAD_STOPPED "_%s", node->details->uname); pe_action_t *load_stopped = get_pseudo_op(load_stopped_task, data_set); if (load_stopped->node == NULL) { load_stopped->node = pe__copy_node(node); pe__clear_action_flags(load_stopped, pe_action_optional); } free(load_stopped_task); return load_stopped; } /*! * \internal * \brief Create utilization-related internal constraints for a resource * * \param[in,out] rsc Resource to create constraints for * \param[in] allowed_nodes List of allowed next nodes for \p rsc */ void pcmk__create_utilization_constraints(pe_resource_t *rsc, const GList *allowed_nodes) { const GList *iter = NULL; const pe_node_t *node = NULL; pe_action_t *load_stopped = NULL; pe_rsc_trace(rsc, "Creating utilization constraints for %s - strategy: %s", rsc->id, rsc->cluster->placement_strategy); // "stop rsc then load_stopped" constraints for current nodes for (iter = rsc->running_on; iter != NULL; iter = iter->next) { node = (const pe_node_t *) iter->data; load_stopped = new_load_stopped_op(node, rsc->cluster); pcmk__new_ordering(rsc, stop_key(rsc), NULL, NULL, NULL, load_stopped, pe_order_load, rsc->cluster); } // "load_stopped then start/migrate_to rsc" constraints for allowed nodes for (iter = allowed_nodes; iter; iter = iter->next) { node = (const pe_node_t *) iter->data; load_stopped = new_load_stopped_op(node, rsc->cluster); pcmk__new_ordering(NULL, NULL, load_stopped, rsc, start_key(rsc), NULL, pe_order_load, rsc->cluster); pcmk__new_ordering(NULL, NULL, load_stopped, rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0), NULL, pe_order_load, rsc->cluster); } } /*! * \internal * \brief Output node capacities if enabled * * \param[in] desc Prefix for output * \param[in,out] data_set Cluster working set */ void pcmk__show_node_capacities(const char *desc, pe_working_set_t *data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_show_utilization)) { return; } for (const GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { const pe_node_t *node = (const pe_node_t *) iter->data; pcmk__output_t *out = data_set->priv; out->message(out, "node-capacity", node, desc); } } diff --git a/lib/pacemaker/pcmk_scheduler.c b/lib/pacemaker/pcmk_scheduler.c index b4e670d865..edd21180f6 100644 --- a/lib/pacemaker/pcmk_scheduler.c +++ b/lib/pacemaker/pcmk_scheduler.c @@ -1,811 +1,811 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include "libpacemaker_private.h" CRM_TRACE_INIT_DATA(pacemaker); /*! * \internal - * \brief Do deferred action checks after allocation + * \brief Do deferred action checks after assignment * * When unpacking the resource history, the scheduler checks for resource * configurations that have changed since an action was run. However, at that * time, bundles using the REMOTE_CONTAINER_HACK don't have their final * parameter information, so instead they add a deferred check to a list. This * function processes one entry in that list. * * \param[in,out] rsc Resource that action history is for * \param[in,out] node Node that action history is for * \param[in] rsc_op Action history entry * \param[in] check Type of deferred check to do */ static void check_params(pe_resource_t *rsc, pe_node_t *node, const xmlNode *rsc_op, enum pe_check_parameters check) { const char *reason = NULL; op_digest_cache_t *digest_data = NULL; switch (check) { case pe_check_active: if (pcmk__check_action_config(rsc, node, rsc_op) && pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL)) { reason = "action definition changed"; } break; case pe_check_last_failure: digest_data = rsc_action_digest_cmp(rsc, rsc_op, node, rsc->cluster); switch (digest_data->rc) { case RSC_DIGEST_UNKNOWN: crm_trace("Resource %s history entry %s on %s has " "no digest to compare", rsc->id, ID(rsc_op), node->details->id); break; case RSC_DIGEST_MATCH: break; default: reason = "resource parameters have changed"; break; } break; } if (reason != NULL) { pe__clear_failcount(rsc, node, reason, rsc->cluster); } } /*! * \internal * \brief Check whether a resource has failcount clearing scheduled on a node * * \param[in] node Node to check * \param[in] rsc Resource to check * * \return true if \p rsc has failcount clearing scheduled on \p node, * otherwise false */ static bool failcount_clear_action_exists(const pe_node_t *node, const pe_resource_t *rsc) { GList *list = pe__resource_actions(rsc, node, CRM_OP_CLEAR_FAILCOUNT, TRUE); if (list != NULL) { g_list_free(list); return true; } return false; } /*! * \internal * \brief Ban a resource from a node if it reached its failure threshold there * * \param[in,out] rsc Resource to check failure threshold for * \param[in] node Node to check \p rsc on */ static void check_failure_threshold(pe_resource_t *rsc, const pe_node_t *node) { // If this is a collective resource, apply recursively to children instead if (rsc->children != NULL) { g_list_foreach(rsc->children, (GFunc) check_failure_threshold, (gpointer) node); return; } else if (failcount_clear_action_exists(node, rsc)) { /* Don't force the resource away from this node due to a failcount * that's going to be cleared. * * @TODO Failcount clearing can be scheduled in * pcmk__handle_rsc_config_changes() via process_rsc_history(), or in * schedule_resource_actions() via check_params(). This runs well before * then, so it cannot detect those, meaning we might check the migration * threshold when we shouldn't. Worst case, we stop or move the * resource, then move it back in the next transition. */ return; } else { pe_resource_t *failed = NULL; if (pcmk__threshold_reached(rsc, node, &failed)) { resource_location(failed, node, -INFINITY, "__fail_limit__", rsc->cluster); } } } /*! * \internal * \brief If resource has exclusive discovery, ban node if not allowed * * Location constraints have a resource-discovery option that allows users to * specify where probes are done for the affected resource. If this is set to * exclusive, probes will only be done on nodes listed in exclusive constraints. * This function bans the resource from the node if the node is not listed. * * \param[in,out] rsc Resource to check * \param[in] node Node to check \p rsc on */ static void apply_exclusive_discovery(pe_resource_t *rsc, const pe_node_t *node) { if (rsc->exclusive_discover || pe__const_top_resource(rsc, false)->exclusive_discover) { pe_node_t *match = NULL; // If this is a collective resource, apply recursively to children g_list_foreach(rsc->children, (GFunc) apply_exclusive_discovery, (gpointer) node); match = g_hash_table_lookup(rsc->allowed_nodes, node->details->id); if ((match != NULL) && (match->rsc_discover_mode != pe_discover_exclusive)) { match->weight = -INFINITY; } } } /*! * \internal * \brief Apply stickiness to a resource if appropriate * * \param[in,out] rsc Resource to check for stickiness * \param[in,out] data_set Cluster working set */ static void apply_stickiness(pe_resource_t *rsc, pe_working_set_t *data_set) { pe_node_t *node = NULL; // If this is a collective resource, apply recursively to children instead if (rsc->children != NULL) { g_list_foreach(rsc->children, (GFunc) apply_stickiness, data_set); return; } /* A resource is sticky if it is managed, has stickiness configured, and is * active on a single node. */ if (!pcmk_is_set(rsc->flags, pe_rsc_managed) || (rsc->stickiness < 1) || !pcmk__list_of_1(rsc->running_on)) { return; } node = rsc->running_on->data; /* In a symmetric cluster, stickiness can always be used. In an * asymmetric cluster, we have to check whether the resource is still * allowed on the node, so we don't keep the resource somewhere it is no * longer explicitly enabled. */ if (!pcmk_is_set(rsc->cluster->flags, pe_flag_symmetric_cluster) && (pe_hash_table_lookup(rsc->allowed_nodes, node->details->id) == NULL)) { pe_rsc_debug(rsc, "Ignoring %s stickiness because the cluster is " "asymmetric and %s is not explicitly allowed", rsc->id, pe__node_name(node)); return; } pe_rsc_debug(rsc, "Resource %s has %d stickiness on %s", rsc->id, rsc->stickiness, pe__node_name(node)); resource_location(rsc, node, rsc->stickiness, "stickiness", data_set); } /*! * \internal * \brief Apply shutdown locks for all resources as appropriate * * \param[in,out] data_set Cluster working set */ static void apply_shutdown_locks(pe_working_set_t *data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { return; } for (GList *iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; rsc->cmds->shutdown_lock(rsc); } } /*! * \internal * \brief Calculate the number of available nodes in the cluster * * \param[in,out] data_set Cluster working set */ static void count_available_nodes(pe_working_set_t *data_set) { if (pcmk_is_set(data_set->flags, pe_flag_no_compat)) { return; } // @COMPAT for API backward compatibility only (cluster does not use value) for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; if ((node != NULL) && (node->weight >= 0) && node->details->online && (node->details->type != node_ping)) { data_set->max_valid_nodes++; } } crm_trace("Online node count: %d", data_set->max_valid_nodes); } /* * \internal * \brief Apply node-specific scheduling criteria * * After the CIB has been unpacked, process node-specific scheduling criteria * including shutdown locks, location constraints, resource stickiness, * migration thresholds, and exclusive resource discovery. */ static void apply_node_criteria(pe_working_set_t *data_set) { crm_trace("Applying node-specific scheduling criteria"); apply_shutdown_locks(data_set); count_available_nodes(data_set); pcmk__apply_locations(data_set); g_list_foreach(data_set->resources, (GFunc) apply_stickiness, data_set); for (GList *node_iter = data_set->nodes; node_iter != NULL; node_iter = node_iter->next) { for (GList *rsc_iter = data_set->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { pe_node_t *node = (pe_node_t *) node_iter->data; pe_resource_t *rsc = (pe_resource_t *) rsc_iter->data; check_failure_threshold(rsc, node); apply_exclusive_discovery(rsc, node); } } } /*! * \internal - * \brief Allocate resources to nodes + * \brief Assign resources to nodes * * \param[in,out] data_set Cluster working set */ static void -allocate_resources(pe_working_set_t *data_set) +assign_resources(pe_working_set_t *data_set) { GList *iter = NULL; - crm_trace("Allocating resources to nodes"); + crm_trace("Assigning resources to nodes"); if (!pcmk__str_eq(data_set->placement_strategy, "default", pcmk__str_casei)) { pcmk__sort_resources(data_set); } pcmk__show_node_capacities("Original", data_set); if (pcmk_is_set(data_set->flags, pe_flag_have_remote_nodes)) { - /* Allocate remote connection resources first (which will also allocate - * any colocation dependencies). If the connection is migrating, always + /* Assign remote connection resources first (which will also assign any + * colocation dependencies). If the connection is migrating, always * prefer the partial migration target. */ for (iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; if (rsc->is_remote_node) { - pe_rsc_trace(rsc, "Allocating remote connection resource '%s'", + pe_rsc_trace(rsc, "Assigning remote connection resource '%s'", rsc->id); rsc->cmds->assign(rsc, rsc->partial_migration_target); } } } /* now do the rest of the resources */ for (iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; if (!rsc->is_remote_node) { - pe_rsc_trace(rsc, "Allocating %s resource '%s'", + pe_rsc_trace(rsc, "Assigning %s resource '%s'", crm_element_name(rsc->xml), rsc->id); rsc->cmds->assign(rsc, NULL); } } pcmk__show_node_capacities("Remaining", data_set); } /*! * \internal * \brief Schedule fail count clearing on online nodes if resource is orphaned * * \param[in,out] rsc Resource to check * \param[in,out] data_set Cluster working set */ static void clear_failcounts_if_orphaned(pe_resource_t *rsc, pe_working_set_t *data_set) { if (!pcmk_is_set(rsc->flags, pe_rsc_orphan)) { return; } crm_trace("Clear fail counts for orphaned resource %s", rsc->id); /* There's no need to recurse into rsc->children because those - * should just be unallocated clone instances. + * should just be unassigned clone instances. */ for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; pe_action_t *clear_op = NULL; if (!node->details->online) { continue; } if (pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL) == 0) { continue; } clear_op = pe__clear_failcount(rsc, node, "it is orphaned", data_set); /* We can't use order_action_then_stop() here because its * pe_order_preserve breaks things */ pcmk__new_ordering(clear_op->rsc, NULL, clear_op, rsc, stop_key(rsc), NULL, pe_order_optional, data_set); } } /*! * \internal * \brief Schedule any resource actions needed * * \param[in,out] data_set Cluster working set */ static void schedule_resource_actions(pe_working_set_t *data_set) { // Process deferred action checks pe__foreach_param_check(data_set, check_params); pe__free_param_checks(data_set); if (pcmk_is_set(data_set->flags, pe_flag_startup_probes)) { crm_trace("Scheduling probes"); pcmk__schedule_probes(data_set); } if (pcmk_is_set(data_set->flags, pe_flag_stop_rsc_orphans)) { g_list_foreach(data_set->resources, (GFunc) clear_failcounts_if_orphaned, data_set); } crm_trace("Scheduling resource actions"); for (GList *iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; rsc->cmds->create_actions(rsc); } } /*! * \internal * \brief Check whether a resource or any of its descendants are managed * * \param[in] rsc Resource to check * * \return true if resource or any descendant is managed, otherwise false */ static bool is_managed(const pe_resource_t *rsc) { if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { return true; } for (GList *iter = rsc->children; iter != NULL; iter = iter->next) { if (is_managed((pe_resource_t *) iter->data)) { return true; } } return false; } /*! * \internal * \brief Check whether any resources in the cluster are managed * * \param[in] data_set Cluster working set * * \return true if any resource is managed, otherwise false */ static bool any_managed_resources(const pe_working_set_t *data_set) { for (const GList *iter = data_set->resources; iter != NULL; iter = iter->next) { if (is_managed((const pe_resource_t *) iter->data)) { return true; } } return false; } /*! * \internal * \brief Check whether a node requires fencing * * \param[in] node Node to check * \param[in] have_managed Whether any resource in cluster is managed * \param[in] data_set Cluster working set * * \return true if \p node should be fenced, otherwise false */ static bool needs_fencing(const pe_node_t *node, bool have_managed, const pe_working_set_t *data_set) { return have_managed && node->details->unclean && pe_can_fence(data_set, node); } /*! * \internal * \brief Check whether a node requires shutdown * * \param[in] node Node to check * * \return true if \p node should be shut down, otherwise false */ static bool needs_shutdown(const pe_node_t *node) { if (pe__is_guest_or_remote_node(node)) { /* Do not send shutdown actions for Pacemaker Remote nodes. * @TODO We might come up with a good use for this in the future. */ return false; } return node->details->online && node->details->shutdown; } /*! * \internal * \brief Track and order non-DC fencing * * \param[in,out] list List of existing non-DC fencing actions * \param[in,out] action Fencing action to prepend to \p list * \param[in] data_set Cluster working set * * \return (Possibly new) head of \p list */ static GList * add_nondc_fencing(GList *list, pe_action_t *action, const pe_working_set_t *data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_concurrent_fencing) && (list != NULL)) { /* Concurrent fencing is disabled, so order each non-DC * fencing in a chain. If there is any DC fencing or * shutdown, it will be ordered after the last action in the * chain later. */ order_actions((pe_action_t *) list->data, action, pe_order_optional); } return g_list_prepend(list, action); } /*! * \internal * \brief Schedule a node for fencing * * \param[in,out] node Node that requires fencing * \param[in,out] data_set Cluster working set */ static pe_action_t * schedule_fencing(pe_node_t *node, pe_working_set_t *data_set) { pe_action_t *fencing = pe_fence_op(node, NULL, FALSE, "node is unclean", FALSE, data_set); pe_warn("Scheduling node %s for fencing", pe__node_name(node)); pcmk__order_vs_fence(fencing, data_set); return fencing; } /*! * \internal * \brief Create and order node fencing and shutdown actions * * \param[in,out] data_set Cluster working set */ static void schedule_fencing_and_shutdowns(pe_working_set_t *data_set) { pe_action_t *dc_down = NULL; bool integrity_lost = false; bool have_managed = any_managed_resources(data_set); GList *fencing_ops = NULL; GList *shutdown_ops = NULL; crm_trace("Scheduling fencing and shutdowns as needed"); if (!have_managed) { crm_notice("No fencing will be done until there are resources to manage"); } // Check each node for whether it needs fencing or shutdown for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; pe_action_t *fencing = NULL; /* Guest nodes are "fenced" by recovering their container resource, * so handle them separately. */ if (pe__is_guest_node(node)) { if (node->details->remote_requires_reset && have_managed && pe_can_fence(data_set, node)) { pcmk__fence_guest(node); } continue; } if (needs_fencing(node, have_managed, data_set)) { fencing = schedule_fencing(node, data_set); // Track DC and non-DC fence actions separately if (node->details->is_dc) { dc_down = fencing; } else { fencing_ops = add_nondc_fencing(fencing_ops, fencing, data_set); } } else if (needs_shutdown(node)) { pe_action_t *down_op = pcmk__new_shutdown_action(node); // Track DC and non-DC shutdown actions separately if (node->details->is_dc) { dc_down = down_op; } else { shutdown_ops = g_list_prepend(shutdown_ops, down_op); } } if ((fencing == NULL) && node->details->unclean) { integrity_lost = true; pe_warn("Node %s is unclean but cannot be fenced", pe__node_name(node)); } } if (integrity_lost) { if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { pe_warn("Resource functionality and data integrity cannot be " "guaranteed (configure, enable, and test fencing to " "correct this)"); } else if (!pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { crm_notice("Unclean nodes will not be fenced until quorum is " "attained or no-quorum-policy is set to ignore"); } } if (dc_down != NULL) { /* Order any non-DC shutdowns before any DC shutdown, to avoid repeated * DC elections. However, we don't want to order non-DC shutdowns before * a DC *fencing*, because even though we don't want a node that's * shutting down to become DC, the DC fencing could be ordered before a * clone stop that's also ordered before the shutdowns, thus leading to * a graph loop. */ if (pcmk__str_eq(dc_down->task, CRM_OP_SHUTDOWN, pcmk__str_none)) { pcmk__order_after_each(dc_down, shutdown_ops); } // Order any non-DC fencing before any DC fencing or shutdown if (pcmk_is_set(data_set->flags, pe_flag_concurrent_fencing)) { /* With concurrent fencing, order each non-DC fencing action * separately before any DC fencing or shutdown. */ pcmk__order_after_each(dc_down, fencing_ops); } else if (fencing_ops != NULL) { /* Without concurrent fencing, the non-DC fencing actions are * already ordered relative to each other, so we just need to order * the DC fencing after the last action in the chain (which is the * first item in the list). */ order_actions((pe_action_t *) fencing_ops->data, dc_down, pe_order_optional); } } g_list_free(fencing_ops); g_list_free(shutdown_ops); } static void log_resource_details(pe_working_set_t *data_set) { pcmk__output_t *out = data_set->priv; GList *all = NULL; /* We need a list of nodes that we are allowed to output information for. * This is necessary because out->message for all the resource-related * messages expects such a list, due to the `crm_mon --node=` feature. Here, * we just make it a list of all the nodes. */ all = g_list_prepend(all, (gpointer) "*"); for (GList *item = data_set->resources; item != NULL; item = item->next) { pe_resource_t *rsc = (pe_resource_t *) item->data; // Log all resources except inactive orphans if (!pcmk_is_set(rsc->flags, pe_rsc_orphan) || (rsc->role != RSC_ROLE_STOPPED)) { out->message(out, crm_map_element_name(rsc->xml), 0, rsc, all, all); } } g_list_free(all); } static void log_all_actions(pe_working_set_t *data_set) { /* This only ever outputs to the log, so ignore whatever output object was * previously set and just log instead. */ pcmk__output_t *prev_out = data_set->priv; pcmk__output_t *out = NULL; if (pcmk__log_output_new(&out) != pcmk_rc_ok) { return; } pe__register_messages(out); pcmk__register_lib_messages(out); pcmk__output_set_log_level(out, LOG_NOTICE); data_set->priv = out; out->begin_list(out, NULL, NULL, "Actions"); pcmk__output_actions(data_set); out->end_list(out); out->finish(out, CRM_EX_OK, true, NULL); pcmk__output_free(out); data_set->priv = prev_out; } /*! * \internal * \brief Log all required but unrunnable actions at trace level * * \param[in] data_set Cluster working set */ static void log_unrunnable_actions(const pe_working_set_t *data_set) { const uint64_t flags = pe_action_optional|pe_action_runnable|pe_action_pseudo; crm_trace("Required but unrunnable actions:"); for (const GList *iter = data_set->actions; iter != NULL; iter = iter->next) { const pe_action_t *action = (const pe_action_t *) iter->data; if (!pcmk_any_flags_set(action->flags, flags)) { pcmk__log_action("\t", action, true); } } } /*! * \internal * \brief Unpack the CIB for scheduling * * \param[in,out] cib CIB XML to unpack (may be NULL if already unpacked) * \param[in] flags Working set flags to set in addition to defaults * \param[in,out] data_set Cluster working set */ static void unpack_cib(xmlNode *cib, unsigned long long flags, pe_working_set_t *data_set) { const char* localhost_save = NULL; if (pcmk_is_set(data_set->flags, pe_flag_have_status)) { crm_trace("Reusing previously calculated cluster status"); pe__set_working_set_flags(data_set, flags); return; } if (data_set->localhost) { localhost_save = data_set->localhost; } CRM_ASSERT(cib != NULL); crm_trace("Calculating cluster status"); /* This will zero the entire struct without freeing anything first, so * callers should never call pcmk__schedule_actions() with a populated data * set unless pe_flag_have_status is set (i.e. cluster_status() was * previously called, whether directly or via pcmk__schedule_actions()). */ set_working_set_defaults(data_set); if (localhost_save) { data_set->localhost = localhost_save; } pe__set_working_set_flags(data_set, flags); data_set->input = cib; cluster_status(data_set); // Sets pe_flag_have_status } /*! * \internal * \brief Run the scheduler for a given CIB * * \param[in,out] cib CIB XML to use as scheduler input * \param[in] flags Working set flags to set in addition to defaults * \param[in,out] data_set Cluster working set */ void pcmk__schedule_actions(xmlNode *cib, unsigned long long flags, pe_working_set_t *data_set) { unpack_cib(cib, flags, data_set); - pcmk__set_allocation_methods(data_set); + pcmk__set_assignment_methods(data_set); pcmk__apply_node_health(data_set); pcmk__unpack_constraints(data_set); if (pcmk_is_set(data_set->flags, pe_flag_check_config)) { return; } if (!pcmk_is_set(data_set->flags, pe_flag_quick_location) && pcmk__is_daemon) { log_resource_details(data_set); } apply_node_criteria(data_set); if (pcmk_is_set(data_set->flags, pe_flag_quick_location)) { return; } pcmk__create_internal_constraints(data_set); pcmk__handle_rsc_config_changes(data_set); - allocate_resources(data_set); + assign_resources(data_set); schedule_resource_actions(data_set); /* Remote ordering constraints need to happen prior to calculating fencing * because it is one more place we can mark nodes as needing fencing. */ pcmk__order_remote_connection_actions(data_set); schedule_fencing_and_shutdowns(data_set); pcmk__apply_orderings(data_set); log_all_actions(data_set); pcmk__create_graph(data_set); if (get_crm_log_level() == LOG_TRACE) { log_unrunnable_actions(data_set); } } diff --git a/lib/pacemaker/pcmk_simulate.c b/lib/pacemaker/pcmk_simulate.c index 165c7d3c15..228909f13c 100644 --- a/lib/pacemaker/pcmk_simulate.c +++ b/lib/pacemaker/pcmk_simulate.c @@ -1,999 +1,999 @@ /* * Copyright 2021-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include "libpacemaker_private.h" static pcmk__output_t *out = NULL; static cib_t *fake_cib = NULL; static GList *fake_resource_list = NULL; static const GList *fake_op_fail_list = NULL; static void set_effective_date(pe_working_set_t *data_set, bool print_original, const char *use_date); /*! * \internal * \brief Create an action name for use in a dot graph * * \param[in] action Action to create name for * \param[in] verbose If true, add action ID to name * * \return Newly allocated string with action name * \note It is the caller's responsibility to free the result. */ static char * create_action_name(const pe_action_t *action, bool verbose) { char *action_name = NULL; const char *prefix = ""; const char *action_host = NULL; const char *clone_name = NULL; const char *task = action->task; if (action->node != NULL) { action_host = action->node->details->uname; } else if (!pcmk_is_set(action->flags, pe_action_pseudo)) { action_host = ""; } if (pcmk__str_eq(action->task, RSC_CANCEL, pcmk__str_none)) { prefix = "Cancel "; task = action->cancel_task; } if (action->rsc != NULL) { clone_name = action->rsc->clone_name; } if (clone_name != NULL) { char *key = NULL; guint interval_ms = 0; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } if (pcmk__strcase_any_of(action->task, RSC_NOTIFY, RSC_NOTIFIED, NULL)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_key_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_key_operation"); CRM_ASSERT(n_type != NULL); CRM_ASSERT(n_task != NULL); key = pcmk__notify_key(clone_name, n_type, n_task); } else { key = pcmk__op_key(clone_name, task, interval_ms); } if (action_host != NULL) { action_name = crm_strdup_printf("%s%s %s", prefix, key, action_host); } else { action_name = crm_strdup_printf("%s%s", prefix, key); } free(key); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); action_name = crm_strdup_printf("%s%s '%s' %s", prefix, action->task, op, action_host); } else if (action->rsc && action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->uuid, action_host); } else if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->task, action_host); } else { action_name = crm_strdup_printf("%s", action->uuid); } if (verbose) { char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id); free(action_name); action_name = with_id; } return action_name; } /*! * \internal * \brief Display the status of a cluster * * \param[in,out] data_set Cluster working set * \param[in] show_opts How to modify display (as pcmk_show_opt_e flags) * \param[in] section_opts Sections to display (as pcmk_section_e flags) * \param[in] title What to use as list title * \param[in] print_spacer Whether to display a spacer first */ static void print_cluster_status(pe_working_set_t *data_set, uint32_t show_opts, uint32_t section_opts, const char *title, bool print_spacer) { pcmk__output_t *out = data_set->priv; GList *all = NULL; crm_exit_t stonith_rc = 0; enum pcmk_pacemakerd_state state = pcmk_pacemakerd_state_invalid; section_opts |= pcmk_section_nodes | pcmk_section_resources; show_opts |= pcmk_show_inactive_rscs | pcmk_show_failed_detail; all = g_list_prepend(all, (gpointer) "*"); PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "%s", title); out->message(out, "cluster-status", data_set, state, stonith_rc, NULL, false, section_opts, show_opts, NULL, all, all); out->end_list(out); g_list_free(all); } /*! * \internal * \brief Display a summary of all actions scheduled in a transition * * \param[in,out] data_set Cluster working set (fully scheduled) * \param[in] print_spacer Whether to display a spacer first */ static void print_transition_summary(pe_working_set_t *data_set, bool print_spacer) { pcmk__output_t *out = data_set->priv; PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "Transition Summary"); pcmk__output_actions(data_set); out->end_list(out); } /*! * \internal * \brief Reset a cluster working set's input, output, date, and flags * * \param[in,out] data_set Cluster working set * \param[in] input What to set as cluster input * \param[in] out What to set as cluster output object * \param[in] use_date What to set as cluster's current timestamp * \param[in] flags Cluster flags to add (pe_flag_*) */ static void reset(pe_working_set_t *data_set, xmlNodePtr input, pcmk__output_t *out, const char *use_date, unsigned int flags) { data_set->input = input; data_set->priv = out; set_effective_date(data_set, true, use_date); if (pcmk_is_set(flags, pcmk_sim_sanitized)) { pe__set_working_set_flags(data_set, pe_flag_sanitized); } if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } } /*! * \brief Write out a file in dot(1) format describing the actions that will * be taken by the scheduler in response to an input CIB file. * * \param[in,out] data_set Working set for the cluster * \param[in] dot_file The filename to write * \param[in] all_actions Write all actions, even those that are optional * or are on unmanaged resources * \param[in] verbose Add extra information, such as action IDs, to the * output * * \return Standard Pacemaker return code */ static int write_sim_dotfile(pe_working_set_t *data_set, const char *dot_file, bool all_actions, bool verbose) { GList *gIter = NULL; FILE *dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { return errno; } fprintf(dot_strm, " digraph \"g\" {\n"); for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; const char *style = "dashed"; const char *font = "black"; const char *color = "black"; char *action_name = create_action_name(action, verbose); if (pcmk_is_set(action->flags, pe_action_pseudo)) { font = "orange"; } if (pcmk_is_set(action->flags, pe_action_dumped)) { style = "bold"; color = "green"; } else if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pe_rsc_managed)) { color = "red"; font = "purple"; if (!all_actions) { goto do_not_write; } } else if (pcmk_is_set(action->flags, pe_action_optional)) { color = "blue"; if (!all_actions) { goto do_not_write; } } else { color = "red"; CRM_LOG_ASSERT(!pcmk_is_set(action->flags, pe_action_runnable)); } pe__set_action_flags(action, pe_action_dumped); fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n", action_name, style, color, font); do_not_write: free(action_name); } for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; GList *gIter2 = NULL; for (gIter2 = action->actions_before; gIter2 != NULL; gIter2 = gIter2->next) { pe_action_wrapper_t *before = (pe_action_wrapper_t *) gIter2->data; char *before_name = NULL; char *after_name = NULL; const char *style = "dashed"; bool optional = true; if (before->state == pe_link_dumped) { optional = false; style = "bold"; } else if (before->type == pe_order_none) { continue; } else if (pcmk_is_set(before->action->flags, pe_action_dumped) && pcmk_is_set(action->flags, pe_action_dumped) && before->type != pe_order_load) { optional = false; } if (all_actions || !optional) { before_name = create_action_name(before->action, verbose); after_name = create_action_name(action, verbose); fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n", before_name, after_name, style); free(before_name); free(after_name); } } } fprintf(dot_strm, "}\n"); fflush(dot_strm); fclose(dot_strm); return pcmk_rc_ok; } /*! * \brief Profile the configuration updates and scheduler actions in a single * CIB file, printing the profiling timings. * * \note \p data_set->priv must have been set to a valid \p pcmk__output_t * object before this function is called. * * \param[in] xml_file The CIB file to profile * \param[in] repeat Number of times to run * \param[in,out] data_set Working set for the cluster * \param[in] use_date The date to set the cluster's time to (may be NULL) */ static void profile_file(const char *xml_file, long long repeat, pe_working_set_t *data_set, const char *use_date) { pcmk__output_t *out = data_set->priv; xmlNode *cib_object = NULL; clock_t start = 0; clock_t end; unsigned long long data_set_flags = pe_flag_no_compat; CRM_ASSERT(out != NULL); cib_object = filename2xml(xml_file); start = clock(); if (pcmk_find_cib_element(cib_object, XML_CIB_TAG_STATUS) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return; } if (pcmk_is_set(data_set->flags, pe_flag_show_scores)) { data_set_flags |= pe_flag_show_scores; } if (pcmk_is_set(data_set->flags, pe_flag_show_utilization)) { data_set_flags |= pe_flag_show_utilization; } for (int i = 0; i < repeat; ++i) { xmlNode *input = (repeat == 1)? cib_object : copy_xml(cib_object); data_set->input = input; set_effective_date(data_set, false, use_date); pcmk__schedule_actions(input, data_set_flags, data_set); pe_reset_working_set(data_set); } end = clock(); out->message(out, "profile", xml_file, start, end); } void pcmk__profile_dir(const char *dir, long long repeat, pe_working_set_t *data_set, const char *use_date) { pcmk__output_t *out = data_set->priv; struct dirent **namelist; int file_num = scandir(dir, &namelist, 0, alphasort); CRM_ASSERT(out != NULL); if (file_num > 0) { struct stat prop; char buffer[FILENAME_MAX]; out->begin_list(out, NULL, NULL, "Timings"); while (file_num--) { if ('.' == namelist[file_num]->d_name[0]) { free(namelist[file_num]); continue; } else if (!pcmk__ends_with_ext(namelist[file_num]->d_name, ".xml")) { free(namelist[file_num]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", dir, namelist[file_num]->d_name); if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) { profile_file(buffer, repeat, data_set, use_date); } free(namelist[file_num]); } free(namelist); out->end_list(out); } } /*! * \brief Set the date of the cluster, either to the value given by * \p use_date, or to the "execution-date" value in the CIB. * * \note \p data_set->priv must have been set to a valid \p pcmk__output_t * object before this function is called. * * \param[in,out] data_set Working set for the cluster * \param[in] print_original If \p true, the "execution-date" should * also be printed * \param[in] use_date The date to set the cluster's time to * (may be NULL) */ static void set_effective_date(pe_working_set_t *data_set, bool print_original, const char *use_date) { pcmk__output_t *out = data_set->priv; time_t original_date = 0; CRM_ASSERT(out != NULL); crm_element_value_epoch(data_set->input, "execution-date", &original_date); if (use_date) { data_set->now = crm_time_new(use_date); out->info(out, "Setting effective cluster time: %s", use_date); crm_time_log(LOG_NOTICE, "Pretending 'now' is", data_set->now, crm_time_log_date | crm_time_log_timeofday); } else if (original_date != 0) { data_set->now = pcmk__copy_timet(original_date); if (print_original) { char *when = crm_time_as_string(data_set->now, crm_time_log_date|crm_time_log_timeofday); out->info(out, "Using the original execution date of: %s", when); free(when); } } } /*! * \internal * \brief Simulate successfully executing a pseudo-action in a graph * * \param[in,out] graph Graph to update with pseudo-action result * \param[in,out] action Pseudo-action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_pseudo_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-pseudo-action", node, task); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate executing a resource action in a graph * * \param[in,out] graph Graph to update with resource action result * \param[in,out] action Resource action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_resource_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { int rc; lrmd_event_data_t *op = NULL; int target_outcome = PCMK_OCF_OK; const char *rtype = NULL; const char *rclass = NULL; const char *resource = NULL; const char *rprovider = NULL; const char *resource_config_name = NULL; const char *operation = crm_element_value(action->xml, "operation"); const char *target_rc_s = crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC); xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; xmlNode *action_rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE); char *node = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET); char *uuid = NULL; const char *router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); // Certain actions don't need to be displayed or history entries if (pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) { crm_debug("No history injection for %s op on %s", operation, node); goto done; // Confirm action and update graph } if (action_rsc == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "Bad"); free(node); return EPROTO; } /* A resource might be known by different names in the configuration and in * the action (for example, a clone instance). Grab the configuration name * (which is preferred when writing history), and if necessary, the instance * name. */ resource_config_name = crm_element_value(action_rsc, XML_ATTR_ID); if (resource_config_name == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "No ID"); free(node); return EPROTO; } resource = resource_config_name; if (pe_find_resource(fake_resource_list, resource) == NULL) { const char *longname = crm_element_value(action_rsc, XML_ATTR_ID_LONG); if ((longname != NULL) && (pe_find_resource(fake_resource_list, longname) != NULL)) { resource = longname; } } // Certain actions need to be displayed but don't need history entries if (pcmk__strcase_any_of(operation, "delete", RSC_METADATA, NULL)) { out->message(out, "inject-rsc-action", resource, operation, node, (guint) 0); goto done; // Confirm action and update graph } rclass = crm_element_value(action_rsc, XML_AGENT_ATTR_CLASS); rtype = crm_element_value(action_rsc, XML_ATTR_TYPE); rprovider = crm_element_value(action_rsc, XML_AGENT_ATTR_PROVIDER); pcmk__scan_min_int(target_rc_s, &target_outcome, 0); CRM_ASSERT(fake_cib->cmds->query(fake_cib, NULL, NULL, cib_sync_call|cib_scope_local) == pcmk_ok); // Ensure the action node is in the CIB uuid = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET_UUID); cib_node = pcmk__inject_node(fake_cib, node, ((router_node == NULL)? uuid: node)); free(uuid); CRM_ASSERT(cib_node != NULL); // Add a history entry for the action cib_resource = pcmk__inject_resource_history(out, cib_node, resource, resource_config_name, rclass, rtype, rprovider); if (cib_resource == NULL) { crm_err("Could not simulate action %d history for resource %s", action->id, resource); free(node); free_xml(cib_node); return EINVAL; } // Simulate and display an executor event for the action result op = pcmk__event_from_graph_action(cib_resource, action, PCMK_EXEC_DONE, target_outcome, "User-injected result"); out->message(out, "inject-rsc-action", resource, op->op_type, node, op->interval_ms); // Check whether action is in a list of desired simulated failures for (const GList *iter = fake_op_fail_list; iter != NULL; iter = iter->next) { const char *spec = (const char *) iter->data; char *key = NULL; const char *match_name = NULL; // Allow user to specify anonymous clone with or without instance number key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource; } free(key); // If not found, try the resource's name in the configuration if ((match_name == NULL) && (strcmp(resource, resource_config_name) != 0)) { key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource_config_name, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource_config_name; } free(key); } if (match_name == NULL) { continue; // This failed action entry doesn't match } // ${match_name}_${task}_${interval_in_ms}@${node}=${rc} rc = sscanf(spec, "%*[^=]=%d", (int *) &op->rc); if (rc != 1) { out->err(out, "Invalid failed operation '%s' " "(result code must be integer)", spec); continue; // Keep checking other list entries } out->info(out, "Pretending action %d failed with rc=%d", action->id, op->rc); pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); graph->abort_priority = INFINITY; pcmk__inject_failcount(out, cib_node, match_name, op->op_type, op->interval_ms, op->rc); break; } pcmk__inject_action_result(cib_resource, op, target_outcome); lrmd_free_event(op); rc = fake_cib->cmds->modify(fake_cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); done: free(node); free_xml(cib_node); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a cluster action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Cluster action to simulate * * \return Standard Pacemaker return code */ static int simulate_cluster_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); xmlNode *rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-cluster-action", node, task, rsc); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a fencing action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Fencing action to simulate * * \return Standard Pacemaker return code */ static int simulate_fencing_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *op = crm_meta_value(action->params, "stonith_action"); char *target = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET); out->message(out, "inject-fencing-action", target, op); if (!pcmk__str_eq(op, "on", pcmk__str_casei)) { int rc = pcmk_ok; GString *xpath = g_string_sized_new(512); // Set node state to offline xmlNode *cib_node = pcmk__inject_node_state_change(fake_cib, target, false); CRM_ASSERT(cib_node != NULL); crm_xml_add(cib_node, XML_ATTR_ORIGIN, __func__); rc = fake_cib->cmds->replace(fake_cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); // Simulate controller clearing node's resource history and attributes pcmk__g_strcat(xpath, "//" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='", target, "']/" XML_CIB_TAG_LRM, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call|cib_scope_local); g_string_truncate(xpath, 0); pcmk__g_strcat(xpath, "//" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='", target, "']" "/" XML_TAG_TRANSIENT_NODEATTRS, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call|cib_scope_local); free_xml(cib_node); g_string_free(xpath, TRUE); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); free(target); return pcmk_rc_ok; } enum pcmk__graph_status pcmk__simulate_transition(pe_working_set_t *data_set, cib_t *cib, const GList *op_fail_list) { pcmk__graph_t *transition = NULL; enum pcmk__graph_status graph_rc; pcmk__graph_functions_t simulation_fns = { simulate_pseudo_action, simulate_resource_action, simulate_cluster_action, simulate_fencing_action, }; out = data_set->priv; fake_cib = cib; fake_op_fail_list = op_fail_list; if (!out->is_quiet(out)) { out->begin_list(out, NULL, NULL, "Executing Cluster Transition"); } pcmk__set_graph_functions(&simulation_fns); transition = pcmk__unpack_graph(data_set->graph, crm_system_name); pcmk__log_graph(LOG_DEBUG, transition); fake_resource_list = data_set->resources; do { graph_rc = pcmk__execute_graph(transition); } while (graph_rc == pcmk__graph_active); fake_resource_list = NULL; if (graph_rc != pcmk__graph_complete) { out->err(out, "Transition failed: %s", pcmk__graph_status2text(graph_rc)); pcmk__log_graph(LOG_ERR, transition); out->err(out, "An invalid transition was produced"); } pcmk__free_graph(transition); if (!out->is_quiet(out)) { // If not quiet, we'll need the resulting CIB for later display xmlNode *cib_object = NULL; int rc = fake_cib->cmds->query(fake_cib, NULL, &cib_object, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); pe_reset_working_set(data_set); data_set->input = cib_object; out->end_list(out); } return graph_rc; } int pcmk__simulate(pe_working_set_t *data_set, pcmk__output_t *out, const pcmk_injections_t *injections, unsigned int flags, uint32_t section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { int printed = pcmk_rc_no_output; int rc = pcmk_rc_ok; xmlNodePtr input = NULL; cib_t *cib = NULL; rc = cib__signon_query(out, &cib, &input); if (rc != pcmk_rc_ok) { goto simulate_done; } reset(data_set, input, out, use_date, flags); cluster_status(data_set); if ((cib->variant == cib_native) && pcmk_is_set(section_opts, pcmk_section_times)) { if (pcmk__our_nodename == NULL) { // Currently used only in the times section pcmk__query_node_name(out, 0, &pcmk__our_nodename, 0); } data_set->localhost = pcmk__our_nodename; } if (!out->is_quiet(out)) { if (pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { printed = out->message(out, "maint-mode", data_set->flags); } if (data_set->disabled_resources || data_set->blocked_resources) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); printed = out->info(out, "%d of %d resource instances DISABLED and " "%d BLOCKED from further action due to failure", data_set->disabled_resources, data_set->ninstances, data_set->blocked_resources); } /* Most formatted output headers use caps for each word, but this one * only has the first word capitalized for compatibility with pcs. */ print_cluster_status(data_set, pcmk_is_set(flags, pcmk_sim_show_pending)? pcmk_show_pending : 0, section_opts, "Current cluster status", (printed == pcmk_rc_ok)); printed = pcmk_rc_ok; } // If the user requested any injections, handle them if ((injections->node_down != NULL) || (injections->node_fail != NULL) || (injections->node_up != NULL) || (injections->op_inject != NULL) || (injections->ticket_activate != NULL) || (injections->ticket_grant != NULL) || (injections->ticket_revoke != NULL) || (injections->ticket_standby != NULL) || (injections->watchdog != NULL)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); pcmk__inject_scheduler_input(data_set, cib, injections); printed = pcmk_rc_ok; rc = cib->cmds->query(cib, NULL, &input, cib_sync_call); if (rc != pcmk_rc_ok) { rc = pcmk_legacy2rc(rc); goto simulate_done; } cleanup_calculations(data_set); reset(data_set, input, out, use_date, flags); cluster_status(data_set); } if (input_file != NULL) { rc = write_xml_file(input, input_file, FALSE); if (rc < 0) { rc = pcmk_legacy2rc(rc); goto simulate_done; } } if (pcmk_any_flags_set(flags, pcmk_sim_process | pcmk_sim_simulate)) { pcmk__output_t *logger_out = NULL; unsigned long long data_set_flags = pe_flag_no_compat; if (pcmk_is_set(data_set->flags, pe_flag_show_scores)) { data_set_flags |= pe_flag_show_scores; } if (pcmk_is_set(data_set->flags, pe_flag_show_utilization)) { data_set_flags |= pe_flag_show_utilization; } if (pcmk_all_flags_set(data_set->flags, pe_flag_show_scores|pe_flag_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, - "Allocation Scores and Utilization Information"); + "Assignment Scores and Utilization Information"); printed = pcmk_rc_ok; } else if (pcmk_is_set(data_set->flags, pe_flag_show_scores)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); - out->begin_list(out, NULL, NULL, "Allocation Scores"); + out->begin_list(out, NULL, NULL, "Assignment Scores"); printed = pcmk_rc_ok; } else if (pcmk_is_set(data_set->flags, pe_flag_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Utilization Information"); printed = pcmk_rc_ok; } else { rc = pcmk__log_output_new(&logger_out); if (rc != pcmk_rc_ok) { goto simulate_done; } pe__register_messages(logger_out); pcmk__register_lib_messages(logger_out); data_set->priv = logger_out; } pcmk__schedule_actions(input, data_set_flags, data_set); if (logger_out == NULL) { out->end_list(out); } else { logger_out->finish(logger_out, CRM_EX_OK, true, NULL); pcmk__output_free(logger_out); data_set->priv = out; } input = NULL; /* Don't try and free it twice */ if (graph_file != NULL) { rc = write_xml_file(data_set->graph, graph_file, FALSE); if (rc < 0) { rc = pcmk_rc_graph_error; goto simulate_done; } } if (dot_file != NULL) { rc = write_sim_dotfile(data_set, dot_file, pcmk_is_set(flags, pcmk_sim_all_actions), pcmk_is_set(flags, pcmk_sim_verbose)); if (rc != pcmk_rc_ok) { rc = pcmk_rc_dot_error; goto simulate_done; } } if (!out->is_quiet(out)) { print_transition_summary(data_set, printed == pcmk_rc_ok); } } rc = pcmk_rc_ok; if (!pcmk_is_set(flags, pcmk_sim_simulate)) { goto simulate_done; } PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); if (pcmk__simulate_transition(data_set, cib, injections->op_fail) != pcmk__graph_complete) { rc = pcmk_rc_invalid_transition; } if (out->is_quiet(out)) { goto simulate_done; } set_effective_date(data_set, true, use_date); if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } cluster_status(data_set); print_cluster_status(data_set, 0, section_opts, "Revised Cluster Status", true); simulate_done: cib__clean_up_connection(&cib); return rc; } int pcmk_simulate(xmlNodePtr *xml, pe_working_set_t *data_set, const pcmk_injections_t *injections, unsigned int flags, unsigned int section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pe__register_messages(out); pcmk__register_lib_messages(out); rc = pcmk__simulate(data_set, out, injections, flags, section_opts, use_date, input_file, graph_file, dot_file); pcmk__xml_output_finish(out, xml); return rc; } diff --git a/lib/pacemaker/pcmk_status.c b/lib/pacemaker/pcmk_status.c index 0e82633fb9..be929fb6b2 100644 --- a/lib/pacemaker/pcmk_status.c +++ b/lib/pacemaker/pcmk_status.c @@ -1,368 +1,369 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include +#include #include #include #include static stonith_t * fencing_connect(void) { stonith_t *st = stonith_api_new(); int rc = pcmk_rc_ok; if (st == NULL) { return NULL; } rc = st->cmds->connect(st, crm_system_name, NULL); if (rc == pcmk_rc_ok) { return st; } else { stonith_api_delete(st); return NULL; } } /*! * \internal * \brief Output the cluster status given a fencer and CIB connection * * \param[in,out] out Output object * \param[in,out] stonith Fencer connection * \param[in,out] cib CIB connection * \param[in] current_cib Current CIB XML * \param[in] pcmkd_state \p pacemakerd state * \param[in] fence_history How much of the fencing history to output * \param[in] show Group of \p pcmk_section_e flags * \param[in] show_opts Group of \p pcmk_show_opt_e flags * \param[in] only_node If a node name or tag, include only the * matching node(s) (if any) in the output. * If \p "*" or \p NULL, include all nodes * in the output. * \param[in] only_rsc If a resource ID or tag, include only the * matching resource(s) (if any) in the * output. If \p "*" or \p NULL, include all * resources in the output. * \param[in] neg_location_prefix Prefix denoting a ban in a constraint ID * \param[in] simple_output Whether to use a simple output format. * Note: This is for use by \p crm_mon only * and is planned to be deprecated. * * \return Standard Pacemaker return code */ int pcmk__output_cluster_status(pcmk__output_t *out, stonith_t *stonith, cib_t *cib, xmlNode *current_cib, enum pcmk_pacemakerd_state pcmkd_state, enum pcmk__fence_history fence_history, uint32_t show, uint32_t show_opts, const char *only_node, const char *only_rsc, const char *neg_location_prefix, bool simple_output) { xmlNode *cib_copy = copy_xml(current_cib); stonith_history_t *stonith_history = NULL; int history_rc = 0; pe_working_set_t *data_set = NULL; GList *unames = NULL; GList *resources = NULL; int rc = pcmk_rc_ok; if (cli_config_update(&cib_copy, NULL, FALSE) == FALSE) { cib__clean_up_connection(&cib); free_xml(cib_copy); rc = pcmk_rc_schema_validation; out->err(out, "Upgrade failed: %s", pcmk_rc_str(rc)); return rc; } /* get the stonith-history if there is evidence we need it */ if (fence_history != pcmk__fence_history_none) { history_rc = pcmk__get_fencing_history(stonith, &stonith_history, fence_history); } data_set = pe_new_working_set(); CRM_ASSERT(data_set != NULL); pe__set_working_set_flags(data_set, pe_flag_no_compat); data_set->input = cib_copy; data_set->priv = out; cluster_status(data_set); if ((cib->variant == cib_native) && pcmk_is_set(show, pcmk_section_times)) { if (pcmk__our_nodename == NULL) { // Currently used only in the times section pcmk__query_node_name(out, 0, &pcmk__our_nodename, 0); } data_set->localhost = pcmk__our_nodename; } /* Unpack constraints if any section will need them * (tickets may be referenced in constraints but not granted yet, * and bans need negative location constraints) */ if (pcmk_is_set(show, pcmk_section_bans) || pcmk_is_set(show, pcmk_section_tickets)) { pcmk__unpack_constraints(data_set); } unames = pe__build_node_name_list(data_set, only_node); resources = pe__build_rsc_list(data_set, only_rsc); /* Always print DC if NULL. */ if (data_set->dc_node == NULL) { show |= pcmk_section_dc; } if (simple_output) { rc = pcmk__output_simple_status(out, data_set); } else { out->message(out, "cluster-status", data_set, pcmkd_state, pcmk_rc2exitc(history_rc), stonith_history, fence_history, show, show_opts, neg_location_prefix, unames, resources); } g_list_free_full(unames, free); g_list_free_full(resources, free); stonith_history_free(stonith_history); stonith_history = NULL; pe_free_working_set(data_set); return rc; } int pcmk_status(xmlNodePtr *xml) { cib_t *cib = NULL; pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; uint32_t show_opts = pcmk_show_pending | pcmk_show_inactive_rscs | pcmk_show_timing; cib = cib_new(); if (cib == NULL) { return pcmk_rc_cib_corrupt; } rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { cib_delete(cib); return rc; } pcmk__register_lib_messages(out); pe__register_messages(out); stonith__register_messages(out); rc = pcmk__status(out, cib, pcmk__fence_history_full, pcmk_section_all, show_opts, NULL, NULL, NULL, false, 0); pcmk__xml_output_finish(out, xml); cib_delete(cib); return rc; } /*! * \internal * \brief Query and output the cluster status * * The operation is considered a success if we're able to get the \p pacemakerd * state. If possible, we'll also try to connect to the fencer and CIB and * output their respective status information. * * \param[in,out] out Output object * \param[in,out] cib CIB connection * \param[in] fence_history How much of the fencing history to output * \param[in] show Group of \p pcmk_section_e flags * \param[in] show_opts Group of \p pcmk_show_opt_e flags * \param[in] only_node If a node name or tag, include only the * matching node(s) (if any) in the output. * If \p "*" or \p NULL, include all nodes * in the output. * \param[in] only_rsc If a resource ID or tag, include only the * matching resource(s) (if any) in the * output. If \p "*" or \p NULL, include all * resources in the output. * \param[in] neg_location_prefix Prefix denoting a ban in a constraint ID * \param[in] simple_output Whether to use a simple output format. * Note: This is for use by \p crm_mon only * and is planned to be deprecated. * \param[in] timeout_ms How long to wait for a reply from the * \p pacemakerd API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * If positive, \p pcmk_ipc_dispatch_main * will be used, and a new mainloop will be * created for this purpose (freed before * return). * * \return Standard Pacemaker return code */ int pcmk__status(pcmk__output_t *out, cib_t *cib, enum pcmk__fence_history fence_history, uint32_t show, uint32_t show_opts, const char *only_node, const char *only_rsc, const char *neg_location_prefix, bool simple_output, unsigned int timeout_ms) { xmlNode *current_cib = NULL; int rc = pcmk_rc_ok; stonith_t *stonith = NULL; enum pcmk_pacemakerd_state pcmkd_state = pcmk_pacemakerd_state_invalid; time_t last_updated = 0; if (cib == NULL) { return ENOTCONN; } if (cib->variant == cib_native) { rc = pcmk__pacemakerd_status(out, crm_system_name, timeout_ms, false, &pcmkd_state); if (rc != pcmk_rc_ok) { return rc; } last_updated = time(NULL); switch (pcmkd_state) { case pcmk_pacemakerd_state_running: case pcmk_pacemakerd_state_shutting_down: case pcmk_pacemakerd_state_remote: /* Fencer and CIB may still be available while shutting down or * running on a Pacemaker Remote node */ break; default: // Fencer and CIB are definitely unavailable out->message(out, "pacemakerd-health", NULL, pcmkd_state, NULL, last_updated); return rc; } if (fence_history != pcmk__fence_history_none) { stonith = fencing_connect(); } } rc = cib__signon_query(out, &cib, ¤t_cib); if (rc != pcmk_rc_ok) { if (pcmkd_state != pcmk_pacemakerd_state_invalid) { // Invalid at this point means we didn't query the pcmkd state out->message(out, "pacemakerd-health", NULL, pcmkd_state, NULL, last_updated); } goto done; } rc = pcmk__output_cluster_status(out, stonith, cib, current_cib, pcmkd_state, fence_history, show, show_opts, only_node, only_rsc, neg_location_prefix, simple_output); if (rc != pcmk_rc_ok) { out->err(out, "Error outputting status info from the fencer or CIB"); } done: stonith_api_delete(stonith); free_xml(current_cib); return pcmk_rc_ok; } /* This is an internal-only function that is planned to be deprecated and removed. * It should only ever be called from crm_mon. */ int pcmk__output_simple_status(pcmk__output_t *out, const pe_working_set_t *data_set) { int nodes_online = 0; int nodes_standby = 0; int nodes_maintenance = 0; GString *offline_nodes = NULL; bool no_dc = false; bool offline = false; bool has_warnings = false; if (data_set->dc_node == NULL) { has_warnings = true; no_dc = true; } for (GList *iter = data_set->nodes; iter != NULL; iter = iter->next) { pe_node_t *node = (pe_node_t *) iter->data; if (node->details->standby && node->details->online) { nodes_standby++; } else if (node->details->maintenance && node->details->online) { nodes_maintenance++; } else if (node->details->online) { nodes_online++; } else { pcmk__add_word(&offline_nodes, 1024, "offline node:"); pcmk__add_word(&offline_nodes, 0, pe__node_name(node)); has_warnings = true; offline = true; } } if (has_warnings) { out->info(out, "CLUSTER WARN: %s%s%s", no_dc ? "No DC" : "", no_dc && offline ? ", " : "", (offline? (const char *) offline_nodes->str : "")); if (offline_nodes != NULL) { g_string_free(offline_nodes, TRUE); } } else { char *nodes_standby_s = NULL; char *nodes_maint_s = NULL; if (nodes_standby > 0) { nodes_standby_s = crm_strdup_printf(", %d standby node%s", nodes_standby, pcmk__plural_s(nodes_standby)); } if (nodes_maintenance > 0) { nodes_maint_s = crm_strdup_printf(", %d maintenance node%s", nodes_maintenance, pcmk__plural_s(nodes_maintenance)); } out->info(out, "CLUSTER OK: %d node%s online%s%s, " "%d resource instance%s configured", nodes_online, pcmk__plural_s(nodes_online), nodes_standby_s != NULL ? nodes_standby_s : "", nodes_maint_s != NULL ? nodes_maint_s : "", data_set->ninstances, pcmk__plural_s(data_set->ninstances)); free(nodes_standby_s); free(nodes_maint_s); } if (has_warnings) { return pcmk_rc_error; } else { return pcmk_rc_ok; } /* coverity[leaked_storage] False positive */ } diff --git a/lib/pengine/pe_notif.c b/lib/pengine/pe_notif.c index 7ed490f7dd..21852987d5 100644 --- a/lib/pengine/pe_notif.c +++ b/lib/pengine/pe_notif.c @@ -1,996 +1,998 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include + +#include #include #include "pe_status_private.h" typedef struct notify_entry_s { const pe_resource_t *rsc; const pe_node_t *node; } notify_entry_t; /*! * \internal * \brief Compare two notification entries * * Compare two notification entries, where the one with the alphabetically first * resource name (or if equal, node name) sorts as first, with NULL sorting as * less than non-NULL. * * \param[in] a First notification entry to compare * \param[in] b Second notification entry to compare * * \return -1 if \p a sorts before \p b, 0 if they are equal, otherwise 1 */ static gint compare_notify_entries(gconstpointer a, gconstpointer b) { int tmp; const notify_entry_t *entry_a = a; const notify_entry_t *entry_b = b; // NULL a or b is not actually possible if ((entry_a == NULL) && (entry_b == NULL)) { return 0; } if (entry_a == NULL) { return 1; } if (entry_b == NULL) { return -1; } // NULL resources sort first if ((entry_a->rsc == NULL) && (entry_b->rsc == NULL)) { return 0; } if (entry_a->rsc == NULL) { return 1; } if (entry_b->rsc == NULL) { return -1; } // Compare resource names tmp = strcmp(entry_a->rsc->id, entry_b->rsc->id); if (tmp != 0) { return tmp; } // Otherwise NULL nodes sort first if ((entry_a->node == NULL) && (entry_b->node == NULL)) { return 0; } if (entry_a->node == NULL) { return 1; } if (entry_b->node == NULL) { return -1; } // Finally, compare node names return strcmp(entry_a->node->details->id, entry_b->node->details->id); } /*! * \internal * \brief Duplicate a notification entry * * \param[in] entry Entry to duplicate * * \return Newly allocated duplicate of \p entry * \note It is the caller's responsibility to free the return value. */ static notify_entry_t * dup_notify_entry(const notify_entry_t *entry) { notify_entry_t *dup = calloc(1, sizeof(notify_entry_t)); CRM_ASSERT(dup != NULL); dup->rsc = entry->rsc; dup->node = entry->node; return dup; } /*! * \internal * \brief Given a list of nodes, create strings with node names * * \param[in] list List of nodes (as pe_node_t *) * \param[out] all_node_names If not NULL, will be set to space-separated list * of the names of all nodes in \p list * \param[out] host_node_names Same as \p all_node_names, except active * guest nodes will list the name of their host * * \note The caller is responsible for freeing the output argument values using * \p g_string_free(). */ static void get_node_names(const GList *list, GString **all_node_names, GString **host_node_names) { if (all_node_names != NULL) { *all_node_names = NULL; } if (host_node_names != NULL) { *host_node_names = NULL; } for (const GList *iter = list; iter != NULL; iter = iter->next) { const pe_node_t *node = (const pe_node_t *) iter->data; if (node->details->uname == NULL) { continue; } // Always add to list of all node names if (all_node_names != NULL) { pcmk__add_word(all_node_names, 1024, node->details->uname); } // Add to host node name list if appropriate if (host_node_names != NULL) { if (pe__is_guest_node(node) && (node->details->remote_rsc->container->running_on != NULL)) { node = pe__current_node(node->details->remote_rsc->container); if (node->details->uname == NULL) { continue; } } pcmk__add_word(host_node_names, 1024, node->details->uname); } } if ((all_node_names != NULL) && (*all_node_names == NULL)) { *all_node_names = g_string_new(" "); } if ((host_node_names != NULL) && (*host_node_names == NULL)) { *host_node_names = g_string_new(" "); } } /*! * \internal * \brief Create strings of instance and node names from notification entries * * \param[in,out] list List of notification entries (will be sorted here) * \param[out] rsc_names If not NULL, will be set to space-separated list * of clone instances from \p list * \param[out] node_names If not NULL, will be set to space-separated list * of node names from \p list * * \return (Possibly new) head of sorted \p list * \note The caller is responsible for freeing the output argument values using * \p g_list_free_full() and \p g_string_free(). */ static GList * notify_entries_to_strings(GList *list, GString **rsc_names, GString **node_names) { const char *last_rsc_id = NULL; // Initialize output lists to NULL if (rsc_names != NULL) { *rsc_names = NULL; } if (node_names != NULL) { *node_names = NULL; } // Sort input list for user-friendliness (and ease of filtering duplicates) list = g_list_sort(list, compare_notify_entries); for (GList *gIter = list; gIter != NULL; gIter = gIter->next) { notify_entry_t *entry = (notify_entry_t *) gIter->data; // Entry must have a resource (with ID) CRM_LOG_ASSERT((entry != NULL) && (entry->rsc != NULL) && (entry->rsc->id != NULL)); if ((entry == NULL) || (entry->rsc == NULL) || (entry->rsc->id == NULL)) { continue; } // Entry must have a node unless listing inactive resources CRM_LOG_ASSERT((node_names == NULL) || (entry->node != NULL)); if ((node_names != NULL) && (entry->node == NULL)) { continue; } // Don't add duplicates of a particular clone instance if (pcmk__str_eq(entry->rsc->id, last_rsc_id, pcmk__str_none)) { continue; } last_rsc_id = entry->rsc->id; if (rsc_names != NULL) { pcmk__add_word(rsc_names, 1024, entry->rsc->id); } if ((node_names != NULL) && (entry->node->details->uname != NULL)) { pcmk__add_word(node_names, 1024, entry->node->details->uname); } } // If there are no entries, return "empty" lists if ((rsc_names != NULL) && (*rsc_names == NULL)) { *rsc_names = g_string_new(" "); } if ((node_names != NULL) && (*node_names == NULL)) { *node_names = g_string_new(" "); } return list; } /*! * \internal * \brief Copy a meta-attribute into a notify action * * \param[in] key Name of meta-attribute to copy * \param[in] value Value of meta-attribute to copy * \param[in,out] user_data Notify action to copy into */ static void copy_meta_to_notify(gpointer key, gpointer value, gpointer user_data) { pe_action_t *notify = (pe_action_t *) user_data; /* Any existing meta-attributes (for example, the action timeout) are for * the notify action itself, so don't override those. */ if (g_hash_table_lookup(notify->meta, (const char *) key) != NULL) { return; } g_hash_table_insert(notify->meta, strdup((const char *) key), strdup((const char *) value)); } static void add_notify_data_to_action_meta(const notify_data_t *n_data, pe_action_t *action) { for (const GSList *item = n_data->keys; item; item = item->next) { const pcmk_nvpair_t *nvpair = (const pcmk_nvpair_t *) item->data; add_hash_param(action->meta, nvpair->name, nvpair->value); } } /*! * \internal * \brief Create a new notify pseudo-action for a clone resource * * \param[in,out] rsc Clone resource that notification is for * \param[in] action Action to use in notify action key * \param[in] notif_action RSC_NOTIFY or RSC_NOTIFIED * \param[in] notif_type "pre", "post", "confirmed-pre", "confirmed-post" * * \return Newly created notify pseudo-action */ static pe_action_t * new_notify_pseudo_action(pe_resource_t *rsc, const pe_action_t *action, const char *notif_action, const char *notif_type) { pe_action_t *notify = NULL; notify = custom_action(rsc, pcmk__notify_key(rsc->id, notif_type, action->task), notif_action, NULL, pcmk_is_set(action->flags, pe_action_optional), TRUE, rsc->cluster); pe__set_action_flags(notify, pe_action_pseudo); add_hash_param(notify->meta, "notify_key_type", notif_type); add_hash_param(notify->meta, "notify_key_operation", action->task); return notify; } /*! * \internal * \brief Create a new notify action for a clone instance * * \param[in,out] rsc Clone instance that notification is for * \param[in] node Node that notification is for * \param[in,out] op Action that notification is for * \param[in,out] notify_done Parent pseudo-action for notifications complete * \param[in] n_data Notification values to add to action meta-data * * \return Newly created notify action */ static pe_action_t * new_notify_action(pe_resource_t *rsc, const pe_node_t *node, pe_action_t *op, pe_action_t *notify_done, const notify_data_t *n_data) { char *key = NULL; pe_action_t *notify_action = NULL; const char *value = NULL; const char *task = NULL; const char *skip_reason = NULL; CRM_CHECK((rsc != NULL) && (node != NULL), return NULL); // Ensure we have all the info we need if (op == NULL) { skip_reason = "no action"; } else if (notify_done == NULL) { skip_reason = "no parent notification"; } else if (!node->details->online) { skip_reason = "node offline"; } else if (!pcmk_is_set(op->flags, pe_action_runnable)) { skip_reason = "original action not runnable"; } if (skip_reason != NULL) { pe_rsc_trace(rsc, "Skipping notify action for %s on %s: %s", rsc->id, pe__node_name(node), skip_reason); return NULL; } value = g_hash_table_lookup(op->meta, "notify_type"); // "pre" or "post" task = g_hash_table_lookup(op->meta, "notify_operation"); // original action pe_rsc_trace(rsc, "Creating notify action for %s on %s (%s-%s)", rsc->id, pe__node_name(node), value, task); // Create the notify action key = pcmk__notify_key(rsc->id, value, task); notify_action = custom_action(rsc, key, op->task, node, pcmk_is_set(op->flags, pe_action_optional), TRUE, rsc->cluster); // Add meta-data to notify action g_hash_table_foreach(op->meta, copy_meta_to_notify, notify_action); add_notify_data_to_action_meta(n_data, notify_action); // Order notify after original action and before parent notification order_actions(op, notify_action, pe_order_optional); order_actions(notify_action, notify_done, pe_order_optional); return notify_action; } /*! * \internal * \brief Create a new "post-" notify action for a clone instance * * \param[in,out] rsc Clone instance that notification is for * \param[in] node Node that notification is for * \param[in,out] n_data Notification values to add to action meta-data */ static void new_post_notify_action(pe_resource_t *rsc, const pe_node_t *node, notify_data_t *n_data) { pe_action_t *notify = NULL; CRM_ASSERT(n_data != NULL); // Create the "post-" notify action for specified instance notify = new_notify_action(rsc, node, n_data->post, n_data->post_done, n_data); if (notify != NULL) { notify->priority = INFINITY; } // Order recurring monitors after all "post-" notifications complete if (n_data->post_done == NULL) { return; } for (GList *iter = rsc->actions; iter != NULL; iter = iter->next) { pe_action_t *mon = (pe_action_t *) iter->data; const char *interval_ms_s = NULL; interval_ms_s = g_hash_table_lookup(mon->meta, XML_LRM_ATTR_INTERVAL_MS); if (pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches) || pcmk__str_eq(mon->task, RSC_CANCEL, pcmk__str_none)) { continue; // Not a recurring monitor } order_actions(n_data->post_done, mon, pe_order_optional); } } /*! * \internal * \brief Create and order notification pseudo-actions for a clone action * * In addition to the actual notify actions needed for each clone instance, * clone notifications also require pseudo-actions to provide ordering points * in the notification process. This creates the notification data, along with * appropriate pseudo-actions and their orderings. * * For example, the ordering sequence for starting a clone is: * * "pre-" notify pseudo-action for clone * -> "pre-" notify actions for each clone instance * -> "pre-" notifications complete pseudo-action for clone * -> start actions for each clone instance * -> "started" pseudo-action for clone * -> "post-" notify pseudo-action for clone * -> "post-" notify actions for each clone instance * -> "post-" notifications complete pseudo-action for clone * * \param[in,out] rsc Clone that notifications are for * \param[in] task Name of action that notifications are for * \param[in,out] action If not NULL, create a "pre-" pseudo-action ordered * before a "pre-" complete pseudo-action, ordered * before this action * \param[in,out] complete If not NULL, create a "post-" pseudo-action ordered * after this action, and a "post-" complete * pseudo-action ordered after that * * \return Newly created notification data */ notify_data_t * pe__action_notif_pseudo_ops(pe_resource_t *rsc, const char *task, pe_action_t *action, pe_action_t *complete) { notify_data_t *n_data = NULL; if (!pcmk_is_set(rsc->flags, pe_rsc_notify)) { return NULL; } n_data = calloc(1, sizeof(notify_data_t)); CRM_ASSERT(n_data != NULL); n_data->action = task; if (action != NULL) { // Need "pre-" pseudo-actions // Create "pre-" notify pseudo-action for clone n_data->pre = new_notify_pseudo_action(rsc, action, RSC_NOTIFY, "pre"); pe__set_action_flags(n_data->pre, pe_action_runnable); add_hash_param(n_data->pre->meta, "notify_type", "pre"); add_hash_param(n_data->pre->meta, "notify_operation", n_data->action); // Create "pre-" notifications complete pseudo-action for clone n_data->pre_done = new_notify_pseudo_action(rsc, action, RSC_NOTIFIED, "confirmed-pre"); pe__set_action_flags(n_data->pre_done, pe_action_runnable); add_hash_param(n_data->pre_done->meta, "notify_type", "pre"); add_hash_param(n_data->pre_done->meta, "notify_operation", n_data->action); // Order "pre-" -> "pre-" complete -> original action order_actions(n_data->pre, n_data->pre_done, pe_order_optional); order_actions(n_data->pre_done, action, pe_order_optional); } if (complete != NULL) { // Need "post-" pseudo-actions // Create "post-" notify pseudo-action for clone n_data->post = new_notify_pseudo_action(rsc, complete, RSC_NOTIFY, "post"); n_data->post->priority = INFINITY; if (pcmk_is_set(complete->flags, pe_action_runnable)) { pe__set_action_flags(n_data->post, pe_action_runnable); } else { pe__clear_action_flags(n_data->post, pe_action_runnable); } add_hash_param(n_data->post->meta, "notify_type", "post"); add_hash_param(n_data->post->meta, "notify_operation", n_data->action); // Create "post-" notifications complete pseudo-action for clone n_data->post_done = new_notify_pseudo_action(rsc, complete, RSC_NOTIFIED, "confirmed-post"); n_data->post_done->priority = INFINITY; if (pcmk_is_set(complete->flags, pe_action_runnable)) { pe__set_action_flags(n_data->post_done, pe_action_runnable); } else { pe__clear_action_flags(n_data->post_done, pe_action_runnable); } add_hash_param(n_data->post_done->meta, "notify_type", "post"); add_hash_param(n_data->post_done->meta, "notify_operation", n_data->action); // Order original action complete -> "post-" -> "post-" complete order_actions(complete, n_data->post, pe_order_implies_then); order_actions(n_data->post, n_data->post_done, pe_order_implies_then); } // If we created both, order "pre-" complete -> "post-" if ((action != NULL) && (complete != NULL)) { order_actions(n_data->pre_done, n_data->post, pe_order_optional); } return n_data; } /*! * \internal * \brief Create a new notification entry * * \param[in] rsc Resource for notification * \param[in] node Node for notification * * \return Newly allocated notification entry * \note The caller is responsible for freeing the return value. */ static notify_entry_t * new_notify_entry(const pe_resource_t *rsc, const pe_node_t *node) { notify_entry_t *entry = calloc(1, sizeof(notify_entry_t)); CRM_ASSERT(entry != NULL); entry->rsc = rsc; entry->node = node; return entry; } /*! * \internal * \brief Add notification data for resource state and optionally actions * * \param[in] rsc Clone or clone instance being notified * \param[in] activity Whether to add notification entries for actions * \param[in,out] n_data Notification data for clone */ static void collect_resource_data(const pe_resource_t *rsc, bool activity, notify_data_t *n_data) { const GList *iter = NULL; notify_entry_t *entry = NULL; const pe_node_t *node = NULL; if (n_data == NULL) { return; } if (n_data->allowed_nodes == NULL) { n_data->allowed_nodes = rsc->allowed_nodes; } // If this is a clone, call recursively for each instance if (rsc->children != NULL) { for (iter = rsc->children; iter != NULL; iter = iter->next) { const pe_resource_t *child = (const pe_resource_t *) iter->data; collect_resource_data(child, activity, n_data); } return; } // This is a notification for a single clone instance if (rsc->running_on != NULL) { node = rsc->running_on->data; // First is sufficient } entry = new_notify_entry(rsc, node); // Add notification indicating the resource state switch (rsc->role) { case RSC_ROLE_STOPPED: n_data->inactive = g_list_prepend(n_data->inactive, entry); break; case RSC_ROLE_STARTED: n_data->active = g_list_prepend(n_data->active, entry); break; case RSC_ROLE_UNPROMOTED: n_data->unpromoted = g_list_prepend(n_data->unpromoted, entry); n_data->active = g_list_prepend(n_data->active, dup_notify_entry(entry)); break; case RSC_ROLE_PROMOTED: n_data->promoted = g_list_prepend(n_data->promoted, entry); n_data->active = g_list_prepend(n_data->active, dup_notify_entry(entry)); break; default: crm_err("Resource %s role on %s (%s) is not supported for " "notifications (bug?)", rsc->id, pe__node_name(node), role2text(rsc->role)); free(entry); break; } if (!activity) { return; } // Add notification entries for each of the resource's actions for (iter = rsc->actions; iter != NULL; iter = iter->next) { const pe_action_t *op = (const pe_action_t *) iter->data; if (!pcmk_is_set(op->flags, pe_action_optional) && (op->node != NULL)) { enum action_tasks task = text2task(op->task); if ((task == stop_rsc) && op->node->details->unclean) { // Create anyway (additional noise if node can't be fenced) } else if (!pcmk_is_set(op->flags, pe_action_runnable)) { continue; } entry = new_notify_entry(rsc, op->node); switch (task) { case start_rsc: n_data->start = g_list_prepend(n_data->start, entry); break; case stop_rsc: n_data->stop = g_list_prepend(n_data->stop, entry); break; case action_promote: n_data->promote = g_list_prepend(n_data->promote, entry); break; case action_demote: n_data->demote = g_list_prepend(n_data->demote, entry); break; default: free(entry); break; } } } } // For (char *) value #define add_notify_env(n_data, key, value) do { \ n_data->keys = pcmk_prepend_nvpair(n_data->keys, key, value); \ } while (0) // For (GString *) value #define add_notify_env_gs(n_data, key, value) do { \ n_data->keys = pcmk_prepend_nvpair(n_data->keys, key, \ (const char *) value->str); \ } while (0) // For (GString *) value #define add_notify_env_free_gs(n_data, key, value) do { \ n_data->keys = pcmk_prepend_nvpair(n_data->keys, key, \ (const char *) value->str); \ g_string_free(value, TRUE); value = NULL; \ } while (0) /*! * \internal * \brief Create notification name/value pairs from structured data * * \param[in] rsc Resource that notification is for * \param[in,out] n_data Notification data */ static void add_notif_keys(const pe_resource_t *rsc, notify_data_t *n_data) { bool required = false; // Whether to make notify actions required GString *rsc_list = NULL; GString *node_list = NULL; GString *metal_list = NULL; const char *source = NULL; GList *nodes = NULL; n_data->stop = notify_entries_to_strings(n_data->stop, &rsc_list, &node_list); if ((strcmp(" ", (const char *) rsc_list->str) != 0) && pcmk__str_eq(n_data->action, RSC_STOP, pcmk__str_none)) { required = true; } add_notify_env_free_gs(n_data, "notify_stop_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_stop_uname", node_list); if ((n_data->start != NULL) && pcmk__str_eq(n_data->action, RSC_START, pcmk__str_none)) { required = true; } n_data->start = notify_entries_to_strings(n_data->start, &rsc_list, &node_list); add_notify_env_free_gs(n_data, "notify_start_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_start_uname", node_list); if ((n_data->demote != NULL) && pcmk__str_eq(n_data->action, RSC_DEMOTE, pcmk__str_none)) { required = true; } n_data->demote = notify_entries_to_strings(n_data->demote, &rsc_list, &node_list); add_notify_env_free_gs(n_data, "notify_demote_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_demote_uname", node_list); if ((n_data->promote != NULL) && pcmk__str_eq(n_data->action, RSC_PROMOTE, pcmk__str_none)) { required = true; } n_data->promote = notify_entries_to_strings(n_data->promote, &rsc_list, &node_list); add_notify_env_free_gs(n_data, "notify_promote_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_promote_uname", node_list); n_data->active = notify_entries_to_strings(n_data->active, &rsc_list, &node_list); add_notify_env_free_gs(n_data, "notify_active_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_active_uname", node_list); n_data->unpromoted = notify_entries_to_strings(n_data->unpromoted, &rsc_list, &node_list); add_notify_env_gs(n_data, "notify_unpromoted_resource", rsc_list); add_notify_env_gs(n_data, "notify_unpromoted_uname", node_list); // Deprecated: kept for backward compatibility with older resource agents add_notify_env_free_gs(n_data, "notify_slave_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_slave_uname", node_list); n_data->promoted = notify_entries_to_strings(n_data->promoted, &rsc_list, &node_list); add_notify_env_gs(n_data, "notify_promoted_resource", rsc_list); add_notify_env_gs(n_data, "notify_promoted_uname", node_list); // Deprecated: kept for backward compatibility with older resource agents add_notify_env_free_gs(n_data, "notify_master_resource", rsc_list); add_notify_env_free_gs(n_data, "notify_master_uname", node_list); n_data->inactive = notify_entries_to_strings(n_data->inactive, &rsc_list, NULL); add_notify_env_free_gs(n_data, "notify_inactive_resource", rsc_list); nodes = g_hash_table_get_values(n_data->allowed_nodes); if (!pcmk__is_daemon) { /* For display purposes, sort the node list, for consistent * regression test output (while avoiding the performance hit * for the live cluster). */ nodes = g_list_sort(nodes, pe__cmp_node_name); } get_node_names(nodes, &node_list, NULL); add_notify_env_free_gs(n_data, "notify_available_uname", node_list); g_list_free(nodes); source = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET); if (pcmk__str_eq("host", source, pcmk__str_none)) { get_node_names(rsc->cluster->nodes, &node_list, &metal_list); add_notify_env_free_gs(n_data, "notify_all_hosts", metal_list); } else { get_node_names(rsc->cluster->nodes, &node_list, NULL); } add_notify_env_free_gs(n_data, "notify_all_uname", node_list); if (required && (n_data->pre != NULL)) { pe__clear_action_flags(n_data->pre, pe_action_optional); pe__clear_action_flags(n_data->pre_done, pe_action_optional); } if (required && (n_data->post != NULL)) { pe__clear_action_flags(n_data->post, pe_action_optional); pe__clear_action_flags(n_data->post_done, pe_action_optional); } } /* * \internal * \brief Find any remote connection start relevant to an action * * \param[in] action Action to check * * \return If action is behind a remote connection, connection's start */ static pe_action_t * find_remote_start(pe_action_t *action) { if ((action != NULL) && (action->node != NULL)) { pe_resource_t *remote_rsc = action->node->details->remote_rsc; if (remote_rsc != NULL) { return find_first_action(remote_rsc->actions, NULL, RSC_START, NULL); } } return NULL; } /*! * \internal * \brief Create notify actions, and add notify data to original actions * * \param[in,out] rsc Clone or clone instance that notification is for * \param[in,out] n_data Clone notification data for some action */ static void create_notify_actions(pe_resource_t *rsc, notify_data_t *n_data) { GList *iter = NULL; pe_action_t *stop = NULL; pe_action_t *start = NULL; enum action_tasks task = text2task(n_data->action); // If this is a clone, call recursively for each instance if (rsc->children != NULL) { g_list_foreach(rsc->children, (GFunc) create_notify_actions, n_data); return; } // Add notification meta-attributes to original actions for (iter = rsc->actions; iter != NULL; iter = iter->next) { pe_action_t *op = (pe_action_t *) iter->data; if (!pcmk_is_set(op->flags, pe_action_optional) && (op->node != NULL)) { switch (text2task(op->task)) { case start_rsc: case stop_rsc: case action_promote: case action_demote: add_notify_data_to_action_meta(n_data, op); break; default: break; } } } // Skip notify action itself if original action was not needed switch (task) { case start_rsc: if (n_data->start == NULL) { pe_rsc_trace(rsc, "No notify action needed for %s %s", rsc->id, n_data->action); return; } break; case action_promote: if (n_data->promote == NULL) { pe_rsc_trace(rsc, "No notify action needed for %s %s", rsc->id, n_data->action); return; } break; case action_demote: if (n_data->demote == NULL) { pe_rsc_trace(rsc, "No notify action needed for %s %s", rsc->id, n_data->action); return; } break; default: // We cannot do same for stop because it might be implied by fencing break; } pe_rsc_trace(rsc, "Creating notify actions for %s %s", rsc->id, n_data->action); // Create notify actions for stop or demote if ((rsc->role != RSC_ROLE_STOPPED) && ((task == stop_rsc) || (task == action_demote))) { stop = find_first_action(rsc->actions, NULL, RSC_STOP, NULL); for (iter = rsc->running_on; iter != NULL; iter = iter->next) { pe_node_t *current_node = (pe_node_t *) iter->data; /* If a stop is a pseudo-action implied by fencing, don't try to * notify the node getting fenced. */ if ((stop != NULL) && pcmk_is_set(stop->flags, pe_action_pseudo) && (current_node->details->unclean || current_node->details->remote_requires_reset)) { continue; } new_notify_action(rsc, current_node, n_data->pre, n_data->pre_done, n_data); if ((task == action_demote) || (stop == NULL) || pcmk_is_set(stop->flags, pe_action_optional)) { new_post_notify_action(rsc, current_node, n_data); } } } // Create notify actions for start or promote if ((rsc->next_role != RSC_ROLE_STOPPED) && ((task == start_rsc) || (task == action_promote))) { start = find_first_action(rsc->actions, NULL, RSC_START, NULL); if (start != NULL) { pe_action_t *remote_start = find_remote_start(start); if ((remote_start != NULL) && !pcmk_is_set(remote_start->flags, pe_action_runnable)) { /* Start and promote actions for a clone instance behind * a Pacemaker Remote connection happen after the * connection starts. If the connection start is blocked, do * not schedule notifications for these actions. */ return; } } if (rsc->allocated_to == NULL) { pe_proc_err("Next role '%s' but %s is not allocated", role2text(rsc->next_role), rsc->id); return; } if ((task != start_rsc) || (start == NULL) || pcmk_is_set(start->flags, pe_action_optional)) { new_notify_action(rsc, rsc->allocated_to, n_data->pre, n_data->pre_done, n_data); } new_post_notify_action(rsc, rsc->allocated_to, n_data); } } /*! * \internal * \brief Create notification data and actions for one clone action * * \param[in,out] rsc Clone resource that notification is for * \param[in,out] n_data Clone notification data for some action */ void pe__create_action_notifications(pe_resource_t *rsc, notify_data_t *n_data) { if ((rsc == NULL) || (n_data == NULL)) { return; } collect_resource_data(rsc, true, n_data); add_notif_keys(rsc, n_data); create_notify_actions(rsc, n_data); } /*! * \internal * \brief Free notification data for one action * * \param[in,out] n_data Notification data to free */ void pe__free_action_notification_data(notify_data_t *n_data) { if (n_data == NULL) { return; } g_list_free_full(n_data->stop, free); g_list_free_full(n_data->start, free); g_list_free_full(n_data->demote, free); g_list_free_full(n_data->promote, free); g_list_free_full(n_data->promoted, free); g_list_free_full(n_data->unpromoted, free); g_list_free_full(n_data->active, free); g_list_free_full(n_data->inactive, free); pcmk_free_nvpairs(n_data->keys); free(n_data); } /*! * \internal * \brief Order clone "notifications complete" pseudo-action after fencing * * If a stop action is implied by fencing, the usual notification pseudo-actions * will not be sufficient to order things properly, or even create all needed * notifications if the clone is also stopping on another node, and another * clone is ordered after it. This function creates new notification * pseudo-actions relative to the fencing to ensure everything works properly. * * \param[in] stop Stop action implied by fencing * \param[in,out] rsc Clone resource that notification is for * \param[in,out] stonith_op Fencing action that implies \p stop */ void pe__order_notifs_after_fencing(const pe_action_t *stop, pe_resource_t *rsc, pe_action_t *stonith_op) { notify_data_t *n_data; crm_info("Ordering notifications for implied %s after fencing", stop->uuid); n_data = pe__action_notif_pseudo_ops(rsc, RSC_STOP, NULL, stonith_op); if (n_data != NULL) { collect_resource_data(rsc, false, n_data); add_notify_env(n_data, "notify_stop_resource", rsc->id); add_notify_env(n_data, "notify_stop_uname", stop->node->details->uname); create_notify_actions(uber_parent(rsc), n_data); pe__free_action_notification_data(n_data); } } diff --git a/lib/pengine/pe_status_private.h b/lib/pengine/pe_status_private.h index e0acc8f74e..017b4bf46a 100644 --- a/lib/pengine/pe_status_private.h +++ b/lib/pengine/pe_status_private.h @@ -1,141 +1,146 @@ /* * Copyright 2018-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PE_STATUS_PRIVATE__H # define PE_STATUS_PRIVATE__H /* This header is for the sole use of libpe_status, so that functions can be * declared with G_GNUC_INTERNAL for efficiency. */ #if defined(PCMK__UNIT_TESTING) #undef G_GNUC_INTERNAL #define G_GNUC_INTERNAL #endif +#include // GSList, GList, GHashTable +#include // xmlNode + +#include // pe_action_t, pe_resource_t, etc. + /*! * \internal * \deprecated This macro will be removed in a future release */ # define status_print(fmt, args...) \ if(options & pe_print_html) { \ FILE *stream = print_data; \ fprintf(stream, fmt, ##args); \ } else if(options & pe_print_printf || options & pe_print_ncurses) { \ FILE *stream = print_data; \ fprintf(stream, fmt, ##args); \ } else if(options & pe_print_xml) { \ FILE *stream = print_data; \ fprintf(stream, fmt, ##args); \ } else if(options & pe_print_log) { \ int log_level = *(int*)print_data; \ do_crm_log(log_level, fmt, ##args); \ } typedef struct notify_data_s { GSList *keys; // Environment variable name/value pairs const char *action; pe_action_t *pre; pe_action_t *post; pe_action_t *pre_done; pe_action_t *post_done; GList *active; /* notify_entry_t* */ GList *inactive; /* notify_entry_t* */ GList *start; /* notify_entry_t* */ GList *stop; /* notify_entry_t* */ GList *demote; /* notify_entry_t* */ GList *promote; /* notify_entry_t* */ GList *promoted; /* notify_entry_t* */ GList *unpromoted; /* notify_entry_t* */ GHashTable *allowed_nodes; } notify_data_t; G_GNUC_INTERNAL pe_resource_t *pe__create_clone_child(pe_resource_t *rsc, pe_working_set_t *data_set); G_GNUC_INTERNAL void pe__create_action_notifications(pe_resource_t *rsc, notify_data_t *n_data); G_GNUC_INTERNAL void pe__free_action_notification_data(notify_data_t *n_data); G_GNUC_INTERNAL notify_data_t *pe__action_notif_pseudo_ops(pe_resource_t *rsc, const char *task, pe_action_t *action, pe_action_t *complete); G_GNUC_INTERNAL void pe__force_anon(const char *standard, pe_resource_t *rsc, const char *rid, pe_working_set_t *data_set); G_GNUC_INTERNAL gint pe__cmp_rsc_priority(gconstpointer a, gconstpointer b); G_GNUC_INTERNAL gboolean pe__unpack_resource(xmlNode *xml_obj, pe_resource_t **rsc, pe_resource_t *parent, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_remote_nodes(xmlNode *xml_resources, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_resources(const xmlNode *xml_resources, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_config(xmlNode *config, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_nodes(xmlNode *xml_nodes, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_tags(xmlNode *xml_tags, pe_working_set_t *data_set); G_GNUC_INTERNAL gboolean unpack_status(xmlNode *status, pe_working_set_t *data_set); G_GNUC_INTERNAL op_digest_cache_t *pe__compare_fencing_digest(pe_resource_t *rsc, const char *agent, pe_node_t *node, pe_working_set_t *data_set); G_GNUC_INTERNAL void pe__unpack_node_health_scores(pe_working_set_t *data_set); // Primitive resource methods G_GNUC_INTERNAL unsigned int pe__primitive_max_per_node(const pe_resource_t *rsc); // Group resource methods G_GNUC_INTERNAL unsigned int pe__group_max_per_node(const pe_resource_t *rsc); // Clone resource methods G_GNUC_INTERNAL unsigned int pe__clone_max_per_node(const pe_resource_t *rsc); // Bundle resource methods G_GNUC_INTERNAL pe_node_t *pe__bundle_active_node(const pe_resource_t *rsc, unsigned int *count_all, unsigned int *count_clean); G_GNUC_INTERNAL unsigned int pe__bundle_max_per_node(const pe_resource_t *rsc); #endif // PE_STATUS_PRIVATE__H diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c index ef0a092dc1..8fb8a3c651 100644 --- a/lib/pengine/utils.c +++ b/lib/pengine/utils.c @@ -1,938 +1,938 @@ /* * Copyright 2004-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include "pe_status_private.h" extern bool pcmk__is_daemon; gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data); /*! * \internal * \brief Check whether we can fence a particular node * * \param[in] data_set Working set for cluster * \param[in] node Name of node to check * * \return true if node can be fenced, false otherwise */ bool pe_can_fence(const pe_working_set_t *data_set, const pe_node_t *node) { if (pe__is_guest_node(node)) { /* Guest nodes are fenced by stopping their container resource. We can * do that if the container's host is either online or fenceable. */ pe_resource_t *rsc = node->details->remote_rsc->container; for (GList *n = rsc->running_on; n != NULL; n = n->next) { pe_node_t *container_node = n->data; if (!container_node->details->online && !pe_can_fence(data_set, container_node)) { return false; } } return true; } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { return false; /* Turned off */ } else if (!pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { return false; /* No devices */ } else if (pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { return true; } else if (data_set->no_quorum_policy == no_quorum_ignore) { return true; } else if(node == NULL) { return false; } else if(node->details->online) { crm_notice("We can fence %s without quorum because they're in our membership", pe__node_name(node)); return true; } crm_trace("Cannot fence %s", pe__node_name(node)); return false; } /*! * \internal * \brief Copy a node object * * \param[in] this_node Node object to copy * * \return Newly allocated shallow copy of this_node * \note This function asserts on errors and is guaranteed to return non-NULL. */ pe_node_t * pe__copy_node(const pe_node_t *this_node) { pe_node_t *new_node = NULL; CRM_ASSERT(this_node != NULL); new_node = calloc(1, sizeof(pe_node_t)); CRM_ASSERT(new_node != NULL); new_node->rsc_discover_mode = this_node->rsc_discover_mode; new_node->weight = this_node->weight; new_node->fixed = this_node->fixed; // @COMPAT deprecated and unused new_node->details = this_node->details; return new_node; } /* any node in list1 or list2 and not in the other gets a score of -INFINITY */ void node_list_exclude(GHashTable * hash, GList *list, gboolean merge_scores) { GHashTable *result = hash; pe_node_t *other_node = NULL; GList *gIter = list; GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { other_node = pe_find_node_id(list, node->details->id); if (other_node == NULL) { node->weight = -INFINITY; crm_trace("Banning dependent from %s (no primary instance)", pe__node_name(node)); } else if (merge_scores) { node->weight = pcmk__add_scores(node->weight, other_node->weight); crm_trace("Added primary's score %s to dependent's score for %s " "(now %s)", pcmk_readable_score(other_node->weight), pe__node_name(node), pcmk_readable_score(node->weight)); } } for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; other_node = pe_hash_table_lookup(result, node->details->id); if (other_node == NULL) { pe_node_t *new_node = pe__copy_node(node); new_node->weight = -INFINITY; g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } } } /*! * \internal * \brief Create a node hash table from a node list * * \param[in] list Node list * * \return Hash table equivalent of node list */ GHashTable * pe__node_list2table(const GList *list) { GHashTable *result = NULL; result = pcmk__strkey_table(NULL, free); for (const GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = pe__copy_node((const pe_node_t *) gIter->data); g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } return result; } /*! * \internal * \brief Compare two nodes by name, with numeric portions sorted numerically * * Sort two node names case-insensitively like strcasecmp(), but with any * numeric portions of the name sorted numerically. For example, "node10" will * sort higher than "node9" but lower than "remotenode9". * * \param[in] a First node to compare (can be \c NULL) * \param[in] b Second node to compare (can be \c NULL) * * \retval -1 \c a comes before \c b (or \c a is \c NULL and \c b is not) * \retval 0 \c a and \c b are equal (or both are \c NULL) * \retval 1 \c a comes after \c b (or \c b is \c NULL and \c a is not) */ gint pe__cmp_node_name(gconstpointer a, gconstpointer b) { const pe_node_t *node1 = (const pe_node_t *) a; const pe_node_t *node2 = (const pe_node_t *) b; if ((node1 == NULL) && (node2 == NULL)) { return 0; } if (node1 == NULL) { return -1; } if (node2 == NULL) { return 1; } return pcmk__numeric_strcasecmp(node1->details->uname, node2->details->uname); } /*! * \internal * \brief Output node weights to stdout * * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes If rsc is not specified, use these nodes * \param[in,out] data_set Cluster working set */ static void pe__output_node_weights(const pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set) { pcmk__output_t *out = data_set->priv; // Sort the nodes so the output is consistent for regression tests GList *list = g_list_sort(g_hash_table_get_values(nodes), pe__cmp_node_name); for (const GList *gIter = list; gIter != NULL; gIter = gIter->next) { const pe_node_t *node = (const pe_node_t *) gIter->data; out->message(out, "node-weight", rsc, comment, node->details->uname, pcmk_readable_score(node->weight)); } g_list_free(list); } /*! * \internal * \brief Log node weights at trace level * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] rsc If not NULL, include this resource's ID in logs * \param[in] comment Text description to prefix lines with * \param[in] nodes Nodes whose scores should be logged */ static void pe__log_node_weights(const char *file, const char *function, int line, const pe_resource_t *rsc, const char *comment, GHashTable *nodes) { GHashTableIter iter; pe_node_t *node = NULL; // Don't waste time if we're not tracing at this point pcmk__if_tracing({}, return); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { if (rsc) { qb_log_from_external_source(function, file, "%s: %s allocation score on %s: %s", LOG_TRACE, line, 0, comment, rsc->id, pe__node_name(node), pcmk_readable_score(node->weight)); } else { qb_log_from_external_source(function, file, "%s: %s = %s", LOG_TRACE, line, 0, comment, pe__node_name(node), pcmk_readable_score(node->weight)); } } } /*! * \internal * \brief Log or output node weights * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] to_log Log if true, otherwise output * \param[in] rsc If not NULL, use this resource's ID in logs, * and show scores recursively for any children * \param[in] comment Text description to prefix lines with * \param[in] nodes Nodes whose scores should be shown * \param[in,out] data_set Cluster working set */ void -pe__show_node_weights_as(const char *file, const char *function, int line, - bool to_log, const pe_resource_t *rsc, - const char *comment, GHashTable *nodes, - pe_working_set_t *data_set) +pe__show_node_scores_as(const char *file, const char *function, int line, + bool to_log, const pe_resource_t *rsc, + const char *comment, GHashTable *nodes, + pe_working_set_t *data_set) { if (rsc != NULL && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { // Don't show allocation scores for orphans return; } if (nodes == NULL) { // Nothing to show return; } if (to_log) { pe__log_node_weights(file, function, line, rsc, comment, nodes); } else { pe__output_node_weights(rsc, comment, nodes, data_set); } // If this resource has children, repeat recursively for each if (rsc && rsc->children) { for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; - pe__show_node_weights_as(file, function, line, to_log, child, - comment, child->allowed_nodes, data_set); + pe__show_node_scores_as(file, function, line, to_log, child, + comment, child->allowed_nodes, data_set); } } } /*! * \internal * \brief Compare two resources by priority * * \param[in] a First resource to compare (can be \c NULL) * \param[in] b Second resource to compare (can be \c NULL) * * \retval -1 \c a->priority > \c b->priority (or \c b is \c NULL and \c a is * not) * \retval 0 \c a->priority == \c b->priority (or both \c a and \c b are * \c NULL) * \retval 1 \c a->priority < \c b->priority (or \c a is \c NULL and \c b is * not) */ gint pe__cmp_rsc_priority(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->priority > resource2->priority) { return -1; } if (resource1->priority < resource2->priority) { return 1; } return 0; } static void resource_node_score(pe_resource_t *rsc, const pe_node_t *node, int score, const char *tag) { pe_node_t *match = NULL; if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never)) && pcmk__str_eq(tag, "symmetric_default", pcmk__str_casei)) { /* This string comparision may be fragile, but exclusive resources and * exclusive nodes should not have the symmetric_default constraint * applied to them. */ return; } else if (rsc->children) { GList *gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; resource_node_score(child_rsc, node, score, tag); } } match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match == NULL) { match = pe__copy_node(node); g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match); } match->weight = pcmk__add_scores(match->weight, score); pe_rsc_trace(rsc, "Enabling %s preference (%s) for %s on %s (now %s)", tag, pcmk_readable_score(score), rsc->id, pe__node_name(node), pcmk_readable_score(match->weight)); } void resource_location(pe_resource_t *rsc, const pe_node_t *node, int score, const char *tag, pe_working_set_t *data_set) { if (node != NULL) { resource_node_score(rsc, node, score, tag); } else if (data_set != NULL) { GList *gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node_iter = (pe_node_t *) gIter->data; resource_node_score(rsc, node_iter, score, tag); } } else { GHashTableIter iter; pe_node_t *node_iter = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) { resource_node_score(rsc, node_iter, score, tag); } } if (node == NULL && score == -INFINITY) { if (rsc->allocated_to) { crm_info("Deallocating %s from %s", rsc->id, pe__node_name(rsc->allocated_to)); free(rsc->allocated_to); rsc->allocated_to = NULL; } } } time_t get_effective_time(pe_working_set_t * data_set) { if(data_set) { if (data_set->now == NULL) { crm_trace("Recording a new 'now'"); data_set->now = crm_time_new(NULL); } return crm_time_get_seconds_since_epoch(data_set->now); } crm_trace("Defaulting to 'now'"); return time(NULL); } gboolean get_target_role(const pe_resource_t *rsc, enum rsc_role_e *role) { enum rsc_role_e local_role = RSC_ROLE_UNKNOWN; const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); CRM_CHECK(role != NULL, return FALSE); if (pcmk__str_eq(value, "started", pcmk__str_null_matches | pcmk__str_casei) || pcmk__str_eq("default", value, pcmk__str_casei)) { return FALSE; } local_role = text2role(value); if (local_role == RSC_ROLE_UNKNOWN) { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' is not valid", rsc->id, value); return FALSE; } else if (local_role > RSC_ROLE_STARTED) { if (pcmk_is_set(pe__const_top_resource(rsc, false)->flags, pe_rsc_promotable)) { if (local_role > RSC_ROLE_UNPROMOTED) { /* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */ return FALSE; } } else { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' only makes sense for promotable " "clones", rsc->id, value); return FALSE; } } *role = local_role; return TRUE; } gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order) { GList *gIter = NULL; pe_action_wrapper_t *wrapper = NULL; GList *list = NULL; if (order == pe_order_none) { return FALSE; } if (lh_action == NULL || rh_action == NULL) { return FALSE; } crm_trace("Creating action wrappers for ordering: %s then %s", lh_action->uuid, rh_action->uuid); /* Ensure we never create a dependency on ourselves... it's happened */ CRM_ASSERT(lh_action != rh_action); /* Filter dups, otherwise update_action_states() has too much work to do */ gIter = lh_action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data; if (after->action == rh_action && (after->type & order)) { return FALSE; } } wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = rh_action; wrapper->type = order; list = lh_action->actions_after; list = g_list_prepend(list, wrapper); lh_action->actions_after = list; wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = lh_action; wrapper->type = order; list = rh_action->actions_before; list = g_list_prepend(list, wrapper); rh_action->actions_before = list; return TRUE; } void destroy_ticket(gpointer data) { pe_ticket_t *ticket = data; if (ticket->state) { g_hash_table_destroy(ticket->state); } free(ticket->id); free(ticket); } pe_ticket_t * ticket_new(const char *ticket_id, pe_working_set_t * data_set) { pe_ticket_t *ticket = NULL; if (pcmk__str_empty(ticket_id)) { return NULL; } if (data_set->tickets == NULL) { data_set->tickets = pcmk__strkey_table(free, destroy_ticket); } ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = calloc(1, sizeof(pe_ticket_t)); if (ticket == NULL) { crm_err("Cannot allocate ticket '%s'", ticket_id); return NULL; } crm_trace("Creaing ticket entry for %s", ticket_id); ticket->id = strdup(ticket_id); ticket->granted = FALSE; ticket->last_granted = -1; ticket->standby = FALSE; ticket->state = pcmk__strkey_table(free, free); g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket); } return ticket; } const char * rsc_printable_id(const pe_resource_t *rsc) { return pcmk_is_set(rsc->flags, pe_rsc_unique)? rsc->id : ID(rsc->xml); } void pe__clear_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__clear_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__clear_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } void pe__clear_resource_flags_on_all(pe_working_set_t *data_set, uint64_t flag) { for (GList *lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; pe__clear_resource_flags_recursive(r, flag); } } void pe__set_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__set_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__set_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } void trigger_unfencing(pe_resource_t *rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t *data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { /* No resources require it */ return; } else if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_fence_device)) { /* Wasn't a stonith device */ return; } else if(node && node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set); if(dependency) { order_actions(unfence, dependency, pe_order_optional); } } else if(rsc) { GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { trigger_unfencing(rsc, node, reason, dependency, data_set); } } } } gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref) { pe_tag_t *tag = NULL; GList *gIter = NULL; gboolean is_existing = FALSE; CRM_CHECK(tags && tag_name && obj_ref, return FALSE); tag = g_hash_table_lookup(tags, tag_name); if (tag == NULL) { tag = calloc(1, sizeof(pe_tag_t)); if (tag == NULL) { return FALSE; } tag->id = strdup(tag_name); tag->refs = NULL; g_hash_table_insert(tags, strdup(tag_name), tag); } for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *existing_ref = (const char *) gIter->data; if (pcmk__str_eq(existing_ref, obj_ref, pcmk__str_none)){ is_existing = TRUE; break; } } if (is_existing == FALSE) { tag->refs = g_list_append(tag->refs, strdup(obj_ref)); crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref); } return TRUE; } /*! * \internal * \brief Check whether shutdown has been requested for a node * * \param[in] node Node to check * * \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise * \note This differs from simply using node->details->shutdown in that it can * be used before that has been determined (and in fact to determine it), * and it can also be used to distinguish requested shutdown from implicit * shutdown of remote nodes by virtue of their connection stopping. */ bool pe__shutdown_requested(const pe_node_t *node) { const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN); return !pcmk__str_eq(shutdown, "0", pcmk__str_null_matches); } /*! * \internal * \brief Update a data set's "recheck by" time * * \param[in] recheck Epoch time when recheck should happen * \param[in,out] data_set Current working set */ void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set) { if ((recheck > get_effective_time(data_set)) && ((data_set->recheck_by == 0) || (data_set->recheck_by > recheck))) { data_set->recheck_by = recheck; } } /*! * \internal * \brief Extract nvpair blocks contained by a CIB XML element into a hash table * * \param[in] xml_obj XML element containing blocks of nvpair elements * \param[in] set_name If not NULL, only use blocks of this element * \param[in] rule_data Matching parameters to use when unpacking * \param[out] hash Where to store extracted name/value pairs * \param[in] always_first If not NULL, process block with this ID first * \param[in] overwrite Whether to replace existing values with same name * \param[in,out] data_set Cluster working set containing \p xml_obj */ void pe__unpack_dataset_nvpairs(const xmlNode *xml_obj, const char *set_name, const pe_rule_eval_data_t *rule_data, GHashTable *hash, const char *always_first, gboolean overwrite, pe_working_set_t *data_set) { crm_time_t *next_change = crm_time_new_undefined(); pe_eval_nvpairs(data_set->input, xml_obj, set_name, rule_data, hash, always_first, overwrite, next_change); if (crm_time_is_defined(next_change)) { time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(recheck, data_set); } crm_time_free(next_change); } bool pe__resource_is_disabled(const pe_resource_t *rsc) { const char *target_role = NULL; CRM_CHECK(rsc != NULL, return false); target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); if (target_role) { enum rsc_role_e target_role_e = text2role(target_role); if ((target_role_e == RSC_ROLE_STOPPED) || ((target_role_e == RSC_ROLE_UNPROMOTED) && pcmk_is_set(pe__const_top_resource(rsc, false)->flags, pe_rsc_promotable))) { return true; } } return false; } /*! * \internal * \brief Check whether a resource is running only on given node * * \param[in] rsc Resource to check * \param[in] node Node to check * * \return true if \p rsc is running only on \p node, otherwise false */ bool pe__rsc_running_on_only(const pe_resource_t *rsc, const pe_node_t *node) { return (rsc != NULL) && pcmk__list_of_1(rsc->running_on) && pe__same_node((const pe_node_t *) rsc->running_on->data, node); } bool pe__rsc_running_on_any(pe_resource_t *rsc, GList *node_list) { for (GList *ele = rsc->running_on; ele; ele = ele->next) { pe_node_t *node = (pe_node_t *) ele->data; if (pcmk__str_in_list(node->details->uname, node_list, pcmk__str_star_matches|pcmk__str_casei)) { return true; } } return false; } bool pcmk__rsc_filtered_by_node(pe_resource_t *rsc, GList *only_node) { return (rsc->fns->active(rsc, FALSE) && !pe__rsc_running_on_any(rsc, only_node)); } GList * pe__filter_rsc_list(GList *rscs, GList *filter) { GList *retval = NULL; for (GList *gIter = rscs; gIter; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; /* I think the second condition is safe here for all callers of this * function. If not, it needs to move into pe__node_text. */ if (pcmk__str_in_list(rsc_printable_id(rsc), filter, pcmk__str_star_matches) || (rsc->parent && pcmk__str_in_list(rsc_printable_id(rsc->parent), filter, pcmk__str_star_matches))) { retval = g_list_prepend(retval, rsc); } } return retval; } GList * pe__build_node_name_list(pe_working_set_t *data_set, const char *s) { GList *nodes = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { /* Nothing was given so return a list of all node names. Or, '*' was * given. This would normally fall into the pe__unames_with_tag branch * where it will return an empty list. Catch it here instead. */ nodes = g_list_prepend(nodes, strdup("*")); } else { pe_node_t *node = pe_find_node(data_set->nodes, s); if (node) { /* The given string was a valid uname for a node. Return a * singleton list containing just that uname. */ nodes = g_list_prepend(nodes, strdup(s)); } else { /* The given string was not a valid uname. It's either a tag or * it's a typo or something. In the first case, we'll return a * list of all the unames of the nodes with the given tag. In the * second case, we'll return a NULL pointer and nothing will * get displayed. */ nodes = pe__unames_with_tag(data_set, s); } } return nodes; } GList * pe__build_rsc_list(pe_working_set_t *data_set, const char *s) { GList *resources = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { resources = g_list_prepend(resources, strdup("*")); } else { pe_resource_t *rsc = pe_find_resource_with_flags(data_set->resources, s, pe_find_renamed|pe_find_any); if (rsc) { /* A colon in the name we were given means we're being asked to filter * on a specific instance of a cloned resource. Put that exact string * into the filter list. Otherwise, use the printable ID of whatever * resource was found that matches what was asked for. */ if (strstr(s, ":") != NULL) { resources = g_list_prepend(resources, strdup(rsc->id)); } else { resources = g_list_prepend(resources, strdup(rsc_printable_id(rsc))); } } else { /* The given string was not a valid resource name. It's a tag or a * typo or something. See pe__build_node_name_list() for more * detail. */ resources = pe__rscs_with_tag(data_set, s); } } return resources; } xmlNode * pe__failed_probe_for_rsc(const pe_resource_t *rsc, const char *name) { const pe_resource_t *parent = pe__const_top_resource(rsc, false); const char *rsc_id = rsc->id; if (parent->variant == pe_clone) { rsc_id = pe__clone_child_id(parent); } for (xmlNode *xml_op = pcmk__xml_first_child(rsc->cluster->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { const char *value = NULL; char *op_id = NULL; /* This resource operation is not a failed probe. */ if (!pcmk_xe_mask_probe_failure(xml_op)) { continue; } /* This resource operation was not run on the given node. Note that if name is * NULL, this will always succeed. */ value = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if (value == NULL || !pcmk__str_eq(value, name, pcmk__str_casei|pcmk__str_null_matches)) { continue; } if (!parse_op_key(pe__xe_history_key(xml_op), &op_id, NULL, NULL)) { continue; // This history entry is missing an operation key } /* This resource operation's ID does not match the rsc_id we are looking for. */ if (!pcmk__str_eq(op_id, rsc_id, pcmk__str_none)) { free(op_id); continue; } free(op_id); return xml_op; } return NULL; } diff --git a/tools/crm_simulate.c b/tools/crm_simulate.c index 932c5bd229..7e12b49a42 100644 --- a/tools/crm_simulate.c +++ b/tools/crm_simulate.c @@ -1,587 +1,588 @@ /* * Copyright 2009-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #define SUMMARY "crm_simulate - simulate a Pacemaker cluster's response to events" struct { char *dot_file; char *graph_file; gchar *input_file; pcmk_injections_t *injections; unsigned int flags; gchar *output_file; long long repeat; gboolean store; gchar *test_dir; char *use_date; char *xml_file; } options = { .flags = pcmk_sim_show_pending | pcmk_sim_sanitized, .repeat = 1 }; uint32_t section_opts = 0; char *temp_shadow = NULL; crm_exit_t exit_code = CRM_EX_OK; #define INDENT " " static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static gboolean all_actions_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_all_actions; return TRUE; } static gboolean attrs_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { section_opts |= pcmk_section_attributes; return TRUE; } static gboolean failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { section_opts |= pcmk_section_failcounts | pcmk_section_failures; return TRUE; } static gboolean in_place_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.store = TRUE; options.flags |= pcmk_sim_process | pcmk_sim_simulate; return TRUE; } static gboolean live_check_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.xml_file) { free(options.xml_file); } options.xml_file = NULL; options.flags &= ~pcmk_sim_sanitized; return TRUE; } static gboolean node_down_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->node_down = g_list_append(options.injections->node_down, g_strdup(optarg)); return TRUE; } static gboolean node_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->node_fail = g_list_append(options.injections->node_fail, g_strdup(optarg)); return TRUE; } static gboolean node_up_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__simulate_node_config = true; options.injections->node_up = g_list_append(options.injections->node_up, g_strdup(optarg)); return TRUE; } static gboolean op_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process | pcmk_sim_simulate; options.injections->op_fail = g_list_append(options.injections->op_fail, g_strdup(optarg)); return TRUE; } static gboolean op_inject_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->op_inject = g_list_append(options.injections->op_inject, g_strdup(optarg)); return TRUE; } static gboolean pending_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_show_pending; return TRUE; } static gboolean process_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process; return TRUE; } static gboolean quorum_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__str_update(&options.injections->quorum, optarg); return TRUE; } static gboolean save_dotfile_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process; pcmk__str_update(&options.dot_file, optarg); return TRUE; } static gboolean save_graph_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process; pcmk__str_update(&options.graph_file, optarg); return TRUE; } static gboolean show_scores_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process | pcmk_sim_show_scores; return TRUE; } static gboolean simulate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process | pcmk_sim_simulate; return TRUE; } static gboolean ticket_activate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->ticket_activate = g_list_append(options.injections->ticket_activate, g_strdup(optarg)); return TRUE; } static gboolean ticket_grant_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->ticket_grant = g_list_append(options.injections->ticket_grant, g_strdup(optarg)); return TRUE; } static gboolean ticket_revoke_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->ticket_revoke = g_list_append(options.injections->ticket_revoke, g_strdup(optarg)); return TRUE; } static gboolean ticket_standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.injections->ticket_standby = g_list_append(options.injections->ticket_standby, g_strdup(optarg)); return TRUE; } static gboolean utilization_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.flags |= pcmk_sim_process | pcmk_sim_show_utilization; return TRUE; } static gboolean watchdog_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__str_update(&options.injections->watchdog, optarg); return TRUE; } static gboolean xml_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__str_update(&options.xml_file, optarg); options.flags |= pcmk_sim_sanitized; return TRUE; } static gboolean xml_pipe_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { pcmk__str_update(&options.xml_file, "-"); options.flags |= pcmk_sim_sanitized; return TRUE; } static GOptionEntry operation_entries[] = { { "run", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, process_cb, "Process the supplied input and show what actions the cluster will take in response", NULL }, { "simulate", 'S', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, simulate_cb, "Like --run, but also simulate taking those actions and show the resulting new status", NULL }, { "in-place", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, in_place_cb, "Like --simulate, but also store the results back to the input file", NULL }, { "show-attrs", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attrs_cb, "Show node attributes", NULL }, { "show-failcounts", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, failcounts_cb, "Show resource fail counts", NULL }, { "show-scores", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_scores_cb, "Show allocation scores", NULL }, { "show-utilization", 'U', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, utilization_cb, "Show utilization information", NULL }, { "profile", 'P', 0, G_OPTION_ARG_FILENAME, &options.test_dir, "Process all the XML files in the named directory to create profiling data", "DIR" }, { "repeat", 'N', 0, G_OPTION_ARG_INT, &options.repeat, "With --profile, repeat each test N times and print timings", "N" }, /* Deprecated */ { "pending", 'j', G_OPTION_FLAG_NO_ARG|G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, pending_cb, "Display pending state if 'record-pending' is enabled", NULL }, { NULL } }; static GOptionEntry synthetic_entries[] = { { "node-up", 'u', 0, G_OPTION_ARG_CALLBACK, node_up_cb, "Simulate bringing a node online", "NODE" }, { "node-down", 'd', 0, G_OPTION_ARG_CALLBACK, node_down_cb, "Simulate taking a node offline", "NODE" }, { "node-fail", 'f', 0, G_OPTION_ARG_CALLBACK, node_fail_cb, "Simulate a node failing", "NODE" }, { "op-inject", 'i', 0, G_OPTION_ARG_CALLBACK, op_inject_cb, "Generate a failure for the cluster to react to in the simulation.\n" INDENT "See `Operation Specification` help for more information.", "OPSPEC" }, { "op-fail", 'F', 0, G_OPTION_ARG_CALLBACK, op_fail_cb, "If the specified task occurs during the simulation, have it fail with return code ${rc}.\n" INDENT "The transition will normally stop at the failed action.\n" INDENT "Save the result with --save-output and re-run with --xml-file.\n" INDENT "See `Operation Specification` help for more information.", "OPSPEC" }, { "set-datetime", 't', 0, G_OPTION_ARG_STRING, &options.use_date, "Set date/time (ISO 8601 format, see https://en.wikipedia.org/wiki/ISO_8601)", "DATETIME" }, { "quorum", 'q', 0, G_OPTION_ARG_CALLBACK, quorum_cb, "Set to '1' (or 'true') to indicate cluster has quorum", "QUORUM" }, { "watchdog", 'w', 0, G_OPTION_ARG_CALLBACK, watchdog_cb, "Set to '1' (or 'true') to indicate cluster has an active watchdog device", "DEVICE" }, { "ticket-grant", 'g', 0, G_OPTION_ARG_CALLBACK, ticket_grant_cb, "Simulate granting a ticket", "TICKET" }, { "ticket-revoke", 'r', 0, G_OPTION_ARG_CALLBACK, ticket_revoke_cb, "Simulate revoking a ticket", "TICKET" }, { "ticket-standby", 'b', 0, G_OPTION_ARG_CALLBACK, ticket_standby_cb, "Simulate making a ticket standby", "TICKET" }, { "ticket-activate", 'e', 0, G_OPTION_ARG_CALLBACK, ticket_activate_cb, "Simulate activating a ticket", "TICKET" }, { NULL } }; static GOptionEntry artifact_entries[] = { { "save-input", 'I', 0, G_OPTION_ARG_FILENAME, &options.input_file, "Save the input configuration to the named file", "FILE" }, { "save-output", 'O', 0, G_OPTION_ARG_FILENAME, &options.output_file, "Save the output configuration to the named file", "FILE" }, { "save-graph", 'G', 0, G_OPTION_ARG_CALLBACK, save_graph_cb, "Save the transition graph (XML format) to the named file", "FILE" }, { "save-dotfile", 'D', 0, G_OPTION_ARG_CALLBACK, save_dotfile_cb, "Save the transition graph (DOT format) to the named file", "FILE" }, { "all-actions", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, all_actions_cb, "Display all possible actions in DOT graph (even if not part of transition)", NULL }, { NULL } }; static GOptionEntry source_entries[] = { { "live-check", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, live_check_cb, "Connect to CIB manager and use the current CIB contents as input", NULL }, { "xml-file", 'x', 0, G_OPTION_ARG_CALLBACK, xml_file_cb, "Retrieve XML from the named file", "FILE" }, { "xml-pipe", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, xml_pipe_cb, "Retrieve XML from stdin", NULL }, { NULL } }; static int setup_input(pcmk__output_t *out, const char *input, const char *output, GError **error) { int rc = pcmk_rc_ok; xmlNode *cib_object = NULL; char *local_output = NULL; if (input == NULL) { /* Use live CIB */ rc = cib__signon_query(out, NULL, &cib_object); if (rc != pcmk_rc_ok) { // cib__signon_query() outputs any relevant error return rc; } } else if (pcmk__str_eq(input, "-", pcmk__str_casei)) { cib_object = filename2xml(NULL); } else { cib_object = filename2xml(input); } if (pcmk_find_cib_element(cib_object, XML_CIB_TAG_STATUS) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return pcmk_rc_transform_failed; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return pcmk_rc_schema_validation; } if (output == NULL) { char *pid = pcmk__getpid_s(); local_output = get_shadow_file(pid); temp_shadow = strdup(local_output); output = local_output; free(pid); } rc = write_xml_file(cib_object, output, FALSE); free_xml(cib_object); cib_object = NULL; if (rc < 0) { rc = pcmk_legacy2rc(rc); g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_CANTCREAT, "Could not create '%s': %s", output, pcmk_rc_str(rc)); return rc; } else { setenv("CIB_file", output, 1); free(local_output); return pcmk_rc_ok; } } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; GOptionEntry extra_prog_entries[] = { { "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet), "Display only essential output", NULL }, { NULL } }; const char *description = "Operation Specification:\n\n" "The OPSPEC in any command line option is of the form\n" "${resource}_${task}_${interval_in_ms}@${node}=${rc}\n" "(memcached_monitor_20000@bart.example.com=7, for example).\n" "${rc} is an OCF return code. For more information on these\n" "return codes, refer to https://clusterlabs.org/pacemaker/doc/2.1/Pacemaker_Administration/html/agents.html#ocf-return-codes\n\n" "Examples:\n\n" "Pretend a recurring monitor action found memcached stopped on node\n" "fred.example.com and, during recovery, that the memcached stop\n" "action failed:\n\n" "\tcrm_simulate -LS --op-inject memcached:0_monitor_20000@bart.example.com=7 " "--op-fail memcached:0_stop_0@fred.example.com=1 --save-output /tmp/memcached-test.xml\n\n" "Now see what the reaction to the stop failed would be:\n\n" "\tcrm_simulate -S --xml-file /tmp/memcached-test.xml\n\n"; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); pcmk__add_main_args(context, extra_prog_entries); g_option_context_set_description(context, description); pcmk__add_arg_group(context, "operations", "Operations:", "Show operations options", operation_entries); pcmk__add_arg_group(context, "synthetic", "Synthetic Cluster Events:", "Show synthetic cluster event options", synthetic_entries); pcmk__add_arg_group(context, "artifact", "Artifact Options:", "Show artifact options", artifact_entries); pcmk__add_arg_group(context, "source", "Data Source:", "Show data source options", source_entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; pe_working_set_t *data_set = NULL; pcmk__output_t *out = NULL; GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "bdefgiqrtuwxDFGINOP"); GOptionContext *context = build_arg_context(args, &output_group); options.injections = calloc(1, sizeof(pcmk_injections_t)); if (options.injections == NULL) { rc = ENOMEM; goto done; } /* This must come before g_option_context_parse_strv. */ options.xml_file = strdup("-"); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("crm_simulate", args->verbosity); rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { fprintf(stderr, "Error creating output format %s: %s\n", args->output_ty, pcmk_rc_str(rc)); exit_code = CRM_EX_ERROR; goto done; } if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches) && !pcmk_is_set(options.flags, pcmk_sim_show_scores) && !pcmk_is_set(options.flags, pcmk_sim_show_utilization)) { pcmk__force_args(context, &error, "%s --text-fancy", g_get_prgname()); } else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) { pcmk__force_args(context, &error, "%s --xml-simple-list --xml-substitute", g_get_prgname()); } pe__register_messages(out); pcmk__register_lib_messages(out); out->quiet = args->quiet; if (args->version) { out->version(out, false); goto done; } if (args->verbosity > 0) { options.flags |= pcmk_sim_verbose; #ifdef PCMK__COMPAT_2_0 /* Redirect stderr to stdout so we can grep the output */ close(STDERR_FILENO); dup2(STDOUT_FILENO, STDERR_FILENO); #endif } data_set = pe_new_working_set(); if (data_set == NULL) { rc = ENOMEM; g_set_error(&error, PCMK__RC_ERROR, rc, "Could not allocate working set"); goto done; } if (pcmk_is_set(options.flags, pcmk_sim_show_scores)) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (pcmk_is_set(options.flags, pcmk_sim_show_utilization)) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } pe__set_working_set_flags(data_set, pe_flag_no_compat); if (options.test_dir != NULL) { data_set->priv = out; pcmk__profile_dir(options.test_dir, options.repeat, data_set, options.use_date); rc = pcmk_rc_ok; goto done; } rc = setup_input(out, options.xml_file, options.store? options.xml_file : options.output_file, &error); if (rc != pcmk_rc_ok) { goto done; } rc = pcmk__simulate(data_set, out, options.injections, options.flags, section_opts, options.use_date, options.input_file, options.graph_file, options.dot_file); done: pcmk__output_and_clear_error(&error, NULL); /* There sure is a lot to free in options. */ free(options.dot_file); free(options.graph_file); g_free(options.input_file); g_free(options.output_file); g_free(options.test_dir); free(options.use_date); free(options.xml_file); pcmk_free_injections(options.injections); pcmk__free_arg_context(context); g_strfreev(processed_args); if (data_set) { pe_free_working_set(data_set); } fflush(stderr); if (temp_shadow) { unlink(temp_shadow); free(temp_shadow); } if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); } if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); crm_exit(exit_code); } diff --git a/tools/crm_ticket.c b/tools/crm_ticket.c index 4cf0045c32..05b026dc0e 100644 --- a/tools/crm_ticket.c +++ b/tools/crm_ticket.c @@ -1,999 +1,1000 @@ /* * Copyright 2012-2023 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include GError *error = NULL; #define SUMMARY "Perform tasks related to cluster tickets\n\n" \ "Allows ticket attributes to be queried, modified and deleted." struct { gchar *attr_default; gchar *attr_id; char *attr_name; char *attr_value; gboolean force; char *get_attr_name; gboolean quiet; gchar *set_name; char ticket_cmd; gchar *ticket_id; gchar *xml_file; } options = { .ticket_cmd = 'S' }; GList *attr_delete; GHashTable *attr_set; bool modified = false; int cib_options = cib_sync_call; #define INDENT " " static gboolean attr_value_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&options.attr_value, optarg); if (!options.attr_name || !options.attr_value) { return TRUE; } g_hash_table_insert(attr_set, strdup(options.attr_name), strdup(options.attr_value)); pcmk__str_update(&options.attr_name, NULL); pcmk__str_update(&options.attr_value, NULL); modified = true; return TRUE; } static gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (pcmk__str_any_of(option_name, "--info", "-l", NULL)) { options.ticket_cmd = 'l'; } else if (pcmk__str_any_of(option_name, "--details", "-L", NULL)) { options.ticket_cmd = 'L'; } else if (pcmk__str_any_of(option_name, "--raw", "-w", NULL)) { options.ticket_cmd = 'w'; } else if (pcmk__str_any_of(option_name, "--query-xml", "-q", NULL)) { options.ticket_cmd = 'q'; } else if (pcmk__str_any_of(option_name, "--constraints", "-c", NULL)) { options.ticket_cmd = 'c'; } else if (pcmk__str_any_of(option_name, "--cleanup", "-C", NULL)) { options.ticket_cmd = 'C'; } return TRUE; } static gboolean delete_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { attr_delete = g_list_append(attr_delete, strdup(optarg)); modified = true; return TRUE; } static gboolean get_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&options.get_attr_name, optarg); options.ticket_cmd = 'G'; return TRUE; } static gboolean grant_standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (pcmk__str_any_of(option_name, "--grant", "-g", NULL)) { g_hash_table_insert(attr_set, strdup("granted"), strdup("true")); modified = true; } else if (pcmk__str_any_of(option_name, "--revoke", "-r", NULL)) { g_hash_table_insert(attr_set, strdup("granted"), strdup("false")); modified = true; } else if (pcmk__str_any_of(option_name, "--standby", "-s", NULL)) { g_hash_table_insert(attr_set, strdup("standby"), strdup("true")); modified = true; } else if (pcmk__str_any_of(option_name, "--activate", "-a", NULL)) { g_hash_table_insert(attr_set, strdup("standby"), strdup("false")); modified = true; } return TRUE; } static gboolean set_attr_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&options.attr_name, optarg); if (!options.attr_name || !options.attr_value) { return TRUE; } g_hash_table_insert(attr_set, strdup(options.attr_name), strdup(options.attr_value)); pcmk__str_update(&options.attr_name, NULL); pcmk__str_update(&options.attr_value, NULL); modified = true; return TRUE; } static GOptionEntry query_entries[] = { { "info", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the information of ticket(s)", NULL }, { "details", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the details of ticket(s)", NULL }, { "raw", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the IDs of ticket(s)", NULL }, { "query-xml", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Query the XML of ticket(s)", NULL }, { "constraints", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the rsc_ticket constraints that apply to ticket(s)", NULL }, { NULL } }; static GOptionEntry command_entries[] = { { "grant", 'g', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb, "Grant a ticket to this cluster site", NULL }, { "revoke", 'r', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb, "Revoke a ticket from this cluster site", NULL }, { "standby", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb, "Tell this cluster site this ticket is standby", NULL }, { "activate", 'a', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, grant_standby_cb, "Tell this cluster site this ticket is active", NULL }, { NULL } }; static GOptionEntry advanced_entries[] = { { "get-attr", 'G', 0, G_OPTION_ARG_CALLBACK, get_attr_cb, "Display the named attribute for a ticket", "ATTRIBUTE" }, { "set-attr", 'S', 0, G_OPTION_ARG_CALLBACK, set_attr_cb, "Set the named attribute for a ticket", "ATTRIBUTE" }, { "delete-attr", 'D', 0, G_OPTION_ARG_CALLBACK, delete_attr_cb, "Delete the named attribute for a ticket", "ATTRIBUTE" }, { "cleanup", 'C', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Delete all state of a ticket at this cluster site", NULL }, { NULL} }; static GOptionEntry addl_entries[] = { { "attr-value", 'v', 0, G_OPTION_ARG_CALLBACK, attr_value_cb, "Attribute value to use with -S", "VALUE" }, { "default", 'd', 0, G_OPTION_ARG_STRING, &options.attr_default, "(Advanced) Default attribute value to display if none is found\n" INDENT "(for use with -G)", "VALUE" }, { "force", 'f', 0, G_OPTION_ARG_NONE, &options.force, "(Advanced) Force the action to be performed", NULL }, { "ticket", 't', 0, G_OPTION_ARG_STRING, &options.ticket_id, "Ticket ID", "ID" }, { "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_STRING, &options.xml_file, NULL, NULL }, { NULL } }; static GOptionEntry deprecated_entries[] = { { "set-name", 'n', 0, G_OPTION_ARG_STRING, &options.set_name, "(Advanced) ID of the instance_attributes object to change", "ID" }, { "nvpair", 'i', 0, G_OPTION_ARG_STRING, &options.attr_id, "(Advanced) ID of the nvpair object to change/delete", "ID" }, { "quiet", 'Q', 0, G_OPTION_ARG_NONE, &options.quiet, "Print only the value on stdout", NULL }, { NULL } }; static pe_ticket_t * find_ticket(gchar *ticket_id, pe_working_set_t * data_set) { return g_hash_table_lookup(data_set->tickets, ticket_id); } static void print_date(time_t time) { int lpc = 0; char date_str[26]; asctime_r(localtime(&time), date_str); for (; lpc < 26; lpc++) { if (date_str[lpc] == '\n') { date_str[lpc] = 0; } } fprintf(stdout, "'%s'", date_str); } static void print_ticket(pe_ticket_t * ticket, bool raw, bool details) { if (raw) { fprintf(stdout, "%s\n", ticket->id); return; } fprintf(stdout, "%s\t%s %s", ticket->id, ticket->granted ? "granted" : "revoked", ticket->standby ? "[standby]" : " "); if (details && g_hash_table_size(ticket->state) > 0) { GHashTableIter iter; const char *name = NULL; const char *value = NULL; int lpc = 0; fprintf(stdout, " ("); g_hash_table_iter_init(&iter, ticket->state); while (g_hash_table_iter_next(&iter, (void **)&name, (void **)&value)) { if (lpc > 0) { fprintf(stdout, ", "); } fprintf(stdout, "%s=", name); if (pcmk__str_any_of(name, "last-granted", "expires", NULL)) { long long time_ll; pcmk__scan_ll(value, &time_ll, 0); print_date((time_t) time_ll); } else { fprintf(stdout, "%s", value); } lpc++; } fprintf(stdout, ")\n"); } else { if (ticket->last_granted > -1) { fprintf(stdout, " last-granted="); print_date(ticket->last_granted); } fprintf(stdout, "\n"); } return; } static void print_ticket_list(pe_working_set_t * data_set, bool raw, bool details) { GHashTableIter iter; pe_ticket_t *ticket = NULL; g_hash_table_iter_init(&iter, data_set->tickets); while (g_hash_table_iter_next(&iter, NULL, (void **)&ticket)) { print_ticket(ticket, raw, details); } } static int find_ticket_state(cib_t * the_cib, gchar *ticket_id, xmlNode ** ticket_state_xml) { int rc = pcmk_rc_ok; xmlNode *xml_search = NULL; GString *xpath = NULL; CRM_ASSERT(ticket_state_xml != NULL); *ticket_state_xml = NULL; xpath = g_string_sized_new(1024); g_string_append(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS "/" XML_CIB_TAG_TICKETS); if (ticket_id != NULL) { pcmk__g_strcat(xpath, "/" XML_CIB_TAG_TICKET_STATE "[@" XML_ATTR_ID "=\"", ticket_id, "\"]", NULL); } rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); rc = pcmk_legacy2rc(rc); g_string_free(xpath, TRUE); if (rc != pcmk_rc_ok) { return rc; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { if (ticket_id) { fprintf(stdout, "Multiple ticket_states match ticket_id=%s\n", ticket_id); } *ticket_state_xml = xml_search; } else { *ticket_state_xml = xml_search; } return rc; } static int find_ticket_constraints(cib_t * the_cib, gchar *ticket_id, xmlNode ** ticket_cons_xml) { int rc = pcmk_rc_ok; xmlNode *xml_search = NULL; GString *xpath = NULL; const char *xpath_base = NULL; CRM_ASSERT(ticket_cons_xml != NULL); *ticket_cons_xml = NULL; xpath_base = pcmk_cib_xpath_for(XML_CIB_TAG_CONSTRAINTS); if (xpath_base == NULL) { crm_err(XML_CIB_TAG_CONSTRAINTS " CIB element not known (bug?)"); return -ENOMSG; } xpath = g_string_sized_new(1024); pcmk__g_strcat(xpath, xpath_base, "/" XML_CONS_TAG_RSC_TICKET, NULL); if (ticket_id != NULL) { pcmk__g_strcat(xpath, "[@" XML_TICKET_ATTR_TICKET "=\"", ticket_id, "\"]", NULL); } rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); rc = pcmk_legacy2rc(rc); g_string_free(xpath, TRUE); if (rc != pcmk_rc_ok) { return rc; } crm_log_xml_debug(xml_search, "Match"); *ticket_cons_xml = xml_search; return rc; } static int dump_ticket_xml(cib_t * the_cib, gchar *ticket_id) { int rc = pcmk_rc_ok; xmlNode *state_xml = NULL; rc = find_ticket_state(the_cib, ticket_id, &state_xml); if (state_xml == NULL) { return rc; } fprintf(stdout, "State XML:\n"); if (state_xml) { fprintf(stdout, "\n"); pcmk__xml2fd(STDOUT_FILENO, state_xml); } return rc; } static int dump_constraints(cib_t * the_cib, gchar *ticket_id) { int rc = pcmk_rc_ok; xmlNode *cons_xml = NULL; rc = find_ticket_constraints(the_cib, ticket_id, &cons_xml); if (cons_xml == NULL) { return rc; } fprintf(stdout, "Constraints XML:\n\n"); pcmk__xml2fd(STDOUT_FILENO, cons_xml); return rc; } static int get_ticket_state_attr(gchar *ticket_id, const char *attr_name, const char **attr_value, pe_working_set_t * data_set) { pe_ticket_t *ticket = NULL; CRM_ASSERT(attr_value != NULL); *attr_value = NULL; ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { return ENXIO; } *attr_value = g_hash_table_lookup(ticket->state, attr_name); if (*attr_value == NULL) { return ENXIO; } return pcmk_rc_ok; } static void ticket_warning(gchar *ticket_id, const char *action) { GString *warning = g_string_sized_new(1024); const char *word = NULL; CRM_ASSERT(action != NULL); if (strcmp(action, "grant") == 0) { pcmk__g_strcat(warning, "This command cannot help you verify whether '", ticket_id, "' has been already granted elsewhere.\n", NULL); word = "to"; } else { pcmk__g_strcat(warning, "Revoking '", ticket_id, "' can trigger the specified " "'loss-policy'(s) relating to '", ticket_id, "'.\n\n" "You can check that with:\n" "crm_ticket --ticket ", ticket_id, " --constraints\n\n" "Otherwise before revoking '", ticket_id, "', " "you may want to make '", ticket_id, "' " "standby with:\n" "crm_ticket --ticket ", ticket_id, " --standby\n\n", NULL); word = "from"; } pcmk__g_strcat(warning, "If you really want to ", action, " '", ticket_id, "' ", word, " this site now, and you know what you are doing,\n" "please specify --force.", NULL); fprintf(stdout, "%s\n", (const char *) warning->str); g_string_free(warning, TRUE); } static bool allow_modification(gchar *ticket_id) { const char *value = NULL; GList *list_iter = NULL; if (options.force) { return true; } if (g_hash_table_lookup_extended(attr_set, "granted", NULL, (gpointer *) & value)) { if (crm_is_true(value)) { ticket_warning(ticket_id, "grant"); return false; } else { ticket_warning(ticket_id, "revoke"); return false; } } for(list_iter = attr_delete; list_iter; list_iter = list_iter->next) { const char *key = (const char *)list_iter->data; if (pcmk__str_eq(key, "granted", pcmk__str_casei)) { ticket_warning(ticket_id, "revoke"); return false; } } return true; } static int modify_ticket_state(gchar * ticket_id, cib_t * cib, pe_working_set_t * data_set) { int rc = pcmk_rc_ok; xmlNode *xml_top = NULL; xmlNode *ticket_state_xml = NULL; bool found = false; GList *list_iter = NULL; GHashTableIter hash_iter; char *key = NULL; char *value = NULL; pe_ticket_t *ticket = NULL; rc = find_ticket_state(cib, ticket_id, &ticket_state_xml); if (rc == pcmk_rc_ok) { crm_debug("Found a match state for ticket: id=%s", ticket_id); xml_top = ticket_state_xml; found = true; } else if (rc != ENXIO) { return rc; } else if (g_hash_table_size(attr_set) == 0){ return pcmk_rc_ok; } else { xmlNode *xml_obj = NULL; xml_top = create_xml_node(NULL, XML_CIB_TAG_STATUS); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_TICKETS); ticket_state_xml = create_xml_node(xml_obj, XML_CIB_TAG_TICKET_STATE); crm_xml_add(ticket_state_xml, XML_ATTR_ID, ticket_id); } for(list_iter = attr_delete; list_iter; list_iter = list_iter->next) { const char *key = (const char *)list_iter->data; xml_remove_prop(ticket_state_xml, key); } ticket = find_ticket(ticket_id, data_set); g_hash_table_iter_init(&hash_iter, attr_set); while (g_hash_table_iter_next(&hash_iter, (gpointer *) & key, (gpointer *) & value)) { crm_xml_add(ticket_state_xml, key, value); if (pcmk__str_eq(key, "granted", pcmk__str_casei) && (ticket == NULL || ticket->granted == FALSE) && crm_is_true(value)) { char *now = pcmk__ttoa(time(NULL)); crm_xml_add(ticket_state_xml, "last-granted", now); free(now); } } if (found && (attr_delete != NULL)) { crm_log_xml_debug(xml_top, "Replace"); rc = cib->cmds->replace(cib, XML_CIB_TAG_STATUS, ticket_state_xml, cib_options); rc = pcmk_legacy2rc(rc); } else { crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, xml_top, cib_options); rc = pcmk_legacy2rc(rc); } free_xml(xml_top); return rc; } static int delete_ticket_state(gchar *ticket_id, cib_t * cib) { xmlNode *ticket_state_xml = NULL; int rc = pcmk_rc_ok; rc = find_ticket_state(cib, ticket_id, &ticket_state_xml); if (rc == ENXIO) { return pcmk_rc_ok; } else if (rc != pcmk_rc_ok) { return rc; } crm_log_xml_debug(ticket_state_xml, "Delete"); rc = cib->cmds->remove(cib, XML_CIB_TAG_STATUS, ticket_state_xml, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { fprintf(stdout, "Cleaned up %s\n", ticket_id); } free_xml(ticket_state_xml); return rc; } static GOptionContext * build_arg_context(pcmk__common_args_t *args) { GOptionContext *context = NULL; const char *description = "Examples:\n\n" "Display the info of tickets:\n\n" "\tcrm_ticket --info\n\n" "Display the detailed info of tickets:\n\n" "\tcrm_ticket --details\n\n" "Display the XML of 'ticketA':\n\n" "\tcrm_ticket --ticket ticketA --query-xml\n\n" "Display the rsc_ticket constraints that apply to 'ticketA':\n\n" "\tcrm_ticket --ticket ticketA --constraints\n\n" "Grant 'ticketA' to this cluster site:\n\n" "\tcrm_ticket --ticket ticketA --grant\n\n" "Revoke 'ticketA' from this cluster site:\n\n" "\tcrm_ticket --ticket ticketA --revoke\n\n" "Make 'ticketA' standby (the cluster site will treat a granted\n" "'ticketA' as 'standby', and the dependent resources will be\n" "stopped or demoted gracefully without triggering loss-policies):\n\n" "\tcrm_ticket --ticket ticketA --standby\n\n" "Activate 'ticketA' from being standby:\n\n" "\tcrm_ticket --ticket ticketA --activate\n\n" "Get the value of the 'granted' attribute for 'ticketA':\n\n" "\tcrm_ticket --ticket ticketA --get-attr granted\n\n" "Set the value of the 'standby' attribute for 'ticketA':\n\n" "\tcrm_ticket --ticket ticketA --set-attr standby --attr-value true\n\n" "Delete the 'granted' attribute for 'ticketA':\n\n" "\tcrm_ticket --ticket ticketA --delete-attr granted\n\n" "Erase the operation history of 'ticketA' at this cluster site,\n" "causing the cluster site to 'forget' the existing ticket state:\n\n" "\tcrm_ticket --ticket ticketA --cleanup\n\n"; context = pcmk__build_arg_context(args, NULL, NULL, NULL); g_option_context_set_description(context, description); pcmk__add_arg_group(context, "queries", "Queries:", "Show queries", query_entries); pcmk__add_arg_group(context, "commands", "Commands:", "Show command options", command_entries); pcmk__add_arg_group(context, "advanced", "Advanced Options:", "Show advanced options", advanced_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); pcmk__add_arg_group(context, "deprecated", "Deprecated Options:", "Show deprecated options", deprecated_entries); return context; } int main(int argc, char **argv) { pe_working_set_t *data_set = NULL; xmlNode *cib_xml_copy = NULL; cib_t *cib_conn = NULL; crm_exit_t exit_code = CRM_EX_OK; int rc = pcmk_rc_ok; pcmk__common_args_t *args = NULL; GOptionContext *context = NULL; gchar **processed_args = NULL; attr_set = pcmk__strkey_table(free, free); attr_delete = NULL; args = pcmk__new_common_args(SUMMARY); context = build_arg_context(args); processed_args = pcmk__cmdline_preproc(argv, "dintvxCDGS"); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("crm_ticket", args->verbosity); if (args->version) { g_strfreev(processed_args); pcmk__free_arg_context(context); /* FIXME: When crm_ticket is converted to use formatted output, this can go. */ pcmk__cli_help('v'); } data_set = pe_new_working_set(); if (data_set == NULL) { crm_perror(LOG_CRIT, "Could not allocate working set"); exit_code = CRM_EX_OSERR; goto done; } pe__set_working_set_flags(data_set, pe_flag_no_counts|pe_flag_no_compat); cib_conn = cib_new(); if (cib_conn == NULL) { exit_code = CRM_EX_DISCONNECT; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to the CIB manager"); goto done; } rc = cib_conn->cmds->signon(cib_conn, crm_system_name, cib_command); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to the CIB: %s", pcmk_rc_str(rc)); goto done; } if (options.xml_file != NULL) { cib_xml_copy = filename2xml(options.xml_file); } else { rc = cib_conn->cmds->query(cib_conn, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not get local CIB: %s", pcmk_rc_str(rc)); goto done; } } if (!cli_config_update(&cib_xml_copy, NULL, FALSE)) { exit_code = CRM_EX_CONFIG; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not update local CIB to latest schema version"); goto done; } data_set->input = cib_xml_copy; data_set->now = crm_time_new(NULL); cluster_status(data_set); /* For recording the tickets that are referenced in rsc_ticket constraints * but have never been granted yet. */ pcmk__unpack_constraints(data_set); if (options.ticket_cmd == 'l' || options.ticket_cmd == 'L' || options.ticket_cmd == 'w') { bool raw = false; bool details = false; if (options.ticket_cmd == 'L') { details = true; } else if (options.ticket_cmd == 'w') { raw = true; } if (options.ticket_id) { pe_ticket_t *ticket = find_ticket(options.ticket_id, data_set); if (ticket == NULL) { exit_code = CRM_EX_NOSUCH; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "No such ticket '%s'", options.ticket_id); goto done; } print_ticket(ticket, raw, details); } else { print_ticket_list(data_set, raw, details); } } else if (options.ticket_cmd == 'q') { rc = dump_ticket_xml(cib_conn, options.ticket_id); exit_code = pcmk_rc2exitc(rc); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not query ticket XML: %s", pcmk_rc_str(rc)); } } else if (options.ticket_cmd == 'c') { rc = dump_constraints(cib_conn, options.ticket_id); exit_code = pcmk_rc2exitc(rc); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not show ticket constraints: %s", pcmk_rc_str(rc)); } } else if (options.ticket_cmd == 'G') { const char *value = NULL; if (options.ticket_id == NULL) { exit_code = CRM_EX_NOSUCH; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply ticket ID with -t"); goto done; } rc = get_ticket_state_attr(options.ticket_id, options.get_attr_name, &value, data_set); if (rc == pcmk_rc_ok) { fprintf(stdout, "%s\n", value); } else if (rc == ENXIO && options.attr_default) { fprintf(stdout, "%s\n", options.attr_default); rc = pcmk_rc_ok; } exit_code = pcmk_rc2exitc(rc); } else if (options.ticket_cmd == 'C') { if (options.ticket_id == NULL) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply ticket ID with -t"); goto done; } if (options.force == FALSE) { pe_ticket_t *ticket = NULL; ticket = find_ticket(options.ticket_id, data_set); if (ticket == NULL) { exit_code = CRM_EX_NOSUCH; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "No such ticket '%s'", options.ticket_id); goto done; } if (ticket->granted) { ticket_warning(options.ticket_id, "revoke"); exit_code = CRM_EX_INSUFFICIENT_PRIV; goto done; } } rc = delete_ticket_state(options.ticket_id, cib_conn); exit_code = pcmk_rc2exitc(rc); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not clean up ticket: %s", pcmk_rc_str(rc)); } } else if (modified) { if (options.ticket_id == NULL) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply ticket ID with -t"); goto done; } if (options.attr_value && (pcmk__str_empty(options.attr_name))) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply attribute name with -S for -v %s", options.attr_value); goto done; } if (options.attr_name && (pcmk__str_empty(options.attr_value))) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply attribute value with -v for -S %s", options.attr_value); goto done; } if (!allow_modification(options.ticket_id)) { exit_code = CRM_EX_INSUFFICIENT_PRIV; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Ticket modification not allowed"); goto done; } rc = modify_ticket_state(options.ticket_id, cib_conn, data_set); exit_code = pcmk_rc2exitc(rc); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not modify ticket: %s", pcmk_rc_str(rc)); } } else if (options.ticket_cmd == 'S') { /* Correct usage was handled in the "if (modified)" block above, so * this is just for reporting usage errors */ if (pcmk__str_empty(options.attr_name)) { // We only get here if ticket_cmd was left as default exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply a command"); goto done; } if (options.ticket_id == NULL) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply ticket ID with -t"); goto done; } if (pcmk__str_empty(options.attr_value)) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Must supply value with -v for -S %s", options.attr_name); goto done; } } else { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Unknown command: %c", options.ticket_cmd); } done: if (attr_set) { g_hash_table_destroy(attr_set); } attr_set = NULL; if (attr_delete) { g_list_free_full(attr_delete, free); } attr_delete = NULL; pe_free_working_set(data_set); data_set = NULL; cib__clean_up_connection(&cib_conn); g_strfreev(processed_args); pcmk__free_arg_context(context); g_free(options.attr_default); g_free(options.attr_id); free(options.attr_name); free(options.attr_value); free(options.get_attr_name); g_free(options.set_name); g_free(options.ticket_id); g_free(options.xml_file); pcmk__output_and_clear_error(&error, NULL); crm_exit(exit_code); }