diff --git a/daemons/controld/controld_execd.c b/daemons/controld/controld_execd.c index fcdadca98b..dbd65dd775 100644 --- a/daemons/controld/controld_execd.c +++ b/daemons/controld/controld_execd.c @@ -1,2898 +1,2905 @@ /* * Copyright 2004-2021 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 // lrmd_event_data_t, lrmd_rsc_info_t, etc. #include #include #include #include +#include #include #include #define START_DELAY_THRESHOLD 5 * 60 * 1000 #define MAX_LRM_REG_FAILS 30 struct delete_event_s { int rc; const char *rsc; lrm_state_t *lrm_state; }; static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id); static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list); static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data); static lrmd_event_data_t *construct_op(lrm_state_t * lrm_state, xmlNode * rsc_op, const char *rsc_id, const char *operation); static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg); static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level); static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time); static void lrm_connection_destroy(void) { if (pcmk_is_set(fsa_input_register, R_LRM_CONNECTED)) { crm_crit("Connection to executor failed"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); controld_clear_fsa_input_flags(R_LRM_CONNECTED); } else { crm_info("Disconnected from executor"); } } static char * make_stop_id(const char *rsc, int call_id) { return crm_strdup_printf("%s:%d", rsc, call_id); } static void copy_instance_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") == NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } static void copy_meta_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") != NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } /*! * \internal * \brief Remove a recurring operation from a resource's history * * \param[in,out] history Resource history to modify * \param[in] op Operation to remove * * \return TRUE if the operation was found and removed, FALSE otherwise */ static gboolean history_remove_recurring_op(rsc_history_t *history, const lrmd_event_data_t *op) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_event_data_t *existing = iter->data; if ((op->interval_ms == existing->interval_ms) && pcmk__str_eq(op->rsc_id, existing->rsc_id, pcmk__str_none) && pcmk__str_eq(op->op_type, existing->op_type, pcmk__str_casei)) { history->recurring_op_list = g_list_delete_link(history->recurring_op_list, iter); lrmd_free_event(existing); return TRUE; } } return FALSE; } /*! * \internal * \brief Free all recurring operations in resource history * * \param[in,out] history Resource history to modify */ static void history_free_recurring_ops(rsc_history_t *history) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_free_event(iter->data); } g_list_free(history->recurring_op_list); history->recurring_op_list = NULL; } /*! * \internal * \brief Free resource history * * \param[in,out] history Resource history to free */ void history_free(gpointer data) { rsc_history_t *history = (rsc_history_t*)data; if (history->stop_params) { g_hash_table_destroy(history->stop_params); } /* Don't need to free history->rsc.id because it's set to history->id */ free(history->rsc.type); free(history->rsc.standard); free(history->rsc.provider); lrmd_free_event(history->failed); lrmd_free_event(history->last); free(history->id); history_free_recurring_ops(history); free(history); } static void update_history_cache(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { int target_rc = 0; rsc_history_t *entry = NULL; if (op->rsc_deleted) { crm_debug("Purged history for '%s' after %s", op->rsc_id, op->op_type); controld_delete_resource_history(op->rsc_id, lrm_state->node_name, NULL, crmd_cib_smart_opt()); return; } if (pcmk__str_eq(op->op_type, RSC_NOTIFY, pcmk__str_casei)) { return; } crm_debug("Updating history for '%s' with %s op", op->rsc_id, op->op_type); entry = g_hash_table_lookup(lrm_state->resource_history, op->rsc_id); if (entry == NULL && rsc) { entry = calloc(1, sizeof(rsc_history_t)); entry->id = strdup(op->rsc_id); g_hash_table_insert(lrm_state->resource_history, entry->id, entry); entry->rsc.id = entry->id; entry->rsc.type = strdup(rsc->type); entry->rsc.standard = strdup(rsc->standard); if (rsc->provider) { entry->rsc.provider = strdup(rsc->provider); } else { entry->rsc.provider = NULL; } } else if (entry == NULL) { crm_info("Resource %s no longer exists, not updating cache", op->rsc_id); return; } entry->last_callid = op->call_id; target_rc = rsc_op_expected_rc(op); if (op->op_status == PCMK_EXEC_CANCELLED) { if (op->interval_ms > 0) { crm_trace("Removing cancelled recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); history_remove_recurring_op(entry, op); return; } else { crm_trace("Skipping " PCMK__OP_FMT " rc=%d, status=%d", op->rsc_id, op->op_type, op->interval_ms, op->rc, op->op_status); } } else if (did_rsc_op_fail(op, target_rc)) { /* Store failed monitors here, otherwise the block below will cause them * to be forgotten when a stop happens. */ if (entry->failed) { lrmd_free_event(entry->failed); } entry->failed = lrmd_copy_event(op); } else if (op->interval_ms == 0) { if (entry->last) { lrmd_free_event(entry->last); } entry->last = lrmd_copy_event(op); if (op->params && pcmk__strcase_any_of(op->op_type, CRMD_ACTION_START, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, CRMD_ACTION_STATUS, NULL)) { if (entry->stop_params) { g_hash_table_destroy(entry->stop_params); } entry->stop_params = pcmk__strkey_table(free, free); g_hash_table_foreach(op->params, copy_instance_keys, entry->stop_params); } } if (op->interval_ms > 0) { /* Ensure there are no duplicates */ history_remove_recurring_op(entry, op); crm_trace("Adding recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); entry->recurring_op_list = g_list_prepend(entry->recurring_op_list, lrmd_copy_event(op)); } else if (entry->recurring_op_list && !pcmk__str_eq(op->op_type, RSC_STATUS, pcmk__str_casei)) { crm_trace("Dropping %d recurring ops because of: " PCMK__OP_FMT, g_list_length(entry->recurring_op_list), op->rsc_id, op->op_type, op->interval_ms); history_free_recurring_ops(entry); } } /*! * \internal * \brief Send a direct OK ack for a resource task * * \param[in] lrm_state LRM connection * \param[in] input Input message being ack'ed * \param[in] rsc_id ID of affected resource * \param[in] rsc Affected resource (if available) * \param[in] task Operation task being ack'ed * \param[in] ack_host Name of host to send ack to * \param[in] ack_sys IPC system name to ack */ static void send_task_ok_ack(lrm_state_t *lrm_state, ha_msg_input_t *input, const char *rsc_id, lrmd_rsc_info_t *rsc, const char *task, const char *ack_host, const char *ack_sys) { lrmd_event_data_t *op = construct_op(lrm_state, input->xml, rsc_id, task); - op->rc = PCMK_OCF_OK; - op->op_status = PCMK_EXEC_DONE; + lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(ack_host, ack_sys, rsc, op, rsc_id); lrmd_free_event(op); } static inline const char * op_node_name(lrmd_event_data_t *op) { return op->remote_nodename? op->remote_nodename : fsa_our_uname; } void lrm_op_callback(lrmd_event_data_t * op) { CRM_CHECK(op != NULL, return); switch (op->type) { case lrmd_event_disconnect: if (op->remote_nodename == NULL) { /* If this is the local executor IPC connection, set the right * bits in the controller when the connection goes down. */ lrm_connection_destroy(); } break; case lrmd_event_exec_complete: { lrm_state_t *lrm_state = lrm_state_find(op_node_name(op)); CRM_ASSERT(lrm_state != NULL); process_lrm_event(lrm_state, op, NULL, NULL); } break; default: break; } } static void try_local_executor_connect(long long action, fsa_data_t *msg_data, lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; crm_debug("Connecting to the local executor"); // If we can connect, great rc = controld_connect_local_executor(lrm_state); if (rc == pcmk_rc_ok) { controld_set_fsa_input_flags(R_LRM_CONNECTED); crm_info("Connection to the local executor established"); return; } // Otherwise, if we can try again, set a timer to do so if (lrm_state->num_lrm_register_fails < MAX_LRM_REG_FAILS) { crm_warn("Failed to connect to the local executor %d time%s " "(%d max): %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), MAX_LRM_REG_FAILS, pcmk_rc_str(rc)); controld_start_timer(wait_timer); crmd_fsa_stall(FALSE); return; } // Otherwise give up crm_err("Failed to connect to the executor the max allowed " "%d time%s: %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), pcmk_rc_str(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } /* A_LRM_CONNECT */ void do_lrm_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { /* This only pertains to local executor connections. Remote connections are * handled as resources within the scheduler. Connecting and disconnecting * from remote executor instances is handled differently. */ lrm_state_t *lrm_state = NULL; if(fsa_our_uname == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(fsa_our_uname); if (lrm_state == NULL) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (action & A_LRM_DISCONNECT) { if (lrm_state_verify_stopped(lrm_state, cur_state, LOG_INFO) == FALSE) { if (action == A_LRM_DISCONNECT) { crmd_fsa_stall(FALSE); return; } } controld_clear_fsa_input_flags(R_LRM_CONNECTED); crm_info("Disconnecting from the executor"); lrm_state_disconnect(lrm_state); lrm_state_reset_tables(lrm_state, FALSE); crm_notice("Disconnected from the executor"); } if (action & A_LRM_CONNECT) { try_local_executor_connect(action, msg_data, lrm_state); } if (action & ~(A_LRM_CONNECT | A_LRM_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __func__); } } static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level) { int counter = 0; gboolean rc = TRUE; const char *when = "lrm disconnect"; GHashTableIter gIter; const char *key = NULL; rsc_history_t *entry = NULL; active_op_t *pending = NULL; crm_debug("Checking for active resources before exit"); if (cur_state == S_TERMINATE) { log_level = LOG_ERR; when = "shutdown"; } else if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { when = "shutdown... waiting"; } if (lrm_state->pending_ops && lrm_state_is_connected(lrm_state) == TRUE) { guint removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_actions, lrm_state); guint nremaining = g_hash_table_size(lrm_state->pending_ops); if (removed || nremaining) { crm_notice("Stopped %u recurring operation%s at %s (%u remaining)", removed, pcmk__plural_s(removed), when, nremaining); } } if (lrm_state->pending_ops) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, NULL, (void **)&pending)) { /* Ignore recurring actions in the shutdown calculations */ if (pending->interval_ms == 0) { counter++; } } } if (counter > 0) { do_crm_log(log_level, "%d pending executor operation%s at %s", counter, pcmk__plural_s(counter), when); if ((cur_state == S_TERMINATE) || !pcmk_is_set(fsa_input_register, R_SENT_RSC_STOP)) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, (gpointer*)&key, (gpointer*)&pending)) { do_crm_log(log_level, "Pending action: %s (%s)", key, pending->op_key); } } else { rc = FALSE; } return rc; } if (lrm_state->resource_history == NULL) { return rc; } if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { /* At this point we're not waiting, we're just shutting down */ when = "shutdown"; } counter = 0; g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (gpointer*)&entry)) { if (is_rsc_active(lrm_state, entry->id) == FALSE) { continue; } counter++; if (log_level == LOG_ERR) { crm_info("Found %s active at %s", entry->id, when); } else { crm_trace("Found %s active at %s", entry->id, when); } if (lrm_state->pending_ops) { GHashTableIter hIter; g_hash_table_iter_init(&hIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&hIter, (gpointer*)&key, (gpointer*)&pending)) { if (pcmk__str_eq(entry->id, pending->rsc_id, pcmk__str_none)) { crm_notice("%sction %s (%s) incomplete at %s", pending->interval_ms == 0 ? "A" : "Recurring a", key, pending->op_key, when); } } } } if (counter) { crm_err("%d resource%s active at %s", counter, (counter == 1)? " was" : "s were", when); } return rc; } /*! * \internal * \brief Build XML and string of parameters meeting some criteria, for digest * * \param[in] op Executor event with parameter table to use * \param[in] metadata Parsed meta-data for executed resource agent * \param[in] param_type Flag used for selection criteria * \param[out] result Will be set to newly created XML with selected * parameters as attributes * * \return Newly allocated space-separated string of parameter names * \note Selection criteria varies by param_type: for the restart digest, we * want parameters that are *not* marked reloadable (OCF 1.1) or that * *are* marked unique (pre-1.1), for both string and XML results; for the * secure digest, we want parameters that *are* marked private for the * string, but parameters that are *not* marked private for the XML. * \note It is the caller's responsibility to free the string return value with * free() and the XML result with free_xml(). */ static char * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, enum ra_param_flags_e param_type, xmlNode **result) { char *list = NULL; size_t len = 0; *result = create_xml_node(NULL, XML_TAG_PARAMS); for (GList *iter = metadata->ra_params; iter != NULL; iter = iter->next) { struct ra_param_s *param = (struct ra_param_s *) iter->data; bool accept_for_list = false; bool accept_for_xml = false; switch (param_type) { case ra_param_reloadable: accept_for_list = !pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_unique: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_private: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = !accept_for_list; break; } if (accept_for_list) { crm_trace("Attr %s is %s", param->rap_name, ra_param_flag2text(param_type)); if (list == NULL) { // We will later search for " WORD ", so start list with a space pcmk__add_word(&list, &len, " "); } pcmk__add_word(&list, &len, param->rap_name); } else { crm_trace("Rejecting %s for %s", param->rap_name, ra_param_flag2text(param_type)); } if (accept_for_xml) { const char *v = g_hash_table_lookup(op->params, param->rap_name); if (v != NULL) { crm_trace("Adding attr %s=%s to the xml result", param->rap_name, v); crm_xml_add(*result, param->rap_name, v); } } } if (list != NULL) { // We will later search for " WORD ", so end list with a space pcmk__add_word(&list, &len, " "); } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *restart = NULL; CRM_LOG_ASSERT(op->params != NULL); if (op->interval_ms > 0) { /* monitors are not reloadable */ return; } if (pcmk_is_set(metadata->ra_flags, ra_supports_reload_agent)) { // Add parameters not marked reloadable to the "op-force-restart" list list = build_parameter_list(op, metadata, ra_param_reloadable, &restart); } else if (pcmk_is_set(metadata->ra_flags, ra_supports_legacy_reload)) { /* @COMPAT pre-OCF-1.1 resource agents * * Before OCF 1.1, Pacemaker abused "unique=0" to indicate * reloadability. Add any parameters with unique="1" to the * "op-force-restart" list. */ list = build_parameter_list(op, metadata, ra_param_unique, &restart); } else { // Resource does not support agent reloads return; } digest = calculate_operation_digest(restart, version); /* Add "op-force-restart" and "op-restart-digest" to indicate the resource supports reload, * no matter if it actually supports any parameters with unique="1"). */ crm_xml_add(update, XML_LRM_ATTR_OP_RESTART, list? list: ""); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(restart, "restart digest source"); free_xml(restart); free(digest); free(list); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *secure = NULL; CRM_LOG_ASSERT(op->params != NULL); /* * To keep XML_LRM_ATTR_OP_SECURE short, we want it to contain the * secure parameters but XML_LRM_ATTR_SECURE_DIGEST to be based on * the insecure ones */ list = build_parameter_list(op, metadata, ra_param_private, &secure); if (list != NULL) { digest = calculate_operation_digest(secure, version); crm_xml_add(update, XML_LRM_ATTR_OP_SECURE, list); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(secure, "secure digest source"); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); free(list); } static gboolean build_operation_update(xmlNode * parent, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *node_name, const char *src) { int target_rc = 0; xmlNode *xml_op = NULL; struct ra_metadata_s *metadata = NULL; const char *caller_version = NULL; lrm_state_t *lrm_state = NULL; uint32_t metadata_source = controld_metadata_from_agent; if (op == NULL) { return FALSE; } target_rc = rsc_op_expected_rc(op); /* there is a small risk in formerly mixed clusters that it will * be sub-optimal. * * however with our upgrade policy, the update we send should * still be completely supported anyway */ caller_version = g_hash_table_lookup(op->params, XML_ATTR_CRM_VERSION); CRM_LOG_ASSERT(caller_version != NULL); if(caller_version == NULL) { caller_version = CRM_FEATURE_SET; } crm_trace("Building %s operation update with originator version: %s", op->rsc_id, caller_version); xml_op = pcmk__create_history_xml(parent, op, caller_version, target_rc, fsa_our_uname, src, LOG_DEBUG); if (xml_op == NULL) { return TRUE; } if ((rsc == NULL) || (op->params == NULL) || !crm_op_needs_metadata(rsc->standard, op->op_type)) { crm_trace("No digests needed for %s action on %s (params=%p rsc=%p)", op->op_type, op->rsc_id, op->params, rsc); return TRUE; } lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { crm_warn("Cannot calculate digests for operation " PCMK__OP_FMT " because we have no connection to executor for %s", op->rsc_id, op->op_type, op->interval_ms, node_name); return TRUE; } /* Getting meta-data from cache is OK unless this is a successful start * action -- always refresh from the agent for those, in case the * resource agent was updated. * * @TODO Only refresh the meta-data after starts if the agent actually * changed (using something like inotify, or a hash or modification time of * the agent executable). */ if ((op->op_status != PCMK_EXEC_DONE) || (op->rc != target_rc) || !pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { metadata_source |= controld_metadata_from_cache; } metadata = controld_get_rsc_metadata(lrm_state, rsc, metadata_source); if (metadata == NULL) { return TRUE; } #if ENABLE_VERSIONED_ATTRS crm_xml_add(xml_op, XML_ATTR_RA_VERSION, metadata->ra_version); #endif crm_trace("Including additional digests for %s:%s:%s", rsc->standard, rsc->provider, rsc->type); append_restart_list(op, metadata, xml_op, caller_version); append_secure_list(op, metadata, xml_op, caller_version); return TRUE; } static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id) { rsc_history_t *entry = NULL; entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->last == NULL) { return FALSE; } crm_trace("Processing %s: %s.%d=%d", rsc_id, entry->last->op_type, entry->last->interval_ms, entry->last->rc); if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_STOP, pcmk__str_casei)) { return FALSE; } else if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { // A stricter check is too complex ... leave that to the scheduler return FALSE; } else if (entry->last->rc == PCMK_OCF_NOT_RUNNING) { return FALSE; } else if ((entry->last->interval_ms == 0) && (entry->last->rc == PCMK_OCF_NOT_CONFIGURED)) { /* Badly configured resources can't be reliably stopped */ return FALSE; } return TRUE; } static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list) { GHashTableIter iter; rsc_history_t *entry = NULL; g_hash_table_iter_init(&iter, lrm_state->resource_history); while (g_hash_table_iter_next(&iter, NULL, (void **)&entry)) { GList *gIter = NULL; xmlNode *xml_rsc = create_xml_node(rsc_list, XML_LRM_TAG_RESOURCE); crm_xml_add(xml_rsc, XML_ATTR_ID, entry->id); crm_xml_add(xml_rsc, XML_ATTR_TYPE, entry->rsc.type); crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, entry->rsc.standard); crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, entry->rsc.provider); if (entry->last && entry->last->params) { const char *container = g_hash_table_lookup(entry->last->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); if (container) { crm_trace("Resource %s is a part of container resource %s", entry->id, container); crm_xml_add(xml_rsc, XML_RSC_ATTR_CONTAINER, container); } } build_operation_update(xml_rsc, &(entry->rsc), entry->failed, lrm_state->node_name, __func__); build_operation_update(xml_rsc, &(entry->rsc), entry->last, lrm_state->node_name, __func__); for (gIter = entry->recurring_op_list; gIter != NULL; gIter = gIter->next) { build_operation_update(xml_rsc, &(entry->rsc), gIter->data, lrm_state->node_name, __func__); } } return FALSE; } static xmlNode * do_lrm_query_internal(lrm_state_t *lrm_state, int update_flags) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; peer = crm_get_peer_full(0, lrm_state->node_name, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return NULL); xml_state = create_node_state_update(peer, update_flags, NULL, __func__); if (xml_state == NULL) { return NULL; } xml_data = create_xml_node(xml_state, XML_CIB_TAG_LRM); crm_xml_add(xml_data, XML_ATTR_ID, peer->uuid); rsc_list = create_xml_node(xml_data, XML_LRM_TAG_RESOURCES); /* Build a list of active (not always running) resources */ build_active_RAs(lrm_state, rsc_list); crm_log_xml_trace(xml_state, "Current executor state"); return xml_state; } xmlNode * controld_query_executor_state(const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (!lrm_state) { crm_err("Could not find executor state for node %s", node_name); return NULL; } return do_lrm_query_internal(lrm_state, node_update_cluster|node_update_peer); } /*! * \internal * \brief Map standard Pacemaker return code to operation status and OCF code * * \param[out] event Executor event whose status and return code should be set * \param[in] rc Standard Pacemaker return code */ void controld_rc2event(lrmd_event_data_t *event, int rc) { + /* This is called for cleanup requests from controller peers/clients, not + * for resource actions, so no exit reason is needed. + */ switch (rc) { case pcmk_rc_ok: - event->rc = PCMK_OCF_OK; - event->op_status = PCMK_EXEC_DONE; + lrmd__set_result(event, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); break; case EACCES: - event->rc = PCMK_OCF_INSUFFICIENT_PRIV; - event->op_status = PCMK_EXEC_ERROR; + lrmd__set_result(event, PCMK_OCF_INSUFFICIENT_PRIV, + PCMK_EXEC_ERROR, NULL); break; default: - event->rc = PCMK_OCF_UNKNOWN_ERROR; - event->op_status = PCMK_EXEC_ERROR; + lrmd__set_result(event, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, + NULL); break; } } /*! * \internal * \brief Trigger a new transition after CIB status was deleted * * If a CIB status delete was not expected (as part of the transition graph), * trigger a new transition by updating the (arbitrary) "last-lrm-refresh" * cluster property. * * \param[in] from_sys IPC name that requested the delete * \param[in] rsc_id Resource whose status was deleted (for logging only) */ void controld_trigger_delete_refresh(const char *from_sys, const char *rsc_id) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { char *now_s = crm_strdup_printf("%lld", (long long) time(NULL)); crm_debug("Triggering a refresh after %s cleaned %s", from_sys, rsc_id); update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, FALSE, NULL, NULL); free(now_s); } } static void notify_deleted(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, int rc) { lrmd_event_data_t *op = NULL; const char *from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); const char *from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "localhost"), rsc_id, ((rc == pcmk_ok)? "" : " not")); op = construct_op(lrm_state, input->xml, rsc_id, CRMD_ACTION_DELETE); controld_rc2event(op, pcmk_legacy2rc(rc)); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); controld_trigger_delete_refresh(from_sys, rsc_id); } static gboolean lrm_remove_deleted_rsc(gpointer key, gpointer value, gpointer user_data) { struct delete_event_s *event = user_data; struct pending_deletion_op_s *op = value; if (pcmk__str_eq(event->rsc, op->rsc, pcmk__str_none)) { notify_deleted(event->lrm_state, op->input, event->rsc, event->rc); return TRUE; } return FALSE; } static gboolean lrm_remove_deleted_op(gpointer key, gpointer value, gpointer user_data) { const char *rsc = user_data; active_op_t *pending = value; if (pcmk__str_eq(rsc, pending->rsc_id, pcmk__str_none)) { crm_info("Removing op %s:%d for deleted resource %s", pending->op_key, pending->call_id, rsc); return TRUE; } return FALSE; } static void delete_rsc_entry(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, GHashTableIter * rsc_gIter, int rc, const char *user_name) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_gIter) { g_hash_table_iter_remove(rsc_gIter); } else { g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); } controld_delete_resource_history(rsc_id_copy, lrm_state->node_name, user_name, crmd_cib_smart_opt()); g_hash_table_foreach_remove(lrm_state->pending_ops, lrm_remove_deleted_op, rsc_id_copy); free(rsc_id_copy); } if (input) { notify_deleted(lrm_state, input, rsc_id, rc); } event.rc = rc; event.rsc = rsc_id; event.lrm_state = lrm_state; g_hash_table_foreach_remove(lrm_state->deletion_ops, lrm_remove_deleted_rsc, &event); } /*! * \internal * \brief Erase an LRM history entry from the CIB, given the operation data * * \param[in] lrm_state LRM state of the desired node * \param[in] op Operation whose history should be deleted */ static void erase_lrm_history_by_op(lrm_state_t *lrm_state, lrmd_event_data_t *op) { xmlNode *xml_top = NULL; CRM_CHECK(op != NULL, return); xml_top = create_xml_node(NULL, XML_LRM_TAG_RSC_OP); crm_xml_add_int(xml_top, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add(xml_top, XML_ATTR_TRANSITION_KEY, op->user_data); if (op->interval_ms > 0) { char *op_id = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); /* Avoid deleting last_failure too (if it was a result of this recurring op failing) */ crm_xml_add(xml_top, XML_ATTR_ID, op_id); free(op_id); } crm_debug("Erasing resource operation history for " PCMK__OP_FMT " (call=%d)", op->rsc_id, op->op_type, op->interval_ms, op->call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, XML_CIB_TAG_STATUS, xml_top, cib_quorum_override); crm_log_xml_trace(xml_top, "op:cancel"); free_xml(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" XML_ATTR_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP /* ... and also by operation key */ #define XPATH_HISTORY_ID XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s']" /* ... and also by operation key and operation call ID */ #define XPATH_HISTORY_CALL XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_CALLID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_TASK_KEY "='%s']" /*! * \internal * \brief Erase an LRM history entry from the CIB, given operation identifiers * * \param[in] lrm_state LRM state of the node to clear history for * \param[in] rsc_id Name of resource to clear history for * \param[in] key Operation key of operation to clear history for * \param[in] orig_op If specified, delete only if it has this original op * \param[in] call_id If specified, delete entry only if it has this call ID */ static void erase_lrm_history_by_id(lrm_state_t *lrm_state, const char *rsc_id, const char *key, const char *orig_op, int call_id) { char *op_xpath = NULL; CRM_CHECK((rsc_id != NULL) && (key != NULL), return); if (call_id > 0) { op_xpath = crm_strdup_printf(XPATH_HISTORY_CALL, lrm_state->node_name, rsc_id, key, call_id); } else if (orig_op) { op_xpath = crm_strdup_printf(XPATH_HISTORY_ORIG, lrm_state->node_name, rsc_id, key, orig_op); } else { op_xpath = crm_strdup_printf(XPATH_HISTORY_ID, lrm_state->node_name, rsc_id, key); } crm_debug("Erasing resource operation history for %s on %s (call=%d)", key, rsc_id, call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, op_xpath, NULL, cib_quorum_override | cib_xpath); free(op_xpath); } static inline gboolean last_failed_matches_op(rsc_history_t *entry, const char *op, guint interval_ms) { if (entry == NULL) { return FALSE; } if (op == NULL) { return TRUE; } return (pcmk__str_eq(op, entry->failed->op_type, pcmk__str_casei) && (interval_ms == entry->failed->interval_ms)); } /*! * \internal * \brief Clear a resource's last failure * * Erase a resource's last failure on a particular node from both the * LRM resource history in the CIB, and the resource history remembered * for the LRM state. * * \param[in] rsc_id Resource name * \param[in] node_name Node name * \param[in] operation If specified, only clear if matching this operation * \param[in] interval_ms If operation is specified, it has this interval */ void lrm_clear_last_failure(const char *rsc_id, const char *node_name, const char *operation, guint interval_ms) { char *op_key = NULL; char *orig_op_key = NULL; lrm_state_t *lrm_state = NULL; lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { return; } /* Erase from CIB */ op_key = pcmk__op_key(rsc_id, "last_failure", 0); if (operation) { orig_op_key = pcmk__op_key(rsc_id, operation, interval_ms); } erase_lrm_history_by_id(lrm_state, rsc_id, op_key, orig_op_key, 0); free(op_key); free(orig_op_key); /* Remove from memory */ if (lrm_state->resource_history) { rsc_history_t *entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (last_failed_matches_op(entry, operation, interval_ms)) { lrmd_free_event(entry->failed); entry->failed = NULL; } } } /* Returns: gboolean - cancellation is in progress */ static gboolean cancel_op(lrm_state_t * lrm_state, const char *rsc_id, const char *key, int op, gboolean remove) { int rc = pcmk_ok; char *local_key = NULL; active_op_t *pending = NULL; CRM_CHECK(op != 0, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); if (key == NULL) { local_key = make_stop_id(rsc_id, op); key = local_key; } pending = g_hash_table_lookup(lrm_state->pending_ops, key); if (pending) { if (remove && !pcmk_is_set(pending->flags, active_op_remove)) { controld_set_active_op_flags(pending, active_op_remove); crm_debug("Scheduling %s for removal", key); } if (pcmk_is_set(pending->flags, active_op_cancelled)) { crm_debug("Operation %s already cancelled", key); free(local_key); return FALSE; } controld_set_active_op_flags(pending, active_op_cancelled); } else { crm_info("No pending op found for %s", key); free(local_key); return FALSE; } crm_debug("Cancelling op %d for %s (%s)", op, rsc_id, key); rc = lrm_state_cancel(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms); if (rc == pcmk_ok) { crm_debug("Op %d for %s (%s): cancelled", op, rsc_id, key); free(local_key); return TRUE; } crm_debug("Op %d for %s (%s): Nothing to cancel", op, rsc_id, key); /* The caller needs to make sure the entry is * removed from the pending_ops list * * Usually by returning TRUE inside the worker function * supplied to g_hash_table_foreach_remove() * * Not removing the entry from pending_ops will block * the node from shutting down */ free(local_key); return FALSE; } struct cancel_data { gboolean done; gboolean remove; const char *key; lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean cancel_action_by_key(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct cancel_data *data = user_data; active_op_t *op = value; if (pcmk__str_eq(op->op_key, data->key, pcmk__str_none)) { data->done = TRUE; remove = !cancel_op(data->lrm_state, data->rsc->id, key, op->call_id, data->remove); } return remove; } static gboolean cancel_op_key(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *key, gboolean remove) { guint removed = 0; struct cancel_data data; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(key != NULL, return FALSE); data.key = key; data.rsc = rsc; data.done = FALSE; data.remove = remove; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->pending_ops, cancel_action_by_key, &data); crm_trace("Removed %u op cache entries, new size: %u", removed, g_hash_table_size(lrm_state->pending_ops)); return data.done; } /*! * \internal * \brief Retrieve resource information from LRM * * \param[in] lrm_state LRM connection to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource with LRM if not already * \param[out] rsc_info Where to store resource information obtained from LRM * * \retval pcmk_ok Success (and rsc_info holds newly allocated result) * \retval -EINVAL Required information is missing from arguments * \retval -ENOTCONN No active connection to LRM * \retval -ENODEV Resource not found * \retval -errno Error communicating with executor when registering resource * * \note Caller is responsible for freeing result on success. */ static int get_lrm_resource(lrm_state_t *lrm_state, xmlNode *rsc_xml, gboolean do_create, lrmd_rsc_info_t **rsc_info) { const char *id = ID(rsc_xml); CRM_CHECK(lrm_state && rsc_xml && rsc_info, return -EINVAL); CRM_CHECK(id, return -EINVAL); if (lrm_state_is_connected(lrm_state) == FALSE) { return -ENOTCONN; } crm_trace("Retrieving resource information for %s from the executor", id); *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); // If resource isn't known by ID, try clone name, if provided if (!*rsc_info) { const char *long_id = crm_element_value(rsc_xml, XML_ATTR_ID_LONG); if (long_id) { *rsc_info = lrm_state_get_rsc_info(lrm_state, long_id, 0); } } if ((*rsc_info == NULL) && do_create) { const char *class = crm_element_value(rsc_xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(rsc_xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(rsc_xml, XML_ATTR_TYPE); int rc; crm_trace("Registering resource %s with the executor", id); rc = lrm_state_register_rsc(lrm_state, id, class, provider, type, lrmd_opt_drop_recurring); if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Could not register resource %s with the executor on %s: %s " CRM_XS " rc=%d", id, lrm_state->node_name, pcmk_strerror(rc), rc); /* Register this as an internal error if this involves the local * executor. Otherwise, we're likely dealing with an unresponsive * remote node, which is not an FSA failure. */ if (lrm_state_is_local(lrm_state) == TRUE) { register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } return rc; } *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); } return *rsc_info? pcmk_ok : -ENODEV; } static void delete_resource(lrm_state_t * lrm_state, const char *id, lrmd_rsc_info_t * rsc, GHashTableIter * gIter, const char *sys, const char *user, ha_msg_input_t * request, gboolean unregister) { int rc = pcmk_ok; crm_info("Removing resource %s from executor for %s%s%s", id, sys, (user? " as " : ""), (user? user : "")); if (rsc && unregister) { rc = lrm_state_unregister_rsc(lrm_state, id, 0); } if (rc == pcmk_ok) { crm_trace("Resource %s deleted from executor", id); } else if (rc == -EINPROGRESS) { crm_info("Deletion of resource '%s' from executor is pending", id); if (request) { struct pending_deletion_op_s *op = NULL; char *ref = crm_element_value_copy(request->msg, XML_ATTR_REFERENCE); op = calloc(1, sizeof(struct pending_deletion_op_s)); op->rsc = strdup(rsc->id); op->input = copy_ha_msg_input(request); g_hash_table_insert(lrm_state->deletion_ops, ref, op); } return; } else { crm_warn("Could not delete '%s' from executor for %s%s%s: %s " CRM_XS " rc=%d", id, sys, (user? " as " : ""), (user? user : ""), pcmk_strerror(rc), rc); } delete_rsc_entry(lrm_state, request, id, gIter, rc, user); } static int get_fake_call_id(lrm_state_t *lrm_state, const char *rsc_id) { int call_id = 999999999; rsc_history_t *entry = NULL; if(lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* Make sure the call id is greater than the last successful operation, * otherwise the failure will not result in a possible recovery of the resource * as it could appear the failure occurred before the successful start */ if (entry) { call_id = entry->last_callid + 1; } if (call_id < 0) { call_id = 1; } return call_id; } static void fake_op_status(lrm_state_t *lrm_state, lrmd_event_data_t *op, int op_status, - enum ocf_exitcode op_exitcode) + enum ocf_exitcode op_exitcode, const char *exit_reason) { op->call_id = get_fake_call_id(lrm_state, op->rsc_id); op->t_run = time(NULL); op->t_rcchange = op->t_run; - op->op_status = op_status; - op->rc = op_exitcode; + lrmd__set_result(op, op_exitcode, op_status, exit_reason); } static void force_reprobe(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { GHashTableIter gIter; rsc_history_t *entry = NULL; crm_info("Clearing resource history on node %s", lrm_state->node_name); g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { /* only unregister the resource during a reprobe if it is not a remote connection * resource. otherwise unregistering the connection will terminate remote-node * membership */ gboolean unregister = TRUE; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state) { /* when forcing a reprobe, make sure to clear remote node before * clearing the remote node's connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE); } unregister = FALSE; } delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, user_name, NULL, unregister); } /* Now delete the copy in the CIB */ controld_delete_node_state(lrm_state->node_name, controld_section_lrm, cib_scope_local); /* Finally, _delete_ the value in pacemaker-attrd -- setting it to FALSE * would result in the scheduler sending us back here again */ update_attrd(lrm_state->node_name, CRM_OP_PROBED, NULL, user_name, is_remote_node); } /*! * \internal * \brief Fail a requested action without actually executing it * * For an action that can't be executed, process it similarly to an actual * execution result, with specified error status (except for notify actions, * which will always be treated as successful). * * \param[in] lrm_state Executor connection that action is for * \param[in] action Action XML from request * \param[in] rc Desired return code to use * \param[in] op_status Desired operation status to use + * \param[in] exit_reason Human-friendly detail, if error */ static void synthesize_lrmd_failure(lrm_state_t *lrm_state, xmlNode *action, - int op_status, enum ocf_exitcode rc) + int op_status, enum ocf_exitcode rc, + const char *exit_reason) { lrmd_event_data_t *op = NULL; const char *operation = crm_element_value(action, XML_LRM_ATTR_TASK); const char *target_node = crm_element_value(action, XML_LRM_ATTR_TARGET); xmlNode *xml_rsc = find_xml_node(action, XML_CIB_TAG_RESOURCE, TRUE); if ((xml_rsc == NULL) || (ID(xml_rsc) == NULL)) { /* @TODO Should we do something else, like direct ack? */ crm_info("Can't fake %s failure (%d) on %s without resource configuration", crm_element_value(action, XML_LRM_ATTR_TASK_KEY), rc, target_node); return; } else if(operation == NULL) { /* This probably came from crm_resource -C, nothing to do */ crm_info("Can't fake %s failure (%d) on %s without operation", ID(xml_rsc), rc, target_node); return; } op = construct_op(lrm_state, action, ID(xml_rsc), operation); if (pcmk__str_eq(operation, RSC_NOTIFY, pcmk__str_casei)) { // Notifications can't fail - fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_OK); + fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_OK, NULL); } else { - fake_op_status(lrm_state, op, op_status, rc); + fake_op_status(lrm_state, op, op_status, rc, exit_reason); } crm_info("Faking " PCMK__OP_FMT " result (%d) on %s", op->rsc_id, op->op_type, op->interval_ms, op->rc, target_node); // Process the result as if it came from the LRM process_lrm_event(lrm_state, op, NULL, action); lrmd_free_event(op); } /*! * \internal * \brief Get target of an LRM operation * * \param[in] xml LRM operation data XML * * \return LRM operation target node name (local node or Pacemaker Remote node) */ static const char * lrm_op_target(xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = fsa_our_uname; } return target; } static void fail_lrm_resource(xmlNode *xml, lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { lrmd_event_data_t *op = NULL; lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(xml, XML_CIB_TAG_RESOURCE, TRUE); CRM_CHECK(xml_rsc != NULL, return); /* The executor simply executes operations and reports the results, without * any concept of success or failure, so to fail a resource, we must fake * what a failure looks like. * * To do this, we create a fake executor operation event for the resource, * and pass that event to the executor client callback so it will be * processed as if it came from the executor. */ op = construct_op(lrm_state, xml, ID(xml_rsc), "asyncmon"); - fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_UNKNOWN_ERROR); free((char*) op->user_data); op->user_data = NULL; op->interval_ms = 0; if (user_name && !pcmk__is_privileged(user_name)) { crm_err("%s does not have permission to fail %s", user_name, ID(xml_rsc)); + fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, + PCMK_OCF_INSUFFICIENT_PRIV, + "Unprivileged user cannot fail resources"); controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); return; } + if (get_lrm_resource(lrm_state, xml_rsc, TRUE, &rsc) == pcmk_ok) { crm_info("Failing resource %s...", rsc->id); - op->exit_reason = strdup("Simulated failure"); + fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_UNKNOWN_ERROR, + "Simulated failure"); process_lrm_event(lrm_state, op, NULL, xml); - op->op_status = PCMK_EXEC_DONE; - op->rc = PCMK_OCF_OK; + op->rc = PCMK_OCF_OK; // The request to fail the resource succeeded lrmd_free_rsc_info(rsc); } else { crm_info("Cannot find/create resource in order to fail it..."); crm_log_xml_warn(xml, "bad input"); + fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, PCMK_OCF_UNKNOWN_ERROR, + "Cannot fail unknown resource"); } controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); } static void handle_refresh_op(lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { int rc = pcmk_ok; xmlNode *fragment = do_lrm_query_internal(lrm_state, node_update_all); fsa_cib_update(XML_CIB_TAG_STATUS, fragment, cib_quorum_override, rc, user_name); crm_info("Forced a local resource history refresh: call=%d", rc); if (!pcmk__str_eq(CRM_SYSTEM_CRMD, from_sys, pcmk__str_casei)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, fragment, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing refresh from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } free_xml(fragment); } static void handle_query_op(xmlNode *msg, lrm_state_t *lrm_state) { xmlNode *data = do_lrm_query_internal(lrm_state, node_update_all); xmlNode *reply = create_reply(msg, data); if (relay_message(reply, TRUE) == FALSE) { crm_err("Unable to route reply"); crm_log_xml_err(reply, "reply"); } free_xml(reply); free_xml(data); } static void handle_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node); if (!pcmk__strcase_any_of(from_sys, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, NULL, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing re-probe from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } } static bool do_lrm_cancel(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_host, const char *from_sys) { char *op_key = NULL; char *meta_key = NULL; int call = 0; const char *call_id = NULL; const char *op_task = NULL; guint interval_ms = 0; gboolean in_progress = FALSE; xmlNode *params = find_xml_node(input->xml, XML_TAG_ATTRS, TRUE); CRM_CHECK(params != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_TASK); op_task = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(op_task != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); if (crm_element_value_ms(params, meta_key, &interval_ms) != pcmk_ok) { free(meta_key); return FALSE; } free(meta_key); op_key = pcmk__op_key(rsc->id, op_task, interval_ms); meta_key = crm_meta_name(XML_LRM_ATTR_CALLID); call_id = crm_element_value(params, meta_key); free(meta_key); crm_debug("Scheduler requested op %s (call=%s) be cancelled", op_key, (call_id? call_id : "NA")); pcmk__scan_min_int(call_id, &call, 0); if (call == 0) { // Normal case when the scheduler cancels a recurring op in_progress = cancel_op_key(lrm_state, rsc, op_key, TRUE); } else { // Normal case when the scheduler cancels an orphan op in_progress = cancel_op(lrm_state, rsc->id, NULL, call, TRUE); } // Acknowledge cancellation operation if for a remote connection resource if (!in_progress || is_remote_lrmd_ra(NULL, NULL, rsc->id)) { char *op_id = make_stop_id(rsc->id, call); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) == FALSE) { crm_info("Nothing known about operation %d for %s", call, op_key); } erase_lrm_history_by_id(lrm_state, rsc->id, op_key, NULL, call); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); /* needed at least for cancellation of a remote operation */ g_hash_table_remove(lrm_state->pending_ops, op_id); free(op_id); } else { /* No ack is needed since abcdaa8, but peers with older versions * in a rolling upgrade need one. We didn't bump the feature set * at that commit, so we can only compare against the previous * CRM version (3.0.8). If any peers have feature set 3.0.9 but * not abcdaa8, they will time out waiting for the ack (no * released versions of Pacemaker are affected). */ const char *peer_version = crm_element_value(params, XML_ATTR_CRM_VERSION); if (compare_version(peer_version, "3.0.8") <= 0) { crm_info("Sending compatibility ack for %s cancellation to %s (CRM version %s)", op_key, from_host, peer_version); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); } } free(op_key); return TRUE; } static void do_lrm_delete(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_sys, const char *from_host, bool crm_rsc_delete, const char *user_name) { gboolean unregister = TRUE; int cib_rc = controld_delete_resource_history(rsc->id, lrm_state->node_name, user_name, cib_dryrun|cib_sync_call); if (cib_rc != pcmk_rc_ok) { lrmd_event_data_t *op = NULL; op = construct_op(lrm_state, input->xml, rsc->id, CRMD_ACTION_DELETE); - op->op_status = PCMK_EXEC_ERROR; - if (cib_rc == EACCES) { - op->rc = PCMK_OCF_INSUFFICIENT_PRIV; - } else { - op->rc = PCMK_OCF_UNKNOWN_ERROR; - } + /* These are resource clean-ups, not actions, so no exit reason is + * needed. + */ + lrmd__set_result(op, pcmk_rc2ocf(cib_rc), PCMK_EXEC_ERROR, NULL); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc->id); lrmd_free_event(op); return; } if (crm_rsc_delete && is_remote_lrmd_ra(NULL, NULL, rsc->id)) { unregister = FALSE; } delete_resource(lrm_state, rsc->id, rsc, NULL, from_sys, user_name, input, unregister); } /* A_LRM_INVOKE */ void do_lrm_invoke(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { lrm_state_t *lrm_state = NULL; const char *crm_op = NULL; const char *from_sys = NULL; const char *from_host = NULL; const char *operation = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *user_name = NULL; const char *target_node = NULL; gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; target_node = lrm_op_target(input->xml); is_remote_node = !pcmk__str_eq(target_node, fsa_our_uname, pcmk__str_casei); lrm_state = lrm_state_find(target_node); if ((lrm_state == NULL) && is_remote_node) { crm_err("Failing action because local node has never had connection to remote node %s", target_node); synthesize_lrmd_failure(NULL, input->xml, PCMK_EXEC_NOT_CONNECTED, - PCMK_OCF_UNKNOWN_ERROR); + PCMK_OCF_UNKNOWN_ERROR, + "Local node has no connection to remote"); return; } CRM_ASSERT(lrm_state != NULL); user_name = pcmk__update_acl_user(input->msg, F_CRM_USER, NULL); crm_op = crm_element_value(input->msg, F_CRM_TASK); from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } crm_trace("Executor %s command from %s as user %s", crm_op, from_sys, user_name); if (pcmk__str_eq(crm_op, CRM_OP_LRM_DELETE, pcmk__str_casei)) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { crm_rsc_delete = TRUE; // from crm_resource } operation = CRMD_ACTION_DELETE; } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_FAIL, pcmk__str_casei)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); return; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } if (pcmk__str_eq(crm_op, CRM_OP_LRM_REFRESH, pcmk__str_casei)) { handle_refresh_op(lrm_state, user_name, from_host, from_sys); } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_QUERY, pcmk__str_casei)) { handle_query_op(input->msg, lrm_state); } else if (pcmk__str_eq(operation, CRM_OP_PROBED, pcmk__str_casei)) { update_attrd(lrm_state->node_name, CRM_OP_PROBED, XML_BOOLEAN_TRUE, user_name, is_remote_node); } else if (pcmk__str_eq(crm_op, CRM_OP_REPROBE, pcmk__str_casei) || pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_casei)) { handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node); } else if (operation != NULL) { lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(input->xml, XML_CIB_TAG_RESOURCE, TRUE); gboolean create_rsc = !pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei); int rc; // We can't return anything meaningful without a resource ID CRM_CHECK(xml_rsc && ID(xml_rsc), return); rc = get_lrm_resource(lrm_state, xml_rsc, create_rsc, &rsc); if (rc == -ENOTCONN) { synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_NOT_CONNECTED, - PCMK_OCF_UNKNOWN_ERROR); + PCMK_OCF_UNKNOWN_ERROR, + "Not connected to remote executor"); return; } else if ((rc < 0) && !create_rsc) { /* Delete of malformed or nonexistent resource * (deleting something that does not exist is a success) */ crm_notice("Not registering resource '%s' for a %s event " CRM_XS " get-rc=%d (%s) transition-key=%s", ID(xml_rsc), operation, rc, pcmk_strerror(rc), ID(input->xml)); delete_rsc_entry(lrm_state, input, ID(xml_rsc), NULL, pcmk_ok, user_name); return; } else if (rc == -EINVAL) { // Resource operation on malformed resource crm_err("Invalid resource definition for %s", ID(xml_rsc)); crm_log_xml_warn(input->msg, "invalid resource"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, - PCMK_OCF_NOT_CONFIGURED); // fatal error + PCMK_OCF_NOT_CONFIGURED, // fatal error + "Invalid resource definition"); return; } else if (rc < 0) { // Error communicating with the executor crm_err("Could not register resource '%s' with executor: %s " CRM_XS " rc=%d", ID(xml_rsc), pcmk_strerror(rc), rc); crm_log_xml_warn(input->msg, "failed registration"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, - PCMK_OCF_INVALID_PARAM); // hard error + PCMK_OCF_INVALID_PARAM, // hard error + "Could not register resource with executor"); return; } if (pcmk__str_eq(operation, CRMD_ACTION_CANCEL, pcmk__str_casei)) { if (!do_lrm_cancel(input, lrm_state, rsc, from_host, from_sys)) { crm_log_xml_warn(input->xml, "Bad command"); } } else if (pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else if (pcmk__str_any_of(operation, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { /* Pre-2.1.0 DCs will schedule reload actions only, and 2.1.0+ DCs * will schedule reload-agent actions only. In either case, we need * to map that to whatever the resource agent actually supports. * Default to the OCF 1.1 name. */ struct ra_metadata_s *md = NULL; const char *reload_name = CRMD_ACTION_RELOAD_AGENT; md = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if ((md != NULL) && pcmk_is_set(md->ra_flags, ra_supports_legacy_reload)) { reload_name = CRMD_ACTION_RELOAD; } do_lrm_rsc_op(lrm_state, rsc, reload_name, input->xml); } else { do_lrm_rsc_op(lrm_state, rsc, operation, input->xml); } lrmd_free_rsc_info(rsc); } else { crm_err("Cannot perform operation %s of unknown type", crm_str(crm_op)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } #if ENABLE_VERSIONED_ATTRS static void resolve_versioned_parameters(lrm_state_t *lrm_state, const char *rsc_id, const xmlNode *rsc_op, GHashTable *params) { /* Resource info *should* already be cached, so we don't get * executor call */ lrmd_rsc_info_t *rsc = lrm_state_get_rsc_info(lrm_state, rsc_id, 0); struct ra_metadata_s *metadata; metadata = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if (metadata) { xmlNode *versioned_attrs = NULL; GHashTable *hash = NULL; char *key = NULL; char *value = NULL; GHashTableIter iter; versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_META); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_replace(params, crm_meta_name(key), strdup(value)); if (pcmk__str_eq(key, XML_ATTR_TIMEOUT, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->timeout, 0); } else if (pcmk__str_eq(key, XML_OP_ATTR_START_DELAY, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->start_delay, 0); } } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_RSC_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); } lrmd_free_rsc_info(rsc); } #endif static lrmd_event_data_t * construct_op(lrm_state_t *lrm_state, xmlNode *rsc_op, const char *rsc_id, const char *operation) { lrmd_event_data_t *op = NULL; const char *op_delay = NULL; const char *op_timeout = NULL; GHashTable *params = NULL; xmlNode *primitive = NULL; const char *class = NULL; const char *transition = NULL; CRM_ASSERT(rsc_id && operation); op = lrmd_new_event(rsc_id, operation, 0); op->type = lrmd_event_exec_complete; - op->op_status = PCMK_EXEC_PENDING; - op->rc = -1; op->timeout = 0; op->start_delay = 0; + lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); if (rsc_op == NULL) { CRM_LOG_ASSERT(pcmk__str_eq(CRMD_ACTION_STOP, operation, pcmk__str_casei)); op->user_data = NULL; /* the stop_all_resources() case * by definition there is no DC (or they'd be shutting * us down). * So we should put our version here. */ op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); crm_trace("Constructed %s op for %s", operation, rsc_id); return op; } params = xml2list(rsc_op); g_hash_table_remove(params, CRM_META "_op_target_rc"); op_delay = crm_meta_value(params, XML_OP_ATTR_START_DELAY); pcmk__scan_min_int(op_delay, &op->start_delay, 0); op_timeout = crm_meta_value(params, XML_ATTR_TIMEOUT); pcmk__scan_min_int(op_timeout, &op->timeout, 0); if (pcmk__guint_from_hash(params, CRM_META "_" XML_LRM_ATTR_INTERVAL_MS, 0, &(op->interval_ms)) != pcmk_rc_ok) { op->interval_ms = 0; } /* Use pcmk_monitor_timeout instead of meta timeout for stonith recurring monitor, if set */ primitive = find_xml_node(rsc_op, XML_CIB_TAG_RESOURCE, FALSE); class = crm_element_value(primitive, XML_AGENT_ATTR_CLASS); if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_fence_params) && pcmk__str_eq(operation, CRMD_ACTION_STATUS, pcmk__str_casei) && (op->interval_ms > 0)) { op_timeout = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (op_timeout != NULL) { op->timeout = crm_get_msec(op_timeout); } } #if ENABLE_VERSIONED_ATTRS if (lrm_state && !is_remote_lrmd_ra(NULL, NULL, rsc_id) && !pcmk__strcase_any_of(op_type, CRMD_ACTION_METADATA, CRMD_ACTION_DELETE, NULL)) { resolve_versioned_parameters(lrm_state, rsc_id, rsc_op, params); } #endif if (!pcmk__str_eq(operation, RSC_STOP, pcmk__str_casei)) { op->params = params; } else { rsc_history_t *entry = NULL; if (lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* If we do not have stop parameters cached, use * whatever we are given */ if (!entry || !entry->stop_params) { op->params = params; } else { /* Copy the cached parameter list so that we stop the resource * with the old attributes, not the new ones */ op->params = pcmk__strkey_table(free, free); g_hash_table_foreach(params, copy_meta_keys, op->params); g_hash_table_foreach(entry->stop_params, copy_instance_keys, op->params); g_hash_table_destroy(params); params = NULL; } } /* sanity */ if (op->timeout <= 0) { op->timeout = op->interval_ms; } if (op->start_delay < 0) { op->start_delay = 0; } transition = crm_element_value(rsc_op, XML_ATTR_TRANSITION_KEY); CRM_CHECK(transition != NULL, return op); op->user_data = strdup(transition); if (op->interval_ms != 0) { if (pcmk__strcase_any_of(operation, CRMD_ACTION_START, CRMD_ACTION_STOP, NULL)) { crm_err("Start and Stop actions cannot have an interval: %u", op->interval_ms); op->interval_ms = 0; } } crm_trace("Constructed %s op for %s: interval=%u", operation, rsc_id, op->interval_ms); return op; } /*! * \internal * \brief Send a (synthesized) event result * * Reply with a synthesized event result directly, as opposed to going through * the executor. * * \param[in] to_host Host to send result to * \param[in] to_sys IPC name to send result to (NULL for transition engine) * \param[in] rsc Type information about resource the result is for * \param[in] op Event with result to send * \param[in] rsc_id ID of resource the result is for */ void controld_ack_event_directly(const char *to_host, const char *to_sys, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, const char *rsc_id) { xmlNode *reply = NULL; xmlNode *update, *iter; crm_node_t *peer = NULL; CRM_CHECK(op != NULL, return); if (op->rsc_id == NULL) { CRM_ASSERT(rsc_id != NULL); op->rsc_id = strdup(rsc_id); } if (to_sys == NULL) { to_sys = CRM_SYSTEM_TENGINE; } peer = crm_get_peer(0, fsa_our_uname); update = create_node_state_update(peer, node_update_none, NULL, __func__); iter = create_xml_node(update, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, fsa_our_uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, fsa_our_uname, __func__); reply = create_request(CRM_OP_INVOKE_LRM, update, to_host, to_sys, CRM_SYSTEM_LRMD, NULL); crm_log_xml_trace(update, "[direct ACK]"); crm_debug("ACK'ing resource op " PCMK__OP_FMT " from %s: %s", op->rsc_id, op->op_type, op->interval_ms, op->user_data, crm_element_value(reply, XML_ATTR_REFERENCE)); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(update); free_xml(reply); } gboolean verify_stopped(enum crmd_fsa_state cur_state, int log_level) { gboolean res = TRUE; GList *lrm_state_list = lrm_state_get_list(); GList *state_entry; for (state_entry = lrm_state_list; state_entry != NULL; state_entry = state_entry->next) { lrm_state_t *lrm_state = state_entry->data; if (!lrm_state_verify_stopped(lrm_state, cur_state, log_level)) { /* keep iterating through all even when false is returned */ res = FALSE; } } controld_set_fsa_input_flags(R_SENT_RSC_STOP); g_list_free(lrm_state_list); lrm_state_list = NULL; return res; } struct stop_recurring_action_s { lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean stop_recurring_action_by_rsc(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct stop_recurring_action_s *event = user_data; active_op_t *op = value; if ((op->interval_ms != 0) && pcmk__str_eq(op->rsc_id, event->rsc->id, pcmk__str_none)) { crm_debug("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (char*)key); remove = !cancel_op(event->lrm_state, event->rsc->id, key, op->call_id, FALSE); } return remove; } static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; lrm_state_t *lrm_state = user_data; active_op_t *op = value; if (op->interval_ms != 0) { crm_info("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (const char *) key); remove = !cancel_op(lrm_state, op->rsc_id, key, op->call_id, FALSE); } return remove; } static void record_pending_op(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op) { const char *record_pending = NULL; CRM_CHECK(node_name != NULL, return); CRM_CHECK(rsc != NULL, return); CRM_CHECK(op != NULL, return); // Never record certain operation types as pending if ((op->op_type == NULL) || (op->params == NULL) || !controld_action_is_recordable(op->op_type)) { return; } // defaults to true record_pending = crm_meta_value(op->params, XML_OP_ATTR_PENDING); if (record_pending && !crm_is_true(record_pending)) { return; } op->call_id = -1; - op->op_status = PCMK_EXEC_PENDING; - op->rc = PCMK_OCF_UNKNOWN; + lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); op->t_run = time(NULL); op->t_rcchange = op->t_run; /* write a "pending" entry to the CIB, inhibit notification */ crm_debug("Recording pending op " PCMK__OP_FMT " on %s in the CIB", op->rsc_id, op->op_type, op->interval_ms, node_name); do_update_resource(node_name, rsc, op, 0); } static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg) { int rc; int call_id = 0; char *op_id = NULL; lrmd_event_data_t *op = NULL; lrmd_key_value_t *params = NULL; fsa_data_t *msg_data = NULL; const char *transition = NULL; gboolean stop_recurring = FALSE; - bool send_nack = FALSE; + const char *nack_reason = NULL; CRM_CHECK(rsc != NULL, return); CRM_CHECK(operation != NULL, return); if (msg != NULL) { transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (transition == NULL) { crm_log_xml_err(msg, "Missing transition number"); } } op = construct_op(lrm_state, msg, rsc->id, operation); CRM_CHECK(op != NULL, return); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) && (op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_MIGRATE) == 0) { /* pcmk remote connections are a special use case. * We never ever want to stop monitoring a connection resource until * the entire migration has completed. If the connection is unexpectedly * severed, even during a migration, this is an event we must detect.*/ stop_recurring = FALSE; } else if ((op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_STATUS) != 0 && strcmp(operation, CRMD_ACTION_NOTIFY) != 0) { /* stop any previous monitor operations before changing the resource state */ stop_recurring = TRUE; } if (stop_recurring == TRUE) { guint removed = 0; struct stop_recurring_action_s data; data.rsc = rsc; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_action_by_rsc, &data); if (removed) { crm_debug("Stopped %u recurring operation%s in preparation for " PCMK__OP_FMT, removed, pcmk__plural_s(removed), rsc->id, operation, op->interval_ms); } } /* now do the op */ crm_notice("Requesting local execution of %s operation for %s on %s " CRM_XS " transition_key=%s op_key=" PCMK__OP_FMT, crm_action_str(op->op_type, op->interval_ms), rsc->id, lrm_state->node_name, transition, rsc->id, operation, op->interval_ms); if (pcmk_is_set(fsa_input_register, R_SHUTDOWN) && pcmk__str_eq(operation, RSC_START, pcmk__str_casei)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); - send_nack = TRUE; + nack_reason = "Not attempting start due to shutdown in progress"; } else if (fsa_state != S_NOT_DC && fsa_state != S_POLICY_ENGINE /* Recalculating */ && fsa_state != S_TRANSITION_ENGINE && !pcmk__str_eq(operation, CRMD_ACTION_STOP, pcmk__str_casei)) { - send_nack = TRUE; + nack_reason = "Controller cannot attempt actions at this time"; } - if(send_nack) { + if (nack_reason != NULL) { crm_notice("Discarding attempt to perform action %s on %s in state %s (shutdown=%s)", operation, rsc->id, fsa_state2string(fsa_state), pcmk__btoa(pcmk_is_set(fsa_input_register, R_SHUTDOWN))); - op->rc = PCMK_OCF_UNKNOWN_ERROR; - op->op_status = PCMK_EXEC_INVALID; + lrmd__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_INVALID, + nack_reason); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); lrmd_free_event(op); free(op_id); return; } record_pending_op(lrm_state->node_name, rsc, op); op_id = pcmk__op_key(rsc->id, op->op_type, op->interval_ms); if (op->interval_ms > 0) { /* cancel it so we can then restart it without conflict */ cancel_op_key(lrm_state, rsc, op_id, FALSE); } if (op->params) { char *key = NULL; char *value = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, op->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { params = lrmd_key_value_add(params, key, value); } } rc = controld_execute_resource_agent(lrm_state, rsc->id, op->op_type, op->user_data, op->interval_ms, op->timeout, op->start_delay, params, &call_id); if (rc == pcmk_rc_ok) { /* record all operations so we can wait * for them to complete during shutdown */ char *call_id_s = make_stop_id(rsc->id, call_id); active_op_t *pending = NULL; pending = calloc(1, sizeof(active_op_t)); crm_trace("Recording pending op: %d - %s %s", call_id, op_id, call_id_s); pending->call_id = call_id; pending->interval_ms = op->interval_ms; pending->op_type = strdup(operation); pending->op_key = strdup(op_id); pending->rsc_id = strdup(rsc->id); pending->start_time = time(NULL); pending->user_data = op->user_data? strdup(op->user_data) : NULL; if (crm_element_value_epoch(msg, XML_CONFIG_ATTR_SHUTDOWN_LOCK, &(pending->lock_time)) != pcmk_ok) { pending->lock_time = 0; } g_hash_table_replace(lrm_state->pending_ops, call_id_s, pending); if ((op->interval_ms > 0) && (op->start_delay > START_DELAY_THRESHOLD)) { - int target_rc = 0; + int target_rc = PCMK_OCF_OK; crm_info("Faking confirmation of %s: execution postponed for over 5 minutes", op_id); decode_transition_key(op->user_data, NULL, NULL, NULL, &target_rc); - op->rc = target_rc; - op->op_status = PCMK_EXEC_DONE; + lrmd__set_result(op, target_rc, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); } pending->params = op->params; op->params = NULL; } else if (lrm_state_is_local(lrm_state)) { crm_err("Could not initiate %s action for resource %s locally: %s " CRM_XS " rc=%d", operation, rsc->id, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, - PCMK_OCF_UNKNOWN_ERROR); + PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } else { crm_err("Could not initiate %s action for resource %s remotely on %s: " "%s " CRM_XS " rc=%d", operation, rsc->id, lrm_state->node_name, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, - PCMK_OCF_UNKNOWN_ERROR); + PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); } free(op_id); lrmd_free_event(op); } int last_resource_update = 0; static void cib_rsc_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { switch (rc) { case pcmk_ok: case -pcmk_err_diff_failed: case -pcmk_err_diff_resync: crm_trace("Resource update %d complete: rc=%d", call_id, rc); break; default: crm_warn("Resource update %d failed: (rc=%d) %s", call_id, rc, pcmk_strerror(rc)); } if (call_id == last_resource_update) { last_resource_update = 0; trigger_fsa(); } } /* Only successful stops, and probes that found the resource inactive, get locks * recorded in the history. This ensures the resource stays locked to the node * until it is active there again after the node comes back up. */ static bool should_preserve_lock(lrmd_event_data_t *op) { if (!controld_shutdown_lock_enabled) { return false; } if (!strcmp(op->op_type, RSC_STOP) && (op->rc == PCMK_OCF_OK)) { return true; } if (!strcmp(op->op_type, RSC_STATUS) && (op->rc == PCMK_OCF_NOT_RUNNING)) { return true; } return false; } static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time) { /* */ int rc = pcmk_ok; xmlNode *update, *iter = NULL; int call_opt = crmd_cib_smart_opt(); const char *uuid = NULL; CRM_CHECK(op != NULL, return 0); iter = create_xml_node(iter, XML_CIB_TAG_STATUS); update = iter; iter = create_xml_node(iter, XML_CIB_TAG_STATE); if (pcmk__str_eq(node_name, fsa_our_uname, pcmk__str_casei)) { uuid = fsa_our_uuid; } else { /* remote nodes uuid and uname are equal */ uuid = node_name; crm_xml_add(iter, XML_NODE_IS_REMOTE, "true"); } CRM_LOG_ASSERT(uuid != NULL); if(uuid == NULL) { rc = -EINVAL; goto done; } crm_xml_add(iter, XML_ATTR_UUID, uuid); crm_xml_add(iter, XML_ATTR_UNAME, node_name); crm_xml_add(iter, XML_ATTR_ORIGIN, __func__); iter = create_xml_node(iter, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, node_name, __func__); if (rsc) { const char *container = NULL; crm_xml_add(iter, XML_ATTR_TYPE, rsc->type); crm_xml_add(iter, XML_AGENT_ATTR_CLASS, rsc->standard); crm_xml_add(iter, XML_AGENT_ATTR_PROVIDER, rsc->provider); if (lock_time != 0) { /* Actions on a locked resource should either preserve the lock by * recording it with the action result, or clear it. */ if (!should_preserve_lock(op)) { lock_time = 0; } crm_xml_add_ll(iter, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) lock_time); } if (op->params) { container = g_hash_table_lookup(op->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); } if (container) { crm_trace("Resource %s is a part of container resource %s", op->rsc_id, container); crm_xml_add(iter, XML_RSC_ATTR_CONTAINER, container); } } else { crm_warn("Resource %s no longer exists in the executor", op->rsc_id); controld_ack_event_directly(NULL, NULL, rsc, op, op->rsc_id); goto cleanup; } crm_log_xml_trace(update, __func__); /* make it an asynchronous call and be done with it * * Best case: * the resource state will be discovered during * the next signup or election. * * Bad case: * we are shutting down and there is no DC at the time, * but then why were we shutting down then anyway? * (probably because of an internal error) * * Worst case: * we get shot for having resources "running" that really weren't * * the alternative however means blocking here for too long, which * isn't acceptable */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, rc, NULL); if (rc > 0) { last_resource_update = rc; } done: /* the return code is a call number, not an error code */ crm_trace("Sent resource state update message: %d for %s=%u on %s", rc, op->op_type, op->interval_ms, op->rsc_id); fsa_register_cib_callback(rc, FALSE, NULL, cib_rsc_callback); cleanup: free_xml(update); return rc; } void do_lrm_event(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input cur_input, fsa_data_t * msg_data) { CRM_CHECK(FALSE, return); } static char * unescape_newlines(const char *string) { char *pch = NULL; char *ret = NULL; static const char *escaped_newline = "\\n"; if (!string) { return NULL; } ret = strdup(string); pch = strstr(ret, escaped_newline); while (pch != NULL) { /* Replace newline escape pattern with actual newline (and a space so we * don't have to shuffle the rest of the buffer) */ pch[0] = '\n'; pch[1] = ' '; pch = strstr(pch, escaped_newline); } return ret; } static bool did_lrm_rsc_op_fail(lrm_state_t *lrm_state, const char * rsc_id, const char * op_type, guint interval_ms) { rsc_history_t *entry = NULL; CRM_CHECK(lrm_state != NULL, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); CRM_CHECK(op_type != NULL, return FALSE); entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->failed == NULL) { return FALSE; } if (pcmk__str_eq(entry->failed->rsc_id, rsc_id, pcmk__str_none) && pcmk__str_eq(entry->failed->op_type, op_type, pcmk__str_casei) && entry->failed->interval_ms == interval_ms) { return TRUE; } return FALSE; } void process_lrm_event(lrm_state_t *lrm_state, lrmd_event_data_t *op, active_op_t *pending, xmlNode *action_xml) { char *op_id = NULL; char *op_key = NULL; int update_id = 0; gboolean remove = FALSE; gboolean removed = FALSE; bool need_direct_ack = FALSE; lrmd_rsc_info_t *rsc = NULL; const char *node_name = NULL; CRM_CHECK(op != NULL, return); CRM_CHECK(op->rsc_id != NULL, return); // Remap new status codes for older DCs if (compare_version(fsa_our_dc_version, "3.2.0") < 0) { switch (op->op_status) { case PCMK_EXEC_NOT_CONNECTED: - op->op_status = PCMK_EXEC_ERROR; - op->rc = PCMK_OCF_CONNECTION_DIED; + lrmd__set_result(op, PCMK_OCF_CONNECTION_DIED, + PCMK_EXEC_ERROR, op->exit_reason); break; case PCMK_EXEC_INVALID: - op->op_status = PCMK_EXEC_ERROR; - op->rc = CRM_DIRECT_NACK_RC; + lrmd__set_result(op, CRM_DIRECT_NACK_RC, PCMK_EXEC_ERROR, + op->exit_reason); break; default: break; } } op_id = make_stop_id(op->rsc_id, op->call_id); op_key = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); // Get resource info if available (from executor state or action XML) if (lrm_state) { rsc = lrm_state_get_rsc_info(lrm_state, op->rsc_id, 0); } if ((rsc == NULL) && action_xml) { xmlNode *xml = find_xml_node(action_xml, XML_CIB_TAG_RESOURCE, TRUE); const char *standard = crm_element_value(xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(xml, XML_ATTR_TYPE); if (standard && type) { crm_info("%s agent information not cached, using %s%s%s:%s from action XML", op->rsc_id, standard, (provider? ":" : ""), (provider? provider : ""), type); rsc = lrmd_new_rsc_info(op->rsc_id, standard, provider, type); } else { crm_err("Can't process %s result because %s agent information not cached or in XML", op_key, op->rsc_id); } } // Get node name if available (from executor state or action XML) if (lrm_state) { node_name = lrm_state->node_name; } else if (action_xml) { node_name = crm_element_value(action_xml, XML_LRM_ATTR_TARGET); } if(pending == NULL) { remove = TRUE; if (lrm_state) { pending = g_hash_table_lookup(lrm_state->pending_ops, op_id); } } if (op->op_status == PCMK_EXEC_ERROR) { switch(op->rc) { case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Leave it to the TE/scheduler to decide if this is an error op->op_status = PCMK_EXEC_DONE; break; default: /* Nothing to do */ break; } } if (op->op_status != PCMK_EXEC_CANCELLED) { /* We might not record the result, so directly acknowledge it to the * originator instead, so it doesn't time out waiting for the result * (especially important if part of a transition). */ need_direct_ack = TRUE; if (controld_action_is_recordable(op->op_type)) { if (node_name && rsc) { // We should record the result, and happily, we can update_id = do_update_resource(node_name, rsc, op, pending? pending->lock_time : 0); need_direct_ack = FALSE; } else if (op->rsc_deleted) { /* We shouldn't record the result (likely the resource was * refreshed, cleaned, or removed while this operation was * in flight). */ crm_notice("Not recording %s result in CIB because " "resource information was removed since it was initiated", op_key); } else { /* This shouldn't be possible; the executor didn't consider the * resource deleted, but we couldn't find resource or node * information. */ crm_err("Unable to record %s result in CIB: %s", op_key, (node_name? "No resource information" : "No node name")); } } } else if (op->interval_ms == 0) { /* A non-recurring operation was cancelled. Most likely, the * never-initiated action was removed from the executor's pending * operations list upon resource removal. */ need_direct_ack = TRUE; } else if (pending == NULL) { /* This recurring operation was cancelled, but was not pending. No * transition actions are waiting on it, nothing needs to be done. */ } else if (op->user_data == NULL) { /* This recurring operation was cancelled and pending, but we don't * have a transition key. This should never happen. */ crm_err("Recurring operation %s was cancelled without transition information", op_key); } else if (pcmk_is_set(pending->flags, active_op_remove)) { /* This recurring operation was cancelled (by us) and pending, and we * have been waiting for it to finish. */ if (lrm_state) { erase_lrm_history_by_op(lrm_state, op); } /* If the recurring operation had failed, the lrm_rsc_op is recorded as * "last_failure" which won't get erased from the cib given the logic on * purpose in erase_lrm_history_by_op(). So that the cancel action won't * have a chance to get confirmed by DC with process_op_deletion(). * Cluster transition would get stuck waiting for the remaining action * timer to time out. * * Directly acknowledge the cancel operation in this case. */ if (did_lrm_rsc_op_fail(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms)) { need_direct_ack = TRUE; } } else if (op->rsc_deleted) { /* This recurring operation was cancelled (but not by us, and the * executor does not have resource information, likely due to resource * cleanup, refresh, or removal) and pending. */ crm_debug("Recurring op %s was cancelled due to resource deletion", op_key); need_direct_ack = TRUE; } else { /* This recurring operation was cancelled (but not by us, likely by the * executor before stopping the resource) and pending. We don't need to * do anything special. */ } if (need_direct_ack) { controld_ack_event_directly(NULL, NULL, NULL, op, op->rsc_id); } if(remove == FALSE) { /* The caller will do this afterwards, but keep the logging consistent */ removed = TRUE; } else if (lrm_state && ((op->interval_ms == 0) || (op->op_status == PCMK_EXEC_CANCELLED))) { gboolean found = g_hash_table_remove(lrm_state->pending_ops, op_id); if (op->interval_ms != 0) { removed = TRUE; } else if (found) { removed = TRUE; crm_trace("Op %s (call=%d, stop-id=%s, remaining=%u): Confirmed", op_key, op->call_id, op_id, g_hash_table_size(lrm_state->pending_ops)); } } if (node_name == NULL) { node_name = "unknown node"; // for logging } switch (op->op_status) { case PCMK_EXEC_CANCELLED: crm_info("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s confirmed=%s", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, pcmk_exec_status_str(op->op_status), op->call_id, op_key, pcmk__btoa(removed)); break; case PCMK_EXEC_DONE: crm_notice("Result of %s operation for %s on %s: %s " CRM_XS " rc=%d call=%d key=%s confirmed=%s cib-update=%d", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, services_ocf_exitcode_str(op->rc), op->rc, op->call_id, op_key, pcmk__btoa(removed), update_id); break; case PCMK_EXEC_TIMEOUT: crm_err("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s timeout=%dms", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, pcmk_exec_status_str(op->op_status), op->call_id, op_key, op->timeout); break; default: crm_err("Result of %s operation for %s on %s: %s " CRM_XS " call=%d key=%s confirmed=%s status=%d cib-update=%d", crm_action_str(op->op_type, op->interval_ms), op->rsc_id, node_name, pcmk_exec_status_str(op->op_status), op->call_id, op_key, pcmk__btoa(removed), op->op_status, update_id); } if (op->output) { char *prefix = crm_strdup_printf("%s-" PCMK__OP_FMT ":%d", node_name, op->rsc_id, op->op_type, op->interval_ms, op->call_id); if (op->rc) { crm_log_output(LOG_NOTICE, prefix, op->output); } else { crm_log_output(LOG_DEBUG, prefix, op->output); } free(prefix); } if (lrm_state) { if (!pcmk__str_eq(op->op_type, RSC_METADATA, pcmk__str_casei)) { crmd_alert_resource_op(lrm_state->node_name, op); } else if (rsc && (op->rc == PCMK_OCF_OK)) { char *metadata = unescape_newlines(op->output); metadata_cache_update(lrm_state->metadata_cache, rsc, metadata); free(metadata); } } if (op->rsc_deleted) { crm_info("Deletion of resource '%s' complete after %s", op->rsc_id, op_key); if (lrm_state) { delete_rsc_entry(lrm_state, NULL, op->rsc_id, NULL, pcmk_ok, NULL); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ mainloop_set_trigger(fsa_source); if (lrm_state && rsc) { update_history_cache(lrm_state, rsc, op); } lrmd_free_rsc_info(rsc); free(op_key); free(op_id); } diff --git a/daemons/controld/controld_execd_state.c b/daemons/controld/controld_execd_state.c index 50e81ef9b1..a0129bd514 100644 --- a/daemons/controld/controld_execd_state.c +++ b/daemons/controld/controld_execd_state.c @@ -1,853 +1,853 @@ /* * Copyright 2012-2021 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 GHashTable *lrm_state_table = NULL; extern GHashTable *proxy_table; int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg); void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)); static void free_rsc_info(gpointer value) { lrmd_rsc_info_t *rsc_info = value; lrmd_free_rsc_info(rsc_info); } static void free_deletion_op(gpointer value) { struct pending_deletion_op_s *op = value; free(op->rsc); delete_ha_msg_input(op->input); free(op); } static void free_recurring_op(gpointer value) { active_op_t *op = value; free(op->user_data); free(op->rsc_id); free(op->op_type); free(op->op_key); if (op->params) { g_hash_table_destroy(op->params); } free(op); } static gboolean fail_pending_op(gpointer key, gpointer value, gpointer user_data) { lrmd_event_data_t event = { 0, }; lrm_state_t *lrm_state = user_data; active_op_t *op = value; crm_trace("Pre-emptively failing " PCMK__OP_FMT " on %s (call=%s, %s)", op->rsc_id, op->op_type, op->interval_ms, lrm_state->node_name, (char*)key, op->user_data); event.type = lrmd_event_exec_complete; event.rsc_id = op->rsc_id; event.op_type = op->op_type; event.user_data = op->user_data; event.timeout = 0; event.interval_ms = op->interval_ms; - event.rc = PCMK_OCF_UNKNOWN_ERROR; - event.op_status = PCMK_EXEC_NOT_CONNECTED; + lrmd__set_result(&event, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_NOT_CONNECTED, + "Action was pending when executor connection was dropped"); event.t_run = (unsigned int) op->start_time; event.t_rcchange = (unsigned int) op->start_time; event.call_id = op->call_id; event.remote_nodename = lrm_state->node_name; event.params = op->params; process_lrm_event(lrm_state, &event, op, NULL); return TRUE; } gboolean lrm_state_is_local(lrm_state_t *lrm_state) { if (lrm_state == NULL || fsa_our_uname == NULL) { return FALSE; } if (strcmp(lrm_state->node_name, fsa_our_uname) != 0) { return FALSE; } return TRUE; } lrm_state_t * lrm_state_create(const char *node_name) { lrm_state_t *state = NULL; if (!node_name) { crm_err("No node name given for lrm state object"); return NULL; } state = calloc(1, sizeof(lrm_state_t)); if (!state) { return NULL; } state->node_name = strdup(node_name); state->rsc_info_cache = pcmk__strkey_table(NULL, free_rsc_info); state->deletion_ops = pcmk__strkey_table(free, free_deletion_op); state->pending_ops = pcmk__strkey_table(free, free_recurring_op); state->resource_history = pcmk__strkey_table(NULL, history_free); state->metadata_cache = metadata_cache_new(); g_hash_table_insert(lrm_state_table, (char *)state->node_name, state); return state; } void lrm_state_destroy(const char *node_name) { g_hash_table_remove(lrm_state_table, node_name); } static gboolean remote_proxy_remove_by_node(gpointer key, gpointer value, gpointer user_data) { remote_proxy_t *proxy = value; const char *node_name = user_data; if (pcmk__str_eq(node_name, proxy->node_name, pcmk__str_casei)) { return TRUE; } return FALSE; } static void internal_lrm_state_destroy(gpointer data) { lrm_state_t *lrm_state = data; if (!lrm_state) { return; } crm_trace("Destroying proxy table %s with %d members", lrm_state->node_name, g_hash_table_size(proxy_table)); g_hash_table_foreach_remove(proxy_table, remote_proxy_remove_by_node, (char *) lrm_state->node_name); remote_ra_cleanup(lrm_state); lrmd_api_delete(lrm_state->conn); if (lrm_state->rsc_info_cache) { crm_trace("Destroying rsc info cache with %d members", g_hash_table_size(lrm_state->rsc_info_cache)); g_hash_table_destroy(lrm_state->rsc_info_cache); } if (lrm_state->resource_history) { crm_trace("Destroying history op cache with %d members", g_hash_table_size(lrm_state->resource_history)); g_hash_table_destroy(lrm_state->resource_history); } if (lrm_state->deletion_ops) { crm_trace("Destroying deletion op cache with %d members", g_hash_table_size(lrm_state->deletion_ops)); g_hash_table_destroy(lrm_state->deletion_ops); } if (lrm_state->pending_ops) { crm_trace("Destroying pending op cache with %d members", g_hash_table_size(lrm_state->pending_ops)); g_hash_table_destroy(lrm_state->pending_ops); } metadata_cache_free(lrm_state->metadata_cache); free((char *)lrm_state->node_name); free(lrm_state); } void lrm_state_reset_tables(lrm_state_t * lrm_state, gboolean reset_metadata) { if (lrm_state->resource_history) { crm_trace("Re-setting history op cache with %d members", g_hash_table_size(lrm_state->resource_history)); g_hash_table_remove_all(lrm_state->resource_history); } if (lrm_state->deletion_ops) { crm_trace("Re-setting deletion op cache with %d members", g_hash_table_size(lrm_state->deletion_ops)); g_hash_table_remove_all(lrm_state->deletion_ops); } if (lrm_state->pending_ops) { crm_trace("Re-setting pending op cache with %d members", g_hash_table_size(lrm_state->pending_ops)); g_hash_table_remove_all(lrm_state->pending_ops); } if (lrm_state->rsc_info_cache) { crm_trace("Re-setting rsc info cache with %d members", g_hash_table_size(lrm_state->rsc_info_cache)); g_hash_table_remove_all(lrm_state->rsc_info_cache); } if (reset_metadata) { metadata_cache_reset(lrm_state->metadata_cache); } } gboolean lrm_state_init_local(void) { if (lrm_state_table) { return TRUE; } lrm_state_table = pcmk__strikey_table(NULL, internal_lrm_state_destroy); if (!lrm_state_table) { return FALSE; } proxy_table = pcmk__strikey_table(NULL, remote_proxy_free); if (!proxy_table) { g_hash_table_destroy(lrm_state_table); lrm_state_table = NULL; return FALSE; } return TRUE; } void lrm_state_destroy_all(void) { if (lrm_state_table) { crm_trace("Destroying state table with %d members", g_hash_table_size(lrm_state_table)); g_hash_table_destroy(lrm_state_table); lrm_state_table = NULL; } if(proxy_table) { crm_trace("Destroying proxy table with %d members", g_hash_table_size(proxy_table)); g_hash_table_destroy(proxy_table); proxy_table = NULL; } } lrm_state_t * lrm_state_find(const char *node_name) { if (!node_name) { return NULL; } return g_hash_table_lookup(lrm_state_table, node_name); } lrm_state_t * lrm_state_find_or_create(const char *node_name) { lrm_state_t *lrm_state; lrm_state = g_hash_table_lookup(lrm_state_table, node_name); if (!lrm_state) { lrm_state = lrm_state_create(node_name); } return lrm_state; } GList * lrm_state_get_list(void) { return g_hash_table_get_values(lrm_state_table); } static remote_proxy_t * find_connected_proxy_by_node(const char * node_name) { GHashTableIter gIter; remote_proxy_t *proxy = NULL; CRM_CHECK(proxy_table != NULL, return NULL); g_hash_table_iter_init(&gIter, proxy_table); while (g_hash_table_iter_next(&gIter, NULL, (gpointer *) &proxy)) { if (proxy->source && pcmk__str_eq(node_name, proxy->node_name, pcmk__str_casei)) { return proxy; } } return NULL; } static void remote_proxy_disconnect_by_node(const char * node_name) { remote_proxy_t *proxy = NULL; CRM_CHECK(proxy_table != NULL, return); while ((proxy = find_connected_proxy_by_node(node_name)) != NULL) { /* mainloop_del_ipc_client() eventually calls remote_proxy_disconnected() * , which removes the entry from proxy_table. * Do not do this in a g_hash_table_iter_next() loop. */ if (proxy->source) { mainloop_del_ipc_client(proxy->source); } } return; } void lrm_state_disconnect_only(lrm_state_t * lrm_state) { int removed = 0; if (!lrm_state->conn) { return; } crm_trace("Disconnecting %s", lrm_state->node_name); remote_proxy_disconnect_by_node(lrm_state->node_name); ((lrmd_t *) lrm_state->conn)->cmds->disconnect(lrm_state->conn); if (!pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { removed = g_hash_table_foreach_remove(lrm_state->pending_ops, fail_pending_op, lrm_state); crm_trace("Synthesized %d operation failures for %s", removed, lrm_state->node_name); } } void lrm_state_disconnect(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return; } lrm_state_disconnect_only(lrm_state); lrmd_api_delete(lrm_state->conn); lrm_state->conn = NULL; } int lrm_state_is_connected(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return FALSE; } return ((lrmd_t *) lrm_state->conn)->cmds->is_connected(lrm_state->conn); } int lrm_state_poke_connection(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return -ENOTCONN; } return ((lrmd_t *) lrm_state->conn)->cmds->poke_connection(lrm_state->conn); } // \return Standard Pacemaker return code int controld_connect_local_executor(lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; if (lrm_state->conn == NULL) { lrmd_t *api = NULL; rc = lrmd__new(&api, NULL, NULL, 0); if (rc != pcmk_rc_ok) { return rc; } api->cmds->set_callback(api, lrm_op_callback); lrm_state->conn = api; } rc = ((lrmd_t *) lrm_state->conn)->cmds->connect(lrm_state->conn, CRM_SYSTEM_CRMD, NULL); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { lrm_state->num_lrm_register_fails = 0; } else { lrm_state->num_lrm_register_fails++; } return rc; } static remote_proxy_t * crmd_remote_proxy_new(lrmd_t *lrmd, const char *node_name, const char *session_id, const char *channel) { struct ipc_client_callbacks proxy_callbacks = { .dispatch = remote_proxy_dispatch, .destroy = remote_proxy_disconnected }; remote_proxy_t *proxy = remote_proxy_new(lrmd, &proxy_callbacks, node_name, session_id, channel); return proxy; } gboolean crmd_is_proxy_session(const char *session) { return g_hash_table_lookup(proxy_table, session) ? TRUE : FALSE; } void crmd_proxy_send(const char *session, xmlNode *msg) { remote_proxy_t *proxy = g_hash_table_lookup(proxy_table, session); lrm_state_t *lrm_state = NULL; if (!proxy) { return; } crm_log_xml_trace(msg, "to-proxy"); lrm_state = lrm_state_find(proxy->node_name); if (lrm_state) { crm_trace("Sending event to %.8s on %s", proxy->session_id, proxy->node_name); remote_proxy_relay_event(proxy, msg); } } static void crmd_proxy_dispatch(const char *session, xmlNode *msg) { crm_trace("Processing proxied IPC message from session %s", session); crm_log_xml_trace(msg, "controller[inbound]"); crm_xml_add(msg, F_CRM_SYS_FROM, session); if (controld_authorize_ipc_message(msg, NULL, session)) { route_message(C_IPC_MESSAGE, msg); } trigger_fsa(); } static void remote_config_check(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc != pcmk_ok) { crm_err("Query resulted in an error: %s", pcmk_strerror(rc)); if (rc == -EACCES || rc == -pcmk_err_schema_validation) { crm_err("The cluster is mis-configured - shutting down and staying down"); } } else { lrmd_t * lrmd = (lrmd_t *)user_data; crm_time_t *now = crm_time_new(NULL); GHashTable *config_hash = pcmk__strkey_table(free, free); crm_debug("Call %d : Parsing CIB options", call_id); pe_unpack_nvpairs(output, output, XML_CIB_TAG_PROPSET, NULL, config_hash, CIB_OPTIONS_FIRST, FALSE, now, NULL); /* Now send it to the remote peer */ lrmd__validate_remote_settings(lrmd, config_hash); g_hash_table_destroy(config_hash); crm_time_free(now); } } static void crmd_remote_proxy_cb(lrmd_t *lrmd, void *userdata, xmlNode *msg) { lrm_state_t *lrm_state = userdata; const char *session = crm_element_value(msg, F_LRMD_IPC_SESSION); remote_proxy_t *proxy = g_hash_table_lookup(proxy_table, session); const char *op = crm_element_value(msg, F_LRMD_IPC_OP); if (pcmk__str_eq(op, LRMD_IPC_OP_NEW, pcmk__str_casei)) { const char *channel = crm_element_value(msg, F_LRMD_IPC_IPC_SERVER); proxy = crmd_remote_proxy_new(lrmd, lrm_state->node_name, session, channel); if (!remote_ra_controlling_guest(lrm_state)) { if (proxy != NULL) { /* Look up stonith-watchdog-timeout and send to the remote peer for validation */ int rc = fsa_cib_conn->cmds->query(fsa_cib_conn, XML_CIB_TAG_CRMCONFIG, NULL, cib_scope_local); fsa_cib_conn->cmds->register_callback_full(fsa_cib_conn, rc, 10, FALSE, lrmd, "remote_config_check", remote_config_check, NULL); } } else { crm_debug("Skipping remote_config_check for guest-nodes"); } } else if (pcmk__str_eq(op, LRMD_IPC_OP_SHUTDOWN_REQ, pcmk__str_casei)) { char *now_s = NULL; crm_notice("%s requested shutdown of its remote connection", lrm_state->node_name); if (!remote_ra_is_in_maintenance(lrm_state)) { now_s = pcmk__ttoa(time(NULL)); update_attrd(lrm_state->node_name, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, TRUE); free(now_s); remote_proxy_ack_shutdown(lrmd); crm_warn("Reconnection attempts to %s may result in failures that must be cleared", lrm_state->node_name); } else { remote_proxy_nack_shutdown(lrmd); crm_notice("Remote resource for %s is not managed so no ordered shutdown happening", lrm_state->node_name); } return; } else if (pcmk__str_eq(op, LRMD_IPC_OP_REQUEST, pcmk__str_casei) && proxy && proxy->is_local) { /* This is for the controller, which we are, so don't try * to send to ourselves over IPC -- do it directly. */ int flags = 0; xmlNode *request = get_message_xml(msg, F_LRMD_IPC_MSG); CRM_CHECK(request != NULL, return); CRM_CHECK(lrm_state->node_name, return); crm_xml_add(request, XML_ACL_TAG_ROLE, "pacemaker-remote"); pcmk__update_acl_user(request, F_LRMD_IPC_USER, lrm_state->node_name); /* Pacemaker Remote nodes don't know their own names (as known to the * cluster). When getting a node info request with no name or ID, add * the name, so we don't return info for ourselves instead of the * Pacemaker Remote node. */ if (pcmk__str_eq(crm_element_value(request, F_CRM_TASK), CRM_OP_NODE_INFO, pcmk__str_casei)) { int node_id = 0; crm_element_value_int(request, XML_ATTR_ID, &node_id); if ((node_id <= 0) && (crm_element_value(request, XML_ATTR_UNAME) == NULL)) { crm_xml_add(request, XML_ATTR_UNAME, lrm_state->node_name); } } crmd_proxy_dispatch(session, request); crm_element_value_int(msg, F_LRMD_IPC_MSG_FLAGS, &flags); if (flags & crm_ipc_client_response) { int msg_id = 0; xmlNode *op_reply = create_xml_node(NULL, "ack"); crm_xml_add(op_reply, "function", __func__); crm_xml_add_int(op_reply, "line", __LINE__); crm_element_value_int(msg, F_LRMD_IPC_MSG_ID, &msg_id); remote_proxy_relay_response(proxy, op_reply, msg_id); free_xml(op_reply); } } else { remote_proxy_cb(lrmd, lrm_state->node_name, msg); } } // \return Standard Pacemaker return code int controld_connect_remote_executor(lrm_state_t *lrm_state, const char *server, int port, int timeout_ms) { int rc = pcmk_rc_ok; if (lrm_state->conn == NULL) { lrmd_t *api = NULL; rc = lrmd__new(&api, lrm_state->node_name, server, port); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%s failed: %s " CRM_XS " rc=%d", server, port, pcmk_rc_str(rc), rc); return rc; } lrm_state->conn = api; api->cmds->set_callback(api, remote_lrm_op_callback); lrmd_internal_set_proxy_callback(api, lrm_state, crmd_remote_proxy_cb); } crm_trace("Initiating remote connection to %s:%d with timeout %dms", server, port, timeout_ms); rc = ((lrmd_t *) lrm_state->conn)->cmds->connect_async(lrm_state->conn, lrm_state->node_name, timeout_ms); if (rc == pcmk_ok) { lrm_state->num_lrm_register_fails = 0; } else { lrm_state->num_lrm_register_fails++; // Ignored for remote connections } return pcmk_legacy2rc(rc); } int lrm_state_get_metadata(lrm_state_t * lrm_state, const char *class, const char *provider, const char *agent, char **output, enum lrmd_call_options options) { lrmd_key_value_t *params = NULL; if (!lrm_state->conn) { return -ENOTCONN; } /* Add the node name to the environment, as is done with normal resource * action calls. Meta-data calls shouldn't need it, but some agents are * written with an ocf_local_nodename call at the beginning regardless of * action. Without the environment variable, the agent would try to contact * the controller to get the node name -- but the controller would be * blocking on the synchronous meta-data call. * * At this point, we have to assume that agents are unlikely to make other * calls that require the controller, such as crm_node --quorum or * --cluster-id. * * @TODO Make meta-data calls asynchronous. (This will be part of a larger * project to make meta-data calls via the executor rather than directly.) */ params = lrmd_key_value_add(params, CRM_META "_" XML_LRM_ATTR_TARGET, lrm_state->node_name); return ((lrmd_t *) lrm_state->conn)->cmds->get_metadata_params(lrm_state->conn, class, provider, agent, output, options, params); } int lrm_state_cancel(lrm_state_t *lrm_state, const char *rsc_id, const char *action, guint interval_ms) { if (!lrm_state->conn) { return -ENOTCONN; } /* Figure out a way to make this async? * NOTICE: Currently it's synced and directly acknowledged in do_lrm_invoke(). */ if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { return remote_ra_cancel(lrm_state, rsc_id, action, interval_ms); } return ((lrmd_t *) lrm_state->conn)->cmds->cancel(lrm_state->conn, rsc_id, action, interval_ms); } lrmd_rsc_info_t * lrm_state_get_rsc_info(lrm_state_t * lrm_state, const char *rsc_id, enum lrmd_call_options options) { lrmd_rsc_info_t *rsc = NULL; if (!lrm_state->conn) { return NULL; } if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { return remote_ra_get_rsc_info(lrm_state, rsc_id); } rsc = g_hash_table_lookup(lrm_state->rsc_info_cache, rsc_id); if (rsc == NULL) { /* only contact the lrmd if we don't already have a cached rsc info */ rsc = ((lrmd_t *) lrm_state->conn)->cmds->get_rsc_info(lrm_state->conn, rsc_id, options); if (rsc == NULL) { return NULL; } /* cache the result */ g_hash_table_insert(lrm_state->rsc_info_cache, rsc->id, rsc); } return lrmd_copy_rsc_info(rsc); } /*! * \internal * \brief Initiate a resource agent action * * \param[in] lrm_state Executor state object * \param[in] rsc_id ID of resource for action * \param[in] action Action to execute * \param[in] userdata String to copy and pass to execution callback * \param[in] interval_ms Action interval (in milliseconds) * \param[in] timeout_ms Action timeout (in milliseconds) * \param[in] start_delay_ms Delay (in milliseconds) before initiating action * \param[in] params Resource parameters * \param[out] call_id Where to store call ID on success * * \return Standard Pacemaker return code * \note This takes ownership of \p params, which should not be used or freed * after calling this function. */ int controld_execute_resource_agent(lrm_state_t *lrm_state, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout_ms, int start_delay_ms, lrmd_key_value_t *params, int *call_id) { int rc = pcmk_rc_ok; if (lrm_state->conn == NULL) { lrmd_key_value_freeall(params); rc = ENOTCONN; } else if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { rc = controld_execute_remote_agent(lrm_state, rsc_id, action, userdata, interval_ms, timeout_ms, start_delay_ms, params, call_id); } else { rc = ((lrmd_t *) lrm_state->conn)->cmds->exec(lrm_state->conn, rsc_id, action, userdata, interval_ms, timeout_ms, start_delay_ms, lrmd_opt_notify_changes_only, params); if (rc < 0) { rc = pcmk_legacy2rc(rc); } else { *call_id = rc; rc = pcmk_rc_ok; } } return rc; } int lrm_state_register_rsc(lrm_state_t * lrm_state, const char *rsc_id, const char *class, const char *provider, const char *agent, enum lrmd_call_options options) { lrmd_t *conn = (lrmd_t *) lrm_state->conn; if (conn == NULL) { return -ENOTCONN; } if (is_remote_lrmd_ra(agent, provider, NULL)) { return lrm_state_find_or_create(rsc_id)? pcmk_ok : -EINVAL; } /* @TODO Implement an asynchronous version of this (currently a blocking * call to the lrmd). */ return conn->cmds->register_rsc(lrm_state->conn, rsc_id, class, provider, agent, options); } int lrm_state_unregister_rsc(lrm_state_t * lrm_state, const char *rsc_id, enum lrmd_call_options options) { if (!lrm_state->conn) { return -ENOTCONN; } if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { lrm_state_destroy(rsc_id); return pcmk_ok; } g_hash_table_remove(lrm_state->rsc_info_cache, rsc_id); /* @TODO Optimize this ... this function is a blocking round trip from * client to daemon. The controld_execd_state.c code path that uses this * function should always treat it as an async operation. The executor API * should make an async version available. */ return ((lrmd_t *) lrm_state->conn)->cmds->unregister_rsc(lrm_state->conn, rsc_id, options); } /* * Functions for sending alerts via local executor connection */ static GList *crmd_alert_list = NULL; void crmd_unpack_alerts(xmlNode *alerts) { pe_free_alert_list(crmd_alert_list); crmd_alert_list = pe_unpack_alerts(alerts); } void crmd_alert_node_event(crm_node_t *node) { lrm_state_t *lrm_state; if (crmd_alert_list == NULL) { return; } lrm_state = lrm_state_find(fsa_our_uname); if (lrm_state == NULL) { return; } lrmd_send_node_alert((lrmd_t *) lrm_state->conn, crmd_alert_list, node->uname, node->id, node->state); } void crmd_alert_fencing_op(stonith_event_t * e) { char *desc; lrm_state_t *lrm_state; if (crmd_alert_list == NULL) { return; } lrm_state = lrm_state_find(fsa_our_uname); if (lrm_state == NULL) { return; } desc = crm_strdup_printf("Operation %s of %s by %s for %s@%s: %s (ref=%s)", e->action, e->target, (e->executioner? e->executioner : ""), e->client_origin, e->origin, pcmk_strerror(e->result), e->id); lrmd_send_fencing_alert((lrmd_t *) lrm_state->conn, crmd_alert_list, e->target, e->operation, desc, e->result); free(desc); } void crmd_alert_resource_op(const char *node, lrmd_event_data_t * op) { lrm_state_t *lrm_state; if (crmd_alert_list == NULL) { return; } lrm_state = lrm_state_find(fsa_our_uname); if (lrm_state == NULL) { return; } lrmd_send_resource_alert((lrmd_t *) lrm_state->conn, crmd_alert_list, node, op); } diff --git a/daemons/controld/controld_remote_ra.c b/daemons/controld/controld_remote_ra.c index fb9cdb3c12..74cbfd6732 100644 --- a/daemons/controld/controld_remote_ra.c +++ b/daemons/controld/controld_remote_ra.c @@ -1,1344 +1,1340 @@ /* * Copyright 2013-2021 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 #define REMOTE_LRMD_RA "remote" /* The max start timeout before cmd retry */ #define MAX_START_TIMEOUT_MS 10000 typedef struct remote_ra_cmd_s { /*! the local node the cmd is issued from */ char *owner; /*! the remote node the cmd is executed on */ char *rsc_id; /*! the action to execute */ char *action; /*! some string the client wants us to give it back */ char *userdata; - char *exit_reason; // descriptive text on error /*! start delay in ms */ int start_delay; /*! timer id used for start delay. */ int delay_id; /*! timeout in ms for cmd */ int timeout; int remaining_timeout; /*! recurring interval in ms */ guint interval_ms; /*! interval timer id */ int interval_id; int reported_success; int monitor_timeout_id; int takeover_timeout_id; /*! action parameters */ lrmd_key_value_t *params; - /*! executed rc */ - int rc; - int op_status; + pcmk__action_result_t result; int call_id; time_t start_time; gboolean cancel; } remote_ra_cmd_t; enum remote_migration_status { expect_takeover = 1, takeover_complete, }; typedef struct remote_ra_data_s { crm_trigger_t *work; remote_ra_cmd_t *cur_cmd; GList *cmds; GList *recurring_cmds; enum remote_migration_status migrate_status; gboolean active; /* Maintenance mode is difficult to determine from the controller's context, * so we have it signalled back with the transition from the scheduler. */ gboolean is_maintenance; /* Similar for whether we are controlling a guest node or remote node. * Fortunately there is a meta-attribute in the transition already and * as the situation doesn't change over time we can use the * resource start for noting down the information for later use when * the attributes aren't at hand. */ gboolean controlling_guest; } remote_ra_data_t; static int handle_remote_ra_start(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd, int timeout_ms); static void handle_remote_ra_stop(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd); static GList *fail_all_monitor_cmds(GList * list); static void free_cmd(gpointer user_data) { remote_ra_cmd_t *cmd = user_data; if (!cmd) { return; } if (cmd->delay_id) { g_source_remove(cmd->delay_id); } if (cmd->interval_id) { g_source_remove(cmd->interval_id); } if (cmd->monitor_timeout_id) { g_source_remove(cmd->monitor_timeout_id); } if (cmd->takeover_timeout_id) { g_source_remove(cmd->takeover_timeout_id); } free(cmd->owner); free(cmd->rsc_id); free(cmd->action); free(cmd->userdata); - free(cmd->exit_reason); + pcmk__reset_result(&(cmd->result)); lrmd_key_value_freeall(cmd->params); free(cmd); } static int generate_callid(void) { static int remote_ra_callid = 0; remote_ra_callid++; if (remote_ra_callid <= 0) { remote_ra_callid = 1; } return remote_ra_callid; } static gboolean recurring_helper(gpointer data) { remote_ra_cmd_t *cmd = data; lrm_state_t *connection_rsc = NULL; cmd->interval_id = 0; connection_rsc = lrm_state_find(cmd->rsc_id); if (connection_rsc && connection_rsc->remote_ra_data) { remote_ra_data_t *ra_data = connection_rsc->remote_ra_data; ra_data->recurring_cmds = g_list_remove(ra_data->recurring_cmds, cmd); ra_data->cmds = g_list_append(ra_data->cmds, cmd); mainloop_set_trigger(ra_data->work); } return FALSE; } static gboolean start_delay_helper(gpointer data) { remote_ra_cmd_t *cmd = data; lrm_state_t *connection_rsc = NULL; cmd->delay_id = 0; connection_rsc = lrm_state_find(cmd->rsc_id); if (connection_rsc && connection_rsc->remote_ra_data) { remote_ra_data_t *ra_data = connection_rsc->remote_ra_data; mainloop_set_trigger(ra_data->work); } return FALSE; } /*! * \internal * \brief Handle cluster communication related to pacemaker_remote node joining * * \param[in] node_name Name of newly integrated pacemaker_remote node */ static void remote_node_up(const char *node_name) { int call_opt, call_id = 0; xmlNode *update, *state; crm_node_t *node; enum controld_section_e section = controld_section_all; CRM_CHECK(node_name != NULL, return); crm_info("Announcing Pacemaker Remote node %s", node_name); /* Clear node's entire state (resource history and transient attributes) * other than shutdown locks. The transient attributes should and normally * will be cleared when the node leaves, but since remote node state has a * number of corner cases, clear them here as well, to be sure. */ call_opt = crmd_cib_smart_opt(); if (controld_shutdown_lock_enabled) { section = controld_section_all_unlocked; } /* Purge node from attrd's memory */ update_attrd_remote_node_removed(node_name, NULL); controld_delete_node_state(node_name, section, call_opt); /* Clear node's probed attribute */ update_attrd(node_name, CRM_OP_PROBED, NULL, NULL, TRUE); /* Ensure node is in the remote peer cache with member status */ node = crm_remote_peer_get(node_name); CRM_CHECK(node != NULL, return); pcmk__update_peer_state(__func__, node, CRM_NODE_MEMBER, 0); /* pacemaker_remote nodes don't participate in the membership layer, * so cluster nodes don't automatically get notified when they come and go. * We send a cluster message to the DC, and update the CIB node state entry, * so the DC will get it sooner (via message) or later (via CIB refresh), * and any other interested parties can query the CIB. */ send_remote_state_message(node_name, TRUE); update = create_xml_node(NULL, XML_CIB_TAG_STATUS); state = create_node_state_update(node, node_update_cluster, update, __func__); /* Clear the XML_NODE_IS_FENCED flag in the node state. If the node ever * needs to be fenced, this flag will allow various actions to determine * whether the fencing has happened yet. */ crm_xml_add(state, XML_NODE_IS_FENCED, "0"); /* TODO: If the remote connection drops, and this (async) CIB update either * failed or has not yet completed, later actions could mistakenly think the * node has already been fenced (if the XML_NODE_IS_FENCED attribute was * previously set, because it won't have been cleared). This could prevent * actual fencing or allow recurring monitor failures to be cleared too * soon. Ideally, we wouldn't rely on the CIB for the fenced status. */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_WARNING, "%s CIB node state setup", node_name); } free_xml(update); } enum down_opts { DOWN_KEEP_LRM, DOWN_ERASE_LRM }; /*! * \internal * \brief Handle cluster communication related to pacemaker_remote node leaving * * \param[in] node_name Name of lost node * \param[in] opts Whether to keep or erase LRM history */ static void remote_node_down(const char *node_name, const enum down_opts opts) { xmlNode *update; int call_id = 0; int call_opt = crmd_cib_smart_opt(); crm_node_t *node; /* Purge node from attrd's memory */ update_attrd_remote_node_removed(node_name, NULL); /* Normally, only node attributes should be erased, and the resource history * should be kept until the node comes back up. However, after a successful * fence, we want to clear the history as well, so we don't think resources * are still running on the node. */ if (opts == DOWN_ERASE_LRM) { controld_delete_node_state(node_name, controld_section_all, call_opt); } else { controld_delete_node_state(node_name, controld_section_attrs, call_opt); } /* Ensure node is in the remote peer cache with lost state */ node = crm_remote_peer_get(node_name); CRM_CHECK(node != NULL, return); pcmk__update_peer_state(__func__, node, CRM_NODE_LOST, 0); /* Notify DC */ send_remote_state_message(node_name, FALSE); /* Update CIB node state */ update = create_xml_node(NULL, XML_CIB_TAG_STATUS); create_node_state_update(node, node_update_cluster, update, __func__); fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_ERR, "%s CIB node state update", node_name); } free_xml(update); } /*! * \internal * \brief Handle effects of a remote RA command on node state * * \param[in] cmd Completed remote RA command */ static void check_remote_node_state(remote_ra_cmd_t *cmd) { /* Only successful actions can change node state */ - if (cmd->rc != PCMK_OCF_OK) { + if (cmd->result.exit_status != PCMK_OCF_OK) { return; } if (pcmk__str_eq(cmd->action, "start", pcmk__str_casei)) { remote_node_up(cmd->rsc_id); } else if (pcmk__str_eq(cmd->action, "migrate_from", pcmk__str_casei)) { /* After a successful migration, we don't need to do remote_node_up() * because the DC already knows the node is up, and we don't want to * clear LRM history etc. We do need to add the remote node to this * host's remote peer cache, because (unless it happens to be DC) * it hasn't been tracking the remote node, and other code relies on * the cache to distinguish remote nodes from unseen cluster nodes. */ crm_node_t *node = crm_remote_peer_get(cmd->rsc_id); CRM_CHECK(node != NULL, return); pcmk__update_peer_state(__func__, node, CRM_NODE_MEMBER, 0); } else if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { lrm_state_t *lrm_state = lrm_state_find(cmd->rsc_id); remote_ra_data_t *ra_data = lrm_state? lrm_state->remote_ra_data : NULL; if (ra_data) { if (ra_data->migrate_status != takeover_complete) { /* Stop means down if we didn't successfully migrate elsewhere */ remote_node_down(cmd->rsc_id, DOWN_KEEP_LRM); } else if (AM_I_DC == FALSE) { /* Only the connection host and DC track node state, * so if the connection migrated elsewhere and we aren't DC, * un-cache the node, so we don't have stale info */ crm_remote_peer_cache_remove(cmd->rsc_id); } } } /* We don't do anything for successful monitors, which is correct for * routine recurring monitors, and for monitors on nodes where the * connection isn't supposed to be (the cluster will stop the connection in * that case). However, if the initial probe finds the connection already * active on the node where we want it, we probably should do * remote_node_up(). Unfortunately, we can't distinguish that case here. * Given that connections have to be initiated by the cluster, the chance of * that should be close to zero. */ } static void report_remote_ra_result(remote_ra_cmd_t * cmd) { lrmd_event_data_t op = { 0, }; check_remote_node_state(cmd); op.type = lrmd_event_exec_complete; op.rsc_id = cmd->rsc_id; op.op_type = cmd->action; op.user_data = cmd->userdata; - op.exit_reason = cmd->exit_reason; op.timeout = cmd->timeout; op.interval_ms = cmd->interval_ms; - op.rc = cmd->rc; - op.op_status = cmd->op_status; op.t_run = (unsigned int) cmd->start_time; op.t_rcchange = (unsigned int) cmd->start_time; - if (cmd->reported_success && cmd->rc != PCMK_OCF_OK) { + + lrmd__set_result(&op, cmd->result.exit_status, cmd->result.execution_status, + cmd->result.exit_reason); + + if (cmd->reported_success && (cmd->result.exit_status != PCMK_OCF_OK)) { op.t_rcchange = (unsigned int) time(NULL); /* This edge case will likely never ever occur, but if it does the * result is that a failure will not be processed correctly. This is only * remotely possible because we are able to detect a connection resource's tcp * connection has failed at any moment after start has completed. The actual * recurring operation is just a connectivity ping. * * basically, we are not guaranteed that the first successful monitor op and * a subsequent failed monitor op will not occur in the same timestamp. We have to * make it look like the operations occurred at separate times though. */ if (op.t_rcchange == op.t_run) { op.t_rcchange++; } } if (cmd->params) { lrmd_key_value_t *tmp; op.params = pcmk__strkey_table(free, free); for (tmp = cmd->params; tmp; tmp = tmp->next) { g_hash_table_insert(op.params, strdup(tmp->key), strdup(tmp->value)); } } op.call_id = cmd->call_id; op.remote_nodename = cmd->owner; lrm_op_callback(&op); if (op.params) { g_hash_table_destroy(op.params); } + lrmd__reset_result(&op); } static void update_remaining_timeout(remote_ra_cmd_t * cmd) { cmd->remaining_timeout = ((cmd->timeout / 1000) - (time(NULL) - cmd->start_time)) * 1000; } static gboolean retry_start_cmd_cb(gpointer data) { lrm_state_t *lrm_state = data; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; remote_ra_cmd_t *cmd = NULL; int rc = pcmk_rc_error; if (!ra_data || !ra_data->cur_cmd) { return FALSE; } cmd = ra_data->cur_cmd; if (!pcmk__strcase_any_of(cmd->action, "start", "migrate_from", NULL)) { return FALSE; } update_remaining_timeout(cmd); if (cmd->remaining_timeout > 0) { rc = handle_remote_ra_start(lrm_state, cmd, cmd->remaining_timeout); } if (rc != pcmk_rc_ok) { report_remote_ra_result(cmd); if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } ra_data->cur_cmd = NULL; free_cmd(cmd); } else { /* wait for connection event */ } return FALSE; } static gboolean connection_takeover_timeout_cb(gpointer data) { lrm_state_t *lrm_state = NULL; remote_ra_cmd_t *cmd = data; crm_info("takeover event timed out for node %s", cmd->rsc_id); cmd->takeover_timeout_id = 0; lrm_state = lrm_state_find(cmd->rsc_id); handle_remote_ra_stop(lrm_state, cmd); free_cmd(cmd); return FALSE; } static gboolean monitor_timeout_cb(gpointer data) { lrm_state_t *lrm_state = NULL; remote_ra_cmd_t *cmd = data; lrm_state = lrm_state_find(cmd->rsc_id); crm_info("Timed out waiting for remote poke response from %s%s", cmd->rsc_id, (lrm_state? "" : " (no LRM state)")); cmd->monitor_timeout_id = 0; - cmd->op_status = PCMK_EXEC_TIMEOUT; - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, + "Remote executor did not respond"); if (lrm_state && lrm_state->remote_ra_data) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; if (ra_data->cur_cmd == cmd) { ra_data->cur_cmd = NULL; } if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } } report_remote_ra_result(cmd); free_cmd(cmd); if(lrm_state) { lrm_state_disconnect(lrm_state); } return FALSE; } static void synthesize_lrmd_success(lrm_state_t *lrm_state, const char *rsc_id, const char *op_type) { lrmd_event_data_t op = { 0, }; if (lrm_state == NULL) { /* if lrm_state not given assume local */ lrm_state = lrm_state_find(fsa_our_uname); } CRM_ASSERT(lrm_state != NULL); op.type = lrmd_event_exec_complete; op.rsc_id = rsc_id; op.op_type = op_type; - op.rc = PCMK_OCF_OK; - op.op_status = PCMK_EXEC_DONE; op.t_run = (unsigned int) time(NULL); op.t_rcchange = op.t_run; op.call_id = generate_callid(); + lrmd__set_result(&op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); process_lrm_event(lrm_state, &op, NULL, NULL); } void remote_lrm_op_callback(lrmd_event_data_t * op) { gboolean cmd_handled = FALSE; lrm_state_t *lrm_state = NULL; remote_ra_data_t *ra_data = NULL; remote_ra_cmd_t *cmd = NULL; crm_debug("Processing '%s%s%s' event on remote connection to %s: %s " "(%d) status=%s (%d)", (op->op_type? op->op_type : ""), (op->op_type? " " : ""), lrmd_event_type2str(op->type), op->remote_nodename, services_ocf_exitcode_str(op->rc), op->rc, pcmk_exec_status_str(op->op_status), op->op_status); lrm_state = lrm_state_find(op->remote_nodename); if (!lrm_state || !lrm_state->remote_ra_data) { crm_debug("No state information found for remote connection event"); return; } ra_data = lrm_state->remote_ra_data; if (op->type == lrmd_event_new_client) { // Another client has connected to the remote daemon if (ra_data->migrate_status == expect_takeover) { // Great, we knew this was coming ra_data->migrate_status = takeover_complete; } else { crm_err("Disconnecting from Pacemaker Remote node %s due to " "unexpected client takeover", op->remote_nodename); /* In this case, lrmd_tls_connection_destroy() will be called under the control of mainloop. */ /* Do not free lrm_state->conn yet. */ /* It'll be freed in the following stop action. */ lrm_state_disconnect_only(lrm_state); } return; } /* filter all EXEC events up */ if (op->type == lrmd_event_exec_complete) { if (ra_data->migrate_status == takeover_complete) { crm_debug("ignoring event, this connection is taken over by another node"); } else { lrm_op_callback(op); } return; } if ((op->type == lrmd_event_disconnect) && (ra_data->cur_cmd == NULL)) { if (ra_data->active == FALSE) { crm_debug("Disconnection from Pacemaker Remote node %s complete", lrm_state->node_name); } else if (!remote_ra_is_in_maintenance(lrm_state)) { crm_err("Lost connection to Pacemaker Remote node %s", lrm_state->node_name); ra_data->recurring_cmds = fail_all_monitor_cmds(ra_data->recurring_cmds); ra_data->cmds = fail_all_monitor_cmds(ra_data->cmds); } else { crm_notice("Unmanaged Pacemaker Remote node %s disconnected", lrm_state->node_name); /* Do roughly what a 'stop' on the remote-resource would do */ handle_remote_ra_stop(lrm_state, NULL); remote_node_down(lrm_state->node_name, DOWN_KEEP_LRM); /* now fake the reply of a successful 'stop' */ synthesize_lrmd_success(NULL, lrm_state->node_name, "stop"); } return; } if (!ra_data->cur_cmd) { crm_debug("no event to match"); return; } cmd = ra_data->cur_cmd; /* Start actions and migrate from actions complete after connection * comes back to us. */ if (op->type == lrmd_event_connect && pcmk__strcase_any_of(cmd->action, "start", "migrate_from", NULL)) { if (op->connection_rc < 0) { update_remaining_timeout(cmd); if (op->connection_rc == -ENOKEY) { // Hard error, don't retry - cmd->op_status = PCMK_EXEC_ERROR; - cmd->rc = PCMK_OCF_INVALID_PARAM; - cmd->exit_reason = strdup("Authentication key not readable"); + pcmk__set_result(&(cmd->result), PCMK_OCF_INVALID_PARAM, + PCMK_EXEC_ERROR, + "Authentication key not readable"); } else if (cmd->remaining_timeout > 3000) { crm_trace("rescheduling start, remaining timeout %d", cmd->remaining_timeout); g_timeout_add(1000, retry_start_cmd_cb, lrm_state); return; } else { crm_trace("can't reschedule start, remaining timeout too small %d", cmd->remaining_timeout); - cmd->op_status = PCMK_EXEC_TIMEOUT; - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_TIMEOUT, + pcmk_strerror(op->connection_rc)); } } else { lrm_state_reset_tables(lrm_state, TRUE); - cmd->rc = PCMK_OCF_OK; - cmd->op_status = PCMK_EXEC_DONE; + pcmk__set_result(&(cmd->result), PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); ra_data->active = TRUE; } crm_debug("Remote connection event matched %s action", cmd->action); report_remote_ra_result(cmd); cmd_handled = TRUE; } else if (op->type == lrmd_event_poke && pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { if (cmd->monitor_timeout_id) { g_source_remove(cmd->monitor_timeout_id); cmd->monitor_timeout_id = 0; } /* Only report success the first time, after that only worry about failures. * For this function, if we get the poke pack, it is always a success. Pokes * only fail if the send fails, or the response times out. */ if (!cmd->reported_success) { - cmd->rc = PCMK_OCF_OK; - cmd->op_status = PCMK_EXEC_DONE; + pcmk__set_result(&(cmd->result), PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); report_remote_ra_result(cmd); cmd->reported_success = 1; } crm_debug("Remote poke event matched %s action", cmd->action); /* success, keep rescheduling if interval is present. */ if (cmd->interval_ms && (cmd->cancel == FALSE)) { ra_data->recurring_cmds = g_list_append(ra_data->recurring_cmds, cmd); cmd->interval_id = g_timeout_add(cmd->interval_ms, recurring_helper, cmd); cmd = NULL; /* prevent free */ } cmd_handled = TRUE; } else if (op->type == lrmd_event_disconnect && pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { if (ra_data->active == TRUE && (cmd->cancel == FALSE)) { - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; - cmd->op_status = PCMK_EXEC_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR, + "Remote connection unexpectedly dropped " + "during monitor"); report_remote_ra_result(cmd); crm_err("Remote connection to %s unexpectedly dropped during monitor", lrm_state->node_name); } cmd_handled = TRUE; } else if (op->type == lrmd_event_new_client && pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { handle_remote_ra_stop(lrm_state, cmd); cmd_handled = TRUE; } else { crm_debug("Event did not match %s action", ra_data->cur_cmd->action); } if (cmd_handled) { ra_data->cur_cmd = NULL; if (ra_data->cmds) { mainloop_set_trigger(ra_data->work); } free_cmd(cmd); } } static void handle_remote_ra_stop(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd) { remote_ra_data_t *ra_data = NULL; CRM_ASSERT(lrm_state); ra_data = lrm_state->remote_ra_data; if (ra_data->migrate_status != takeover_complete) { /* delete pending ops when ever the remote connection is intentionally stopped */ g_hash_table_remove_all(lrm_state->pending_ops); } else { /* we no longer hold the history if this connection has been migrated, * however, we keep metadata cache for future use */ lrm_state_reset_tables(lrm_state, FALSE); } ra_data->active = FALSE; lrm_state_disconnect(lrm_state); if (ra_data->cmds) { g_list_free_full(ra_data->cmds, free_cmd); } if (ra_data->recurring_cmds) { g_list_free_full(ra_data->recurring_cmds, free_cmd); } ra_data->cmds = NULL; ra_data->recurring_cmds = NULL; ra_data->cur_cmd = NULL; if (cmd) { - cmd->rc = PCMK_OCF_OK; - cmd->op_status = PCMK_EXEC_DONE; + pcmk__set_result(&(cmd->result), PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); report_remote_ra_result(cmd); } } // \return Standard Pacemaker return code static int handle_remote_ra_start(lrm_state_t * lrm_state, remote_ra_cmd_t * cmd, int timeout_ms) { const char *server = NULL; lrmd_key_value_t *tmp = NULL; int port = 0; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; int timeout_used = timeout_ms > MAX_START_TIMEOUT_MS ? MAX_START_TIMEOUT_MS : timeout_ms; int rc = pcmk_rc_ok; for (tmp = cmd->params; tmp; tmp = tmp->next) { if (pcmk__strcase_any_of(tmp->key, XML_RSC_ATTR_REMOTE_RA_ADDR, XML_RSC_ATTR_REMOTE_RA_SERVER, NULL)) { server = tmp->value; } else if (pcmk__str_eq(tmp->key, XML_RSC_ATTR_REMOTE_RA_PORT, pcmk__str_casei)) { port = atoi(tmp->value); } else if (pcmk__str_eq(tmp->key, CRM_META "_" XML_RSC_ATTR_CONTAINER, pcmk__str_casei)) { ra_data->controlling_guest = TRUE; } } rc = controld_connect_remote_executor(lrm_state, server, port, timeout_used); if (rc != pcmk_rc_ok) { - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; - cmd->op_status = PCMK_EXEC_ERROR; - cmd->exit_reason = strdup(pcmk_rc_str(rc)); + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR, pcmk_rc_str(rc)); } return rc; } static gboolean handle_remote_ra_exec(gpointer user_data) { int rc = 0; lrm_state_t *lrm_state = user_data; remote_ra_data_t *ra_data = lrm_state->remote_ra_data; remote_ra_cmd_t *cmd; GList *first = NULL; if (ra_data->cur_cmd) { /* still waiting on previous cmd */ return TRUE; } while (ra_data->cmds) { first = ra_data->cmds; cmd = first->data; if (cmd->delay_id) { /* still waiting for start delay timer to trip */ return TRUE; } ra_data->cmds = g_list_remove_link(ra_data->cmds, first); g_list_free_1(first); if (!strcmp(cmd->action, "start") || !strcmp(cmd->action, "migrate_from")) { ra_data->migrate_status = 0; if (handle_remote_ra_start(lrm_state, cmd, cmd->timeout) == pcmk_rc_ok) { /* take care of this later when we get async connection result */ crm_debug("Initiated async remote connection, %s action will complete after connect event", cmd->action); ra_data->cur_cmd = cmd; return TRUE; } report_remote_ra_result(cmd); } else if (!strcmp(cmd->action, "monitor")) { if (lrm_state_is_connected(lrm_state) == TRUE) { rc = lrm_state_poke_connection(lrm_state); if (rc < 0) { - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; - cmd->op_status = PCMK_EXEC_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR, pcmk_strerror(rc)); } } else { rc = -1; - cmd->op_status = PCMK_EXEC_DONE; - cmd->rc = PCMK_OCF_NOT_RUNNING; + pcmk__set_result(&(cmd->result), PCMK_OCF_NOT_RUNNING, + PCMK_EXEC_DONE, "Remote connection inactive"); } if (rc == 0) { crm_debug("Poked Pacemaker Remote at node %s, waiting for async response", cmd->rsc_id); ra_data->cur_cmd = cmd; cmd->monitor_timeout_id = g_timeout_add(cmd->timeout, monitor_timeout_cb, cmd); return TRUE; } report_remote_ra_result(cmd); } else if (!strcmp(cmd->action, "stop")) { if (ra_data->migrate_status == expect_takeover) { /* briefly wait on stop for the takeover event to occur. If the * takeover event does not occur during the wait period, that's fine. * It just means that the remote-node's lrm_status section is going to get * cleared which will require all the resources running in the remote-node * to be explicitly re-detected via probe actions. If the takeover does occur * successfully, then we can leave the status section intact. */ cmd->takeover_timeout_id = g_timeout_add((cmd->timeout/2), connection_takeover_timeout_cb, cmd); ra_data->cur_cmd = cmd; return TRUE; } handle_remote_ra_stop(lrm_state, cmd); } else if (!strcmp(cmd->action, "migrate_to")) { ra_data->migrate_status = expect_takeover; - cmd->rc = PCMK_OCF_OK; - cmd->op_status = PCMK_EXEC_DONE; + pcmk__set_result(&(cmd->result), PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); report_remote_ra_result(cmd); } else if (pcmk__str_any_of(cmd->action, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { /* Currently the only reloadable parameter is reconnect_interval, * which is only used by the scheduler via the CIB, so reloads are a * no-op. * * @COMPAT DC <2.1.0: We only need to check for "reload" in case * we're in a rolling upgrade with a DC scheduling "reload" instead * of "reload-agent". An OCF 1.1 "reload" would be a no-op anyway, * so this would work for that purpose as well. */ - cmd->rc = PCMK_OCF_OK; - cmd->op_status = PCMK_EXEC_DONE; + pcmk__set_result(&(cmd->result), PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); report_remote_ra_result(cmd); } free_cmd(cmd); } return TRUE; } static void remote_ra_data_init(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = NULL; if (lrm_state->remote_ra_data) { return; } ra_data = calloc(1, sizeof(remote_ra_data_t)); ra_data->work = mainloop_add_trigger(G_PRIORITY_HIGH, handle_remote_ra_exec, lrm_state); lrm_state->remote_ra_data = ra_data; } void remote_ra_cleanup(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; if (!ra_data) { return; } if (ra_data->cmds) { g_list_free_full(ra_data->cmds, free_cmd); } if (ra_data->recurring_cmds) { g_list_free_full(ra_data->recurring_cmds, free_cmd); } mainloop_destroy_trigger(ra_data->work); free(ra_data); lrm_state->remote_ra_data = NULL; } gboolean is_remote_lrmd_ra(const char *agent, const char *provider, const char *id) { if (agent && provider && !strcmp(agent, REMOTE_LRMD_RA) && !strcmp(provider, "pacemaker")) { return TRUE; } if (id && lrm_state_find(id) && !pcmk__str_eq(id, fsa_our_uname, pcmk__str_casei)) { return TRUE; } return FALSE; } lrmd_rsc_info_t * remote_ra_get_rsc_info(lrm_state_t * lrm_state, const char *rsc_id) { lrmd_rsc_info_t *info = NULL; if ((lrm_state_find(rsc_id))) { info = calloc(1, sizeof(lrmd_rsc_info_t)); info->id = strdup(rsc_id); info->type = strdup(REMOTE_LRMD_RA); info->standard = strdup(PCMK_RESOURCE_CLASS_OCF); info->provider = strdup("pacemaker"); } return info; } static gboolean is_remote_ra_supported_action(const char *action) { return pcmk__str_any_of(action, CRMD_ACTION_START, CRMD_ACTION_STOP, CRMD_ACTION_STATUS, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, CRMD_ACTION_RELOAD_AGENT, CRMD_ACTION_RELOAD, NULL); } static GList * fail_all_monitor_cmds(GList * list) { GList *rm_list = NULL; remote_ra_cmd_t *cmd = NULL; GList *gIter = NULL; for (gIter = list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms > 0) && pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { rm_list = g_list_append(rm_list, cmd); } } for (gIter = rm_list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; - cmd->rc = PCMK_OCF_UNKNOWN_ERROR; - cmd->op_status = PCMK_EXEC_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR, "Lost connection to remote executor"); crm_trace("Pre-emptively failing %s %s (interval=%u, %s)", cmd->action, cmd->rsc_id, cmd->interval_ms, cmd->userdata); report_remote_ra_result(cmd); list = g_list_remove(list, cmd); free_cmd(cmd); } /* frees only the list data, not the cmds */ g_list_free(rm_list); return list; } static GList * remove_cmd(GList * list, const char *action, guint interval_ms) { remote_ra_cmd_t *cmd = NULL; GList *gIter = NULL; for (gIter = list; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && pcmk__str_eq(cmd->action, action, pcmk__str_casei)) { break; } cmd = NULL; } if (cmd) { list = g_list_remove(list, cmd); free_cmd(cmd); } return list; } int remote_ra_cancel(lrm_state_t *lrm_state, const char *rsc_id, const char *action, guint interval_ms) { lrm_state_t *connection_rsc = NULL; remote_ra_data_t *ra_data = NULL; connection_rsc = lrm_state_find(rsc_id); if (!connection_rsc || !connection_rsc->remote_ra_data) { return -EINVAL; } ra_data = connection_rsc->remote_ra_data; ra_data->cmds = remove_cmd(ra_data->cmds, action, interval_ms); ra_data->recurring_cmds = remove_cmd(ra_data->recurring_cmds, action, interval_ms); if (ra_data->cur_cmd && (ra_data->cur_cmd->interval_ms == interval_ms) && (pcmk__str_eq(ra_data->cur_cmd->action, action, pcmk__str_casei))) { ra_data->cur_cmd->cancel = TRUE; } return 0; } static remote_ra_cmd_t * handle_dup_monitor(remote_ra_data_t *ra_data, guint interval_ms, const char *userdata) { GList *gIter = NULL; remote_ra_cmd_t *cmd = NULL; /* there are 3 places a potential duplicate monitor operation * could exist. * 1. recurring_cmds list. where the op is waiting for its next interval * 2. cmds list, where the op is queued to get executed immediately * 3. cur_cmd, which means the monitor op is in flight right now. */ if (interval_ms == 0) { return NULL; } if (ra_data->cur_cmd && ra_data->cur_cmd->cancel == FALSE && (ra_data->cur_cmd->interval_ms == interval_ms) && pcmk__str_eq(ra_data->cur_cmd->action, "monitor", pcmk__str_casei)) { cmd = ra_data->cur_cmd; goto handle_dup; } for (gIter = ra_data->recurring_cmds; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { goto handle_dup; } } for (gIter = ra_data->cmds; gIter != NULL; gIter = gIter->next) { cmd = gIter->data; if ((cmd->interval_ms == interval_ms) && pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { goto handle_dup; } } return NULL; handle_dup: crm_trace("merging duplicate monitor cmd " PCMK__OP_FMT, cmd->rsc_id, "monitor", interval_ms); /* update the userdata */ if (userdata) { free(cmd->userdata); cmd->userdata = strdup(userdata); } /* if we've already reported success, generate a new call id */ if (cmd->reported_success) { cmd->start_time = time(NULL); cmd->call_id = generate_callid(); cmd->reported_success = 0; } /* if we have an interval_id set, that means we are in the process of * waiting for this cmd's next interval. instead of waiting, cancel * the timer and execute the action immediately */ if (cmd->interval_id) { g_source_remove(cmd->interval_id); cmd->interval_id = 0; recurring_helper(cmd); } return cmd; } /*! * \internal * \brief Execute an action using the (internal) ocf:pacemaker:remote agent * * \param[in] lrm_state Executor state object for remote connection * \param[in] rsc_id Connection resource ID * \param[in] action Action to execute * \param[in] userdata String to copy and pass to execution callback * \param[in] interval_ms Action interval (in milliseconds) * \param[in] timeout_ms Action timeout (in milliseconds) * \param[in] start_delay_ms Delay (in milliseconds) before initiating action * \param[in] params Connection resource parameters * \param[out] call_id Where to store call ID on success * * \return Standard Pacemaker return code * \note This takes ownership of \p params, which should not be used or freed * after calling this function. */ int controld_execute_remote_agent(lrm_state_t *lrm_state, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout_ms, int start_delay_ms, lrmd_key_value_t *params, int *call_id) { lrm_state_t *connection_rsc = NULL; remote_ra_cmd_t *cmd = NULL; remote_ra_data_t *ra_data = NULL; *call_id = 0; CRM_CHECK((lrm_state != NULL) && (rsc_id != NULL) && (action != NULL) && (userdata != NULL) && (call_id != NULL), lrmd_key_value_freeall(params); return EINVAL); if (!is_remote_ra_supported_action(action)) { lrmd_key_value_freeall(params); return EOPNOTSUPP; } connection_rsc = lrm_state_find(rsc_id); if (connection_rsc == NULL) { lrmd_key_value_freeall(params); return ENOTCONN; } remote_ra_data_init(connection_rsc); ra_data = connection_rsc->remote_ra_data; cmd = handle_dup_monitor(ra_data, interval_ms, userdata); if (cmd) { *call_id = cmd->call_id; lrmd_key_value_freeall(params); return pcmk_rc_ok; } cmd = calloc(1, sizeof(remote_ra_cmd_t)); if (cmd == NULL) { lrmd_key_value_freeall(params); return ENOMEM; } cmd->owner = strdup(lrm_state->node_name); cmd->rsc_id = strdup(rsc_id); cmd->action = strdup(action); cmd->userdata = strdup(userdata); if ((cmd->owner == NULL) || (cmd->rsc_id == NULL) || (cmd->action == NULL) || (cmd->userdata == NULL)) { free_cmd(cmd); lrmd_key_value_freeall(params); return ENOMEM; } cmd->interval_ms = interval_ms; cmd->timeout = timeout_ms; cmd->start_delay = start_delay_ms; cmd->params = params; cmd->start_time = time(NULL); cmd->call_id = generate_callid(); if (cmd->start_delay) { cmd->delay_id = g_timeout_add(cmd->start_delay, start_delay_helper, cmd); } ra_data->cmds = g_list_append(ra_data->cmds, cmd); mainloop_set_trigger(ra_data->work); *call_id = cmd->call_id; return pcmk_rc_ok; } /*! * \internal * \brief Immediately fail all monitors of a remote node, if proxied here * * \param[in] node_name Name of pacemaker_remote node */ void remote_ra_fail(const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (lrm_state && lrm_state_is_connected(lrm_state)) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; crm_info("Failing monitors on Pacemaker Remote node %s", node_name); ra_data->recurring_cmds = fail_all_monitor_cmds(ra_data->recurring_cmds); ra_data->cmds = fail_all_monitor_cmds(ra_data->cmds); } } /* A guest node fencing implied by host fencing looks like: * * * * * * * */ #define XPATH_PSEUDO_FENCE "//" XML_GRAPH_TAG_PSEUDO_EVENT \ "[@" XML_LRM_ATTR_TASK "='stonith']/" XML_GRAPH_TAG_DOWNED \ "/" XML_CIB_TAG_NODE /*! * \internal * \brief Check a pseudo-action for Pacemaker Remote node side effects * * \param[in] xml XML of pseudo-action to check */ void remote_ra_process_pseudo(xmlNode *xml) { xmlXPathObjectPtr search = xpath_search(xml, XPATH_PSEUDO_FENCE); if (numXpathResults(search) == 1) { xmlNode *result = getXpathResult(search, 0); /* Normally, we handle the necessary side effects of a guest node stop * action when reporting the remote agent's result. However, if the stop * is implied due to fencing, it will be a fencing pseudo-event, and * there won't be a result to report. Handle that case here. * * This will result in a duplicate call to remote_node_down() if the * guest stop was real instead of implied, but that shouldn't hurt. * * There is still one corner case that isn't handled: if a guest node * isn't running any resources when its host is fenced, it will appear * to be cleanly stopped, so there will be no pseudo-fence, and our * peer cache state will be incorrect unless and until the guest is * recovered. */ if (result) { const char *remote = ID(result); if (remote) { remote_node_down(remote, DOWN_ERASE_LRM); } } } freeXpathObject(search); } static void remote_ra_maintenance(lrm_state_t * lrm_state, gboolean maintenance) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; xmlNode *update, *state; int call_opt, call_id = 0; crm_node_t *node; call_opt = crmd_cib_smart_opt(); node = crm_remote_peer_get(lrm_state->node_name); CRM_CHECK(node != NULL, return); update = create_xml_node(NULL, XML_CIB_TAG_STATUS); state = create_node_state_update(node, node_update_none, update, __func__); crm_xml_add(state, XML_NODE_IS_MAINTENANCE, maintenance?"1":"0"); fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, call_id, NULL); if (call_id < 0) { crm_perror(LOG_WARNING, "%s CIB node state update failed", lrm_state->node_name); } else { /* TODO: still not 100% sure that async update will succeed ... */ ra_data->is_maintenance = maintenance; } free_xml(update); } #define XPATH_PSEUDO_MAINTENANCE "//" XML_GRAPH_TAG_PSEUDO_EVENT \ "[@" XML_LRM_ATTR_TASK "='" CRM_OP_MAINTENANCE_NODES "']/" \ XML_GRAPH_TAG_MAINTENANCE /*! * \internal * \brief Check a pseudo-action holding updates for maintenance state * * \param[in] xml XML of pseudo-action to check */ void remote_ra_process_maintenance_nodes(xmlNode *xml) { xmlXPathObjectPtr search = xpath_search(xml, XPATH_PSEUDO_MAINTENANCE); if (numXpathResults(search) == 1) { xmlNode *node; int cnt = 0, cnt_remote = 0; for (node = first_named_child(getXpathResult(search, 0), XML_CIB_TAG_NODE); node != NULL; node = pcmk__xml_next(node)) { lrm_state_t *lrm_state = lrm_state_find(ID(node)); cnt++; if (lrm_state && lrm_state->remote_ra_data && ((remote_ra_data_t *) lrm_state->remote_ra_data)->active) { int is_maint; cnt_remote++; pcmk__scan_min_int(crm_element_value(node, XML_NODE_IS_MAINTENANCE), &is_maint, 0); remote_ra_maintenance(lrm_state, is_maint); } } crm_trace("Action holds %d nodes (%d remotes found) " "adjusting maintenance-mode", cnt, cnt_remote); } freeXpathObject(search); } gboolean remote_ra_is_in_maintenance(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; return ra_data->is_maintenance; } gboolean remote_ra_controlling_guest(lrm_state_t * lrm_state) { remote_ra_data_t *ra_data = lrm_state->remote_ra_data; return ra_data->controlling_guest; } diff --git a/daemons/controld/controld_te_actions.c b/daemons/controld/controld_te_actions.c index 282f7ff683..80a5236c95 100644 --- a/daemons/controld/controld_te_actions.c +++ b/daemons/controld/controld_te_actions.c @@ -1,643 +1,641 @@ /* * Copyright 2004-2021 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 // lrmd_event_data_t, lrmd_free_event() #include #include #include #include #include char *te_uuid = NULL; GHashTable *te_targets = NULL; void send_rsc_command(crm_action_t * action); static void te_update_job_count(crm_action_t * action, int offset); static void te_start_action_timer(crm_graph_t * graph, crm_action_t * action) { action->timer = calloc(1, sizeof(crm_action_timer_t)); action->timer->timeout = action->timeout; action->timer->action = action; action->timer->source_id = g_timeout_add(action->timer->timeout + graph->network_delay, action_timer_callback, (void *)action->timer); CRM_ASSERT(action->timer->source_id != 0); } static gboolean te_pseudo_action(crm_graph_t * graph, crm_action_t * pseudo) { const char *task = crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK); /* send to peers as well? */ if (pcmk__str_eq(task, CRM_OP_MAINTENANCE_NODES, pcmk__str_casei)) { GHashTableIter iter; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *cmd = NULL; if (pcmk__str_eq(fsa_our_uname, node->uname, pcmk__str_casei)) { continue; } cmd = create_request(task, pseudo->xml, node->uname, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); send_cluster_message(node, crm_msg_crmd, cmd, FALSE); free_xml(cmd); } remote_ra_process_maintenance_nodes(pseudo->xml); } else { /* Check action for Pacemaker Remote node side effects */ remote_ra_process_pseudo(pseudo->xml); } crm_debug("Pseudo-action %d (%s) fired and confirmed", pseudo->id, crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK_KEY)); te_action_confirmed(pseudo, graph); return TRUE; } static int get_target_rc(crm_action_t * action) { int exit_status; pcmk__scan_min_int(crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC), &exit_status, 0); return exit_status; } static gboolean te_crm_command(crm_graph_t * graph, crm_action_t * action) { char *counter = NULL; xmlNode *cmd = NULL; gboolean is_local = FALSE; const char *id = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; id = ID(action->xml); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; if (pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { const char *mode = crm_element_value(action->xml, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_casei)) { router_node = fsa_our_uname; } } } CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command (id=%s) %s: no node", crm_str(id), crm_str(task)); return FALSE); if (pcmk__str_eq(router_node, fsa_our_uname, pcmk__str_casei)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_info("Executing crm-event (%s)%s%s: %s on %s", crm_str(id), (is_local? " locally" : ""), (no_wait? " without waiting" : ""), crm_str(task), on_node); if (is_local && pcmk__str_eq(task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { /* defer until everything else completes */ crm_info("crm-event (%s) is a local shutdown", crm_str(id)); graph->completion_action = tg_shutdown; graph->abort_reason = "local shutdown"; te_action_confirmed(action, graph); return TRUE; } else if (pcmk__str_eq(task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { crm_node_t *peer = crm_get_peer(0, router_node); pcmk__update_peer_expected(__func__, peer, CRMD_JOINSTATE_DOWN); } cmd = create_request(task, action->xml, router_node, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); counter = pcmk__transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(cmd, XML_ATTR_TRANSITION_KEY, counter); rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_crmd, cmd, TRUE); free(counter); free_xml(cmd); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { te_action_confirmed(action, graph); } else { if (action->timeout <= 0) { crm_err("Action %d: %s on %s had an invalid timeout (%dms). Using %ums instead", action->id, task, on_node, action->timeout, graph->network_delay); action->timeout = (int) graph->network_delay; } te_start_action_timer(graph, action); } return TRUE; } void controld_record_action_timeout(crm_action_t *action) { lrmd_event_data_t *op = NULL; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *xml_op = NULL; xmlNode *action_rsc = NULL; int rc = pcmk_ok; const char *rsc_id = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); const char *target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); int call_options = cib_quorum_override | cib_scope_local; int target_rc = get_target_rc(action); crm_warn("%s %d: %s on %s timed out", crm_element_name(action->xml), action->id, task_uuid, target); action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if (action_rsc == NULL) { return; } rsc_id = ID(action_rsc); CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); return); /* update the CIB */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, XML_ATTR_UUID, target_uuid); crm_xml_add(state, XML_ATTR_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, XML_ATTR_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, XML_ATTR_ID, rsc_id); crm_copy_xml_element(action_rsc, rsc, XML_ATTR_TYPE); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_CLASS); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_PROVIDER); /* If the executor gets a timeout while waiting for the action to complete, * that will be reported via the usual callback. This timeout means that we * didn't hear from the executor or the controller that relayed the action * to the executor. - * - * @TODO Using PCMK_OCF_UNKNOWN_ERROR instead of PCMK_OCF_TIMEOUT is one way - * to distinguish those situations, but perhaps PCMK_OCF_TIMEOUT would be - * preferable anyway. */ op = pcmk__event_from_graph_action(NULL, action, PCMK_EXEC_TIMEOUT, - PCMK_OCF_UNKNOWN_ERROR); + PCMK_OCF_TIMEOUT, + "Cluster communication timeout " + "(no response from executor)"); op->call_id = -1; op->user_data = pcmk__transition_key(transition_graph->id, action->id, target_rc, te_uuid); xml_op = pcmk__create_history_xml(rsc, op, CRM_FEATURE_SET, target_rc, target, __func__, LOG_INFO); lrmd_free_event(op); crm_log_xml_trace(xml_op, "Action timeout"); rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, state, call_options); fsa_register_cib_callback(rc, FALSE, NULL, cib_action_updated); free_xml(state); crm_trace("Sent CIB update (call ID %d) for timeout of action %d (%s on %s)", rc, action->id, task_uuid, target); action->sent_update = TRUE; } static gboolean te_rsc_command(crm_graph_t * graph, crm_action_t * action) { /* never overwrite stop actions in the CIB with * anything other than completed results * * Writing pending stops makes it look like the * resource is running again */ xmlNode *cmd = NULL; xmlNode *rsc_op = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; gboolean is_local = FALSE; char *counter = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; const char *task_uuid = NULL; CRM_ASSERT(action != NULL); CRM_ASSERT(action->xml != NULL); action->executed = FALSE; on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), crm_str(task)); return FALSE); rsc_op = action->xml; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); router_node = crm_element_value(rsc_op, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } counter = pcmk__transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (pcmk__str_eq(router_node, fsa_our_uname, pcmk__str_casei)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_notice("Initiating %s operation %s%s on %s%s "CRM_XS" action %d", task, task_uuid, (is_local? " locally" : ""), on_node, (no_wait? " without waiting" : ""), action->id); cmd = create_request(CRM_OP_INVOKE_LRM, rsc_op, router_node, CRM_SYSTEM_LRMD, CRM_SYSTEM_TENGINE, NULL); if (is_local) { /* shortcut local resource commands */ ha_msg_input_t data = { .msg = cmd, .xml = rsc_op, }; fsa_data_t msg = { .id = 0, .data = &data, .data_type = fsa_dt_ha_msg, .fsa_input = I_NULL, .fsa_cause = C_FSA_INTERNAL, .actions = A_LRM_INVOKE, .origin = __func__, }; do_lrm_invoke(A_LRM_INVOKE, C_FSA_INTERNAL, fsa_state, I_NULL, &msg); } else { rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_lrmd, cmd, TRUE); } free(counter); free_xml(cmd); action->executed = TRUE; if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { crm_info("Action %d confirmed - no wait", action->id); action->confirmed = TRUE; /* Just mark confirmed. * Don't bump the job count only to immediately decrement it */ pcmk__update_graph(transition_graph, action); trigger_graph(); } else if (action->confirmed == TRUE) { crm_debug("Action %d: %s %s on %s(timeout %dms) was already confirmed.", action->id, task, task_uuid, on_node, action->timeout); } else { if (action->timeout <= 0) { crm_err("Action %d: %s %s on %s had an invalid timeout (%dms). Using %ums instead", action->id, task, task_uuid, on_node, action->timeout, graph->network_delay); action->timeout = (int) graph->network_delay; } te_update_job_count(action, 1); te_start_action_timer(graph, action); } return TRUE; } struct te_peer_s { char *name; int jobs; int migrate_jobs; }; static void te_peer_free(gpointer p) { struct te_peer_s *peer = p; free(peer->name); free(peer); } void te_reset_job_counts(void) { GHashTableIter iter; struct te_peer_s *peer = NULL; if(te_targets == NULL) { te_targets = pcmk__strkey_table(NULL, te_peer_free); } g_hash_table_iter_init(&iter, te_targets); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & peer)) { peer->jobs = 0; peer->migrate_jobs = 0; } } static void te_update_job_count_on(const char *target, int offset, bool migrate) { struct te_peer_s *r = NULL; if(target == NULL || te_targets == NULL) { return; } r = g_hash_table_lookup(te_targets, target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } r->jobs += offset; if(migrate) { r->migrate_jobs += offset; } crm_trace("jobs[%s] = %d", target, r->jobs); } static void te_update_job_count(crm_action_t * action, int offset) { const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (action->type != action_type_rsc || target == NULL) { /* No limit on these */ return; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && pcmk__strcase_any_of(task, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { const char *t1 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); const char *t2 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); te_update_job_count_on(t1, offset, TRUE); te_update_job_count_on(t2, offset, TRUE); return; } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } te_update_job_count_on(target, offset, FALSE); } static gboolean te_should_perform_action_on(crm_graph_t * graph, crm_action_t * action, const char *target) { int limit = 0; struct te_peer_s *r = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *id = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); if(target == NULL) { /* No limit on these */ return TRUE; } else if(te_targets == NULL) { return FALSE; } r = g_hash_table_lookup(te_targets, target); limit = throttle_get_job_limit(target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } if(limit <= r->jobs) { crm_trace("Peer %s is over their job limit of %d (%d): deferring %s", target, limit, r->jobs, id); return FALSE; } else if(graph->migration_limit > 0 && r->migrate_jobs >= graph->migration_limit) { if (pcmk__strcase_any_of(task, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { crm_trace("Peer %s is over their migration job limit of %d (%d): deferring %s", target, graph->migration_limit, r->migrate_jobs, id); return FALSE; } } crm_trace("Peer %s has not hit their limit yet. current jobs = %d limit= %d limit", target, r->jobs, limit); return TRUE; } static gboolean te_should_perform_action(crm_graph_t * graph, crm_action_t * action) { const char *target = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != action_type_rsc) { /* No limit on these */ return TRUE; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && pcmk__strcase_any_of(task, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, NULL)) { target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); if(te_should_perform_action_on(graph, action, target) == FALSE) { return FALSE; } target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } return te_should_perform_action_on(graph, action, target); } /*! * \brief Confirm a graph action (and optionally update graph) * * \param[in] action Action to confirm * \param[in] graph Update and trigger this graph (if non-NULL) */ void te_action_confirmed(crm_action_t *action, crm_graph_t *graph) { if (action->confirmed == FALSE) { if ((action->type == action_type_rsc) && (crm_element_value(action->xml, XML_LRM_ATTR_TARGET) != NULL)) { te_update_job_count(action, -1); } action->confirmed = TRUE; } if (graph) { pcmk__update_graph(graph, action); trigger_graph(); } } crm_graph_functions_t te_graph_fns = { te_pseudo_action, te_rsc_command, te_crm_command, te_fence_node, te_should_perform_action, }; void notify_crmd(crm_graph_t * graph) { const char *type = "unknown"; enum crmd_fsa_input event = I_NULL; crm_debug("Processing transition completion in state %s", fsa_state2string(fsa_state)); if (graph->complete == FALSE) { CRM_CHECK(graph->complete,); graph->complete = TRUE; } switch (graph->completion_action) { case tg_stop: type = "stop"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_done: type = "done"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_restart: type = "restart"; if (fsa_state == S_TRANSITION_ENGINE) { if (transition_timer->period_ms > 0) { controld_stop_timer(transition_timer); controld_start_timer(transition_timer); } else { event = I_PE_CALC; } } else if (fsa_state == S_POLICY_ENGINE) { controld_set_fsa_action_flags(A_PE_INVOKE); trigger_fsa(); } break; case tg_shutdown: type = "shutdown"; if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { event = I_STOP; } else { crm_err("We didn't ask to be shut down, yet the scheduler is telling us to"); event = I_TERMINATE; } } crm_debug("Transition %d status: %s - %s", graph->id, type, crm_str(graph->abort_reason)); graph->abort_reason = NULL; graph->completion_action = tg_done; controld_clear_fsa_input_flags(R_IN_TRANSITION); if (event != I_NULL) { register_fsa_input(C_FSA_INTERNAL, event, NULL); } else if (fsa_source) { mainloop_set_trigger(fsa_source); } } diff --git a/daemons/execd/execd_commands.c b/daemons/execd/execd_commands.c index 80449ab71e..4be1ce7ee9 100644 --- a/daemons/execd/execd_commands.c +++ b/daemons/execd/execd_commands.c @@ -1,1995 +1,2014 @@ /* * Copyright 2012-2021 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 // Check whether we have a high-resolution monotonic clock #undef PCMK__TIME_USE_CGT #if HAVE_DECL_CLOCK_MONOTONIC && defined(CLOCK_MONOTONIC) # define PCMK__TIME_USE_CGT # include /* clock_gettime */ #endif #include #include #include #include #include #include #include #include #include "pacemaker-execd.h" #define EXIT_REASON_MAX_LEN 128 GHashTable *rsc_list = NULL; typedef struct lrmd_cmd_s { int timeout; guint interval_ms; int start_delay; int timeout_orig; int call_id; - int exec_rc; - int lrmd_op_status; int call_opts; /* Timer ids, must be removed on cmd destruction. */ int delay_id; int stonith_recurring_id; int rsc_deleted; int service_flags; char *client_id; char *origin; char *rsc_id; char *action; char *real_action; - char *exit_reason; - char *output; char *userdata_str; + pcmk__action_result_t result; + /* We can track operation queue time and run time, to be saved with the CIB * resource history (and displayed in cluster status). We need * high-resolution monotonic time for this purpose, so we use * clock_gettime(CLOCK_MONOTONIC, ...) (if available, otherwise this feature * is disabled). * * However, we also need epoch timestamps for recording the time the command * last ran and the time its return value last changed, for use in time * displays (as opposed to interval calculations). We keep time_t values for * this purpose. * * The last run time is used for both purposes, so we keep redundant * monotonic and epoch values for this. Technically the two could represent * different times, but since time_t has only second resolution and the * values are used for distinct purposes, that is not significant. */ #ifdef PCMK__TIME_USE_CGT /* Recurring and systemd operations may involve more than one executor * command per operation, so they need info about the original and the most * recent. */ struct timespec t_first_run; // When op first ran struct timespec t_run; // When op most recently ran struct timespec t_first_queue; // When op was first queued struct timespec t_queue; // When op was most recently queued #endif time_t epoch_last_run; // Epoch timestamp of when op last ran time_t epoch_rcchange; // Epoch timestamp of when rc last changed bool first_notify_sent; int last_notify_rc; int last_notify_op_status; int last_pid; GHashTable *params; } lrmd_cmd_t; static void cmd_finalize(lrmd_cmd_t * cmd, lrmd_rsc_t * rsc); static gboolean lrmd_rsc_dispatch(gpointer user_data); static void cancel_all_recurring(lrmd_rsc_t * rsc, const char *client_id); #ifdef PCMK__TIME_USE_CGT /*! * \internal * \brief Check whether a struct timespec has been set * * \param[in] timespec Time to check * * \return true if timespec has been set (i.e. is nonzero), false otherwise */ static inline bool time_is_set(struct timespec *timespec) { return (timespec != NULL) && ((timespec->tv_sec != 0) || (timespec->tv_nsec != 0)); } /* * \internal * \brief Set a timespec (and its original if unset) to the current time * * \param[out] t_current Where to store current time * \param[out] t_orig Where to copy t_current if unset */ static void get_current_time(struct timespec *t_current, struct timespec *t_orig) { clock_gettime(CLOCK_MONOTONIC, t_current); if ((t_orig != NULL) && !time_is_set(t_orig)) { *t_orig = *t_current; } } /*! * \internal * \brief Return difference between two times in milliseconds * * \param[in] now More recent time (or NULL to use current time) * \param[in] old Earlier time * * \return milliseconds difference (or 0 if old is NULL or unset) * * \note Can overflow on 32bit machines when the differences is around * 24 days or more. */ static int time_diff_ms(struct timespec *now, struct timespec *old) { int diff_ms = 0; if (time_is_set(old)) { struct timespec local_now = { 0, }; if (now == NULL) { clock_gettime(CLOCK_MONOTONIC, &local_now); now = &local_now; } diff_ms = (now->tv_sec - old->tv_sec) * 1000 + (now->tv_nsec - old->tv_nsec) / 1000000; } return diff_ms; } /*! * \internal * \brief Reset a command's operation times to their original values. * * Reset a command's run and queued timestamps to the timestamps of the original * command, so we report the entire time since then and not just the time since * the most recent command (for recurring and systemd operations). * * \param[in] cmd Executor command object to reset * * \note It's not obvious what the queued time should be for a systemd * start/stop operation, which might go like this: * initial command queued 5ms, runs 3s * monitor command queued 10ms, runs 10s * monitor command queued 10ms, runs 10s * Is the queued time for that operation 5ms, 10ms or 25ms? The current * implementation will report 5ms. If it's 25ms, then we need to * subtract 20ms from the total exec time so as not to count it twice. * We can implement that later if it matters to anyone ... */ static void cmd_original_times(lrmd_cmd_t * cmd) { cmd->t_run = cmd->t_first_run; cmd->t_queue = cmd->t_first_queue; } #endif static inline bool action_matches(lrmd_cmd_t *cmd, const char *action, guint interval_ms) { return (cmd->interval_ms == interval_ms) && pcmk__str_eq(cmd->action, action, pcmk__str_casei); } static void log_finished(lrmd_cmd_t * cmd, int exec_time, int queue_time) { char pid_str[32] = { 0, }; int log_level = LOG_INFO; if (cmd->last_pid) { snprintf(pid_str, 32, "%d", cmd->last_pid); } if (pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { log_level = LOG_DEBUG; } #ifdef PCMK__TIME_USE_CGT do_crm_log(log_level, "%s %s (call %d%s%s) exited with status %d" " (execution time %dms, queue time %dms)", cmd->rsc_id, cmd->action, cmd->call_id, - (cmd->last_pid? ", PID " : ""), pid_str, cmd->exec_rc, - exec_time, queue_time); + (cmd->last_pid? ", PID " : ""), pid_str, + cmd->result.exit_status, exec_time, queue_time); #else do_crm_log(log_level, "%s %s (call %d%s%s) exited with status %d", cmd->rsc_id, cmd->action, cmd->call_id, - (cmd->last_pid? ", PID " : ""), pid_str, cmd->exec_rc); + (cmd->last_pid? ", PID " : ""), pid_str, + cmd->result.exit_status); #endif } static void log_execute(lrmd_cmd_t * cmd) { int log_level = LOG_INFO; if (pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { log_level = LOG_DEBUG; } do_crm_log(log_level, "executing - rsc:%s action:%s call_id:%d", cmd->rsc_id, cmd->action, cmd->call_id); } static const char * normalize_action_name(lrmd_rsc_t * rsc, const char *action) { if (pcmk__str_eq(action, "monitor", pcmk__str_casei) && pcmk_is_set(pcmk_get_ra_caps(rsc->class), pcmk_ra_cap_status)) { return "status"; } return action; } static lrmd_rsc_t * build_rsc_from_xml(xmlNode * msg) { xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, msg, LOG_ERR); lrmd_rsc_t *rsc = NULL; rsc = calloc(1, sizeof(lrmd_rsc_t)); crm_element_value_int(msg, F_LRMD_CALLOPTS, &rsc->call_opts); rsc->rsc_id = crm_element_value_copy(rsc_xml, F_LRMD_RSC_ID); rsc->class = crm_element_value_copy(rsc_xml, F_LRMD_CLASS); rsc->provider = crm_element_value_copy(rsc_xml, F_LRMD_PROVIDER); rsc->type = crm_element_value_copy(rsc_xml, F_LRMD_TYPE); rsc->work = mainloop_add_trigger(G_PRIORITY_HIGH, lrmd_rsc_dispatch, rsc); rsc->st_probe_rc = -ENODEV; // if stonith, initialize to "not running" return rsc; } static lrmd_cmd_t * create_lrmd_cmd(xmlNode *msg, pcmk__client_t *client) { int call_options = 0; xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, msg, LOG_ERR); lrmd_cmd_t *cmd = NULL; cmd = calloc(1, sizeof(lrmd_cmd_t)); crm_element_value_int(msg, F_LRMD_CALLOPTS, &call_options); cmd->call_opts = call_options; cmd->client_id = strdup(client->id); crm_element_value_int(msg, F_LRMD_CALLID, &cmd->call_id); crm_element_value_ms(rsc_xml, F_LRMD_RSC_INTERVAL, &cmd->interval_ms); crm_element_value_int(rsc_xml, F_LRMD_TIMEOUT, &cmd->timeout); crm_element_value_int(rsc_xml, F_LRMD_RSC_START_DELAY, &cmd->start_delay); cmd->timeout_orig = cmd->timeout; cmd->origin = crm_element_value_copy(rsc_xml, F_LRMD_ORIGIN); cmd->action = crm_element_value_copy(rsc_xml, F_LRMD_RSC_ACTION); cmd->userdata_str = crm_element_value_copy(rsc_xml, F_LRMD_RSC_USERDATA_STR); cmd->rsc_id = crm_element_value_copy(rsc_xml, F_LRMD_RSC_ID); cmd->params = xml2list(rsc_xml); if (pcmk__str_eq(g_hash_table_lookup(cmd->params, "CRM_meta_on_fail"), "block", pcmk__str_casei)) { crm_debug("Setting flag to leave pid group on timeout and " "only kill action pid for " PCMK__OP_FMT, cmd->rsc_id, cmd->action, cmd->interval_ms); cmd->service_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, "Action", cmd->action, 0, SVC_ACTION_LEAVE_GROUP, "SVC_ACTION_LEAVE_GROUP"); } return cmd; } static void stop_recurring_timer(lrmd_cmd_t *cmd) { if (cmd) { if (cmd->stonith_recurring_id) { g_source_remove(cmd->stonith_recurring_id); } cmd->stonith_recurring_id = 0; } } static void free_lrmd_cmd(lrmd_cmd_t * cmd) { stop_recurring_timer(cmd); if (cmd->delay_id) { g_source_remove(cmd->delay_id); } if (cmd->params) { g_hash_table_destroy(cmd->params); } + pcmk__reset_result(&(cmd->result)); free(cmd->origin); free(cmd->action); free(cmd->real_action); free(cmd->userdata_str); free(cmd->rsc_id); - free(cmd->output); - free(cmd->exit_reason); free(cmd->client_id); free(cmd); } static gboolean stonith_recurring_op_helper(gpointer data) { lrmd_cmd_t *cmd = data; lrmd_rsc_t *rsc; cmd->stonith_recurring_id = 0; if (!cmd->rsc_id) { return FALSE; } rsc = g_hash_table_lookup(rsc_list, cmd->rsc_id); CRM_ASSERT(rsc != NULL); /* take it out of recurring_ops list, and put it in the pending ops * to be executed */ rsc->recurring_ops = g_list_remove(rsc->recurring_ops, cmd); rsc->pending_ops = g_list_append(rsc->pending_ops, cmd); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_queue), &(cmd->t_first_queue)); #endif mainloop_set_trigger(rsc->work); return FALSE; } static inline void start_recurring_timer(lrmd_cmd_t *cmd) { if (cmd && (cmd->interval_ms > 0)) { cmd->stonith_recurring_id = g_timeout_add(cmd->interval_ms, stonith_recurring_op_helper, cmd); } } static gboolean start_delay_helper(gpointer data) { lrmd_cmd_t *cmd = data; lrmd_rsc_t *rsc = NULL; cmd->delay_id = 0; rsc = cmd->rsc_id ? g_hash_table_lookup(rsc_list, cmd->rsc_id) : NULL; if (rsc) { mainloop_set_trigger(rsc->work); } return FALSE; } /*! * \internal * \brief Check whether a list already contains the equivalent of a given action */ static lrmd_cmd_t * find_duplicate_action(GList *action_list, lrmd_cmd_t *cmd) { for (GList *item = action_list; item != NULL; item = item->next) { lrmd_cmd_t *dup = item->data; if (action_matches(cmd, dup->action, dup->interval_ms)) { return dup; } } return NULL; } static bool merge_recurring_duplicate(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { lrmd_cmd_t * dup = NULL; bool dup_pending = true; if (cmd->interval_ms == 0) { return false; } // Search for a duplicate of this action (in-flight or not) dup = find_duplicate_action(rsc->pending_ops, cmd); if (dup == NULL) { dup_pending = false; dup = find_duplicate_action(rsc->recurring_ops, cmd); if (dup == NULL) { return false; } } /* Do not merge fencing monitors marked for cancellation, so we can reply to * the cancellation separately. */ if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei) - && (dup->lrmd_op_status == PCMK_EXEC_CANCELLED)) { + && (dup->result.execution_status == PCMK_EXEC_CANCELLED)) { return false; } /* This should not occur. If it does, we need to investigate how something * like this is possible in the controller. */ crm_warn("Duplicate recurring op entry detected (" PCMK__OP_FMT "), merging with previous op entry", rsc->rsc_id, normalize_action_name(rsc, dup->action), dup->interval_ms); // Merge new action's call ID and user data into existing action dup->first_notify_sent = false; free(dup->userdata_str); dup->userdata_str = cmd->userdata_str; cmd->userdata_str = NULL; dup->call_id = cmd->call_id; free_lrmd_cmd(cmd); cmd = NULL; /* If dup is not pending, that means it has already executed at least once * and is waiting in the interval. In that case, stop waiting and initiate * a new instance now. */ if (!dup_pending) { if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { stop_recurring_timer(dup); stonith_recurring_op_helper(dup); } else { services_action_kick(rsc->rsc_id, normalize_action_name(rsc, dup->action), dup->interval_ms); } } return true; } static void schedule_lrmd_cmd(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { CRM_CHECK(cmd != NULL, return); CRM_CHECK(rsc != NULL, return); crm_trace("Scheduling %s on %s", cmd->action, rsc->rsc_id); if (merge_recurring_duplicate(rsc, cmd)) { // Equivalent of cmd has already been scheduled return; } /* The controller expects the executor to automatically cancel * recurring operations before a resource stops. */ if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { cancel_all_recurring(rsc, NULL); } rsc->pending_ops = g_list_append(rsc->pending_ops, cmd); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_queue), &(cmd->t_first_queue)); #endif mainloop_set_trigger(rsc->work); if (cmd->start_delay) { cmd->delay_id = g_timeout_add(cmd->start_delay, start_delay_helper, cmd); } } static xmlNode * create_lrmd_reply(const char *origin, int rc, int call_id) { xmlNode *reply = create_xml_node(NULL, T_LRMD_REPLY); crm_xml_add(reply, F_LRMD_ORIGIN, origin); crm_xml_add_int(reply, F_LRMD_RC, rc); crm_xml_add_int(reply, F_LRMD_CALLID, call_id); return reply; } static void send_client_notify(gpointer key, gpointer value, gpointer user_data) { xmlNode *update_msg = user_data; pcmk__client_t *client = value; int rc; int log_level = LOG_WARNING; const char *msg = NULL; CRM_CHECK(client != NULL, return); if (client->name == NULL) { crm_trace("Skipping notification to client without name"); return; } if (pcmk_is_set(client->flags, pcmk__client_to_proxy)) { /* We only want to notify clients of the executor IPC API. If we are * running as Pacemaker Remote, we may have clients proxied to other * IPC services in the cluster, so skip those. */ crm_trace("Skipping executor API notification to client %s", pcmk__client_name(client)); return; } rc = lrmd_server_send_notify(client, update_msg); if (rc == pcmk_rc_ok) { return; } switch (rc) { case ENOTCONN: case EPIPE: // Client exited without waiting for notification log_level = LOG_INFO; msg = "Disconnected"; break; default: msg = pcmk_rc_str(rc); break; } do_crm_log(log_level, "Could not notify client %s: %s " CRM_XS " rc=%d", pcmk__client_name(client), msg, rc); } static void send_cmd_complete_notify(lrmd_cmd_t * cmd) { xmlNode *notify = NULL; #ifdef PCMK__TIME_USE_CGT int exec_time = time_diff_ms(NULL, &(cmd->t_run)); int queue_time = time_diff_ms(&cmd->t_run, &(cmd->t_queue)); log_finished(cmd, exec_time, queue_time); #else log_finished(cmd, 0, 0); #endif /* if the first notify result for a cmd has already been sent earlier, and the * the option to only send notifies on result changes is set. Check to see * if the last result is the same as the new one. If so, suppress this update */ if (cmd->first_notify_sent && (cmd->call_opts & lrmd_opt_notify_changes_only)) { - if (cmd->last_notify_rc == cmd->exec_rc && - cmd->last_notify_op_status == cmd->lrmd_op_status) { + if ((cmd->last_notify_rc == cmd->result.exit_status) && + (cmd->last_notify_op_status == cmd->result.execution_status)) { /* only send changes */ return; } } cmd->first_notify_sent = true; - cmd->last_notify_rc = cmd->exec_rc; - cmd->last_notify_op_status = cmd->lrmd_op_status; + cmd->last_notify_rc = cmd->result.exit_status; + cmd->last_notify_op_status = cmd->result.execution_status; notify = create_xml_node(NULL, T_LRMD_NOTIFY); crm_xml_add(notify, F_LRMD_ORIGIN, __func__); crm_xml_add_int(notify, F_LRMD_TIMEOUT, cmd->timeout); crm_xml_add_ms(notify, F_LRMD_RSC_INTERVAL, cmd->interval_ms); crm_xml_add_int(notify, F_LRMD_RSC_START_DELAY, cmd->start_delay); - crm_xml_add_int(notify, F_LRMD_EXEC_RC, cmd->exec_rc); - crm_xml_add_int(notify, F_LRMD_OP_STATUS, cmd->lrmd_op_status); + crm_xml_add_int(notify, F_LRMD_EXEC_RC, cmd->result.exit_status); + crm_xml_add_int(notify, F_LRMD_OP_STATUS, cmd->result.execution_status); crm_xml_add_int(notify, F_LRMD_CALLID, cmd->call_id); crm_xml_add_int(notify, F_LRMD_RSC_DELETED, cmd->rsc_deleted); crm_xml_add_ll(notify, F_LRMD_RSC_RUN_TIME, (long long) cmd->epoch_last_run); crm_xml_add_ll(notify, F_LRMD_RSC_RCCHANGE_TIME, (long long) cmd->epoch_rcchange); #ifdef PCMK__TIME_USE_CGT crm_xml_add_int(notify, F_LRMD_RSC_EXEC_TIME, exec_time); crm_xml_add_int(notify, F_LRMD_RSC_QUEUE_TIME, queue_time); #endif crm_xml_add(notify, F_LRMD_OPERATION, LRMD_OP_RSC_EXEC); crm_xml_add(notify, F_LRMD_RSC_ID, cmd->rsc_id); if(cmd->real_action) { crm_xml_add(notify, F_LRMD_RSC_ACTION, cmd->real_action); } else { crm_xml_add(notify, F_LRMD_RSC_ACTION, cmd->action); } crm_xml_add(notify, F_LRMD_RSC_USERDATA_STR, cmd->userdata_str); - crm_xml_add(notify, F_LRMD_RSC_OUTPUT, cmd->output); - crm_xml_add(notify, F_LRMD_RSC_EXIT_REASON, cmd->exit_reason); + crm_xml_add(notify, F_LRMD_RSC_EXIT_REASON, cmd->result.exit_reason); + + if (cmd->result.action_stderr != NULL) { + crm_xml_add(notify, F_LRMD_RSC_OUTPUT, cmd->result.action_stderr); + + } else if (cmd->result.action_stdout != NULL) { + crm_xml_add(notify, F_LRMD_RSC_OUTPUT, cmd->result.action_stdout); + } if (cmd->params) { char *key = NULL; char *value = NULL; GHashTableIter iter; xmlNode *args = create_xml_node(notify, XML_TAG_ATTRS); g_hash_table_iter_init(&iter, cmd->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { hash2smartfield((gpointer) key, (gpointer) value, args); } } if (cmd->client_id && (cmd->call_opts & lrmd_opt_notify_orig_only)) { pcmk__client_t *client = pcmk__find_client_by_id(cmd->client_id); if (client) { send_client_notify(client->id, client, notify); } } else { pcmk__foreach_ipc_client(send_client_notify, notify); } free_xml(notify); } static void send_generic_notify(int rc, xmlNode * request) { if (pcmk__ipc_client_count() != 0) { int call_id = 0; xmlNode *notify = NULL; xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, request, LOG_ERR); const char *rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); const char *op = crm_element_value(request, F_LRMD_OPERATION); crm_element_value_int(request, F_LRMD_CALLID, &call_id); notify = create_xml_node(NULL, T_LRMD_NOTIFY); crm_xml_add(notify, F_LRMD_ORIGIN, __func__); crm_xml_add_int(notify, F_LRMD_RC, rc); crm_xml_add_int(notify, F_LRMD_CALLID, call_id); crm_xml_add(notify, F_LRMD_OPERATION, op); crm_xml_add(notify, F_LRMD_RSC_ID, rsc_id); pcmk__foreach_ipc_client(send_client_notify, notify); free_xml(notify); } } static void cmd_reset(lrmd_cmd_t * cmd) { - cmd->lrmd_op_status = 0; cmd->last_pid = 0; #ifdef PCMK__TIME_USE_CGT memset(&cmd->t_run, 0, sizeof(cmd->t_run)); memset(&cmd->t_queue, 0, sizeof(cmd->t_queue)); #endif cmd->epoch_last_run = 0; - free(cmd->exit_reason); - cmd->exit_reason = NULL; - free(cmd->output); - cmd->output = NULL; + + pcmk__reset_result(&(cmd->result)); + cmd->result.execution_status = PCMK_EXEC_DONE; } static void cmd_finalize(lrmd_cmd_t * cmd, lrmd_rsc_t * rsc) { crm_trace("Resource operation rsc:%s action:%s completed (%p %p)", cmd->rsc_id, cmd->action, rsc ? rsc->active : NULL, cmd); if (rsc && (rsc->active == cmd)) { rsc->active = NULL; mainloop_set_trigger(rsc->work); } if (!rsc) { cmd->rsc_deleted = 1; } /* reset original timeout so client notification has correct information */ cmd->timeout = cmd->timeout_orig; send_cmd_complete_notify(cmd); - if (cmd->interval_ms && (cmd->lrmd_op_status == PCMK_EXEC_CANCELLED)) { + if ((cmd->interval_ms != 0) + && (cmd->result.execution_status == PCMK_EXEC_CANCELLED)) { + if (rsc) { rsc->recurring_ops = g_list_remove(rsc->recurring_ops, cmd); rsc->pending_ops = g_list_remove(rsc->pending_ops, cmd); } free_lrmd_cmd(cmd); } else if (cmd->interval_ms == 0) { if (rsc) { rsc->pending_ops = g_list_remove(rsc->pending_ops, cmd); } free_lrmd_cmd(cmd); } else { /* Clear all the values pertaining just to the last iteration of a recurring op. */ cmd_reset(cmd); } } static int ocf2uniform_rc(int rc) { switch (rc) { case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: break; default: if (rc < 0 || rc > PCMK_OCF_FAILED_PROMOTED) { return PCMK_OCF_UNKNOWN_ERROR; } } return rc; } static int stonith2uniform_rc(const char *action, int rc) { switch (rc) { case pcmk_ok: rc = PCMK_OCF_OK; break; case -ENODEV: /* This should be possible only for probes in practice, but * interpret for all actions to be safe. */ if (pcmk__str_eq(action, "monitor", pcmk__str_casei)) { rc = PCMK_OCF_NOT_RUNNING; } else if (pcmk__str_eq(action, "stop", pcmk__str_casei)) { rc = PCMK_OCF_OK; } else { rc = PCMK_OCF_NOT_INSTALLED; } break; case -EOPNOTSUPP: rc = PCMK_OCF_UNIMPLEMENT_FEATURE; break; case -ETIME: case -ETIMEDOUT: rc = PCMK_OCF_TIMEOUT; break; default: rc = PCMK_OCF_UNKNOWN_ERROR; break; } return rc; } #if SUPPORT_NAGIOS static int nagios2uniform_rc(const char *action, int rc) { if (rc < 0) { return PCMK_OCF_UNKNOWN_ERROR; } switch (rc) { case NAGIOS_STATE_OK: return PCMK_OCF_OK; case NAGIOS_INSUFFICIENT_PRIV: return PCMK_OCF_INSUFFICIENT_PRIV; case NAGIOS_NOT_INSTALLED: return PCMK_OCF_NOT_INSTALLED; case NAGIOS_STATE_WARNING: case NAGIOS_STATE_CRITICAL: case NAGIOS_STATE_UNKNOWN: case NAGIOS_STATE_DEPENDENT: default: return PCMK_OCF_UNKNOWN_ERROR; } return PCMK_OCF_UNKNOWN_ERROR; } #endif static int get_uniform_rc(const char *standard, const char *action, int rc) { if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_casei)) { return ocf2uniform_rc(rc); } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { return stonith2uniform_rc(action, rc); } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { return rc; } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_UPSTART, pcmk__str_casei)) { return rc; #if SUPPORT_NAGIOS } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return nagios2uniform_rc(action, rc); #endif } else { return services_get_ocf_exitcode(action, rc); } } static int action_get_uniform_rc(svc_action_t * action) { lrmd_cmd_t *cmd = action->cb_data; return get_uniform_rc(action->standard, cmd->action, action->rc); } struct notify_new_client_data { xmlNode *notify; pcmk__client_t *new_client; }; static void notify_one_client(gpointer key, gpointer value, gpointer user_data) { pcmk__client_t *client = value; struct notify_new_client_data *data = user_data; if (!pcmk__str_eq(client->id, data->new_client->id, pcmk__str_casei)) { send_client_notify(key, (gpointer) client, (gpointer) data->notify); } } void notify_of_new_client(pcmk__client_t *new_client) { struct notify_new_client_data data; data.new_client = new_client; data.notify = create_xml_node(NULL, T_LRMD_NOTIFY); crm_xml_add(data.notify, F_LRMD_ORIGIN, __func__); crm_xml_add(data.notify, F_LRMD_OPERATION, LRMD_OP_NEW_CLIENT); pcmk__foreach_ipc_client(notify_one_client, &data); free_xml(data.notify); } static char * parse_exit_reason(const char *output) { const char *cur = NULL; const char *last = NULL; static int cookie_len = 0; char *eol = NULL; size_t reason_len = EXIT_REASON_MAX_LEN; if (output == NULL) { return NULL; } if (!cookie_len) { cookie_len = strlen(PCMK_OCF_REASON_PREFIX); } cur = strstr(output, PCMK_OCF_REASON_PREFIX); for (; cur != NULL; cur = strstr(cur, PCMK_OCF_REASON_PREFIX)) { /* skip over the cookie delimiter string */ cur += cookie_len; last = cur; } if (last == NULL) { return NULL; } // Truncate everything after a new line, and limit reason string size eol = strchr(last, '\n'); if (eol) { reason_len = QB_MIN(reason_len, eol - last); } return strndup(last, reason_len); } void client_disconnect_cleanup(const char *client_id) { GHashTableIter iter; lrmd_rsc_t *rsc = NULL; char *key = NULL; g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & rsc)) { if (rsc->call_opts & lrmd_opt_drop_recurring) { /* This client is disconnecting, drop any recurring operations * it may have initiated on the resource */ cancel_all_recurring(rsc, client_id); } } } static void action_complete(svc_action_t * action) { lrmd_rsc_t *rsc; lrmd_cmd_t *cmd = action->cb_data; #ifdef PCMK__TIME_USE_CGT const char *rclass = NULL; bool goagain = false; #endif if (!cmd) { crm_err("Completed executor action (%s) does not match any known operations", action->id); return; } #ifdef PCMK__TIME_USE_CGT - if (cmd->exec_rc != action->rc) { + if (cmd->result.exit_status != action->rc) { cmd->epoch_rcchange = time(NULL); } #endif cmd->last_pid = action->pid; - cmd->exec_rc = action_get_uniform_rc(action); - cmd->lrmd_op_status = action->status; + pcmk__set_result(&(cmd->result), action_get_uniform_rc(action), + action->status, NULL); rsc = cmd->rsc_id ? g_hash_table_lookup(rsc_list, cmd->rsc_id) : NULL; #ifdef PCMK__TIME_USE_CGT if (rsc && pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_SERVICE, pcmk__str_casei)) { rclass = resources_find_service_class(rsc->type); } else if(rsc) { rclass = rsc->class; } if (pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { - if ((cmd->exec_rc == PCMK_OCF_OK) + if ((cmd->result.exit_status == PCMK_OCF_OK) && pcmk__strcase_any_of(cmd->action, "start", "stop", NULL)) { /* systemd returns from start and stop actions after the action * begins, not after it completes. We have to jump through a few * hoops so that we don't report 'complete' to the rest of pacemaker * until it's actually done. */ goagain = true; cmd->real_action = cmd->action; cmd->action = strdup("monitor"); } else if (cmd->real_action != NULL) { // This is follow-up monitor to check whether start/stop completed - if ((cmd->lrmd_op_status == PCMK_EXEC_DONE) - && (cmd->exec_rc == PCMK_OCF_PENDING)) { + if ((cmd->result.execution_status == PCMK_EXEC_DONE) + && (cmd->result.exit_status == PCMK_OCF_PENDING)) { goagain = true; - } else if ((cmd->exec_rc == PCMK_OCF_OK) + } else if ((cmd->result.exit_status == PCMK_OCF_OK) && pcmk__str_eq(cmd->real_action, "stop", pcmk__str_casei)) { goagain = true; } else { int time_sum = time_diff_ms(NULL, &(cmd->t_first_run)); int timeout_left = cmd->timeout_orig - time_sum; crm_debug("%s systemd %s is now complete (elapsed=%dms, " "remaining=%dms): %s (%d)", cmd->rsc_id, cmd->real_action, time_sum, timeout_left, - services_ocf_exitcode_str(cmd->exec_rc), - cmd->exec_rc); + services_ocf_exitcode_str(cmd->result.exit_status), + cmd->result.exit_status); cmd_original_times(cmd); // Monitors may return "not running", but start/stop shouldn't - if ((cmd->lrmd_op_status == PCMK_EXEC_DONE) - && (cmd->exec_rc == PCMK_OCF_NOT_RUNNING)) { + if ((cmd->result.execution_status == PCMK_EXEC_DONE) + && (cmd->result.exit_status == PCMK_OCF_NOT_RUNNING)) { if (pcmk__str_eq(cmd->real_action, "start", pcmk__str_casei)) { - cmd->exec_rc = PCMK_OCF_UNKNOWN_ERROR; + cmd->result.exit_status = PCMK_OCF_UNKNOWN_ERROR; } else if (pcmk__str_eq(cmd->real_action, "stop", pcmk__str_casei)) { - cmd->exec_rc = PCMK_OCF_OK; + cmd->result.exit_status = PCMK_OCF_OK; } } } } } #endif #if SUPPORT_NAGIOS if (rsc && pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { if (action_matches(cmd, "monitor", 0) - && (cmd->exec_rc == PCMK_OCF_OK)) { + && (cmd->result.exit_status == PCMK_OCF_OK)) { /* Successfully executed --version for the nagios plugin */ - cmd->exec_rc = PCMK_OCF_NOT_RUNNING; + cmd->result.exit_status = PCMK_OCF_NOT_RUNNING; - } else if (pcmk__str_eq(cmd->action, "start", pcmk__str_casei) && cmd->exec_rc != PCMK_OCF_OK) { + } else if (pcmk__str_eq(cmd->action, "start", pcmk__str_casei) + && (cmd->result.exit_status != PCMK_OCF_OK)) { #ifdef PCMK__TIME_USE_CGT goagain = true; #endif } } #endif #ifdef PCMK__TIME_USE_CGT if (goagain) { int time_sum = time_diff_ms(NULL, &(cmd->t_first_run)); int timeout_left = cmd->timeout_orig - time_sum; int delay = cmd->timeout_orig / 10; if(delay >= timeout_left && timeout_left > 20) { delay = timeout_left/2; } delay = QB_MIN(2000, delay); if (delay < timeout_left) { cmd->start_delay = delay; cmd->timeout = timeout_left; - if(cmd->exec_rc == PCMK_OCF_OK) { + if (cmd->result.exit_status == PCMK_OCF_OK) { crm_debug("%s %s may still be in progress: re-scheduling (elapsed=%dms, remaining=%dms, start_delay=%dms)", cmd->rsc_id, cmd->real_action, time_sum, timeout_left, delay); - } else if(cmd->exec_rc == PCMK_OCF_PENDING) { + } else if (cmd->result.exit_status == PCMK_OCF_PENDING) { crm_info("%s %s is still in progress: re-scheduling (elapsed=%dms, remaining=%dms, start_delay=%dms)", cmd->rsc_id, cmd->action, time_sum, timeout_left, delay); } else { crm_notice("%s %s failed '%s' (%d): re-scheduling (elapsed=%dms, remaining=%dms, start_delay=%dms)", - cmd->rsc_id, cmd->action, services_ocf_exitcode_str(cmd->exec_rc), cmd->exec_rc, time_sum, timeout_left, delay); + cmd->rsc_id, cmd->action, + services_ocf_exitcode_str(cmd->result.exit_status), + cmd->result.exit_status, time_sum, timeout_left, + delay); } cmd_reset(cmd); if(rsc) { rsc->active = NULL; } schedule_lrmd_cmd(rsc, cmd); /* Don't finalize cmd, we're not done with it yet */ return; } else { crm_notice("Giving up on %s %s (rc=%d): timeout (elapsed=%dms, remaining=%dms)", - cmd->rsc_id, cmd->real_action?cmd->real_action:cmd->action, cmd->exec_rc, time_sum, timeout_left); - cmd->lrmd_op_status = PCMK_EXEC_TIMEOUT; - cmd->exec_rc = PCMK_OCF_TIMEOUT; + cmd->rsc_id, + (cmd->real_action? cmd->real_action : cmd->action), + cmd->result.exit_status, time_sum, timeout_left); + pcmk__set_result(&(cmd->result), PCMK_OCF_TIMEOUT, + PCMK_EXEC_TIMEOUT, + "Investigate reason for timeout, and adjust " + "configured operation timeout if necessary"); cmd_original_times(cmd); } } #endif + pcmk__set_result_output(&(cmd->result), + action->stdout_data, action->stderr_data); if (action->stderr_data) { - cmd->output = strdup(action->stderr_data); - cmd->exit_reason = parse_exit_reason(action->stderr_data); - - } else if (action->stdout_data) { - cmd->output = strdup(action->stdout_data); + cmd->result.exit_reason = parse_exit_reason(action->stderr_data); } cmd_finalize(cmd, rsc); } /*! * \internal * \brief Determine operation status of a stonith operation * * Non-stonith resource operations get their operation status directly from the * service library, but the fencer does not have an equivalent, so we must infer * an operation status from the fencer API's return code. * * \param[in] action Name of action performed on stonith resource * \param[in] interval_ms Action interval * \param[in] rc Action result from fencer * * \return Operation status corresponding to fencer API return code */ static int stonith_rc2status(const char *action, guint interval_ms, int rc) { int status = PCMK_EXEC_DONE; switch (rc) { case pcmk_ok: break; case -EOPNOTSUPP: case -EPROTONOSUPPORT: status = PCMK_EXEC_NOT_SUPPORTED; break; case -ETIME: case -ETIMEDOUT: status = PCMK_EXEC_TIMEOUT; break; case -ENOTCONN: case -ECOMM: // Couldn't talk to fencer status = PCMK_EXEC_ERROR; break; case -ENODEV: // The device is not registered with the fencer status = PCMK_EXEC_ERROR; break; default: break; } return status; } static void stonith_action_complete(lrmd_cmd_t * cmd, int rc) { // This can be NULL if resource was removed before command completed lrmd_rsc_t *rsc = g_hash_table_lookup(rsc_list, cmd->rsc_id); - cmd->exec_rc = stonith2uniform_rc(cmd->action, rc); + cmd->result.exit_status = stonith2uniform_rc(cmd->action, rc); /* This function may be called with status already set to cancelled, if a * pending action was aborted. Otherwise, we need to determine status from * the fencer return code. */ - if (cmd->lrmd_op_status != PCMK_EXEC_CANCELLED) { - cmd->lrmd_op_status = stonith_rc2status(cmd->action, cmd->interval_ms, - rc); + if (cmd->result.execution_status != PCMK_EXEC_CANCELLED) { + cmd->result.execution_status = stonith_rc2status(cmd->action, + cmd->interval_ms, rc); // Certain successful actions change the known state of the resource - if (rsc && (cmd->exec_rc == PCMK_OCF_OK)) { + if ((rsc != NULL) && (cmd->result.exit_status == PCMK_OCF_OK)) { if (pcmk__str_eq(cmd->action, "start", pcmk__str_casei)) { rsc->st_probe_rc = pcmk_ok; // maps to PCMK_OCF_OK } else if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { rsc->st_probe_rc = -ENODEV; // maps to PCMK_OCF_NOT_RUNNING } } } + // Give the user more detail than an OCF code + if (rc != -pcmk_err_generic) { + cmd->result.exit_reason = strdup(pcmk_strerror(rc)); + } + /* The recurring timer should not be running at this point in any case, but * as a failsafe, stop it if it is. */ stop_recurring_timer(cmd); /* Reschedule this command if appropriate. If a recurring command is *not* * rescheduled, its status must be PCMK_EXEC_CANCELLED, otherwise it will * not be removed from recurring_ops by cmd_finalize(). */ if (rsc && (cmd->interval_ms > 0) - && (cmd->lrmd_op_status != PCMK_EXEC_CANCELLED)) { + && (cmd->result.execution_status != PCMK_EXEC_CANCELLED)) { start_recurring_timer(cmd); } cmd_finalize(cmd, rsc); } static void lrmd_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data) { stonith_action_complete(data->userdata, data->rc); } void stonith_connection_failed(void) { GHashTableIter iter; GList *cmd_list = NULL; GList *cmd_iter = NULL; lrmd_rsc_t *rsc = NULL; char *key = NULL; g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & rsc)) { if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { /* If we registered this fence device, we don't know whether the * fencer still has the registration or not. Cause future probes to * return PCMK_OCF_UNKNOWN_ERROR until the resource is stopped or * started successfully. This is especially important if the * controller also went away (possibly due to a cluster layer * restart) and won't receive our client notification of any * monitors finalized below. */ if (rsc->st_probe_rc == pcmk_ok) { rsc->st_probe_rc = pcmk_err_generic; } if (rsc->active) { cmd_list = g_list_append(cmd_list, rsc->active); } if (rsc->recurring_ops) { cmd_list = g_list_concat(cmd_list, rsc->recurring_ops); } if (rsc->pending_ops) { cmd_list = g_list_concat(cmd_list, rsc->pending_ops); } rsc->pending_ops = rsc->recurring_ops = NULL; } } if (!cmd_list) { return; } crm_err("Connection to fencer failed, finalizing %d pending operations", g_list_length(cmd_list)); for (cmd_iter = cmd_list; cmd_iter; cmd_iter = cmd_iter->next) { stonith_action_complete(cmd_iter->data, -ENOTCONN); } g_list_free(cmd_list); } /*! * \internal * \brief Execute a stonith resource "start" action * * Start a stonith resource by registering it with the fencer. * (Stonith agents don't have a start command.) * * \param[in] stonith_api Connection to fencer * \param[in] rsc Stonith resource to start * \param[in] cmd Start command to execute * * \return pcmk_ok on success, -errno otherwise */ static int execd_stonith_start(stonith_t *stonith_api, lrmd_rsc_t *rsc, lrmd_cmd_t *cmd) { char *key = NULL; char *value = NULL; stonith_key_value_t *device_params = NULL; int rc = pcmk_ok; // Convert command parameters to stonith API key/values if (cmd->params) { GHashTableIter iter; g_hash_table_iter_init(&iter, cmd->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { device_params = stonith_key_value_add(device_params, key, value); } } /* The fencer will automatically register devices via CIB notifications * when the CIB changes, but to avoid a possible race condition between * the fencer receiving the notification and the executor requesting that * resource, the executor registers the device as well. The fencer knows how * to handle duplicate registrations. */ rc = stonith_api->cmds->register_device(stonith_api, st_opt_sync_call, cmd->rsc_id, rsc->provider, rsc->type, device_params); stonith_key_value_freeall(device_params, 1, 1); return rc; } /*! * \internal * \brief Execute a stonith resource "stop" action * * Stop a stonith resource by unregistering it with the fencer. * (Stonith agents don't have a stop command.) * * \param[in] stonith_api Connection to fencer * \param[in] rsc Stonith resource to stop * * \return pcmk_ok on success, -errno otherwise */ static inline int execd_stonith_stop(stonith_t *stonith_api, const lrmd_rsc_t *rsc) { /* @TODO Failure would indicate a problem communicating with fencer; * perhaps we should try reconnecting and retrying a few times? */ return stonith_api->cmds->remove_device(stonith_api, st_opt_sync_call, rsc->rsc_id); } /*! * \internal * \brief Initiate a stonith resource agent recurring "monitor" action * * \param[in] stonith_api Connection to fencer * \param[in] rsc Stonith resource to monitor * \param[in] cmd Monitor command being executed * * \return pcmk_ok if monitor was successfully initiated, -errno otherwise */ static inline int execd_stonith_monitor(stonith_t *stonith_api, lrmd_rsc_t *rsc, lrmd_cmd_t *cmd) { int rc = stonith_api->cmds->monitor(stonith_api, 0, cmd->rsc_id, cmd->timeout / 1000); rc = stonith_api->cmds->register_callback(stonith_api, rc, 0, 0, cmd, "lrmd_stonith_callback", lrmd_stonith_callback); if (rc == TRUE) { rsc->active = cmd; rc = pcmk_ok; } else { rc = -pcmk_err_generic; } return rc; } static void lrmd_rsc_execute_stonith(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { int rc = 0; bool do_monitor = FALSE; stonith_t *stonith_api = get_stonith_connection(); if (!stonith_api) { rc = -ENOTCONN; } else if (pcmk__str_eq(cmd->action, "start", pcmk__str_casei)) { rc = execd_stonith_start(stonith_api, rsc, cmd); if (rc == 0) { do_monitor = TRUE; } } else if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { rc = execd_stonith_stop(stonith_api, rsc); } else if (pcmk__str_eq(cmd->action, "monitor", pcmk__str_casei)) { if (cmd->interval_ms > 0) { do_monitor = TRUE; } else { rc = rsc->st_probe_rc; } } if (do_monitor) { rc = execd_stonith_monitor(stonith_api, rsc, cmd); if (rc == pcmk_ok) { // Don't clean up yet, we will find out result of the monitor later return; } } stonith_action_complete(cmd, rc); } static int lrmd_rsc_execute_service_lib(lrmd_rsc_t * rsc, lrmd_cmd_t * cmd) { svc_action_t *action = NULL; GHashTable *params_copy = NULL; CRM_ASSERT(rsc); CRM_ASSERT(cmd); crm_trace("Creating action, resource:%s action:%s class:%s provider:%s agent:%s", rsc->rsc_id, cmd->action, rsc->class, rsc->provider, rsc->type); #if SUPPORT_NAGIOS /* Recurring operations are cancelled anyway for a stop operation */ if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei) && pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { - cmd->exec_rc = PCMK_OCF_OK; + cmd->result.exit_status = PCMK_OCF_OK; goto exec_done; } #endif params_copy = pcmk__str_table_dup(cmd->params); action = services__create_resource_action(rsc->rsc_id, rsc->class, rsc->provider, rsc->type, normalize_action_name(rsc, cmd->action), cmd->interval_ms, cmd->timeout, params_copy, cmd->service_flags); if (!action) { + // Invalid arguments (which would be a bug) or out-of-memory crm_err("Failed to create action, action:%s on resource %s", cmd->action, rsc->rsc_id); - cmd->exec_rc = PCMK_OCF_UNKNOWN_ERROR; - cmd->lrmd_op_status = PCMK_EXEC_ERROR; + pcmk__set_result(&(cmd->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR, "Internal Pacemaker error"); goto exec_done; } - if (action->rc != 0) { - cmd->exec_rc = action->rc; - cmd->lrmd_op_status = action->status; + if (action->rc != PCMK_OCF_OK) { + pcmk__set_result(&(cmd->result), action->rc, action->status, NULL); services_action_free(action); goto exec_done; } action->cb_data = cmd; /* 'cmd' may not be valid after this point if * services_action_async() returned TRUE * * Upstart and systemd both synchronously determine monitor/status * results and call action_complete (which may free 'cmd') if necessary. */ if (services_action_async(action, action_complete)) { return TRUE; } - cmd->exec_rc = action->rc; - if (action->status != PCMK_EXEC_DONE) { - cmd->lrmd_op_status = action->status; - } else { - cmd->lrmd_op_status = PCMK_EXEC_ERROR; + /* Asynchronous execution could not be initiated. services_action_async() + * should have already set an appropriate execution status, but as a + * fail-safe for cases where we've neglected to do so, make sure this is + * considered an execution error. + */ + if (action->status == PCMK_EXEC_DONE) { + action->status = PCMK_EXEC_ERROR; } + + pcmk__set_result(&(cmd->result), action->rc, action->status, NULL); services_action_free(action); action = NULL; exec_done: cmd_finalize(cmd, rsc); return TRUE; } static gboolean lrmd_rsc_execute(lrmd_rsc_t * rsc) { lrmd_cmd_t *cmd = NULL; CRM_CHECK(rsc != NULL, return FALSE); if (rsc->active) { crm_trace("%s is still active", rsc->rsc_id); return TRUE; } if (rsc->pending_ops) { GList *first = rsc->pending_ops; cmd = first->data; if (cmd->delay_id) { crm_trace ("Command %s %s was asked to run too early, waiting for start_delay timeout of %dms", cmd->rsc_id, cmd->action, cmd->start_delay); return TRUE; } rsc->pending_ops = g_list_remove_link(rsc->pending_ops, first); g_list_free_1(first); #ifdef PCMK__TIME_USE_CGT get_current_time(&(cmd->t_run), &(cmd->t_first_run)); #endif cmd->epoch_last_run = time(NULL); } if (!cmd) { crm_trace("Nothing further to do for %s", rsc->rsc_id); return TRUE; } rsc->active = cmd; /* only one op at a time for a rsc */ if (cmd->interval_ms) { rsc->recurring_ops = g_list_append(rsc->recurring_ops, cmd); } log_execute(cmd); if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { lrmd_rsc_execute_stonith(rsc, cmd); } else { lrmd_rsc_execute_service_lib(rsc, cmd); } return TRUE; } static gboolean lrmd_rsc_dispatch(gpointer user_data) { return lrmd_rsc_execute(user_data); } void free_rsc(gpointer data) { GList *gIter = NULL; lrmd_rsc_t *rsc = data; int is_stonith = pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei); gIter = rsc->pending_ops; while (gIter != NULL) { GList *next = gIter->next; lrmd_cmd_t *cmd = gIter->data; /* command was never executed */ - cmd->lrmd_op_status = PCMK_EXEC_CANCELLED; + cmd->result.execution_status = PCMK_EXEC_CANCELLED; cmd_finalize(cmd, NULL); gIter = next; } /* frees list, but not list elements. */ g_list_free(rsc->pending_ops); gIter = rsc->recurring_ops; while (gIter != NULL) { GList *next = gIter->next; lrmd_cmd_t *cmd = gIter->data; if (is_stonith) { - cmd->lrmd_op_status = PCMK_EXEC_CANCELLED; + cmd->result.execution_status = PCMK_EXEC_CANCELLED; /* If a stonith command is in-flight, just mark it as cancelled; * it is not safe to finalize/free the cmd until the stonith api * says it has either completed or timed out. */ if (rsc->active != cmd) { cmd_finalize(cmd, NULL); } } else { /* This command is already handed off to service library, * let service library cancel it and tell us via the callback * when it is cancelled. The rsc can be safely destroyed * even if we are waiting for the cancel result */ services_action_cancel(rsc->rsc_id, normalize_action_name(rsc, cmd->action), cmd->interval_ms); } gIter = next; } /* frees list, but not list elements. */ g_list_free(rsc->recurring_ops); free(rsc->rsc_id); free(rsc->class); free(rsc->provider); free(rsc->type); mainloop_destroy_trigger(rsc->work); free(rsc); } static int process_lrmd_signon(pcmk__client_t *client, xmlNode *request, int call_id, xmlNode **reply) { int rc = pcmk_ok; const char *is_ipc_provider = crm_element_value(request, F_LRMD_IS_IPC_PROVIDER); const char *protocol_version = crm_element_value(request, F_LRMD_PROTOCOL_VERSION); if (compare_version(protocol_version, LRMD_MIN_PROTOCOL_VERSION) < 0) { crm_err("Cluster API version must be greater than or equal to %s, not %s", LRMD_MIN_PROTOCOL_VERSION, protocol_version); rc = -EPROTO; } if (crm_is_true(is_ipc_provider)) { #ifdef PCMK__COMPILE_REMOTE if ((client->remote != NULL) && client->remote->tls_handshake_complete) { // This is a remote connection from a cluster node's controller ipc_proxy_add_provider(client); } else { rc = -EACCES; } #else rc = -EPROTONOSUPPORT; #endif } *reply = create_lrmd_reply(__func__, rc, call_id); crm_xml_add(*reply, F_LRMD_OPERATION, CRM_OP_REGISTER); crm_xml_add(*reply, F_LRMD_CLIENTID, client->id); crm_xml_add(*reply, F_LRMD_PROTOCOL_VERSION, LRMD_PROTOCOL_VERSION); return rc; } static int process_lrmd_rsc_register(pcmk__client_t *client, uint32_t id, xmlNode *request) { int rc = pcmk_ok; lrmd_rsc_t *rsc = build_rsc_from_xml(request); lrmd_rsc_t *dup = g_hash_table_lookup(rsc_list, rsc->rsc_id); if (dup && pcmk__str_eq(rsc->class, dup->class, pcmk__str_casei) && pcmk__str_eq(rsc->provider, dup->provider, pcmk__str_casei) && pcmk__str_eq(rsc->type, dup->type, pcmk__str_casei)) { crm_notice("Ignoring duplicate registration of '%s'", rsc->rsc_id); free_rsc(rsc); return rc; } g_hash_table_replace(rsc_list, rsc->rsc_id, rsc); crm_info("Cached agent information for '%s'", rsc->rsc_id); return rc; } static xmlNode * process_lrmd_get_rsc_info(xmlNode *request, int call_id) { int rc = pcmk_ok; xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, request, LOG_ERR); const char *rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); xmlNode *reply = NULL; lrmd_rsc_t *rsc = NULL; if (rsc_id == NULL) { rc = -ENODEV; } else { rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Agent information for '%s' not in cache", rsc_id); rc = -ENODEV; } } reply = create_lrmd_reply(__func__, rc, call_id); if (rsc) { crm_xml_add(reply, F_LRMD_RSC_ID, rsc->rsc_id); crm_xml_add(reply, F_LRMD_CLASS, rsc->class); crm_xml_add(reply, F_LRMD_PROVIDER, rsc->provider); crm_xml_add(reply, F_LRMD_TYPE, rsc->type); } return reply; } static int process_lrmd_rsc_unregister(pcmk__client_t *client, uint32_t id, xmlNode *request) { int rc = pcmk_ok; lrmd_rsc_t *rsc = NULL; xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, request, LOG_ERR); const char *rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); if (!rsc_id) { return -ENODEV; } rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Ignoring unregistration of resource '%s', which is not registered", rsc_id); return pcmk_ok; } if (rsc->active) { /* let the caller know there are still active ops on this rsc to watch for */ crm_trace("Operation (0x%p) still in progress for unregistered resource %s", rsc->active, rsc_id); rc = -EINPROGRESS; } g_hash_table_remove(rsc_list, rsc_id); return rc; } static int process_lrmd_rsc_exec(pcmk__client_t *client, uint32_t id, xmlNode *request) { lrmd_rsc_t *rsc = NULL; lrmd_cmd_t *cmd = NULL; xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, request, LOG_ERR); const char *rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); int call_id; if (!rsc_id) { return -EINVAL; } if (!(rsc = g_hash_table_lookup(rsc_list, rsc_id))) { crm_info("Resource '%s' not found (%d active resources)", rsc_id, g_hash_table_size(rsc_list)); return -ENODEV; } cmd = create_lrmd_cmd(request, client); call_id = cmd->call_id; /* Don't reference cmd after handing it off to be scheduled. * The cmd could get merged and freed. */ schedule_lrmd_cmd(rsc, cmd); return call_id; } static int cancel_op(const char *rsc_id, const char *action, guint interval_ms) { GList *gIter = NULL; lrmd_rsc_t *rsc = g_hash_table_lookup(rsc_list, rsc_id); /* How to cancel an action. * 1. Check pending ops list, if it hasn't been handed off * to the service library or stonith recurring list remove * it there and that will stop it. * 2. If it isn't in the pending ops list, then it's either a * recurring op in the stonith recurring list, or the service * library's recurring list. Stop it there * 3. If not found in any lists, then this operation has either * been executed already and is not a recurring operation, or * never existed. */ if (!rsc) { return -ENODEV; } for (gIter = rsc->pending_ops; gIter != NULL; gIter = gIter->next) { lrmd_cmd_t *cmd = gIter->data; if (action_matches(cmd, action, interval_ms)) { - cmd->lrmd_op_status = PCMK_EXEC_CANCELLED; + cmd->result.execution_status = PCMK_EXEC_CANCELLED; cmd_finalize(cmd, rsc); return pcmk_ok; } } if (pcmk__str_eq(rsc->class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { /* The service library does not handle stonith operations. * We have to handle recurring stonith operations ourselves. */ for (gIter = rsc->recurring_ops; gIter != NULL; gIter = gIter->next) { lrmd_cmd_t *cmd = gIter->data; if (action_matches(cmd, action, interval_ms)) { - cmd->lrmd_op_status = PCMK_EXEC_CANCELLED; + cmd->result.execution_status = PCMK_EXEC_CANCELLED; if (rsc->active != cmd) { cmd_finalize(cmd, rsc); } return pcmk_ok; } } } else if (services_action_cancel(rsc_id, normalize_action_name(rsc, action), interval_ms) == TRUE) { /* The service library will tell the action_complete callback function * this action was cancelled, which will destroy the cmd and remove * it from the recurring_op list. Do not do that in this function * if the service library says it cancelled it. */ return pcmk_ok; } return -EOPNOTSUPP; } static void cancel_all_recurring(lrmd_rsc_t * rsc, const char *client_id) { GList *cmd_list = NULL; GList *cmd_iter = NULL; /* Notice a copy of each list is created when concat is called. * This prevents odd behavior from occurring when the cmd_list * is iterated through later on. It is possible the cancel_op * function may end up modifying the recurring_ops and pending_ops * lists. If we did not copy those lists, our cmd_list iteration * could get messed up.*/ if (rsc->recurring_ops) { cmd_list = g_list_concat(cmd_list, g_list_copy(rsc->recurring_ops)); } if (rsc->pending_ops) { cmd_list = g_list_concat(cmd_list, g_list_copy(rsc->pending_ops)); } if (!cmd_list) { return; } for (cmd_iter = cmd_list; cmd_iter; cmd_iter = cmd_iter->next) { lrmd_cmd_t *cmd = cmd_iter->data; if (cmd->interval_ms == 0) { continue; } if (client_id && !pcmk__str_eq(cmd->client_id, client_id, pcmk__str_casei)) { continue; } cancel_op(rsc->rsc_id, cmd->action, cmd->interval_ms); } /* frees only the copied list data, not the cmds */ g_list_free(cmd_list); } static int process_lrmd_rsc_cancel(pcmk__client_t *client, uint32_t id, xmlNode *request) { xmlNode *rsc_xml = get_xpath_object("//" F_LRMD_RSC, request, LOG_ERR); const char *rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); const char *action = crm_element_value(rsc_xml, F_LRMD_RSC_ACTION); guint interval_ms = 0; crm_element_value_ms(rsc_xml, F_LRMD_RSC_INTERVAL, &interval_ms); if (!rsc_id || !action) { return -EINVAL; } return cancel_op(rsc_id, action, interval_ms); } static void add_recurring_op_xml(xmlNode *reply, lrmd_rsc_t *rsc) { xmlNode *rsc_xml = create_xml_node(reply, F_LRMD_RSC); crm_xml_add(rsc_xml, F_LRMD_RSC_ID, rsc->rsc_id); for (GList *item = rsc->recurring_ops; item != NULL; item = item->next) { lrmd_cmd_t *cmd = item->data; xmlNode *op_xml = create_xml_node(rsc_xml, T_LRMD_RSC_OP); crm_xml_add(op_xml, F_LRMD_RSC_ACTION, (cmd->real_action? cmd->real_action : cmd->action)); crm_xml_add_ms(op_xml, F_LRMD_RSC_INTERVAL, cmd->interval_ms); crm_xml_add_int(op_xml, F_LRMD_TIMEOUT, cmd->timeout_orig); } } static xmlNode * process_lrmd_get_recurring(xmlNode *request, int call_id) { int rc = pcmk_ok; const char *rsc_id = NULL; lrmd_rsc_t *rsc = NULL; xmlNode *reply = NULL; xmlNode *rsc_xml = NULL; // Resource ID is optional rsc_xml = first_named_child(request, F_LRMD_CALLDATA); if (rsc_xml) { rsc_xml = first_named_child(rsc_xml, F_LRMD_RSC); } if (rsc_xml) { rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); } // If resource ID is specified, resource must exist if (rsc_id != NULL) { rsc = g_hash_table_lookup(rsc_list, rsc_id); if (rsc == NULL) { crm_info("Resource '%s' not found (%d active resources)", rsc_id, g_hash_table_size(rsc_list)); rc = -ENODEV; } } reply = create_lrmd_reply(__func__, rc, call_id); // If resource ID is not specified, check all resources if (rsc_id == NULL) { GHashTableIter iter; char *key = NULL; g_hash_table_iter_init(&iter, rsc_list); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &rsc)) { add_recurring_op_xml(reply, rsc); } } else if (rsc) { add_recurring_op_xml(reply, rsc); } return reply; } void process_lrmd_message(pcmk__client_t *client, uint32_t id, xmlNode *request) { int rc = pcmk_ok; int call_id = 0; const char *op = crm_element_value(request, F_LRMD_OPERATION); int do_reply = 0; int do_notify = 0; xmlNode *reply = NULL; /* Certain IPC commands may be done only by privileged users (i.e. root or * hacluster), because they would otherwise provide a means of bypassing * ACLs. */ bool allowed = pcmk_is_set(client->flags, pcmk__client_privileged); crm_trace("Processing %s operation from %s", op, client->id); crm_element_value_int(request, F_LRMD_CALLID, &call_id); if (pcmk__str_eq(op, CRM_OP_IPC_FWD, pcmk__str_none)) { #ifdef PCMK__COMPILE_REMOTE if (allowed) { ipc_proxy_forward_client(client, request); } else { rc = -EACCES; } #else rc = -EPROTONOSUPPORT; #endif do_reply = 1; } else if (pcmk__str_eq(op, CRM_OP_REGISTER, pcmk__str_none)) { rc = process_lrmd_signon(client, request, call_id, &reply); do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_RSC_REG, pcmk__str_none)) { if (allowed) { rc = process_lrmd_rsc_register(client, id, request); do_notify = 1; } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_RSC_INFO, pcmk__str_none)) { if (allowed) { reply = process_lrmd_get_rsc_info(request, call_id); } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_RSC_UNREG, pcmk__str_none)) { if (allowed) { rc = process_lrmd_rsc_unregister(client, id, request); /* don't notify anyone about failed un-registers */ if (rc == pcmk_ok || rc == -EINPROGRESS) { do_notify = 1; } } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_RSC_EXEC, pcmk__str_none)) { if (allowed) { rc = process_lrmd_rsc_exec(client, id, request); } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_RSC_CANCEL, pcmk__str_none)) { if (allowed) { rc = process_lrmd_rsc_cancel(client, id, request); } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_POKE, pcmk__str_none)) { do_notify = 1; do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_CHECK, pcmk__str_none)) { if (allowed) { xmlNode *data = get_message_xml(request, F_LRMD_CALLDATA); CRM_LOG_ASSERT(data != NULL); pcmk__valid_sbd_timeout(crm_element_value(data, F_LRMD_WATCHDOG)); } else { rc = -EACCES; } } else if (pcmk__str_eq(op, LRMD_OP_ALERT_EXEC, pcmk__str_none)) { if (allowed) { rc = process_lrmd_alert_exec(client, id, request); } else { rc = -EACCES; } do_reply = 1; } else if (pcmk__str_eq(op, LRMD_OP_GET_RECURRING, pcmk__str_none)) { if (allowed) { reply = process_lrmd_get_recurring(request, call_id); } else { rc = -EACCES; } do_reply = 1; } else { rc = -EOPNOTSUPP; do_reply = 1; crm_err("Unknown IPC request '%s' from client %s", op, pcmk__client_name(client)); } if (rc == -EACCES) { crm_warn("Rejecting IPC request '%s' from unprivileged client %s", op, pcmk__client_name(client)); } crm_debug("Processed %s operation from %s: rc=%d, reply=%d, notify=%d", op, client->id, rc, do_reply, do_notify); if (do_reply) { int send_rc = pcmk_rc_ok; if (reply == NULL) { reply = create_lrmd_reply(__func__, rc, call_id); } send_rc = lrmd_server_send_reply(client, id, reply); free_xml(reply); if (send_rc != pcmk_rc_ok) { crm_warn("Reply to client %s failed: %s " CRM_XS " rc=%d", pcmk__client_name(client), pcmk_rc_str(send_rc), send_rc); } } if (do_notify) { send_generic_notify(rc, request); } } diff --git a/daemons/fenced/fenced_remote.c b/daemons/fenced/fenced_remote.c index f77c7eb20c..ffaf60018a 100644 --- a/daemons/fenced/fenced_remote.c +++ b/daemons/fenced/fenced_remote.c @@ -1,2212 +1,2211 @@ /* * Copyright 2009-2021 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 #include #include #define TIMEOUT_MULTIPLY_FACTOR 1.2 /* When one fencer queries its peers for devices able to handle a fencing * request, each peer will reply with a list of such devices available to it. * Each reply will be parsed into a st_query_result_t, with each device's * information kept in a device_properties_t. */ typedef struct device_properties_s { /* Whether access to this device has been verified */ gboolean verified; /* The remaining members are indexed by the operation's "phase" */ /* Whether this device has been executed in each phase */ gboolean executed[st_phase_max]; /* Whether this device is disallowed from executing in each phase */ gboolean disallowed[st_phase_max]; /* Action-specific timeout for each phase */ int custom_action_timeout[st_phase_max]; /* Action-specific maximum random delay for each phase */ int delay_max[st_phase_max]; /* Action-specific base delay for each phase */ int delay_base[st_phase_max]; } device_properties_t; typedef struct st_query_result_s { /* Name of peer that sent this result */ char *host; /* Only try peers for non-topology based operations once */ gboolean tried; /* Number of entries in the devices table */ int ndevices; /* Devices available to this host that are capable of fencing the target */ GHashTable *devices; } st_query_result_t; GHashTable *stonith_remote_op_list = NULL; void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer, int rc); static void remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup); extern xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options); static void report_timeout_period(remote_fencing_op_t * op, int op_timeout); static int get_op_total_timeout(const remote_fencing_op_t *op, const st_query_result_t *chosen_peer); static gint sort_strings(gconstpointer a, gconstpointer b) { return strcmp(a, b); } static void free_remote_query(gpointer data) { if (data) { st_query_result_t *query = data; crm_trace("Free'ing query result from %s", query->host); g_hash_table_destroy(query->devices); free(query->host); free(query); } } void free_stonith_remote_op_list() { if (stonith_remote_op_list != NULL) { g_hash_table_destroy(stonith_remote_op_list); stonith_remote_op_list = NULL; } } struct peer_count_data { const remote_fencing_op_t *op; gboolean verified_only; int count; }; /*! * \internal * \brief Increment a counter if a device has not been executed yet * * \param[in] key Device ID (ignored) * \param[in] value Device properties * \param[in] user_data Peer count data */ static void count_peer_device(gpointer key, gpointer value, gpointer user_data) { device_properties_t *props = (device_properties_t*)value; struct peer_count_data *data = user_data; if (!props->executed[data->op->phase] && (!data->verified_only || props->verified)) { ++(data->count); } } /*! * \internal * \brief Check the number of available devices in a peer's query results * * \param[in] op Operation that results are for * \param[in] peer Peer to count * \param[in] verified_only Whether to count only verified devices * * \return Number of devices available to peer that were not already executed */ static int count_peer_devices(const remote_fencing_op_t *op, const st_query_result_t *peer, gboolean verified_only) { struct peer_count_data data; data.op = op; data.verified_only = verified_only; data.count = 0; if (peer) { g_hash_table_foreach(peer->devices, count_peer_device, &data); } return data.count; } /*! * \internal * \brief Search for a device in a query result * * \param[in] op Operation that result is for * \param[in] peer Query result for a peer * \param[in] device Device ID to search for * * \return Device properties if found, NULL otherwise */ static device_properties_t * find_peer_device(const remote_fencing_op_t *op, const st_query_result_t *peer, const char *device) { device_properties_t *props = g_hash_table_lookup(peer->devices, device); return (props && !props->executed[op->phase] && !props->disallowed[op->phase])? props : NULL; } /*! * \internal * \brief Find a device in a peer's device list and mark it as executed * * \param[in] op Operation that peer result is for * \param[in,out] peer Peer with results to search * \param[in] device ID of device to mark as done * \param[in] verified_devices_only Only consider verified devices * * \return TRUE if device was found and marked, FALSE otherwise */ static gboolean grab_peer_device(const remote_fencing_op_t *op, st_query_result_t *peer, const char *device, gboolean verified_devices_only) { device_properties_t *props = find_peer_device(op, peer, device); if ((props == NULL) || (verified_devices_only && !props->verified)) { return FALSE; } crm_trace("Removing %s from %s (%d remaining)", device, peer->host, count_peer_devices(op, peer, FALSE)); props->executed[op->phase] = TRUE; return TRUE; } static void clear_remote_op_timers(remote_fencing_op_t * op) { if (op->query_timer) { g_source_remove(op->query_timer); op->query_timer = 0; } if (op->op_timer_total) { g_source_remove(op->op_timer_total); op->op_timer_total = 0; } if (op->op_timer_one) { g_source_remove(op->op_timer_one); op->op_timer_one = 0; } } static void free_remote_op(gpointer data) { remote_fencing_op_t *op = data; crm_log_xml_debug(op->request, "Destroying"); clear_remote_op_timers(op); free(op->id); free(op->action); free(op->delegate); free(op->target); free(op->client_id); free(op->client_name); free(op->originator); if (op->query_results) { g_list_free_full(op->query_results, free_remote_query); } if (op->request) { free_xml(op->request); op->request = NULL; } if (op->devices_list) { g_list_free_full(op->devices_list, free); op->devices_list = NULL; } g_list_free_full(op->automatic_list, free); g_list_free(op->duplicates); free(op); } void init_stonith_remote_op_hash_table(GHashTable **table) { if (*table == NULL) { *table = pcmk__strkey_table(NULL, free_remote_op); } } /*! * \internal * \brief Return an operation's originally requested action (before any remap) * * \param[in] op Operation to check * * \return Operation's original action */ static const char * op_requested_action(const remote_fencing_op_t *op) { return ((op->phase > st_phase_requested)? "reboot" : op->action); } /*! * \internal * \brief Remap a "reboot" operation to the "off" phase * * \param[in,out] op Operation to remap */ static void op_phase_off(remote_fencing_op_t *op) { crm_info("Remapping multiple-device reboot targeting %s to 'off' " CRM_XS " id=%.8s", op->target, op->id); op->phase = st_phase_off; /* Happily, "off" and "on" are shorter than "reboot", so we can reuse the * memory allocation at each phase. */ strcpy(op->action, "off"); } /*! * \internal * \brief Advance a remapped reboot operation to the "on" phase * * \param[in,out] op Operation to remap */ static void op_phase_on(remote_fencing_op_t *op) { GList *iter = NULL; crm_info("Remapped 'off' targeting %s complete, " "remapping to 'on' for %s " CRM_XS " id=%.8s", op->target, op->client_name, op->id); op->phase = st_phase_on; strcpy(op->action, "on"); /* Skip devices with automatic unfencing, because the cluster will handle it * when the node rejoins. */ for (iter = op->automatic_list; iter != NULL; iter = iter->next) { GList *match = g_list_find_custom(op->devices_list, iter->data, sort_strings); if (match) { op->devices_list = g_list_remove(op->devices_list, match->data); } } g_list_free_full(op->automatic_list, free); op->automatic_list = NULL; /* Rewind device list pointer */ op->devices = op->devices_list; } /*! * \internal * \brief Reset a remapped reboot operation * * \param[in,out] op Operation to reset */ static void undo_op_remap(remote_fencing_op_t *op) { if (op->phase > 0) { crm_info("Undoing remap of reboot targeting %s for %s " CRM_XS " id=%.8s", op->target, op->client_name, op->id); op->phase = st_phase_requested; strcpy(op->action, "reboot"); } } static xmlNode * create_op_done_notify(remote_fencing_op_t * op, int rc) { xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); crm_xml_add_int(notify_data, "state", op->state); crm_xml_add_int(notify_data, F_STONITH_RC, rc); crm_xml_add(notify_data, F_STONITH_TARGET, op->target); crm_xml_add(notify_data, F_STONITH_ACTION, op->action); crm_xml_add(notify_data, F_STONITH_DELEGATE, op->delegate); crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(notify_data, F_STONITH_ORIGIN, op->originator); crm_xml_add(notify_data, F_STONITH_CLIENTID, op->client_id); crm_xml_add(notify_data, F_STONITH_CLIENTNAME, op->client_name); return notify_data; } void stonith_bcast_result_to_peers(remote_fencing_op_t * op, int rc, gboolean op_merged) { static int count = 0; xmlNode *bcast = create_xml_node(NULL, T_STONITH_REPLY); xmlNode *notify_data = create_op_done_notify(op, rc); count++; crm_trace("Broadcasting result to peers"); crm_xml_add(bcast, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(bcast, F_SUBTYPE, "broadcast"); crm_xml_add(bcast, F_STONITH_OPERATION, T_STONITH_NOTIFY); crm_xml_add_int(bcast, "count", count); if (op_merged) { crm_xml_add(bcast, F_STONITH_MERGED, "true"); } add_message_xml(bcast, F_STONITH_CALLDATA, notify_data); send_cluster_message(NULL, crm_msg_stonith_ng, bcast, FALSE); free_xml(notify_data); free_xml(bcast); return; } static void handle_local_reply_and_notify(remote_fencing_op_t * op, xmlNode * data, int rc) { xmlNode *notify_data = NULL; xmlNode *reply = NULL; if (op->notify_sent == TRUE) { /* nothing to do */ return; } /* Do notification with a clean data object */ notify_data = create_op_done_notify(op, rc); crm_xml_add_int(data, "state", op->state); crm_xml_add(data, F_STONITH_TARGET, op->target); crm_xml_add(data, F_STONITH_OPERATION, op->action); reply = stonith_construct_reply(op->request, NULL, data, rc); crm_xml_add(reply, F_STONITH_DELEGATE, op->delegate); /* Send fencing OP reply to local client that initiated fencing */ do_local_reply(reply, op->client_id, op->call_options & st_opt_sync_call, FALSE); /* bcast to all local clients that the fencing operation happend */ do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data); do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); /* mark this op as having notify's already sent */ op->notify_sent = TRUE; free_xml(reply); free_xml(notify_data); } static void handle_duplicates(remote_fencing_op_t * op, xmlNode * data, int rc) { GList *iter = NULL; for (iter = op->duplicates; iter != NULL; iter = iter->next) { remote_fencing_op_t *other = iter->data; if (other->state == st_duplicate) { other->state = op->state; crm_debug("Performing duplicate notification for %s@%s: %s " CRM_XS " id=%.8s", other->client_name, other->originator, pcmk_strerror(rc), other->id); remote_op_done(other, data, rc, TRUE); } else { // Possible if (for example) it timed out already crm_err("Skipping duplicate notification for %s@%s " CRM_XS " state=%s id=%.8s", other->client_name, other->originator, stonith_op_state_str(other->state), other->id); } } } static char * delegate_from_xml(xmlNode *xml) { xmlNode *match = get_xpath_object("//@" F_STONITH_DELEGATE, xml, LOG_NEVER); if (match == NULL) { return crm_element_value_copy(xml, F_ORIG); } else { return crm_element_value_copy(match, F_STONITH_DELEGATE); } } /*! * \internal * \brief Finalize a remote operation. * * \description This function has two code paths. * * Path 1. This node is the owner of the operation and needs * to notify the cpg group via a broadcast as to the operation's * results. * * Path 2. The cpg broadcast is received. All nodes notify their local * stonith clients the operation results. * * So, The owner of the operation first notifies the cluster of the result, * and once that cpg notify is received back it notifies all the local clients. * * Nodes that are passive watchers of the operation will receive the * broadcast and only need to notify their local clients the operation finished. * * \param op, The fencing operation to finalize * \param data, The xml msg reply (if present) of the last delegated fencing * operation. * \param dup, Is this operation a duplicate, if so treat it a little differently * making sure the broadcast is not sent out. */ static void remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup) { int level = LOG_ERR; const char *subt = NULL; xmlNode *local_data = NULL; gboolean op_merged = FALSE; set_fencing_completed(op); clear_remote_op_timers(op); undo_op_remap(op); if (op->notify_sent == TRUE) { crm_err("Already sent notifications for '%s' targeting %s by %s for " "client %s@%s: %s " CRM_XS " rc=%d state=%s id=%.8s", op->action, op->target, (op->delegate? op->delegate : "unknown node"), op->client_name, op->originator, pcmk_strerror(rc), rc, stonith_op_state_str(op->state), op->id); goto remote_op_done_cleanup; } - if (op->delegate == NULL) { + if (data == NULL) { + data = create_xml_node(NULL, "remote-op"); + local_data = data; + + } else if (op->delegate == NULL) { switch (rc) { case -ENODEV: case -EHOSTUNREACH: break; default: op->delegate = delegate_from_xml(data); break; } } - if (data == NULL) { - data = create_xml_node(NULL, "remote-op"); - local_data = data; - } - if(dup) { op_merged = TRUE; } else if (crm_element_value(data, F_STONITH_MERGED)) { op_merged = TRUE; } /* Tell everyone the operation is done, we will continue * with doing the local notifications once we receive * the broadcast back. */ subt = crm_element_value(data, F_SUBTYPE); if (dup == FALSE && !pcmk__str_eq(subt, "broadcast", pcmk__str_casei)) { /* Defer notification until the bcast message arrives */ stonith_bcast_result_to_peers(op, rc, (op_merged? TRUE: FALSE)); goto remote_op_done_cleanup; } if (rc == pcmk_ok || dup) { level = LOG_NOTICE; } else if (!pcmk__str_eq(op->originator, stonith_our_uname, pcmk__str_casei)) { level = LOG_NOTICE; } do_crm_log(level, "Operation '%s'%s%s by %s for %s@%s%s: %s " CRM_XS " id=%.8s", op->action, (op->target? " targeting " : ""), (op->target? op->target : ""), (op->delegate? op->delegate : "unknown node"), op->client_name, op->originator, (op_merged? " (merged)" : ""), pcmk_strerror(rc), op->id); handle_local_reply_and_notify(op, data, rc); if (dup == FALSE) { handle_duplicates(op, data, rc); } /* Free non-essential parts of the record * Keep the record around so we can query the history */ if (op->query_results) { g_list_free_full(op->query_results, free_remote_query); op->query_results = NULL; } if (op->request) { free_xml(op->request); op->request = NULL; } remote_op_done_cleanup: free_xml(local_data); } static gboolean remote_op_watchdog_done(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_one = 0; crm_notice("Self-fencing (%s) by %s for %s assumed complete " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); op->state = st_done; remote_op_done(op, NULL, pcmk_ok, FALSE); return FALSE; } static gboolean remote_op_timeout_one(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_one = 0; crm_notice("Peer's '%s' action targeting %s for client %s timed out " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); call_remote_stonith(op, NULL, pcmk_ok); return FALSE; } static gboolean remote_op_timeout(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_total = 0; if (op->state == st_done) { crm_debug("Action '%s' targeting %s for client %s already completed " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); return FALSE; } crm_debug("Action '%s' targeting %s for client %s timed out " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); if (op->phase == st_phase_on) { /* A remapped reboot operation timed out in the "on" phase, but the * "off" phase completed successfully, so quit trying any further * devices, and return success. */ remote_op_done(op, NULL, pcmk_ok, FALSE); return FALSE; } op->state = st_failed; remote_op_done(op, NULL, -ETIME, FALSE); return FALSE; } static gboolean remote_op_query_timeout(gpointer data) { remote_fencing_op_t *op = data; op->query_timer = 0; if (op->state == st_done) { crm_debug("Operation %.8s targeting %s already completed", op->id, op->target); } else if (op->state == st_exec) { crm_debug("Operation %.8s targeting %s already in progress", op->id, op->target); } else if (op->query_results) { crm_debug("Query %.8s targeting %s complete (state=%s)", op->id, op->target, stonith_op_state_str(op->state)); call_remote_stonith(op, NULL, pcmk_ok); } else { crm_debug("Query %.8s targeting %s timed out (state=%s)", op->id, op->target, stonith_op_state_str(op->state)); if (op->op_timer_total) { g_source_remove(op->op_timer_total); op->op_timer_total = 0; } remote_op_timeout(op); } return FALSE; } static gboolean topology_is_empty(stonith_topology_t *tp) { int i; if (tp == NULL) { return TRUE; } for (i = 0; i < ST_LEVEL_MAX; i++) { if (tp->levels[i] != NULL) { return FALSE; } } return TRUE; } /*! * \internal * \brief Add a device to an operation's automatic unfencing list * * \param[in,out] op Operation to modify * \param[in] device Device ID to add */ static void add_required_device(remote_fencing_op_t *op, const char *device) { GList *match = g_list_find_custom(op->automatic_list, device, sort_strings); if (!match) { op->automatic_list = g_list_prepend(op->automatic_list, strdup(device)); } } /*! * \internal * \brief Remove a device from the automatic unfencing list * * \param[in,out] op Operation to modify * \param[in] device Device ID to remove */ static void remove_required_device(remote_fencing_op_t *op, const char *device) { GList *match = g_list_find_custom(op->automatic_list, device, sort_strings); if (match) { op->automatic_list = g_list_remove(op->automatic_list, match->data); } } /* deep copy the device list */ static void set_op_device_list(remote_fencing_op_t * op, GList *devices) { GList *lpc = NULL; if (op->devices_list) { g_list_free_full(op->devices_list, free); op->devices_list = NULL; } for (lpc = devices; lpc != NULL; lpc = lpc->next) { op->devices_list = g_list_append(op->devices_list, strdup(lpc->data)); } op->devices = op->devices_list; } /*! * \internal * \brief Check whether a node matches a topology target * * \param[in] tp Topology table entry to check * \param[in] node Name of node to check * * \return TRUE if node matches topology target */ static gboolean topology_matches(const stonith_topology_t *tp, const char *node) { regex_t r_patt; CRM_CHECK(node && tp && tp->target, return FALSE); switch(tp->kind) { case 2: /* This level targets by attribute, so tp->target is a NAME=VALUE pair * of a permanent attribute applied to targeted nodes. The test below * relies on the locally cached copy of the CIB, so if fencing needs to * be done before the initial CIB is received or after a malformed CIB * is received, then the topology will be unable to be used. */ if (node_has_attr(node, tp->target_attribute, tp->target_value)) { crm_notice("Matched %s with %s by attribute", node, tp->target); return TRUE; } break; case 1: /* This level targets by name, so tp->target is a regular expression * matching names of nodes to be targeted. */ if (regcomp(&r_patt, tp->target_pattern, REG_EXTENDED|REG_NOSUB)) { crm_info("Bad regex '%s' for fencing level", tp->target); } else { int status = regexec(&r_patt, node, 0, NULL, 0); regfree(&r_patt); if (status == 0) { crm_notice("Matched %s with %s by name", node, tp->target); return TRUE; } } break; case 0: crm_trace("Testing %s against %s", node, tp->target); return pcmk__str_eq(tp->target, node, pcmk__str_casei); } crm_trace("No match for %s with %s", node, tp->target); return FALSE; } stonith_topology_t * find_topology_for_host(const char *host) { GHashTableIter tIter; stonith_topology_t *tp = g_hash_table_lookup(topology, host); if(tp != NULL) { crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology)); return tp; } g_hash_table_iter_init(&tIter, topology); while (g_hash_table_iter_next(&tIter, NULL, (gpointer *) & tp)) { if (topology_matches(tp, host)) { crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology)); return tp; } } crm_trace("No matches for %s in %d topology entries", host, g_hash_table_size(topology)); return NULL; } /*! * \internal * \brief Set fencing operation's device list to target's next topology level * * \param[in,out] op Remote fencing operation to modify * \param[in] empty_ok If true, an operation without a target (i.e. * queries) or a target without a topology will get a * pcmk_rc_ok return value instead of ENODEV * * \return Standard Pacemaker return value */ static int advance_topology_level(remote_fencing_op_t *op, bool empty_ok) { stonith_topology_t *tp = NULL; if (op->target) { tp = find_topology_for_host(op->target); } if (topology_is_empty(tp)) { return empty_ok? pcmk_rc_ok : ENODEV; } CRM_ASSERT(tp->levels != NULL); stonith__set_call_options(op->call_options, op->id, st_opt_topology); /* This is a new level, so undo any remapping left over from previous */ undo_op_remap(op); do { op->level++; } while (op->level < ST_LEVEL_MAX && tp->levels[op->level] == NULL); if (op->level < ST_LEVEL_MAX) { crm_trace("Attempting fencing level %d targeting %s (%d devices) " "for client %s@%s (id=%.8s)", op->level, op->target, g_list_length(tp->levels[op->level]), op->client_name, op->originator, op->id); set_op_device_list(op, tp->levels[op->level]); // The requested delay has been applied for the first fencing level if (op->level > 1 && op->delay > 0) { op->delay = 0; } if (g_list_next(op->devices_list) && pcmk__str_eq(op->action, "reboot", pcmk__str_casei)) { /* A reboot has been requested for a topology level with multiple * devices. Instead of rebooting the devices sequentially, we will * turn them all off, then turn them all on again. (Think about * switched power outlets for redundant power supplies.) */ op_phase_off(op); } return pcmk_rc_ok; } crm_notice("All fencing options targeting %s for client %s@%s failed " CRM_XS " id=%.8s", op->target, op->client_name, op->originator, op->id); return ENODEV; } /*! * \brief Check to see if this operation is a duplicate of another in flight * operation. If so merge this operation into the inflight operation, and mark * it as a duplicate. */ static void merge_duplicates(remote_fencing_op_t * op) { GHashTableIter iter; remote_fencing_op_t *other = NULL; time_t now = time(NULL); g_hash_table_iter_init(&iter, stonith_remote_op_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&other)) { const char *other_action = op_requested_action(other); if (!strcmp(op->id, other->id)) { continue; // Don't compare against self } if (other->state > st_exec) { crm_trace("%.8s not duplicate of %.8s: not in progress", op->id, other->id); continue; } if (!pcmk__str_eq(op->target, other->target, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: node %s vs. %s", op->id, other->id, op->target, other->target); continue; } if (!pcmk__str_eq(op->action, other_action, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: action %s vs. %s", op->id, other->id, op->action, other_action); continue; } if (pcmk__str_eq(op->client_name, other->client_name, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: same client %s", op->id, other->id, op->client_name); continue; } if (pcmk__str_eq(other->target, other->originator, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: suicide for %s", op->id, other->id, other->target); continue; } if (!fencing_peer_active(crm_get_peer(0, other->originator))) { crm_notice("Failing action '%s' targeting %s originating from " "client %s@%s: Originator is dead " CRM_XS " id=%.8s", other->action, other->target, other->client_name, other->originator, other->id); crm_trace("%.8s not duplicate of %.8s: originator dead", op->id, other->id); other->state = st_failed; continue; } if ((other->total_timeout > 0) && (now > (other->total_timeout + other->created))) { crm_trace("%.8s not duplicate of %.8s: old (%ld vs. %ld + %d)", op->id, other->id, now, other->created, other->total_timeout); continue; } /* There is another in-flight request to fence the same host * Piggyback on that instead. If it fails, so do we. */ other->duplicates = g_list_append(other->duplicates, op); if (other->total_timeout == 0) { other->total_timeout = op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * get_op_total_timeout(op, NULL); crm_trace("Best guess as to timeout used for %.8s: %d", other->id, other->total_timeout); } crm_notice("Merging fencing action '%s' targeting %s originating from " "client %s with identical request from %s@%s " CRM_XS " original=%.8s duplicate=%.8s total_timeout=%ds", op->action, op->target, op->client_name, other->client_name, other->originator, op->id, other->id, other->total_timeout); report_timeout_period(op, other->total_timeout); op->state = st_duplicate; } } static uint32_t fencing_active_peers(void) { uint32_t count = 0; crm_node_t *entry; GHashTableIter gIter; g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { if(fencing_peer_active(entry)) { count++; } } return count; } int stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR); op->state = st_done; set_fencing_completed(op); op->delegate = strdup("a human"); crm_notice("Injecting manual confirmation that %s is safely off/down", crm_element_value(dev, F_STONITH_TARGET)); remote_op_done(op, msg, pcmk_ok, FALSE); /* Replies are sent via done_cb->stonith_send_async_reply()->do_local_reply() */ return -EINPROGRESS; } /*! * \internal * \brief Create a new remote stonith operation * * \param[in] client ID of local stonith client that initiated the operation * \param[in] request The request from the client that started the operation * \param[in] peer TRUE if this operation is owned by another stonith peer * (an operation owned by one peer is stored on all peers, * but only the owner executes it; all nodes get the results * once the owner finishes execution) */ void * create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer) { remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_NEVER); int call_options = 0; const char *operation = NULL; init_stonith_remote_op_hash_table(&stonith_remote_op_list); /* If this operation is owned by another node, check to make * sure we haven't already created this operation. */ if (peer && dev) { const char *op_id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(op_id != NULL, return NULL); op = g_hash_table_lookup(stonith_remote_op_list, op_id); if (op) { crm_debug("Reusing existing remote fencing op %.8s for %s", op_id, ((client == NULL)? "unknown client" : client)); return op; } } op = calloc(1, sizeof(remote_fencing_op_t)); CRM_ASSERT(op != NULL); crm_element_value_int(request, F_STONITH_TIMEOUT, &(op->base_timeout)); // Value -1 means disable any static/random fencing delays crm_element_value_int(request, F_STONITH_DELAY, &(op->delay)); if (peer && dev) { op->id = crm_element_value_copy(dev, F_STONITH_REMOTE_OP_ID); } else { op->id = crm_generate_uuid(); } g_hash_table_replace(stonith_remote_op_list, op->id, op); op->state = st_query; op->replies_expected = fencing_active_peers(); op->action = crm_element_value_copy(dev, F_STONITH_ACTION); op->originator = crm_element_value_copy(dev, F_STONITH_ORIGIN); op->delegate = crm_element_value_copy(dev, F_STONITH_DELEGATE); /* May not be set */ op->created = time(NULL); if (op->originator == NULL) { /* Local or relayed request */ op->originator = strdup(stonith_our_uname); } CRM_LOG_ASSERT(client != NULL); if (client) { op->client_id = strdup(client); } /* For a RELAY operation, set fenced on the client. */ operation = crm_element_value(request, F_STONITH_OPERATION); if (pcmk__str_eq(operation, STONITH_OP_RELAY, pcmk__str_none)) { op->client_name = crm_strdup_printf("%s.%lu", crm_system_name, (unsigned long) getpid()); } else { op->client_name = crm_element_value_copy(request, F_STONITH_CLIENTNAME); } op->target = crm_element_value_copy(dev, F_STONITH_TARGET); op->request = copy_xml(request); /* TODO: Figure out how to avoid this */ crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); op->call_options = call_options; crm_element_value_int(request, F_STONITH_CALLID, &(op->client_callid)); crm_trace("%s new fencing op %s ('%s' targeting %s for client %s, " "base timeout %d, %u %s expected)", (peer && dev)? "Recorded" : "Generated", op->id, op->action, op->target, op->client_name, op->base_timeout, op->replies_expected, pcmk__plural_alt(op->replies_expected, "reply", "replies")); if (op->call_options & st_opt_cs_nodeid) { int nodeid; crm_node_t *node; pcmk__scan_min_int(op->target, &nodeid, 0); node = pcmk__search_known_node_cache(nodeid, NULL, CRM_GET_PEER_ANY); /* Ensure the conversion only happens once */ stonith__clear_call_options(op->call_options, op->id, st_opt_cs_nodeid); if (node && node->uname) { free(op->target); op->target = strdup(node->uname); } else { crm_warn("Could not expand nodeid '%s' into a host name", op->target); } } /* check to see if this is a duplicate operation of another in-flight operation */ merge_duplicates(op); if (op->state != st_duplicate) { /* kick history readers */ do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); } /* safe to trim as long as that doesn't touch pending ops */ stonith_fence_history_trim(); return op; } remote_fencing_op_t * initiate_remote_stonith_op(pcmk__client_t *client, xmlNode *request, gboolean manual_ack) { int query_timeout = 0; xmlNode *query = NULL; const char *client_id = NULL; remote_fencing_op_t *op = NULL; const char *relay_op_id = NULL; const char *operation = NULL; if (client) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } CRM_LOG_ASSERT(client_id != NULL); op = create_remote_stonith_op(client_id, request, FALSE); op->owner = TRUE; if (manual_ack) { crm_notice("Processing manual confirmation of fencing targeting %s " CRM_XS " id=%.8s", op->target, op->id); return op; } CRM_CHECK(op->action, return NULL); if (advance_topology_level(op, true) != pcmk_rc_ok) { op->state = st_failed; } switch (op->state) { case st_failed: crm_warn("Could not request peer fencing (%s) targeting %s " CRM_XS " id=%.8s", op->action, op->target, op->id); remote_op_done(op, NULL, -EINVAL, FALSE); return op; case st_duplicate: crm_info("Requesting peer fencing (%s) targeting %s (duplicate) " CRM_XS " id=%.8s", op->action, op->target, op->id); return op; default: crm_notice("Requesting peer fencing (%s) targeting %s " CRM_XS " id=%.8s state=%s base_timeout=%d", op->action, op->target, op->id, stonith_op_state_str(op->state), op->base_timeout); } query = stonith_create_op(op->client_callid, op->id, STONITH_OP_QUERY, NULL, op->call_options); crm_xml_add(query, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(query, F_STONITH_TARGET, op->target); crm_xml_add(query, F_STONITH_ACTION, op_requested_action(op)); crm_xml_add(query, F_STONITH_ORIGIN, op->originator); crm_xml_add(query, F_STONITH_CLIENTID, op->client_id); crm_xml_add(query, F_STONITH_CLIENTNAME, op->client_name); crm_xml_add_int(query, F_STONITH_TIMEOUT, op->base_timeout); /* In case of RELAY operation, RELAY information is added to the query to delete the original operation of RELAY. */ operation = crm_element_value(request, F_STONITH_OPERATION); if (pcmk__str_eq(operation, STONITH_OP_RELAY, pcmk__str_none)) { relay_op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID); if (relay_op_id) { crm_xml_add(query, F_STONITH_REMOTE_OP_ID_RELAY, relay_op_id); } } send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); query_timeout = op->base_timeout * TIMEOUT_MULTIPLY_FACTOR; op->query_timer = g_timeout_add((1000 * query_timeout), remote_op_query_timeout, op); return op; } enum find_best_peer_options { /*! Skip checking the target peer for capable fencing devices */ FIND_PEER_SKIP_TARGET = 0x0001, /*! Only check the target peer for capable fencing devices */ FIND_PEER_TARGET_ONLY = 0x0002, /*! Skip peers and devices that are not verified */ FIND_PEER_VERIFIED_ONLY = 0x0004, }; static st_query_result_t * find_best_peer(const char *device, remote_fencing_op_t * op, enum find_best_peer_options options) { GList *iter = NULL; gboolean verified_devices_only = (options & FIND_PEER_VERIFIED_ONLY) ? TRUE : FALSE; if (!device && pcmk_is_set(op->call_options, st_opt_topology)) { return NULL; } for (iter = op->query_results; iter != NULL; iter = iter->next) { st_query_result_t *peer = iter->data; crm_trace("Testing result from %s targeting %s with %d device%s: %d %x", peer->host, op->target, peer->ndevices, pcmk__plural_s(peer->ndevices), peer->tried, options); if ((options & FIND_PEER_SKIP_TARGET) && pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } if ((options & FIND_PEER_TARGET_ONLY) && !pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } if (pcmk_is_set(op->call_options, st_opt_topology)) { if (grab_peer_device(op, peer, device, verified_devices_only)) { return peer; } } else if ((peer->tried == FALSE) && count_peer_devices(op, peer, verified_devices_only)) { /* No topology: Use the current best peer */ crm_trace("Simple fencing"); return peer; } } return NULL; } static st_query_result_t * stonith_choose_peer(remote_fencing_op_t * op) { const char *device = NULL; st_query_result_t *peer = NULL; uint32_t active = fencing_active_peers(); do { if (op->devices) { device = op->devices->data; crm_trace("Checking for someone to fence (%s) %s using %s", op->action, op->target, device); } else { crm_trace("Checking for someone to fence (%s) %s", op->action, op->target); } /* Best choice is a peer other than the target with verified access */ peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET|FIND_PEER_VERIFIED_ONLY); if (peer) { crm_trace("Found verified peer %s for %s", peer->host, device?device:""); return peer; } if(op->query_timer != 0 && op->replies < QB_MIN(op->replies_expected, active)) { crm_trace("Waiting before looking for unverified devices to fence %s", op->target); return NULL; } /* If no other peer has verified access, next best is unverified access */ peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET); if (peer) { crm_trace("Found best unverified peer %s", peer->host); return peer; } /* If no other peer can do it, last option is self-fencing * (which is never allowed for the "on" phase of a remapped reboot) */ if (op->phase != st_phase_on) { peer = find_best_peer(device, op, FIND_PEER_TARGET_ONLY); if (peer) { crm_trace("%s will fence itself", peer->host); return peer; } } /* Try the next fencing level if there is one (unless we're in the "on" * phase of a remapped "reboot", because we ignore errors in that case) */ } while ((op->phase != st_phase_on) && pcmk_is_set(op->call_options, st_opt_topology) && (advance_topology_level(op, false) == pcmk_rc_ok)); crm_notice("Couldn't find anyone to fence (%s) %s using %s", op->action, op->target, (device? device : "any device")); return NULL; } static int get_device_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer, const char *device) { device_properties_t *props; if (!peer || !device) { return op->base_timeout; } props = g_hash_table_lookup(peer->devices, device); if (!props) { return op->base_timeout; } return (props->custom_action_timeout[op->phase]? props->custom_action_timeout[op->phase] : op->base_timeout) + props->delay_max[op->phase]; } struct timeout_data { const remote_fencing_op_t *op; const st_query_result_t *peer; int total_timeout; }; /*! * \internal * \brief Add timeout to a total if device has not been executed yet * * \param[in] key GHashTable key (device ID) * \param[in] value GHashTable value (device properties) * \param[in] user_data Timeout data */ static void add_device_timeout(gpointer key, gpointer value, gpointer user_data) { const char *device_id = key; device_properties_t *props = value; struct timeout_data *timeout = user_data; if (!props->executed[timeout->op->phase] && !props->disallowed[timeout->op->phase]) { timeout->total_timeout += get_device_timeout(timeout->op, timeout->peer, device_id); } } static int get_peer_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer) { struct timeout_data timeout; timeout.op = op; timeout.peer = peer; timeout.total_timeout = 0; g_hash_table_foreach(peer->devices, add_device_timeout, &timeout); return (timeout.total_timeout? timeout.total_timeout : op->base_timeout); } static int get_op_total_timeout(const remote_fencing_op_t *op, const st_query_result_t *chosen_peer) { int total_timeout = 0; stonith_topology_t *tp = find_topology_for_host(op->target); if (pcmk_is_set(op->call_options, st_opt_topology) && tp) { int i; GList *device_list = NULL; GList *iter = NULL; /* Yep, this looks scary, nested loops all over the place. * Here is what is going on. * Loop1: Iterate through fencing levels. * Loop2: If a fencing level has devices, loop through each device * Loop3: For each device in a fencing level, see what peer owns it * and what that peer has reported the timeout is for the device. */ for (i = 0; i < ST_LEVEL_MAX; i++) { if (!tp->levels[i]) { continue; } for (device_list = tp->levels[i]; device_list; device_list = device_list->next) { for (iter = op->query_results; iter != NULL; iter = iter->next) { const st_query_result_t *peer = iter->data; if (find_peer_device(op, peer, device_list->data)) { total_timeout += get_device_timeout(op, peer, device_list->data); break; } } /* End Loop3: match device with peer that owns device, find device's timeout period */ } /* End Loop2: iterate through devices at a specific level */ } /*End Loop1: iterate through fencing levels */ } else if (chosen_peer) { total_timeout = get_peer_timeout(op, chosen_peer); } else { total_timeout = op->base_timeout; } return total_timeout ? total_timeout : op->base_timeout; } static void report_timeout_period(remote_fencing_op_t * op, int op_timeout) { GList *iter = NULL; xmlNode *update = NULL; const char *client_node = NULL; const char *client_id = NULL; const char *call_id = NULL; if (op->call_options & st_opt_sync_call) { /* There is no reason to report the timeout for a synchronous call. It * is impossible to use the reported timeout to do anything when the client * is blocking for the response. This update is only important for * async calls that require a callback to report the results in. */ return; } else if (!op->request) { return; } crm_trace("Reporting timeout for %s (id=%.8s)", op->client_name, op->id); client_node = crm_element_value(op->request, F_STONITH_CLIENTNODE); call_id = crm_element_value(op->request, F_STONITH_CALLID); client_id = crm_element_value(op->request, F_STONITH_CLIENTID); if (!client_node || !call_id || !client_id) { return; } if (pcmk__str_eq(client_node, stonith_our_uname, pcmk__str_casei)) { /* The client is connected to this node, send the update direclty to them */ do_stonith_async_timeout_update(client_id, call_id, op_timeout); return; } /* The client is connected to another node, relay this update to them */ update = stonith_create_op(op->client_callid, op->id, STONITH_OP_TIMEOUT_UPDATE, NULL, 0); crm_xml_add(update, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(update, F_STONITH_CLIENTID, client_id); crm_xml_add(update, F_STONITH_CALLID, call_id); crm_xml_add_int(update, F_STONITH_TIMEOUT, op_timeout); send_cluster_message(crm_get_peer(0, client_node), crm_msg_stonith_ng, update, FALSE); free_xml(update); for (iter = op->duplicates; iter != NULL; iter = iter->next) { remote_fencing_op_t *dup = iter->data; crm_trace("Reporting timeout for duplicate %.8s to client %s", dup->id, dup->client_name); report_timeout_period(iter->data, op_timeout); } } /*! * \internal * \brief Advance an operation to the next device in its topology * * \param[in,out] op Operation to advance * \param[in] device ID of device just completed * \param[in] msg XML reply that contained device result (if available) * \param[in] rc Return code of device's execution */ static void advance_topology_device_in_level(remote_fencing_op_t *op, const char *device, xmlNode *msg, int rc) { /* Advance to the next device at this topology level, if any */ if (op->devices) { op->devices = op->devices->next; } /* Handle automatic unfencing if an "on" action was requested */ if ((op->phase == st_phase_requested) && pcmk__str_eq(op->action, "on", pcmk__str_casei)) { /* If the device we just executed was required, it's not anymore */ remove_required_device(op, device); /* If there are no more devices at this topology level, run through any * remaining devices with automatic unfencing */ if (op->devices == NULL) { op->devices = op->automatic_list; } } if ((op->devices == NULL) && (op->phase == st_phase_off)) { /* We're done with this level and with required devices, but we had * remapped "reboot" to "off", so start over with "on". If any devices * need to be turned back on, op->devices will be non-NULL after this. */ op_phase_on(op); } if (op->devices) { /* Necessary devices remain, so execute the next one */ crm_trace("Next targeting %s on behalf of %s@%s (rc was %d)", op->target, op->client_name, op->originator, rc); // The requested delay has been applied for the first device if (op->delay > 0) { op->delay = 0; } call_remote_stonith(op, NULL, pcmk_ok); } else { /* We're done with all devices and phases, so finalize operation */ crm_trace("Marking complex fencing op targeting %s as complete", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); } } static gboolean check_watchdog_fencing_and_wait(remote_fencing_op_t * op) { if (node_does_watchdog_fencing(op->target)) { crm_notice("Waiting %lds for %s to self-fence (%s) for " "client %s " CRM_XS " id=%.8s", (stonith_watchdog_timeout_ms / 1000), op->target, op->action, op->client_name, op->id); op->op_timer_one = g_timeout_add(stonith_watchdog_timeout_ms, remote_op_watchdog_done, op); return TRUE; } else { crm_debug("Skipping fallback to watchdog-fencing as %s is " "not in host-list", op->target); } return FALSE; } void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer, int rc) { const char *device = NULL; int timeout = op->base_timeout; crm_trace("Action %.8s targeting %s for %s is %s", op->id, op->target, op->client_name, stonith_op_state_str(op->state)); if ((peer == NULL) && !pcmk_is_set(op->call_options, st_opt_topology)) { peer = stonith_choose_peer(op); } if (!op->op_timer_total) { int total_timeout = get_op_total_timeout(op, peer); op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * total_timeout; op->op_timer_total = g_timeout_add(1000 * op->total_timeout, remote_op_timeout, op); report_timeout_period(op, op->total_timeout); crm_info("Total timeout set to %d for peer's fencing targeting %s for %s" CRM_XS "id=%.8s", total_timeout, op->target, op->client_name, op->id); } if (pcmk_is_set(op->call_options, st_opt_topology) && op->devices) { /* Ignore any peer preference, they might not have the device we need */ /* When using topology, stonith_choose_peer() removes the device from * further consideration, so be sure to calculate timeout beforehand */ peer = stonith_choose_peer(op); device = op->devices->data; timeout = get_device_timeout(op, peer, device); } if (peer) { int timeout_one = 0; xmlNode *remote_op = stonith_create_op(op->client_callid, op->id, STONITH_OP_FENCE, NULL, 0); crm_xml_add(remote_op, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(remote_op, F_STONITH_TARGET, op->target); crm_xml_add(remote_op, F_STONITH_ACTION, op->action); crm_xml_add(remote_op, F_STONITH_ORIGIN, op->originator); crm_xml_add(remote_op, F_STONITH_CLIENTID, op->client_id); crm_xml_add(remote_op, F_STONITH_CLIENTNAME, op->client_name); crm_xml_add_int(remote_op, F_STONITH_TIMEOUT, timeout); crm_xml_add_int(remote_op, F_STONITH_CALLOPTS, op->call_options); crm_xml_add_int(remote_op, F_STONITH_DELAY, op->delay); if (device) { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_device_timeout(op, peer, device); crm_notice("Requesting that %s perform '%s' action targeting %s " "using %s " CRM_XS " for client %s (%ds)", peer->host, op->action, op->target, device, op->client_name, timeout_one); crm_xml_add(remote_op, F_STONITH_DEVICE, device); } else { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_peer_timeout(op, peer); crm_notice("Requesting that %s perform '%s' action targeting %s " CRM_XS " for client %s (%ds, %lds)", peer->host, op->action, op->target, op->client_name, timeout_one, stonith_watchdog_timeout_ms); } op->state = st_exec; if (op->op_timer_one) { g_source_remove(op->op_timer_one); } if (!(stonith_watchdog_timeout_ms > 0 && ( (pcmk__str_eq(device, STONITH_WATCHDOG_ID, pcmk__str_none)) || (pcmk__str_eq(peer->host, op->target, pcmk__str_casei) && !pcmk__str_eq(op->action, "on", pcmk__str_casei))) && check_watchdog_fencing_and_wait(op))) { /* Some thoughts about self-fencing cases reaching this point: - Actually check in check_watchdog_fencing_and_wait shouldn't fail if STONITH_WATCHDOG_ID is chosen as fencing-device and it being present implies watchdog-fencing is enabled anyway - If watchdog-fencing is disabled either in general or for a specific target - detected in check_watchdog_fencing_and_wait - for some other kind of self-fencing we can't expect a success answer but timeout is fine if the node doesn't come back in between - Delicate might be the case where we have watchdog-fencing enabled for a node but the watchdog-fencing-device isn't explicitly chosen for suicide. Local pe-execution in sbd may detect the node as unclean and lead to timely suicide. Otherwise the selection of stonith-watchdog-timeout at least is questionable. */ op->op_timer_one = g_timeout_add((1000 * timeout_one), remote_op_timeout_one, op); } send_cluster_message(crm_get_peer(0, peer->host), crm_msg_stonith_ng, remote_op, FALSE); peer->tried = TRUE; free_xml(remote_op); return; } else if (op->phase == st_phase_on) { /* A remapped "on" cannot be executed, but the node was already * turned off successfully, so ignore the error and continue. */ crm_warn("Ignoring %s 'on' failure (no capable peers) targeting %s " "after successful 'off'", device, op->target); advance_topology_device_in_level(op, device, NULL, pcmk_ok); return; } else if (op->owner == FALSE) { crm_err("Fencing (%s) targeting %s for client %s is not ours to control", op->action, op->target, op->client_name); } else if (op->query_timer == 0) { /* We've exhausted all available peers */ crm_info("No remaining peers capable of fencing (%s) %s for client %s " CRM_XS " state=%s", op->action, op->target, op->client_name, stonith_op_state_str(op->state)); CRM_LOG_ASSERT(op->state < st_done); remote_op_timeout(op); } else if(op->replies >= op->replies_expected || op->replies >= fencing_active_peers()) { // int rc = -EHOSTUNREACH; /* if the operation never left the query state, * but we have all the expected replies, then no devices * are available to execute the fencing operation. */ if(stonith_watchdog_timeout_ms > 0 && pcmk__str_eq(device, STONITH_WATCHDOG_ID, pcmk__str_null_matches)) { if (check_watchdog_fencing_and_wait(op)) { return; } } if (op->state == st_query) { crm_info("No peers (out of %d) have devices capable of fencing " "(%s) %s for client %s " CRM_XS " state=%s", op->replies, op->action, op->target, op->client_name, stonith_op_state_str(op->state)); rc = -ENODEV; } else { if (pcmk_is_set(op->call_options, st_opt_topology)) { rc = -EHOSTUNREACH; } crm_info("No peers (out of %d) are capable of fencing (%s) %s " "for client %s " CRM_XS " state=%s", op->replies, op->action, op->target, op->client_name, stonith_op_state_str(op->state)); } op->state = st_failed; remote_op_done(op, NULL, rc, FALSE); } else { crm_info("Waiting for additional peers capable of fencing (%s) %s%s%s " "for client %s " CRM_XS " id=%.8s", op->action, op->target, (device? " using " : ""), (device? device : ""), op->client_name, op->id); } } /*! * \internal * \brief Comparison function for sorting query results * * \param[in] a GList item to compare * \param[in] b GList item to compare * * \return Per the glib documentation, "a negative integer if the first value * comes before the second, 0 if they are equal, or a positive integer * if the first value comes after the second." */ static gint sort_peers(gconstpointer a, gconstpointer b) { const st_query_result_t *peer_a = a; const st_query_result_t *peer_b = b; return (peer_b->ndevices - peer_a->ndevices); } /*! * \internal * \brief Determine if all the devices in the topology are found or not */ static gboolean all_topology_devices_found(remote_fencing_op_t * op) { GList *device = NULL; GList *iter = NULL; device_properties_t *match = NULL; stonith_topology_t *tp = NULL; gboolean skip_target = FALSE; int i; tp = find_topology_for_host(op->target); if (!tp) { return FALSE; } if (pcmk__strcase_any_of(op->action, "off", "reboot", NULL)) { /* Don't count the devices on the target node if we are killing * the target node. */ skip_target = TRUE; } for (i = 0; i < ST_LEVEL_MAX; i++) { for (device = tp->levels[i]; device; device = device->next) { match = NULL; for (iter = op->query_results; iter && !match; iter = iter->next) { st_query_result_t *peer = iter->data; if (skip_target && pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } match = find_peer_device(op, peer, device->data); } if (!match) { return FALSE; } } } return TRUE; } /*! * \internal * \brief Parse action-specific device properties from XML * * \param[in] msg XML element containing the properties * \param[in] peer Name of peer that sent XML (for logs) * \param[in] device Device ID (for logs) * \param[in] action Action the properties relate to (for logs) * \param[in] phase Phase the properties relate to * \param[in,out] props Device properties to update */ static void parse_action_specific(xmlNode *xml, const char *peer, const char *device, const char *action, remote_fencing_op_t *op, enum st_remap_phase phase, device_properties_t *props) { props->custom_action_timeout[phase] = 0; crm_element_value_int(xml, F_STONITH_ACTION_TIMEOUT, &props->custom_action_timeout[phase]); if (props->custom_action_timeout[phase]) { crm_trace("Peer %s with device %s returned %s action timeout %d", peer, device, action, props->custom_action_timeout[phase]); } props->delay_max[phase] = 0; crm_element_value_int(xml, F_STONITH_DELAY_MAX, &props->delay_max[phase]); if (props->delay_max[phase]) { crm_trace("Peer %s with device %s returned maximum of random delay %d for %s", peer, device, props->delay_max[phase], action); } props->delay_base[phase] = 0; crm_element_value_int(xml, F_STONITH_DELAY_BASE, &props->delay_base[phase]); if (props->delay_base[phase]) { crm_trace("Peer %s with device %s returned base delay %d for %s", peer, device, props->delay_base[phase], action); } /* Handle devices with automatic unfencing */ if (pcmk__str_eq(action, "on", pcmk__str_casei)) { int required = 0; crm_element_value_int(xml, F_STONITH_DEVICE_REQUIRED, &required); if (required) { crm_trace("Peer %s requires device %s to execute for action %s", peer, device, action); add_required_device(op, device); } } /* If a reboot is remapped to off+on, it's possible that a node is allowed * to perform one action but not another. */ if (crm_is_true(crm_element_value(xml, F_STONITH_ACTION_DISALLOWED))) { props->disallowed[phase] = TRUE; crm_trace("Peer %s is disallowed from executing %s for device %s", peer, action, device); } } /*! * \internal * \brief Parse one device's properties from peer's XML query reply * * \param[in] xml XML node containing device properties * \param[in,out] op Operation that query and reply relate to * \param[in,out] result Peer's results * \param[in] device ID of device being parsed */ static void add_device_properties(xmlNode *xml, remote_fencing_op_t *op, st_query_result_t *result, const char *device) { xmlNode *child; int verified = 0; device_properties_t *props = calloc(1, sizeof(device_properties_t)); /* Add a new entry to this result's devices list */ CRM_ASSERT(props != NULL); g_hash_table_insert(result->devices, strdup(device), props); /* Peers with verified (monitored) access will be preferred */ crm_element_value_int(xml, F_STONITH_DEVICE_VERIFIED, &verified); if (verified) { crm_trace("Peer %s has confirmed a verified device %s", result->host, device); props->verified = TRUE; } /* Parse action-specific device properties */ parse_action_specific(xml, result->host, device, op_requested_action(op), op, st_phase_requested, props); for (child = pcmk__xml_first_child(xml); child != NULL; child = pcmk__xml_next(child)) { /* Replies for "reboot" operations will include the action-specific * values for "off" and "on" in child elements, just in case the reboot * winds up getting remapped. */ if (pcmk__str_eq(ID(child), "off", pcmk__str_casei)) { parse_action_specific(child, result->host, device, "off", op, st_phase_off, props); } else if (pcmk__str_eq(ID(child), "on", pcmk__str_casei)) { parse_action_specific(child, result->host, device, "on", op, st_phase_on, props); } } } /*! * \internal * \brief Parse a peer's XML query reply and add it to operation's results * * \param[in,out] op Operation that query and reply relate to * \param[in] host Name of peer that sent this reply * \param[in] ndevices Number of devices expected in reply * \param[in] xml XML node containing device list * * \return Newly allocated result structure with parsed reply */ static st_query_result_t * add_result(remote_fencing_op_t *op, const char *host, int ndevices, xmlNode *xml) { st_query_result_t *result = calloc(1, sizeof(st_query_result_t)); xmlNode *child; // cppcheck seems not to understand the abort logic in CRM_CHECK // cppcheck-suppress memleak CRM_CHECK(result != NULL, return NULL); result->host = strdup(host); result->devices = pcmk__strkey_table(free, free); /* Each child element describes one capable device available to the peer */ for (child = pcmk__xml_first_child(xml); child != NULL; child = pcmk__xml_next(child)) { const char *device = ID(child); if (device) { add_device_properties(child, op, result, device); } } result->ndevices = g_hash_table_size(result->devices); CRM_CHECK(ndevices == result->ndevices, crm_err("Query claimed to have %d device%s but %d found", ndevices, pcmk__plural_s(ndevices), result->ndevices)); op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers); return result; } /*! * \internal * \brief Handle a peer's reply to our fencing query * * Parse a query result from XML and store it in the remote operation * table, and when enough replies have been received, issue a fencing request. * * \param[in] msg XML reply received * * \return pcmk_ok on success, -errno on error * * \note See initiate_remote_stonith_op() for how the XML query was initially * formed, and stonith_query() for how the peer formed its XML reply. */ int process_remote_stonith_query(xmlNode * msg) { int ndevices = 0; gboolean host_is_target = FALSE; gboolean have_all_replies = FALSE; const char *id = NULL; const char *host = NULL; remote_fencing_op_t *op = NULL; st_query_result_t *result = NULL; uint32_t replies_expected; xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@" F_STONITH_AVAILABLE_DEVICES, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, F_STONITH_AVAILABLE_DEVICES, &ndevices); op = g_hash_table_lookup(stonith_remote_op_list, id); if (op == NULL) { crm_debug("Received query reply for unknown or expired operation %s", id); return -EOPNOTSUPP; } replies_expected = fencing_active_peers(); if (op->replies_expected < replies_expected) { replies_expected = op->replies_expected; } if ((++op->replies >= replies_expected) && (op->state == st_query)) { have_all_replies = TRUE; } host = crm_element_value(msg, F_ORIG); host_is_target = pcmk__str_eq(host, op->target, pcmk__str_casei); crm_info("Query result %d of %d from %s for %s/%s (%d device%s) %s", op->replies, replies_expected, host, op->target, op->action, ndevices, pcmk__plural_s(ndevices), id); if (ndevices > 0) { result = add_result(op, host, ndevices, dev); } if (pcmk_is_set(op->call_options, st_opt_topology)) { /* If we start the fencing before all the topology results are in, * it is possible fencing levels will be skipped because of the missing * query results. */ if (op->state == st_query && all_topology_devices_found(op)) { /* All the query results are in for the topology, start the fencing ops. */ crm_trace("All topology devices found"); call_remote_stonith(op, result, pcmk_ok); } else if (have_all_replies) { crm_info("All topology query replies have arrived, continuing (%d expected/%d received) ", replies_expected, op->replies); call_remote_stonith(op, NULL, pcmk_ok); } } else if (op->state == st_query) { int nverified = count_peer_devices(op, result, TRUE); /* We have a result for a non-topology fencing op that looks promising, * go ahead and start fencing before query timeout */ if (result && (host_is_target == FALSE) && nverified) { /* we have a verified device living on a peer that is not the target */ crm_trace("Found %d verified device%s", nverified, pcmk__plural_s(nverified)); call_remote_stonith(op, result, pcmk_ok); } else if (have_all_replies) { crm_info("All query replies have arrived, continuing (%d expected/%d received) ", replies_expected, op->replies); call_remote_stonith(op, NULL, pcmk_ok); } else { crm_trace("Waiting for more peer results before launching fencing operation"); } } else if (result && (op->state == st_done)) { crm_info("Discarding query result from %s (%d device%s): " "Operation is %s", result->host, result->ndevices, pcmk__plural_s(result->ndevices), stonith_op_state_str(op->state)); } return pcmk_ok; } /*! * \internal * \brief Handle a peer's reply to a fencing request * * Parse a fencing reply from XML, and either finalize the operation * or attempt another device as appropriate. * * \param[in] msg XML reply received * * \return pcmk_ok on success, -errno on error */ int process_remote_stonith_exec(xmlNode * msg) { int rc = 0; const char *id = NULL; const char *device = NULL; remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@" F_STONITH_RC, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, F_STONITH_RC, &rc); device = crm_element_value(dev, F_STONITH_DEVICE); if (stonith_remote_op_list) { op = g_hash_table_lookup(stonith_remote_op_list, id); } if (op == NULL && rc == pcmk_ok) { /* Record successful fencing operations */ const char *client_id = crm_element_value(dev, F_STONITH_CLIENTID); op = create_remote_stonith_op(client_id, dev, TRUE); } if (op == NULL) { /* Could be for an event that began before we started */ /* TODO: Record the op for later querying */ crm_info("Received peer result of unknown or expired operation %s", id); return -EOPNOTSUPP; } if (op->devices && device && !pcmk__str_eq(op->devices->data, device, pcmk__str_casei)) { crm_err("Received outdated reply for device %s (instead of %s) to " "fence (%s) %s. Operation already timed out at peer level.", device, (const char *) op->devices->data, op->action, op->target); return rc; } if (pcmk__str_eq(crm_element_value(msg, F_SUBTYPE), "broadcast", pcmk__str_casei)) { crm_debug("Finalizing action '%s' targeting %s on behalf of %s@%s: %s " CRM_XS " rc=%d id=%.8s", op->action, op->target, op->client_name, op->originator, pcmk_strerror(rc), rc, op->id); if (rc == pcmk_ok) { op->state = st_done; } else { op->state = st_failed; } remote_op_done(op, msg, rc, FALSE); return pcmk_ok; } else if (!pcmk__str_eq(op->originator, stonith_our_uname, pcmk__str_casei)) { /* If this isn't a remote level broadcast, and we are not the * originator of the operation, we should not be receiving this msg. */ crm_err("Received non-broadcast fencing result for operation %.8s " "we do not own (device %s targeting %s)", op->id, device, op->target); return rc; } if (pcmk_is_set(op->call_options, st_opt_topology)) { const char *device = crm_element_value(msg, F_STONITH_DEVICE); crm_notice("Action '%s' targeting %s using %s on behalf of %s@%s: %s " CRM_XS " rc=%d", op->action, op->target, device, op->client_name, op->originator, pcmk_strerror(rc), rc); /* We own the op, and it is complete. broadcast the result to all nodes * and notify our local clients. */ if (op->state == st_done) { remote_op_done(op, msg, rc, FALSE); return rc; } if ((op->phase == 2) && (rc != pcmk_ok)) { /* A remapped "on" failed, but the node was already turned off * successfully, so ignore the error and continue. */ crm_warn("Ignoring %s 'on' failure (exit code %d) targeting %s " "after successful 'off'", device, rc, op->target); rc = pcmk_ok; } if (rc == pcmk_ok) { /* An operation completed successfully. Try another device if * necessary, otherwise mark the operation as done. */ advance_topology_device_in_level(op, device, msg, rc); return rc; } else { /* This device failed, time to try another topology level. If no other * levels are available, mark this operation as failed and report results. */ if (advance_topology_level(op, false) != pcmk_rc_ok) { op->state = st_failed; remote_op_done(op, msg, rc, FALSE); return rc; } } } else if (rc == pcmk_ok && op->devices == NULL) { crm_trace("All done for %s", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); return rc; } else if (rc == -ETIME && op->devices == NULL) { /* If the operation timed out don't bother retrying other peers. */ op->state = st_failed; remote_op_done(op, msg, rc, FALSE); return rc; } else { /* fall-through and attempt other fencing action using another peer */ } /* Retry on failure */ crm_trace("Next for %s on behalf of %s@%s (rc was %d)", op->target, op->originator, op->client_name, rc); call_remote_stonith(op, NULL, rc); return rc; } gboolean stonith_check_fence_tolerance(int tolerance, const char *target, const char *action) { GHashTableIter iter; time_t now = time(NULL); remote_fencing_op_t *rop = NULL; if (tolerance <= 0 || !stonith_remote_op_list || target == NULL || action == NULL) { return FALSE; } g_hash_table_iter_init(&iter, stonith_remote_op_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&rop)) { if (strcmp(rop->target, target) != 0) { continue; } else if (rop->state != st_done) { continue; /* We don't have to worry about remapped reboots here * because if state is done, any remapping has been undone */ } else if (strcmp(rop->action, action) != 0) { continue; } else if ((rop->completed + tolerance) < now) { continue; } crm_notice("Target %s was fenced (%s) less than %ds ago by %s on behalf of %s", target, action, tolerance, rop->delegate, rop->originator); return TRUE; } return FALSE; } diff --git a/include/crm/common/results.h b/include/crm/common/results.h index 457afe387f..3b6cf4e05d 100644 --- a/include/crm/common/results.h +++ b/include/crm/common/results.h @@ -1,332 +1,333 @@ /* * Copyright 2012-2021 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 CRM_RESULTS__H # define CRM_RESULTS__H #ifdef __cplusplus extern "C" { #endif /*! * \file * \brief Function and executable result codes * \ingroup core */ // Lifted from config.h /* The _Noreturn keyword of C11. */ #ifndef _Noreturn # if (defined __cplusplus \ && ((201103 <= __cplusplus && !(__GNUC__ == 4 && __GNUC_MINOR__ == 7)) \ || (defined _MSC_VER && 1900 <= _MSC_VER))) # define _Noreturn [[noreturn]] # elif ((!defined __cplusplus || defined __clang__) \ && (201112 <= (defined __STDC_VERSION__ ? __STDC_VERSION__ : 0) \ || 4 < __GNUC__ + (7 <= __GNUC_MINOR__))) /* _Noreturn works as-is. */ # elif 2 < __GNUC__ + (8 <= __GNUC_MINOR__) || 0x5110 <= __SUNPRO_C # define _Noreturn __attribute__ ((__noreturn__)) # elif 1200 <= (defined _MSC_VER ? _MSC_VER : 0) # define _Noreturn __declspec (noreturn) # else # define _Noreturn # endif #endif # define CRM_ASSERT(expr) do { \ if (!(expr)) { \ crm_abort(__FILE__, __func__, __LINE__, #expr, TRUE, FALSE); \ abort(); /* crm_abort() doesn't always abort! */ \ } \ } while(0) /* * Function return codes * * Most Pacemaker API functions return an integer return code. There are two * alternative interpretations. The legacy interpration is that the absolute * value of the return code is either a system error number or a custom * pcmk_err_* number. This is less than ideal because system error numbers are * constrained only to the positive int range, so there's the possibility that * system errors and custom errors could collide (which did in fact happen * already on one architecture). The new intepretation is that negative values * are from the pcmk_rc_e enum, and positive values are system error numbers. * Both use 0 for success. * * For system error codes, see: * - /usr/include/asm-generic/errno.h * - /usr/include/asm-generic/errno-base.h */ // Legacy custom return codes for Pacemaker API functions (deprecated) # define pcmk_ok 0 # define PCMK_ERROR_OFFSET 190 /* Replacements on non-linux systems, see include/portability.h */ # define PCMK_CUSTOM_OFFSET 200 /* Purely custom codes */ # define pcmk_err_generic 201 # define pcmk_err_no_quorum 202 # define pcmk_err_schema_validation 203 # define pcmk_err_transform_failed 204 # define pcmk_err_old_data 205 # define pcmk_err_diff_failed 206 # define pcmk_err_diff_resync 207 # define pcmk_err_cib_modified 208 # define pcmk_err_cib_backup 209 # define pcmk_err_cib_save 210 # define pcmk_err_schema_unchanged 211 # define pcmk_err_cib_corrupt 212 # define pcmk_err_multiple 213 # define pcmk_err_node_unknown 214 # define pcmk_err_already 215 /* On HPPA 215 is ENOSYM (Unknown error 215), which hopefully never happens. */ #ifdef __hppa__ # define pcmk_err_bad_nvpair 250 /* 216 is ENOTSOCK */ # define pcmk_err_unknown_format 252 /* 217 is EDESTADDRREQ */ #else # define pcmk_err_bad_nvpair 216 # define pcmk_err_unknown_format 217 #endif /*! * \enum pcmk_rc_e * \brief Return codes for Pacemaker API functions * * Any Pacemaker API function documented as returning a "standard Pacemaker * return code" will return pcmk_rc_ok (0) on success, and one of this * enumeration's other (negative) values or a (positive) system error number * otherwise. The custom codes are at -1001 and lower, so that the caller may * use -1 through -1000 for their own custom values if desired. While generally * referred to as "errors", nonzero values simply indicate a result, which might * or might not be an error depending on the calling context. */ enum pcmk_rc_e { /* When adding new values, use consecutively lower numbers, update the array * in lib/common/results.c, and test with crm_error. */ pcmk_rc_invalid_transition = -1031, pcmk_rc_graph_error = -1030, pcmk_rc_dot_error = -1029, pcmk_rc_underflow = -1028, pcmk_rc_no_input = -1027, pcmk_rc_no_output = -1026, pcmk_rc_after_range = -1025, pcmk_rc_within_range = -1024, pcmk_rc_before_range = -1023, pcmk_rc_undetermined = -1022, pcmk_rc_op_unsatisfied = -1021, pcmk_rc_ipc_pid_only = -1020, pcmk_rc_ipc_unresponsive = -1019, pcmk_rc_ipc_unauthorized = -1018, pcmk_rc_no_quorum = -1017, pcmk_rc_schema_validation = -1016, pcmk_rc_schema_unchanged = -1015, pcmk_rc_transform_failed = -1014, pcmk_rc_old_data = -1013, pcmk_rc_diff_failed = -1012, pcmk_rc_diff_resync = -1011, pcmk_rc_cib_modified = -1010, pcmk_rc_cib_backup = -1009, pcmk_rc_cib_save = -1008, pcmk_rc_cib_corrupt = -1007, pcmk_rc_multiple = -1006, pcmk_rc_node_unknown = -1005, pcmk_rc_already = -1004, pcmk_rc_bad_nvpair = -1003, pcmk_rc_unknown_format = -1002, // Developers: Use a more specific code than pcmk_rc_error whenever possible pcmk_rc_error = -1001, // Values -1 through -1000 reserved for caller use pcmk_rc_ok = 0 // Positive values reserved for system error numbers }; /*! * \enum ocf_exitcode * \brief Exit status codes for resource agents * * The OCF Resource Agent API standard enumerates the possible exit status codes * that agents should return. Pacemaker adds some more to cover problems with * executing the agent, and to use a single set of codes for all agent types. */ enum ocf_exitcode { PCMK_OCF_OK = 0, //!< Success PCMK_OCF_UNKNOWN_ERROR = 1, //!< Unspecified error PCMK_OCF_INVALID_PARAM = 2, //!< Parameter invalid (in local context) PCMK_OCF_UNIMPLEMENT_FEATURE = 3, //!< Requested action not implemented PCMK_OCF_INSUFFICIENT_PRIV = 4, //!< Insufficient privileges PCMK_OCF_NOT_INSTALLED = 5, //!< Dependencies not available locally PCMK_OCF_NOT_CONFIGURED = 6, //!< Parameter invalid (inherently) PCMK_OCF_NOT_RUNNING = 7, //!< Service safely stopped PCMK_OCF_RUNNING_PROMOTED = 8, //!< Service active and promoted PCMK_OCF_FAILED_PROMOTED = 9, //!< Service failed and possibly in promoted role PCMK_OCF_DEGRADED = 190, //!< Service active but more likely to fail soon PCMK_OCF_DEGRADED_PROMOTED = 191, //!< Service promoted but more likely to fail soon // Pacemaker extensions PCMK_OCF_CONNECTION_DIED = 189, //!< \deprecated See PCMK_EXEC_NOT_CONNECTED PCMK_OCF_EXEC_ERROR = 192, //!< Error executing the agent PCMK_OCF_UNKNOWN = 193, //!< Action is pending PCMK_OCF_SIGNAL = 194, //!< Agent terminated due to signal PCMK_OCF_NOT_SUPPORTED = 195, //!< \deprecated (Unused) PCMK_OCF_PENDING = 196, //!< Multi-stage execution in progress PCMK_OCF_CANCELLED = 197, //!< \deprecated (Unused) PCMK_OCF_TIMEOUT = 198, //!< Action did not complete in time PCMK_OCF_OTHER_ERROR = 199, //!< \deprecated (Unused) #if !defined(PCMK_ALLOW_DEPRECATED) || (PCMK_ALLOW_DEPRECATED == 1) //! \deprecated Use PCMK_OCF_RUNNING_PROMOTED instead PCMK_OCF_RUNNING_MASTER = PCMK_OCF_RUNNING_PROMOTED, //! \deprecated Use PCMK_OCF_FAILED_PROMOTED instead PCMK_OCF_FAILED_MASTER = PCMK_OCF_FAILED_PROMOTED, //! \deprecated Use PCMK_OCF_DEGRADED_PROMOTED instead PCMK_OCF_DEGRADED_MASTER = PCMK_OCF_DEGRADED_PROMOTED, #endif }; /*! * \enum crm_exit_e * \brief Exit status codes for tools and daemons * * We want well-specified (i.e. OS-invariant) exit status codes for our daemons * and applications so they can be relied on by callers. (Function return codes * and errno's do not make good exit statuses.) * * The only hard rule is that exit statuses must be between 0 and 255; all else * is convention. Universally, 0 is success, and 1 is generic error (excluding * OSes we don't support -- for example, OpenVMS considers 1 success!). * * For init scripts, the LSB gives meaning to 0-7, and sets aside 150-199 for * application use. OCF adds 8-9 and 189-199. * * sysexits.h was an attempt to give additional meanings, but never really * caught on. It uses 0 and 64-78. * * Bash reserves 2 ("incorrect builtin usage") and 126-255 (126 is "command * found but not executable", 127 is "command not found", 128 + n is * "interrupted by signal n"). * * tldp.org recommends 64-113 for application use. * * We try to overlap with the above conventions when practical. */ typedef enum crm_exit_e { // Common convention CRM_EX_OK = 0, //!< Success CRM_EX_ERROR = 1, //!< Unspecified error // LSB + OCF CRM_EX_INVALID_PARAM = 2, //!< Parameter invalid (in local context) CRM_EX_UNIMPLEMENT_FEATURE = 3, //!< Requested action not implemented CRM_EX_INSUFFICIENT_PRIV = 4, //!< Insufficient privileges CRM_EX_NOT_INSTALLED = 5, //!< Dependencies not available locally CRM_EX_NOT_CONFIGURED = 6, //!< Parameter invalid (inherently) CRM_EX_NOT_RUNNING = 7, //!< Service safely stopped // sysexits.h CRM_EX_USAGE = 64, //!< Command line usage error CRM_EX_DATAERR = 65, //!< User-supplied data incorrect CRM_EX_NOINPUT = 66, //!< Input file not available CRM_EX_NOUSER = 67, //!< User does not exist CRM_EX_NOHOST = 68, //!< Host unknown CRM_EX_UNAVAILABLE = 69, //!< Needed service unavailable CRM_EX_SOFTWARE = 70, //!< Internal software bug CRM_EX_OSERR = 71, //!< External (OS/environmental) problem CRM_EX_OSFILE = 72, //!< System file not usable CRM_EX_CANTCREAT = 73, //!< File couldn't be created CRM_EX_IOERR = 74, //!< File I/O error CRM_EX_TEMPFAIL = 75, //!< Try again CRM_EX_PROTOCOL = 76, //!< Protocol violated CRM_EX_NOPERM = 77, //!< Non-file permission issue CRM_EX_CONFIG = 78, //!< Misconfiguration // Custom CRM_EX_FATAL = 100, //!< Do not respawn CRM_EX_PANIC = 101, //!< Panic the local host CRM_EX_DISCONNECT = 102, //!< Lost connection to something CRM_EX_OLD = 103, //!< Update older than existing config CRM_EX_DIGEST = 104, //!< Digest comparison failed CRM_EX_NOSUCH = 105, //!< Requested item does not exist CRM_EX_QUORUM = 106, //!< Local partition does not have quorum CRM_EX_UNSAFE = 107, //!< Requires --force or new conditions CRM_EX_EXISTS = 108, //!< Requested item already exists CRM_EX_MULTIPLE = 109, //!< Requested item has multiple matches CRM_EX_EXPIRED = 110, //!< Requested item has expired CRM_EX_NOT_YET_IN_EFFECT = 111, //!< Requested item is not in effect CRM_EX_INDETERMINATE = 112, //!< Could not determine status CRM_EX_UNSATISFIED = 113, //!< Requested item does not satisfy constraints // Other CRM_EX_TIMEOUT = 124, //!< Convention from timeout(1) CRM_EX_MAX = 255, //!< Ensure crm_exit_t can hold this } crm_exit_t; /*! * \enum pcmk_exec_status * \brief Execution status * * These codes are used to specify the result of the attempt to execute an * agent, rather than the agent's result itself. */ enum pcmk_exec_status { PCMK_EXEC_UNKNOWN = -2, //!< Used only to initialize variables PCMK_EXEC_PENDING = -1, //!< Action is in progress PCMK_EXEC_DONE, //!< Action completed, result is known PCMK_EXEC_CANCELLED, //!< Action was cancelled PCMK_EXEC_TIMEOUT, //!< Action did not complete in time PCMK_EXEC_NOT_SUPPORTED, //!< Agent does not implement requested action PCMK_EXEC_ERROR, //!< Execution failed, may be retried PCMK_EXEC_ERROR_HARD, //!< Execution failed, do not retry on node PCMK_EXEC_ERROR_FATAL, //!< Execution failed, do not retry anywhere PCMK_EXEC_NOT_INSTALLED, //!< Agent or dependency not available locally PCMK_EXEC_NOT_CONNECTED, //!< No connection to executor PCMK_EXEC_INVALID, //!< Action cannot be attempted (e.g. shutdown) }; const char *pcmk_rc_name(int rc); const char *pcmk_rc_str(int rc); crm_exit_t pcmk_rc2exitc(int rc); +enum ocf_exitcode pcmk_rc2ocf(int rc); int pcmk_rc2legacy(int rc); int pcmk_legacy2rc(int legacy_rc); const char *pcmk_strerror(int rc); const char *pcmk_errorname(int rc); const char *bz2_strerror(int rc); crm_exit_t crm_errno2exit(int rc); const char *crm_exit_name(crm_exit_t exit_code); const char *crm_exit_str(crm_exit_t exit_code); _Noreturn crm_exit_t crm_exit(crm_exit_t rc); static inline const char * pcmk_exec_status_str(enum pcmk_exec_status status) { switch (status) { case PCMK_EXEC_PENDING: return "pending"; case PCMK_EXEC_DONE: return "complete"; case PCMK_EXEC_CANCELLED: return "Cancelled"; case PCMK_EXEC_TIMEOUT: return "Timed Out"; case PCMK_EXEC_NOT_SUPPORTED: return "NOT SUPPORTED"; case PCMK_EXEC_ERROR: return "Error"; case PCMK_EXEC_ERROR_HARD: return "Hard error"; case PCMK_EXEC_ERROR_FATAL: return "Fatal error"; case PCMK_EXEC_NOT_INSTALLED: return "Not installed"; case PCMK_EXEC_NOT_CONNECTED: return "No executor connection"; case PCMK_EXEC_INVALID: return "Cannot execute now"; default: return "UNKNOWN!"; } } #ifdef __cplusplus } #endif #endif diff --git a/include/crm/common/results_internal.h b/include/crm/common/results_internal.h index 4a90305b82..b06e921404 100644 --- a/include/crm/common/results_internal.h +++ b/include/crm/common/results_internal.h @@ -1,23 +1,42 @@ /* * Copyright 2020-2021 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__COMMON_RESULTS_INTERNAL__H #define PCMK__COMMON_RESULTS_INTERNAL__H #include // GQuark /* Error domains for use with g_set_error */ GQuark pcmk__rc_error_quark(void); GQuark pcmk__exitc_error_quark(void); #define PCMK__RC_ERROR pcmk__rc_error_quark() #define PCMK__EXITC_ERROR pcmk__exitc_error_quark() +/* Action results */ + +typedef struct { + int exit_status; // Child exit status + enum pcmk_exec_status execution_status; // Execution status + char *exit_reason; // Brief, human-friendly explanation + char *action_stdout; // Action output + char *action_stderr; // Action error output +} pcmk__action_result_t; + +void pcmk__set_result(pcmk__action_result_t *result, int exit_status, + enum pcmk_exec_status exec_status, + const char *exit_reason); + +void pcmk__set_result_output(pcmk__action_result_t *result, + const char *out, const char *err); + +void pcmk__reset_result(pcmk__action_result_t *result); + #endif // PCMK__COMMON_RESULTS_INTERNAL__H diff --git a/include/crm/lrmd_internal.h b/include/crm/lrmd_internal.h index 4480285589..284c4d6a9a 100644 --- a/include/crm/lrmd_internal.h +++ b/include/crm/lrmd_internal.h @@ -1,78 +1,83 @@ /* * Copyright 2015-2021 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 LRMD_INTERNAL__H #define LRMD_INTERNAL__H #include // uint32_t #include // GList, GHashTable, gpointer #include // xmlNode #include // crm_ipc_t #include // mainloop_io_t, ipc_client_callbacks #include // pcmk__output_t #include // pcmk__remote_t #include // lrmd_t, lrmd_event_data_t int lrmd__new(lrmd_t **api, const char *nodename, const char *server, int port); int lrmd_send_attribute_alert(lrmd_t *lrmd, GList *alert_list, const char *node, uint32_t nodeid, const char *attr_name, const char *attr_value); int lrmd_send_node_alert(lrmd_t *lrmd, GList *alert_list, const char *node, uint32_t nodeid, const char *state); int lrmd_send_fencing_alert(lrmd_t *lrmd, GList *alert_list, const char *target, const char *task, const char *desc, int op_rc); int lrmd_send_resource_alert(lrmd_t *lrmd, GList *alert_list, const char *node, lrmd_event_data_t *op); int lrmd__remote_send_xml(pcmk__remote_t *session, xmlNode *msg, uint32_t id, const char *msg_type); +void lrmd__set_result(lrmd_event_data_t *event, enum ocf_exitcode rc, + int op_status, const char *exit_reason); + +void lrmd__reset_result(lrmd_event_data_t *event); + /* Shared functions for IPC proxy back end */ typedef struct remote_proxy_s { char *node_name; char *session_id; gboolean is_local; crm_ipc_t *ipc; mainloop_io_t *source; uint32_t last_request_id; lrmd_t *lrm; } remote_proxy_t; remote_proxy_t *remote_proxy_new(lrmd_t *lrmd, struct ipc_client_callbacks *proxy_callbacks, const char *node_name, const char *session_id, const char *channel); int lrmd__validate_remote_settings(lrmd_t *lrmd, GHashTable *hash); void remote_proxy_cb(lrmd_t *lrmd, const char *node_name, xmlNode *msg); void remote_proxy_ack_shutdown(lrmd_t *lrmd); void remote_proxy_nack_shutdown(lrmd_t *lrmd); int remote_proxy_dispatch(const char *buffer, ssize_t length, gpointer userdata); void remote_proxy_disconnected(gpointer data); void remote_proxy_free(gpointer data); void remote_proxy_relay_event(remote_proxy_t *proxy, xmlNode *msg); void remote_proxy_relay_response(remote_proxy_t *proxy, xmlNode *msg, int msg_id); void lrmd__register_messages(pcmk__output_t *out); #ifdef HAVE_GNUTLS_GNUTLS_H int lrmd__init_remote_key(gnutls_datum_t *key); #endif #endif diff --git a/include/pcmki/pcmki_transition.h b/include/pcmki/pcmki_transition.h index 26d331145f..17373565d7 100644 --- a/include/pcmki/pcmki_transition.h +++ b/include/pcmki/pcmki_transition.h @@ -1,140 +1,141 @@ /* * Copyright 2004-2021 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 CRM_TRANSITION__H # define CRM_TRANSITION__H #ifdef __cplusplus extern "C" { #endif #include #include #include #include typedef enum { action_type_pseudo, action_type_rsc, action_type_crm } action_type_e; typedef struct te_timer_s crm_action_timer_t; typedef struct crm_graph_s crm_graph_t; typedef struct synapse_s { int id; int priority; gboolean ready; gboolean failed; gboolean executed; gboolean confirmed; GList *actions; /* crm_action_t* */ GList *inputs; /* crm_action_t* */ } synapse_t; typedef struct crm_action_s { int id; int timeout; guint interval_ms; GHashTable *params; action_type_e type; crm_action_timer_t *timer; synapse_t *synapse; gboolean sent_update; /* sent to the CIB */ gboolean executed; /* sent to the CRM */ gboolean confirmed; gboolean failed; gboolean can_fail; //! \deprecated Will be removed in a future release xmlNode *xml; } crm_action_t; struct te_timer_s { int source_id; int timeout; crm_action_t *action; }; /* order matters here */ enum transition_action { tg_done, tg_stop, tg_restart, tg_shutdown, }; struct crm_graph_s { int id; char *source; int abort_priority; gboolean complete; const char *abort_reason; enum transition_action 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; /* synapse_t* */ int migration_limit; }; typedef struct crm_graph_functions_s { gboolean(*pseudo) (crm_graph_t * graph, crm_action_t * action); gboolean(*rsc) (crm_graph_t * graph, crm_action_t * action); gboolean(*crmd) (crm_graph_t * graph, crm_action_t * action); gboolean(*stonith) (crm_graph_t * graph, crm_action_t * action); gboolean(*allowed) (crm_graph_t * graph, crm_action_t * action); } crm_graph_functions_t; enum transition_status { transition_active, transition_pending, /* active but no actions performed this time */ transition_complete, transition_stopped, transition_terminated, transition_action_failed, transition_failed, }; void pcmk__set_graph_functions(crm_graph_functions_t *fns); crm_graph_t *pcmk__unpack_graph(xmlNode *xml_graph, const char *reference); enum transition_status pcmk__execute_graph(crm_graph_t *graph); void pcmk__update_graph(crm_graph_t *graph, crm_action_t *action); void pcmk__free_graph(crm_graph_t *graph); const char *pcmk__graph_status2text(enum transition_status state); void pcmk__log_graph(unsigned int log_level, crm_graph_t *graph); void pcmk__log_graph_action(int log_level, crm_action_t *action); lrmd_event_data_t *pcmk__event_from_graph_action(xmlNode *resource, crm_action_t *action, - int status, int rc); + int status, int rc, + const char *exit_reason); #ifdef __cplusplus } #endif #endif diff --git a/lib/common/results.c b/lib/common/results.c index dd901fb62b..cd1ebe3a67 100644 --- a/lib/common/results.c +++ b/lib/common/results.c @@ -1,796 +1,904 @@ /* * Copyright 2004-2020 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include G_DEFINE_QUARK(pcmk-rc-error-quark, pcmk__rc_error) G_DEFINE_QUARK(pcmk-exitc-error-quark, pcmk__exitc_error) // @COMPAT Legacy function return codes //! \deprecated Use standard return codes and pcmk_rc_name() instead const char * pcmk_errorname(int rc) { rc = abs(rc); switch (rc) { case pcmk_err_generic: return "pcmk_err_generic"; case pcmk_err_no_quorum: return "pcmk_err_no_quorum"; case pcmk_err_schema_validation: return "pcmk_err_schema_validation"; case pcmk_err_transform_failed: return "pcmk_err_transform_failed"; case pcmk_err_old_data: return "pcmk_err_old_data"; case pcmk_err_diff_failed: return "pcmk_err_diff_failed"; case pcmk_err_diff_resync: return "pcmk_err_diff_resync"; case pcmk_err_cib_modified: return "pcmk_err_cib_modified"; case pcmk_err_cib_backup: return "pcmk_err_cib_backup"; case pcmk_err_cib_save: return "pcmk_err_cib_save"; case pcmk_err_cib_corrupt: return "pcmk_err_cib_corrupt"; case pcmk_err_multiple: return "pcmk_err_multiple"; case pcmk_err_node_unknown: return "pcmk_err_node_unknown"; case pcmk_err_already: return "pcmk_err_already"; case pcmk_err_bad_nvpair: return "pcmk_err_bad_nvpair"; case pcmk_err_unknown_format: return "pcmk_err_unknown_format"; default: return pcmk_rc_name(rc); // system errno } } //! \deprecated Use standard return codes and pcmk_rc_str() instead const char * pcmk_strerror(int rc) { if (rc == 0) { return "OK"; } rc = abs(rc); // Of course rc > 0 ... unless someone passed INT_MIN as rc if ((rc > 0) && (rc < PCMK_ERROR_OFFSET)) { return strerror(rc); } switch (rc) { case pcmk_err_generic: return "Generic Pacemaker error"; case pcmk_err_no_quorum: return "Operation requires quorum"; case pcmk_err_schema_validation: return "Update does not conform to the configured schema"; case pcmk_err_transform_failed: return "Schema transform failed"; case pcmk_err_old_data: return "Update was older than existing configuration"; case pcmk_err_diff_failed: return "Application of an update diff failed"; case pcmk_err_diff_resync: return "Application of an update diff failed, requesting a full refresh"; case pcmk_err_cib_modified: return "The on-disk configuration was manually modified"; case pcmk_err_cib_backup: return "Could not archive the previous configuration"; case pcmk_err_cib_save: return "Could not save the new configuration to disk"; case pcmk_err_cib_corrupt: return "Could not parse on-disk configuration"; case pcmk_err_multiple: return "Resource active on multiple nodes"; case pcmk_err_node_unknown: return "Node not found"; case pcmk_err_already: return "Situation already as requested"; case pcmk_err_bad_nvpair: return "Bad name/value pair given"; case pcmk_err_schema_unchanged: return "Schema is already the latest available"; case pcmk_err_unknown_format: return "Unknown output format"; /* The following cases will only be hit on systems for which they are non-standard */ /* coverity[dead_error_condition] False positive on non-Linux */ case ENOTUNIQ: return "Name not unique on network"; /* coverity[dead_error_condition] False positive on non-Linux */ case ECOMM: return "Communication error on send"; /* coverity[dead_error_condition] False positive on non-Linux */ case ELIBACC: return "Can not access a needed shared library"; /* coverity[dead_error_condition] False positive on non-Linux */ case EREMOTEIO: return "Remote I/O error"; /* coverity[dead_error_condition] False positive on non-Linux */ case EUNATCH: return "Protocol driver not attached"; /* coverity[dead_error_condition] False positive on non-Linux */ case ENOKEY: return "Required key not available"; } crm_err("Unknown error code: %d", rc); return "Unknown error"; } // Standard Pacemaker API return codes /* This array is used only for nonzero values of pcmk_rc_e. Its values must be * kept in the exact reverse order of the enum value numbering (i.e. add new * values to the end of the array). */ static struct pcmk__rc_info { const char *name; const char *desc; int legacy_rc; } pcmk__rcs[] = { { "pcmk_rc_error", "Error", -pcmk_err_generic, }, { "pcmk_rc_unknown_format", "Unknown output format", -pcmk_err_unknown_format, }, { "pcmk_rc_bad_nvpair", "Bad name/value pair given", -pcmk_err_bad_nvpair, }, { "pcmk_rc_already", "Already in requested state", -pcmk_err_already, }, { "pcmk_rc_node_unknown", "Node not found", -pcmk_err_node_unknown, }, { "pcmk_rc_multiple", "Resource active on multiple nodes", -pcmk_err_multiple, }, { "pcmk_rc_cib_corrupt", "Could not parse on-disk configuration", -pcmk_err_cib_corrupt, }, { "pcmk_rc_cib_save", "Could not save new configuration to disk", -pcmk_err_cib_save, }, { "pcmk_rc_cib_backup", "Could not archive previous configuration", -pcmk_err_cib_backup, }, { "pcmk_rc_cib_modified", "On-disk configuration was manually modified", -pcmk_err_cib_modified, }, { "pcmk_rc_diff_resync", "Application of update diff failed, requesting full refresh", -pcmk_err_diff_resync, }, { "pcmk_rc_diff_failed", "Application of update diff failed", -pcmk_err_diff_failed, }, { "pcmk_rc_old_data", "Update was older than existing configuration", -pcmk_err_old_data, }, { "pcmk_rc_transform_failed", "Schema transform failed", -pcmk_err_transform_failed, }, { "pcmk_rc_schema_unchanged", "Schema is already the latest available", -pcmk_err_schema_unchanged, }, { "pcmk_rc_schema_validation", "Update does not conform to the configured schema", -pcmk_err_schema_validation, }, { "pcmk_rc_no_quorum", "Operation requires quorum", -pcmk_err_no_quorum, }, { "pcmk_rc_ipc_pid_only", "IPC server process is active but not accepting connections", -pcmk_err_generic, }, { "pcmk_rc_ipc_unresponsive", "IPC server is unresponsive", -pcmk_err_generic, }, { "pcmk_rc_ipc_unauthorized", "IPC server is blocked by unauthorized process", -pcmk_err_generic, }, { "pcmk_rc_op_unsatisifed", "Not applicable under current conditions", -pcmk_err_generic, }, { "pcmk_rc_undetermined", "Result undetermined", -pcmk_err_generic, }, { "pcmk_rc_before_range", "Result occurs before given range", -pcmk_err_generic, }, { "pcmk_rc_within_range", "Result occurs within given range", -pcmk_err_generic, }, { "pcmk_rc_after_range", "Result occurs after given range", -pcmk_err_generic, }, { "pcmk_rc_no_output", "Output message produced no output", -pcmk_err_generic, }, { "pcmk_rc_no_input", "Input file not available", -pcmk_err_generic, }, { "pcmk_rc_underflow", "Value too small to be stored in data type", -pcmk_err_generic, }, { "pcmk_rc_dot_error", "Error writing dot(1) file", -pcmk_err_generic, }, { "pcmk_rc_graph_error", "Error writing graph file", -pcmk_err_generic, }, { "pcmk_rc_invalid_transition", "Cluster simulation produced invalid transition", -pcmk_err_generic, }, }; #define PCMK__N_RC (sizeof(pcmk__rcs) / sizeof(struct pcmk__rc_info)) /*! * \brief Get a return code constant name as a string * * \param[in] rc Integer return code to convert * * \return String of constant name corresponding to rc */ const char * pcmk_rc_name(int rc) { if ((rc <= pcmk_rc_error) && ((pcmk_rc_error - rc) < PCMK__N_RC)) { return pcmk__rcs[pcmk_rc_error - rc].name; } switch (rc) { case pcmk_rc_ok: return "pcmk_rc_ok"; case E2BIG: return "E2BIG"; case EACCES: return "EACCES"; case EADDRINUSE: return "EADDRINUSE"; case EADDRNOTAVAIL: return "EADDRNOTAVAIL"; case EAFNOSUPPORT: return "EAFNOSUPPORT"; case EAGAIN: return "EAGAIN"; case EALREADY: return "EALREADY"; case EBADF: return "EBADF"; case EBADMSG: return "EBADMSG"; case EBUSY: return "EBUSY"; case ECANCELED: return "ECANCELED"; case ECHILD: return "ECHILD"; case ECOMM: return "ECOMM"; case ECONNABORTED: return "ECONNABORTED"; case ECONNREFUSED: return "ECONNREFUSED"; case ECONNRESET: return "ECONNRESET"; /* case EDEADLK: return "EDEADLK"; */ case EDESTADDRREQ: return "EDESTADDRREQ"; case EDOM: return "EDOM"; case EDQUOT: return "EDQUOT"; case EEXIST: return "EEXIST"; case EFAULT: return "EFAULT"; case EFBIG: return "EFBIG"; case EHOSTDOWN: return "EHOSTDOWN"; case EHOSTUNREACH: return "EHOSTUNREACH"; case EIDRM: return "EIDRM"; case EILSEQ: return "EILSEQ"; case EINPROGRESS: return "EINPROGRESS"; case EINTR: return "EINTR"; case EINVAL: return "EINVAL"; case EIO: return "EIO"; case EISCONN: return "EISCONN"; case EISDIR: return "EISDIR"; case ELIBACC: return "ELIBACC"; case ELOOP: return "ELOOP"; case EMFILE: return "EMFILE"; case EMLINK: return "EMLINK"; case EMSGSIZE: return "EMSGSIZE"; #ifdef EMULTIHOP // Not available on OpenBSD case EMULTIHOP: return "EMULTIHOP"; #endif case ENAMETOOLONG: return "ENAMETOOLONG"; case ENETDOWN: return "ENETDOWN"; case ENETRESET: return "ENETRESET"; case ENETUNREACH: return "ENETUNREACH"; case ENFILE: return "ENFILE"; case ENOBUFS: return "ENOBUFS"; case ENODATA: return "ENODATA"; case ENODEV: return "ENODEV"; case ENOENT: return "ENOENT"; case ENOEXEC: return "ENOEXEC"; case ENOKEY: return "ENOKEY"; case ENOLCK: return "ENOLCK"; #ifdef ENOLINK // Not available on OpenBSD case ENOLINK: return "ENOLINK"; #endif case ENOMEM: return "ENOMEM"; case ENOMSG: return "ENOMSG"; case ENOPROTOOPT: return "ENOPROTOOPT"; case ENOSPC: return "ENOSPC"; case ENOSR: return "ENOSR"; case ENOSTR: return "ENOSTR"; case ENOSYS: return "ENOSYS"; case ENOTBLK: return "ENOTBLK"; case ENOTCONN: return "ENOTCONN"; case ENOTDIR: return "ENOTDIR"; case ENOTEMPTY: return "ENOTEMPTY"; case ENOTSOCK: return "ENOTSOCK"; #if ENOTSUP != EOPNOTSUPP case ENOTSUP: return "ENOTSUP"; #endif case ENOTTY: return "ENOTTY"; case ENOTUNIQ: return "ENOTUNIQ"; case ENXIO: return "ENXIO"; case EOPNOTSUPP: return "EOPNOTSUPP"; case EOVERFLOW: return "EOVERFLOW"; case EPERM: return "EPERM"; case EPFNOSUPPORT: return "EPFNOSUPPORT"; case EPIPE: return "EPIPE"; case EPROTO: return "EPROTO"; case EPROTONOSUPPORT: return "EPROTONOSUPPORT"; case EPROTOTYPE: return "EPROTOTYPE"; case ERANGE: return "ERANGE"; case EREMOTE: return "EREMOTE"; case EREMOTEIO: return "EREMOTEIO"; case EROFS: return "EROFS"; case ESHUTDOWN: return "ESHUTDOWN"; case ESPIPE: return "ESPIPE"; case ESOCKTNOSUPPORT: return "ESOCKTNOSUPPORT"; case ESRCH: return "ESRCH"; case ESTALE: return "ESTALE"; case ETIME: return "ETIME"; case ETIMEDOUT: return "ETIMEDOUT"; case ETXTBSY: return "ETXTBSY"; case EUNATCH: return "EUNATCH"; case EUSERS: return "EUSERS"; /* case EWOULDBLOCK: return "EWOULDBLOCK"; */ case EXDEV: return "EXDEV"; #ifdef EBADE // Not available on OS X case EBADE: return "EBADE"; case EBADFD: return "EBADFD"; case EBADSLT: return "EBADSLT"; case EDEADLOCK: return "EDEADLOCK"; case EBADR: return "EBADR"; case EBADRQC: return "EBADRQC"; case ECHRNG: return "ECHRNG"; #ifdef EISNAM // Not available on OS X, Illumos, Solaris case EISNAM: return "EISNAM"; case EKEYEXPIRED: return "EKEYEXPIRED"; case EKEYREJECTED: return "EKEYREJECTED"; case EKEYREVOKED: return "EKEYREVOKED"; #endif case EL2HLT: return "EL2HLT"; case EL2NSYNC: return "EL2NSYNC"; case EL3HLT: return "EL3HLT"; case EL3RST: return "EL3RST"; case ELIBBAD: return "ELIBBAD"; case ELIBMAX: return "ELIBMAX"; case ELIBSCN: return "ELIBSCN"; case ELIBEXEC: return "ELIBEXEC"; #ifdef ENOMEDIUM // Not available on OS X, Illumos, Solaris case ENOMEDIUM: return "ENOMEDIUM"; case EMEDIUMTYPE: return "EMEDIUMTYPE"; #endif case ENONET: return "ENONET"; case ENOPKG: return "ENOPKG"; case EREMCHG: return "EREMCHG"; case ERESTART: return "ERESTART"; case ESTRPIPE: return "ESTRPIPE"; #ifdef EUCLEAN // Not available on OS X, Illumos, Solaris case EUCLEAN: return "EUCLEAN"; #endif case EXFULL: return "EXFULL"; #endif // EBADE default: return "Unknown"; } } /*! * \brief Get a user-friendly description of a return code * * \param[in] rc Integer return code to convert * * \return String description of rc */ const char * pcmk_rc_str(int rc) { if (rc == pcmk_rc_ok) { return "OK"; } if ((rc <= pcmk_rc_error) && ((pcmk_rc_error - rc) < PCMK__N_RC)) { return pcmk__rcs[pcmk_rc_error - rc].desc; } if (rc < 0) { return "Unknown error"; } return strerror(rc); } // This returns negative values for errors //! \deprecated Use standard return codes instead int pcmk_rc2legacy(int rc) { if (rc >= 0) { return -rc; // OK or system errno } if ((rc <= pcmk_rc_error) && ((pcmk_rc_error - rc) < PCMK__N_RC)) { return pcmk__rcs[pcmk_rc_error - rc].legacy_rc; } return -pcmk_err_generic; } //! \deprecated Use standard return codes instead int pcmk_legacy2rc(int legacy_rc) { legacy_rc = abs(legacy_rc); switch (legacy_rc) { case pcmk_err_no_quorum: return pcmk_rc_no_quorum; case pcmk_err_schema_validation: return pcmk_rc_schema_validation; case pcmk_err_schema_unchanged: return pcmk_rc_schema_unchanged; case pcmk_err_transform_failed: return pcmk_rc_transform_failed; case pcmk_err_old_data: return pcmk_rc_old_data; case pcmk_err_diff_failed: return pcmk_rc_diff_failed; case pcmk_err_diff_resync: return pcmk_rc_diff_resync; case pcmk_err_cib_modified: return pcmk_rc_cib_modified; case pcmk_err_cib_backup: return pcmk_rc_cib_backup; case pcmk_err_cib_save: return pcmk_rc_cib_save; case pcmk_err_cib_corrupt: return pcmk_rc_cib_corrupt; case pcmk_err_multiple: return pcmk_rc_multiple; case pcmk_err_node_unknown: return pcmk_rc_node_unknown; case pcmk_err_already: return pcmk_rc_already; case pcmk_err_bad_nvpair: return pcmk_rc_bad_nvpair; case pcmk_err_unknown_format: return pcmk_rc_unknown_format; case pcmk_err_generic: return pcmk_rc_error; case pcmk_ok: return pcmk_rc_ok; default: return legacy_rc; // system errno } } // Exit status codes const char * crm_exit_name(crm_exit_t exit_code) { switch (exit_code) { case CRM_EX_OK: return "CRM_EX_OK"; case CRM_EX_ERROR: return "CRM_EX_ERROR"; case CRM_EX_INVALID_PARAM: return "CRM_EX_INVALID_PARAM"; case CRM_EX_UNIMPLEMENT_FEATURE: return "CRM_EX_UNIMPLEMENT_FEATURE"; case CRM_EX_INSUFFICIENT_PRIV: return "CRM_EX_INSUFFICIENT_PRIV"; case CRM_EX_NOT_INSTALLED: return "CRM_EX_NOT_INSTALLED"; case CRM_EX_NOT_CONFIGURED: return "CRM_EX_NOT_CONFIGURED"; case CRM_EX_NOT_RUNNING: return "CRM_EX_NOT_RUNNING"; case CRM_EX_USAGE: return "CRM_EX_USAGE"; case CRM_EX_DATAERR: return "CRM_EX_DATAERR"; case CRM_EX_NOINPUT: return "CRM_EX_NOINPUT"; case CRM_EX_NOUSER: return "CRM_EX_NOUSER"; case CRM_EX_NOHOST: return "CRM_EX_NOHOST"; case CRM_EX_UNAVAILABLE: return "CRM_EX_UNAVAILABLE"; case CRM_EX_SOFTWARE: return "CRM_EX_SOFTWARE"; case CRM_EX_OSERR: return "CRM_EX_OSERR"; case CRM_EX_OSFILE: return "CRM_EX_OSFILE"; case CRM_EX_CANTCREAT: return "CRM_EX_CANTCREAT"; case CRM_EX_IOERR: return "CRM_EX_IOERR"; case CRM_EX_TEMPFAIL: return "CRM_EX_TEMPFAIL"; case CRM_EX_PROTOCOL: return "CRM_EX_PROTOCOL"; case CRM_EX_NOPERM: return "CRM_EX_NOPERM"; case CRM_EX_CONFIG: return "CRM_EX_CONFIG"; case CRM_EX_FATAL: return "CRM_EX_FATAL"; case CRM_EX_PANIC: return "CRM_EX_PANIC"; case CRM_EX_DISCONNECT: return "CRM_EX_DISCONNECT"; case CRM_EX_DIGEST: return "CRM_EX_DIGEST"; case CRM_EX_NOSUCH: return "CRM_EX_NOSUCH"; case CRM_EX_QUORUM: return "CRM_EX_QUORUM"; case CRM_EX_UNSAFE: return "CRM_EX_UNSAFE"; case CRM_EX_EXISTS: return "CRM_EX_EXISTS"; case CRM_EX_MULTIPLE: return "CRM_EX_MULTIPLE"; case CRM_EX_EXPIRED: return "CRM_EX_EXPIRED"; case CRM_EX_NOT_YET_IN_EFFECT: return "CRM_EX_NOT_YET_IN_EFFECT"; case CRM_EX_INDETERMINATE: return "CRM_EX_INDETERMINATE"; case CRM_EX_UNSATISFIED: return "CRM_EX_UNSATISFIED"; case CRM_EX_OLD: return "CRM_EX_OLD"; case CRM_EX_TIMEOUT: return "CRM_EX_TIMEOUT"; case CRM_EX_MAX: return "CRM_EX_UNKNOWN"; } return "CRM_EX_UNKNOWN"; } const char * crm_exit_str(crm_exit_t exit_code) { switch (exit_code) { case CRM_EX_OK: return "OK"; case CRM_EX_ERROR: return "Error occurred"; case CRM_EX_INVALID_PARAM: return "Invalid parameter"; case CRM_EX_UNIMPLEMENT_FEATURE: return "Unimplemented"; case CRM_EX_INSUFFICIENT_PRIV: return "Insufficient privileges"; case CRM_EX_NOT_INSTALLED: return "Not installed"; case CRM_EX_NOT_CONFIGURED: return "Not configured"; case CRM_EX_NOT_RUNNING: return "Not running"; case CRM_EX_USAGE: return "Incorrect usage"; case CRM_EX_DATAERR: return "Invalid data given"; case CRM_EX_NOINPUT: return "Input file not available"; case CRM_EX_NOUSER: return "User does not exist"; case CRM_EX_NOHOST: return "Host does not exist"; case CRM_EX_UNAVAILABLE: return "Necessary service unavailable"; case CRM_EX_SOFTWARE: return "Internal software bug"; case CRM_EX_OSERR: return "Operating system error occurred"; case CRM_EX_OSFILE: return "System file not available"; case CRM_EX_CANTCREAT: return "Cannot create output file"; case CRM_EX_IOERR: return "I/O error occurred"; case CRM_EX_TEMPFAIL: return "Temporary failure, try again"; case CRM_EX_PROTOCOL: return "Protocol violated"; case CRM_EX_NOPERM: return "Insufficient privileges"; case CRM_EX_CONFIG: return "Invalid configuration"; case CRM_EX_FATAL: return "Fatal error occurred, will not respawn"; case CRM_EX_PANIC: return "System panic required"; case CRM_EX_DISCONNECT: return "Not connected"; case CRM_EX_DIGEST: return "Digest mismatch"; case CRM_EX_NOSUCH: return "No such object"; case CRM_EX_QUORUM: return "Quorum required"; case CRM_EX_UNSAFE: return "Operation not safe"; case CRM_EX_EXISTS: return "Requested item already exists"; case CRM_EX_MULTIPLE: return "Multiple items match request"; case CRM_EX_EXPIRED: return "Requested item has expired"; case CRM_EX_NOT_YET_IN_EFFECT: return "Requested item is not yet in effect"; case CRM_EX_INDETERMINATE: return "Could not determine status"; case CRM_EX_UNSATISFIED: return "Not applicable under current conditions"; case CRM_EX_OLD: return "Update was older than existing configuration"; case CRM_EX_TIMEOUT: return "Timeout occurred"; case CRM_EX_MAX: return "Error occurred"; } if ((exit_code > 128) && (exit_code < CRM_EX_MAX)) { return "Interrupted by signal"; } return "Unknown exit status"; } //! \deprecated Use standard return codes and pcmk_rc2exitc() instead crm_exit_t crm_errno2exit(int rc) { rc = abs(rc); // Convenience for functions that return -errno switch (rc) { case pcmk_ok: return CRM_EX_OK; case pcmk_err_no_quorum: return CRM_EX_QUORUM; case pcmk_err_old_data: return CRM_EX_OLD; case pcmk_err_schema_validation: case pcmk_err_transform_failed: return CRM_EX_CONFIG; case pcmk_err_bad_nvpair: return CRM_EX_INVALID_PARAM; case pcmk_err_already: return CRM_EX_EXISTS; case pcmk_err_multiple: return CRM_EX_MULTIPLE; case pcmk_err_node_unknown: case pcmk_err_unknown_format: return CRM_EX_NOSUCH; default: return pcmk_rc2exitc(rc); // system errno } } /*! * \brief Map a function return code to the most similar exit code * * \param[in] rc Function return code * * \return Most similar exit code */ crm_exit_t pcmk_rc2exitc(int rc) { switch (rc) { case pcmk_rc_ok: return CRM_EX_OK; case pcmk_rc_no_quorum: return CRM_EX_QUORUM; case pcmk_rc_old_data: return CRM_EX_OLD; case pcmk_rc_schema_validation: case pcmk_rc_transform_failed: return CRM_EX_CONFIG; case pcmk_rc_bad_nvpair: return CRM_EX_INVALID_PARAM; case EACCES: return CRM_EX_INSUFFICIENT_PRIV; case EBADF: case EINVAL: case EFAULT: case ENOSYS: case EOVERFLOW: case pcmk_rc_underflow: return CRM_EX_SOFTWARE; case EBADMSG: case EMSGSIZE: case ENOMSG: case ENOPROTOOPT: case EPROTO: case EPROTONOSUPPORT: case EPROTOTYPE: return CRM_EX_PROTOCOL; case ECOMM: case ENOMEM: return CRM_EX_OSERR; case ECONNABORTED: case ECONNREFUSED: case ECONNRESET: case ENOTCONN: return CRM_EX_DISCONNECT; case EEXIST: case pcmk_rc_already: return CRM_EX_EXISTS; case EIO: case pcmk_rc_no_output: case pcmk_rc_dot_error: case pcmk_rc_graph_error: return CRM_EX_IOERR; case ENOTSUP: #if EOPNOTSUPP != ENOTSUP case EOPNOTSUPP: #endif return CRM_EX_UNIMPLEMENT_FEATURE; case ENOTUNIQ: case pcmk_rc_multiple: return CRM_EX_MULTIPLE; case ENXIO: case pcmk_rc_node_unknown: case pcmk_rc_unknown_format: return CRM_EX_NOSUCH; case ETIME: case ETIMEDOUT: return CRM_EX_TIMEOUT; case EAGAIN: case EBUSY: return CRM_EX_UNSATISFIED; case pcmk_rc_before_range: return CRM_EX_NOT_YET_IN_EFFECT; case pcmk_rc_after_range: return CRM_EX_EXPIRED; case pcmk_rc_undetermined: return CRM_EX_INDETERMINATE; case pcmk_rc_op_unsatisfied: return CRM_EX_UNSATISFIED; case pcmk_rc_within_range: return CRM_EX_OK; case pcmk_rc_no_input: return CRM_EX_NOINPUT; default: return CRM_EX_ERROR; } } +/*! + * \brief Map a function return code to the most similar OCF exit code + * + * \param[in] rc Function return code + * + * \return Most similar OCF exit code + */ +enum ocf_exitcode +pcmk_rc2ocf(int rc) +{ + switch (rc) { + case pcmk_rc_ok: + return PCMK_OCF_OK; + + case pcmk_rc_bad_nvpair: + return PCMK_OCF_INVALID_PARAM; + + case EACCES: + return PCMK_OCF_INSUFFICIENT_PRIV; + + case ENOTSUP: +#if EOPNOTSUPP != ENOTSUP + case EOPNOTSUPP: +#endif + return PCMK_OCF_UNIMPLEMENT_FEATURE; + + default: + return PCMK_OCF_UNKNOWN_ERROR; + } +} + + // Other functions const char * bz2_strerror(int rc) { // See ftp://sources.redhat.com/pub/bzip2/docs/manual_3.html#SEC17 switch (rc) { case BZ_OK: case BZ_RUN_OK: case BZ_FLUSH_OK: case BZ_FINISH_OK: case BZ_STREAM_END: return "Ok"; case BZ_CONFIG_ERROR: return "libbz2 has been improperly compiled on your platform"; case BZ_SEQUENCE_ERROR: return "library functions called in the wrong order"; case BZ_PARAM_ERROR: return "parameter is out of range or otherwise incorrect"; case BZ_MEM_ERROR: return "memory allocation failed"; case BZ_DATA_ERROR: return "data integrity error is detected during decompression"; case BZ_DATA_ERROR_MAGIC: return "the compressed stream does not start with the correct magic bytes"; case BZ_IO_ERROR: return "error reading or writing in the compressed file"; case BZ_UNEXPECTED_EOF: return "compressed file finishes before the logical end of stream is detected"; case BZ_OUTBUFF_FULL: return "output data will not fit into the buffer provided"; } return "Unknown error"; } crm_exit_t crm_exit(crm_exit_t rc) { /* A compiler could theoretically use any type for crm_exit_t, but an int * should always hold it, so cast to int to keep static analysis happy. */ if ((((int) rc) < 0) || (((int) rc) > CRM_EX_MAX)) { rc = CRM_EX_ERROR; } mainloop_cleanup(); crm_xml_cleanup(); pcmk__cli_option_cleanup(); if (crm_system_name) { crm_info("Exiting %s " CRM_XS " with status %d", crm_system_name, rc); free(crm_system_name); } else { crm_trace("Exiting with status %d", rc); } qb_log_fini(); // Don't log anything after this point exit(rc); } + +/* + * External action results + */ + +/*! + * \internal + * \brief Set the result of an action + * + * \param[out] result Where to set action result + * \param[in] exit_status OCF exit status to set + * \param[in] exec_status Execution status to set + * \param[in] exit_reason Human-friendly description of event to set + */ +void +pcmk__set_result(pcmk__action_result_t *result, int exit_status, + enum pcmk_exec_status exec_status, const char *exit_reason) +{ + if (result == NULL) { + return; + } + + result->exit_status = exit_status; + result->execution_status = exec_status; + + if (!pcmk__str_eq(result->exit_reason, exit_reason, pcmk__str_none)) { + free(result->exit_reason); + result->exit_reason = (exit_reason == NULL)? NULL : strdup(exit_reason); + } +} + +/*! + * \internal + * \brief Set the output of an action + * + * \param[out] result Action result to set output for + * \param[in] out Action output to copy + * \param[in] err Action error output to copy + */ +void +pcmk__set_result_output(pcmk__action_result_t *result, + const char *out, const char *err) +{ + if (result == NULL) { + return; + } + + free(result->action_stdout); + result->action_stdout = (out == NULL)? NULL : strdup(out); + + free(result->action_stderr); + result->action_stderr = (err == NULL)? NULL : strdup(err); +} + +/*! + * \internal + * \brief Clear a result's exit reason, output, and error output + * + * \param[in] result Result to reset + */ +void +pcmk__reset_result(pcmk__action_result_t *result) +{ + if (result == NULL) { + return; + } + + free(result->exit_reason); + result->exit_reason = NULL; + + free(result->action_stdout); + result->action_stdout = NULL; + + free(result->action_stderr); + result->action_stderr = NULL; +} diff --git a/lib/lrmd/lrmd_client.c b/lib/lrmd/lrmd_client.c index 11a8a7b62c..eb42478c98 100644 --- a/lib/lrmd/lrmd_client.c +++ b/lib/lrmd/lrmd_client.c @@ -1,2347 +1,2393 @@ /* * Copyright 2012-2021 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 // uint32_t, uint64_t #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef HAVE_GNUTLS_GNUTLS_H # undef KEYFILE # include #endif #include #include #include #include #include #define MAX_TLS_RECV_WAIT 10000 CRM_TRACE_INIT_DATA(lrmd); static int lrmd_api_disconnect(lrmd_t * lrmd); static int lrmd_api_is_connected(lrmd_t * lrmd); /* IPC proxy functions */ int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg); static void lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg); void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)); #ifdef HAVE_GNUTLS_GNUTLS_H # define LRMD_CLIENT_HANDSHAKE_TIMEOUT 5000 /* 5 seconds */ gnutls_psk_client_credentials_t psk_cred_s; static void lrmd_tls_disconnect(lrmd_t * lrmd); static int global_remote_msg_id = 0; static void lrmd_tls_connection_destroy(gpointer userdata); #endif typedef struct lrmd_private_s { uint64_t type; char *token; mainloop_io_t *source; /* IPC parameters */ crm_ipc_t *ipc; pcmk__remote_t *remote; /* Extra TLS parameters */ char *remote_nodename; #ifdef HAVE_GNUTLS_GNUTLS_H char *server; int port; gnutls_psk_client_credentials_t psk_cred_c; /* while the async connection is occurring, this is the id * of the connection timeout timer. */ int async_timer; int sock; /* since tls requires a round trip across the network for a * request/reply, there are times where we just want to be able * to send a request from the client and not wait around (or even care * about) what the reply is. */ int expected_late_replies; GList *pending_notify; crm_trigger_t *process_notify; #endif lrmd_event_callback callback; /* Internal IPC proxy msg passing for remote guests */ void (*proxy_callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg); void *proxy_callback_userdata; char *peer_version; } lrmd_private_t; static lrmd_list_t * lrmd_list_add(lrmd_list_t * head, const char *value) { lrmd_list_t *p, *end; p = calloc(1, sizeof(lrmd_list_t)); p->val = strdup(value); end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void lrmd_list_freeall(lrmd_list_t * head) { lrmd_list_t *p; while (head) { char *val = (char *)head->val; p = head->next; free(val); free(head); head = p; } } lrmd_key_value_t * lrmd_key_value_add(lrmd_key_value_t * head, const char *key, const char *value) { lrmd_key_value_t *p, *end; p = calloc(1, sizeof(lrmd_key_value_t)); p->key = strdup(key); p->value = strdup(value); end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void lrmd_key_value_freeall(lrmd_key_value_t * head) { lrmd_key_value_t *p; while (head) { p = head->next; free(head->key); free(head->value); free(head); head = p; } } /*! * Create a new lrmd_event_data_t object * * \param[in] rsc_id ID of resource involved in event * \param[in] task Action name * \param[in] interval_ms Action interval * * \return Newly allocated and initialized lrmd_event_data_t * \note This functions asserts on memory errors, so the return value is * guaranteed to be non-NULL. The caller is responsible for freeing the * result with lrmd_free_event(). */ lrmd_event_data_t * lrmd_new_event(const char *rsc_id, const char *task, guint interval_ms) { lrmd_event_data_t *event = calloc(1, sizeof(lrmd_event_data_t)); CRM_ASSERT(event != NULL); if (rsc_id != NULL) { event->rsc_id = strdup(rsc_id); CRM_ASSERT(event->rsc_id != NULL); } if (task != NULL) { event->op_type = strdup(task); CRM_ASSERT(event->op_type != NULL); } event->interval_ms = interval_ms; return event; } lrmd_event_data_t * lrmd_copy_event(lrmd_event_data_t * event) { lrmd_event_data_t *copy = NULL; copy = calloc(1, sizeof(lrmd_event_data_t)); /* This will get all the int values. * we just have to be careful not to leave any * dangling pointers to strings. */ memcpy(copy, event, sizeof(lrmd_event_data_t)); copy->rsc_id = event->rsc_id ? strdup(event->rsc_id) : NULL; copy->op_type = event->op_type ? strdup(event->op_type) : NULL; copy->user_data = event->user_data ? strdup(event->user_data) : NULL; copy->output = event->output ? strdup(event->output) : NULL; copy->exit_reason = event->exit_reason ? strdup(event->exit_reason) : NULL; copy->remote_nodename = event->remote_nodename ? strdup(event->remote_nodename) : NULL; copy->params = pcmk__str_table_dup(event->params); return copy; } /*! * \brief Free an executor event * * \param[in] Executor event object to free */ void lrmd_free_event(lrmd_event_data_t *event) { if (event == NULL) { return; } // @TODO Why are these const char *? free((void *) event->rsc_id); free((void *) event->op_type); free((void *) event->user_data); - free((void *) event->output); - free((void *) event->exit_reason); free((void *) event->remote_nodename); + lrmd__reset_result(event); if (event->params != NULL) { g_hash_table_destroy(event->params); } free(event); } static void lrmd_dispatch_internal(lrmd_t * lrmd, xmlNode * msg) { const char *type; const char *proxy_session = crm_element_value(msg, F_LRMD_IPC_SESSION); lrmd_private_t *native = lrmd->lrmd_private; lrmd_event_data_t event = { 0, }; if (proxy_session != NULL) { /* this is proxy business */ lrmd_internal_proxy_dispatch(lrmd, msg); return; } else if (!native->callback) { /* no callback set */ crm_trace("notify event received but client has not set callback"); return; } event.remote_nodename = native->remote_nodename; type = crm_element_value(msg, F_LRMD_OPERATION); crm_element_value_int(msg, F_LRMD_CALLID, &event.call_id); event.rsc_id = crm_element_value(msg, F_LRMD_RSC_ID); if (pcmk__str_eq(type, LRMD_OP_RSC_REG, pcmk__str_none)) { event.type = lrmd_event_register; } else if (pcmk__str_eq(type, LRMD_OP_RSC_UNREG, pcmk__str_none)) { event.type = lrmd_event_unregister; } else if (pcmk__str_eq(type, LRMD_OP_RSC_EXEC, pcmk__str_none)) { time_t epoch = 0; crm_element_value_int(msg, F_LRMD_TIMEOUT, &event.timeout); crm_element_value_ms(msg, F_LRMD_RSC_INTERVAL, &event.interval_ms); crm_element_value_int(msg, F_LRMD_RSC_START_DELAY, &event.start_delay); crm_element_value_int(msg, F_LRMD_EXEC_RC, (int *)&event.rc); crm_element_value_int(msg, F_LRMD_OP_STATUS, &event.op_status); crm_element_value_int(msg, F_LRMD_RSC_DELETED, &event.rsc_deleted); crm_element_value_epoch(msg, F_LRMD_RSC_RUN_TIME, &epoch); event.t_run = (unsigned int) epoch; crm_element_value_epoch(msg, F_LRMD_RSC_RCCHANGE_TIME, &epoch); event.t_rcchange = (unsigned int) epoch; crm_element_value_int(msg, F_LRMD_RSC_EXEC_TIME, (int *)&event.exec_time); crm_element_value_int(msg, F_LRMD_RSC_QUEUE_TIME, (int *)&event.queue_time); event.op_type = crm_element_value(msg, F_LRMD_RSC_ACTION); event.user_data = crm_element_value(msg, F_LRMD_RSC_USERDATA_STR); + event.type = lrmd_event_exec_complete; + + // No need to duplicate the memory, so don't use setter functions event.output = crm_element_value(msg, F_LRMD_RSC_OUTPUT); event.exit_reason = crm_element_value(msg, F_LRMD_RSC_EXIT_REASON); - event.type = lrmd_event_exec_complete; event.params = xml2list(msg); } else if (pcmk__str_eq(type, LRMD_OP_NEW_CLIENT, pcmk__str_none)) { event.type = lrmd_event_new_client; } else if (pcmk__str_eq(type, LRMD_OP_POKE, pcmk__str_none)) { event.type = lrmd_event_poke; } else { return; } crm_trace("op %s notify event received", type); native->callback(&event); if (event.params) { g_hash_table_destroy(event.params); } } // \return Always 0, to indicate that IPC mainloop source should be kept static int lrmd_ipc_dispatch(const char *buffer, ssize_t length, gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; if (native->callback != NULL) { xmlNode *msg = string2xml(buffer); lrmd_dispatch_internal(lrmd, msg); free_xml(msg); } return 0; } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_free_xml(gpointer userdata) { free_xml((xmlNode *) userdata); } static bool remote_executor_connected(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; return (native->remote->tls_session != NULL); } /*! * \internal * \brief TLS dispatch function (for both trigger and file descriptor sources) * * \param[in] userdata API connection * * \return Always return a nonnegative value, which as a file descriptor * dispatch function means keep the mainloop source, and as a * trigger dispatch function, 0 means remove the trigger from the * mainloop while 1 means keep it (and job completed) */ static int lrmd_tls_dispatch(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *xml = NULL; int rc = pcmk_rc_ok; if (!remote_executor_connected(lrmd)) { crm_trace("TLS dispatch triggered after disconnect"); return 0; } crm_trace("TLS dispatch triggered"); /* First check if there are any pending notifies to process that came * while we were waiting for replies earlier. */ if (native->pending_notify) { GList *iter = NULL; crm_trace("Processing pending notifies"); for (iter = native->pending_notify; iter; iter = iter->next) { lrmd_dispatch_internal(lrmd, iter->data); } g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } /* Next read the current buffer and see if there are any messages to handle. */ switch (pcmk__remote_ready(native->remote, 0)) { case pcmk_rc_ok: rc = pcmk__read_remote_message(native->remote, -1); xml = pcmk__remote_message_xml(native->remote); break; case ETIME: // Nothing to read, check if a full message is already in buffer xml = pcmk__remote_message_xml(native->remote); break; default: rc = ENOTCONN; break; } while (xml) { const char *msg_type = crm_element_value(xml, F_LRMD_REMOTE_MSG_TYPE); if (pcmk__str_eq(msg_type, "notify", pcmk__str_casei)) { lrmd_dispatch_internal(lrmd, xml); } else if (pcmk__str_eq(msg_type, "reply", pcmk__str_casei)) { if (native->expected_late_replies > 0) { native->expected_late_replies--; } else { int reply_id = 0; crm_element_value_int(xml, F_LRMD_CALLID, &reply_id); /* if this happens, we want to know about it */ crm_err("Got outdated Pacemaker Remote reply %d", reply_id); } } free_xml(xml); xml = pcmk__remote_message_xml(native->remote); } if (rc == ENOTCONN) { crm_info("Lost %s executor connection while reading data", (native->remote_nodename? native->remote_nodename : "local")); lrmd_tls_disconnect(lrmd); return 0; } return 1; } #endif /* Not used with mainloop */ int lrmd_poll(lrmd_t * lrmd, int timeout) { lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: return crm_ipc_ready(native->ipc); #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: if (native->pending_notify) { return 1; } else { int rc = pcmk__remote_ready(native->remote, 0); switch (rc) { case pcmk_rc_ok: return 1; case ETIME: return 0; default: return pcmk_rc2legacy(rc); } } #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); return -EPROTONOSUPPORT; } } /* Not used with mainloop */ bool lrmd_dispatch(lrmd_t * lrmd) { lrmd_private_t *private = NULL; CRM_ASSERT(lrmd != NULL); private = lrmd->lrmd_private; switch (private->type) { case pcmk__client_ipc: while (crm_ipc_ready(private->ipc)) { if (crm_ipc_read(private->ipc) > 0) { const char *msg = crm_ipc_buffer(private->ipc); lrmd_ipc_dispatch(msg, strlen(msg), lrmd); } } break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: lrmd_tls_dispatch(lrmd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", private->type); } if (lrmd_api_is_connected(lrmd) == FALSE) { crm_err("Connection closed"); return FALSE; } return TRUE; } static xmlNode * lrmd_create_op(const char *token, const char *op, xmlNode *data, int timeout, enum lrmd_call_options options) { xmlNode *op_msg = create_xml_node(NULL, "lrmd_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "lrmd_command"); crm_xml_add(op_msg, F_TYPE, T_LRMD); crm_xml_add(op_msg, F_LRMD_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_LRMD_OPERATION, op); crm_xml_add_int(op_msg, F_LRMD_TIMEOUT, timeout); crm_xml_add_int(op_msg, F_LRMD_CALLOPTS, options); if (data != NULL) { add_message_xml(op_msg, F_LRMD_CALLDATA, data); } crm_trace("Created executor %s command with call options %.8lx (%d)", op, (long)options, options); return op_msg; } static void lrmd_ipc_connection_destroy(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; crm_info("IPC connection destroyed"); /* Prevent these from being cleaned up in lrmd_api_disconnect() */ native->ipc = NULL; native->source = NULL; if (native->callback) { lrmd_event_data_t event = { 0, }; event.type = lrmd_event_disconnect; event.remote_nodename = native->remote_nodename; native->callback(&event); } } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_tls_connection_destroy(gpointer userdata) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; crm_info("TLS connection destroyed"); if (native->remote->tls_session) { gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); } if (native->psk_cred_c) { gnutls_psk_free_client_credentials(native->psk_cred_c); } if (native->sock) { close(native->sock); } if (native->process_notify) { mainloop_destroy_trigger(native->process_notify); native->process_notify = NULL; } if (native->pending_notify) { g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } free(native->remote->buffer); native->remote->buffer = NULL; native->source = 0; native->sock = 0; native->psk_cred_c = NULL; native->remote->tls_session = NULL; native->sock = 0; if (native->callback) { lrmd_event_data_t event = { 0, }; event.remote_nodename = native->remote_nodename; event.type = lrmd_event_disconnect; native->callback(&event); } return; } // \return Standard Pacemaker return code int lrmd__remote_send_xml(pcmk__remote_t *session, xmlNode *msg, uint32_t id, const char *msg_type) { crm_xml_add_int(msg, F_LRMD_REMOTE_MSG_ID, id); crm_xml_add(msg, F_LRMD_REMOTE_MSG_TYPE, msg_type); return pcmk__remote_send_xml(session, msg); } // \return Standard Pacemaker return code static int read_remote_reply(lrmd_t *lrmd, int total_timeout, int expected_reply_id, xmlNode **reply) { lrmd_private_t *native = lrmd->lrmd_private; time_t start = time(NULL); const char *msg_type = NULL; int reply_id = 0; int remaining_timeout = 0; int rc = pcmk_rc_ok; /* A timeout of 0 here makes no sense. We have to wait a period of time * for the response to come back. If -1 or 0, default to 10 seconds. */ if (total_timeout <= 0 || total_timeout > MAX_TLS_RECV_WAIT) { total_timeout = MAX_TLS_RECV_WAIT; } for (*reply = NULL; *reply == NULL; ) { *reply = pcmk__remote_message_xml(native->remote); if (*reply == NULL) { /* read some more off the tls buffer if we still have time left. */ if (remaining_timeout) { remaining_timeout = total_timeout - ((time(NULL) - start) * 1000); } else { remaining_timeout = total_timeout; } if (remaining_timeout <= 0) { return ETIME; } rc = pcmk__read_remote_message(native->remote, remaining_timeout); if (rc != pcmk_rc_ok) { return rc; } *reply = pcmk__remote_message_xml(native->remote); if (*reply == NULL) { return ENOMSG; } } crm_element_value_int(*reply, F_LRMD_REMOTE_MSG_ID, &reply_id); msg_type = crm_element_value(*reply, F_LRMD_REMOTE_MSG_TYPE); if (!msg_type) { crm_err("Empty msg type received while waiting for reply"); free_xml(*reply); *reply = NULL; } else if (pcmk__str_eq(msg_type, "notify", pcmk__str_casei)) { /* got a notify while waiting for reply, trigger the notify to be processed later */ crm_info("queueing notify"); native->pending_notify = g_list_append(native->pending_notify, *reply); if (native->process_notify) { crm_info("notify trigger set."); mainloop_set_trigger(native->process_notify); } *reply = NULL; } else if (!pcmk__str_eq(msg_type, "reply", pcmk__str_casei)) { /* msg isn't a reply, make some noise */ crm_err("Expected a reply, got %s", msg_type); free_xml(*reply); *reply = NULL; } else if (reply_id != expected_reply_id) { if (native->expected_late_replies > 0) { native->expected_late_replies--; } else { crm_err("Got outdated reply, expected id %d got id %d", expected_reply_id, reply_id); } free_xml(*reply); *reply = NULL; } } if (native->remote->buffer && native->process_notify) { mainloop_set_trigger(native->process_notify); } return rc; } // \return Standard Pacemaker return code static int send_remote_message(lrmd_t *lrmd, xmlNode *msg) { int rc = pcmk_rc_ok; lrmd_private_t *native = lrmd->lrmd_private; global_remote_msg_id++; if (global_remote_msg_id <= 0) { global_remote_msg_id = 1; } rc = lrmd__remote_send_xml(native->remote, msg, global_remote_msg_id, "request"); if (rc != pcmk_rc_ok) { crm_err("Disconnecting because TLS message could not be sent to " "Pacemaker Remote: %s", pcmk_rc_str(rc)); lrmd_tls_disconnect(lrmd); } return rc; } static int lrmd_tls_send_recv(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply) { int rc = 0; xmlNode *xml = NULL; if (!remote_executor_connected(lrmd)) { return -ENOTCONN; } rc = send_remote_message(lrmd, msg); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } rc = read_remote_reply(lrmd, timeout, global_remote_msg_id, &xml); if (rc != pcmk_rc_ok) { crm_err("Disconnecting remote after request %d reply not received: %s " CRM_XS " rc=%d timeout=%dms", global_remote_msg_id, pcmk_rc_str(rc), rc, timeout); lrmd_tls_disconnect(lrmd); } if (reply) { *reply = xml; } else { free_xml(xml); } return pcmk_rc2legacy(rc); } #endif static int lrmd_send_xml(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = crm_ipc_send(native->ipc, msg, crm_ipc_client_response, timeout, reply); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_send_recv(lrmd, msg, timeout, reply); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static int lrmd_send_xml_no_reply(lrmd_t * lrmd, xmlNode * msg) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = crm_ipc_send(native->ipc, msg, crm_ipc_flags_none, 0, NULL); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = send_remote_message(lrmd, msg); if (rc == pcmk_rc_ok) { /* we don't want to wait around for the reply, but * since the request/reply protocol needs to behave the same * as libqb, a reply will eventually come later anyway. */ native->expected_late_replies++; } rc = pcmk_rc2legacy(rc); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static int lrmd_api_is_connected(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: return crm_ipc_connected(native->ipc); #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: return remote_executor_connected(lrmd); #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); return 0; } } /*! * \internal * \brief Send a prepared API command to the executor * * \param[in] lrmd Existing connection to the executor * \param[in] op Name of API command to send * \param[in] data Command data XML to add to the sent command * \param[out] output_data If expecting a reply, it will be stored here * \param[in] timeout Timeout in milliseconds (if 0, defaults to * a sensible value per the type of connection, * standard vs. pacemaker remote); * also propagated to the command XML * \param[in] call_options Call options to pass to server when sending * \param[in] expect_reply If TRUE, wait for a reply from the server; * must be TRUE for IPC (as opposed to TLS) clients * * \return pcmk_ok on success, -errno on error */ static int lrmd_send_command(lrmd_t *lrmd, const char *op, xmlNode *data, xmlNode **output_data, int timeout, enum lrmd_call_options options, gboolean expect_reply) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; if (!lrmd_api_is_connected(lrmd)) { return -ENOTCONN; } if (op == NULL) { crm_err("No operation specified"); return -EINVAL; } CRM_CHECK(native->token != NULL,; ); crm_trace("Sending %s op to executor", op); op_msg = lrmd_create_op(native->token, op, data, timeout, options); if (op_msg == NULL) { return -EINVAL; } if (expect_reply) { rc = lrmd_send_xml(lrmd, op_msg, timeout, &op_reply); } else { rc = lrmd_send_xml_no_reply(lrmd, op_msg); goto done; } if (rc < 0) { crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%d): %d", op, timeout, rc); - rc = -ECOMM; goto done; } else if(op_reply == NULL) { rc = -ENOMSG; goto done; } rc = pcmk_ok; crm_trace("%s op reply received", op); if (crm_element_value_int(op_reply, F_LRMD_RC, &rc) != 0) { rc = -ENOMSG; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (output_data) { *output_data = op_reply; op_reply = NULL; /* Prevent subsequent free */ } done: if (lrmd_api_is_connected(lrmd) == FALSE) { crm_err("Executor disconnected"); } free_xml(op_msg); free_xml(op_reply); return rc; } static int lrmd_api_poke_connection(lrmd_t * lrmd) { int rc; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); rc = lrmd_send_command(lrmd, LRMD_OP_POKE, data, NULL, 0, 0, (native->type == pcmk__client_ipc)); free_xml(data); return rc < 0 ? rc : pcmk_ok; } // \return Standard Pacemaker return code int lrmd__validate_remote_settings(lrmd_t *lrmd, GHashTable *hash) { int rc = pcmk_rc_ok; const char *value; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *data = create_xml_node(NULL, F_LRMD_OPERATION); crm_xml_add(data, F_LRMD_ORIGIN, __func__); value = g_hash_table_lookup(hash, "stonith-watchdog-timeout"); if ((value) && (stonith__watchdog_fencing_enabled_for_node(native->remote_nodename))) { crm_xml_add(data, F_LRMD_WATCHDOG, value); } rc = lrmd_send_command(lrmd, LRMD_OP_CHECK, data, NULL, 0, 0, (native->type == pcmk__client_ipc)); free_xml(data); return (rc < 0)? pcmk_legacy2rc(rc) : pcmk_rc_ok; } static int lrmd_handshake(lrmd_t * lrmd, const char *name) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "lrmd_command"); crm_xml_add(hello, F_TYPE, T_LRMD); crm_xml_add(hello, F_LRMD_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_LRMD_CLIENTNAME, name); crm_xml_add(hello, F_LRMD_PROTOCOL_VERSION, LRMD_PROTOCOL_VERSION); /* advertise that we are a proxy provider */ if (native->proxy_callback) { crm_xml_add(hello, F_LRMD_IS_IPC_PROVIDER, "true"); } rc = lrmd_send_xml(lrmd, hello, -1, &reply); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't complete registration with the executor API: %d", rc); rc = -ECOMM; } else if (reply == NULL) { crm_err("Did not receive registration reply"); rc = -EPROTO; } else { const char *version = crm_element_value(reply, F_LRMD_PROTOCOL_VERSION); const char *msg_type = crm_element_value(reply, F_LRMD_OPERATION); const char *tmp_ticket = crm_element_value(reply, F_LRMD_CLIENTID); crm_element_value_int(reply, F_LRMD_RC, &rc); if (rc == -EPROTO) { crm_err("Executor protocol version mismatch between client (%s) and server (%s)", LRMD_PROTOCOL_VERSION, version); crm_log_xml_err(reply, "Protocol Error"); } else if (!pcmk__str_eq(msg_type, CRM_OP_REGISTER, pcmk__str_casei)) { crm_err("Invalid registration message: %s", msg_type); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else if (tmp_ticket == NULL) { crm_err("No registration token provided"); crm_log_xml_err(reply, "Bad reply"); rc = -EPROTO; } else { crm_trace("Obtained registration token: %s", tmp_ticket); native->token = strdup(tmp_ticket); native->peer_version = strdup(version?version:"1.0"); /* Included since 1.1 */ rc = pcmk_ok; } } free_xml(reply); free_xml(hello); if (rc != pcmk_ok) { lrmd_api_disconnect(lrmd); } return rc; } static int lrmd_ipc_connect(lrmd_t * lrmd, int *fd) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; struct ipc_client_callbacks lrmd_callbacks = { .dispatch = lrmd_ipc_dispatch, .destroy = lrmd_ipc_connection_destroy }; crm_info("Connecting to executor"); if (fd) { /* No mainloop */ native->ipc = crm_ipc_new(CRM_SYSTEM_LRMD, 0); if (native->ipc && crm_ipc_connect(native->ipc)) { *fd = crm_ipc_get_fd(native->ipc); } else if (native->ipc) { crm_perror(LOG_ERR, "Connection to executor failed"); rc = -ENOTCONN; } } else { native->source = mainloop_add_ipc_client(CRM_SYSTEM_LRMD, G_PRIORITY_HIGH, 0, lrmd, &lrmd_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { crm_debug("Could not connect to the executor API"); rc = -ENOTCONN; } return rc; } #ifdef HAVE_GNUTLS_GNUTLS_H static void copy_gnutls_datum(gnutls_datum_t *dest, gnutls_datum_t *source) { dest->data = gnutls_malloc(source->size); CRM_ASSERT(dest->data); memcpy(dest->data, source->data, source->size); dest->size = source->size; } static void clear_gnutls_datum(gnutls_datum_t *datum) { gnutls_free(datum->data); datum->data = NULL; datum->size = 0; } #define KEY_READ_LEN 256 // Chunk size for reading key from file // \return Standard Pacemaker return code static int read_gnutls_key(const char *location, gnutls_datum_t *key) { FILE *stream = NULL; size_t buf_len = KEY_READ_LEN; if ((location == NULL) || (key == NULL)) { return EINVAL; } stream = fopen(location, "r"); if (stream == NULL) { return errno; } key->data = gnutls_malloc(buf_len); key->size = 0; while (!feof(stream)) { int next = fgetc(stream); if (next == EOF) { if (!feof(stream)) { crm_warn("Pacemaker Remote key read was partially successful " "(copy in memory may be corrupted)"); } break; } if (key->size == buf_len) { buf_len = key->size + KEY_READ_LEN; key->data = gnutls_realloc(key->data, buf_len); CRM_ASSERT(key->data); } key->data[key->size++] = (unsigned char) next; } fclose(stream); if (key->size == 0) { clear_gnutls_datum(key); return ENOKEY; } return pcmk_rc_ok; } // Cache the most recently used Pacemaker Remote authentication key struct key_cache_s { time_t updated; // When cached key was read (valid for 1 minute) const char *location; // Where cached key was read from gnutls_datum_t key; // Cached key }; static bool key_is_cached(struct key_cache_s *key_cache) { return key_cache->updated != 0; } static bool key_cache_expired(struct key_cache_s *key_cache) { return (time(NULL) - key_cache->updated) >= 60; } static void clear_key_cache(struct key_cache_s *key_cache) { clear_gnutls_datum(&(key_cache->key)); if ((key_cache->updated != 0) || (key_cache->location != NULL)) { key_cache->updated = 0; key_cache->location = NULL; crm_debug("Cleared Pacemaker Remote key cache"); } } static void get_cached_key(struct key_cache_s *key_cache, gnutls_datum_t *key) { copy_gnutls_datum(key, &(key_cache->key)); crm_debug("Using cached Pacemaker Remote key from %s", crm_str(key_cache->location)); } static void cache_key(struct key_cache_s *key_cache, gnutls_datum_t *key, const char *location) { key_cache->updated = time(NULL); key_cache->location = location; copy_gnutls_datum(&(key_cache->key), key); crm_debug("Using (and cacheing) Pacemaker Remote key from %s", crm_str(location)); } /*! * \internal * \brief Get Pacemaker Remote authentication key from file or cache * * \param[in] location Path to key file to try (this memory must * persist across all calls of this function) * \param[out] key Key from location or cache * * \return Standard Pacemaker return code */ static int get_remote_key(const char *location, gnutls_datum_t *key) { static struct key_cache_s key_cache = { 0, }; int rc = pcmk_rc_ok; if ((location == NULL) || (key == NULL)) { return EINVAL; } if (key_is_cached(&key_cache)) { if (key_cache_expired(&key_cache)) { clear_key_cache(&key_cache); } else { get_cached_key(&key_cache, key); return pcmk_rc_ok; } } rc = read_gnutls_key(location, key); if (rc != pcmk_rc_ok) { return rc; } cache_key(&key_cache, key, location); return pcmk_rc_ok; } /*! * \internal * \brief Initialize the Pacemaker Remote authentication key * * Try loading the Pacemaker Remote authentication key from cache if available, * otherwise from these locations, in order of preference: the value of the * PCMK_authkey_location environment variable, if set; the Pacemaker default key * file location; or (for historical reasons) /etc/corosync/authkey. * * \param[out] key Where to store key * * \return Standard Pacemaker return code */ int lrmd__init_remote_key(gnutls_datum_t *key) { static const char *env_location = NULL; static bool need_env = true; int env_rc = pcmk_rc_ok; int default_rc = pcmk_rc_ok; int alt_rc = pcmk_rc_ok; bool env_is_default = false; bool env_is_fallback = false; if (need_env) { env_location = getenv("PCMK_authkey_location"); need_env = false; } // Try location in environment variable, if set if (env_location != NULL) { env_rc = get_remote_key(env_location, key); if (env_rc == pcmk_rc_ok) { return pcmk_rc_ok; } env_is_default = !strcmp(env_location, DEFAULT_REMOTE_KEY_LOCATION); env_is_fallback = !strcmp(env_location, ALT_REMOTE_KEY_LOCATION); /* @TODO It would be more secure to fail, rather than fall back to the * default, if an explicitly set key location is not readable, and it * would be better to never use the Corosync location as a fallback. * However, that would break any deployments currently working with the * fallbacks. */ } // Try default location, if environment wasn't explicitly set to it if (env_is_default) { default_rc = env_rc; } else { default_rc = get_remote_key(DEFAULT_REMOTE_KEY_LOCATION, key); } // Try fallback location, if environment wasn't set to it and default failed if (env_is_fallback) { alt_rc = env_rc; } else if (default_rc != pcmk_rc_ok) { alt_rc = get_remote_key(ALT_REMOTE_KEY_LOCATION, key); } // We have all results, so log and return if ((env_rc != pcmk_rc_ok) && (default_rc != pcmk_rc_ok) && (alt_rc != pcmk_rc_ok)) { // Environment set, everything failed crm_warn("Could not read Pacemaker Remote key from %s (%s%s%s%s%s): %s", env_location, env_is_default? "" : "or default location ", env_is_default? "" : DEFAULT_REMOTE_KEY_LOCATION, !env_is_default && !env_is_fallback? " " : "", env_is_fallback? "" : "or fallback location ", env_is_fallback? "" : ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(env_rc)); return ENOKEY; } if (env_rc != pcmk_rc_ok) { // Environment set but failed, using a default crm_warn("Could not read Pacemaker Remote key from %s " "(using %s location %s instead): %s", env_location, (default_rc == pcmk_rc_ok)? "default" : "fallback", (default_rc == pcmk_rc_ok)? DEFAULT_REMOTE_KEY_LOCATION : ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(env_rc)); return pcmk_rc_ok; } if ((default_rc != pcmk_rc_ok) && (alt_rc != pcmk_rc_ok)) { // Environment unset, defaults failed crm_warn("Could not read Pacemaker Remote key from default location %s" " (or fallback location %s): %s", DEFAULT_REMOTE_KEY_LOCATION, ALT_REMOTE_KEY_LOCATION, pcmk_rc_str(default_rc)); return ENOKEY; } return pcmk_rc_ok; // Environment variable unset, a default worked } static void lrmd_gnutls_global_init(void) { static int gnutls_init = 0; if (!gnutls_init) { crm_gnutls_global_init(); } gnutls_init = 1; } #endif static void report_async_connection_result(lrmd_t * lrmd, int rc) { lrmd_private_t *native = lrmd->lrmd_private; if (native->callback) { lrmd_event_data_t event = { 0, }; event.type = lrmd_event_connect; event.remote_nodename = native->remote_nodename; event.connection_rc = rc; native->callback(&event); } } #ifdef HAVE_GNUTLS_GNUTLS_H static inline int lrmd__tls_client_handshake(pcmk__remote_t *remote) { return pcmk__tls_client_handshake(remote, LRMD_CLIENT_HANDSHAKE_TIMEOUT); } /*! * \internal * \brief Add trigger and file descriptor mainloop sources for TLS * * \param[in] lrmd API connection with established TLS session * \param[in] do_handshake Whether to perform executor handshake * * \return Standard Pacemaker return code */ static int add_tls_to_mainloop(lrmd_t *lrmd, bool do_handshake) { lrmd_private_t *native = lrmd->lrmd_private; int rc = pcmk_rc_ok; char *name = crm_strdup_printf("pacemaker-remote-%s:%d", native->server, native->port); struct mainloop_fd_callbacks tls_fd_callbacks = { .dispatch = lrmd_tls_dispatch, .destroy = lrmd_tls_connection_destroy, }; native->process_notify = mainloop_add_trigger(G_PRIORITY_HIGH, lrmd_tls_dispatch, lrmd); native->source = mainloop_add_fd(name, G_PRIORITY_HIGH, native->sock, lrmd, &tls_fd_callbacks); /* Async connections lose the client name provided by the API caller, so we * have to use our generated name here to perform the executor handshake. * * @TODO Keep track of the caller-provided name. Perhaps we should be using * that name in this function instead of generating one anyway. */ if (do_handshake) { rc = lrmd_handshake(lrmd, name); rc = pcmk_legacy2rc(rc); } free(name); return rc; } static void lrmd_tcp_connect_cb(void *userdata, int rc, int sock) { lrmd_t *lrmd = userdata; lrmd_private_t *native = lrmd->lrmd_private; gnutls_datum_t psk_key = { NULL, 0 }; native->async_timer = 0; if (rc != pcmk_rc_ok) { lrmd_tls_connection_destroy(lrmd); crm_info("Could not connect to Pacemaker Remote at %s:%d: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); return; } /* The TCP connection was successful, so establish the TLS connection. * @TODO make this async to avoid blocking code in client */ native->sock = sock; rc = lrmd__init_remote_key(&psk_key); if (rc != pcmk_rc_ok) { crm_info("Could not connect to Pacemaker Remote at %s:%d: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); return; } gnutls_psk_allocate_client_credentials(&native->psk_cred_c); gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW); gnutls_free(psk_key.data); native->remote->tls_session = pcmk__new_tls_session(sock, GNUTLS_CLIENT, GNUTLS_CRD_PSK, native->psk_cred_c); if (native->remote->tls_session == NULL) { lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, -EPROTO); return; } if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) { crm_warn("Disconnecting after TLS handshake with Pacemaker Remote server %s:%d failed", native->server, native->port); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = NULL; lrmd_tls_connection_destroy(lrmd); report_async_connection_result(lrmd, -EKEYREJECTED); return; } crm_info("TLS connection to Pacemaker Remote server %s:%d succeeded", native->server, native->port); rc = add_tls_to_mainloop(lrmd, true); report_async_connection_result(lrmd, pcmk_rc2legacy(rc)); } static int lrmd_tls_connect_async(lrmd_t * lrmd, int timeout /*ms */ ) { int rc; int timer_id = 0; lrmd_private_t *native = lrmd->lrmd_private; lrmd_gnutls_global_init(); native->sock = -1; rc = pcmk__connect_remote(native->server, native->port, timeout, &timer_id, &(native->sock), lrmd, lrmd_tcp_connect_cb); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%d failed: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); return pcmk_rc2legacy(rc); } native->async_timer = timer_id; return pcmk_ok; } static int lrmd_tls_connect(lrmd_t * lrmd, int *fd) { int rc; lrmd_private_t *native = lrmd->lrmd_private; gnutls_datum_t psk_key = { NULL, 0 }; lrmd_gnutls_global_init(); native->sock = -1; rc = pcmk__connect_remote(native->server, native->port, 0, NULL, &(native->sock), NULL, NULL); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%d failed: %s " CRM_XS " rc=%d", native->server, native->port, pcmk_rc_str(rc), rc); lrmd_tls_connection_destroy(lrmd); return -ENOTCONN; } rc = lrmd__init_remote_key(&psk_key); if (rc != pcmk_rc_ok) { lrmd_tls_connection_destroy(lrmd); return pcmk_rc2legacy(rc); } gnutls_psk_allocate_client_credentials(&native->psk_cred_c); gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW); gnutls_free(psk_key.data); native->remote->tls_session = pcmk__new_tls_session(native->sock, GNUTLS_CLIENT, GNUTLS_CRD_PSK, native->psk_cred_c); if (native->remote->tls_session == NULL) { lrmd_tls_connection_destroy(lrmd); return -EPROTO; } if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) { crm_err("Session creation for %s:%d failed", native->server, native->port); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = NULL; lrmd_tls_connection_destroy(lrmd); return -EKEYREJECTED; } crm_info("Client TLS connection established with Pacemaker Remote server %s:%d", native->server, native->port); if (fd) { *fd = native->sock; } else { add_tls_to_mainloop(lrmd, false); } return pcmk_ok; } #endif static int lrmd_api_connect(lrmd_t * lrmd, const char *name, int *fd) { int rc = -ENOTCONN; lrmd_private_t *native = lrmd->lrmd_private; switch (native->type) { case pcmk__client_ipc: rc = lrmd_ipc_connect(lrmd, fd); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_connect(lrmd, fd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } if (rc == pcmk_ok) { rc = lrmd_handshake(lrmd, name); } return rc; } static int lrmd_api_connect_async(lrmd_t * lrmd, const char *name, int timeout) { int rc = pcmk_ok; lrmd_private_t *native = lrmd->lrmd_private; CRM_CHECK(native && native->callback, return -EINVAL); switch (native->type) { case pcmk__client_ipc: /* fake async connection with ipc. it should be fast * enough that we gain very little from async */ rc = lrmd_api_connect(lrmd, name, NULL); if (!rc) { report_async_connection_result(lrmd, rc); } break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: rc = lrmd_tls_connect_async(lrmd, timeout); if (rc) { /* connection failed, report rc now */ report_async_connection_result(lrmd, rc); } break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } return rc; } static void lrmd_ipc_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } } #ifdef HAVE_GNUTLS_GNUTLS_H static void lrmd_tls_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; if (native->remote->tls_session) { gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR); gnutls_deinit(*native->remote->tls_session); gnutls_free(native->remote->tls_session); native->remote->tls_session = 0; } if (native->async_timer) { g_source_remove(native->async_timer); native->async_timer = 0; } if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; } else if (native->sock) { close(native->sock); native->sock = 0; } if (native->pending_notify) { g_list_free_full(native->pending_notify, lrmd_free_xml); native->pending_notify = NULL; } } #endif static int lrmd_api_disconnect(lrmd_t * lrmd) { lrmd_private_t *native = lrmd->lrmd_private; int rc = pcmk_ok; crm_info("Disconnecting %s %s executor connection", pcmk__client_type_str(native->type), (native->remote_nodename? native->remote_nodename : "local")); switch (native->type) { case pcmk__client_ipc: lrmd_ipc_disconnect(lrmd); break; #ifdef HAVE_GNUTLS_GNUTLS_H case pcmk__client_tls: lrmd_tls_disconnect(lrmd); break; #endif default: crm_err("Unsupported executor connection type (bug?): %d", native->type); rc = -EPROTONOSUPPORT; } free(native->token); native->token = NULL; free(native->peer_version); native->peer_version = NULL; return rc; } static int lrmd_api_register_rsc(lrmd_t * lrmd, const char *rsc_id, const char *class, const char *provider, const char *type, enum lrmd_call_options options) { int rc = pcmk_ok; xmlNode *data = NULL; if (!class || !type || !rsc_id) { return -EINVAL; } if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) && (provider == NULL)) { return -EINVAL; } data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add(data, F_LRMD_CLASS, class); crm_xml_add(data, F_LRMD_PROVIDER, provider); crm_xml_add(data, F_LRMD_TYPE, type); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_REG, data, NULL, 0, options, TRUE); free_xml(data); return rc; } static int lrmd_api_unregister_rsc(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_UNREG, data, NULL, 0, options, TRUE); free_xml(data); return rc; } lrmd_rsc_info_t * lrmd_new_rsc_info(const char *rsc_id, const char *standard, const char *provider, const char *type) { lrmd_rsc_info_t *rsc_info = calloc(1, sizeof(lrmd_rsc_info_t)); CRM_ASSERT(rsc_info); if (rsc_id) { rsc_info->id = strdup(rsc_id); CRM_ASSERT(rsc_info->id); } if (standard) { rsc_info->standard = strdup(standard); CRM_ASSERT(rsc_info->standard); } if (provider) { rsc_info->provider = strdup(provider); CRM_ASSERT(rsc_info->provider); } if (type) { rsc_info->type = strdup(type); CRM_ASSERT(rsc_info->type); } return rsc_info; } lrmd_rsc_info_t * lrmd_copy_rsc_info(lrmd_rsc_info_t * rsc_info) { return lrmd_new_rsc_info(rsc_info->id, rsc_info->standard, rsc_info->provider, rsc_info->type); } void lrmd_free_rsc_info(lrmd_rsc_info_t * rsc_info) { if (!rsc_info) { return; } free(rsc_info->id); free(rsc_info->type); free(rsc_info->standard); free(rsc_info->provider); free(rsc_info); } static lrmd_rsc_info_t * lrmd_api_get_rsc_info(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options) { lrmd_rsc_info_t *rsc_info = NULL; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); xmlNode *output = NULL; const char *class = NULL; const char *provider = NULL; const char *type = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); lrmd_send_command(lrmd, LRMD_OP_RSC_INFO, data, &output, 0, options, TRUE); free_xml(data); if (!output) { return NULL; } class = crm_element_value(output, F_LRMD_CLASS); provider = crm_element_value(output, F_LRMD_PROVIDER); type = crm_element_value(output, F_LRMD_TYPE); if (!class || !type) { free_xml(output); return NULL; } else if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) && !provider) { free_xml(output); return NULL; } rsc_info = lrmd_new_rsc_info(rsc_id, class, provider, type); free_xml(output); return rsc_info; } void lrmd_free_op_info(lrmd_op_info_t *op_info) { if (op_info) { free(op_info->rsc_id); free(op_info->action); free(op_info->interval_ms_s); free(op_info->timeout_ms_s); free(op_info); } } static int lrmd_api_get_recurring_ops(lrmd_t *lrmd, const char *rsc_id, int timeout_ms, enum lrmd_call_options options, GList **output) { xmlNode *data = NULL; xmlNode *output_xml = NULL; int rc = pcmk_ok; if (output == NULL) { return -EINVAL; } *output = NULL; // Send request if (rsc_id) { data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); } rc = lrmd_send_command(lrmd, LRMD_OP_GET_RECURRING, data, &output_xml, timeout_ms, options, TRUE); if (data) { free_xml(data); } // Process reply if ((rc != pcmk_ok) || (output_xml == NULL)) { return rc; } for (xmlNode *rsc_xml = first_named_child(output_xml, F_LRMD_RSC); (rsc_xml != NULL) && (rc == pcmk_ok); rsc_xml = crm_next_same_xml(rsc_xml)) { rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID); if (rsc_id == NULL) { crm_err("Could not parse recurring operation information from executor"); continue; } for (xmlNode *op_xml = first_named_child(rsc_xml, T_LRMD_RSC_OP); op_xml != NULL; op_xml = crm_next_same_xml(op_xml)) { lrmd_op_info_t *op_info = calloc(1, sizeof(lrmd_op_info_t)); if (op_info == NULL) { rc = -ENOMEM; break; } op_info->rsc_id = strdup(rsc_id); op_info->action = crm_element_value_copy(op_xml, F_LRMD_RSC_ACTION); op_info->interval_ms_s = crm_element_value_copy(op_xml, F_LRMD_RSC_INTERVAL); op_info->timeout_ms_s = crm_element_value_copy(op_xml, F_LRMD_TIMEOUT); *output = g_list_prepend(*output, op_info); } } free_xml(output_xml); return rc; } static void lrmd_api_set_callback(lrmd_t * lrmd, lrmd_event_callback callback) { lrmd_private_t *native = lrmd->lrmd_private; native->callback = callback; } void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)) { lrmd_private_t *native = lrmd->lrmd_private; native->proxy_callback = callback; native->proxy_callback_userdata = userdata; } void lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg) { lrmd_private_t *native = lrmd->lrmd_private; if (native->proxy_callback) { crm_log_xml_trace(msg, "PROXY_INBOUND"); native->proxy_callback(lrmd, native->proxy_callback_userdata, msg); } } int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg) { if (lrmd == NULL) { return -ENOTCONN; } crm_xml_add(msg, F_LRMD_OPERATION, CRM_OP_IPC_FWD); crm_log_xml_trace(msg, "PROXY_OUTBOUND"); return lrmd_send_xml_no_reply(lrmd, msg); } static int stonith_get_metadata(const char *provider, const char *type, char **output) { int rc = pcmk_ok; stonith_t *stonith_api = stonith_api_new(); if (stonith_api == NULL) { crm_err("Could not get fence agent meta-data: API memory allocation failed"); return -ENOMEM; } rc = stonith_api->cmds->metadata(stonith_api, st_opt_sync_call, type, provider, output, 0); if ((rc == pcmk_ok) && (*output == NULL)) { rc = -EIO; } stonith_api->cmds->free(stonith_api); return rc; } static int lrmd_api_get_metadata(lrmd_t *lrmd, const char *standard, const char *provider, const char *type, char **output, enum lrmd_call_options options) { return lrmd->cmds->get_metadata_params(lrmd, standard, provider, type, output, options, NULL); } static int lrmd_api_get_metadata_params(lrmd_t *lrmd, const char *standard, const char *provider, const char *type, char **output, enum lrmd_call_options options, lrmd_key_value_t *params) { svc_action_t *action = NULL; GHashTable *params_table = NULL; if (!standard || !type) { lrmd_key_value_freeall(params); return -EINVAL; } if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { lrmd_key_value_freeall(params); return stonith_get_metadata(provider, type, output); } params_table = pcmk__strkey_table(free, free); for (const lrmd_key_value_t *param = params; param; param = param->next) { g_hash_table_insert(params_table, strdup(param->key), strdup(param->value)); } action = services__create_resource_action(type, standard, provider, type, CRMD_ACTION_METADATA, 0, CRMD_METADATA_CALL_TIMEOUT, params_table, 0); lrmd_key_value_freeall(params); if (action == NULL) { return -ENOMEM; } if (action->rc != PCMK_OCF_OK) { services_action_free(action); return -EINVAL; } if (!services_action_sync(action)) { crm_err("Failed to retrieve meta-data for %s:%s:%s", standard, provider, type); services_action_free(action); return -EIO; } if (!action->stdout_data) { crm_err("Failed to receive meta-data for %s:%s:%s", standard, provider, type); services_action_free(action); return -EIO; } *output = strdup(action->stdout_data); services_action_free(action); return pcmk_ok; } static int lrmd_api_exec(lrmd_t *lrmd, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout, /* ms */ int start_delay, /* ms */ enum lrmd_call_options options, lrmd_key_value_t * params) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); lrmd_key_value_t *tmp = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add(data, F_LRMD_RSC_ACTION, action); crm_xml_add(data, F_LRMD_RSC_USERDATA_STR, userdata); crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms); crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout); crm_xml_add_int(data, F_LRMD_RSC_START_DELAY, start_delay); for (tmp = params; tmp; tmp = tmp->next) { hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args); } rc = lrmd_send_command(lrmd, LRMD_OP_RSC_EXEC, data, NULL, timeout, options, TRUE); free_xml(data); lrmd_key_value_freeall(params); return rc; } /* timeout is in ms */ static int lrmd_api_exec_alert(lrmd_t *lrmd, const char *alert_id, const char *alert_path, int timeout, lrmd_key_value_t *params) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_ALERT); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); lrmd_key_value_t *tmp = NULL; crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_ALERT_ID, alert_id); crm_xml_add(data, F_LRMD_ALERT_PATH, alert_path); crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout); for (tmp = params; tmp; tmp = tmp->next) { hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args); } rc = lrmd_send_command(lrmd, LRMD_OP_ALERT_EXEC, data, NULL, timeout, lrmd_opt_notify_orig_only, TRUE); free_xml(data); lrmd_key_value_freeall(params); return rc; } static int lrmd_api_cancel(lrmd_t *lrmd, const char *rsc_id, const char *action, guint interval_ms) { int rc = pcmk_ok; xmlNode *data = create_xml_node(NULL, F_LRMD_RSC); crm_xml_add(data, F_LRMD_ORIGIN, __func__); crm_xml_add(data, F_LRMD_RSC_ACTION, action); crm_xml_add(data, F_LRMD_RSC_ID, rsc_id); crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms); rc = lrmd_send_command(lrmd, LRMD_OP_RSC_CANCEL, data, NULL, 0, 0, TRUE); free_xml(data); return rc; } static int list_stonith_agents(lrmd_list_t ** resources) { int rc = 0; stonith_t *stonith_api = stonith_api_new(); stonith_key_value_t *stonith_resources = NULL; stonith_key_value_t *dIter = NULL; if (stonith_api == NULL) { crm_err("Could not list fence agents: API memory allocation failed"); return -ENOMEM; } stonith_api->cmds->list_agents(stonith_api, st_opt_sync_call, NULL, &stonith_resources, 0); stonith_api->cmds->free(stonith_api); for (dIter = stonith_resources; dIter; dIter = dIter->next) { rc++; if (resources) { *resources = lrmd_list_add(*resources, dIter->value); } } stonith_key_value_freeall(stonith_resources, 1, 0); return rc; } static int lrmd_api_list_agents(lrmd_t * lrmd, lrmd_list_t ** resources, const char *class, const char *provider) { int rc = 0; int stonith_count = 0; // Initially, whether to include stonith devices if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { stonith_count = 1; } else { GList *gIter = NULL; GList *agents = resources_list_agents(class, provider); for (gIter = agents; gIter != NULL; gIter = gIter->next) { *resources = lrmd_list_add(*resources, (const char *)gIter->data); rc++; } g_list_free_full(agents, free); if (!class) { stonith_count = 1; } } if (stonith_count) { // Now, if stonith devices are included, how many there are stonith_count = list_stonith_agents(resources); if (stonith_count > 0) { rc += stonith_count; } } if (rc == 0) { crm_notice("No agents found for class %s", class); rc = -EPROTONOSUPPORT; } return rc; } static bool does_provider_have_agent(const char *agent, const char *provider, const char *class) { bool found = false; GList *agents = NULL; GList *gIter2 = NULL; agents = resources_list_agents(class, provider); for (gIter2 = agents; gIter2 != NULL; gIter2 = gIter2->next) { if (pcmk__str_eq(agent, gIter2->data, pcmk__str_casei)) { found = true; } } g_list_free_full(agents, free); return found; } static int lrmd_api_list_ocf_providers(lrmd_t * lrmd, const char *agent, lrmd_list_t ** providers) { int rc = pcmk_ok; char *provider = NULL; GList *ocf_providers = NULL; GList *gIter = NULL; ocf_providers = resources_list_providers(PCMK_RESOURCE_CLASS_OCF); for (gIter = ocf_providers; gIter != NULL; gIter = gIter->next) { provider = gIter->data; if (!agent || does_provider_have_agent(agent, provider, PCMK_RESOURCE_CLASS_OCF)) { *providers = lrmd_list_add(*providers, (const char *)gIter->data); rc++; } } g_list_free_full(ocf_providers, free); return rc; } static int lrmd_api_list_standards(lrmd_t * lrmd, lrmd_list_t ** supported) { int rc = 0; GList *standards = NULL; GList *gIter = NULL; standards = resources_list_standards(); for (gIter = standards; gIter != NULL; gIter = gIter->next) { *supported = lrmd_list_add(*supported, (const char *)gIter->data); rc++; } if (list_stonith_agents(NULL) > 0) { *supported = lrmd_list_add(*supported, PCMK_RESOURCE_CLASS_STONITH); rc++; } g_list_free_full(standards, free); return rc; } /*! * \internal * \brief Create an executor API object * * \param[out] api Will be set to newly created API object (it is the * caller's responsibility to free this value with * lrmd_api_delete() if this function succeeds) * \param[in] nodename If the object will be used for a remote connection, * the node name to use in cluster for remote executor * \param[in] server If the object will be used for a remote connection, * the resolvable host name to connect to * \param[in] port If the object will be used for a remote connection, * port number on \p server to connect to * * \return Standard Pacemaker return code * \note If the caller leaves one of \p nodename or \p server NULL, the other's * value will be used for both. If the caller leaves both NULL, an API * object will be created for a local executor connection. */ int lrmd__new(lrmd_t **api, const char *nodename, const char *server, int port) { lrmd_private_t *pvt = NULL; if (api == NULL) { return EINVAL; } *api = NULL; // Allocate all memory needed *api = calloc(1, sizeof(lrmd_t)); if (*api == NULL) { return ENOMEM; } pvt = calloc(1, sizeof(lrmd_private_t)); if (pvt == NULL) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } (*api)->lrmd_private = pvt; // @TODO Do we need to do this for local connections? pvt->remote = calloc(1, sizeof(pcmk__remote_t)); (*api)->cmds = calloc(1, sizeof(lrmd_api_operations_t)); if ((pvt->remote == NULL) || ((*api)->cmds == NULL)) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } // Set methods (*api)->cmds->connect = lrmd_api_connect; (*api)->cmds->connect_async = lrmd_api_connect_async; (*api)->cmds->is_connected = lrmd_api_is_connected; (*api)->cmds->poke_connection = lrmd_api_poke_connection; (*api)->cmds->disconnect = lrmd_api_disconnect; (*api)->cmds->register_rsc = lrmd_api_register_rsc; (*api)->cmds->unregister_rsc = lrmd_api_unregister_rsc; (*api)->cmds->get_rsc_info = lrmd_api_get_rsc_info; (*api)->cmds->get_recurring_ops = lrmd_api_get_recurring_ops; (*api)->cmds->set_callback = lrmd_api_set_callback; (*api)->cmds->get_metadata = lrmd_api_get_metadata; (*api)->cmds->exec = lrmd_api_exec; (*api)->cmds->cancel = lrmd_api_cancel; (*api)->cmds->list_agents = lrmd_api_list_agents; (*api)->cmds->list_ocf_providers = lrmd_api_list_ocf_providers; (*api)->cmds->list_standards = lrmd_api_list_standards; (*api)->cmds->exec_alert = lrmd_api_exec_alert; (*api)->cmds->get_metadata_params = lrmd_api_get_metadata_params; if ((nodename == NULL) && (server == NULL)) { pvt->type = pcmk__client_ipc; } else { #ifdef HAVE_GNUTLS_GNUTLS_H if (nodename == NULL) { nodename = server; } else if (server == NULL) { server = nodename; } pvt->type = pcmk__client_tls; pvt->remote_nodename = strdup(nodename); pvt->server = strdup(server); if ((pvt->remote_nodename == NULL) || (pvt->server == NULL)) { lrmd_api_delete(*api); *api = NULL; return ENOMEM; } pvt->port = port; if (pvt->port == 0) { pvt->port = crm_default_remote_port(); } #else crm_err("Cannot communicate with Pacemaker Remote " "because GnuTLS is not enabled for this build"); lrmd_api_delete(*api); *api = NULL; return EOPNOTSUPP; #endif } return pcmk_rc_ok; } lrmd_t * lrmd_api_new(void) { lrmd_t *api = NULL; CRM_ASSERT(lrmd__new(&api, NULL, NULL, 0) == pcmk_rc_ok); return api; } lrmd_t * lrmd_remote_api_new(const char *nodename, const char *server, int port) { lrmd_t *api = NULL; CRM_ASSERT(lrmd__new(&api, nodename, server, port) == pcmk_rc_ok); return api; } void lrmd_api_delete(lrmd_t * lrmd) { if (!lrmd) { return; } lrmd->cmds->disconnect(lrmd); /* no-op if already disconnected */ free(lrmd->cmds); if (lrmd->lrmd_private) { lrmd_private_t *native = lrmd->lrmd_private; #ifdef HAVE_GNUTLS_GNUTLS_H free(native->server); #endif free(native->remote_nodename); free(native->remote); free(native->token); free(native->peer_version); } free(lrmd->lrmd_private); free(lrmd); } + +/*! + * \internal + * \brief Set the result of an executor event + * + * \param[in,out] event Executor event to set + * \param[in] rc OCF exit status of event + * \param[in] op_status Executor status of event + * \param[in] exit_reason Human-friendly description of event + */ +void +lrmd__set_result(lrmd_event_data_t *event, enum ocf_exitcode rc, int op_status, + const char *exit_reason) +{ + if (event == NULL) { + return; + } + + event->rc = rc; + event->op_status = op_status; + + if (!pcmk__str_eq(event->exit_reason, exit_reason, pcmk__str_none)) { + free((void *) event->exit_reason); + event->exit_reason = (exit_reason == NULL)? NULL : strdup(exit_reason); + } +} + +/*! + * \internal + * \brief Clear an executor event's exit reason, output, and error output + * + * \param[in] event Executor event to reset + */ +void +lrmd__reset_result(lrmd_event_data_t *event) +{ + if (event == NULL) { + return; + } + + free((void *) event->exit_reason); + event->exit_reason = NULL; + + free((void *) event->output); + event->output = NULL; +} diff --git a/lib/pacemaker/pcmk_graph_consumer.c b/lib/pacemaker/pcmk_graph_consumer.c index 08198a850f..ceb731f769 100644 --- a/lib/pacemaker/pcmk_graph_consumer.c +++ b/lib/pacemaker/pcmk_graph_consumer.c @@ -1,845 +1,846 @@ /* * Copyright 2004-2021 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] 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 synapse->ready, so the only * thing that uses the synapse->ready value from here is * synapse_state_str(). */ static void update_synapse_ready(synapse_t *synapse, int action_id) { if (synapse->ready) { return; // All inputs have already been confirmed } synapse->ready = TRUE; // Presume ready until proven otherwise for (GList *lpc = synapse->inputs; lpc != NULL; lpc = lpc->next) { crm_action_t *prereq = (crm_action_t *) lpc->data; if (prereq->id == action_id) { crm_trace("Confirming input %d of synapse %d", action_id, synapse->id); prereq->confirmed = TRUE; } else if (!(prereq->confirmed)) { synapse->ready = FALSE; crm_trace("Synapse %d still not ready after action %d", synapse->id, action_id); } } if (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] synapse Transition graph synapse that action belongs to * \param[in] action_id ID of action that completed */ static void update_synapse_confirmed(synapse_t *synapse, int action_id) { bool all_confirmed = true; for (GList *lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { crm_action_t *action = (crm_action_t *) lpc->data; if (action->id == action_id) { crm_trace("Confirmed action %d of synapse %d", action_id, synapse->id); action->confirmed = TRUE; } else if (all_confirmed && !(action->confirmed)) { all_confirmed = false; crm_trace("Synapse %d still not confirmed after action %d", synapse->id, action_id); } } if (all_confirmed && !(synapse->confirmed)) { crm_trace("Confirmed synapse %d", synapse->id); synapse->confirmed = TRUE; } } /*! * \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(crm_graph_t *graph, crm_action_t *action) { for (GList *lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { synapse_t *synapse = (synapse_t *) lpc->data; if (synapse->confirmed || synapse->failed) { continue; // This synapse already completed } else if (synapse->executed) { update_synapse_confirmed(synapse, action->id); } else if (!(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 crm_graph_functions_t *graph_fns = NULL; /*! * \internal * \brief Set transition graph execution functions * * \param[in] Execution functions to use */ void pcmk__set_graph_functions(crm_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->crmd != NULL); CRM_ASSERT(graph_fns->pseudo != NULL); CRM_ASSERT(graph_fns->stonith != NULL); } /*! * \internal * \brief Check whether a graph synapse is ready to be executed * * \param[in] graph Transition graph that synapse is part of * \param[in] synapse Synapse to check * * \return true if synapse is ready, false otherwise */ static bool should_fire_synapse(crm_graph_t *graph, synapse_t *synapse) { GList *lpc = NULL; synapse->ready = TRUE; for (lpc = synapse->inputs; lpc != NULL; lpc = lpc->next) { crm_action_t *prereq = (crm_action_t *) lpc->data; if (!(prereq->confirmed)) { crm_trace("Input %d for synapse %d not yet confirmed", prereq->id, synapse->id); synapse->ready = FALSE; break; } else if (prereq->failed && !(prereq->can_fail)) { crm_trace("Input %d for synapse %d confirmed but failed", prereq->id, synapse->id); synapse->ready = FALSE; break; } } if (synapse->ready) { crm_trace("Synapse %d is ready to execute", synapse->id); } else { return false; } for (lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { crm_action_t *a = (crm_action_t *) lpc->data; if (a->type == action_type_pseudo) { /* 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] graph Transition graph containing action * \parma[in] action Action to execute * * \return TRUE if action was initiated, FALSE otherwise */ static gboolean initiate_action(crm_graph_t *graph, crm_action_t *action) { const char *id = ID(action->xml); CRM_CHECK(!(action->executed), return FALSE); CRM_CHECK(id != NULL, return FALSE); action->executed = TRUE; switch (action->type) { case action_type_pseudo: crm_trace("Executing pseudo-action %d (%s)", action->id, id); return graph_fns->pseudo(graph, action); case action_type_rsc: crm_trace("Executing resource action %d (%s)", action->id, id); return graph_fns->rsc(graph, action); case action_type_crm: 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->stonith(graph, action); } crm_trace("Executing control action %d (%s)", action->id, id); return graph_fns->crmd(graph, action); default: crm_err("Unsupported graph action type <%s id='%s'> (bug?)", crm_element_name(action->xml), id); return FALSE; } } /*! * \internal * \brief Execute a graph synapse * * \param[in] graph Transition graph with synapse to execute * \param[in] synapse Synapse to execute * * \return Standard Pacemaker return value */ static int fire_synapse(crm_graph_t *graph, synapse_t *synapse) { synapse->executed = TRUE; for (GList *lpc = synapse->actions; lpc != NULL; lpc = lpc->next) { crm_action_t *action = (crm_action_t *) lpc->data; if (!initiate_action(graph, action)) { crm_err("Failed initiating <%s id=%d> in synapse %d", crm_element_name(action->xml), action->id, synapse->id); synapse->confirmed = TRUE; action->confirmed = TRUE; action->failed = TRUE; return pcmk_rc_error; } } return pcmk_rc_ok; } /*! * \internal * \brief Dummy graph method that can be used with simulations * * \param[in] graph Transition graph containing action * \param[in] action Action to be initiated * * \retval TRUE Action initiation was (simulated to be) successful * \retval FALSE Action initiation was (simulated to be) failed (due to the * PE_fail environment variable being set to the action ID) */ static gboolean pseudo_action_dummy(crm_graph_t * graph, crm_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); action->failed = TRUE; graph->abort_priority = INFINITY; } else { crm_trace("Dummy event handler: action %d initiated", action->id); } action->confirmed = TRUE; pcmk__update_graph(graph, action); return TRUE; } static crm_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] graph Transition graph to execute * * \return Status of transition after execution */ enum transition_status pcmk__execute_graph(crm_graph_t *graph) { GList *lpc = NULL; int log_level = LOG_DEBUG; enum transition_status pass_result = transition_active; const char *status = "In progress"; if (graph_fns == NULL) { graph_fns = &default_fns; } if (graph == NULL) { return transition_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) { synapse_t *synapse = (synapse_t *) lpc->data; if (synapse->confirmed) { graph->completed++; } else if (!(synapse->failed) && 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) { synapse_t *synapse = (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 (synapse->failed) { graph->skipped++; continue; } else if (synapse->confirmed || 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 (!(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 = transition_terminated; status = "Terminated"; } else if (graph->skipped != 0) { log_level = LOG_NOTICE; pass_result = transition_complete; status = "Stopped"; } else { log_level = LOG_NOTICE; pass_result = transition_complete; status = "Complete"; } } else if (graph->fired == 0) { pass_result = transition_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 crm_action_t * unpack_action(synapse_t *parent, xmlNode *xml_action) { action_type_e action_type; crm_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 = action_type_rsc; } else if (pcmk__str_eq(element, XML_GRAPH_TAG_PSEUDO_EVENT, pcmk__str_casei)) { action_type = action_type_pseudo; } else if (pcmk__str_eq(element, XML_GRAPH_TAG_CRM_EVENT, pcmk__str_casei)) { action_type = action_type_crm; } 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(crm_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 = action_type_rsc; 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) { crm_str_to_boolean(value, &(action->can_fail)); #ifndef PCMK__COMPAT_2_0 if (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] 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 synapse_t * unpack_synapse(crm_graph_t *new_graph, xmlNode *xml_synapse) { const char *value = NULL; xmlNode *action_set = NULL; synapse_t *new_synapse = NULL; crm_trace("Unpacking synapse %s", ID(xml_synapse)); new_synapse = calloc(1, sizeof(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)) { crm_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)) { crm_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: ... ... */ crm_graph_t * pcmk__unpack_graph(xmlNode *xml_graph, const char *reference) { crm_graph_t *new_graph = NULL; const char *t_id = NULL; const char *time = NULL; new_graph = calloc(1, sizeof(crm_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 = tg_done; // Parse top-level attributes from if (xml_graph != NULL) { t_id = crm_element_value(xml_graph, "transition_id"); CRM_CHECK(t_id != NULL, free(new_graph); return NULL); pcmk__scan_min_int(t_id, &(new_graph->id), -1); time = crm_element_value(xml_graph, "cluster-delay"); CRM_CHECK(time != NULL, free(new_graph); return NULL); new_graph->network_delay = crm_parse_interval_spec(time); time = crm_element_value(xml_graph, "stonith-timeout"); if (time == NULL) { new_graph->stonith_timeout = new_graph->network_delay; } else { new_graph->stonith_timeout = crm_parse_interval_spec(time); } // Use 0 (dynamic limit) as default/invalid, -1 (no limit) as minimum t_id = crm_element_value(xml_graph, "batch-limit"); if ((t_id == NULL) || (pcmk__scan_min_int(t_id, &(new_graph->batch_limit), -1) != pcmk_rc_ok)) { new_graph->batch_limit = 0; } t_id = crm_element_value(xml_graph, "migration-limit"); pcmk__scan_min_int(t_id, &(new_graph->migration_limit), -1); } // Unpack each child element for (xmlNode *synapse_xml = first_named_child(xml_graph, "synapse"); synapse_xml != NULL; synapse_xml = crm_next_same_xml(synapse_xml)) { 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] user_data Action to free */ static void free_graph_action(gpointer user_data) { crm_action_t *action = user_data; if ((action->timer != NULL) && (action->timer->source_id != 0)) { crm_warn("Cancelling timer for graph action %d", action->id); g_source_remove(action->timer->source_id); } if (action->params != NULL) { g_hash_table_destroy(action->params); } free_xml(action->xml); free(action->timer); free(action); } /*! * \internal * \brief Free a transition graph synapse object * * \param[in] user_data Synapse to free */ static void free_graph_synapse(gpointer user_data) { 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] graph Transition graph to free */ void pcmk__free_graph(crm_graph_t *graph) { if (graph != NULL) { g_list_free_full(graph->synapses, free_graph_synapse); free(graph->source); 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] 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(xmlNode *resource, crm_action_t *action, - int status, int rc) + 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 == action_type_rsc, 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); - op->rc = rc; - op->op_status = status; + 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_sched_transition.c b/lib/pacemaker/pcmk_sched_transition.c index efb4a81cde..0dcd5bb37b 100644 --- a/lib/pacemaker/pcmk_sched_transition.c +++ b/lib/pacemaker/pcmk_sched_transition.c @@ -1,850 +1,850 @@ /* * Copyright 2009-2021 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 // lrmd_event_data_t, lrmd_free_event() #include #include #include #include +#include #include #include static pcmk__output_t *out = NULL; static cib_t *fake_cib = NULL; static GList *fake_resource_list = NULL; static GList *fake_op_fail_list = NULL; gboolean bringing_nodes_online = FALSE; #define STATUS_PATH_MAX 512 #define NEW_NODE_TEMPLATE "//"XML_CIB_TAG_NODE"[@uname='%s']" #define NODE_TEMPLATE "//"XML_CIB_TAG_STATE"[@uname='%s']" #define RSC_TEMPLATE "//"XML_CIB_TAG_STATE"[@uname='%s']//"XML_LRM_TAG_RESOURCE"[@id='%s']" static void inject_transient_attr(xmlNode * cib_node, const char *name, const char *value) { xmlNode *attrs = NULL; xmlNode *instance_attrs = NULL; const char *node_uuid = ID(cib_node); out->message(out, "inject-attr", name, value, cib_node); attrs = first_named_child(cib_node, XML_TAG_TRANSIENT_NODEATTRS); if (attrs == NULL) { attrs = create_xml_node(cib_node, XML_TAG_TRANSIENT_NODEATTRS); crm_xml_add(attrs, XML_ATTR_ID, node_uuid); } instance_attrs = first_named_child(attrs, XML_TAG_ATTR_SETS); if (instance_attrs == NULL) { instance_attrs = create_xml_node(attrs, XML_TAG_ATTR_SETS); crm_xml_add(instance_attrs, XML_ATTR_ID, node_uuid); } crm_create_nvpair_xml(instance_attrs, NULL, name, value); } static void update_failcounts(xmlNode * cib_node, const char *resource, const char *task, guint interval_ms, int rc) { if (rc == 0) { return; } else if ((rc == 7) && (interval_ms == 0)) { return; } else { char *name = NULL; char *now = pcmk__ttoa(time(NULL)); name = pcmk__failcount_name(resource, task, interval_ms); inject_transient_attr(cib_node, name, "value++"); free(name); name = pcmk__lastfailure_name(resource, task, interval_ms); inject_transient_attr(cib_node, name, now); free(name); free(now); } } static void create_node_entry(cib_t * cib_conn, const char *node) { int rc = pcmk_ok; char *xpath = crm_strdup_printf(NEW_NODE_TEMPLATE, node); rc = cib_conn->cmds->query(cib_conn, xpath, NULL, cib_xpath | cib_sync_call | cib_scope_local); if (rc == -ENXIO) { xmlNode *cib_object = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(cib_object, XML_ATTR_ID, node); // Use node name as ID crm_xml_add(cib_object, XML_ATTR_UNAME, node); cib_conn->cmds->create(cib_conn, XML_CIB_TAG_NODES, cib_object, cib_sync_call | cib_scope_local); /* Not bothering with subsequent query to see if it exists, we'll bomb out later in the call to query_node_uuid()... */ free_xml(cib_object); } free(xpath); } static lrmd_event_data_t * create_op(xmlNode *cib_resource, const char *task, guint interval_ms, int outcome) { lrmd_event_data_t *op = NULL; xmlNode *xop = NULL; op = lrmd_new_event(ID(cib_resource), task, interval_ms); - op->rc = outcome; - op->op_status = PCMK_EXEC_DONE; + lrmd__set_result(op, outcome, PCMK_EXEC_DONE, "Simulated action result"); op->params = NULL; /* TODO: Fill me in */ op->t_run = (unsigned int) time(NULL); op->t_rcchange = op->t_run; op->call_id = 0; for (xop = pcmk__xe_first_child(cib_resource); xop != NULL; xop = pcmk__xe_next(xop)) { int tmp = 0; crm_element_value_int(xop, XML_LRM_ATTR_CALLID, &tmp); if (tmp > op->call_id) { op->call_id = tmp; } } op->call_id++; return op; } static xmlNode * inject_op(xmlNode * cib_resource, lrmd_event_data_t * op, int target_rc) { return pcmk__create_history_xml(cib_resource, op, CRM_FEATURE_SET, target_rc, NULL, crm_system_name, LOG_TRACE); } static xmlNode * inject_node_state(cib_t * cib_conn, const char *node, const char *uuid) { int rc = pcmk_ok; xmlNode *cib_object = NULL; char *xpath = crm_strdup_printf(NODE_TEMPLATE, node); if (bringing_nodes_online) { create_node_entry(cib_conn, node); } rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object, cib_xpath | cib_sync_call | cib_scope_local); if (cib_object && ID(cib_object) == NULL) { crm_err("Detected multiple node_state entries for xpath=%s, bailing", xpath); crm_log_xml_warn(cib_object, "Duplicates"); free(xpath); crm_exit(CRM_EX_SOFTWARE); return NULL; // not reached, but makes static analysis happy } if (rc == -ENXIO) { char *found_uuid = NULL; if (uuid == NULL) { query_node_uuid(cib_conn, node, &found_uuid, NULL); } else { found_uuid = strdup(uuid); } cib_object = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(cib_object, XML_ATTR_UUID, found_uuid); crm_xml_add(cib_object, XML_ATTR_UNAME, node); cib_conn->cmds->create(cib_conn, XML_CIB_TAG_STATUS, cib_object, cib_sync_call | cib_scope_local); free_xml(cib_object); free(found_uuid); rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object, cib_xpath | cib_sync_call | cib_scope_local); crm_trace("injecting node state for %s. rc is %d", node, rc); } free(xpath); CRM_ASSERT(rc == pcmk_ok); return cib_object; } static xmlNode * modify_node(cib_t * cib_conn, char *node, gboolean up) { xmlNode *cib_node = inject_node_state(cib_conn, node, NULL); if (up) { crm_xml_add(cib_node, XML_NODE_IN_CLUSTER, XML_BOOLEAN_YES); crm_xml_add(cib_node, XML_NODE_IS_PEER, ONLINESTATUS); crm_xml_add(cib_node, XML_NODE_JOIN_STATE, CRMD_JOINSTATE_MEMBER); crm_xml_add(cib_node, XML_NODE_EXPECTED, CRMD_JOINSTATE_MEMBER); } else { crm_xml_add(cib_node, XML_NODE_IN_CLUSTER, XML_BOOLEAN_NO); crm_xml_add(cib_node, XML_NODE_IS_PEER, OFFLINESTATUS); crm_xml_add(cib_node, XML_NODE_JOIN_STATE, CRMD_JOINSTATE_DOWN); crm_xml_add(cib_node, XML_NODE_EXPECTED, CRMD_JOINSTATE_DOWN); } crm_xml_add(cib_node, XML_ATTR_ORIGIN, crm_system_name); return cib_node; } static xmlNode * find_resource_xml(xmlNode * cib_node, const char *resource) { xmlNode *match = NULL; const char *node = crm_element_value(cib_node, XML_ATTR_UNAME); char *xpath = crm_strdup_printf(RSC_TEMPLATE, node, resource); match = get_xpath_object(xpath, cib_node, LOG_TRACE); free(xpath); return match; } static xmlNode * inject_resource(xmlNode * cib_node, const char *resource, const char *lrm_name, const char *rclass, const char *rtype, const char *rprovider) { xmlNode *lrm = NULL; xmlNode *container = NULL; xmlNode *cib_resource = NULL; char *xpath = NULL; cib_resource = find_resource_xml(cib_node, resource); if (cib_resource != NULL) { /* If an existing LRM history entry uses the resource name, * continue using it, even if lrm_name is different. */ return cib_resource; } // Check for history entry under preferred name if (strcmp(resource, lrm_name)) { cib_resource = find_resource_xml(cib_node, lrm_name); if (cib_resource != NULL) { return cib_resource; } } /* One day, add query for class, provider, type */ if (rclass == NULL || rtype == NULL) { out->err(out, "Resource %s not found in the status section of %s." " Please supply the class and type to continue", resource, ID(cib_node)); return NULL; } else if (!pcmk__strcase_any_of(rclass, PCMK_RESOURCE_CLASS_OCF, PCMK_RESOURCE_CLASS_STONITH, PCMK_RESOURCE_CLASS_SERVICE, PCMK_RESOURCE_CLASS_UPSTART, PCMK_RESOURCE_CLASS_SYSTEMD, PCMK_RESOURCE_CLASS_LSB, NULL)) { out->err(out, "Invalid class for %s: %s", resource, rclass); return NULL; } else if (pcmk_is_set(pcmk_get_ra_caps(rclass), pcmk_ra_cap_provider) && (rprovider == NULL)) { out->err(out, "Please specify the provider for resource %s", resource); return NULL; } xpath = (char *)xmlGetNodePath(cib_node); crm_info("Injecting new resource %s into %s '%s'", lrm_name, xpath, ID(cib_node)); free(xpath); lrm = first_named_child(cib_node, XML_CIB_TAG_LRM); if (lrm == NULL) { const char *node_uuid = ID(cib_node); lrm = create_xml_node(cib_node, XML_CIB_TAG_LRM); crm_xml_add(lrm, XML_ATTR_ID, node_uuid); } container = first_named_child(lrm, XML_LRM_TAG_RESOURCES); if (container == NULL) { container = create_xml_node(lrm, XML_LRM_TAG_RESOURCES); } cib_resource = create_xml_node(container, XML_LRM_TAG_RESOURCE); // If we're creating a new entry, use the preferred name crm_xml_add(cib_resource, XML_ATTR_ID, lrm_name); crm_xml_add(cib_resource, XML_AGENT_ATTR_CLASS, rclass); crm_xml_add(cib_resource, XML_AGENT_ATTR_PROVIDER, rprovider); crm_xml_add(cib_resource, XML_ATTR_TYPE, rtype); return cib_resource; } #define XPATH_MAX 1024 static int find_ticket_state(cib_t * the_cib, const char *ticket_id, xmlNode ** ticket_state_xml) { int offset = 0; int rc = pcmk_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; CRM_ASSERT(ticket_state_xml != NULL); *ticket_state_xml = NULL; xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", "/cib/status/tickets"); if (ticket_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s[@id=\"%s\"]", XML_CIB_TAG_TICKET_STATE, ticket_id); } CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); if (rc != pcmk_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { if (ticket_id) { out->err(out, "Multiple ticket_states match ticket_id=%s", ticket_id); } *ticket_state_xml = xml_search; } else { *ticket_state_xml = xml_search; } bail: free(xpath_string); return rc; } static int set_ticket_state_attr(const char *ticket_id, const char *attr_name, const char *attr_value, cib_t * cib, int cib_options) { int rc = pcmk_ok; xmlNode *xml_top = NULL; xmlNode *ticket_state_xml = NULL; rc = find_ticket_state(cib, ticket_id, &ticket_state_xml); if (rc == pcmk_ok) { crm_debug("Found a match state for ticket: id=%s", ticket_id); xml_top = ticket_state_xml; } else if (rc != -ENXIO) { return rc; } 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); } crm_xml_add(ticket_state_xml, attr_name, attr_value); crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, xml_top, cib_options); free_xml(xml_top); return rc; } void modify_configuration(pe_working_set_t * data_set, cib_t *cib, pcmk_injections_t *injections) { int rc = pcmk_ok; GList *gIter = NULL; xmlNode *cib_op = NULL; xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; lrmd_event_data_t *op = NULL; out = data_set->priv; out->message(out, "inject-modify-config", injections->quorum, injections->watchdog); if (injections->quorum) { xmlNode *top = create_xml_node(NULL, XML_TAG_CIB); /* crm_xml_add(top, XML_ATTR_DC_UUID, dc_uuid); */ crm_xml_add(top, XML_ATTR_HAVE_QUORUM, injections->quorum); rc = cib->cmds->modify(cib, NULL, top, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } if (injections->watchdog) { rc = update_attr_delegate(cib, cib_sync_call | cib_scope_local, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, XML_ATTR_HAVE_WATCHDOG, injections->watchdog, FALSE, NULL, NULL); CRM_ASSERT(rc == pcmk_ok); } for (gIter = injections->node_up; gIter != NULL; gIter = gIter->next) { char *node = (char *)gIter->data; out->message(out, "inject-modify-node", "Online", node); cib_node = modify_node(cib, node, TRUE); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); } for (gIter = injections->node_down; gIter != NULL; gIter = gIter->next) { char xpath[STATUS_PATH_MAX]; char *node = (char *)gIter->data; out->message(out, "inject-modify-node", "Offline", node); cib_node = modify_node(cib, node, FALSE); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); snprintf(xpath, STATUS_PATH_MAX, "//node_state[@uname='%s']/%s", node, XML_CIB_TAG_LRM); cib->cmds->remove(cib, xpath, NULL, cib_xpath | cib_sync_call | cib_scope_local); snprintf(xpath, STATUS_PATH_MAX, "//node_state[@uname='%s']/%s", node, XML_TAG_TRANSIENT_NODEATTRS); cib->cmds->remove(cib, xpath, NULL, cib_xpath | cib_sync_call | cib_scope_local); } for (gIter = injections->node_fail; gIter != NULL; gIter = gIter->next) { char *node = (char *)gIter->data; out->message(out, "inject-modify-node", "Failing", node); cib_node = modify_node(cib, node, TRUE); crm_xml_add(cib_node, XML_NODE_IN_CLUSTER, XML_BOOLEAN_NO); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); } for (gIter = injections->ticket_grant; gIter != NULL; gIter = gIter->next) { char *ticket_id = (char *)gIter->data; out->message(out, "inject-modify-ticket", "Granting", ticket_id); rc = set_ticket_state_attr(ticket_id, "granted", "true", cib, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } for (gIter = injections->ticket_revoke; gIter != NULL; gIter = gIter->next) { char *ticket_id = (char *)gIter->data; out->message(out, "inject-modify-ticket", "Revoking", ticket_id); rc = set_ticket_state_attr(ticket_id, "granted", "false", cib, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } for (gIter = injections->ticket_standby; gIter != NULL; gIter = gIter->next) { char *ticket_id = (char *)gIter->data; out->message(out, "inject-modify-ticket", "Standby", ticket_id); rc = set_ticket_state_attr(ticket_id, "standby", "true", cib, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } for (gIter = injections->ticket_activate; gIter != NULL; gIter = gIter->next) { char *ticket_id = (char *)gIter->data; out->message(out, "inject-modify-ticket", "Activating", ticket_id); rc = set_ticket_state_attr(ticket_id, "standby", "false", cib, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } for (gIter = injections->op_inject; gIter != NULL; gIter = gIter->next) { char *spec = (char *)gIter->data; int rc = 0; - int outcome = 0; + int outcome = PCMK_OCF_OK; guint interval_ms = 0; char *key = NULL; char *node = NULL; char *task = NULL; char *resource = NULL; const char *rtype = NULL; const char *rclass = NULL; const char *rprovider = NULL; pe_resource_t *rsc = NULL; out->message(out, "inject-spec", spec); key = calloc(1, strlen(spec) + 1); node = calloc(1, strlen(spec) + 1); rc = sscanf(spec, "%[^@]@%[^=]=%d", key, node, &outcome); if (rc != 3) { out->err(out, "Invalid operation spec: %s. Only found %d fields", spec, rc); free(key); free(node); continue; } parse_op_key(key, &resource, &task, &interval_ms); rsc = pe_find_resource(data_set->resources, resource); if (rsc == NULL) { out->err(out, "Invalid resource name: %s", resource); } else { rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); rtype = crm_element_value(rsc->xml, XML_ATTR_TYPE); rprovider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER); cib_node = inject_node_state(cib, node, NULL); CRM_ASSERT(cib_node != NULL); update_failcounts(cib_node, resource, task, interval_ms, outcome); cib_resource = inject_resource(cib_node, resource, resource, rclass, rtype, rprovider); CRM_ASSERT(cib_resource != NULL); op = create_op(cib_resource, task, interval_ms, outcome); CRM_ASSERT(op != NULL); cib_op = inject_op(cib_resource, op, 0); CRM_ASSERT(cib_op != NULL); lrmd_free_event(op); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } free(task); free(node); free(key); } if (!out->is_quiet(out)) { out->end_list(out); } } static gboolean exec_pseudo_action(crm_graph_t * graph, crm_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); action->confirmed = TRUE; out->message(out, "inject-pseudo-action", node, task); pcmk__update_graph(graph, action); return TRUE; } static gboolean exec_rsc_action(crm_graph_t * graph, crm_action_t * action) { int rc = 0; GList *gIter = NULL; lrmd_event_data_t *op = NULL; - int target_outcome = 0; + int target_outcome = PCMK_OCF_OK; const char *rtype = NULL; const char *rclass = NULL; const char *resource = NULL; const char *rprovider = NULL; const char *lrm_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 = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET_UUID); const char *router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (pcmk__strcase_any_of(operation, CRM_OP_PROBED, CRM_OP_REPROBE, NULL)) { crm_info("Skipping %s op for %s", operation, node); goto done; } if (action_rsc == NULL) { crm_log_xml_err(action->xml, "Bad"); free(node); free(uuid); return FALSE; } /* Look for the preferred name * If not found, try the expected 'local' name * If not found use the preferred name anyway */ resource = crm_element_value(action_rsc, XML_ATTR_ID); CRM_ASSERT(resource != NULL); // makes static analysis happy lrm_name = resource; // Preferred name when writing history if (pe_find_resource(fake_resource_list, resource) == NULL) { const char *longname = crm_element_value(action_rsc, XML_ATTR_ID_LONG); if (longname && pe_find_resource(fake_resource_list, longname)) { resource = longname; } } if (pcmk__strcase_any_of(operation, "delete", RSC_METADATA, NULL)) { out->message(out, "inject-rsc-action", resource, operation, node, (guint) 0); goto done; } 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); cib_node = inject_node_state(fake_cib, node, (router_node? node : uuid)); CRM_ASSERT(cib_node != NULL); cib_resource = inject_resource(cib_node, resource, lrm_name, rclass, rtype, rprovider); if (cib_resource == NULL) { crm_err("invalid resource in transition"); free(node); free(uuid); free_xml(cib_node); return FALSE; } op = pcmk__event_from_graph_action(cib_resource, action, PCMK_EXEC_DONE, - target_outcome); + target_outcome, "User-injected result"); out->message(out, "inject-rsc-action", resource, op->op_type, node, op->interval_ms); for (gIter = fake_op_fail_list; gIter != NULL; gIter = gIter->next) { char *spec = (char *)gIter->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 ((match_name == NULL) && strcmp(resource, lrm_name)) { key = crm_strdup_printf(PCMK__OP_FMT "@%s=", lrm_name, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = lrm_name; } free(key); } if (match_name != NULL) { rc = sscanf(spec, "%*[^=]=%d", (int *) &op->rc); // ${match_name}_${task}_${interval_in_ms}@${node}=${rc} if (rc != 1) { out->err(out, "Invalid failed operation spec: %s. Result code must be integer", spec); continue; } action->failed = TRUE; graph->abort_priority = INFINITY; out->info(out, "Pretending action %d failed with rc=%d", action->id, op->rc); update_failcounts(cib_node, match_name, op->op_type, op->interval_ms, op->rc); break; } } inject_op(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(uuid); free_xml(cib_node); action->confirmed = TRUE; pcmk__update_graph(graph, action); return TRUE; } static gboolean exec_crmd_action(crm_graph_t * graph, crm_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); action->confirmed = TRUE; out->message(out, "inject-cluster-action", node, task, rsc); pcmk__update_graph(graph, action); return TRUE; } static gboolean exec_stonith_action(crm_graph_t * graph, crm_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 = 0; char xpath[STATUS_PATH_MAX]; xmlNode *cib_node = modify_node(fake_cib, target, FALSE); crm_xml_add(cib_node, XML_ATTR_ORIGIN, __func__); CRM_ASSERT(cib_node != NULL); rc = fake_cib->cmds->replace(fake_cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call | cib_scope_local); CRM_ASSERT(rc == pcmk_ok); snprintf(xpath, STATUS_PATH_MAX, "//node_state[@uname='%s']/%s", target, XML_CIB_TAG_LRM); fake_cib->cmds->remove(fake_cib, xpath, NULL, cib_xpath | cib_sync_call | cib_scope_local); snprintf(xpath, STATUS_PATH_MAX, "//node_state[@uname='%s']/%s", target, XML_TAG_TRANSIENT_NODEATTRS); fake_cib->cmds->remove(fake_cib, xpath, NULL, cib_xpath | cib_sync_call | cib_scope_local); free_xml(cib_node); } action->confirmed = TRUE; pcmk__update_graph(graph, action); free(target); return TRUE; } enum transition_status run_simulation(pe_working_set_t * data_set, cib_t *cib, GList *op_fail_list) { crm_graph_t *transition = NULL; enum transition_status graph_rc; crm_graph_functions_t exec_fns = { exec_pseudo_action, exec_rsc_action, exec_crmd_action, exec_stonith_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(&exec_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 == transition_active); fake_resource_list = NULL; if (graph_rc != transition_complete) { out->err(out, "Transition failed: %s", pcmk__graph_status2text(graph_rc)); pcmk__log_graph(LOG_ERR, transition); } pcmk__free_graph(transition); if (graph_rc != transition_complete) { out->err(out, "An invalid transition was produced"); } if (!out->is_quiet(out)) { 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; } diff --git a/lib/pacemaker/pcmk_sched_utils.c b/lib/pacemaker/pcmk_sched_utils.c index 39b13ffe9c..8e764ea630 100644 --- a/lib/pacemaker/pcmk_sched_utils.c +++ b/lib/pacemaker/pcmk_sched_utils.c @@ -1,828 +1,828 @@ /* * Copyright 2004-2021 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 #include "libpacemaker_private.h" gboolean can_run_resources(const pe_node_t * node) { if (node == NULL) { return FALSE; } #if 0 if (node->weight < 0) { return FALSE; } #endif if (node->details->online == FALSE || node->details->shutdown || node->details->unclean || node->details->standby || node->details->maintenance) { crm_trace("%s: online=%d, unclean=%d, standby=%d, maintenance=%d", node->details->uname, node->details->online, node->details->unclean, node->details->standby, node->details->maintenance); 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; } struct node_weight_s { pe_node_t *active; pe_working_set_t *data_set; }; /* return -1 if 'a' is more preferred * return 1 if 'b' is more preferred */ static gint sort_node_weight(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; struct node_weight_s *nw = data; int node1_weight = 0; int node2_weight = 0; int result = 0; if (a == NULL) { return 1; } if (b == NULL) { return -1; } node1_weight = node1->weight; node2_weight = node2->weight; if (can_run_resources(node1) == FALSE) { node1_weight = -INFINITY; } if (can_run_resources(node2) == FALSE) { node2_weight = -INFINITY; } if (node1_weight > node2_weight) { crm_trace("%s (%d) > %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return -1; } if (node1_weight < node2_weight) { crm_trace("%s (%d) < %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return 1; } crm_trace("%s (%d) == %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); if (pcmk__str_eq(nw->data_set->placement_strategy, "minimal", pcmk__str_casei)) { goto equal; } if (pcmk__str_eq(nw->data_set->placement_strategy, "balanced", pcmk__str_casei)) { result = compare_capacity(node1, node2); if (result < 0) { crm_trace("%s > %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return -1; } else if (result > 0) { crm_trace("%s < %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return 1; } } /* now try to balance resources across the cluster */ if (node1->details->num_resources < node2->details->num_resources) { crm_trace("%s (%d) > %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (node1->details->num_resources > node2->details->num_resources) { crm_trace("%s (%d) < %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } if (nw->active && nw->active->details == node1->details) { crm_trace("%s (%d) > %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (nw->active && nw->active->details == node2->details) { crm_trace("%s (%d) < %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } equal: crm_trace("%s = %s", node1->details->uname, node2->details->uname); return strcmp(node1->details->uname, node2->details->uname); } GList * sort_nodes_by_weight(GList *nodes, pe_node_t *active_node, pe_working_set_t *data_set) { struct node_weight_s nw = { active_node, data_set }; return g_list_sort_with_data(nodes, sort_node_weight, &nw); } void native_deallocate(pe_resource_t * rsc) { if (rsc->allocated_to) { pe_node_t *old = rsc->allocated_to; crm_info("Deallocating %s from %s", rsc->id, old->details->uname); pe__set_resource_flags(rsc, pe_rsc_provisional); rsc->allocated_to = NULL; old->details->allocated_rsc = g_list_remove(old->details->allocated_rsc, rsc); old->details->num_resources--; /* old->count--; */ calculate_utilization(old->details->utilization, rsc->utilization, TRUE); free(old); } } gboolean native_assign_node(pe_resource_t *rsc, pe_node_t *chosen, gboolean force) { pcmk__output_t *out = rsc->cluster->priv; CRM_ASSERT(rsc->variant == pe_native); if (force == FALSE && chosen != NULL) { bool unset = FALSE; if(chosen->weight < 0) { unset = TRUE; // Allow the graph to assume that the remote resource will come up } else if (!can_run_resources(chosen) && !pe__is_guest_node(chosen)) { unset = TRUE; } if(unset) { crm_debug("All nodes for resource %s are unavailable" ", unclean or shutting down (%s: %d, %d)", rsc->id, chosen->details->uname, can_run_resources(chosen), chosen->weight); pe__set_next_role(rsc, RSC_ROLE_STOPPED, "node availability"); chosen = NULL; } } /* todo: update the old node for each resource to reflect its * new resource count */ native_deallocate(rsc); pe__clear_resource_flags(rsc, pe_rsc_provisional); if (chosen == NULL) { GList *gIter = NULL; char *rc_inactive = pcmk__itoa(PCMK_OCF_NOT_RUNNING); crm_debug("Could not allocate a node for %s", rsc->id); pe__set_next_role(rsc, RSC_ROLE_STOPPED, "unable to allocate"); for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; const char *interval_ms_s = g_hash_table_lookup(op->meta, XML_LRM_ATTR_INTERVAL_MS); crm_debug("Processing %s", op->uuid); if(pcmk__str_eq(RSC_STOP, op->task, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_optional); } else if(pcmk__str_eq(RSC_START, op->task, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_runnable); //pe__set_resource_flags(rsc, pe_rsc_block); } else if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { if(pcmk__str_eq(rc_inactive, g_hash_table_lookup(op->meta, XML_ATTR_TE_TARGET_RC), pcmk__str_casei)) { /* This is a recurring monitor for the stopped state, leave it alone */ } else { /* Normal monitor operation, cancel it */ pe__clear_action_flags(op, pe_action_runnable); } } } free(rc_inactive); return FALSE; } crm_debug("Assigning %s to %s", chosen->details->uname, rsc->id); 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++; calculate_utilization(chosen->details->utilization, rsc->utilization, FALSE); if (pcmk_is_set(rsc->cluster->flags, pe_flag_show_utilization)) { out->message(out, "resource-util", rsc, chosen, __func__); } return TRUE; } void log_action(unsigned int log_level, const char *pre_text, pe_action_t * action, gboolean details) { const char *node_uname = NULL; const char *node_uuid = NULL; const char *desc = NULL; if (action == NULL) { crm_trace("%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": "); return; } if (pcmk_is_set(action->flags, pe_action_pseudo)) { node_uname = NULL; node_uuid = NULL; } else if (action->node != NULL) { node_uname = action->node->details->uname; node_uuid = action->node->details->id; } else { node_uname = ""; node_uuid = NULL; } 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) { GList *gIter = NULL; crm_trace("\t\t====== Preceding Actions"); gIter = action->actions_before; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\t\t", other->action, FALSE); } crm_trace("\t\t====== Subsequent Actions"); gIter = action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\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)); } } gboolean can_run_any(GHashTable * nodes) { GHashTableIter iter; 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 (can_run_resources(node) && node->weight >= 0) { return TRUE; } } return FALSE; } pe_action_t * create_pseudo_resource_op(pe_resource_t * rsc, const char *task, bool optional, bool runnable, pe_working_set_t *data_set) { pe_action_t *action = custom_action(rsc, pcmk__op_key(rsc->id, task, 0), task, NULL, optional, TRUE, data_set); pe__set_action_flags(action, pe_action_pseudo); if(runnable) { pe__set_action_flags(action, pe_action_runnable); } return action; } /*! * \internal * \brief Create an executor cancel op * * \param[in] rsc Resource of action to cancel * \param[in] task Name of action to cancel * \param[in] interval_ms Interval of action to cancel * \param[in] node Node of action to cancel * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * pe_cancel_op(pe_resource_t *rsc, const char *task, guint interval_ms, pe_node_t *node, pe_working_set_t *data_set) { pe_action_t *cancel_op; char *interval_ms_s = crm_strdup_printf("%u", interval_ms); // @TODO dangerous if possible to schedule another action with this key char *key = pcmk__op_key(rsc->id, task, interval_ms); cancel_op = custom_action(rsc, key, RSC_CANCEL, node, FALSE, TRUE, data_set); free(cancel_op->task); cancel_op->task = strdup(RSC_CANCEL); free(cancel_op->cancel_task); cancel_op->cancel_task = strdup(task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_TASK, task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_INTERVAL_MS, interval_ms_s); free(interval_ms_s); return cancel_op; } /*! * \internal * \brief Create a shutdown op for a scheduler transition * * \param[in] node Node being shut down * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * sched_shutdown_op(pe_node_t *node, pe_working_set_t *data_set) { char *shutdown_id = crm_strdup_printf("%s-%s", CRM_OP_SHUTDOWN, node->details->uname); pe_action_t *shutdown_op = custom_action(NULL, shutdown_id, CRM_OP_SHUTDOWN, node, FALSE, TRUE, data_set); pcmk__order_stops_before_shutdown(node, shutdown_op, data_set); add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); return shutdown_op; } static char * generate_transition_magic(const char *transition_key, int op_status, int op_rc) { CRM_CHECK(transition_key != NULL, return NULL); return crm_strdup_printf("%d:%d;%s", op_status, op_rc, transition_key); } static void append_digest(lrmd_event_data_t *op, xmlNode *update, const char *version, const char *magic, int level) { /* this will enable us to later determine that the * resource's parameters have changed and we should force * a restart */ 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, version); #if 0 if (level < get_crm_log_level() && op->interval_ms == 0 && pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { char *digest_source = dump_xml_unformatted(args_xml); do_crm_log(level, "Calculated digest %s for %s (%s). Source: %s\n", digest, ID(update), magic, digest_source); free(digest_source); } #endif 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 * \param[in] level A log message will be logged at this level * * \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, int level) { 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); do_crm_log(level, "%s: Updating resource %s after %s op %s (interval=%u)", origin, op->rsc_id, op->op_type, pcmk_exec_status_str(op->op_status), op->interval_ms); crm_trace("DC version: %s", caller_version); 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(). */ - op->op_status = PCMK_EXEC_DONE; - op->rc = 0; + lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } } 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 = generate_transition_magic(op->user_data, op->op_status, op->rc); } 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)); } append_digest(op, xml_op, caller_version, magic, LOG_DEBUG); 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; } pcmk__output_t * pcmk__new_logger(void) { int rc = pcmk_rc_ok; pcmk__output_t *out = NULL; const char* argv[] = { "", NULL }; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_LOG, { NULL, NULL, NULL } }; pcmk__register_formats(NULL, formats); rc = pcmk__output_new(&out, "log", NULL, (char**)argv); if ((rc != pcmk_rc_ok) || (out == NULL)) { crm_err("Can't log resource details due to internal error: %s\n", pcmk_rc_str(rc)); return NULL; } pe__register_messages(out); pcmk__register_lib_messages(out); return out; } /*! * \internal * \brief Check whether a resource has reached its migration threshold on a node * * \param[in] rsc Resource to check * \param[in] node Node to check * \param[in] data_set Cluster working set * \param[out] failed If the threshold has been reached, this will be set to * the 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, pe_node_t *node, pe_working_set_t *data_set, 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, data_set); 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, node->details->uname, 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), node->details->uname, rsc->migration_threshold); return false; } void pcmk_free_injections(pcmk_injections_t *injections) { if (injections == NULL) { return; } g_list_free_full(injections->node_up, g_free); g_list_free_full(injections->node_down, g_free); g_list_free_full(injections->node_fail, g_free); g_list_free_full(injections->op_fail, g_free); g_list_free_full(injections->op_inject, g_free); g_list_free_full(injections->ticket_grant, g_free); g_list_free_full(injections->ticket_revoke, g_free); g_list_free_full(injections->ticket_standby, g_free); g_list_free_full(injections->ticket_activate, g_free); free(injections->quorum); free(injections->watchdog); free(injections); } diff --git a/lib/pengine/pe_output.c b/lib/pengine/pe_output.c index 2fe537598d..8dc733b5bf 100644 --- a/lib/pengine/pe_output.c +++ b/lib/pengine/pe_output.c @@ -1,2800 +1,2800 @@ /* * Copyright 2019-2021 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 /* Never display node attributes whose name starts with one of these prefixes */ #define FILTER_STR { PCMK__FAIL_COUNT_PREFIX, PCMK__LAST_FAILURE_PREFIX, \ "shutdown", "terminate", "standby", "probe_complete", \ "#", NULL } static int compare_attribute(gconstpointer a, gconstpointer b) { int rc; rc = strcmp((const char *)a, (const char *)b); return rc; } /*! * \internal * \brief Determine whether extended information about an attribute should be added. * * \param[in] node Node that ran this resource. * \param[in] rsc_list The list of resources for this node. * \param[in] attrname The attribute to find. * \param[out] expected_score The expected value for this attribute. * * \return TRUE if extended information should be printed, FALSE otherwise * \note Currently, extended information is only supported for ping/pingd * resources, for which a message will be printed if connectivity is lost * or degraded. */ static gboolean add_extra_info(pe_node_t *node, GList *rsc_list, pe_working_set_t *data_set, const char *attrname, int *expected_score) { GList *gIter = NULL; for (gIter = rsc_list; gIter != NULL; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; const char *type = g_hash_table_lookup(rsc->meta, "type"); const char *name = NULL; GHashTable *params = NULL; if (rsc->children != NULL) { if (add_extra_info(node, rsc->children, data_set, attrname, expected_score)) { return TRUE; } } if (!pcmk__strcase_any_of(type, "ping", "pingd", NULL)) { continue; } params = pe_rsc_params(rsc, node, data_set); name = g_hash_table_lookup(params, "name"); if (name == NULL) { name = "pingd"; } /* To identify the resource with the attribute name. */ if (pcmk__str_eq(name, attrname, pcmk__str_casei)) { int host_list_num = 0; const char *hosts = g_hash_table_lookup(params, "host_list"); const char *multiplier = g_hash_table_lookup(params, "multiplier"); int multiplier_i; if (hosts) { char **host_list = g_strsplit(hosts, " ", 0); host_list_num = g_strv_length(host_list); g_strfreev(host_list); } if ((multiplier == NULL) || (pcmk__scan_min_int(multiplier, &multiplier_i, INT_MIN) != pcmk_rc_ok)) { /* The ocf:pacemaker:ping resource agent defaults multiplier to * 1. The agent currently does not handle invalid text, but it * should, and this would be a reasonable choice ... */ multiplier_i = 1; } *expected_score = host_list_num * multiplier_i; return TRUE; } } return FALSE; } static GList * filter_attr_list(GList *attr_list, char *name) { int i; const char *filt_str[] = FILTER_STR; CRM_CHECK(name != NULL, return attr_list); /* filtering automatic attributes */ for (i = 0; filt_str[i] != NULL; i++) { if (g_str_has_prefix(name, filt_str[i])) { return attr_list; } } return g_list_insert_sorted(attr_list, name, compare_attribute); } static GList * get_operation_list(xmlNode *rsc_entry) { GList *op_list = NULL; xmlNode *rsc_op = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op)) { const char *task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(rsc_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC); int op_rc_i; pcmk__scan_min_int(op_rc, &op_rc_i, 0); /* Display 0-interval monitors as "probe" */ if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei) && pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches | pcmk__str_casei)) { task = "probe"; } /* Ignore notifies and some probes */ if (pcmk__str_eq(task, CRMD_ACTION_NOTIFY, pcmk__str_casei) || (pcmk__str_eq(task, "probe", pcmk__str_casei) && (op_rc_i == 7))) { continue; } if (pcmk__str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, pcmk__str_none)) { op_list = g_list_append(op_list, rsc_op); } } op_list = g_list_sort(op_list, sort_op_by_callid); return op_list; } static void add_dump_node(gpointer key, gpointer value, gpointer user_data) { xmlNodePtr node = user_data; pcmk_create_xml_text_node(node, (const char *) key, (const char *) value); } static void append_dump_text(gpointer key, gpointer value, gpointer user_data) { char **dump_text = user_data; char *new_text = crm_strdup_printf("%s %s=%s", *dump_text, (char *)key, (char *)value); free(*dump_text); *dump_text = new_text; } static const char * get_cluster_stack(pe_working_set_t *data_set) { xmlNode *stack = get_xpath_object("//nvpair[@name='cluster-infrastructure']", data_set->input, LOG_DEBUG); return stack? crm_element_value(stack, XML_NVPAIR_ATTR_VALUE) : "unknown"; } static char * last_changed_string(const char *last_written, const char *user, const char *client, const char *origin) { if (last_written != NULL || user != NULL || client != NULL || origin != NULL) { return crm_strdup_printf("%s%s%s%s%s%s%s", last_written ? last_written : "", user ? " by " : "", user ? user : "", client ? " via " : "", client ? client : "", origin ? " on " : "", origin ? origin : ""); } else { return strdup(""); } } static char * op_history_string(xmlNode *xml_op, const char *task, const char *interval_ms_s, int rc, gboolean print_timing) { const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); char *interval_str = NULL; char *buf = NULL; if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *pair = pcmk__format_nvpair("interval", interval_ms_s, "ms"); interval_str = crm_strdup_printf(" %s", pair); free(pair); } if (print_timing) { char *last_change_str = NULL; char *exec_str = NULL; char *queue_str = NULL; const char *value = NULL; time_t epoch = 0; if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { char *time = pcmk__format_named_time(XML_RSC_OP_LAST_CHANGE, epoch); last_change_str = crm_strdup_printf(" %s", time); free(time); } value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); if (value) { char *pair = pcmk__format_nvpair(XML_RSC_OP_T_EXEC, value, "ms"); exec_str = crm_strdup_printf(" %s", pair); free(pair); } value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); if (value) { char *pair = pcmk__format_nvpair(XML_RSC_OP_T_QUEUE, value, "ms"); queue_str = crm_strdup_printf(" %s", pair); free(pair); } buf = crm_strdup_printf("(%s) %s:%s%s%s%s rc=%d (%s)", call, task, interval_str ? interval_str : "", last_change_str ? last_change_str : "", exec_str ? exec_str : "", queue_str ? queue_str : "", rc, services_ocf_exitcode_str(rc)); if (last_change_str) { free(last_change_str); } if (exec_str) { free(exec_str); } if (queue_str) { free(queue_str); } } else { buf = crm_strdup_printf("(%s) %s%s%s", call, task, interval_str ? ":" : "", interval_str ? interval_str : ""); } if (interval_str) { free(interval_str); } return buf; } static char * resource_history_string(pe_resource_t *rsc, const char *rsc_id, gboolean all, int failcount, time_t last_failure) { char *buf = NULL; if (rsc == NULL) { buf = crm_strdup_printf("%s: orphan", rsc_id); } else if (all || failcount || last_failure > 0) { char *failcount_s = NULL; char *lastfail_s = NULL; if (failcount > 0) { failcount_s = crm_strdup_printf(" %s=%d", PCMK__FAIL_COUNT_PREFIX, failcount); } else { failcount_s = strdup(""); } if (last_failure > 0) { lastfail_s = crm_strdup_printf(" %s='%s'", PCMK__LAST_FAILURE_PREFIX, pcmk__epoch2str(&last_failure)); } buf = crm_strdup_printf("%s: migration-threshold=%d%s%s", rsc_id, rsc->migration_threshold, failcount_s, lastfail_s? lastfail_s : ""); free(failcount_s); free(lastfail_s); } else { buf = crm_strdup_printf("%s:", rsc_id); } return buf; } PCMK__OUTPUT_ARGS("cluster-summary", "pe_working_set_t *", "unsigned int", "unsigned int") static int cluster_summary(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); unsigned int section_opts = va_arg(args, unsigned int); unsigned int show_opts = va_arg(args, unsigned int); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(data_set); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s); } if (pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']", data_set->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE) : NULL; const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM); char *dc_name = data_set->dc_node ? pe__node_display_name(data_set->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-dc", data_set->dc_node, quorum, dc_version_s, dc_name); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN); const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER); const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT); const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG); PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-times", last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(data_set->nodes), data_set->ninstances, data_set->disabled_resources, data_set->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-options", data_set); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", data_set->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } PCMK__OUTPUT_ARGS("cluster-summary", "pe_working_set_t *", "unsigned int", "unsigned int") static int cluster_summary_html(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); unsigned int section_opts = va_arg(args, unsigned int); unsigned int show_opts = va_arg(args, unsigned int); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(data_set); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s); } /* Always print DC if none, even if not requested */ if (data_set->dc_node == NULL || pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']", data_set->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE) : NULL; const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM); char *dc_name = data_set->dc_node ? pe__node_display_name(data_set->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-dc", data_set->dc_node, quorum, dc_version_s, dc_name); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN); const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER); const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT); const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG); PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-times", last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(data_set->nodes), data_set->ninstances, data_set->disabled_resources, data_set->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { /* Kind of a hack - close the list we may have opened earlier in this * function so we can put all the options into their own list. We * only want to do this on HTML output, though. */ PCMK__OUTPUT_LIST_FOOTER(out, rc); out->begin_list(out, NULL, NULL, "Config Options"); out->message(out, "cluster-options", data_set); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", data_set->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } char * pe__node_display_name(pe_node_t *node, bool print_detail) { char *node_name; const char *node_host = NULL; const char *node_id = NULL; int name_len; CRM_ASSERT((node != NULL) && (node->details != NULL) && (node->details->uname != NULL)); /* Host is displayed only if this is a guest node */ if (pe__is_guest_node(node)) { pe_node_t *host_node = pe__current_node(node->details->remote_rsc); if (host_node && host_node->details) { node_host = host_node->details->uname; } if (node_host == NULL) { node_host = ""; /* so we at least get "uname@" to indicate guest */ } } /* Node ID is displayed if different from uname and detail is requested */ if (print_detail && !pcmk__str_eq(node->details->uname, node->details->id, pcmk__str_casei)) { node_id = node->details->id; } /* Determine name length */ name_len = strlen(node->details->uname) + 1; if (node_host) { name_len += strlen(node_host) + 1; /* "@node_host" */ } if (node_id) { name_len += strlen(node_id) + 3; /* + " (node_id)" */ } /* Allocate and populate display name */ node_name = malloc(name_len); CRM_ASSERT(node_name != NULL); strcpy(node_name, node->details->uname); if (node_host) { strcat(node_name, "@"); strcat(node_name, node_host); } if (node_id) { strcat(node_name, " ("); strcat(node_name, node_id); strcat(node_name, ")"); } return node_name; } int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name , size_t pairs_count, ...) { xmlNodePtr xml_node = NULL; va_list args; CRM_ASSERT(tag_name != NULL); xml_node = pcmk__output_xml_peek_parent(out); CRM_ASSERT(xml_node != NULL); xml_node = is_list ? create_xml_node(xml_node, tag_name) : xmlNewChild(xml_node, NULL, (pcmkXmlStr) tag_name, NULL); va_start(args, pairs_count); while(pairs_count--) { const char *param_name = va_arg(args, const char *); const char *param_value = va_arg(args, const char *); if (param_name && param_value) { crm_xml_add(xml_node, param_name, param_value); } }; va_end(args); if (is_list) { pcmk__output_xml_push_parent(out, xml_node); } return pcmk_rc_ok; } static const char * role_desc(enum rsc_role_e role) { if (role == RSC_ROLE_PROMOTED) { #ifdef PCMK__COMPAT_2_0 return "as " RSC_ROLE_PROMOTED_LEGACY_S " "; #else return "in " RSC_ROLE_PROMOTED_S " role "; #endif } return ""; } PCMK__OUTPUT_ARGS("ban", "pe_node_t *", "pe__location_t *", "unsigned int") static int ban_html(pcmk__output_t *out, va_list args) { pe_node_t *pe_node = va_arg(args, pe_node_t *); pe__location_t *location = va_arg(args, pe__location_t *); unsigned int show_opts = va_arg(args, unsigned int); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); char *buf = crm_strdup_printf("%s\tprevents %s from running %son %s", location->id, location->rsc_lh->id, role_desc(location->role_filter), node_name); pcmk__output_create_html_node(out, "li", NULL, NULL, buf); free(node_name); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pe_node_t *", "pe__location_t *", "unsigned int") static int ban_text(pcmk__output_t *out, va_list args) { pe_node_t *pe_node = va_arg(args, pe_node_t *); pe__location_t *location = va_arg(args, pe__location_t *); unsigned int show_opts = va_arg(args, unsigned int); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->list_item(out, NULL, "%s\tprevents %s from running %son %s", location->id, location->rsc_lh->id, role_desc(location->role_filter), node_name); free(node_name); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pe_node_t *", "pe__location_t *", "unsigned int") static int ban_xml(pcmk__output_t *out, va_list args) { pe_node_t *pe_node = va_arg(args, pe_node_t *); pe__location_t *location = va_arg(args, pe__location_t *); unsigned int show_opts G_GNUC_UNUSED = va_arg(args, unsigned int); const char *promoted_only = pcmk__btoa(location->role_filter == RSC_ROLE_PROMOTED); char *weight_s = pcmk__itoa(pe_node->weight); pcmk__output_create_xml_node(out, "ban", "id", location->id, "resource", location->rsc_lh->id, "node", pe_node->details->uname, "weight", weight_s, "promoted-only", promoted_only, /* This is a deprecated alias for * promoted_only. Removing it will break * backward compatibility of the API schema, * which will require an API schema major * version bump. */ "master_only", promoted_only, NULL); free(weight_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban-list", "pe_working_set_t *", "const char *", "GList *", "unsigned int", "gboolean") static int ban_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); const char *prefix = va_arg(args, const char *); GList *only_rsc = va_arg(args, GList *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer = va_arg(args, gboolean); GList *gIter, *gIter2; int rc = pcmk_rc_no_output; /* Print each ban */ for (gIter = data_set->placement_constraints; gIter != NULL; gIter = gIter->next) { pe__location_t *location = gIter->data; if (prefix != NULL && !g_str_has_prefix(location->id, prefix)) { continue; } if (!pcmk__str_in_list(rsc_printable_id(location->rsc_lh), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(uber_parent(location->rsc_lh)), only_rsc, pcmk__str_star_matches)) { continue; } for (gIter2 = location->node_list_rh; gIter2 != NULL; gIter2 = gIter2->next) { pe_node_t *node = (pe_node_t *) gIter2->data; if (node->weight < 0) { PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Negative Location Constraints"); out->message(out, "ban", node, location, show_opts); } } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_html(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "li", NULL); char *nnodes_str = crm_strdup_printf("%d node%s configured", nnodes, pcmk__plural_s(nnodes)); pcmk_create_html_node(nodes_node, "span", NULL, NULL, nnodes_str); free(nnodes_str); if (ndisabled && nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED"); s = crm_strdup_printf(", %d ", nblocked); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, " from further action due to failure)"); } else if (ndisabled && !nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, ")"); } else if (!ndisabled && nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), nblocked); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, " from further action due to failure)"); } else { char *s = crm_strdup_printf("%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_text(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); out->list_item(out, NULL, "%d node%s configured", nnodes, pcmk__plural_s(nnodes)); if (ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED, %d BLOCKED from " "further action due to failure)", nresources, pcmk__plural_s(nresources), ndisabled, nblocked); } else if (ndisabled && !nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED)", nresources, pcmk__plural_s(nresources), ndisabled); } else if (!ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d BLOCKED from further action " "due to failure)", nresources, pcmk__plural_s(nresources), nblocked); } else { out->list_item(out, NULL, "%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_xml(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "nodes_configured", NULL); xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "resources_configured", NULL); char *s = pcmk__itoa(nnodes); crm_xml_add(nodes_node, "number", s); free(s); s = pcmk__itoa(nresources); crm_xml_add(resources_node, "number", s); free(s); s = pcmk__itoa(ndisabled); crm_xml_add(resources_node, "disabled", s); free(s); s = pcmk__itoa(nblocked); crm_xml_add(resources_node, "blocked", s); free(s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pe_node_t *", "const char *", "const char *", "char *") static int cluster_dc_html(pcmk__output_t *out, va_list args) { pe_node_t *dc = va_arg(args, pe_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, "bold", "Current DC: "); if (dc) { if (crm_is_true(quorum)) { char *buf = crm_strdup_printf("%s (version %s) - partition with quorum", dc_name, dc_version_s ? dc_version_s : "unknown"); pcmk_create_html_node(node, "span", NULL, NULL, buf); free(buf); } else { char *buf = crm_strdup_printf("%s (version %s) - partition", dc_name, dc_version_s ? dc_version_s : "unknown"); pcmk_create_html_node(node, "span", NULL, NULL, buf); free(buf); pcmk_create_html_node(node, "span", NULL, "warning", "WITHOUT"); pcmk_create_html_node(node, "span", NULL, NULL, "quorum"); } } else { pcmk_create_html_node(node ,"span", NULL, "warning", "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pe_node_t *", "const char *", "const char *", "char *") static int cluster_dc_text(pcmk__output_t *out, va_list args) { pe_node_t *dc = va_arg(args, pe_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); if (dc) { out->list_item(out, "Current DC", "%s (version %s) - partition %s quorum", dc_name, dc_version_s ? dc_version_s : "unknown", crm_is_true(quorum) ? "with" : "WITHOUT"); } else { out->list_item(out, "Current DC", "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pe_node_t *", "const char *", "const char *", "char *") static int cluster_dc_xml(pcmk__output_t *out, va_list args) { pe_node_t *dc = va_arg(args, pe_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name G_GNUC_UNUSED = va_arg(args, char *); if (dc) { pcmk__output_create_xml_node(out, "current_dc", "present", "true", "version", dc_version_s ? dc_version_s : "", "name", dc->details->uname, "id", dc->details->id, "with_quorum", pcmk__btoa(crm_is_true(quorum)), NULL); } else { pcmk__output_create_xml_node(out, "current_dc", "present", "false", NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("maint-mode", "unsigned long long int") static int cluster_maint_mode_text(pcmk__output_t *out, va_list args) { unsigned long long flags = va_arg(args, unsigned long long); if (pcmk_is_set(flags, pe_flag_maintenance_mode)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will not attempt to start, stop or recover services\n"); return pcmk_rc_ok; } else if (pcmk_is_set(flags, pe_flag_stop_everything)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will keep all resources stopped\n"); return pcmk_rc_ok; } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pe_working_set_t *") static int cluster_options_html(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); out->list_item(out, NULL, "STONITH of failed nodes %s", pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled"); out->list_item(out, NULL, "Cluster is %s", pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster) ? "symmetric" : "asymmetric"); switch (data_set->no_quorum_policy) { case no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case no_quorum_suicide: out->list_item(out, NULL, "No quorum policy: Suicide"); break; } if (pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, NULL, "Resource management: "); pcmk_create_html_node(node, "span", NULL, "bold", "DISABLED"); pcmk_create_html_node(node, "span", NULL, NULL, " (the cluster will not attempt to start, stop, or recover services)"); } else if (pcmk_is_set(data_set->flags, pe_flag_stop_everything)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, NULL, "Resource management: "); pcmk_create_html_node(node, "span", NULL, "bold", "STOPPED"); pcmk_create_html_node(node, "span", NULL, NULL, " (the cluster will keep all resources stopped)"); } else { out->list_item(out, NULL, "Resource management: enabled"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-options", "pe_working_set_t *") static int cluster_options_log(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); if (pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { return out->info(out, "Resource management is DISABLED. The cluster will not attempt to start, stop or recover services."); } else if (pcmk_is_set(data_set->flags, pe_flag_stop_everything)) { return out->info(out, "Resource management is DISABLED. The cluster has stopped all resources."); } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pe_working_set_t *") static int cluster_options_text(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); out->list_item(out, NULL, "STONITH of failed nodes %s", pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled"); out->list_item(out, NULL, "Cluster is %s", pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster) ? "symmetric" : "asymmetric"); switch (data_set->no_quorum_policy) { case no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case no_quorum_suicide: out->list_item(out, NULL, "No quorum policy: Suicide"); break; } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-options", "pe_working_set_t *") static int cluster_options_xml(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); const char *no_quorum_policy = NULL; char *stonith_timeout_str = pcmk__itoa(data_set->stonith_timeout); char *priority_fencing_delay_str = pcmk__itoa(data_set->priority_fencing_delay * 1000); switch (data_set->no_quorum_policy) { case no_quorum_freeze: no_quorum_policy = "freeze"; break; case no_quorum_stop: no_quorum_policy = "stop"; break; case no_quorum_demote: no_quorum_policy = "demote"; break; case no_quorum_ignore: no_quorum_policy = "ignore"; break; case no_quorum_suicide: no_quorum_policy = "suicide"; break; } pcmk__output_create_xml_node(out, "cluster_options", "stonith-enabled", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)), "symmetric-cluster", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster)), "no-quorum-policy", no_quorum_policy, "maintenance-mode", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)), "stop-all-resources", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_stop_everything)), "stonith-timeout-ms", stonith_timeout_str, "priority-fencing-delay-ms", priority_fencing_delay_str, NULL); free(stonith_timeout_str); free(priority_fencing_delay_str); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *") static int cluster_stack_html(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, "bold", "Stack: "); pcmk_create_html_node(node, "span", NULL, NULL, stack_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *") static int cluster_stack_text(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); out->list_item(out, "Stack", "%s", stack_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *") static int cluster_stack_xml(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); pcmk__output_create_xml_node(out, "stack", "type", stack_s, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *") static int cluster_times_html(pcmk__output_t *out, va_list args) { const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); xmlNodePtr updated_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr changed_node = pcmk__output_create_xml_node(out, "li", NULL); char *buf = last_changed_string(last_written, user, client, origin); pcmk_create_html_node(updated_node, "span", NULL, "bold", "Last updated: "); pcmk_create_html_node(updated_node, "span", NULL, NULL, pcmk__epoch2str(NULL)); pcmk_create_html_node(changed_node, "span", NULL, "bold", "Last change: "); pcmk_create_html_node(changed_node, "span", NULL, NULL, buf); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *") static int cluster_times_xml(pcmk__output_t *out, va_list args) { const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); pcmk__output_create_xml_node(out, "last_update", "time", pcmk__epoch2str(NULL), NULL); pcmk__output_create_xml_node(out, "last_change", "time", last_written ? last_written : "", "user", user ? user : "", "client", client ? client : "", "origin", origin ? origin : "", NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *") static int cluster_times_text(pcmk__output_t *out, va_list args) { const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); char *buf = last_changed_string(last_written, user, client, origin); out->list_item(out, "Last updated", "%s", pcmk__epoch2str(NULL)); out->list_item(out, "Last change", " %s", buf); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action", "xmlNodePtr", "unsigned int") static int failed_action_default(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); unsigned int show_opts G_GNUC_UNUSED = va_arg(args, unsigned int); const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); const char *node_name = crm_element_value(xml_op, XML_ATTR_UNAME); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); const char *call_id = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); const char *queue_time = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); const char *exec_time = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); int rc; int status; const char *exit_status = NULL; const char *lrm_status = NULL; const char *last_change_str = NULL; time_t last_change_epoch = 0; GString *str = NULL; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), &rc, 0); exit_status = services_ocf_exitcode_str(rc); pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); lrm_status = pcmk_exec_status_str(status); if (pcmk__str_empty(op_key)) { op_key = ID(xml_op); if (pcmk__str_empty(op_key)) { op_key = "unknown operation"; } } if (pcmk__str_empty(node_name)) { node_name = "unknown node"; } if (pcmk__str_empty(exit_status)) { exit_status = "unknown exit status"; } if (pcmk__str_empty(call_id)) { call_id = "unknown"; } str = g_string_sized_new(strlen(op_key) + strlen(node_name) + strlen(exit_status) + strlen(call_id) + strlen(lrm_status) + 50); // rough estimate - g_string_printf(str, "%s on %s '%s' (%d): call=%s, status=%s", + g_string_printf(str, "%s on %s '%s' (%d): call=%s, status='%s'", op_key, node_name, exit_status, rc, call_id, lrm_status); if (!pcmk__str_empty(exit_reason)) { g_string_append_printf(str, ", exitreason='%s'", exit_reason); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change_epoch) == pcmk_ok) { last_change_str = pcmk__epoch2str(&last_change_epoch); if (last_change_str != NULL) { g_string_append_printf(str, ", " XML_RSC_OP_LAST_CHANGE "='%s'", last_change_str); } } if (!pcmk__str_empty(queue_time)) { g_string_append_printf(str, ", queued=%sms", queue_time); } if (!pcmk__str_empty(exec_time)) { g_string_append_printf(str, ", exec=%sms", exec_time); } out->list_item(out, NULL, "%s", str->str); g_string_free(str, TRUE); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action", "xmlNodePtr", "unsigned int") static int failed_action_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); unsigned int show_opts G_GNUC_UNUSED = va_arg(args, unsigned int); const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); int rc; int status; const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); time_t epoch = 0; char *rc_s = NULL; char *reason_s = crm_xml_escape(exit_reason ? exit_reason : "none"); xmlNodePtr node = NULL; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), &rc, 0); pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); rc_s = pcmk__itoa(rc); node = pcmk__output_create_xml_node(out, "failure", (op_key == NULL)? "id" : "op_key", (op_key == NULL)? ID(xml_op) : op_key, "node", crm_element_value(xml_op, XML_ATTR_UNAME), "exitstatus", services_ocf_exitcode_str(rc), "exitreason", crm_str(reason_s), "exitcode", rc_s, "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "status", pcmk_exec_status_str(status), NULL); free(rc_s); if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { guint interval_ms = 0; char *s = NULL; crm_time_t *crm_when = crm_time_new_undefined(); char *rc_change = NULL; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); s = pcmk__itoa(interval_ms); crm_time_set_timet(crm_when, &epoch); rc_change = crm_time_as_string(crm_when, crm_time_log_date | crm_time_log_timeofday | crm_time_log_with_timezone); pcmk__xe_set_props(node, XML_RSC_OP_LAST_CHANGE, rc_change, "queued", crm_element_value(xml_op, XML_RSC_OP_T_QUEUE), "exec", crm_element_value(xml_op, XML_RSC_OP_T_EXEC), "interval", s, "task", crm_element_value(xml_op, XML_LRM_ATTR_TASK), NULL); free(s); free(rc_change); crm_time_free(crm_when); } free(reason_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action-list", "pe_working_set_t *", "GList *", "GList *", "unsigned int", "gboolean") static int failed_action_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int show_opts = va_arg(args, gboolean); gboolean print_spacer = va_arg(args, gboolean); xmlNode *xml_op = NULL; int rc = pcmk_rc_no_output; const char *id = NULL; if (xmlChildElementCount(data_set->failed) == 0) { return rc; } for (xml_op = pcmk__xml_first_child(data_set->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { char *rsc = NULL; if (!pcmk__str_in_list(crm_element_value(xml_op, XML_ATTR_UNAME), only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } id = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if (parse_op_key(id ? id : ID(xml_op), &rsc, NULL, NULL) == FALSE) { continue; } if (!pcmk__str_in_list(rsc, only_rsc, pcmk__str_star_matches)) { free(rsc); continue; } free(rsc); PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Failed Resource Actions"); out->message(out, "failed-action", xml_op, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } static void status_node(pe_node_t *node, xmlNodePtr parent) { if (node->details->standby_onfail && node->details->online) { pcmk_create_html_node(parent, "span", NULL, "standby", " standby (on-fail)"); } else if (node->details->standby && node->details->online) { char *s = crm_strdup_printf(" standby%s", node->details->running_rsc ? " (with active resources)" : ""); pcmk_create_html_node(parent, "span", NULL, " standby", s); free(s); } else if (node->details->standby) { pcmk_create_html_node(parent, "span", NULL, "offline", " OFFLINE (standby)"); } else if (node->details->maintenance && node->details->online) { pcmk_create_html_node(parent, "span", NULL, "maint", " maintenance"); } else if (node->details->maintenance) { pcmk_create_html_node(parent, "span", NULL, "offline", " OFFLINE (maintenance)"); } else if (node->details->online) { pcmk_create_html_node(parent, "span", NULL, "online", " online"); } else { pcmk_create_html_node(parent, "span", NULL, "offline", " OFFLINE"); } } PCMK__OUTPUT_ARGS("node", "pe_node_t *", "unsigned int", "gboolean", "const char *", "GList *", "GList *") static int node_html(pcmk__output_t *out, va_list args) { pe_node_t *node = va_arg(args, pe_node_t *); unsigned int show_opts = va_arg(args, unsigned int); gboolean full = va_arg(args, gboolean); const char *node_mode G_GNUC_UNUSED = va_arg(args, const char *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (full) { xmlNodePtr item_node; if (pcmk_all_flags_set(show_opts, pcmk_show_brief | pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); out->begin_list(out, NULL, NULL, "Node: %s", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, NULL, "Status:"); status_node(node, item_node); if (rscs != NULL) { unsigned int new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); } pcmk__output_xml_pop_parent(out); out->end_list(out); } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc2 = NULL; int rc = pcmk_rc_no_output; out->begin_list(out, NULL, NULL, "Node: %s", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, NULL, "Status:"); status_node(node, item_node); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { pe_resource_t *rsc = (pe_resource_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Resources"); out->message(out, crm_map_element_name(rsc->xml), show_opts | pcmk_show_rsc_only, rsc, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); pcmk__output_xml_pop_parent(out); out->end_list(out); } else { char *buf = crm_strdup_printf("Node: %s", node_name); item_node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, NULL, buf); status_node(node, item_node); free(buf); } } else { out->begin_list(out, NULL, NULL, "Node: %s", node_name); } free(node_name); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node", "pe_node_t *", "unsigned int", "gboolean", "const char *", "GList *", "GList *") static int node_text(pcmk__output_t *out, va_list args) { pe_node_t *node = va_arg(args, pe_node_t *); unsigned int show_opts = va_arg(args, unsigned int); gboolean full = va_arg(args, gboolean); const char *node_mode = va_arg(args, const char *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); char *buf = NULL; /* Print the node name and status */ if (pe__is_guest_node(node)) { buf = crm_strdup_printf("GuestNode %s: %s", node_name, node_mode); } else if (pe__is_remote_node(node)) { buf = crm_strdup_printf("RemoteNode %s: %s", node_name, node_mode); } else { buf = crm_strdup_printf("Node %s: %s", node_name, node_mode); } /* If we're grouping by node, print its resources */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (pcmk_is_set(show_opts, pcmk_show_brief)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); if (rscs != NULL) { unsigned int new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "%s", buf); out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); out->end_list(out); g_list_free(rscs); } } else { GList *gIter2 = NULL; out->begin_list(out, NULL, NULL, "%s", buf); out->begin_list(out, NULL, NULL, "Resources"); for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) { pe_resource_t *rsc = (pe_resource_t *) gIter2->data; out->message(out, crm_map_element_name(rsc->xml), show_opts | pcmk_show_rsc_only, rsc, only_node, only_rsc); } out->end_list(out); out->end_list(out); } } else { out->list_item(out, NULL, "%s", buf); } free(buf); free(node_name); } else { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->begin_list(out, NULL, NULL, "Node: %s", node_name); free(node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node", "pe_node_t *", "unsigned int", "gboolean", "const char *", "GList *", "GList *") static int node_xml(pcmk__output_t *out, va_list args) { pe_node_t *node = va_arg(args, pe_node_t *); unsigned int show_opts G_GNUC_UNUSED = va_arg(args, unsigned int); gboolean full = va_arg(args, gboolean); const char *node_mode G_GNUC_UNUSED = va_arg(args, const char *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { const char *node_type = "unknown"; char *length_s = pcmk__itoa(g_list_length(node->details->running_rsc)); switch (node->details->type) { case node_member: node_type = "member"; break; case node_remote: node_type = "remote"; break; case node_ping: node_type = "ping"; break; } pe__name_and_nvpairs_xml(out, true, "node", 13, "name", node->details->uname, "id", node->details->id, "online", pcmk__btoa(node->details->online), "standby", pcmk__btoa(node->details->standby), "standby_onfail", pcmk__btoa(node->details->standby_onfail), "maintenance", pcmk__btoa(node->details->maintenance), "pending", pcmk__btoa(node->details->pending), "unclean", pcmk__btoa(node->details->unclean), "shutdown", pcmk__btoa(node->details->shutdown), "expected_up", pcmk__btoa(node->details->expected_up), "is_dc", pcmk__btoa(node->details->is_dc), "resources_running", length_s, "type", node_type); if (pe__is_guest_node(node)) { xmlNodePtr xml_node = pcmk__output_xml_peek_parent(out); crm_xml_add(xml_node, "id_as_resource", node->details->remote_rsc->container->id); } if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc = NULL; for (lpc = node->details->running_rsc; lpc != NULL; lpc = lpc->next) { pe_resource_t *rsc = (pe_resource_t *) lpc->data; out->message(out, crm_map_element_name(rsc->xml), show_opts | pcmk_show_rsc_only, rsc, only_node, only_rsc); } } free(length_s); out->end_list(out); } else { pcmk__output_xml_create_parent(out, "node", "name", node->details->uname, NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "gboolean", "int") static int node_attribute_text(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); gboolean add_extra = va_arg(args, gboolean); int expected_score = va_arg(args, int); if (add_extra) { int v; if (value == NULL) { v = 0; } else { pcmk__scan_min_int(value, &v, INT_MIN); } if (v <= 0) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is lost", name, value); } else if (v < expected_score) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is degraded (Expected=%d)", name, value, expected_score); } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "gboolean", "int") static int node_attribute_html(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); gboolean add_extra = va_arg(args, gboolean); int expected_score = va_arg(args, int); if (add_extra) { int v; char *s = crm_strdup_printf("%s: %s", name, value); xmlNodePtr item_node = pcmk__output_create_xml_node(out, "li", NULL); if (value == NULL) { v = 0; } else { pcmk__scan_min_int(value, &v, INT_MIN); } pcmk_create_html_node(item_node, "span", NULL, NULL, s); free(s); if (v <= 0) { pcmk_create_html_node(item_node, "span", NULL, "bold", "(connectivity is lost)"); } else if (v < expected_score) { char *buf = crm_strdup_printf("(connectivity is degraded -- expected %d", expected_score); pcmk_create_html_node(item_node, "span", NULL, "bold", buf); free(buf); } } else { out->list_item(out, NULL, "%s: %s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pe_working_set_t *", "xmlNodePtr") static int node_and_op(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pe_resource_t *rsc = NULL; gchar *node_str = NULL; char *last_change_str = NULL; const char *op_rsc = crm_element_value(xml_op, "resource"); const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); int status; time_t last_change = 0; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); rsc = pe_find_resource(data_set->resources, op_rsc); if (rsc) { pe_node_t *node = pe__current_node(rsc); const char *target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); unsigned int show_opts = pcmk_show_rsc_only | pcmk_show_pending; if (node == NULL) { node = rsc->pending_node; } node_str = pcmk__native_output_string(rsc, rsc_printable_id(rsc), node, show_opts, target_role, false); } else { node_str = crm_strdup_printf("Unknown resource %s", op_rsc); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change) == pcmk_ok) { last_change_str = crm_strdup_printf(", %s=%s, exec=%sms", XML_RSC_OP_LAST_CHANGE, pcmk__trim(ctime(&last_change)), crm_element_value(xml_op, XML_RSC_OP_T_EXEC)); } out->list_item(out, NULL, "%s: %s (node=%s, call=%s, rc=%s%s): %s", node_str, op_key ? op_key : ID(xml_op), crm_element_value(xml_op, XML_ATTR_UNAME), crm_element_value(xml_op, XML_LRM_ATTR_CALLID), crm_element_value(xml_op, XML_LRM_ATTR_RC), last_change_str ? last_change_str : "", pcmk_exec_status_str(status)); g_free(node_str); free(last_change_str); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pe_working_set_t *", "xmlNodePtr") static int node_and_op_xml(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pe_resource_t *rsc = NULL; const char *op_rsc = crm_element_value(xml_op, "resource"); const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); int status; time_t last_change = 0; xmlNode *node = NULL; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); node = pcmk__output_create_xml_node(out, "operation", "op", op_key ? op_key : ID(xml_op), "node", crm_element_value(xml_op, XML_ATTR_UNAME), "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "rc", crm_element_value(xml_op, XML_LRM_ATTR_RC), "status", pcmk_exec_status_str(status), NULL); rsc = pe_find_resource(data_set->resources, op_rsc); if (rsc) { const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS); const char *kind = crm_element_value(rsc->xml, XML_ATTR_TYPE); char *agent_tuple = NULL; agent_tuple = crm_strdup_printf("%s:%s:%s", class, pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) ? crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER) : "", kind); pcmk__xe_set_props(node, "rsc", rsc_printable_id(rsc), "agent", agent_tuple, NULL); free(agent_tuple); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change) == pcmk_ok) { pcmk__xe_set_props(node, XML_RSC_OP_LAST_CHANGE, pcmk__trim(ctime(&last_change)), XML_RSC_OP_T_EXEC, crm_element_value(xml_op, XML_RSC_OP_T_EXEC), NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "gboolean", "int") static int node_attribute_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); gboolean add_extra = va_arg(args, gboolean); int expected_score = va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "attribute", "name", name, "value", value, NULL); if (add_extra) { char *buf = pcmk__itoa(expected_score); crm_xml_add(node, "expected", buf); free(buf); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute-list", "pe_working_set_t *", "unsigned int", "gboolean", "GList *", "GList *") static int node_attribute_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer = va_arg(args, gboolean); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); int rc = pcmk_rc_no_output; /* Display each node's attributes */ for (GList *gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = gIter->data; GList *attr_list = NULL; GHashTableIter iter; gpointer key; if (!node || !node->details || !node->details->online) { continue; } g_hash_table_iter_init(&iter, node->details->attrs); while (g_hash_table_iter_next (&iter, &key, NULL)) { attr_list = filter_attr_list(attr_list, key); } if (attr_list == NULL) { continue; } if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { g_list_free(attr_list); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Node Attributes"); out->message(out, "node", node, show_opts, FALSE, NULL, only_node, only_rsc); for (GList *aIter = attr_list; aIter != NULL; aIter = aIter->next) { const char *name = aIter->data; const char *value = NULL; int expected_score = 0; gboolean add_extra = FALSE; value = pe_node_attribute_raw(node, name); add_extra = add_extra_info(node, node->details->running_rsc, data_set, name, &expected_score); /* Print attribute name and value */ out->message(out, "node-attribute", name, value, add_extra, expected_score); } g_list_free(attr_list); out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-capacity", "pe_node_t *", "const char *") static int node_capacity(pcmk__output_t *out, va_list args) { pe_node_t *node = va_arg(args, pe_node_t *); const char *comment = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s capacity:", comment, node->details->uname); g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-capacity", "pe_node_t *", "const char *") static int node_capacity_xml(pcmk__output_t *out, va_list args) { pe_node_t *node = va_arg(args, pe_node_t *); const char *comment = va_arg(args, const char *); xmlNodePtr xml_node = pcmk__output_create_xml_node(out, "capacity", "node", node->details->uname, "comment", comment, NULL); g_hash_table_foreach(node->details->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-history-list", "pe_working_set_t *", "pe_node_t *", "xmlNodePtr", "GList *", "GList *", "unsigned int", "unsigned int") static int node_history_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); pe_node_t *node = va_arg(args, pe_node_t *); xmlNode *node_state = va_arg(args, xmlNode *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int section_opts = va_arg(args, unsigned int); unsigned int show_opts = va_arg(args, unsigned int); xmlNode *lrm_rsc = NULL; xmlNode *rsc_entry = NULL; int rc = pcmk_rc_no_output; lrm_rsc = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE); /* Print history of each of the node's resources */ for (rsc_entry = first_named_child(lrm_rsc, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); pe_resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); /* We can't use is_filtered here to filter group resources. For is_filtered, * we have to decide whether to check the parent or not. If we check the * parent, all elements of a group will always be printed because that's how * is_filtered works for groups. If we do not check the parent, sometimes * this will filter everything out. * * For other resource types, is_filtered is okay. */ if (uber_parent(rsc)->variant == pe_group) { if (!pcmk__str_in_list(rsc_printable_id(rsc), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(uber_parent(rsc)), only_rsc, pcmk__str_star_matches)) { continue; } } else { if (rsc->fns->is_filtered(rsc, only_rsc, TRUE)) { continue; } } if (!pcmk_is_set(section_opts, pcmk_section_operations)) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pe_fc_default, NULL, data_set); if (failcount <= 0) { continue; } if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, FALSE, NULL, only_node, only_rsc); } out->message(out, "resource-history", rsc, rsc_id, FALSE, failcount, last_failure, FALSE); } else { GList *op_list = get_operation_list(rsc_entry); pe_resource_t *rsc = pe_find_resource(data_set->resources, crm_element_value(rsc_entry, XML_ATTR_ID)); if (op_list == NULL) { continue; } if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, FALSE, NULL, only_node, only_rsc); } out->message(out, "resource-operation-list", data_set, rsc, node, op_list, show_opts); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "unsigned int", "gboolean") static int node_list_html(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer G_GNUC_UNUSED = va_arg(args, gboolean); int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, FALSE, rc, "Node List"); out->message(out, "node", node, show_opts, TRUE, NULL, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "unsigned int", "gboolean") static int node_list_text(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer = va_arg(args, gboolean); /* space-separated lists of node names */ char *online_nodes = NULL; char *online_remote_nodes = NULL; char *online_guest_nodes = NULL; char *offline_nodes = NULL; char *offline_remote_nodes = NULL; size_t online_nodes_len = 0; size_t online_remote_nodes_len = 0; size_t online_guest_nodes_len = 0; size_t offline_nodes_len = 0; size_t offline_remote_nodes_len = 0; int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; const char *node_mode = NULL; char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { free(node_name); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer == TRUE, rc, "Node List"); /* Get node mode */ if (node->details->unclean) { if (node->details->online) { node_mode = "UNCLEAN (online)"; } else if (node->details->pending) { node_mode = "UNCLEAN (pending)"; } else { node_mode = "UNCLEAN (offline)"; } } else if (node->details->pending) { node_mode = "pending"; } else if (node->details->standby_onfail && node->details->online) { node_mode = "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { if (node->details->running_rsc) { node_mode = "standby (with active resources)"; } else { node_mode = "standby"; } } else { node_mode = "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { node_mode = "maintenance"; } else { node_mode = "OFFLINE (maintenance)"; } } else if (node->details->online) { node_mode = "online"; if (!pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (pe__is_guest_node(node)) { pcmk__add_word(&online_guest_nodes, &online_guest_nodes_len, node_name); } else if (pe__is_remote_node(node)) { pcmk__add_word(&online_remote_nodes, &online_remote_nodes_len, node_name); } else { pcmk__add_word(&online_nodes, &online_nodes_len, node_name); } free(node_name); continue; } } else { node_mode = "OFFLINE"; if (!pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (pe__is_remote_node(node)) { pcmk__add_word(&offline_remote_nodes, &offline_remote_nodes_len, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline guest nodes */ } else { pcmk__add_word(&offline_nodes, &offline_nodes_len, node_name); } free(node_name); continue; } } /* If we get here, node is in bad state, or we're grouping by node */ out->message(out, "node", node, show_opts, TRUE, node_mode, only_node, only_rsc); free(node_name); } /* If we're not grouping by node, summarize nodes by status */ if (online_nodes) { out->list_item(out, "Online", "[ %s ]", online_nodes); free(online_nodes); } if (offline_nodes) { out->list_item(out, "OFFLINE", "[ %s ]", offline_nodes); free(offline_nodes); } if (online_remote_nodes) { out->list_item(out, "RemoteOnline", "[ %s ]", online_remote_nodes); free(online_remote_nodes); } if (offline_remote_nodes) { out->list_item(out, "RemoteOFFLINE", "[ %s ]", offline_remote_nodes); free(offline_remote_nodes); } if (online_guest_nodes) { out->list_item(out, "GuestOnline", "[ %s ]", online_guest_nodes); free(online_guest_nodes); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "unsigned int", "gboolean") static int node_list_xml(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer G_GNUC_UNUSED = va_arg(args, gboolean); out->begin_list(out, NULL, NULL, "nodes"); for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } out->message(out, "node", node, show_opts, TRUE, NULL, only_node, only_rsc); } out->end_list(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-summary", "pe_working_set_t *", "GList *", "GList *", "unsigned int", "unsigned int", "gboolean") static int node_summary(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); unsigned int section_opts = va_arg(args, unsigned int); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_spacer = va_arg(args, gboolean); xmlNode *node_state = NULL; xmlNode *cib_status = get_object_root(XML_CIB_TAG_STATUS, data_set->input); int rc = pcmk_rc_no_output; if (xmlChildElementCount(cib_status) == 0) { return rc; } for (node_state = first_named_child(cib_status, XML_CIB_TAG_STATE); node_state != NULL; node_state = crm_next_same_xml(node_state)) { pe_node_t *node = pe_find_node_id(data_set->nodes, ID(node_state)); if (!node || !node->details || !node->details->online) { continue; } if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, pcmk_is_set(section_opts, pcmk_section_operations) ? "Operations" : "Migration Summary"); out->message(out, "node-history-list", data_set, node, node_state, only_node, only_rsc, section_opts, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-weight", "pe_resource_t *", "const char *", "const char *", "char *") static int node_weight(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); char *score = va_arg(args, char *); if (rsc) { out->list_item(out, NULL, "%s: %s allocation score on %s: %s", prefix, rsc->id, uname, score); } else { out->list_item(out, NULL, "%s: %s = %s", prefix, uname, score); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-weight", "pe_resource_t *", "const char *", "const char *", "char *") static int node_weight_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); char *score = va_arg(args, char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "node_weight", "function", prefix, "node", uname, "score", score, NULL); if (rsc) { crm_xml_add(node, "id", rsc->id); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNodePtr", "const char *", "const char *", "int", "unsigned int") static int op_history_text(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); unsigned int show_opts = va_arg(args, unsigned int); char *buf = op_history_string(xml_op, task, interval_ms_s, rc, pcmk_is_set(show_opts, pcmk_show_timing)); out->list_item(out, NULL, "%s", buf); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNodePtr", "const char *", "const char *", "int", "unsigned int") static int op_history_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); unsigned int show_opts = va_arg(args, unsigned int); char *rc_s = pcmk__itoa(rc); xmlNodePtr node = pcmk__output_create_xml_node(out, "operation_history", "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "task", task, "rc", rc_s, "rc_text", services_ocf_exitcode_str(rc), NULL); free(rc_s); if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *s = crm_strdup_printf("%sms", interval_ms_s); crm_xml_add(node, "interval", s); free(s); } if (pcmk_is_set(show_opts, pcmk_show_timing)) { const char *value = NULL; time_t epoch = 0; if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { crm_xml_add(node, XML_RSC_OP_LAST_CHANGE, pcmk__epoch2str(&epoch)); } value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, XML_RSC_OP_T_EXEC, s); free(s); } value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, XML_RSC_OP_T_QUEUE, s); free(s); } } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pe_resource_t *", "pe_node_t *", "char *") static int promotion_score(pcmk__output_t *out, va_list args) { pe_resource_t *child_rsc = va_arg(args, pe_resource_t *); pe_node_t *chosen = va_arg(args, pe_node_t *); char *score = va_arg(args, char *); out->list_item(out, NULL, "%s promotion score on %s: %s", child_rsc->id, chosen? chosen->details->uname : "none", score); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pe_resource_t *", "pe_node_t *", "char *") static int promotion_score_xml(pcmk__output_t *out, va_list args) { pe_resource_t *child_rsc = va_arg(args, pe_resource_t *); pe_node_t *chosen = va_arg(args, pe_node_t *); char *score = va_arg(args, char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "promotion_score", "id", child_rsc->id, "score", score, NULL); if (chosen) { crm_xml_add(node, "node", chosen->details->uname); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-config", "pe_resource_t *", "gboolean") static int resource_config(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); gboolean raw = va_arg(args, gboolean); char *rsc_xml = NULL; if (raw) { rsc_xml = dump_xml_formatted(rsc->orig_xml ? rsc->orig_xml : rsc->xml); } else { rsc_xml = dump_xml_formatted(rsc->xml); } pcmk__formatted_printf(out, "Resource XML:\n"); out->output_xml(out, "xml", rsc_xml); free(rsc_xml); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-history", "pe_resource_t *", "const char *", "gboolean", "int", "time_t", "gboolean") static int resource_history_text(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); const char *rsc_id = va_arg(args, const char *); gboolean all = va_arg(args, gboolean); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, int); gboolean as_header = va_arg(args, gboolean); char *buf = resource_history_string(rsc, rsc_id, all, failcount, last_failure); if (as_header) { out->begin_list(out, NULL, NULL, "%s", buf); } else { out->list_item(out, NULL, "%s", buf); } free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-history", "pe_resource_t *", "const char *", "gboolean", "int", "time_t", "gboolean") static int resource_history_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); const char *rsc_id = va_arg(args, const char *); gboolean all = va_arg(args, gboolean); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, int); gboolean as_header = va_arg(args, gboolean); xmlNodePtr node = pcmk__output_xml_create_parent(out, "resource_history", "id", rsc_id, NULL); if (rsc == NULL) { crm_xml_add(node, "orphan", "true"); } else if (all || failcount || last_failure > 0) { char *migration_s = pcmk__itoa(rsc->migration_threshold); pcmk__xe_set_props(node, "orphan", "false", "migration-threshold", migration_s, NULL); free(migration_s); if (failcount > 0) { char *s = pcmk__itoa(failcount); crm_xml_add(node, PCMK__FAIL_COUNT_PREFIX, s); free(s); } if (last_failure > 0) { crm_xml_add(node, PCMK__LAST_FAILURE_PREFIX, pcmk__epoch2str(&last_failure)); } } if (as_header == FALSE) { pcmk__output_xml_pop_parent(out); } return pcmk_rc_ok; } static void print_resource_header(pcmk__output_t *out, unsigned int show_opts) { if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { /* Active resources have already been printed by node */ out->begin_list(out, NULL, NULL, "Inactive Resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->begin_list(out, NULL, NULL, "Full List of Resources"); } else { out->begin_list(out, NULL, NULL, "Active Resources"); } } PCMK__OUTPUT_ARGS("resource-list", "pe_working_set_t *", "unsigned int", "gboolean", "GList *", "GList *", "gboolean") static int resource_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); unsigned int show_opts = va_arg(args, unsigned int); gboolean print_summary = va_arg(args, gboolean); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); gboolean print_spacer = va_arg(args, gboolean); GList *rsc_iter; int rc = pcmk_rc_no_output; bool printed_header = false; /* If we already showed active resources by node, and * we're not showing inactive resources, we have nothing to do */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node) && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { return rc; } /* If we haven't already printed resources grouped by node, * and brief output was requested, print resource summary */ if (pcmk_is_set(show_opts, pcmk_show_brief) && !pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(data_set->resources, only_rsc); PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; rc = pe__rscs_brief_output(out, rscs, show_opts); g_list_free(rscs); } /* For each resource, display it if appropriate */ for (rsc_iter = data_set->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { pe_resource_t *rsc = (pe_resource_t *) rsc_iter->data; int x; /* Complex resources may have some sub-resources active and some inactive */ gboolean is_active = rsc->fns->active(rsc, TRUE); gboolean partially_active = rsc->fns->active(rsc, FALSE); /* Skip inactive orphans (deleted but still in CIB) */ if (pcmk_is_set(rsc->flags, pe_rsc_orphan) && !is_active) { continue; /* Skip active resources if we already displayed them by node */ } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (is_active) { continue; } /* Skip primitives already counted in a brief summary */ } else if (pcmk_is_set(show_opts, pcmk_show_brief) && (rsc->variant == pe_native)) { continue; /* Skip resources that aren't at least partially active, * unless we're displaying inactive resources */ } else if (!partially_active && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { continue; } else if (partially_active && !pe__rsc_running_on_any(rsc, only_node)) { continue; } if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } /* Print this resource */ x = out->message(out, crm_map_element_name(rsc->xml), show_opts, rsc, only_node, only_rsc); if (x == pcmk_rc_ok) { rc = pcmk_rc_ok; } } if (print_summary && rc != pcmk_rc_ok) { if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { out->list_item(out, NULL, "No inactive resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->list_item(out, NULL, "No resources"); } else { out->list_item(out, NULL, "No active resources"); } } if (printed_header) { out->end_list(out); } return rc; } PCMK__OUTPUT_ARGS("resource-operation-list", "pe_working_set_t *", "pe_resource_t *", "pe_node_t *", "GList *", "unsigned int") static int resource_operation_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *node = va_arg(args, pe_node_t *); GList *op_list = va_arg(args, GList *); unsigned int show_opts = va_arg(args, unsigned int); GList *gIter = NULL; int rc = pcmk_rc_no_output; /* Print each operation */ for (gIter = op_list; gIter != NULL; gIter = gIter->next) { xmlNode *xml_op = (xmlNode *) gIter->data; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(xml_op, XML_LRM_ATTR_RC); int op_rc_i; pcmk__scan_min_int(op_rc, &op_rc_i, 0); /* Display 0-interval monitors as "probe" */ if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei) && pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches | pcmk__str_casei)) { task = "probe"; } /* If this is the first printed operation, print heading for resource */ if (rc == pcmk_rc_no_output) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pe_fc_default, NULL, data_set); out->message(out, "resource-history", rsc, rsc_printable_id(rsc), TRUE, failcount, last_failure, TRUE); rc = pcmk_rc_ok; } /* Print the operation */ out->message(out, "op-history", xml_op, task, interval_ms_s, op_rc_i, show_opts); } /* Free the list we created (no need to free the individual items) */ g_list_free(op_list); PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("resource-util", "pe_resource_t *", "pe_node_t *", "const char *") static int resource_util(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *node = va_arg(args, pe_node_t *); const char *fn = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s utilization on %s:", fn, rsc->id, node->details->uname); g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-util", "pe_resource_t *", "pe_node_t *", "const char *") static int resource_util_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *node = va_arg(args, pe_node_t *); const char *fn = va_arg(args, const char *); xmlNodePtr xml_node = pcmk__output_create_xml_node(out, "utilization", "resource", rsc->id, "node", node->details->uname, "function", fn, NULL); g_hash_table_foreach(rsc->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pe_ticket_t *") static int ticket_html(pcmk__output_t *out, va_list args) { pe_ticket_t *ticket = va_arg(args, pe_ticket_t *); if (ticket->last_granted > -1) { char *time = pcmk__format_named_time("last-granted", ticket->last_granted); out->list_item(out, NULL, "%s:\t%s%s %s", ticket->id, ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : "", time); free(time); } else { out->list_item(out, NULL, "%s:\t%s%s", ticket->id, ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : ""); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pe_ticket_t *") static int ticket_text(pcmk__output_t *out, va_list args) { pe_ticket_t *ticket = va_arg(args, pe_ticket_t *); if (ticket->last_granted > -1) { char *time = pcmk__format_named_time("last-granted", ticket->last_granted); out->list_item(out, ticket->id, "%s%s %s", ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : "", time); free(time); } else { out->list_item(out, ticket->id, "%s%s", ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : ""); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pe_ticket_t *") static int ticket_xml(pcmk__output_t *out, va_list args) { pe_ticket_t *ticket = va_arg(args, pe_ticket_t *); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, "ticket", "id", ticket->id, "status", ticket->granted ? "granted" : "revoked", "standby", pcmk__btoa(ticket->standby), NULL); if (ticket->last_granted > -1) { crm_xml_add(node, "last-granted", pcmk__epoch2str(&ticket->last_granted)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket-list", "pe_working_set_t *", "gboolean") static int ticket_list(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); gboolean print_spacer = va_arg(args, gboolean); GHashTableIter iter; gpointer key, value; if (g_hash_table_size(data_set->tickets) == 0) { return pcmk_rc_no_output; } PCMK__OUTPUT_SPACER_IF(out, print_spacer); /* Print section heading */ out->begin_list(out, NULL, NULL, "Tickets"); /* Print each ticket */ g_hash_table_iter_init(&iter, data_set->tickets); while (g_hash_table_iter_next(&iter, &key, &value)) { pe_ticket_t *ticket = (pe_ticket_t *) value; out->message(out, "ticket", ticket); } /* Close section */ out->end_list(out); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "ban", "default", ban_text }, { "ban", "html", ban_html }, { "ban", "xml", ban_xml }, { "ban-list", "default", ban_list }, { "bundle", "default", pe__bundle_text }, { "bundle", "xml", pe__bundle_xml }, { "bundle", "html", pe__bundle_html }, { "clone", "default", pe__clone_default }, { "clone", "xml", pe__clone_xml }, { "cluster-counts", "default", cluster_counts_text }, { "cluster-counts", "html", cluster_counts_html }, { "cluster-counts", "xml", cluster_counts_xml }, { "cluster-dc", "default", cluster_dc_text }, { "cluster-dc", "html", cluster_dc_html }, { "cluster-dc", "xml", cluster_dc_xml }, { "cluster-options", "default", cluster_options_text }, { "cluster-options", "html", cluster_options_html }, { "cluster-options", "log", cluster_options_log }, { "cluster-options", "xml", cluster_options_xml }, { "cluster-summary", "default", cluster_summary }, { "cluster-summary", "html", cluster_summary_html }, { "cluster-stack", "default", cluster_stack_text }, { "cluster-stack", "html", cluster_stack_html }, { "cluster-stack", "xml", cluster_stack_xml }, { "cluster-times", "default", cluster_times_text }, { "cluster-times", "html", cluster_times_html }, { "cluster-times", "xml", cluster_times_xml }, { "failed-action", "default", failed_action_default }, { "failed-action", "xml", failed_action_xml }, { "failed-action-list", "default", failed_action_list }, { "group", "default", pe__group_default}, { "group", "xml", pe__group_xml }, { "maint-mode", "text", cluster_maint_mode_text }, { "node", "default", node_text }, { "node", "html", node_html }, { "node", "xml", node_xml }, { "node-and-op", "default", node_and_op }, { "node-and-op", "xml", node_and_op_xml }, { "node-capacity", "default", node_capacity }, { "node-capacity", "xml", node_capacity_xml }, { "node-history-list", "default", node_history_list }, { "node-list", "default", node_list_text }, { "node-list", "html", node_list_html }, { "node-list", "xml", node_list_xml }, { "node-weight", "default", node_weight }, { "node-weight", "xml", node_weight_xml }, { "node-attribute", "default", node_attribute_text }, { "node-attribute", "html", node_attribute_html }, { "node-attribute", "xml", node_attribute_xml }, { "node-attribute-list", "default", node_attribute_list }, { "node-summary", "default", node_summary }, { "op-history", "default", op_history_text }, { "op-history", "xml", op_history_xml }, { "primitive", "default", pe__resource_text }, { "primitive", "xml", pe__resource_xml }, { "primitive", "html", pe__resource_html }, { "promotion-score", "default", promotion_score }, { "promotion-score", "xml", promotion_score_xml }, { "resource-config", "default", resource_config }, { "resource-history", "default", resource_history_text }, { "resource-history", "xml", resource_history_xml }, { "resource-list", "default", resource_list }, { "resource-operation-list", "default", resource_operation_list }, { "resource-util", "default", resource_util }, { "resource-util", "xml", resource_util_xml }, { "ticket", "default", ticket_text }, { "ticket", "html", ticket_html }, { "ticket", "xml", ticket_xml }, { "ticket-list", "default", ticket_list }, { NULL, NULL, NULL } }; void pe__register_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } void pe__output_node(pe_node_t *node, gboolean details, pcmk__output_t *out) { if (node == NULL) { crm_trace(""); return; } CRM_ASSERT(node->details); crm_trace("%sNode %s: (weight=%d, fixed=%s)", node->details->online ? "" : "Unavailable/Unclean ", node->details->uname, node->weight, node->fixed ? "True" : "False"); if (details) { char *pe_mutable = strdup("\t\t"); GList *gIter = node->details->running_rsc; GList *all = NULL; all = g_list_prepend(all, (gpointer) "*"); crm_trace("\t\t===Node Attributes"); g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable); free(pe_mutable); crm_trace("\t\t=== Resources"); for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; out->message(out, crm_map_element_name(rsc->xml), pe_print_pending, rsc, all, all); } g_list_free(all); } }