diff --git a/daemons/controld/controld_execd.c b/daemons/controld/controld_execd.c index 01dbf8b4ed..477f5c3deb 100644 --- a/daemons/controld/controld_execd.c +++ b/daemons/controld/controld_execd.c @@ -1,2968 +1,2968 @@ /* * Copyright 2004-2022 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(const lrm_state_t *lrm_state, const 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, xmlNode *msg, struct ra_metadata_s *md); 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(controld_globals.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); pcmk__str_update(&entry->rsc.provider, rsc->provider); } 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(const lrm_state_t *lrm_state, const ha_msg_input_t *input, const char *rsc_id, const 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); 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 pcmk__s(op->remote_nodename, controld_globals.our_nodename); } 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_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 (controld_globals.our_nodename == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(controld_globals.our_nodename); 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(controld_globals.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(controld_globals.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(controld_globals.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 * \p g_string_free() and the XML result with \p free_xml(). */ static GString * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, enum ra_param_flags_e param_type, xmlNode **result) { GString *list = NULL; *result = create_xml_node(NULL, XML_TAG_PARAMS); /* Consider all parameters only except private ones to be consistent with * what scheduler does with calculate_secure_digest(). */ if (param_type == ra_param_private && compare_version(controld_globals.dc_version, "3.16.0") >= 0) { g_hash_table_foreach(op->params, hash2field, *result); pcmk__filter_op_for_digest(*result); } 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, 256, " "); } pcmk__add_word(&list, 0, 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); } } else { crm_trace("Removing attr %s from the xml result", param->rap_name); xml_remove_prop(*result, param->rap_name); } } if (list != NULL) { // We will later search for " WORD ", so end list with a space pcmk__add_word(&list, 0, " "); } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { GString *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 == NULL)? "" : (const char *) list->str); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); if ((list != NULL) && (list->len > 0)) { crm_trace("%s: %s, %s", op->rsc_id, digest, (const char *) list->str); } else { crm_trace("%s: %s", op->rsc_id, digest); } if (list != NULL) { g_string_free(list, TRUE); } free_xml(restart); free(digest); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { GString *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, (const char *) list->str); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, (const char *) list->str); g_string_free(list, TRUE); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); } static gboolean build_operation_update(xmlNode * parent, const 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; if (op == NULL) { return FALSE; } target_rc = rsc_op_expected_rc(op); caller_version = g_hash_table_lookup(op->params, XML_ATTR_CRM_VERSION); CRM_CHECK(caller_version != NULL, caller_version = CRM_FEATURE_SET); xml_op = pcmk__create_history_xml(parent, op, caller_version, target_rc, controld_globals.our_nodename, src); 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; } /* Ideally the metadata is cached, and the agent is just a fallback. * * @TODO Go through all callers and ensure they get metadata asynchronously * first. */ metadata = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_agent |controld_metadata_from_cache); if (metadata == NULL) { return TRUE; } 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; } xmlNode * controld_query_executor_state(void) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; lrm_state_t *lrm_state = lrm_state_find(controld_globals.our_nodename); if (!lrm_state) { crm_err("Could not find executor state for node %s", controld_globals.our_nodename); return 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, node_update_cluster|node_update_peer, 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; } /*! * \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: lrmd__set_result(event, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); break; case EACCES: lrmd__set_result(event, PCMK_OCF_INSUFFICIENT_PRIV, PCMK_EXEC_ERROR, NULL); break; default: 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); cib__update_node_attr(controld_globals.logger_out, controld_globals.cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, 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_iter, int rc, const char *user_name, bool from_cib) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_iter) { g_hash_table_iter_remove(rsc_iter); } else { g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); } if (from_cib) { 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] op Operation whose history should be deleted */ static void erase_lrm_history_by_op(const 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); controld_globals.cib_conn->cmds->remove(controld_globals.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(const 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); controld_globals.cib_conn->cmds->remove(controld_globals.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,out] lrm_state Executor connection state to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource if not already * \param[out] rsc_info Where to store information obtained from executor * * \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, const 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 *iter, const char *sys, const char *user, ha_msg_input_t *request, bool unregister, bool from_cib) { 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, iter, rc, user, from_cib); } 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, 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; 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, bool reprobe_all_nodes) { 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 */ bool unregister = true; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { unregister = false; if (reprobe_all_nodes) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state != NULL) { /* If reprobing all nodes, be sure to reprobe the remote * node before clearing its connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE, reprobe_all_nodes); } } } /* Don't delete from the CIB, since we'll delete the whole node's LRM * state from the CIB soon */ delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, user_name, NULL, unregister, false); } /* Now delete the copy in the CIB */ controld_delete_node_state(lrm_state->node_name, controld_section_lrm, cib_scope_local); // @COMPAT DCs < 1.1.14 need this deleted (in case it was explicitly false) 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,out] 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, const xmlNode *action, 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, NULL); } else { 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 (replacing \p NULL with local node * name) * * \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(const xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = controld_globals.our_nodename; } 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"); 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); fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_UNKNOWN_ERROR, "Simulated failure"); process_lrm_event(lrm_state, op, NULL, xml); 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_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node, bool reprobe_all_nodes) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node, reprobe_all_nodes); 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, controld_globals.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) { bool 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); /* 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, true); } // User data for asynchronous metadata execution struct metadata_cb_data { lrmd_rsc_info_t *rsc; // Copy of resource information xmlNode *input_xml; // Copy of FSA input XML }; static struct metadata_cb_data * new_metadata_cb_data(lrmd_rsc_info_t *rsc, xmlNode *input_xml) { struct metadata_cb_data *data = NULL; data = calloc(1, sizeof(struct metadata_cb_data)); CRM_ASSERT(data != NULL); data->input_xml = copy_xml(input_xml); data->rsc = lrmd_copy_rsc_info(rsc); return data; } static void free_metadata_cb_data(struct metadata_cb_data *data) { lrmd_free_rsc_info(data->rsc); free_xml(data->input_xml); free(data); } /*! * \internal * \brief Execute an action after metadata has been retrieved * * \param[in] pid Ignored * \param[in] result Result of metadata action * \param[in] user_data Metadata callback data */ static void metadata_complete(int pid, const pcmk__action_result_t *result, void *user_data) { struct metadata_cb_data *data = (struct metadata_cb_data *) user_data; struct ra_metadata_s *md = NULL; lrm_state_t *lrm_state = lrm_state_find(lrm_op_target(data->input_xml)); if ((lrm_state != NULL) && pcmk__result_ok(result)) { md = controld_cache_metadata(lrm_state->metadata_cache, data->rsc, result->action_stdout); } do_lrm_rsc_op(lrm_state, data->rsc, data->input_xml, md); free_metadata_cb_data(data); } /* 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 = lrm_op_target(input->xml); gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; // Message routed to the local node is targeting a specific, non-local node is_remote_node = !pcmk__str_eq(target_node, controld_globals.our_nodename, 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, "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_none)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } if (pcmk__str_eq(crm_op, CRM_OP_LRM_DELETE, pcmk__str_none)) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { crm_rsc_delete = TRUE; // from crm_resource } operation = CRMD_ACTION_DELETE; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } CRM_CHECK(!pcmk__str_empty(crm_op) || !pcmk__str_empty(operation), return); crm_trace("'%s' execution request from %s as %s user", pcmk__s(crm_op, operation), pcmk__s(from_sys, "unknown subsystem"), pcmk__s(user_name, "current")); if (pcmk__str_eq(crm_op, CRM_OP_LRM_FAIL, pcmk__str_none)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_REFRESH, pcmk__str_none)) { /* @COMPAT This can only be sent by crm_resource --refresh on a * Pacemaker Remote node running Pacemaker 1.1.9, which is extremely * unlikely. It previously would cause the controller to re-write its * resource history to the CIB. Just ignore it. */ crm_notice("Ignoring refresh request from Pacemaker Remote 1.1.9 node"); // @COMPAT DCs <1.1.14 in a rolling upgrade might schedule this op } else if (pcmk__str_eq(operation, CRM_OP_PROBED, pcmk__str_none)) { 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_none) || pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) { const char *raw_target = NULL; if (input->xml != NULL) { // For CRM_OP_REPROBE, a NULL target means we're targeting all nodes raw_target = crm_element_value(input->xml, XML_LRM_ATTR_TARGET); } handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node, (raw_target == NULL)); } 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_none); 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, "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, true); 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 "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 "Could not register resource with executor"); return; } if (pcmk__str_eq(operation, CRMD_ACTION_CANCEL, pcmk__str_none)) { 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_none)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else { struct ra_metadata_s *md = NULL; /* Getting metadata from cache is OK except for start actions -- * always refresh from the agent for those, in case the resource * agent was updated. * * @TODO Only refresh metadata for starts if the agent actually * changed (using something like inotify, or a hash or modification * time of the agent executable). */ if (strcmp(operation, CRMD_ACTION_START) != 0) { md = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); } if ((md == NULL) && crm_op_needs_metadata(rsc->standard, operation)) { /* Most likely, we'll need the agent metadata to record the * pending operation and the operation result. Get it now rather * than wait until then, so the metadata action doesn't eat into * the real action's timeout. * * @TODO Metadata is retrieved via direct execution of the * agent, which has a couple of related issues: the executor * should execute agents, not the controller; and metadata for * Pacemaker Remote nodes should be collected on those nodes, * not locally. */ struct metadata_cb_data *data = NULL; data = new_metadata_cb_data(rsc, input->xml); crm_info("Retrieving metadata for %s (%s%s%s:%s) asynchronously", rsc->id, rsc->standard, ((rsc->provider == NULL)? "" : ":"), ((rsc->provider == NULL)? "" : rsc->provider), rsc->type); (void) lrmd__metadata_async(rsc, metadata_complete, (void *) data); } else { do_lrm_rsc_op(lrm_state, rsc, input->xml, md); } } lrmd_free_rsc_info(rsc); } else { crm_err("Invalid execution request: unknown command '%s' (bug?)", crm_op); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } static lrmd_event_data_t * construct_op(const lrm_state_t *lrm_state, const 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->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 (!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 (NULL for transition engine) * \param[in] rsc Type information about resource the result is for * \param[in,out] 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, const 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, controld_globals.our_nodename); 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, controld_globals.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, controld_globals.our_nodename, __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; 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, xmlNode *msg, struct ra_metadata_s *md) { 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; const char *operation = NULL; gboolean stop_recurring = FALSE; const char *nack_reason = NULL; CRM_CHECK((rsc != NULL) && (msg != NULL), return); operation = crm_element_value(msg, XML_LRM_ATTR_TASK); CRM_CHECK(!pcmk__str_empty(operation), return); transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (pcmk__str_empty(transition)) { crm_log_xml_err(msg, "Missing transition number"); } if (lrm_state == NULL) { // This shouldn't be possible, but provide a failsafe just in case crm_err("Cannot execute %s of %s: No executor connection " CRM_XS " transition_key=%s", operation, rsc->id, pcmk__s(transition, "")); synthesize_lrmd_failure(NULL, msg, PCMK_EXEC_INVALID, PCMK_OCF_UNKNOWN_ERROR, "No executor connection"); return; } 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. */ if ((md != NULL) && pcmk_is_set(md->ra_flags, ra_supports_legacy_reload)) { operation = CRMD_ACTION_RELOAD; } else { operation = CRMD_ACTION_RELOAD_AGENT; } } 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, pcmk__s(transition, ""), rsc->id, operation, op->interval_ms); if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN) && pcmk__str_eq(operation, RSC_START, pcmk__str_casei)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); nack_reason = "Not attempting start due to shutdown in progress"; } else { switch (controld_globals.fsa_state) { case S_NOT_DC: case S_POLICY_ENGINE: // Recalculating case S_TRANSITION_ENGINE: break; default: if (!pcmk__str_eq(operation, CRMD_ACTION_STOP, pcmk__str_none)) { nack_reason = "Controller cannot attempt actions at this " "time"; } break; } } if (nack_reason != NULL) { crm_notice("Discarding attempt to perform action %s on %s in state %s " "(shutdown=%s)", operation, rsc->id, fsa_state2string(controld_globals.fsa_state), pcmk__btoa(pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN))); 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); pcmk__str_update(&pending->user_data, op->user_data); 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 = 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); 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_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_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); } free(op_id); lrmd_free_event(op); } 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 == controld_globals.resource_update) { controld_globals.resource_update = 0; controld_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 (!pcmk_is_set(controld_globals.flags, 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, controld_globals.our_nodename, pcmk__str_casei)) { uuid = controld_globals.our_uuid; } else { /* remote nodes uuid and uname are equal */ uuid = node_name; pcmk__xe_set_bool_attr(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_ID, 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) { controld_globals.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; } /*! * \internal * \brief Log the result of an executor action (actual or synthesized) * * \param[in] op Executor action to log result for * \param[in] op_key Operation key for action * \param[in] node_name Name of node action was performed on, if known * \param[in] update_id Call ID for CIB update (or 0 if none) * \param[in] confirmed Whether to log that graph action was confirmed */ static void log_executor_event(const lrmd_event_data_t *op, const char *op_key, const char *node_name, int update_id, gboolean confirmed) { int log_level = LOG_ERR; GString *str = g_string_sized_new(100); // reasonable starting size pcmk__g_strcat(str, "Result of ", crm_action_str(op->op_type, op->interval_ms), " operation for ", op->rsc_id, NULL); if (node_name != NULL) { pcmk__g_strcat(str, " on ", node_name, NULL); } switch (op->op_status) { case PCMK_EXEC_DONE: log_level = LOG_NOTICE; pcmk__g_strcat(str, ": ", services_ocf_exitcode_str(op->rc), NULL); break; case PCMK_EXEC_TIMEOUT: pcmk__g_strcat(str, ": ", pcmk_exec_status_str(op->op_status), " after ", pcmk__readable_interval(op->timeout), NULL); break; case PCMK_EXEC_CANCELLED: log_level = LOG_INFO; /* order of __attribute__ and Fall through comment is IMPORTANT! * do not change it without proper testing with both clang and gcc * in multiple versions. * the clang check allows to build with all versions of clang. * the has_c_attribute check is to workaround a bug in clang version * in rhel7. has_attribute would happily return "YES SIR WE GOT IT" * and fail the build the next line. */ #ifdef __clang__ #ifdef __has_c_attribute #if __has_attribute(fallthrough) __attribute__((fallthrough)); #endif #endif #endif // Fall through default: pcmk__g_strcat(str, ": ", pcmk_exec_status_str(op->op_status), NULL); } if ((op->exit_reason != NULL) && ((op->op_status != PCMK_EXEC_DONE) || (op->rc != PCMK_OCF_OK))) { pcmk__g_strcat(str, " (", op->exit_reason, ")", NULL); } g_string_append(str, " " CRM_XS); if (update_id != 0) { g_string_append_printf(str, " CIB update %d,", update_id); } g_string_append_printf(str, " graph action %sconfirmed; call=%d key=%s", (confirmed? "" : "un"), op->call_id, op_key); if (op->op_status == PCMK_EXEC_DONE) { g_string_append_printf(str, " rc=%d", op->rc); } do_crm_log(log_level, "%s", str->str); g_string_free(str, TRUE); /* The services library has already logged the output at info or debug * level, so just raise to notice if it looks like a failure. */ if ((op->output != NULL) && (op->rc != PCMK_OCF_OK)) { char *prefix = crm_strdup_printf(PCMK__OP_FMT "@%s output", op->rsc_id, op->op_type, op->interval_ms, node_name); crm_log_output(LOG_NOTICE, prefix, op->output); free(prefix); } } void process_lrm_event(lrm_state_t *lrm_state, lrmd_event_data_t *op, active_op_t *pending, const 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(controld_globals.dc_version, "3.2.0") < 0) { switch (op->op_status) { case PCMK_EXEC_NOT_CONNECTED: lrmd__set_result(op, PCMK_OCF_CONNECTION_DIED, PCMK_EXEC_ERROR, op->exit_reason); break; case PCMK_EXEC_INVALID: 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(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)); } } log_executor_event(op, op_key, node_name, update_id, removed); 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); controld_cache_metadata(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, true); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ controld_trigger_fsa(); 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_fencing.c b/daemons/controld/controld_fencing.c index a8fbf29072..3f08e2f8e7 100644 --- a/daemons/controld/controld_fencing.c +++ b/daemons/controld/controld_fencing.c @@ -1,1108 +1,1108 @@ /* * Copyright 2004-2022 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 static void tengine_stonith_history_synced(stonith_t *st, stonith_event_t *st_event); /* * stonith failure counting * * We don't want to get stuck in a permanent fencing loop. Keep track of the * number of fencing failures for each target node, and the most we'll restart a * transition for. */ struct st_fail_rec { int count; }; static bool fence_reaction_panic = false; static unsigned long int stonith_max_attempts = 10; static GHashTable *stonith_failures = NULL; /*! * \internal * \brief Update max fencing attempts before giving up * * \param[in] value New max fencing attempts */ static void update_stonith_max_attempts(const char *value) { stonith_max_attempts = char2score(value); if (stonith_max_attempts < 1UL) { stonith_max_attempts = 10UL; } } /*! * \internal * \brief Configure reaction to notification of local node being fenced * * \param[in] reaction_s Reaction type */ static void set_fence_reaction(const char *reaction_s) { if (pcmk__str_eq(reaction_s, "panic", pcmk__str_casei)) { fence_reaction_panic = true; } else { if (!pcmk__str_eq(reaction_s, "stop", pcmk__str_casei)) { crm_warn("Invalid value '%s' for %s, using 'stop'", reaction_s, XML_CONFIG_ATTR_FENCE_REACTION); } fence_reaction_panic = false; } } /*! * \internal * \brief Configure fencing options based on the CIB * * \param[in,out] options Name/value pairs for configured options */ void controld_configure_fencing(GHashTable *options) { const char *value = NULL; value = g_hash_table_lookup(options, XML_CONFIG_ATTR_FENCE_REACTION); set_fence_reaction(value); value = g_hash_table_lookup(options, "stonith-max-attempts"); update_stonith_max_attempts(value); } static gboolean too_many_st_failures(const char *target) { GHashTableIter iter; const char *key = NULL; struct st_fail_rec *value = NULL; if (stonith_failures == NULL) { return FALSE; } if (target == NULL) { g_hash_table_iter_init(&iter, stonith_failures); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { if (value->count >= stonith_max_attempts) { target = (const char*)key; goto too_many; } } } else { value = g_hash_table_lookup(stonith_failures, target); if ((value != NULL) && (value->count >= stonith_max_attempts)) { goto too_many; } } return FALSE; too_many: crm_warn("Too many failures (%d) to fence %s, giving up", value->count, target); return TRUE; } /*! * \internal * \brief Reset a stonith fail count * * \param[in] target Name of node to reset, or NULL for all */ void st_fail_count_reset(const char *target) { if (stonith_failures == NULL) { return; } if (target) { struct st_fail_rec *rec = NULL; rec = g_hash_table_lookup(stonith_failures, target); if (rec) { rec->count = 0; } } else { GHashTableIter iter; const char *key = NULL; struct st_fail_rec *rec = NULL; g_hash_table_iter_init(&iter, stonith_failures); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &rec)) { rec->count = 0; } } } static void st_fail_count_increment(const char *target) { struct st_fail_rec *rec = NULL; if (stonith_failures == NULL) { stonith_failures = pcmk__strkey_table(free, free); } rec = g_hash_table_lookup(stonith_failures, target); if (rec) { rec->count++; } else { rec = malloc(sizeof(struct st_fail_rec)); if(rec == NULL) { return; } rec->count = 1; g_hash_table_insert(stonith_failures, strdup(target), rec); } } /* end stonith fail count functions */ static void cib_fencing_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { if (rc < pcmk_ok) { crm_err("Fencing update %d for %s: failed - %s (%d)", call_id, (char *)user_data, pcmk_strerror(rc), rc); crm_log_xml_warn(msg, "Failed update"); abort_transition(INFINITY, pcmk__graph_shutdown, "CIB update failed", NULL); } else { crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data); } } static void send_stonith_update(pcmk__graph_action_t *action, const char *target, const char *uuid) { int rc = pcmk_ok; crm_node_t *peer = NULL; /* We (usually) rely on the membership layer to do node_update_cluster, * and the peer status callback to do node_update_peer, because the node * might have already rejoined before we get the stonith result here. */ int flags = node_update_join | node_update_expected; /* zero out the node-status & remove all LRM status info */ xmlNode *node_state = NULL; CRM_CHECK(target != NULL, return); CRM_CHECK(uuid != NULL, return); /* Make sure the membership and join caches are accurate */ peer = crm_get_peer_full(0, target, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return); if (peer->state == NULL) { /* Usually, we rely on the membership layer to update the cluster state * in the CIB. However, if the node has never been seen, do it here, so * the node is not considered unclean. */ flags |= node_update_cluster; } if (peer->uuid == NULL) { crm_info("Recording uuid '%s' for node '%s'", uuid, target); peer->uuid = strdup(uuid); } crmd_peer_down(peer, TRUE); /* Generate a node state update for the CIB */ node_state = create_node_state_update(peer, flags, NULL, __func__); /* we have to mark whether or not remote nodes have already been fenced */ if (peer->flags & crm_remote_node) { char *now_s = pcmk__ttoa(time(NULL)); crm_xml_add(node_state, XML_NODE_IS_FENCED, now_s); free(now_s); } /* Force our known ID */ - crm_xml_add(node_state, XML_ATTR_UUID, uuid); + crm_xml_add(node_state, XML_ATTR_ID, uuid); rc = controld_globals.cib_conn->cmds->update(controld_globals.cib_conn, XML_CIB_TAG_STATUS, node_state, cib_quorum_override |cib_scope_local |cib_can_create); /* Delay processing the trigger until the update completes */ crm_debug("Sending fencing update %d for %s", rc, target); fsa_register_cib_callback(rc, FALSE, strdup(target), cib_fencing_updated); // Make sure it sticks /* controld_globals.cib_conn->cmds->bump_epoch(controld_globals.cib_conn, * cib_quorum_override * |cib_scope_local); */ controld_delete_node_state(peer->uname, controld_section_all, cib_scope_local); free_xml(node_state); return; } /*! * \internal * \brief Abort transition due to stonith failure * * \param[in] abort_action Whether to restart or stop transition * \param[in] target Don't restart if this (NULL for any) has too many failures * \param[in] reason Log this stonith action XML as abort reason (or NULL) */ static void abort_for_stonith_failure(enum pcmk__graph_next abort_action, const char *target, const xmlNode *reason) { /* If stonith repeatedly fails, we eventually give up on starting a new * transition for that reason. */ if ((abort_action != pcmk__graph_wait) && too_many_st_failures(target)) { abort_action = pcmk__graph_wait; } abort_transition(INFINITY, abort_action, "Stonith failed", reason); } /* * stonith cleanup list * * If the DC is shot, proper notifications might not go out. * The stonith cleanup list allows the cluster to (re-)send * notifications once a new DC is elected. */ static GList *stonith_cleanup_list = NULL; /*! * \internal * \brief Add a node to the stonith cleanup list * * \param[in] target Name of node to add */ void add_stonith_cleanup(const char *target) { stonith_cleanup_list = g_list_append(stonith_cleanup_list, strdup(target)); } /*! * \internal * \brief Remove a node from the stonith cleanup list * * \param[in] Name of node to remove */ void remove_stonith_cleanup(const char *target) { GList *iter = stonith_cleanup_list; while (iter != NULL) { GList *tmp = iter; char *iter_name = tmp->data; iter = iter->next; if (pcmk__str_eq(target, iter_name, pcmk__str_casei)) { crm_trace("Removing %s from the cleanup list", iter_name); stonith_cleanup_list = g_list_delete_link(stonith_cleanup_list, tmp); free(iter_name); } } } /*! * \internal * \brief Purge all entries from the stonith cleanup list */ void purge_stonith_cleanup(void) { if (stonith_cleanup_list) { GList *iter = NULL; for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { char *target = iter->data; crm_info("Purging %s from stonith cleanup list", target); free(target); } g_list_free(stonith_cleanup_list); stonith_cleanup_list = NULL; } } /*! * \internal * \brief Send stonith updates for all entries in cleanup list, then purge it */ void execute_stonith_cleanup(void) { GList *iter; for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { char *target = iter->data; crm_node_t *target_node = crm_get_peer(0, target); const char *uuid = crm_peer_uuid(target_node); crm_notice("Marking %s, target of a previous stonith action, as clean", target); send_stonith_update(NULL, target, uuid); free(target); } g_list_free(stonith_cleanup_list); stonith_cleanup_list = NULL; } /* end stonith cleanup list functions */ /* stonith API client * * Functions that need to interact directly with the fencer via its API */ static stonith_t *stonith_api = NULL; static crm_trigger_t *stonith_reconnect = NULL; static char *te_client_id = NULL; static gboolean fail_incompletable_stonith(pcmk__graph_t *graph) { GList *lpc = NULL; const char *task = NULL; xmlNode *last_action = NULL; if (graph == NULL) { return FALSE; } for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { GList *lpc2 = NULL; pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) lpc->data; if (pcmk_is_set(synapse->flags, pcmk__synapse_confirmed)) { continue; } for (lpc2 = synapse->actions; lpc2 != NULL; lpc2 = lpc2->next) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) lpc2->data; if ((action->type != pcmk__cluster_graph_action) || pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { continue; } task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (task && pcmk__str_eq(task, CRM_OP_FENCE, pcmk__str_casei)) { pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); last_action = action->xml; pcmk__update_graph(graph, action); crm_notice("Failing action %d (%s): fencer terminated", action->id, ID(action->xml)); } } } if (last_action != NULL) { crm_warn("Fencer failure resulted in unrunnable actions"); abort_for_stonith_failure(pcmk__graph_restart, NULL, last_action); return TRUE; } return FALSE; } static void tengine_stonith_connection_destroy(stonith_t *st, stonith_event_t *e) { te_cleanup_stonith_history_sync(st, FALSE); if (pcmk_is_set(controld_globals.fsa_input_register, R_ST_REQUIRED)) { crm_crit("Fencing daemon connection failed"); mainloop_set_trigger(stonith_reconnect); } else { crm_info("Fencing daemon disconnected"); } if (stonith_api) { /* the client API won't properly reconnect notifications * if they are still in the table - so remove them */ if (stonith_api->state != stonith_disconnected) { stonith_api->cmds->disconnect(st); } stonith_api->cmds->remove_notification(stonith_api, NULL); } if (AM_I_DC) { fail_incompletable_stonith(controld_globals.transition_graph); trigger_graph(); } } /*! * \internal * \brief Handle an event notification from the fencing API * * \param[in] st Fencing API connection (ignored) * \param[in] event Fencing API event notification */ static void handle_fence_notification(stonith_t *st, stonith_event_t *event) { bool succeeded = true; const char *executioner = "the cluster"; const char *client = "a client"; const char *reason = NULL; int exec_status; if (te_client_id == NULL) { te_client_id = crm_strdup_printf("%s.%lu", crm_system_name, (unsigned long) getpid()); } if (event == NULL) { crm_err("Notify data not found"); return; } if (event->executioner != NULL) { executioner = event->executioner; } if (event->client_origin != NULL) { client = event->client_origin; } exec_status = stonith__event_execution_status(event); if ((stonith__event_exit_status(event) != CRM_EX_OK) || (exec_status != PCMK_EXEC_DONE)) { succeeded = false; if (exec_status == PCMK_EXEC_DONE) { exec_status = PCMK_EXEC_ERROR; } } reason = stonith__event_exit_reason(event); crmd_alert_fencing_op(event); if (pcmk__str_eq("on", event->action, pcmk__str_none)) { // Unfencing doesn't need special handling, just a log message if (succeeded) { crm_notice("%s was unfenced by %s at the request of %s@%s", event->target, executioner, client, event->origin); } else { crm_err("Unfencing of %s by %s failed (%s%s%s) with exit status %d", event->target, executioner, pcmk_exec_status_str(exec_status), ((reason == NULL)? "" : ": "), ((reason == NULL)? "" : reason), stonith__event_exit_status(event)); } return; } if (succeeded && pcmk__str_eq(event->target, controld_globals.our_nodename, pcmk__str_casei)) { /* We were notified of our own fencing. Most likely, either fencing was * misconfigured, or fabric fencing that doesn't cut cluster * communication is in use. * * Either way, shutting down the local host is a good idea, to require * administrator intervention. Also, other nodes would otherwise likely * set our status to lost because of the fencing callback and discard * our subsequent election votes as "not part of our cluster". */ crm_crit("We were allegedly just fenced by %s for %s!", executioner, event->origin); // Dumps blackbox if enabled if (fence_reaction_panic) { pcmk__panic(__func__); } else { crm_exit(CRM_EX_FATAL); } return; // Should never get here } /* Update the count of fencing failures for this target, in case we become * DC later. The current DC has already updated its fail count in * tengine_stonith_callback(). */ if (!AM_I_DC) { if (succeeded) { st_fail_count_reset(event->target); } else { st_fail_count_increment(event->target); } } crm_notice("Peer %s was%s terminated (%s) by %s on behalf of %s@%s: " "%s%s%s%s " CRM_XS " event=%s", event->target, (succeeded? "" : " not"), event->action, executioner, client, event->origin, (succeeded? "OK" : pcmk_exec_status_str(exec_status)), ((reason == NULL)? "" : " ("), ((reason == NULL)? "" : reason), ((reason == NULL)? "" : ")"), event->id); if (succeeded) { crm_node_t *peer = pcmk__search_known_node_cache(0, event->target, CRM_GET_PEER_ANY); const char *uuid = NULL; if (peer == NULL) { return; } uuid = crm_peer_uuid(peer); if (AM_I_DC) { /* The DC always sends updates */ send_stonith_update(NULL, event->target, uuid); /* @TODO Ideally, at this point, we'd check whether the fenced node * hosted any guest nodes, and call remote_node_down() for them. * Unfortunately, the controller doesn't have a simple, reliable way * to map hosts to guests. It might be possible to track this in the * peer cache via crm_remote_peer_cache_refresh(). For now, we rely * on the scheduler creating fence pseudo-events for the guests. */ if (!pcmk__str_eq(client, te_client_id, pcmk__str_casei)) { /* Abort the current transition if it wasn't the cluster that * initiated fencing. */ crm_info("External fencing operation from %s fenced %s", client, event->target); abort_transition(INFINITY, pcmk__graph_restart, "External Fencing Operation", NULL); } } else if (pcmk__str_eq(controld_globals.dc_name, event->target, pcmk__str_null_matches|pcmk__str_casei) && !pcmk_is_set(peer->flags, crm_remote_node)) { // Assume the target was our DC if we don't currently have one if (controld_globals.dc_name != NULL) { crm_notice("Fencing target %s was our DC", event->target); } else { crm_notice("Fencing target %s may have been our DC", event->target); } /* Given the CIB resyncing that occurs around elections, * have one node update the CIB now and, if the new DC is different, * have them do so too after the election */ if (pcmk__str_eq(event->executioner, controld_globals.our_nodename, pcmk__str_casei)) { send_stonith_update(NULL, event->target, uuid); } add_stonith_cleanup(event->target); } /* If the target is a remote node, and we host its connection, * immediately fail all monitors so it can be recovered quickly. * The connection won't necessarily drop when a remote node is fenced, * so the failure might not otherwise be detected until the next poke. */ if (pcmk_is_set(peer->flags, crm_remote_node)) { remote_ra_fail(event->target); } crmd_peer_down(peer, TRUE); } } /*! * \brief Connect to fencer * * \param[in] user_data If NULL, retry failures now, otherwise retry in main loop * * \return TRUE * \note If user_data is NULL, this will wait 2s between attempts, for up to * 30 attempts, meaning the controller could be blocked as long as 58s. */ static gboolean te_connect_stonith(gpointer user_data) { int rc = pcmk_ok; if (stonith_api == NULL) { stonith_api = stonith_api_new(); if (stonith_api == NULL) { crm_err("Could not connect to fencer: API memory allocation failed"); return TRUE; } } if (stonith_api->state != stonith_disconnected) { crm_trace("Already connected to fencer, no need to retry"); return TRUE; } if (user_data == NULL) { // Blocking (retry failures now until successful) rc = stonith_api_connect_retry(stonith_api, crm_system_name, 30); if (rc != pcmk_ok) { crm_err("Could not connect to fencer in 30 attempts: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } } else { // Non-blocking (retry failures later in main loop) rc = stonith_api->cmds->connect(stonith_api, crm_system_name, NULL); if (rc != pcmk_ok) { if (pcmk_is_set(controld_globals.fsa_input_register, R_ST_REQUIRED)) { crm_notice("Fencer connection failed (will retry): %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); mainloop_set_trigger(stonith_reconnect); } else { crm_info("Fencer connection failed (ignoring because no longer required): %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } return TRUE; } } if (rc == pcmk_ok) { stonith_api->cmds->register_notification(stonith_api, T_STONITH_NOTIFY_DISCONNECT, tengine_stonith_connection_destroy); stonith_api->cmds->register_notification(stonith_api, T_STONITH_NOTIFY_FENCE, handle_fence_notification); stonith_api->cmds->register_notification(stonith_api, T_STONITH_NOTIFY_HISTORY_SYNCED, tengine_stonith_history_synced); te_trigger_stonith_history_sync(TRUE); crm_notice("Fencer successfully connected"); } return TRUE; } /*! \internal \brief Schedule fencer connection attempt in main loop */ void controld_trigger_fencer_connect(void) { if (stonith_reconnect == NULL) { stonith_reconnect = mainloop_add_trigger(G_PRIORITY_LOW, te_connect_stonith, GINT_TO_POINTER(TRUE)); } controld_set_fsa_input_flags(R_ST_REQUIRED); mainloop_set_trigger(stonith_reconnect); } void controld_disconnect_fencer(bool destroy) { if (stonith_api) { // Prevent fencer connection from coming up again controld_clear_fsa_input_flags(R_ST_REQUIRED); if (stonith_api->state != stonith_disconnected) { stonith_api->cmds->disconnect(stonith_api); } stonith_api->cmds->remove_notification(stonith_api, NULL); } if (destroy) { if (stonith_api) { stonith_api->cmds->free(stonith_api); stonith_api = NULL; } if (stonith_reconnect) { mainloop_destroy_trigger(stonith_reconnect); stonith_reconnect = NULL; } if (te_client_id) { free(te_client_id); te_client_id = NULL; } } } static gboolean do_stonith_history_sync(gpointer user_data) { if (stonith_api && (stonith_api->state != stonith_disconnected)) { stonith_history_t *history = NULL; te_cleanup_stonith_history_sync(stonith_api, FALSE); stonith_api->cmds->history(stonith_api, st_opt_sync_call | st_opt_broadcast, NULL, &history, 5); stonith_history_free(history); return TRUE; } else { crm_info("Skip triggering stonith history-sync as stonith is disconnected"); return FALSE; } } static void tengine_stonith_callback(stonith_t *stonith, stonith_callback_data_t *data) { char *uuid = NULL; int stonith_id = -1; int transition_id = -1; pcmk__graph_action_t *action = NULL; const char *target = NULL; if ((data == NULL) || (data->userdata == NULL)) { crm_err("Ignoring fence operation %d result: " "No transition key given (bug?)", ((data == NULL)? -1 : data->call_id)); return; } if (!AM_I_DC) { const char *reason = stonith__exit_reason(data); if (reason == NULL) { reason = pcmk_exec_status_str(stonith__execution_status(data)); } crm_notice("Result of fence operation %d: %d (%s) " CRM_XS " key=%s", data->call_id, stonith__exit_status(data), reason, (const char *) data->userdata); return; } CRM_CHECK(decode_transition_key(data->userdata, &uuid, &transition_id, &stonith_id, NULL), goto bail); if (controld_globals.transition_graph->complete || (stonith_id < 0) || !pcmk__str_eq(uuid, controld_globals.te_uuid, pcmk__str_none) || (controld_globals.transition_graph->id != transition_id)) { crm_info("Ignoring fence operation %d result: " "Not from current transition " CRM_XS " complete=%s action=%d uuid=%s (vs %s) transition=%d (vs %d)", data->call_id, pcmk__btoa(controld_globals.transition_graph->complete), stonith_id, uuid, controld_globals.te_uuid, transition_id, controld_globals.transition_graph->id); goto bail; } action = controld_get_action(stonith_id); if (action == NULL) { crm_err("Ignoring fence operation %d result: " "Action %d not found in transition graph (bug?) " CRM_XS " uuid=%s transition=%d", data->call_id, stonith_id, uuid, transition_id); goto bail; } target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (target == NULL) { crm_err("Ignoring fence operation %d result: No target given (bug?)", data->call_id); goto bail; } stop_te_timer(action); if (stonith__exit_status(data) == CRM_EX_OK) { const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); const char *op = crm_meta_value(action->params, "stonith_action"); crm_info("Fence operation %d for %s succeeded", data->call_id, target); if (!(pcmk_is_set(action->flags, pcmk__graph_action_confirmed))) { te_action_confirmed(action, NULL); if (pcmk__str_eq("on", op, pcmk__str_casei)) { const char *value = NULL; char *now = pcmk__ttoa(time(NULL)); gboolean is_remote_node = FALSE; /* This check is not 100% reliable, since this node is not * guaranteed to have the remote node cached. However, it * doesn't have to be reliable, since the attribute manager can * learn a node's "remoteness" by other means sooner or later. * This allows it to learn more quickly if this node does have * the information. */ if (g_hash_table_lookup(crm_remote_peer_cache, uuid) != NULL) { is_remote_node = TRUE; } update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, is_remote_node); free(now); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL); update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, is_remote_node); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE); update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, is_remote_node); } else if (!(pcmk_is_set(action->flags, pcmk__graph_action_sent_update))) { send_stonith_update(action, target, uuid); pcmk__set_graph_action_flags(action, pcmk__graph_action_sent_update); } } st_fail_count_reset(target); } else { enum pcmk__graph_next abort_action = pcmk__graph_restart; int status = stonith__execution_status(data); const char *reason = stonith__exit_reason(data); if (reason == NULL) { if (status == PCMK_EXEC_DONE) { reason = "Agent returned error"; } else { reason = pcmk_exec_status_str(status); } } pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); /* If no fence devices were available, there's no use in immediately * checking again, so don't start a new transition in that case. */ if (status == PCMK_EXEC_NO_FENCE_DEVICE) { crm_warn("Fence operation %d for %s failed: %s " "(aborting transition and giving up for now)", data->call_id, target, reason); abort_action = pcmk__graph_wait; } else { crm_notice("Fence operation %d for %s failed: %s " "(aborting transition)", data->call_id, target, reason); } /* Increment the fail count now, so abort_for_stonith_failure() can * check it. Non-DC nodes will increment it in * handle_fence_notification(). */ st_fail_count_increment(target); abort_for_stonith_failure(abort_action, target, NULL); } pcmk__update_graph(controld_globals.transition_graph, action); trigger_graph(); bail: free(data->userdata); free(uuid); return; } static int fence_with_delay(const char *target, const char *type, const char *delay) { uint32_t options = st_opt_none; // Group of enum stonith_call_options int timeout_sec = (int) (controld_globals.transition_graph->stonith_timeout / 1000); int delay_i; if (crmd_join_phase_count(crm_join_confirmed) == 1) { stonith__set_call_options(options, target, st_opt_allow_suicide); } pcmk__scan_min_int(delay, &delay_i, 0); return stonith_api->cmds->fence_with_delay(stonith_api, options, target, type, timeout_sec, 0, delay_i); } /*! * \internal * \brief Execute a fencing action from a transition graph * * \param[in] graph Transition graph being executed (ignored) * \param[in] action Fencing action to execute * * \return Standard Pacemaker return code */ int controld_execute_fence_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { int rc = 0; const char *id = ID(action->xml); const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *type = crm_meta_value(action->params, "stonith_action"); char *transition_key = NULL; const char *priority_delay = NULL; gboolean invalid_action = FALSE; guint stonith_timeout = controld_globals.transition_graph->stonith_timeout; CRM_CHECK(id != NULL, invalid_action = TRUE); CRM_CHECK(uuid != NULL, invalid_action = TRUE); CRM_CHECK(type != NULL, invalid_action = TRUE); CRM_CHECK(target != NULL, invalid_action = TRUE); if (invalid_action) { crm_log_xml_warn(action->xml, "BadAction"); return EPROTO; } priority_delay = crm_meta_value(action->params, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY); crm_notice("Requesting fencing (%s) of node %s " CRM_XS " action=%s timeout=%u%s%s", type, target, id, stonith_timeout, priority_delay ? " priority_delay=" : "", priority_delay ? priority_delay : ""); /* Passing NULL means block until we can connect... */ te_connect_stonith(NULL); rc = fence_with_delay(target, type, priority_delay); transition_key = pcmk__transition_key(controld_globals.transition_graph->id, action->id, 0, controld_globals.te_uuid), stonith_api->cmds->register_callback(stonith_api, rc, (int) (stonith_timeout / 1000), st_opt_timeout_updates, transition_key, "tengine_stonith_callback", tengine_stonith_callback); return pcmk_rc_ok; } bool controld_verify_stonith_watchdog_timeout(const char *value) { const char *our_nodename = controld_globals.our_nodename; gboolean rv = TRUE; if (stonith_api && (stonith_api->state != stonith_disconnected) && stonith__watchdog_fencing_enabled_for_node_api(stonith_api, our_nodename)) { rv = pcmk__valid_sbd_timeout(value); } return rv; } /* end stonith API client functions */ /* * stonith history synchronization * * Each node's fencer keeps track of a cluster-wide fencing history. When a node * joins or leaves, we need to synchronize the history across all nodes. */ static crm_trigger_t *stonith_history_sync_trigger = NULL; static mainloop_timer_t *stonith_history_sync_timer_short = NULL; static mainloop_timer_t *stonith_history_sync_timer_long = NULL; void te_cleanup_stonith_history_sync(stonith_t *st, bool free_timers) { if (free_timers) { mainloop_timer_del(stonith_history_sync_timer_short); stonith_history_sync_timer_short = NULL; mainloop_timer_del(stonith_history_sync_timer_long); stonith_history_sync_timer_long = NULL; } else { mainloop_timer_stop(stonith_history_sync_timer_short); mainloop_timer_stop(stonith_history_sync_timer_long); } if (st) { st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY_SYNCED); } } static void tengine_stonith_history_synced(stonith_t *st, stonith_event_t *st_event) { te_cleanup_stonith_history_sync(st, FALSE); crm_debug("Fence-history synced - cancel all timers"); } static gboolean stonith_history_sync_set_trigger(gpointer user_data) { mainloop_set_trigger(stonith_history_sync_trigger); return FALSE; } void te_trigger_stonith_history_sync(bool long_timeout) { /* trigger a sync in 5s to give more nodes the * chance to show up so that we don't create * unnecessary stonith-history-sync traffic * * the long timeout of 30s is there as a fallback * so that after a successful connection to fenced * we will wait for 30s for the DC to trigger a * history-sync * if this doesn't happen we trigger a sync locally * (e.g. fenced segfaults and is restarted by pacemakerd) */ /* as we are finally checking the stonith-connection * in do_stonith_history_sync we should be fine * leaving stonith_history_sync_time & stonith_history_sync_trigger * around */ if (stonith_history_sync_trigger == NULL) { stonith_history_sync_trigger = mainloop_add_trigger(G_PRIORITY_LOW, do_stonith_history_sync, NULL); } if (long_timeout) { if(stonith_history_sync_timer_long == NULL) { stonith_history_sync_timer_long = mainloop_timer_add("history_sync_long", 30000, FALSE, stonith_history_sync_set_trigger, NULL); } crm_info("Fence history will be synchronized cluster-wide within 30 seconds"); mainloop_timer_start(stonith_history_sync_timer_long); } else { if(stonith_history_sync_timer_short == NULL) { stonith_history_sync_timer_short = mainloop_timer_add("history_sync_short", 5000, FALSE, stonith_history_sync_set_trigger, NULL); } crm_info("Fence history will be synchronized cluster-wide within 5 seconds"); mainloop_timer_start(stonith_history_sync_timer_short); } } /* end stonith history synchronization functions */ diff --git a/daemons/controld/controld_join_dc.c b/daemons/controld/controld_join_dc.c index 30817ba196..fb55e84394 100644 --- a/daemons/controld/controld_join_dc.c +++ b/daemons/controld/controld_join_dc.c @@ -1,958 +1,958 @@ /* * Copyright 2004-2022 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 static char *max_generation_from = NULL; static xmlNodePtr max_generation_xml = NULL; /*! * \internal * \brief Nodes from which a CIB sync has failed since the peer joined * * This table is of the form (node_name -> join_id). \p node_name is * the name of a client node from which a CIB \p sync_from() call has failed in * \p do_dc_join_finalize() since the client joined the cluster as a peer. * \p join_id is the ID of the join round in which the \p sync_from() failed, * and is intended for use in nack log messages. */ static GHashTable *failed_sync_nodes = NULL; void finalize_join_for(gpointer key, gpointer value, gpointer user_data); void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data); gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); /* Numeric counter used to identify join rounds (an unsigned int would be * appropriate, except we get and set it in XML as int) */ static int current_join_id = 0; /*! * \internal * \brief Destroy the hash table containing failed sync nodes */ void controld_destroy_failed_sync_table(void) { if (failed_sync_nodes != NULL) { g_hash_table_destroy(failed_sync_nodes); failed_sync_nodes = NULL; } } /*! * \internal * \brief Remove a node from the failed sync nodes table if present * * \param[in] node_name Node name to remove */ void controld_remove_failed_sync_node(const char *node_name) { if (failed_sync_nodes != NULL) { g_hash_table_remove(failed_sync_nodes, (gchar *) node_name); } } /*! * \internal * \brief Add to a hash table a node whose CIB failed to sync * * \param[in] node_name Name of node whose CIB failed to sync * \param[in] join_id Join round when the failure occurred */ static void record_failed_sync_node(const char *node_name, gint join_id) { if (failed_sync_nodes == NULL) { failed_sync_nodes = pcmk__strikey_table(g_free, NULL); } /* If the node is already in the table then we failed to nack it during the * filter offer step */ CRM_LOG_ASSERT(g_hash_table_insert(failed_sync_nodes, g_strdup(node_name), GINT_TO_POINTER(join_id))); } /*! * \internal * \brief Look up a node name in the failed sync table * * \param[in] node_name Name of node to look up * \param[out] join_id Where to store the join ID of when the sync failed * * \return Standard Pacemaker return code. Specifically, \p pcmk_rc_ok if the * node name was found, or \p pcmk_rc_node_unknown otherwise. * \note \p *join_id is set to -1 if the node is not found. */ static int lookup_failed_sync_node(const char *node_name, gint *join_id) { *join_id = -1; if (failed_sync_nodes != NULL) { gpointer result = g_hash_table_lookup(failed_sync_nodes, (gchar *) node_name); if (result != NULL) { *join_id = GPOINTER_TO_INT(result); return pcmk_rc_ok; } } return pcmk_rc_node_unknown; } void crm_update_peer_join(const char *source, crm_node_t * node, enum crm_join_phase phase) { enum crm_join_phase last = 0; CRM_CHECK(node != NULL, return); /* Remote nodes do not participate in joins */ if (pcmk_is_set(node->flags, crm_remote_node)) { return; } last = node->join; if(phase == last) { crm_trace("Node %s join-%d phase is still %s " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(last), node->id, source); } else if ((phase <= crm_join_none) || (phase == (last + 1))) { node->join = phase; crm_trace("Node %s join-%d phase is now %s (was %s) " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(phase), crm_join_phase_str(last), node->id, source); } else { crm_warn("Rejecting join-%d phase update for node %s because " "can't go from %s to %s " CRM_XS " nodeid=%u source=%s", current_join_id, node->uname, crm_join_phase_str(last), crm_join_phase_str(phase), node->id, source); } } static void start_join_round(void) { GHashTableIter iter; crm_node_t *peer = NULL; crm_debug("Starting new join round join-%d", current_join_id); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { crm_update_peer_join(__func__, peer, crm_join_none); } if (max_generation_from != NULL) { free(max_generation_from); max_generation_from = NULL; } if (max_generation_xml != NULL) { free_xml(max_generation_xml); max_generation_xml = NULL; } controld_clear_fsa_input_flags(R_HAVE_CIB|R_CIB_ASKED); } /*! * \internal * \brief Create a join message from the DC * * \param[in] join_op Join operation name * \param[in] host_to Recipient of message */ static xmlNode * create_dc_message(const char *join_op, const char *host_to) { xmlNode *msg = create_request(join_op, NULL, host_to, CRM_SYSTEM_CRMD, CRM_SYSTEM_DC, NULL); /* Identify which election this is a part of */ crm_xml_add_int(msg, F_CRM_JOIN_ID, current_join_id); /* Add a field specifying whether the DC is shutting down. This keeps the * joining node from fencing the old DC if it becomes the new DC. */ pcmk__xe_set_bool_attr(msg, F_CRM_DC_LEAVING, pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)); return msg; } static void join_make_offer(gpointer key, gpointer value, gpointer user_data) { xmlNode *offer = NULL; crm_node_t *member = (crm_node_t *)value; CRM_ASSERT(member != NULL); if (crm_is_peer_active(member) == FALSE) { crm_info("Not making join-%d offer to inactive node %s", current_join_id, (member->uname? member->uname : "with unknown name")); if(member->expected == NULL && pcmk__str_eq(member->state, CRM_NODE_LOST, pcmk__str_casei)) { /* You would think this unsafe, but in fact this plus an * active resource is what causes it to be fenced. * * Yes, this does mean that any node that dies at the same * time as the old DC and is not running resource (still) * won't be fenced. * * I'm not happy about this either. */ pcmk__update_peer_expected(__func__, member, CRMD_JOINSTATE_DOWN); } return; } if (member->uname == NULL) { crm_info("Not making join-%d offer to node uuid %s with unknown name", current_join_id, member->uuid); return; } if (controld_globals.membership_id != crm_peer_seq) { controld_globals.membership_id = crm_peer_seq; crm_info("Making join-%d offers based on membership event %llu", current_join_id, crm_peer_seq); } if(user_data && member->join > crm_join_none) { crm_info("Not making join-%d offer to already known node %s (%s)", current_join_id, member->uname, crm_join_phase_str(member->join)); return; } crm_update_peer_join(__func__, (crm_node_t*)member, crm_join_none); offer = create_dc_message(CRM_OP_JOIN_OFFER, member->uname); // Advertise our feature set so the joining node can bail if not compatible crm_xml_add(offer, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); crm_info("Sending join-%d offer to %s", current_join_id, member->uname); send_cluster_message(member, crm_msg_crmd, offer, TRUE); free_xml(offer); crm_update_peer_join(__func__, member, crm_join_welcomed); } /* A_DC_JOIN_OFFER_ALL */ void do_dc_join_offer_all(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) { int count; /* Reset everyone's status back to down or in_ccm in the CIB. * Any nodes that are active in the CIB but not in the cluster membership * will be seen as offline by the scheduler anyway. */ current_join_id++; start_join_round(); // do_update_cib_nodes(TRUE, __func__); update_dc(NULL); if (cause == C_HA_MESSAGE && current_input == I_NODE_JOIN) { crm_info("A new node joined the cluster"); } g_hash_table_foreach(crm_peer_cache, join_make_offer, NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } /* A_DC_JOIN_OFFER_ONE */ void do_dc_join_offer_one(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) { crm_node_t *member; ha_msg_input_t *welcome = NULL; int count; const char *join_to = NULL; if (msg_data->data == NULL) { crm_info("Making join-%d offers to any unconfirmed nodes " "because an unknown node joined", current_join_id); g_hash_table_foreach(crm_peer_cache, join_make_offer, &member); check_join_state(cur_state, __func__); return; } welcome = fsa_typed_data(fsa_dt_ha_msg); if (welcome == NULL) { // fsa_typed_data() already logged an error return; } join_to = crm_element_value(welcome->msg, F_CRM_HOST_FROM); if (join_to == NULL) { crm_err("Can't make join-%d offer to unknown node", current_join_id); return; } member = crm_get_peer(0, join_to); /* It is possible that a node will have been sick or starting up when the * original offer was made. However, it will either re-announce itself in * due course, or we can re-store the original offer on the client. */ crm_update_peer_join(__func__, member, crm_join_none); join_make_offer(NULL, member, NULL); /* If the offer isn't to the local node, make an offer to the local node as * well, to ensure the correct value for max_generation_from. */ if (strcasecmp(join_to, controld_globals.our_nodename) != 0) { member = crm_get_peer(0, controld_globals.our_nodename); join_make_offer(NULL, member, NULL); } /* This was a genuine join request; cancel any existing transition and * invoke the scheduler. */ abort_transition(INFINITY, pcmk__graph_restart, "Node join", NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } static int compare_int_fields(xmlNode * left, xmlNode * right, const char *field) { const char *elem_l = crm_element_value(left, field); const char *elem_r = crm_element_value(right, field); long long int_elem_l; long long int_elem_r; pcmk__scan_ll(elem_l, &int_elem_l, -1LL); pcmk__scan_ll(elem_r, &int_elem_r, -1LL); if (int_elem_l < int_elem_r) { return -1; } else if (int_elem_l > int_elem_r) { return 1; } return 0; } /* A_DC_JOIN_PROCESS_REQ */ void do_dc_join_filter_offer(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) { xmlNode *generation = NULL; int cmp = 0; int join_id = -1; int count = 0; gint value = 0; gboolean ack_nack_bool = TRUE; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *join_from = crm_element_value(join_ack->msg, F_CRM_HOST_FROM); const char *ref = crm_element_value(join_ack->msg, F_CRM_REFERENCE); const char *join_version = crm_element_value(join_ack->msg, XML_ATTR_CRM_VERSION); crm_node_t *join_node = NULL; if (join_from == NULL) { crm_err("Ignoring invalid join request without node name"); return; } join_node = crm_get_peer(0, join_from); crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id); if (join_id != current_join_id) { crm_debug("Ignoring join-%d request from %s because we are on join-%d", join_id, join_from, current_join_id); check_join_state(cur_state, __func__); return; } generation = join_ack->xml; if (max_generation_xml != NULL && generation != NULL) { int lpc = 0; const char *attributes[] = { XML_ATTR_GENERATION_ADMIN, XML_ATTR_GENERATION, XML_ATTR_NUMUPDATES, }; for (lpc = 0; cmp == 0 && lpc < PCMK__NELEM(attributes); lpc++) { cmp = compare_int_fields(max_generation_xml, generation, attributes[lpc]); } } if (ref == NULL) { ref = "none"; // for logging only } if (lookup_failed_sync_node(join_from, &value) == pcmk_rc_ok) { crm_err("Rejecting join-%d request from node %s because we failed to " "sync its CIB in join-%d " CRM_XS " ref=%s", join_id, join_from, value, ref); ack_nack_bool = FALSE; } else if (!crm_is_peer_active(join_node)) { if (match_down_event(join_from) != NULL) { /* The join request was received after the node was fenced or * otherwise shutdown in a way that we're aware of. No need to log * an error in this rare occurrence; we know the client was recently * shut down, and receiving a lingering in-flight request is not * cause for alarm. */ crm_debug("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } else { crm_err("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } ack_nack_bool = FALSE; } else if (generation == NULL) { crm_err("Rejecting invalid join-%d request from node %s " "missing CIB generation " CRM_XS " ref=%s", join_id, join_from, ref); ack_nack_bool = FALSE; } else if ((join_version == NULL) || !feature_set_compatible(CRM_FEATURE_SET, join_version)) { crm_err("Rejecting join-%d request from node %s because feature set %s" " is incompatible with ours (%s) " CRM_XS " ref=%s", join_id, join_from, (join_version? join_version : "pre-3.1.0"), CRM_FEATURE_SET, ref); ack_nack_bool = FALSE; } else if (max_generation_xml == NULL) { const char *validation = crm_element_value(generation, XML_ATTR_VALIDATION); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with first CIB " "generation) due to unknown schema version %s " CRM_XS " ref=%s", join_id, join_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with first CIB " "generation) " CRM_XS " ref=%s", join_id, join_from, ref); max_generation_xml = copy_xml(generation); pcmk__str_update(&max_generation_from, join_from); } } else if ((cmp < 0) || ((cmp == 0) && pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei))) { const char *validation = crm_element_value(generation, XML_ATTR_VALIDATION); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with better CIB " "generation than current best from %s) due to unknown " "schema version %s " CRM_XS " ref=%s", join_id, join_from, max_generation_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with better CIB " "generation than current best from %s) " CRM_XS " ref=%s", join_id, join_from, max_generation_from, ref); crm_log_xml_debug(max_generation_xml, "Old max generation"); crm_log_xml_debug(generation, "New max generation"); free_xml(max_generation_xml); max_generation_xml = copy_xml(join_ack->xml); pcmk__str_update(&max_generation_from, join_from); } } else { crm_debug("Accepting join-%d request from %s " CRM_XS " ref=%s", join_id, join_from, ref); } if (!ack_nack_bool) { if (compare_version(join_version, "3.17.0") < 0) { /* Clients with CRM_FEATURE_SET < 3.17.0 may respawn infinitely * after a nack message, don't send one */ crm_update_peer_join(__func__, join_node, crm_join_nack_quiet); } else { crm_update_peer_join(__func__, join_node, crm_join_nack); } pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_NACK); } else { crm_update_peer_join(__func__, join_node, crm_join_integrated); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); } count = crmd_join_phase_count(crm_join_integrated); crm_debug("%d node%s currently integrated in join-%d", count, pcmk__plural_s(count), join_id); if (check_join_state(cur_state, __func__) == FALSE) { // Don't waste time by invoking the scheduler yet count = crmd_join_phase_count(crm_join_welcomed); crm_debug("Waiting on join-%d requests from %d outstanding node%s", join_id, count, pcmk__plural_s(count)); } } /* A_DC_JOIN_FINALIZE */ void do_dc_join_finalize(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) { char *sync_from = NULL; int rc = pcmk_ok; int count_welcomed = crmd_join_phase_count(crm_join_welcomed); int count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); /* This we can do straight away and avoid clients timing us out * while we compute the latest CIB */ if (count_welcomed != 0) { crm_debug("Waiting on join-%d requests from %d outstanding node%s " "before finalizing join", current_join_id, count_welcomed, pcmk__plural_s(count_welcomed)); crmd_join_phase_log(LOG_DEBUG); /* crmd_fsa_stall(FALSE); Needed? */ return; } else if (count_finalizable == 0) { crm_debug("Finalization not needed for join-%d at the current time", current_join_id); crmd_join_phase_log(LOG_DEBUG); check_join_state(controld_globals.fsa_state, __func__); return; } controld_clear_fsa_input_flags(R_HAVE_CIB); if (pcmk__str_eq(max_generation_from, controld_globals.our_nodename, pcmk__str_null_matches|pcmk__str_casei)) { controld_set_fsa_input_flags(R_HAVE_CIB); } if (pcmk_is_set(controld_globals.fsa_input_register, R_IN_TRANSITION)) { crm_warn("Delaying join-%d finalization while transition in progress", current_join_id); crmd_join_phase_log(LOG_DEBUG); crmd_fsa_stall(FALSE); return; } if ((max_generation_from != NULL) && !pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { /* ask for the agreed best CIB */ pcmk__str_update(&sync_from, max_generation_from); controld_set_fsa_input_flags(R_CIB_ASKED); crm_notice("Finalizing join-%d for %d node%s (sync'ing CIB from %s)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable), sync_from); crm_log_xml_notice(max_generation_xml, "Requested CIB version"); } else { /* Send _our_ CIB out to everyone */ pcmk__str_update(&sync_from, controld_globals.our_nodename); crm_debug("Finalizing join-%d for %d node%s (sync'ing from local CIB)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable)); crm_log_xml_debug(max_generation_xml, "Requested CIB version"); } crmd_join_phase_log(LOG_DEBUG); rc = controld_globals.cib_conn->cmds->sync_from(controld_globals.cib_conn, sync_from, NULL, cib_quorum_override); fsa_register_cib_callback(rc, FALSE, sync_from, finalize_sync_callback); } void free_max_generation(void) { free(max_generation_from); max_generation_from = NULL; free_xml(max_generation_xml); max_generation_xml = NULL; } void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { CRM_LOG_ASSERT(-EPERM != rc); controld_clear_fsa_input_flags(R_CIB_ASKED); if (rc != pcmk_ok) { const char *sync_from = (const char *) user_data; do_crm_log(((rc == -pcmk_err_old_data)? LOG_WARNING : LOG_ERR), "Could not sync CIB from %s in join-%d: %s", sync_from, current_join_id, pcmk_strerror(rc)); record_failed_sync_node(sync_from, current_join_id); /* restart the whole join process */ register_fsa_error_adv(C_FSA_INTERNAL, I_ELECTION_DC, NULL, NULL, __func__); } else if (!AM_I_DC) { crm_debug("Sync'ed CIB for join-%d but no longer DC", current_join_id); } else if (controld_globals.fsa_state != S_FINALIZE_JOIN) { crm_debug("Sync'ed CIB for join-%d but no longer in S_FINALIZE_JOIN " "(%s)", current_join_id, fsa_state2string(controld_globals.fsa_state)); } else { controld_set_fsa_input_flags(R_HAVE_CIB); controld_clear_fsa_input_flags(R_CIB_ASKED); /* make sure dc_uuid is re-set to us */ if (!check_join_state(controld_globals.fsa_state, __func__)) { int count_finalizable = 0; count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); crm_debug("Notifying %d node%s of join-%d results", count_finalizable, pcmk__plural_s(count_finalizable), current_join_id); g_hash_table_foreach(crm_peer_cache, finalize_join_for, NULL); } } } static void join_update_complete_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if (rc == pcmk_ok) { crm_debug("join-%d node history update (via CIB call %d) complete", current_join_id, call_id); check_join_state(controld_globals.fsa_state, __func__); } else { crm_err("join-%d node history update (via CIB call %d) failed: %s " "(next transition may determine resource status incorrectly)", current_join_id, call_id, pcmk_strerror(rc)); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } /* A_DC_JOIN_PROCESS_ACK */ void do_dc_join_ack(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) { int join_id = -1; int call_id = 0; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); enum controld_section_e section = controld_section_lrm; const char *op = crm_element_value(join_ack->msg, F_CRM_TASK); const char *join_from = crm_element_value(join_ack->msg, F_CRM_HOST_FROM); crm_node_t *peer = NULL; // Sanity checks if (join_from == NULL) { crm_warn("Ignoring message received without node identification"); return; } if (op == NULL) { crm_warn("Ignoring message received from %s without task", join_from); return; } if (strcmp(op, CRM_OP_JOIN_CONFIRM)) { crm_debug("Ignoring '%s' message from %s while waiting for '%s'", op, join_from, CRM_OP_JOIN_CONFIRM); return; } if (crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id) != 0) { crm_warn("Ignoring join confirmation from %s without valid join ID", join_from); return; } peer = crm_get_peer(0, join_from); if (peer->join != crm_join_finalized) { crm_info("Ignoring out-of-sequence join-%d confirmation from %s " "(currently %s not %s)", join_id, join_from, crm_join_phase_str(peer->join), crm_join_phase_str(crm_join_finalized)); return; } if (join_id != current_join_id) { crm_err("Rejecting join-%d confirmation from %s " "because currently on join-%d", join_id, join_from, current_join_id); crm_update_peer_join(__func__, peer, crm_join_nack); return; } crm_update_peer_join(__func__, peer, crm_join_confirmed); /* Update CIB with node's current executor state. A new transition will be * triggered later, when the CIB notifies us of the change. */ if (pcmk_is_set(controld_globals.flags, controld_shutdown_lock_enabled)) { section = controld_section_lrm_unlocked; } controld_delete_node_state(join_from, section, cib_scope_local); if (pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei)) { xmlNode *now_dc_lrmd_state = controld_query_executor_state(); if (now_dc_lrmd_state != NULL) { fsa_cib_update(XML_CIB_TAG_STATUS, now_dc_lrmd_state, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); free_xml(now_dc_lrmd_state); crm_debug("Updating local node history for join-%d " "from query result (via CIB call %d)", join_id, call_id); } else { fsa_cib_update(XML_CIB_TAG_STATUS, join_ack->xml, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); crm_warn("Updating local node history from join-%d confirmation " "because query failed (via CIB call %d)", join_id, call_id); } } else { fsa_cib_update(XML_CIB_TAG_STATUS, join_ack->xml, cib_scope_local | cib_quorum_override | cib_can_create, call_id, NULL); crm_debug("Updating node history for %s from join-%d confirmation " "(via CIB call %d)", join_from, join_id, call_id); } fsa_register_cib_callback(call_id, FALSE, NULL, join_update_complete_callback); } void finalize_join_for(gpointer key, gpointer value, gpointer user_data) { xmlNode *acknak = NULL; xmlNode *tmp1 = NULL; crm_node_t *join_node = value; const char *join_to = join_node->uname; bool integrated = false; switch (join_node->join) { case crm_join_integrated: integrated = true; break; case crm_join_nack: case crm_join_nack_quiet: break; default: crm_trace("Not updating non-integrated and non-nacked node %s (%s) " "for join-%d", join_to, crm_join_phase_str(join_node->join), current_join_id); return; } /* Update the element with the node's name and UUID, in case they * weren't known before */ crm_trace("Updating node name and UUID in CIB for %s", join_to); tmp1 = create_xml_node(NULL, XML_CIB_TAG_NODE); - set_uuid(tmp1, XML_ATTR_UUID, join_node); + set_uuid(tmp1, XML_ATTR_ID, join_node); crm_xml_add(tmp1, XML_ATTR_UNAME, join_to); fsa_cib_anon_update(XML_CIB_TAG_NODES, tmp1); free_xml(tmp1); if (join_node->join == crm_join_nack_quiet) { crm_trace("Not sending nack message to node %s with feature set older " "than 3.17.0", join_to); return; } join_node = crm_get_peer(0, join_to); if (!crm_is_peer_active(join_node)) { /* * NACK'ing nodes that the membership layer doesn't know about yet * simply creates more churn * * Better to leave them waiting and let the join restart when * the new membership event comes in * * All other NACKs (due to versions etc) should still be processed */ pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_PENDING); return; } // Acknowledge or nack node's join request crm_debug("%sing join-%d request from %s", integrated? "Acknowledg" : "Nack", current_join_id, join_to); acknak = create_dc_message(CRM_OP_JOIN_ACKNAK, join_to); pcmk__xe_set_bool_attr(acknak, CRM_OP_JOIN_ACKNAK, integrated); if (integrated) { // No change needed for a nacked node crm_update_peer_join(__func__, join_node, crm_join_finalized); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); } send_cluster_message(join_node, crm_msg_crmd, acknak, TRUE); free_xml(acknak); return; } gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source) { static unsigned long long highest_seq = 0; if (controld_globals.membership_id != crm_peer_seq) { crm_debug("join-%d: Membership changed from %llu to %llu " CRM_XS " highest=%llu state=%s for=%s", current_join_id, controld_globals.membership_id, crm_peer_seq, highest_seq, fsa_state2string(cur_state), source); if(highest_seq < crm_peer_seq) { /* Don't spam the FSA with duplicates */ highest_seq = crm_peer_seq; register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } } else if (cur_state == S_INTEGRATION) { if (crmd_join_phase_count(crm_join_welcomed) == 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Integration of %d peer%s complete " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); register_fsa_input_before(C_FSA_INTERNAL, I_INTEGRATED, NULL); return TRUE; } } else if (cur_state == S_FINALIZE_JOIN) { if (!pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { crm_debug("join-%d: Delaying finalization until we have CIB " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); return TRUE; } else if (crmd_join_phase_count(crm_join_welcomed) != 0) { int count = crmd_join_phase_count(crm_join_welcomed); crm_debug("join-%d: Still waiting on %d welcomed node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_integrated) != 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Still waiting on %d integrated node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_finalized) != 0) { int count = crmd_join_phase_count(crm_join_finalized); crm_debug("join-%d: Still waiting on %d finalized node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else { crm_debug("join-%d: Complete " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); register_fsa_input_later(C_FSA_INTERNAL, I_FINALIZED, NULL); return TRUE; } } return FALSE; } void do_dc_join_final(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) { crm_debug("Ensuring DC, quorum and node attributes are up-to-date"); crm_update_quorum(crm_have_quorum, TRUE); } int crmd_join_phase_count(enum crm_join_phase phase) { int count = 0; crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { if(peer->join == phase) { count++; } } return count; } void crmd_join_phase_log(int level) { crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { do_crm_log(level, "join-%d: %s=%s", current_join_id, peer->uname, crm_join_phase_str(peer->join)); } } diff --git a/daemons/controld/controld_membership.c b/daemons/controld/controld_membership.c index 69bec96466..4c940bcf76 100644 --- a/daemons/controld/controld_membership.c +++ b/daemons/controld/controld_membership.c @@ -1,467 +1,467 @@ /* * Copyright 2004-2022 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. */ /* put these first so that uuid_t is defined without conflicts */ #include #include #include #include #include #include #include #include void post_cache_update(int instance); extern gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); static void reap_dead_nodes(gpointer key, gpointer value, gpointer user_data) { crm_node_t *node = value; if (crm_is_peer_active(node) == FALSE) { crm_update_peer_join(__func__, node, crm_join_none); if(node && node->uname) { if (pcmk__str_eq(controld_globals.our_nodename, node->uname, pcmk__str_casei)) { crm_err("We're not part of the cluster anymore"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); } else if (!AM_I_DC && pcmk__str_eq(node->uname, controld_globals.dc_name, pcmk__str_casei)) { crm_warn("Our DC node (%s) left the cluster", node->uname); register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL); } } if ((controld_globals.fsa_state == S_INTEGRATION) || (controld_globals.fsa_state == S_FINALIZE_JOIN)) { check_join_state(controld_globals.fsa_state, __func__); } if ((node != NULL) && (node->uuid != NULL)) { fail_incompletable_actions(controld_globals.transition_graph, node->uuid); } } } void post_cache_update(int instance) { xmlNode *no_op = NULL; crm_peer_seq = instance; crm_debug("Updated cache after membership event %d.", instance); g_hash_table_foreach(crm_peer_cache, reap_dead_nodes, NULL); controld_set_fsa_input_flags(R_MEMBERSHIP); if (AM_I_DC) { populate_cib_nodes(node_update_quick | node_update_cluster | node_update_peer | node_update_expected, __func__); } /* * If we lost nodes, we should re-check the election status * Safe to call outside of an election */ controld_set_fsa_action_flags(A_ELECTION_CHECK); controld_trigger_fsa(); /* Membership changed, remind everyone we're here. * This will aid detection of duplicate DCs */ no_op = create_request(CRM_OP_NOOP, NULL, NULL, CRM_SYSTEM_CRMD, AM_I_DC ? CRM_SYSTEM_DC : CRM_SYSTEM_CRMD, NULL); send_cluster_message(NULL, crm_msg_crmd, no_op, FALSE); free_xml(no_op); } static void crmd_node_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if (rc == pcmk_ok) { crm_trace("Node update %d complete", call_id); } else if(call_id < pcmk_ok) { crm_err("Node update failed: %s (%d)", pcmk_strerror(call_id), call_id); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } else { crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } /*! * \internal * \brief Create an XML node state tag with updates * * \param[in,out] node Node whose state will be used for update * \param[in] flags Bitmask of node_update_flags indicating what to update * \param[in,out] parent XML node to contain update (or NULL) * \param[in] source Who requested the update (only used for logging) * * \return Pointer to created node state tag */ xmlNode * create_node_state_update(crm_node_t *node, int flags, xmlNode *parent, const char *source) { const char *value = NULL; xmlNode *node_state; if (!node->state) { crm_info("Node update for %s cancelled: no state, not seen yet", node->uname); return NULL; } node_state = create_xml_node(parent, XML_CIB_TAG_STATE); if (pcmk_is_set(node->flags, crm_remote_node)) { pcmk__xe_set_bool_attr(node_state, XML_NODE_IS_REMOTE, true); } - set_uuid(node_state, XML_ATTR_UUID, node); + set_uuid(node_state, XML_ATTR_ID, node); - if (crm_element_value(node_state, XML_ATTR_UUID) == NULL) { + if (crm_element_value(node_state, XML_ATTR_ID) == NULL) { crm_info("Node update for %s cancelled: no id", node->uname); free_xml(node_state); return NULL; } crm_xml_add(node_state, XML_ATTR_UNAME, node->uname); if ((flags & node_update_cluster) && node->state) { pcmk__xe_set_bool_attr(node_state, XML_NODE_IN_CLUSTER, pcmk__str_eq(node->state, CRM_NODE_MEMBER, pcmk__str_casei)); } if (!pcmk_is_set(node->flags, crm_remote_node)) { if (flags & node_update_peer) { value = OFFLINESTATUS; if (pcmk_is_set(node->processes, crm_get_cluster_proc())) { value = ONLINESTATUS; } crm_xml_add(node_state, XML_NODE_IS_PEER, value); } if (flags & node_update_join) { if (node->join <= crm_join_none) { value = CRMD_JOINSTATE_DOWN; } else { value = CRMD_JOINSTATE_MEMBER; } crm_xml_add(node_state, XML_NODE_JOIN_STATE, value); } if (flags & node_update_expected) { crm_xml_add(node_state, XML_NODE_EXPECTED, node->expected); } } crm_xml_add(node_state, XML_ATTR_ORIGIN, source); return node_state; } static void remove_conflicting_node_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *node_uuid = user_data; do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE, "Deletion of the unknown conflicting node \"%s\": %s (rc=%d)", node_uuid, pcmk_strerror(rc), rc); } static void search_conflicting_node_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *new_node_uuid = user_data; xmlNode *node_xml = NULL; if (rc != pcmk_ok) { if (rc != -ENXIO) { crm_notice("Searching conflicting nodes for %s failed: %s (%d)", new_node_uuid, pcmk_strerror(rc), rc); } return; } else if (output == NULL) { return; } if (pcmk__str_eq(crm_element_name(output), XML_CIB_TAG_NODE, pcmk__str_casei)) { node_xml = output; } else { node_xml = pcmk__xml_first_child(output); } for (; node_xml != NULL; node_xml = pcmk__xml_next(node_xml)) { const char *node_uuid = NULL; const char *node_uname = NULL; GHashTableIter iter; crm_node_t *node = NULL; gboolean known = FALSE; if (!pcmk__str_eq(crm_element_name(node_xml), XML_CIB_TAG_NODE, pcmk__str_casei)) { continue; } node_uuid = crm_element_value(node_xml, XML_ATTR_ID); node_uname = crm_element_value(node_xml, XML_ATTR_UNAME); if (node_uuid == NULL || node_uname == NULL) { continue; } g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (node->uuid && pcmk__str_eq(node->uuid, node_uuid, pcmk__str_casei) && node->uname && pcmk__str_eq(node->uname, node_uname, pcmk__str_casei)) { known = TRUE; break; } } if (known == FALSE) { cib_t *cib_conn = controld_globals.cib_conn; int delete_call_id = 0; xmlNode *node_state_xml = NULL; crm_notice("Deleting unknown node %s/%s which has conflicting uname with %s", node_uuid, node_uname, new_node_uuid); delete_call_id = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_NODES, node_xml, cib_scope_local |cib_quorum_override); fsa_register_cib_callback(delete_call_id, FALSE, strdup(node_uuid), remove_conflicting_node_callback); node_state_xml = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(node_state_xml, XML_ATTR_ID, node_uuid); crm_xml_add(node_state_xml, XML_ATTR_UNAME, node_uname); delete_call_id = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_STATUS, node_state_xml, cib_scope_local |cib_quorum_override); fsa_register_cib_callback(delete_call_id, FALSE, strdup(node_uuid), remove_conflicting_node_callback); free_xml(node_state_xml); } } } static void node_list_update_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if(call_id < pcmk_ok) { crm_err("Node list update failed: %s (%d)", pcmk_strerror(call_id), call_id); crm_log_xml_debug(msg, "update:failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } else if(rc < pcmk_ok) { crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "update:failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } void populate_cib_nodes(enum node_update_flags flags, const char *source) { cib_t *cib_conn = controld_globals.cib_conn; int call_id = 0; gboolean from_hashtable = TRUE; int call_options = cib_scope_local | cib_quorum_override; xmlNode *node_list = create_xml_node(NULL, XML_CIB_TAG_NODES); #if SUPPORT_COROSYNC if (!pcmk_is_set(flags, node_update_quick) && is_corosync_cluster()) { from_hashtable = pcmk__corosync_add_nodes(node_list); } #endif if (from_hashtable) { GHashTableIter iter; crm_node_t *node = NULL; GString *xpath = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *new_node = NULL; if ((node->uuid != NULL) && (node->uname != NULL)) { crm_trace("Creating node entry for %s/%s", node->uname, node->uuid); if (xpath == NULL) { xpath = g_string_sized_new(512); } else { g_string_truncate(xpath, 0); } /* We need both to be valid */ new_node = create_xml_node(node_list, XML_CIB_TAG_NODE); crm_xml_add(new_node, XML_ATTR_ID, node->uuid); crm_xml_add(new_node, XML_ATTR_UNAME, node->uname); /* Search and remove unknown nodes with the conflicting uname from CIB */ pcmk__g_strcat(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE "[@" XML_ATTR_UNAME "='", node->uname, "']" "[@" XML_ATTR_ID "!='", node->uuid, "']", NULL); call_id = cib_conn->cmds->query(cib_conn, (const char *) xpath->str, NULL, cib_scope_local|cib_xpath); fsa_register_cib_callback(call_id, FALSE, strdup(node->uuid), search_conflicting_node_callback); } } if (xpath != NULL) { g_string_free(xpath, TRUE); } } crm_trace("Populating section from %s", from_hashtable ? "hashtable" : "cluster"); fsa_cib_update(XML_CIB_TAG_NODES, node_list, call_options, call_id, NULL); fsa_register_cib_callback(call_id, FALSE, NULL, node_list_update_callback); free_xml(node_list); if (call_id >= pcmk_ok && crm_peer_cache != NULL && AM_I_DC) { /* * There is no need to update the local CIB with our values if * we've not seen valid membership data */ GHashTableIter iter; crm_node_t *node = NULL; node_list = create_xml_node(NULL, XML_CIB_TAG_STATUS); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { create_node_state_update(node, flags, node_list, source); } if (crm_remote_peer_cache) { g_hash_table_iter_init(&iter, crm_remote_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { create_node_state_update(node, flags, node_list, source); } } fsa_cib_update(XML_CIB_TAG_STATUS, node_list, call_options, call_id, NULL); fsa_register_cib_callback(call_id, FALSE, NULL, crmd_node_update_complete); free_xml(node_list); } } static void cib_quorum_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if (rc == pcmk_ok) { crm_trace("Quorum update %d complete", call_id); } else { crm_err("Quorum update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } void crm_update_quorum(gboolean quorum, gboolean force_update) { bool has_quorum = pcmk_is_set(controld_globals.flags, controld_has_quorum); if (quorum) { controld_set_global_flags(controld_ever_had_quorum); } else if (pcmk_all_flags_set(controld_globals.flags, controld_ever_had_quorum |controld_no_quorum_suicide)) { pcmk__panic(__func__); } if (AM_I_DC && ((has_quorum && !quorum) || (!has_quorum && quorum) || force_update)) { int call_id = 0; xmlNode *update = NULL; int call_options = cib_scope_local | cib_quorum_override; update = create_xml_node(NULL, XML_TAG_CIB); crm_xml_add_int(update, XML_ATTR_HAVE_QUORUM, quorum); crm_xml_add(update, XML_ATTR_DC_UUID, controld_globals.our_uuid); fsa_cib_update(XML_TAG_CIB, update, call_options, call_id, NULL); crm_debug("Updating quorum status to %s (call=%d)", pcmk__btoa(quorum), call_id); fsa_register_cib_callback(call_id, FALSE, NULL, cib_quorum_update_complete); free_xml(update); /* Quorum changes usually cause a new transition via other activity: * quorum gained via a node joining will abort via the node join, * and quorum lost via a node leaving will usually abort via resource * activity and/or fencing. * * However, it is possible that nothing else causes a transition (e.g. * someone forces quorum via corosync-cmaptcl, or quorum is lost due to * a node in standby shutting down cleanly), so here ensure a new * transition is triggered. */ if (quorum) { /* If quorum was gained, abort after a short delay, in case multiple * nodes are joining around the same time, so the one that brings us * to quorum doesn't cause all the remaining ones to be fenced. */ abort_after_delay(INFINITY, pcmk__graph_restart, "Quorum gained", 5000); } else { abort_transition(INFINITY, pcmk__graph_restart, "Quorum lost", NULL); } } if (quorum) { controld_set_global_flags(controld_has_quorum); } else { controld_clear_global_flags(controld_has_quorum); } } diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c index b4ca55bab5..58da54cdaa 100644 --- a/daemons/controld/controld_messages.c +++ b/daemons/controld/controld_messages.c @@ -1,1302 +1,1301 @@ /* * Copyright 2004-2022 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 extern void crm_shutdown(int nsig); static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause); static void handle_response(xmlNode *stored_msg); static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause); static enum crmd_fsa_input handle_shutdown_request(xmlNode *stored_msg); static void send_msg_via_ipc(xmlNode * msg, const char *sys); /* debug only, can wrap all it likes */ static int last_data_id = 0; void register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, fsa_data_t * cur_data, void *new_data, const char *raised_from) { /* save the current actions if any */ if (controld_globals.fsa_actions != A_NOTHING) { register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL, I_NULL, cur_data ? cur_data->data : NULL, controld_globals.fsa_actions, TRUE, __func__); } /* reset the action list */ crm_info("Resetting the current action list"); fsa_dump_actions(controld_globals.fsa_actions, "Drop"); controld_globals.fsa_actions = A_NOTHING; /* register the error */ register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from); } void register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, void *data, uint64_t with_actions, gboolean prepend, const char *raised_from) { unsigned old_len = g_list_length(controld_globals.fsa_message_queue); fsa_data_t *fsa_data = NULL; if (raised_from == NULL) { raised_from = ""; } if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) { /* no point doing anything */ crm_err("Cannot add entry to queue: no input and no action"); return; } if (input == I_WAIT_FOR_EVENT) { controld_set_global_flags(controld_fsa_is_stalled); crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d", raised_from, fsa_cause2string(cause), data, old_len); if (old_len > 0) { fsa_dump_queue(LOG_TRACE); prepend = FALSE; } if (data == NULL) { controld_set_fsa_action_flags(with_actions); fsa_dump_actions(with_actions, "Restored"); return; } /* Store everything in the new event and reset * controld_globals.fsa_actions */ with_actions |= controld_globals.fsa_actions; controld_globals.fsa_actions = A_NOTHING; } last_data_id++; crm_trace("%s %s FSA input %d (%s) due to %s, %s data", raised_from, (prepend? "prepended" : "appended"), last_data_id, fsa_input2string(input), fsa_cause2string(cause), (data? "with" : "without")); fsa_data = calloc(1, sizeof(fsa_data_t)); fsa_data->id = last_data_id; fsa_data->fsa_input = input; fsa_data->fsa_cause = cause; fsa_data->origin = raised_from; fsa_data->data = NULL; fsa_data->data_type = fsa_dt_none; fsa_data->actions = with_actions; if (with_actions != A_NOTHING) { crm_trace("Adding actions %.16llx to input", (unsigned long long) with_actions); } if (data != NULL) { switch (cause) { case C_FSA_INTERNAL: case C_CRMD_STATUS_CALLBACK: case C_IPC_MESSAGE: case C_HA_MESSAGE: CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL, crm_err("Bogus data from %s", raised_from)); crm_trace("Copying %s data from %s as cluster message data", fsa_cause2string(cause), raised_from); fsa_data->data = copy_ha_msg_input(data); fsa_data->data_type = fsa_dt_ha_msg; break; case C_LRM_OP_CALLBACK: crm_trace("Copying %s data from %s as lrmd_event_data_t", fsa_cause2string(cause), raised_from); fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data); fsa_data->data_type = fsa_dt_lrm; break; case C_TIMER_POPPED: case C_SHUTDOWN: case C_UNKNOWN: case C_STARTUP: crm_crit("Copying %s data (from %s) is not yet implemented", fsa_cause2string(cause), raised_from); crmd_exit(CRM_EX_SOFTWARE); break; } } /* make sure to free it properly later */ if (prepend) { controld_globals.fsa_message_queue = g_list_prepend(controld_globals.fsa_message_queue, fsa_data); } else { controld_globals.fsa_message_queue = g_list_append(controld_globals.fsa_message_queue, fsa_data); } crm_trace("FSA message queue length is %d", g_list_length(controld_globals.fsa_message_queue)); /* fsa_dump_queue(LOG_TRACE); */ if (old_len == g_list_length(controld_globals.fsa_message_queue)) { crm_err("Couldn't add message to the queue"); } if (input != I_WAIT_FOR_EVENT) { controld_trigger_fsa(); } } void fsa_dump_queue(int log_level) { int offset = 0; for (GList *iter = controld_globals.fsa_message_queue; iter != NULL; iter = iter->next) { fsa_data_t *data = (fsa_data_t *) iter->data; do_crm_log_unlikely(log_level, "queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)", offset++, data->id, fsa_input2string(data->fsa_input), data->origin, data->data, data->data_type, fsa_cause2string(data->fsa_cause)); } } ha_msg_input_t * copy_ha_msg_input(ha_msg_input_t * orig) { ha_msg_input_t *copy = calloc(1, sizeof(ha_msg_input_t)); CRM_ASSERT(copy != NULL); copy->msg = (orig && orig->msg)? copy_xml(orig->msg) : NULL; copy->xml = get_message_xml(copy->msg, F_CRM_DATA); return copy; } void delete_fsa_input(fsa_data_t * fsa_data) { lrmd_event_data_t *op = NULL; xmlNode *foo = NULL; if (fsa_data == NULL) { return; } crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause)); if (fsa_data->data != NULL) { switch (fsa_data->data_type) { case fsa_dt_ha_msg: delete_ha_msg_input(fsa_data->data); break; case fsa_dt_xml: foo = fsa_data->data; free_xml(foo); break; case fsa_dt_lrm: op = (lrmd_event_data_t *) fsa_data->data; lrmd_free_event(op); break; case fsa_dt_none: if (fsa_data->data != NULL) { crm_err("Don't know how to free %s data from %s", fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin); crmd_exit(CRM_EX_SOFTWARE); } break; } crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause)); } free(fsa_data); } /* returns the next message */ fsa_data_t * get_message(void) { fsa_data_t *message = (fsa_data_t *) controld_globals.fsa_message_queue->data; controld_globals.fsa_message_queue = g_list_remove(controld_globals.fsa_message_queue, message); crm_trace("Processing input %d", message->id); return message; } void * fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller) { void *ret_val = NULL; if (fsa_data == NULL) { crm_err("%s: No FSA data available", caller); } else if (fsa_data->data == NULL) { crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin); } else if (fsa_data->data_type != a_type) { crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s", caller, fsa_data->data_type, a_type, fsa_data->origin); CRM_ASSERT(fsa_data->data_type == a_type); } else { ret_val = fsa_data->data; } return ret_val; } /* A_MSG_ROUTE */ void do_msg_route(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) { ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); route_message(msg_data->fsa_cause, input->msg); } void route_message(enum crmd_fsa_cause cause, xmlNode * input) { ha_msg_input_t fsa_input; enum crmd_fsa_input result = I_NULL; fsa_input.msg = input; CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return); /* try passing the buck first */ if (relay_message(input, cause == C_IPC_MESSAGE)) { return; } /* handle locally */ result = handle_message(input, cause); /* done or process later? */ switch (result) { case I_NULL: case I_CIB_OP: case I_ROUTER: case I_NODE_JOIN: case I_JOIN_REQUEST: case I_JOIN_RESULT: break; default: /* Defering local processing of message */ register_fsa_input_later(cause, result, &fsa_input); return; } if (result != I_NULL) { /* add to the front of the queue */ register_fsa_input(cause, result, &fsa_input); } } gboolean relay_message(xmlNode * msg, gboolean originated_locally) { int dest = 1; bool is_for_dc = false; bool is_for_dcib = false; bool is_for_te = false; bool is_for_crm = false; bool is_for_cib = false; bool is_local = false; const char *host_to = crm_element_value(msg, F_CRM_HOST_TO); const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM); const char *type = crm_element_value(msg, F_TYPE); const char *task = crm_element_value(msg, F_CRM_TASK); const char *ref = crm_element_value(msg, XML_ATTR_REFERENCE); if (ref == NULL) { ref = "without reference ID"; } if (msg == NULL) { crm_warn("Cannot route empty message"); return TRUE; } else if (pcmk__str_eq(task, CRM_OP_HELLO, pcmk__str_casei)) { crm_trace("No routing needed for hello message %s", ref); return TRUE; } else if (!pcmk__str_eq(type, T_CRM, pcmk__str_casei)) { crm_warn("Received invalid message %s: type '%s' not '" T_CRM "'", ref, pcmk__s(type, "")); crm_log_xml_warn(msg, "[bad message type]"); return TRUE; } else if (sys_to == NULL) { crm_warn("Received invalid message %s: no subsystem", ref); crm_log_xml_warn(msg, "[no subsystem]"); return TRUE; } is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0); is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0); is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0); is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0); is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0); is_local = false; if (pcmk__str_empty(host_to)) { if (is_for_dc || is_for_te) { is_local = false; } else if (is_for_crm) { if (pcmk__strcase_any_of(task, CRM_OP_NODE_INFO, PCMK__CONTROLD_CMD_NODES, NULL)) { /* Node info requests do not specify a host, which is normally * treated as "all hosts", because the whole point is that the * client may not know the local node name. Always handle these * requests locally. */ is_local = true; } else { is_local = !originated_locally; } } else { is_local = true; } } else if (pcmk__str_eq(controld_globals.our_nodename, host_to, pcmk__str_casei)) { is_local = true; } else if (is_for_crm && pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { xmlNode *msg_data = get_message_xml(msg, F_CRM_DATA); const char *mode = crm_element_value(msg_data, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_casei)) { // Local delete of an offline node's resource history is_local = true; } } if (is_for_dc || is_for_dcib || is_for_te) { if (AM_I_DC && is_for_te) { crm_trace("Route message %s locally as transition request", ref); send_msg_via_ipc(msg, sys_to); } else if (AM_I_DC) { crm_trace("Route message %s locally as DC request", ref); return FALSE; // More to be done by caller } else if (originated_locally && !pcmk__strcase_any_of(sys_from, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { #if SUPPORT_COROSYNC if (is_corosync_cluster()) { dest = text2msg_type(sys_to); } #endif crm_trace("Relay message %s to DC", ref); send_cluster_message(host_to ? crm_get_peer(0, host_to) : NULL, dest, msg, TRUE); } else { /* Neither the TE nor the scheduler should be sending messages * to DCs on other nodes. By definition, if we are no longer the DC, * then the scheduler's or TE's data should be discarded. */ crm_trace("Discard message %s because we are not DC", ref); } } else if (is_local && (is_for_crm || is_for_cib)) { crm_trace("Route message %s locally as controller request", ref); return FALSE; // More to be done by caller } else if (is_local) { crm_trace("Relay message %s locally to %s", ref, (sys_to? sys_to : "unknown client")); crm_log_xml_trace(msg, "[IPC relay]"); send_msg_via_ipc(msg, sys_to); } else { crm_node_t *node_to = NULL; #if SUPPORT_COROSYNC if (is_corosync_cluster()) { dest = text2msg_type(sys_to); if (dest == crm_msg_none || dest > crm_msg_stonith_ng) { dest = crm_msg_crmd; } } #endif if (host_to) { node_to = pcmk__search_cluster_node_cache(0, host_to); if (node_to == NULL) { crm_warn("Cannot route message %s: Unknown node %s", ref, host_to); return TRUE; } crm_trace("Relay message %s to %s", ref, (node_to->uname? node_to->uname : "peer")); } else { crm_trace("Broadcast message %s to all peers", ref); } send_cluster_message(host_to ? node_to : NULL, dest, msg, TRUE); } return TRUE; // No further processing of message is needed } // Return true if field contains a positive integer static bool authorize_version(xmlNode *message_data, const char *field, const char *client_name, const char *ref, const char *uuid) { const char *version = crm_element_value(message_data, field); long long version_num; if ((pcmk__scan_ll(version, &version_num, -1LL) != pcmk_rc_ok) || (version_num < 0LL)) { crm_warn("Rejected IPC hello from %s: '%s' is not a valid protocol %s " CRM_XS " ref=%s uuid=%s", client_name, ((version == NULL)? "" : version), field, (ref? ref : "none"), uuid); return false; } return true; } /*! * \internal * \brief Check whether a client IPC message is acceptable * * If a given client IPC message is a hello, "authorize" it by ensuring it has * valid information such as a protocol version, and return false indicating * that nothing further needs to be done with the message. If the message is not * a hello, just return true to indicate it needs further processing. * * \param[in] client_msg XML of IPC message * \param[in,out] curr_client If IPC is not proxied, client that sent message * \param[in] proxy_session If IPC is proxied, the session ID * * \return true if message needs further processing, false if it doesn't */ bool controld_authorize_ipc_message(const xmlNode *client_msg, pcmk__client_t *curr_client, const char *proxy_session) { xmlNode *message_data = NULL; const char *client_name = NULL; const char *op = crm_element_value(client_msg, F_CRM_TASK); const char *ref = crm_element_value(client_msg, XML_ATTR_REFERENCE); const char *uuid = (curr_client? curr_client->id : proxy_session); if (uuid == NULL) { crm_warn("IPC message from client rejected: No client identifier " CRM_XS " ref=%s", (ref? ref : "none")); goto rejected; } if (!pcmk__str_eq(CRM_OP_HELLO, op, pcmk__str_casei)) { // Only hello messages need to be authorized return true; } message_data = get_message_xml(client_msg, F_CRM_DATA); client_name = crm_element_value(message_data, "client_name"); if (pcmk__str_empty(client_name)) { crm_warn("IPC hello from client rejected: No client name", CRM_XS " ref=%s uuid=%s", (ref? ref : "none"), uuid); goto rejected; } if (!authorize_version(message_data, "major_version", client_name, ref, uuid)) { goto rejected; } if (!authorize_version(message_data, "minor_version", client_name, ref, uuid)) { goto rejected; } crm_trace("Validated IPC hello from client %s", client_name); if (curr_client) { curr_client->userdata = strdup(client_name); } controld_trigger_fsa(); return false; rejected: if (curr_client) { qb_ipcs_disconnect(curr_client->ipcs); } return false; } static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause) { const char *type = NULL; CRM_CHECK(msg != NULL, return I_NULL); type = crm_element_value(msg, F_CRM_MSG_TYPE); if (pcmk__str_eq(type, XML_ATTR_REQUEST, pcmk__str_none)) { return handle_request(msg, cause); } else if (pcmk__str_eq(type, XML_ATTR_RESPONSE, pcmk__str_none)) { handle_response(msg); return I_NULL; } crm_err("Unknown message type: %s", type); return I_NULL; } static enum crmd_fsa_input handle_failcount_op(xmlNode * stored_msg) { const char *rsc = NULL; const char *uname = NULL; const char *op = NULL; char *interval_spec = NULL; guint interval_ms = 0; gboolean is_remote_node = FALSE; xmlNode *xml_op = get_message_xml(stored_msg, F_CRM_DATA); if (xml_op) { xmlNode *xml_rsc = first_named_child(xml_op, XML_CIB_TAG_RESOURCE); xmlNode *xml_attrs = first_named_child(xml_op, XML_TAG_ATTRS); if (xml_rsc) { rsc = ID(xml_rsc); } if (xml_attrs) { op = crm_element_value(xml_attrs, CRM_META "_" XML_RSC_ATTR_CLEAR_OP); crm_element_value_ms(xml_attrs, CRM_META "_" XML_RSC_ATTR_CLEAR_INTERVAL, &interval_ms); } } uname = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if ((rsc == NULL) || (uname == NULL)) { crm_log_xml_warn(stored_msg, "invalid failcount op"); return I_NULL; } if (crm_element_value(xml_op, XML_LRM_ATTR_ROUTER_NODE)) { is_remote_node = TRUE; } if (interval_ms) { interval_spec = crm_strdup_printf("%ums", interval_ms); } update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node); free(interval_spec); lrm_clear_last_failure(rsc, uname, op, interval_ms); return I_NULL; } static enum crmd_fsa_input handle_lrm_delete(xmlNode *stored_msg) { const char *mode = NULL; xmlNode *msg_data = get_message_xml(stored_msg, F_CRM_DATA); CRM_CHECK(msg_data != NULL, return I_NULL); /* CRM_OP_LRM_DELETE has two distinct modes. The default behavior is to * relay the operation to the affected node, which will unregister the * resource from the local executor, clear the resource's history from the * CIB, and do some bookkeeping in the controller. * * However, if the affected node is offline, the client will specify * mode="cib" which means the controller receiving the operation should * clear the resource's history from the CIB and nothing else. This is used * to clear shutdown locks. */ mode = crm_element_value(msg_data, PCMK__XA_MODE); if ((mode == NULL) || strcmp(mode, XML_TAG_CIB)) { // Relay to affected node crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else { // Delete CIB history locally (compare with do_lrm_delete()) const char *from_sys = NULL; const char *user_name = NULL; const char *rsc_id = NULL; const char *node = NULL; xmlNode *rsc_xml = NULL; int rc = pcmk_rc_ok; rsc_xml = first_named_child(msg_data, XML_CIB_TAG_RESOURCE); CRM_CHECK(rsc_xml != NULL, return I_NULL); rsc_id = ID(rsc_xml); from_sys = crm_element_value(stored_msg, F_CRM_SYS_FROM); node = crm_element_value(msg_data, XML_LRM_ATTR_TARGET); user_name = pcmk__update_acl_user(stored_msg, F_CRM_USER, NULL); crm_debug("Handling " CRM_OP_LRM_DELETE " for %s on %s locally%s%s " "(clearing CIB resource history only)", rsc_id, node, (user_name? " for user " : ""), (user_name? user_name : "")); rc = controld_delete_resource_history(rsc_id, node, user_name, cib_dryrun|cib_sync_call); if (rc == pcmk_rc_ok) { rc = controld_delete_resource_history(rsc_id, node, user_name, crmd_cib_smart_opt()); } //Notify client and tengine.(Only notify tengine if mode = "cib" and CRM_OP_LRM_DELETE.) if (from_sys) { lrmd_event_data_t *op = NULL; const char *from_host = crm_element_value(stored_msg, F_CRM_HOST_FROM); const char *transition; if (strcmp(from_sys, CRM_SYSTEM_TENGINE)) { transition = crm_element_value(msg_data, XML_ATTR_TRANSITION_KEY); } else { transition = crm_element_value(stored_msg, XML_ATTR_TRANSITION_KEY); } crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "local node"), rsc_id, ((rc == pcmk_rc_ok)? "" : " not")); op = lrmd_new_event(rsc_id, CRMD_ACTION_DELETE, 0); op->type = lrmd_event_exec_complete; op->user_data = strdup(transition? transition : FAKE_TE_ID); op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); controld_rc2event(op, 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); } return I_NULL; } } /*! * \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_remote_state(const xmlNode *msg) { const char *remote_uname = ID(msg); crm_node_t *remote_peer; bool remote_is_up = false; int rc = pcmk_rc_ok; rc = pcmk__xe_get_bool_attr(msg, XML_NODE_IN_CLUSTER, &remote_is_up); CRM_CHECK(remote_uname && rc == pcmk_rc_ok, return I_NULL); remote_peer = crm_remote_peer_get(remote_uname); CRM_CHECK(remote_peer, return I_NULL); pcmk__update_peer_state(__func__, remote_peer, remote_is_up ? CRM_NODE_MEMBER : CRM_NODE_LOST, 0); return I_NULL; } /*! * \brief Handle a CRM_OP_PING message * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_ping(const xmlNode *msg) { const char *value = NULL; xmlNode *ping = NULL; xmlNode *reply = NULL; // Build reply ping = create_xml_node(NULL, XML_CRM_TAG_PING); value = crm_element_value(msg, F_CRM_SYS_TO); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value); // Add controller state value = fsa_state2string(controld_globals.fsa_state); crm_xml_add(ping, XML_PING_ATTR_CRMDSTATE, value); crm_notice("Current ping state: %s", value); // CTS needs this // Add controller health // @TODO maybe do some checks to determine meaningful status crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); // Send reply reply = create_reply(msg, ping); free_xml(ping); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a PCMK__CONTROLD_CMD_NODES message * * \param[in] request Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_list(const xmlNode *request) { GHashTableIter iter; crm_node_t *node = NULL; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Create message data for reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODES); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { xmlNode *xml = create_xml_node(reply_data, XML_CIB_TAG_NODE); crm_xml_add_ll(xml, XML_ATTR_ID, (long long) node->id); // uint32_t crm_xml_add(xml, XML_ATTR_UNAME, node->uname); crm_xml_add(xml, XML_NODE_IN_CLUSTER, node->state); } // Create and send reply reply = create_reply(request, reply_data); free_xml(reply_data); if (reply) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a CRM_OP_NODE_INFO request * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_info_request(const xmlNode *msg) { const char *value = NULL; crm_node_t *node = NULL; int node_id = 0; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Build reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(reply_data, XML_PING_ATTR_SYSFROM, CRM_SYSTEM_CRMD); // Add whether current partition has quorum pcmk__xe_set_bool_attr(reply_data, XML_ATTR_HAVE_QUORUM, pcmk_is_set(controld_globals.flags, controld_has_quorum)); // Check whether client requested node info by ID and/or name crm_element_value_int(msg, XML_ATTR_ID, &node_id); if (node_id < 0) { node_id = 0; } value = crm_element_value(msg, XML_ATTR_UNAME); // Default to local node if none given if ((node_id == 0) && (value == NULL)) { value = controld_globals.our_nodename; } node = pcmk__search_node_caches(node_id, value, CRM_GET_PEER_ANY); if (node) { - crm_xml_add_int(reply_data, XML_ATTR_ID, node->id); - crm_xml_add(reply_data, XML_ATTR_UUID, node->uuid); + crm_xml_add(reply_data, XML_ATTR_ID, node->uuid); crm_xml_add(reply_data, XML_ATTR_UNAME, node->uname); crm_xml_add(reply_data, XML_NODE_IS_PEER, node->state); pcmk__xe_set_bool_attr(reply_data, XML_NODE_IS_REMOTE, pcmk_is_set(node->flags, crm_remote_node)); } // Send reply reply = create_reply(msg, reply_data); free_xml(reply_data); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } static void verify_feature_set(xmlNode *msg) { const char *dc_version = crm_element_value(msg, XML_ATTR_CRM_VERSION); if (dc_version == NULL) { /* All we really know is that the DC feature set is older than 3.1.0, * but that's also all that really matters. */ dc_version = "3.0.14"; } if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) { crm_trace("Local feature set (%s) is compatible with DC's (%s)", CRM_FEATURE_SET, dc_version); } else { crm_err("Local feature set (%s) is incompatible with DC's (%s)", CRM_FEATURE_SET, dc_version); // Nothing is likely to improve without administrator involvement controld_set_fsa_input_flags(R_STAYDOWN); crmd_exit(CRM_EX_FATAL); } } // DC gets own shutdown all-clear static enum crmd_fsa_input handle_shutdown_self_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { // The expected case -- we initiated own shutdown sequence crm_info("Shutting down controller"); return I_STOP; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_casei)) { // Must be logic error -- DC confirming its own unrequested shutdown crm_err("Shutting down controller immediately due to " "unexpected shutdown confirmation"); return I_TERMINATE; } if (controld_globals.fsa_state != S_STOPPING) { // Shouldn't happen -- non-DC confirming unrequested shutdown crm_err("Starting new DC election because %s is " "confirming shutdown we did not request", (host_from? host_from : "another node")); return I_ELECTION; } // Shouldn't happen, but we are already stopping anyway crm_debug("Ignoring unexpected shutdown confirmation from %s", (host_from? host_from : "another node")); return I_NULL; } // Non-DC gets shutdown all-clear from DC static enum crmd_fsa_input handle_shutdown_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { crm_warn("Ignoring shutdown request without origin specified"); return I_NULL; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_null_matches|pcmk__str_casei)) { if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { crm_info("Shutting down controller after confirmation from %s", host_from); } else { crm_err("Shutting down controller after unexpected " "shutdown request from %s", host_from); controld_set_fsa_input_flags(R_STAYDOWN); } return I_STOP; } crm_warn("Ignoring shutdown request from %s because DC is %s", host_from, controld_globals.dc_name); return I_NULL; } static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause) { xmlNode *msg = NULL; const char *op = crm_element_value(stored_msg, F_CRM_TASK); /* Optimize this for the DC - it has the most to do */ if (op == NULL) { crm_log_xml_warn(stored_msg, "[request without " F_CRM_TASK "]"); return I_NULL; } if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { const char *from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_node_t *node = pcmk__search_cluster_node_cache(0, from); pcmk__update_peer_expected(__func__, node, CRMD_JOINSTATE_DOWN); if(AM_I_DC == FALSE) { return I_NULL; /* Done */ } } /*========== DC-Only Actions ==========*/ if (AM_I_DC) { if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) { return I_NODE_JOIN; } else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) { return I_JOIN_REQUEST; } else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) { return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_self_ack(stored_msg); } else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { // Another controller wants to shut down its node return handle_shutdown_request(stored_msg); } else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) { /* a remote connection host is letting us know the node state */ return handle_remote_state(stored_msg); } } /*========== common actions ==========*/ if (strcmp(op, CRM_OP_NOVOTE) == 0) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); } else if (strcmp(op, CRM_OP_THROTTLE) == 0) { throttle_update(stored_msg); if (AM_I_DC && (controld_globals.transition_graph != NULL) && !controld_globals.transition_graph->complete) { crm_debug("The throttle changed. Trigger a graph."); trigger_graph(); } return I_NULL; } else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) { return handle_failcount_op(stored_msg); } else if (strcmp(op, CRM_OP_VOTE) == 0) { /* count the vote and decide what to do after that */ ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); /* Sometimes we _must_ go into S_ELECTION */ if (controld_globals.fsa_state == S_HALT) { crm_debug("Forcing an election from S_HALT"); return I_ELECTION; #if 0 } else if (AM_I_DC) { /* This is the old way of doing things but what is gained? */ return I_ELECTION; #endif } } else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) { verify_feature_set(stored_msg); crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_OFFER; } else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) { crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_LRM_DELETE) == 0) { return handle_lrm_delete(stored_msg); } else if ((strcmp(op, CRM_OP_LRM_FAIL) == 0) || (strcmp(op, CRM_OP_LRM_REFRESH) == 0) // @COMPAT || (strcmp(op, CRM_OP_REPROBE) == 0)) { crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else if (strcmp(op, CRM_OP_NOOP) == 0) { return I_NULL; } else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) { crm_shutdown(SIGTERM); /*return I_SHUTDOWN; */ return I_NULL; } else if (strcmp(op, CRM_OP_PING) == 0) { return handle_ping(stored_msg); } else if (strcmp(op, CRM_OP_NODE_INFO) == 0) { return handle_node_info_request(stored_msg); } else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) { int id = 0; const char *name = NULL; crm_element_value_int(stored_msg, XML_ATTR_ID, &id); name = crm_element_value(stored_msg, XML_ATTR_UNAME); if(cause == C_IPC_MESSAGE) { msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) { crm_err("Could not instruct peers to remove references to node %s/%u", name, id); } else { crm_notice("Instructing peers to remove references to node %s/%u", name, id); } free_xml(msg); } else { reap_crm_member(id, name); /* If we're forgetting this node, also forget any failures to fence * it, so we don't carry that over to any node added later with the * same name. */ st_fail_count_reset(name); } } else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) { xmlNode *xml = get_message_xml(stored_msg, F_CRM_DATA); remote_ra_process_maintenance_nodes(xml); } else if (strcmp(op, PCMK__CONTROLD_CMD_NODES) == 0) { return handle_node_list(stored_msg); /*========== (NOT_DC)-Only Actions ==========*/ } else if (!AM_I_DC) { if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_ack(stored_msg); } } else { crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node"); crm_log_xml_err(stored_msg, "Unexpected"); } return I_NULL; } static void handle_response(xmlNode *stored_msg) { const char *op = crm_element_value(stored_msg, F_CRM_TASK); if (op == NULL) { crm_log_xml_err(stored_msg, "Bad message"); } else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) { // Check whether scheduler answer been superseded by subsequent request const char *msg_ref = crm_element_value(stored_msg, XML_ATTR_REFERENCE); if (msg_ref == NULL) { crm_err("%s - Ignoring calculation with no reference", op); } else if (pcmk__str_eq(msg_ref, controld_globals.fsa_pe_ref, pcmk__str_none)) { ha_msg_input_t fsa_input; controld_stop_sched_timer(); fsa_input.msg = stored_msg; register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input); } else { crm_info("%s calculation %s is obsolete", op, msg_ref); } } else if (strcmp(op, CRM_OP_VOTE) == 0 || strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) { } else { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_err("Unexpected response (op=%s, src=%s) sent to the %s", op, host_from, AM_I_DC ? "DC" : "controller"); } } static enum crmd_fsa_input handle_shutdown_request(xmlNode * stored_msg) { /* handle here to avoid potential version issues * where the shutdown message/procedure may have * been changed in later versions. * * This way the DC is always in control of the shutdown */ char *now_s = NULL; const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { /* we're shutting down and the DC */ host_from = controld_globals.our_nodename; } crm_info("Creating shutdown request for %s (state=%s)", host_from, fsa_state2string(controld_globals.fsa_state)); crm_log_xml_trace(stored_msg, "message"); now_s = pcmk__ttoa(time(NULL)); update_attrd(host_from, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, FALSE); free(now_s); /* will be picked up by the TE as long as its running */ return I_NULL; } static void send_msg_via_ipc(xmlNode * msg, const char *sys) { pcmk__client_t *client_channel = NULL; CRM_CHECK(sys != NULL, return); client_channel = pcmk__find_client_by_id(sys); if (crm_element_value(msg, F_CRM_HOST_FROM) == NULL) { crm_xml_add(msg, F_CRM_HOST_FROM, controld_globals.our_nodename); } if (client_channel != NULL) { /* Transient clients such as crmadmin */ pcmk__ipc_send_xml(client_channel, 0, msg, crm_ipc_server_event); } else if (pcmk__str_eq(sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { xmlNode *data = get_message_xml(msg, F_CRM_DATA); process_te_message(msg, data); } else if (pcmk__str_eq(sys, CRM_SYSTEM_LRMD, pcmk__str_none)) { fsa_data_t fsa_data; ha_msg_input_t fsa_input; fsa_input.msg = msg; fsa_input.xml = get_message_xml(msg, F_CRM_DATA); fsa_data.id = 0; fsa_data.actions = 0; fsa_data.data = &fsa_input; fsa_data.fsa_input = I_MESSAGE; fsa_data.fsa_cause = C_IPC_MESSAGE; fsa_data.origin = __func__; fsa_data.data_type = fsa_dt_ha_msg; do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, controld_globals.fsa_state, I_MESSAGE, &fsa_data); } else if (crmd_is_proxy_session(sys)) { crmd_proxy_send(sys, msg); } else { crm_info("Received invalid request: unknown subsystem '%s'", sys); } } void delete_ha_msg_input(ha_msg_input_t * orig) { if (orig == NULL) { return; } free_xml(orig->msg); free(orig); } /*! * \internal * \brief Notify the DC of a remote node state change * * \param[in] node_name Node's name * \param[in] node_up TRUE if node is up, FALSE if down */ void send_remote_state_message(const char *node_name, gboolean node_up) { /* If we don't have a DC, or the message fails, we have a failsafe: * the DC will eventually pick up the change via the CIB node state. * The message allows it to happen sooner if possible. */ if (controld_globals.dc_name != NULL) { xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, controld_globals.dc_name, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); crm_info("Notifying DC %s of Pacemaker Remote node %s %s", controld_globals.dc_name, node_name, node_up? "coming up" : "going down"); crm_xml_add(msg, XML_ATTR_ID, node_name); pcmk__xe_set_bool_attr(msg, XML_NODE_IN_CLUSTER, node_up); send_cluster_message(crm_get_peer(0, controld_globals.dc_name), crm_msg_crmd, msg, TRUE); free_xml(msg); } else { crm_debug("No DC to notify of Pacemaker Remote node %s %s", node_name, (node_up? "coming up" : "going down")); } } diff --git a/daemons/controld/controld_te_actions.c b/daemons/controld/controld_te_actions.c index 214a99f96d..375f2d2f49 100644 --- a/daemons/controld/controld_te_actions.c +++ b/daemons/controld/controld_te_actions.c @@ -1,748 +1,748 @@ /* * Copyright 2004-2022 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 static GHashTable *te_targets = NULL; void send_rsc_command(pcmk__graph_action_t *action); static void te_update_job_count(pcmk__graph_action_t *action, int offset); static void te_start_action_timer(const pcmk__graph_t *graph, pcmk__graph_action_t *action) { action->timer = g_timeout_add(action->timeout + graph->network_delay, action_timer_callback, (void *) action); CRM_ASSERT(action->timer != 0); } /*! * \internal * \brief Execute a graph pseudo-action * * \param[in,out] graph Transition graph being executed * \param[in,out] pseudo Pseudo-action to execute * * \return Standard Pacemaker return code */ static int execute_pseudo_action(pcmk__graph_t *graph, pcmk__graph_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(controld_globals.our_nodename, 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 pcmk_rc_ok; } static int get_target_rc(pcmk__graph_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; } /*! * \internal * \brief Execute a cluster action from a transition graph * * \param[in,out] graph Transition graph being executed * \param[in,out] action Cluster action to execute * * \return Standard Pacemaker return code */ static int execute_cluster_action(pcmk__graph_t *graph, pcmk__graph_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); CRM_CHECK(!pcmk__str_empty(id), return EPROTO); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); CRM_CHECK(!pcmk__str_empty(task), return EPROTO); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(!pcmk__str_empty(on_node), return pcmk_rc_node_unknown); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (router_node == NULL) { router_node = on_node; if (pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_none)) { const char *mode = crm_element_value(action->xml, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_none)) { router_node = controld_globals.our_nodename; } } } if (pcmk__str_eq(router_node, controld_globals.our_nodename, 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("Handling controller request '%s' (%s on %s)%s%s", id, task, on_node, (is_local? " locally" : ""), (no_wait? " without waiting" : "")); if (is_local && pcmk__str_eq(task, CRM_OP_SHUTDOWN, pcmk__str_none)) { /* defer until everything else completes */ crm_info("Controller request '%s' is a local shutdown", id); graph->completion_action = pcmk__graph_shutdown; graph->abort_reason = "local shutdown"; te_action_confirmed(action, graph); return pcmk_rc_ok; } else if (pcmk__str_eq(task, CRM_OP_SHUTDOWN, pcmk__str_none)) { 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(controld_globals.transition_graph->id, action->id, get_target_rc(action), controld_globals.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 ECOMM; } 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 pcmk_rc_ok; } /*! * \internal * \brief Synthesize an executor event for a resource action timeout * * \param[in] action Resource action that timed out * \param[in] target_rc Expected result of action that timed out * * Synthesize an executor event for a resource action timeout. (If the executor * gets a timeout while waiting for a resource action to complete, that will be * reported via the usual callback. This timeout means we didn't hear from the * executor itself or the controller that relayed the action to the executor.) * * \return Newly created executor event for result of \p action * \note The caller is responsible for freeing the return value using * lrmd_free_event(). */ static lrmd_event_data_t * synthesize_timeout_event(const pcmk__graph_action_t *action, int target_rc) { lrmd_event_data_t *op = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *reason = NULL; char *dynamic_reason = NULL; if (pcmk__str_eq(target, get_local_node_name(), pcmk__str_casei)) { reason = "Local executor did not return result in time"; } else { const char *router_node = NULL; router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (router_node == NULL) { router_node = target; } dynamic_reason = crm_strdup_printf("Controller on %s did not return " "result in time", router_node); reason = dynamic_reason; } op = pcmk__event_from_graph_action(NULL, action, PCMK_EXEC_TIMEOUT, PCMK_OCF_UNKNOWN_ERROR, reason); op->call_id = -1; op->user_data = pcmk__transition_key(controld_globals.transition_graph->id, action->id, target_rc, controld_globals.te_uuid); free(dynamic_reason); return op; } static void controld_record_action_event(pcmk__graph_action_t *action, lrmd_event_data_t *op) { cib_t *cib_conn = controld_globals.cib_conn; xmlNode *state = NULL; xmlNode *rsc = 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); 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_ID, 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); pcmk__create_history_xml(rsc, op, CRM_FEATURE_SET, target_rc, target, __func__); rc = cib_conn->cmds->update(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 synthesized event of action %d (%s on %s)", rc, action->id, task_uuid, target); pcmk__set_graph_action_flags(action, pcmk__graph_action_sent_update); } void controld_record_action_timeout(pcmk__graph_action_t *action) { lrmd_event_data_t *op = 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); 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); op = synthesize_timeout_event(action, target_rc); controld_record_action_event(action, op); lrmd_free_event(op); } /*! * \internal * \brief Execute a resource action from a transition graph * * \param[in,out] graph Transition graph being executed * \param[in,out] action Resource action to execute * * \return Standard Pacemaker return code */ static int execute_rsc_action(pcmk__graph_t *graph, pcmk__graph_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); pcmk__clear_graph_action_flags(action, pcmk__graph_action_executed); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(!pcmk__str_empty(on_node), crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), pcmk__s(task, "without task")); return pcmk_rc_node_unknown); 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(controld_globals.transition_graph->id, action->id, get_target_rc(action), controld_globals.te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (pcmk__str_eq(router_node, controld_globals.our_nodename, 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, controld_globals.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); pcmk__set_graph_action_flags(action, pcmk__graph_action_executed); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return ECOMM; } else if (no_wait) { /* Just mark confirmed. Don't bump the job count only to immediately * decrement it. */ crm_info("Action %d confirmed - no wait", action->id); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(controld_globals.transition_graph, action); trigger_graph(); } else if (pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { 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 pcmk_rc_ok; } 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(pcmk__graph_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 != pcmk__rsc_graph_action) || (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); } /*! * \internal * \brief Check whether a graph action is allowed to be executed on a node * * \param[in] graph Transition graph being executed * \param[in] action Graph action being executed * \param[in] target Name of node where action should be executed * * \return true if action is allowed, otherwise false */ static bool allowed_on_node(const pcmk__graph_t *graph, const pcmk__graph_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; } /*! * \internal * \brief Check whether a graph action is allowed to be executed * * \param[in] graph Transition graph being executed * \param[in] action Graph action being executed * * \return true if action is allowed, otherwise false */ static bool graph_action_allowed(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *target = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != pcmk__rsc_graph_action) { /* 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 (!allowed_on_node(graph, action, target)) { 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 allowed_on_node(graph, action, target); } /*! * \brief Confirm a graph action (and optionally update graph) * * \param[in,out] action Action to confirm * \param[in,out] graph Update and trigger this graph (if non-NULL) */ void te_action_confirmed(pcmk__graph_action_t *action, pcmk__graph_t *graph) { if (!pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { if ((action->type == pcmk__rsc_graph_action) && (crm_element_value(action->xml, XML_LRM_ATTR_TARGET) != NULL)) { te_update_job_count(action, -1); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); } if (graph) { pcmk__update_graph(graph, action); trigger_graph(); } } static pcmk__graph_functions_t te_graph_fns = { execute_pseudo_action, execute_rsc_action, execute_cluster_action, controld_execute_fence_action, graph_action_allowed, }; /* * \internal * \brief Register the transitioner's graph functions with \p libpacemaker */ void controld_register_graph_functions(void) { pcmk__set_graph_functions(&te_graph_fns); } void notify_crmd(pcmk__graph_t *graph) { const char *type = "unknown"; enum crmd_fsa_input event = I_NULL; crm_debug("Processing transition completion in state %s", fsa_state2string(controld_globals.fsa_state)); CRM_CHECK(graph->complete, graph->complete = true); switch (graph->completion_action) { case pcmk__graph_wait: type = "stop"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case pcmk__graph_done: type = "done"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case pcmk__graph_restart: type = "restart"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { if (controld_get_period_transition_timer() > 0) { controld_stop_transition_timer(); controld_start_transition_timer(); } else { event = I_PE_CALC; } } else if (controld_globals.fsa_state == S_POLICY_ENGINE) { controld_set_fsa_action_flags(A_PE_INVOKE); controld_trigger_fsa(); } break; case pcmk__graph_shutdown: type = "shutdown"; if (pcmk_is_set(controld_globals.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, pcmk__s(graph->abort_reason, "unspecified reason")); graph->abort_reason = NULL; graph->completion_action = pcmk__graph_done; controld_clear_fsa_input_flags(R_IN_TRANSITION); if (event != I_NULL) { register_fsa_input(C_FSA_INTERNAL, event, NULL); } else { controld_trigger_fsa(); } } diff --git a/daemons/controld/controld_te_events.c b/daemons/controld/controld_te_events.c index 1e6765840a..1dfa4ceaf2 100644 --- a/daemons/controld/controld_te_events.c +++ b/daemons/controld/controld_te_events.c @@ -1,589 +1,589 @@ /* * Copyright 2004-2022 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 /*! * \internal * \brief Action numbers of outside events processed in current update diff * * This table is to be used as a set. It should be empty when the transitioner * begins processing a CIB update diff. It ensures that if there are multiple * events (for example, "_last_0" and "_last_failure_0") for the same action, * only one of them updates the failcount. Events that originate outside the * cluster can't be confirmed, since they're not in the transition graph. */ static GHashTable *outside_events = NULL; /*! * \internal * \brief Destroy the hash table containing action numbers of outside events */ void controld_destroy_outside_event_table(void) { if (outside_events != NULL) { g_hash_table_destroy(outside_events); outside_events = NULL; } } /*! * \internal * \brief Add an outside event's action number to a set * * \return Standard Pacemaker return code. Specifically, \p pcmk_rc_ok if the * event was not already in the set, or \p pcmk_rc_already otherwise. */ static int record_outside_event(gint action_num) { if (outside_events == NULL) { outside_events = g_hash_table_new(NULL, NULL); } if (g_hash_table_add(outside_events, GINT_TO_POINTER(action_num))) { return pcmk_rc_ok; } return pcmk_rc_already; } gboolean fail_incompletable_actions(pcmk__graph_t *graph, const char *down_node) { const char *target_uuid = NULL; const char *router = NULL; const char *router_uuid = NULL; xmlNode *last_action = NULL; GList *gIter = NULL; GList *gIter2 = NULL; if (graph == NULL || graph->complete) { return FALSE; } gIter = graph->synapses; for (; gIter != NULL; gIter = gIter->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) gIter->data; if (pcmk_any_flags_set(synapse->flags, pcmk__synapse_confirmed|pcmk__synapse_failed)) { /* We've already been here */ continue; } gIter2 = synapse->actions; for (; gIter2 != NULL; gIter2 = gIter2->next) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) gIter2->data; if ((action->type == pcmk__pseudo_graph_action) || pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { continue; } else if (action->type == pcmk__cluster_graph_action) { const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (pcmk__str_eq(task, CRM_OP_FENCE, pcmk__str_casei)) { continue; } } target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); router = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (router) { crm_node_t *node = crm_get_peer(0, router); if (node) { router_uuid = node->uuid; } } if (pcmk__str_eq(target_uuid, down_node, pcmk__str_casei) || pcmk__str_eq(router_uuid, down_node, pcmk__str_casei)) { pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); pcmk__set_synapse_flags(synapse, pcmk__synapse_failed); last_action = action->xml; stop_te_timer(action); pcmk__update_graph(graph, action); if (pcmk_is_set(synapse->flags, pcmk__synapse_executed)) { crm_notice("Action %d (%s) was pending on %s (offline)", action->id, crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY), down_node); } else { crm_info("Action %d (%s) is scheduled for %s (offline)", action->id, crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY), down_node); } } } } if (last_action != NULL) { crm_info("Node %s shutdown resulted in un-runnable actions", down_node); abort_transition(INFINITY, pcmk__graph_restart, "Node failure", last_action); return TRUE; } return FALSE; } /*! * \internal * \brief Update failure-related node attributes if warranted * * \param[in] event XML describing operation that (maybe) failed * \param[in] event_node_uuid Node that event occurred on * \param[in] rc Actual operation return code * \param[in] target_rc Expected operation return code * \param[in] do_update If TRUE, do update regardless of operation type * \param[in] ignore_failures If TRUE, update last failure but not fail count * * \return TRUE if this was not a direct nack, success or lrm status refresh */ static gboolean update_failcount(const xmlNode *event, const char *event_node_uuid, int rc, int target_rc, gboolean do_update, gboolean ignore_failures) { guint interval_ms = 0; char *task = NULL; char *rsc_id = NULL; const char *value = NULL; const char *id = crm_element_value(event, XML_LRM_ATTR_TASK_KEY); const char *on_uname = crm_peer_uname(event_node_uuid); const char *origin = crm_element_value(event, XML_ATTR_ORIGIN); // Nothing needs to be done for success or status refresh if (rc == target_rc) { return FALSE; } else if (pcmk__str_eq(origin, "build_active_RAs", pcmk__str_casei)) { crm_debug("No update for %s (rc=%d) on %s: Old failure from lrm status refresh", id, rc, on_uname); return FALSE; } /* Sanity check */ CRM_CHECK(on_uname != NULL, return TRUE); CRM_CHECK(parse_op_key(id, &rsc_id, &task, &interval_ms), crm_err("Couldn't parse: %s", ID(event)); goto bail); /* Decide whether update is necessary and what value to use */ if ((interval_ms > 0) || pcmk__str_eq(task, CRMD_ACTION_PROMOTE, pcmk__str_none) || pcmk__str_eq(task, CRMD_ACTION_DEMOTE, pcmk__str_none)) { do_update = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_START, pcmk__str_none)) { do_update = TRUE; value = pcmk__s(controld_globals.transition_graph->failed_start_offset, CRM_INFINITY_S); } else if (pcmk__str_eq(task, CRMD_ACTION_STOP, pcmk__str_none)) { do_update = TRUE; value = pcmk__s(controld_globals.transition_graph->failed_stop_offset, CRM_INFINITY_S); } if (do_update) { pcmk__attrd_query_pair_t *fail_pair = NULL; pcmk__attrd_query_pair_t *last_pair = NULL; char *fail_name = NULL; char *last_name = NULL; GList *attrs = NULL; uint32_t opts = pcmk__node_attr_none; char *now = pcmk__ttoa(time(NULL)); // Fail count will be either incremented or set to infinity if (!pcmk_str_is_infinity(value)) { value = XML_NVPAIR_ATTR_VALUE "++"; } if (g_hash_table_lookup(crm_remote_peer_cache, event_node_uuid)) { opts |= pcmk__node_attr_remote; } crm_info("Updating %s for %s on %s after failed %s: rc=%d (update=%s, time=%s)", (ignore_failures? "last failure" : "failcount"), rsc_id, on_uname, task, rc, value, now); /* Update the fail count, if we're not ignoring failures */ if (!ignore_failures) { fail_pair = calloc(1, sizeof(pcmk__attrd_query_pair_t)); CRM_ASSERT(fail_pair != NULL); fail_name = pcmk__failcount_name(rsc_id, task, interval_ms); fail_pair->name = fail_name; fail_pair->value = value; fail_pair->node = on_uname; attrs = g_list_prepend(attrs, fail_pair); } /* Update the last failure time (even if we're ignoring failures, * so that failure can still be detected and shown, e.g. by crm_mon) */ last_pair = calloc(1, sizeof(pcmk__attrd_query_pair_t)); CRM_ASSERT(last_pair != NULL); last_name = pcmk__lastfailure_name(rsc_id, task, interval_ms); last_pair->name = last_name; last_pair->value = now; last_pair->node = on_uname; attrs = g_list_prepend(attrs, last_pair); update_attrd_list(attrs, opts); free(fail_name); free(fail_pair); free(last_name); free(last_pair); g_list_free(attrs); free(now); } bail: free(rsc_id); free(task); return TRUE; } pcmk__graph_action_t * controld_get_action(int id) { for (GList *item = controld_globals.transition_graph->synapses; item != NULL; item = item->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) item->data; for (GList *item2 = synapse->actions; item2; item2 = item2->next) { pcmk__graph_action_t *action = (pcmk__graph_action_t *) item2->data; if (action->id == id) { return action; } } } return NULL; } pcmk__graph_action_t * get_cancel_action(const char *id, const char *node) { GList *gIter = NULL; GList *gIter2 = NULL; gIter = controld_globals.transition_graph->synapses; for (; gIter != NULL; gIter = gIter->next) { pcmk__graph_synapse_t *synapse = (pcmk__graph_synapse_t *) gIter->data; gIter2 = synapse->actions; for (; gIter2 != NULL; gIter2 = gIter2->next) { const char *task = NULL; const char *target = NULL; pcmk__graph_action_t *action = (pcmk__graph_action_t *) gIter2->data; task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (!pcmk__str_eq(CRMD_ACTION_CANCEL, task, pcmk__str_casei)) { continue; } task = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); if (!pcmk__str_eq(task, id, pcmk__str_casei)) { crm_trace("Wrong key %s for %s on %s", task, id, node); continue; } target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); if (node && !pcmk__str_eq(target, node, pcmk__str_casei)) { crm_trace("Wrong node %s for %s on %s", target, id, node); continue; } crm_trace("Found %s on %s", id, node); return action; } } return NULL; } bool confirm_cancel_action(const char *id, const char *node_id) { const char *op_key = NULL; const char *node_name = NULL; pcmk__graph_action_t *cancel = get_cancel_action(id, node_id); if (cancel == NULL) { return FALSE; } op_key = crm_element_value(cancel->xml, XML_LRM_ATTR_TASK_KEY); node_name = crm_element_value(cancel->xml, XML_LRM_ATTR_TARGET); stop_te_timer(cancel); te_action_confirmed(cancel, controld_globals.transition_graph); crm_info("Cancellation of %s on %s confirmed (action %d)", op_key, node_name, cancel->id); return TRUE; } /* downed nodes are listed like: ... */ #define XPATH_DOWNED "//" XML_GRAPH_TAG_DOWNED \ - "/" XML_CIB_TAG_NODE "[@" XML_ATTR_UUID "='%s']" + "/" XML_CIB_TAG_NODE "[@" XML_ATTR_ID "='%s']" /*! * \brief Find a transition event that would have made a specified node down * * \param[in] target UUID of node to match * * \return Matching event if found, NULL otherwise */ pcmk__graph_action_t * match_down_event(const char *target) { pcmk__graph_action_t *match = NULL; xmlXPathObjectPtr xpath_ret = NULL; GList *gIter, *gIter2; char *xpath = crm_strdup_printf(XPATH_DOWNED, target); for (gIter = controld_globals.transition_graph->synapses; gIter != NULL && match == NULL; gIter = gIter->next) { for (gIter2 = ((pcmk__graph_synapse_t * ) gIter->data)->actions; gIter2 != NULL && match == NULL; gIter2 = gIter2->next) { match = (pcmk__graph_action_t *) gIter2->data; if (pcmk_is_set(match->flags, pcmk__graph_action_executed)) { xpath_ret = xpath_search(match->xml, xpath); if (numXpathResults(xpath_ret) < 1) { match = NULL; } freeXpathObject(xpath_ret); } else { // Only actions that were actually started can match match = NULL; } } } free(xpath); if (match != NULL) { crm_debug("Shutdown action %d (%s) found for node %s", match->id, crm_element_value(match->xml, XML_LRM_ATTR_TASK_KEY), target); } else { crm_debug("No reason to expect node %s to be down", target); } return match; } void process_graph_event(xmlNode *event, const char *event_node) { int rc = -1; // Actual result int target_rc = -1; // Expected result int status = -1; // Executor status int callid = -1; // Executor call ID int transition_num = -1; // Transition number int action_num = -1; // Action number within transition char *update_te_uuid = NULL; bool ignore_failures = FALSE; const char *id = NULL; const char *desc = NULL; const char *magic = NULL; const char *uname = NULL; CRM_ASSERT(event != NULL); /* */ magic = crm_element_value(event, XML_ATTR_TRANSITION_KEY); if (magic == NULL) { /* non-change */ return; } crm_element_value_int(event, XML_LRM_ATTR_OPSTATUS, &status); if (status == PCMK_EXEC_PENDING) { return; } id = crm_element_value(event, XML_LRM_ATTR_TASK_KEY); crm_element_value_int(event, XML_LRM_ATTR_RC, &rc); crm_element_value_int(event, XML_LRM_ATTR_CALLID, &callid); rc = pcmk__effective_rc(rc); if (decode_transition_key(magic, &update_te_uuid, &transition_num, &action_num, &target_rc) == FALSE) { // decode_transition_key() already logged the bad key crm_err("Can't process action %s result: Incompatible versions? " CRM_XS " call-id=%d", id, callid); abort_transition(INFINITY, pcmk__graph_restart, "Bad event", event); return; } if (transition_num == -1) { // E.g. crm_resource --fail if (record_outside_event(action_num) != pcmk_rc_ok) { crm_debug("Outside event with transition key '%s' has already been " "processed", magic); goto bail; } desc = "initiated outside of the cluster"; abort_transition(INFINITY, pcmk__graph_restart, "Unexpected event", event); } else if ((action_num < 0) || !pcmk__str_eq(update_te_uuid, controld_globals.te_uuid, pcmk__str_none)) { desc = "initiated by a different DC"; abort_transition(INFINITY, pcmk__graph_restart, "Foreign event", event); } else if ((controld_globals.transition_graph->id != transition_num) || controld_globals.transition_graph->complete) { // Action is not from currently active transition guint interval_ms = 0; if (parse_op_key(id, NULL, NULL, &interval_ms) && (interval_ms != 0)) { /* Recurring actions have the transition number they were first * scheduled in. */ if (status == PCMK_EXEC_CANCELLED) { confirm_cancel_action(id, get_node_id(event)); goto bail; } desc = "arrived after initial scheduling"; abort_transition(INFINITY, pcmk__graph_restart, "Change in recurring result", event); } else if (controld_globals.transition_graph->id != transition_num) { desc = "arrived really late"; abort_transition(INFINITY, pcmk__graph_restart, "Old event", event); } else { desc = "arrived late"; abort_transition(INFINITY, pcmk__graph_restart, "Inactive graph", event); } } else { // Event is result of an action from currently active transition pcmk__graph_action_t *action = controld_get_action(action_num); if (action == NULL) { // Should never happen desc = "unknown"; abort_transition(INFINITY, pcmk__graph_restart, "Unknown event", event); } else if (pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { /* Nothing further needs to be done if the action has already been * confirmed. This can happen e.g. when processing both an * "xxx_last_0" or "xxx_last_failure_0" record as well as the main * history record, which would otherwise result in incorrectly * bumping the fail count twice. */ crm_log_xml_debug(event, "Event already confirmed:"); goto bail; } else { /* An action result needs to be confirmed. * (This is the only case where desc == NULL.) */ if (pcmk__str_eq(crm_meta_value(action->params, XML_OP_ATTR_ON_FAIL), "ignore", pcmk__str_casei)) { ignore_failures = TRUE; } else if (rc != target_rc) { pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); } stop_te_timer(action); te_action_confirmed(action, controld_globals.transition_graph); if (pcmk_is_set(action->flags, pcmk__graph_action_failed)) { abort_transition(action->synapse->priority + 1, pcmk__graph_restart, "Event failed", event); } } } if (id == NULL) { id = "unknown action"; } uname = crm_element_value(event, XML_LRM_ATTR_TARGET); if (uname == NULL) { uname = "unknown node"; } if (status == PCMK_EXEC_INVALID) { // We couldn't attempt the action crm_info("Transition %d action %d (%s on %s): %s", transition_num, action_num, id, uname, pcmk_exec_status_str(status)); } else if (desc && update_failcount(event, event_node, rc, target_rc, (transition_num == -1), FALSE)) { crm_notice("Transition %d action %d (%s on %s): expected '%s' but got '%s' " CRM_XS " target-rc=%d rc=%d call-id=%d event='%s'", transition_num, action_num, id, uname, services_ocf_exitcode_str(target_rc), services_ocf_exitcode_str(rc), target_rc, rc, callid, desc); } else if (desc) { crm_info("Transition %d action %d (%s on %s): %s " CRM_XS " rc=%d target-rc=%d call-id=%d", transition_num, action_num, id, uname, desc, rc, target_rc, callid); } else if (rc == target_rc) { crm_info("Transition %d action %d (%s on %s) confirmed: %s " CRM_XS " rc=%d call-id=%d", transition_num, action_num, id, uname, services_ocf_exitcode_str(rc), rc, callid); } else { update_failcount(event, event_node, rc, target_rc, (transition_num == -1), ignore_failures); crm_notice("Transition %d action %d (%s on %s): expected '%s' but got '%s' " CRM_XS " target-rc=%d rc=%d call-id=%d", transition_num, action_num, id, uname, services_ocf_exitcode_str(target_rc), services_ocf_exitcode_str(rc), target_rc, rc, callid); } bail: free(update_te_uuid); } diff --git a/lib/cib/cib_attrs.c b/lib/cib/cib_attrs.c index 953322cfad..672dbf2ac5 100644 --- a/lib/cib/cib_attrs.c +++ b/lib/cib/cib_attrs.c @@ -1,732 +1,732 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static pcmk__output_t * new_output_object(const char *ty) { int rc = pcmk_rc_ok; pcmk__output_t *out = NULL; const char* argv[] = { "", NULL }; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_LOG, PCMK__SUPPORTED_FORMAT_TEXT, { NULL, NULL, NULL } }; pcmk__register_formats(NULL, formats); rc = pcmk__output_new(&out, ty, NULL, (char**)argv); if ((rc != pcmk_rc_ok) || (out == NULL)) { crm_err("Can't out due to internal error: %s", pcmk_rc_str(rc)); return NULL; } return out; } static int find_attr(cib_t *cib, const char *section, const char *node_uuid, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *user_name, xmlNode **result) { int rc = pcmk_rc_ok; const char *xpath_base = NULL; GString *xpath = NULL; xmlNode *xml_search = NULL; const char *set_type = NULL; const char *node_type = NULL; if (attr_set_type) { set_type = attr_set_type; } else { set_type = XML_TAG_ATTR_SETS; } if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { node_uuid = NULL; set_type = XML_CIB_TAG_PROPSET; } else if (pcmk__strcase_any_of(section, XML_CIB_TAG_OPCONFIG, XML_CIB_TAG_RSCCONFIG, NULL)) { node_uuid = NULL; set_type = XML_TAG_META_SETS; } else if (pcmk__str_eq(section, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { node_uuid = NULL; section = XML_CIB_TAG_STATUS; node_type = XML_CIB_TAG_TICKETS; } else if (node_uuid == NULL) { return EINVAL; } xpath_base = pcmk_cib_xpath_for(section); if (xpath_base == NULL) { crm_warn("%s CIB section not known", section); return ENOMSG; } xpath = g_string_sized_new(1024); g_string_append(xpath, xpath_base); if (pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { pcmk__g_strcat(xpath, "//", node_type, NULL); } else if (node_uuid) { const char *node_type = XML_CIB_TAG_NODE; if (pcmk__str_eq(section, XML_CIB_TAG_STATUS, pcmk__str_casei)) { node_type = XML_CIB_TAG_STATE; set_type = XML_TAG_TRANSIENT_NODEATTRS; } pcmk__g_strcat(xpath, "//", node_type, "[@" XML_ATTR_ID "='", node_uuid, "']", NULL); } pcmk__g_strcat(xpath, "//", set_type, NULL); if (set_name) { pcmk__g_strcat(xpath, "[@" XML_ATTR_ID "='", set_name, "']", NULL); } g_string_append(xpath, "//nvpair"); if (attr_id && attr_name) { pcmk__g_strcat(xpath, "[@" XML_ATTR_ID "='", attr_id, "' " "and @" XML_ATTR_NAME "='", attr_name, "']", NULL); } else if (attr_id) { pcmk__g_strcat(xpath, "[@" XML_ATTR_ID "='", attr_id, "']", NULL); } else if (attr_name) { pcmk__g_strcat(xpath, "[@" XML_ATTR_NAME "='", attr_name, "']", NULL); } rc = cib_internal_op(cib, PCMK__CIB_REQUEST_QUERY, NULL, (const char *) xpath->str, NULL, &xml_search, cib_sync_call|cib_scope_local|cib_xpath, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); crm_trace("Query failed for attribute %s (section=%s, node=%s, set=%s, xpath=%s): %s", attr_name, section, pcmk__s(node_uuid, ""), pcmk__s(set_name, ""), (const char *) xpath->str, pcmk_rc_str(rc)); } else { rc = pcmk_rc_ok; crm_log_xml_debug(xml_search, "Match"); } g_string_free(xpath, TRUE); *result = xml_search; return rc; } static int handle_multiples(pcmk__output_t *out, xmlNode *search, const char *attr_name) { if (xml_has_children(search)) { xmlNode *child = NULL; out->info(out, "Multiple attributes match name=%s", attr_name); for (child = pcmk__xml_first_child(search); child != NULL; child = pcmk__xml_next(child)) { out->info(out, " Value: %s \t(id=%s)", crm_element_value(child, XML_NVPAIR_ATTR_VALUE), ID(child)); } return ENOTUNIQ; } else { return pcmk_rc_ok; } } int cib__update_node_attr(pcmk__output_t *out, cib_t *cib, int call_options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, const char *user_name, const char *node_type) { const char *tag = NULL; int rc = pcmk_rc_ok; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; xmlNode *xml_search = NULL; char *local_attr_id = NULL; char *local_set_name = NULL; CRM_CHECK(section != NULL, return EINVAL); CRM_CHECK(attr_value != NULL, return EINVAL); CRM_CHECK(attr_name != NULL || attr_id != NULL, return EINVAL); rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc == pcmk_rc_ok) { if (handle_multiples(out, xml_search, attr_name) == ENOTUNIQ) { free_xml(xml_search); return ENOTUNIQ; } else { pcmk__str_update(&local_attr_id, crm_element_value(xml_search, XML_ATTR_ID)); attr_id = local_attr_id; free_xml(xml_search); goto do_modify; } } else if (rc != ENXIO) { free_xml(xml_search); return rc; /* } else if(attr_id == NULL) { */ /* return EINVAL; */ } else { free_xml(xml_search); crm_trace("%s does not exist, create it", attr_name); if (pcmk__str_eq(section, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { node_uuid = NULL; section = XML_CIB_TAG_STATUS; node_type = XML_CIB_TAG_TICKETS; xml_top = create_xml_node(xml_obj, XML_CIB_TAG_STATUS); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_TICKETS); } else if (pcmk__str_eq(section, XML_CIB_TAG_NODES, pcmk__str_casei)) { if (node_uuid == NULL) { return EINVAL; } if (pcmk__str_eq(node_type, "remote", pcmk__str_casei)) { xml_top = create_xml_node(xml_obj, XML_CIB_TAG_NODES); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_NODE); crm_xml_add(xml_obj, XML_ATTR_TYPE, "remote"); crm_xml_add(xml_obj, XML_ATTR_ID, node_uuid); crm_xml_add(xml_obj, XML_ATTR_UNAME, node_uuid); } else { tag = XML_CIB_TAG_NODE; } } else if (pcmk__str_eq(section, XML_CIB_TAG_STATUS, pcmk__str_casei)) { tag = XML_TAG_TRANSIENT_NODEATTRS; if (node_uuid == NULL) { return EINVAL; } xml_top = create_xml_node(xml_obj, XML_CIB_TAG_STATE); crm_xml_add(xml_top, XML_ATTR_ID, node_uuid); xml_obj = xml_top; } else { tag = section; node_uuid = NULL; } if (set_name == NULL) { if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { local_set_name = strdup(CIB_OPTIONS_FIRST); } else if (pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { local_set_name = crm_strdup_printf("%s-%s", section, XML_CIB_TAG_TICKETS); } else if (node_uuid) { local_set_name = crm_strdup_printf("%s-%s", section, node_uuid); if (set_type) { char *tmp_set_name = local_set_name; local_set_name = crm_strdup_printf("%s-%s", tmp_set_name, set_type); free(tmp_set_name); } } else { local_set_name = crm_strdup_printf("%s-options", section); } set_name = local_set_name; } if (attr_id == NULL) { local_attr_id = crm_strdup_printf("%s-%s", set_name, attr_name); crm_xml_sanitize_id(local_attr_id); attr_id = local_attr_id; } else if (attr_name == NULL) { attr_name = attr_id; } crm_trace("Creating %s/%s", section, tag); if (tag != NULL) { xml_obj = create_xml_node(xml_obj, tag); crm_xml_add(xml_obj, XML_ATTR_ID, node_uuid); if (xml_top == NULL) { xml_top = xml_obj; } } if (node_uuid == NULL && !pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { xml_obj = create_xml_node(xml_obj, XML_CIB_TAG_PROPSET); } else { xml_obj = create_xml_node(xml_obj, XML_TAG_META_SETS); } } else if (set_type) { xml_obj = create_xml_node(xml_obj, set_type); } else { xml_obj = create_xml_node(xml_obj, XML_TAG_ATTR_SETS); } crm_xml_add(xml_obj, XML_ATTR_ID, set_name); if (xml_top == NULL) { xml_top = xml_obj; } } do_modify: xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_trace(xml_top, "update_attr"); rc = cib_internal_op(cib, PCMK__CIB_REQUEST_MODIFY, NULL, section, xml_top, NULL, call_options|cib_quorum_override, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); out->err(out, "Error setting %s=%s (section=%s, set=%s): %s", attr_name, attr_value, section, pcmk__s(set_name, ""), pcmk_rc_str(rc)); crm_log_xml_info(xml_top, "Update"); } else { rc = pcmk_rc_ok; } free(local_set_name); free(local_attr_id); free_xml(xml_top); return rc; } int cib__get_node_attrs(pcmk__output_t *out, cib_t *cib, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *user_name, xmlNode **result) { int rc = pcmk_rc_ok; CRM_ASSERT(result != NULL); CRM_CHECK(section != NULL, return EINVAL); *result = NULL; rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, result); if (rc != pcmk_rc_ok) { crm_trace("Query failed for attribute %s (section=%s node=%s set=%s): %s", pcmk__s(attr_name, "with unspecified name"), section, pcmk__s(set_name, ""), pcmk__s(node_uuid, ""), pcmk_strerror(rc)); } return rc; } int cib__delete_node_attr(pcmk__output_t *out, cib_t *cib, int options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, const char *user_name) { int rc = pcmk_rc_ok; xmlNode *xml_obj = NULL; xmlNode *xml_search = NULL; char *local_attr_id = NULL; CRM_CHECK(section != NULL, return EINVAL); CRM_CHECK(attr_name != NULL || attr_id != NULL, return EINVAL); if (attr_id == NULL) { rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc != pcmk_rc_ok || handle_multiples(out, xml_search, attr_name) == ENOTUNIQ) { free_xml(xml_search); return rc; } else { pcmk__str_update(&local_attr_id, crm_element_value(xml_search, XML_ATTR_ID)); attr_id = local_attr_id; free_xml(xml_search); } } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, attr_value); rc = cib_internal_op(cib, PCMK__CIB_REQUEST_DELETE, NULL, section, xml_obj, NULL, options|cib_quorum_override, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); } else { rc = pcmk_rc_ok; out->info(out, "Deleted %s %s: id=%s%s%s%s%s", section, node_uuid ? "attribute" : "option", local_attr_id, set_name ? " set=" : "", set_name ? set_name : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(local_attr_id); free_xml(xml_obj); return rc; } int find_nvpair_attr_delegate(cib_t *cib, const char *attr, const char *section, const char *node_uuid, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, gboolean to_console, char **value, const char *user_name) { pcmk__output_t *out = NULL; xmlNode *xml_search = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = find_attr(cib, section, node_uuid, attr_set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc == pcmk_rc_ok) { rc = handle_multiples(out, xml_search, attr_name); if (rc == pcmk_rc_ok) { pcmk__str_update(value, crm_element_value(xml_search, attr)); } } out->finish(out, CRM_EX_OK, true, NULL); free_xml(xml_search); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int update_attr_delegate(cib_t *cib, int call_options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, gboolean to_console, const char *user_name, const char *node_type) { pcmk__output_t *out = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__update_node_attr(out, cib, call_options, section, node_uuid, set_type, set_name, attr_id, attr_name, attr_value, user_name, node_type); out->finish(out, CRM_EX_OK, true, NULL); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int read_attr_delegate(cib_t *cib, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, char **attr_value, gboolean to_console, const char *user_name) { pcmk__output_t *out = NULL; xmlNode *result = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__get_node_attrs(out, cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &result); if (rc == pcmk_rc_ok) { if (!xml_has_children(result)) { pcmk__str_update(attr_value, crm_element_value(result, XML_NVPAIR_ATTR_VALUE)); } else { rc = ENOTUNIQ; } } out->finish(out, CRM_EX_OK, true, NULL); free_xml(result); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int delete_attr_delegate(cib_t *cib, int options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, gboolean to_console, const char *user_name) { pcmk__output_t *out = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__delete_node_attr(out, cib, options, section, node_uuid, set_type, set_name, attr_id, attr_name, attr_value, user_name); out->finish(out, CRM_EX_OK, true, NULL); pcmk__output_free(out); return pcmk_rc2legacy(rc); } /*! * \internal * \brief Parse node UUID from search result * * \param[in] result XML search result * \param[out] uuid If non-NULL, where to store parsed UUID * \param[out] is_remote If non-NULL, set TRUE if result is remote node * * \return pcmk_ok if UUID was successfully parsed, -ENXIO otherwise */ static int get_uuid_from_result(xmlNode *result, char **uuid, int *is_remote) { int rc = -ENXIO; const char *tag; const char *parsed_uuid = NULL; int parsed_is_remote = FALSE; if (result == NULL) { return rc; } /* If there are multiple results, the first is sufficient */ tag = (const char *) (result->name); if (pcmk__str_eq(tag, "xpath-query", pcmk__str_casei)) { result = pcmk__xml_first_child(result); CRM_CHECK(result != NULL, return rc); tag = (const char *) (result->name); } if (pcmk__str_eq(tag, XML_CIB_TAG_NODE, pcmk__str_casei)) { /* Result is tag from section */ if (pcmk__str_eq(crm_element_value(result, XML_ATTR_TYPE), "remote", pcmk__str_casei)) { parsed_uuid = crm_element_value(result, XML_ATTR_UNAME); parsed_is_remote = TRUE; } else { parsed_uuid = ID(result); parsed_is_remote = FALSE; } } else if (pcmk__str_eq(tag, XML_CIB_TAG_RESOURCE, pcmk__str_casei)) { /* Result is for ocf:pacemaker:remote resource */ parsed_uuid = ID(result); parsed_is_remote = TRUE; } else if (pcmk__str_eq(tag, XML_CIB_TAG_NVPAIR, pcmk__str_casei)) { /* Result is remote-node parameter of for guest node */ parsed_uuid = crm_element_value(result, XML_NVPAIR_ATTR_VALUE); parsed_is_remote = TRUE; } else if (pcmk__str_eq(tag, XML_CIB_TAG_STATE, pcmk__str_casei)) { /* Result is tag from section */ parsed_uuid = crm_element_value(result, XML_ATTR_UNAME); if (pcmk__xe_attr_is_true(result, XML_NODE_IS_REMOTE)) { parsed_is_remote = TRUE; } } if (parsed_uuid) { if (uuid) { *uuid = strdup(parsed_uuid); } if (is_remote) { *is_remote = parsed_is_remote; } rc = pcmk_ok; } return rc; } /* Search string to find a node by name, as: * - cluster or remote node in nodes section * - remote node in resources section * - guest node in resources section * - orphaned remote node or bundle guest node in status section */ #define XPATH_UPPER_TRANS "ABCDEFGHIJKLMNOPQRSTUVWXYZ" #define XPATH_LOWER_TRANS "abcdefghijklmnopqrstuvwxyz" #define XPATH_NODE \ "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_NODES \ "/" XML_CIB_TAG_NODE "[translate(@" XML_ATTR_UNAME ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "/" XML_CIB_TAG_RESOURCE \ "[@class='ocf'][@provider='pacemaker'][@type='remote'][translate(@id,'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "/" XML_CIB_TAG_RESOURCE "/" XML_TAG_META_SETS "/" XML_CIB_TAG_NVPAIR \ "[@name='" XML_RSC_ATTR_REMOTE_NODE "'][translate(@value,'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS "/" XML_CIB_TAG_STATE \ - "[@" XML_NODE_IS_REMOTE "='true'][translate(@" XML_ATTR_UUID ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" + "[@" XML_NODE_IS_REMOTE "='true'][translate(@" XML_ATTR_ID ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" int query_node_uuid(cib_t * the_cib, const char *uname, char **uuid, int *is_remote_node) { int rc = pcmk_ok; char *xpath_string; xmlNode *xml_search = NULL; char *host_lowercase = NULL; CRM_ASSERT(uname != NULL); host_lowercase = g_ascii_strdown(uname, -1); if (uuid) { *uuid = NULL; } if (is_remote_node) { *is_remote_node = FALSE; } xpath_string = crm_strdup_printf(XPATH_NODE, host_lowercase, host_lowercase, host_lowercase, host_lowercase); if (cib_internal_op(the_cib, PCMK__CIB_REQUEST_QUERY, NULL, xpath_string, NULL, &xml_search, cib_sync_call|cib_scope_local|cib_xpath, NULL) == pcmk_ok) { rc = get_uuid_from_result(xml_search, uuid, is_remote_node); } else { rc = -ENXIO; } free(xpath_string); free_xml(xml_search); g_free(host_lowercase); if (rc != pcmk_ok) { crm_debug("Could not map node name '%s' to a UUID: %s", uname, pcmk_strerror(rc)); } else { crm_info("Mapped node name '%s' to UUID %s", uname, (uuid? *uuid : "")); } return rc; } int query_node_uname(cib_t * the_cib, const char *uuid, char **uname) { int rc = pcmk_ok; xmlNode *a_child = NULL; xmlNode *xml_obj = NULL; xmlNode *fragment = NULL; const char *child_name = NULL; CRM_ASSERT(uname != NULL); CRM_ASSERT(uuid != NULL); rc = the_cib->cmds->query(the_cib, XML_CIB_TAG_NODES, &fragment, cib_sync_call | cib_scope_local); if (rc != pcmk_ok) { return rc; } xml_obj = fragment; CRM_CHECK(pcmk__str_eq(crm_element_name(xml_obj), XML_CIB_TAG_NODES, pcmk__str_casei), return -ENOMSG); CRM_ASSERT(xml_obj != NULL); crm_log_xml_trace(xml_obj, "Result section"); rc = -ENXIO; *uname = NULL; for (a_child = pcmk__xml_first_child(xml_obj); a_child != NULL; a_child = pcmk__xml_next(a_child)) { if (pcmk__str_eq((const char *)a_child->name, XML_CIB_TAG_NODE, pcmk__str_none)) { child_name = ID(a_child); if (pcmk__str_eq(uuid, child_name, pcmk__str_casei)) { child_name = crm_element_value(a_child, XML_ATTR_UNAME); if (child_name != NULL) { *uname = strdup(child_name); rc = pcmk_ok; } break; } } } free_xml(fragment); return rc; } int set_standby(cib_t * the_cib, const char *uuid, const char *scope, const char *standby_value) { int rc = pcmk_ok; char *attr_id = NULL; CRM_CHECK(uuid != NULL, return -EINVAL); CRM_CHECK(standby_value != NULL, return -EINVAL); if (pcmk__strcase_any_of(scope, "reboot", XML_CIB_TAG_STATUS, NULL)) { scope = XML_CIB_TAG_STATUS; attr_id = crm_strdup_printf("transient-standby-%.256s", uuid); } else { scope = XML_CIB_TAG_NODES; attr_id = crm_strdup_printf("standby-%.256s", uuid); } rc = update_attr_delegate(the_cib, cib_sync_call, scope, uuid, NULL, NULL, attr_id, "standby", standby_value, TRUE, NULL, NULL); free(attr_id); return rc; } diff --git a/lib/common/ipc_controld.c b/lib/common/ipc_controld.c index d6a247b19b..b6672fc643 100644 --- a/lib/common/ipc_controld.c +++ b/lib/common/ipc_controld.c @@ -1,664 +1,671 @@ /* * Copyright 2020-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include "crmcommon_private.h" struct controld_api_private_s { char *client_uuid; unsigned int replies_expected; }; /*! * \internal * \brief Get a string representation of a controller API reply type * * \param[in] reply Controller API reply type * * \return String representation of a controller API reply type */ const char * pcmk__controld_api_reply2str(enum pcmk_controld_api_reply reply) { switch (reply) { case pcmk_controld_reply_reprobe: return "reprobe"; case pcmk_controld_reply_info: return "info"; case pcmk_controld_reply_resource: return "resource"; case pcmk_controld_reply_ping: return "ping"; case pcmk_controld_reply_nodes: return "nodes"; default: return "unknown"; } } // \return Standard Pacemaker return code static int new_data(pcmk_ipc_api_t *api) { struct controld_api_private_s *private = NULL; api->api_data = calloc(1, sizeof(struct controld_api_private_s)); if (api->api_data == NULL) { return errno; } private = api->api_data; /* This is set to the PID because that's how it was always done, but PIDs * are not unique because clients can be remote. The value appears to be * unused other than as part of F_CRM_SYS_FROM in IPC requests, which is * only compared against the internal system names (CRM_SYSTEM_TENGINE, * etc.), so it shouldn't be a problem. */ private->client_uuid = pcmk__getpid_s(); /* @TODO Implement a call ID model similar to the CIB, executor, and fencer * IPC APIs, so that requests and replies can be matched, and * duplicate replies can be discarded. */ return pcmk_rc_ok; } static void free_data(void *data) { free(((struct controld_api_private_s *) data)->client_uuid); free(data); } // \return Standard Pacemaker return code static int post_connect(pcmk_ipc_api_t *api) { /* The controller currently requires clients to register via a hello * request, but does not reply back. */ struct controld_api_private_s *private = api->api_data; const char *client_name = crm_system_name? crm_system_name : "client"; xmlNode *hello; int rc; hello = create_hello_message(private->client_uuid, client_name, PCMK__CONTROLD_API_MAJOR, PCMK__CONTROLD_API_MINOR); rc = pcmk__send_ipc_request(api, hello); free_xml(hello); if (rc != pcmk_rc_ok) { crm_info("Could not send IPC hello to %s: %s " CRM_XS " rc=%s", pcmk_ipc_name(api, true), pcmk_rc_str(rc), rc); } else { crm_debug("Sent IPC hello to %s", pcmk_ipc_name(api, true)); } return rc; } static void set_node_info_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_info; if (msg_data == NULL) { return; } data->data.node_info.have_quorum = pcmk__xe_attr_is_true(msg_data, XML_ATTR_HAVE_QUORUM); data->data.node_info.is_remote = pcmk__xe_attr_is_true(msg_data, XML_NODE_IS_REMOTE); + + /* Integer node_info.id is currently valid only for Corosync nodes. + * + * @TODO: Improve handling after crm_peer_t is refactored to handle layer- + * specific data better. + */ crm_element_value_int(msg_data, XML_ATTR_ID, &(data->data.node_info.id)); - data->data.node_info.uuid = crm_element_value(msg_data, XML_ATTR_UUID); + + data->data.node_info.uuid = crm_element_value(msg_data, XML_ATTR_ID); data->data.node_info.uname = crm_element_value(msg_data, XML_ATTR_UNAME); data->data.node_info.state = crm_element_value(msg_data, XML_NODE_IS_PEER); } static void set_ping_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_ping; if (msg_data == NULL) { return; } data->data.ping.sys_from = crm_element_value(msg_data, XML_PING_ATTR_SYSFROM); data->data.ping.fsa_state = crm_element_value(msg_data, XML_PING_ATTR_CRMDSTATE); data->data.ping.result = crm_element_value(msg_data, XML_PING_ATTR_STATUS); } static void set_nodes_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { pcmk_controld_api_node_t *node_info; data->reply_type = pcmk_controld_reply_nodes; for (xmlNode *node = first_named_child(msg_data, XML_CIB_TAG_NODE); node != NULL; node = crm_next_same_xml(node)) { long long id_ll = 0; node_info = calloc(1, sizeof(pcmk_controld_api_node_t)); crm_element_value_ll(node, XML_ATTR_ID, &id_ll); if (id_ll > 0) { node_info->id = id_ll; } node_info->uname = crm_element_value(node, XML_ATTR_UNAME); node_info->state = crm_element_value(node, XML_NODE_IN_CLUSTER); data->data.nodes = g_list_prepend(data->data.nodes, node_info); } } static bool reply_expected(pcmk_ipc_api_t *api, xmlNode *request) { const char *command = crm_element_value(request, F_CRM_TASK); if (command == NULL) { return false; } // We only need to handle commands that functions in this file can send return !strcmp(command, CRM_OP_REPROBE) || !strcmp(command, CRM_OP_NODE_INFO) || !strcmp(command, CRM_OP_PING) || !strcmp(command, CRM_OP_LRM_FAIL) || !strcmp(command, CRM_OP_LRM_DELETE); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { struct controld_api_private_s *private = api->api_data; crm_exit_t status = CRM_EX_OK; xmlNode *msg_data = NULL; const char *value = NULL; pcmk_controld_api_reply_t reply_data = { pcmk_controld_reply_unknown, NULL, NULL, }; /* If we got an ACK, return true so the caller knows to expect more responses * from the IPC server. We do this before decrementing replies_expected because * ACKs are not going to be included in that value. * * Note that we cannot do the same kind of status checking here that we do in * ipc_pacemakerd.c. The ACK message we receive does not necessarily contain * a status attribute. That is, we may receive this: * * * * Instead of this: * * */ if (pcmk__str_eq(crm_element_name(reply), "ack", pcmk__str_none)) { return true; // More replies needed } if (private->replies_expected > 0) { private->replies_expected--; } // Do some basic validation of the reply /* @TODO We should be able to verify that value is always a response, but * currently the controller doesn't always properly set the type. Even * if we fix the controller, we'll still need to handle replies from * old versions (feature set could be used to differentiate). */ value = crm_element_value(reply, F_CRM_MSG_TYPE); if (pcmk__str_empty(value) || !pcmk__str_any_of(value, XML_ATTR_REQUEST, XML_ATTR_RESPONSE, NULL)) { crm_info("Unrecognizable message from controller: " "invalid message type '%s'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } if (pcmk__str_empty(crm_element_value(reply, XML_ATTR_REFERENCE))) { crm_info("Unrecognizable message from controller: no reference"); status = CRM_EX_PROTOCOL; goto done; } value = crm_element_value(reply, F_CRM_TASK); if (pcmk__str_empty(value)) { crm_info("Unrecognizable message from controller: no command name"); status = CRM_EX_PROTOCOL; goto done; } // Parse useful info from reply reply_data.feature_set = crm_element_value(reply, XML_ATTR_VERSION); reply_data.host_from = crm_element_value(reply, F_CRM_HOST_FROM); msg_data = get_message_xml(reply, F_CRM_DATA); if (!strcmp(value, CRM_OP_REPROBE)) { reply_data.reply_type = pcmk_controld_reply_reprobe; } else if (!strcmp(value, CRM_OP_NODE_INFO)) { set_node_info_data(&reply_data, msg_data); } else if (!strcmp(value, CRM_OP_INVOKE_LRM)) { reply_data.reply_type = pcmk_controld_reply_resource; reply_data.data.resource.node_state = msg_data; } else if (!strcmp(value, CRM_OP_PING)) { set_ping_data(&reply_data, msg_data); } else if (!strcmp(value, PCMK__CONTROLD_CMD_NODES)) { set_nodes_data(&reply_data, msg_data); } else { crm_info("Unrecognizable message from controller: unknown command '%s'", value); status = CRM_EX_PROTOCOL; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); // Free any reply data that was allocated if (pcmk__str_eq(value, PCMK__CONTROLD_CMD_NODES, pcmk__str_casei)) { g_list_free_full(reply_data.data.nodes, free); } return false; // No further replies needed } pcmk__ipc_methods_t * pcmk__controld_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = new_data; cmds->free_data = free_data; cmds->post_connect = post_connect; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; } return cmds; } /*! * \internal * \brief Create XML for a controller IPC request * * \param[in] api Controller connection * \param[in] op Controller IPC command name * \param[in] node Node name to set as destination host * \param[in] msg_data XML to attach to request as message data * * \return Newly allocated XML for request */ static xmlNode * create_controller_request(const pcmk_ipc_api_t *api, const char *op, const char *node, xmlNode *msg_data) { struct controld_api_private_s *private = NULL; const char *sys_to = NULL; if (api == NULL) { return NULL; } private = api->api_data; if ((node == NULL) && !strcmp(op, CRM_OP_PING)) { sys_to = CRM_SYSTEM_DC; } else { sys_to = CRM_SYSTEM_CRMD; } return create_request(op, msg_data, node, sys_to, (crm_system_name? crm_system_name : "client"), private->client_uuid); } // \return Standard Pacemaker return code static int send_controller_request(pcmk_ipc_api_t *api, xmlNode *request, bool reply_is_expected) { int rc; if (crm_element_value(request, XML_ATTR_REFERENCE) == NULL) { return EINVAL; } rc = pcmk__send_ipc_request(api, request); if ((rc == pcmk_rc_ok) && reply_is_expected) { struct controld_api_private_s *private = api->api_data; private->replies_expected++; } return rc; } static xmlNode * create_reprobe_message_data(const char *target_node, const char *router_node) { xmlNode *msg_data; msg_data = create_xml_node(NULL, "data_for_" CRM_OP_REPROBE); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if ((router_node != NULL) && !pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } return msg_data; } /*! * \brief Send a reprobe controller operation * * \param[in,out] api Controller connection * \param[in] target_node Name of node to reprobe * \param[in] router_node Router node for host * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_reprobe. */ int pcmk_controld_api_reprobe(pcmk_ipc_api_t *api, const char *target_node, const char *router_node) { xmlNode *request; xmlNode *msg_data; int rc = pcmk_rc_ok; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } crm_debug("Sending %s IPC request to reprobe %s via %s", pcmk_ipc_name(api, true), pcmk__s(target_node, "local node"), pcmk__s(router_node, "local node")); msg_data = create_reprobe_message_data(target_node, router_node); request = create_controller_request(api, CRM_OP_REPROBE, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Send a "node info" controller operation * * \param[in,out] api Controller connection * \param[in] nodeid ID of node to get info for (or 0 for local node) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_info. */ int pcmk_controld_api_node_info(pcmk_ipc_api_t *api, uint32_t nodeid) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_NODE_INFO, NULL, NULL); if (request == NULL) { return EINVAL; } if (nodeid > 0) { crm_xml_set_id(request, "%lu", (unsigned long) nodeid); } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for status * * \param[in,out] api Controller connection * \param[in] node_name Name of node whose status is desired (NULL for DC) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_ping. */ int pcmk_controld_api_ping(pcmk_ipc_api_t *api, const char *node_name) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_PING, node_name, NULL); if (request == NULL) { return EINVAL; } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for cluster information * * \param[in,out] api Controller connection * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_nodes. */ int pcmk_controld_api_list_nodes(pcmk_ipc_api_t *api) { xmlNode *request; int rc = EINVAL; request = create_controller_request(api, PCMK__CONTROLD_CMD_NODES, NULL, NULL); if (request != NULL) { rc = send_controller_request(api, request, true); free_xml(request); } return rc; } // \return Standard Pacemaker return code static int controller_resource_op(pcmk_ipc_api_t *api, const char *op, const char *target_node, const char *router_node, bool cib_only, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { int rc = pcmk_rc_ok; char *key; xmlNode *request, *msg_data, *xml_rsc, *params; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } msg_data = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); /* The controller logs the transition key from resource op requests, so we * need to have *something* for it. * @TODO don't use "crm-resource" */ key = pcmk__transition_key(0, getpid(), 0, "xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx"); crm_xml_add(msg_data, XML_ATTR_TRANSITION_KEY, key); free(key); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if (!pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } if (cib_only) { // Indicate that only the CIB needs to be cleaned crm_xml_add(msg_data, PCMK__XA_MODE, XML_TAG_CIB); } xml_rsc = create_xml_node(msg_data, XML_CIB_TAG_RESOURCE); crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id); crm_xml_add(xml_rsc, XML_ATTR_ID_LONG, rsc_long_id); crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, standard); crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, provider); crm_xml_add(xml_rsc, XML_ATTR_TYPE, type); params = create_xml_node(msg_data, XML_TAG_ATTRS); crm_xml_add(params, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); // The controller parses the timeout from the request key = crm_meta_name(XML_ATTR_TIMEOUT); crm_xml_add(params, key, "60000"); /* 1 minute */ //@TODO pass as arg free(key); request = create_controller_request(api, op, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Ask the controller to fail a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to fail * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource to fail * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_fail(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { crm_debug("Sending %s IPC request to fail %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_FAIL, target_node, router_node, false, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Ask the controller to refresh a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to refresh * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource * \param[in] cib_only If true, clean resource from CIB only * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_refresh(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type, bool cib_only) { crm_debug("Sending %s IPC request to refresh %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_DELETE, target_node, router_node, cib_only, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Get the number of IPC replies currently expected from the controller * * \param[in] api Controller IPC API connection * * \return Number of replies expected */ unsigned int pcmk_controld_api_replies_expected(const pcmk_ipc_api_t *api) { struct controld_api_private_s *private = api->api_data; return private->replies_expected; } /*! * \brief Create XML for a controller IPC "hello" message * * \deprecated This function is deprecated as part of the public C API. */ // \todo make this static to this file when breaking API backward compatibility xmlNode * create_hello_message(const char *uuid, const char *client_name, const char *major_version, const char *minor_version) { xmlNode *hello_node = NULL; xmlNode *hello = NULL; if (pcmk__str_empty(uuid) || pcmk__str_empty(client_name) || pcmk__str_empty(major_version) || pcmk__str_empty(minor_version)) { crm_err("Could not create IPC hello message from %s (UUID %s): " "missing information", client_name? client_name : "unknown client", uuid? uuid : "unknown"); return NULL; } hello_node = create_xml_node(NULL, XML_TAG_OPTIONS); if (hello_node == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Message data creation failed", client_name, uuid); return NULL; } crm_xml_add(hello_node, "major_version", major_version); crm_xml_add(hello_node, "minor_version", minor_version); crm_xml_add(hello_node, "client_name", client_name); crm_xml_add(hello_node, "client_uuid", uuid); hello = create_request(CRM_OP_HELLO, hello_node, NULL, NULL, client_name, uuid); if (hello == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Request creation failed", client_name, uuid); return NULL; } free_xml(hello_node); crm_trace("Created hello message from %s (UUID %s)", client_name, uuid); return hello; } diff --git a/lib/pacemaker/pcmk_graph_producer.c b/lib/pacemaker/pcmk_graph_producer.c index d5f320d6a1..d129ca4861 100644 --- a/lib/pacemaker/pcmk_graph_producer.c +++ b/lib/pacemaker/pcmk_graph_producer.c @@ -1,1078 +1,1078 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include "libpacemaker_private.h" // Convenience macros for logging action properties #define action_type_str(flags) \ (pcmk_is_set((flags), pe_action_pseudo)? "pseudo-action" : "action") #define action_optional_str(flags) \ (pcmk_is_set((flags), pe_action_optional)? "optional" : "required") #define action_runnable_str(flags) \ (pcmk_is_set((flags), pe_action_runnable)? "runnable" : "unrunnable") #define action_node_str(a) \ (((a)->node == NULL)? "no node" : (a)->node->details->uname) /*! * \internal * \brief Add an XML node tag for a specified ID * * \param[in] id Node UUID to add * \param[in,out] xml Parent XML tag to add to */ static xmlNode* add_node_to_xml_by_id(const char *id, xmlNode *xml) { xmlNode *node_xml; node_xml = create_xml_node(xml, XML_CIB_TAG_NODE); - crm_xml_add(node_xml, XML_ATTR_UUID, id); + crm_xml_add(node_xml, XML_ATTR_ID, id); return node_xml; } /*! * \internal * \brief Add an XML node tag for a specified node * * \param[in] node Node to add * \param[in,out] xml XML to add node to */ static void add_node_to_xml(const pe_node_t *node, void *xml) { add_node_to_xml_by_id(node->details->id, (xmlNode *) xml); } /*! * \internal * \brief Add XML with nodes that need an update of their maintenance state * * \param[in,out] xml Parent XML tag to add to * \param[in] data_set Working set for cluster */ static int add_maintenance_nodes(xmlNode *xml, const pe_working_set_t *data_set) { GList *gIter = NULL; xmlNode *maintenance = xml?create_xml_node(xml, XML_GRAPH_TAG_MAINTENANCE):NULL; int count = 0; for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; struct pe_node_shared_s *details = node->details; if (!pe__is_guest_or_remote_node(node)) { continue; /* just remote nodes need to know atm */ } if (details->maintenance != details->remote_maintenance) { if (maintenance) { crm_xml_add( add_node_to_xml_by_id(node->details->id, maintenance), XML_NODE_IS_MAINTENANCE, details->maintenance?"1":"0"); } count++; } } crm_trace("%s %d nodes to adjust maintenance-mode " "to transition", maintenance?"Added":"Counted", count); return count; } /*! * \internal * \brief Add pseudo action with nodes needing maintenance state update * * \param[in,out] data_set Working set for cluster */ static void add_maintenance_update(pe_working_set_t *data_set) { pe_action_t *action = NULL; if (add_maintenance_nodes(NULL, data_set)) { crm_trace("adding maintenance state update pseudo action"); action = get_pseudo_op(CRM_OP_MAINTENANCE_NODES, data_set); pe__set_action_flags(action, pe_action_print_always); } } /*! * \internal * \brief Add XML with nodes that an action is expected to bring down * * If a specified action is expected to bring any nodes down, add an XML block * with their UUIDs. When a node is lost, this allows the controller to * determine whether it was expected. * * \param[in,out] xml Parent XML tag to add to * \param[in] action Action to check for downed nodes * \param[in] data_set Working set for cluster */ static void add_downed_nodes(xmlNode *xml, const pe_action_t *action, const pe_working_set_t *data_set) { CRM_CHECK(xml && action && action->node && data_set, return); if (pcmk__str_eq(action->task, CRM_OP_SHUTDOWN, pcmk__str_casei)) { /* Shutdown makes the action's node down */ xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* Fencing makes the action's node and any hosted guest nodes down */ const char *fence = g_hash_table_lookup(action->meta, "stonith_action"); if (pcmk__is_fencing_action(fence)) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->node->details->id, downed); pe_foreach_guest_node(data_set, action->node, add_node_to_xml, downed); } } else if (action->rsc && action->rsc->is_remote_node && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { /* Stopping a remote connection resource makes connected node down, * unless it's part of a migration */ GList *iter; pe_action_t *input; gboolean migrating = FALSE; for (iter = action->actions_before; iter != NULL; iter = iter->next) { input = ((pe_action_wrapper_t *) iter->data)->action; if (input->rsc && pcmk__str_eq(action->rsc->id, input->rsc->id, pcmk__str_casei) && pcmk__str_eq(input->task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { migrating = TRUE; break; } } if (!migrating) { xmlNode *downed = create_xml_node(xml, XML_GRAPH_TAG_DOWNED); add_node_to_xml_by_id(action->rsc->id, downed); } } } /*! * \internal * \brief Create a transition graph operation key for a clone action * * \param[in] action Clone action * \param[in] interval_ms Action interval in milliseconds * * \return Newly allocated string with transition graph operation key */ static char * clone_op_key(pe_action_t *action, guint interval_ms) { if (pcmk__str_eq(action->task, RSC_NOTIFY, pcmk__str_none)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_operation"); CRM_LOG_ASSERT((n_type != NULL) && (n_task != NULL)); return pcmk__notify_key(action->rsc->clone_name, n_type, n_task); } else if (action->cancel_task != NULL) { return pcmk__op_key(action->rsc->clone_name, action->cancel_task, interval_ms); } else { return pcmk__op_key(action->rsc->clone_name, action->task, interval_ms); } } /*! * \internal * \brief Add node details to transition graph action XML * * \param[in] action Scheduled action * \param[in] xml Transition graph action XML for \p action */ static void add_node_details(pe_action_t *action, xmlNode *xml) { pe_node_t *router_node = pcmk__connection_host_for_action(action); crm_xml_add(xml, XML_LRM_ATTR_TARGET, action->node->details->uname); crm_xml_add(xml, XML_LRM_ATTR_TARGET_UUID, action->node->details->id); if (router_node != NULL) { crm_xml_add(xml, XML_LRM_ATTR_ROUTER_NODE, router_node->details->uname); } } /*! * \internal * \brief Add resource details to transition graph action XML * * \param[in] action Scheduled action * \param[in] action_xml Transition graph action XML for \p action */ static void add_resource_details(pe_action_t *action, xmlNode *action_xml) { xmlNode *rsc_xml = NULL; const char *attr_list[] = { XML_AGENT_ATTR_CLASS, XML_AGENT_ATTR_PROVIDER, XML_ATTR_TYPE }; /* If a resource is locked to a node via shutdown-lock, mark its actions * so the controller can preserve the lock when the action completes. */ if (pcmk__action_locks_rsc_to_node(action)) { crm_xml_add_ll(action_xml, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) action->rsc->lock_time); } // List affected resource rsc_xml = create_xml_node(action_xml, crm_element_name(action->rsc->xml)); if (pcmk_is_set(action->rsc->flags, pe_rsc_orphan) && (action->rsc->clone_name != NULL)) { /* Use the numbered instance name here, because if there is more * than one instance on a node, we need to make sure the command * goes to the right one. * * This is important even for anonymous clones, because the clone's * unique meta-attribute might have just been toggled from on to * off. */ crm_debug("Using orphan clone name %s instead of %s", action->rsc->id, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->clone_name); crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } else if (!pcmk_is_set(action->rsc->flags, pe_rsc_unique)) { const char *xml_id = ID(action->rsc->xml); crm_debug("Using anonymous clone name %s for %s (aka %s)", xml_id, action->rsc->id, action->rsc->clone_name); /* ID is what we'd like client to use * ID_LONG is what they might know it as instead * * ID_LONG is only strictly needed /here/ during the * transition period until all nodes in the cluster * are running the new software /and/ have rebooted * once (meaning that they've only ever spoken to a DC * supporting this feature). * * If anyone toggles the unique flag to 'on', the * 'instance free' name will correspond to an orphan * and fall into the clause above instead */ crm_xml_add(rsc_xml, XML_ATTR_ID, xml_id); if ((action->rsc->clone_name != NULL) && !pcmk__str_eq(xml_id, action->rsc->clone_name, pcmk__str_none)) { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->clone_name); } else { crm_xml_add(rsc_xml, XML_ATTR_ID_LONG, action->rsc->id); } } else { CRM_ASSERT(action->rsc->clone_name == NULL); crm_xml_add(rsc_xml, XML_ATTR_ID, action->rsc->id); } for (int lpc = 0; lpc < PCMK__NELEM(attr_list); lpc++) { crm_xml_add(rsc_xml, attr_list[lpc], g_hash_table_lookup(action->rsc->meta, attr_list[lpc])); } } /*! * \internal * \brief Add action attributes to transition graph action XML * * \param[in] action Scheduled action * \param[in] action_xml Transition graph action XML for \p action */ static void add_action_attributes(pe_action_t *action, xmlNode *action_xml) { xmlNode *args_xml = NULL; /* We create free-standing XML to start, so we can sort the attributes * before adding it to action_xml, which keeps the scheduler regression * test graphs comparable. */ args_xml = create_xml_node(NULL, XML_TAG_ATTRS); crm_xml_add(args_xml, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); g_hash_table_foreach(action->extra, hash2field, args_xml); if ((action->rsc != NULL) && (action->node != NULL)) { // Get the resource instance attributes, evaluated properly for node GHashTable *params = pe_rsc_params(action->rsc, action->node, action->rsc->cluster); pcmk__substitute_remote_addr(action->rsc, params); g_hash_table_foreach(params, hash2smartfield, args_xml); } else if ((action->rsc != NULL) && (action->rsc->variant <= pe_native)) { GHashTable *params = pe_rsc_params(action->rsc, NULL, action->rsc->cluster); g_hash_table_foreach(params, hash2smartfield, args_xml); } g_hash_table_foreach(action->meta, hash2metafield, args_xml); if (action->rsc != NULL) { pe_resource_t *parent = action->rsc; while (parent != NULL) { parent->cmds->add_graph_meta(parent, args_xml); parent = parent->parent; } pcmk__add_bundle_meta_to_xml(args_xml, action); } else if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_none) && (action->node != NULL)) { /* Pass the node's attributes as meta-attributes. * * @TODO: Determine whether it is still necessary to do this. It was * added in 33d99707, probably for the libfence-based implementation in * c9a90bd, which is no longer used. */ g_hash_table_foreach(action->node->details->attrs, hash2metafield, args_xml); } sorted_xml(args_xml, action_xml, FALSE); free_xml(args_xml); } /*! * \internal * \brief Create the transition graph XML for a scheduled action * * \param[in] parent Parent XML element to add action to * \param[in] action Scheduled action * \param[in] skip_details If false, add action details as sub-elements * \param[in] data_set Cluster working set */ static void create_graph_action(xmlNode *parent, pe_action_t *action, bool skip_details, pe_working_set_t *data_set) { bool needs_node_info = true; bool needs_maintenance_info = false; xmlNode *action_xml = NULL; if ((action == NULL) || (data_set == NULL)) { return; } // Create the top-level element based on task if (pcmk__str_eq(action->task, CRM_OP_FENCE, pcmk__str_casei)) { /* All fences need node info; guest node fences are pseudo-events */ action_xml = create_xml_node(parent, pcmk_is_set(action->flags, pe_action_pseudo)? XML_GRAPH_TAG_PSEUDO_EVENT : XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_any_of(action->task, CRM_OP_SHUTDOWN, CRM_OP_CLEAR_FAILCOUNT, NULL)) { action_xml = create_xml_node(parent, XML_GRAPH_TAG_CRM_EVENT); } else if (pcmk__str_eq(action->task, CRM_OP_LRM_DELETE, pcmk__str_none)) { // CIB-only clean-up for shutdown locks action_xml = create_xml_node(parent, XML_GRAPH_TAG_CRM_EVENT); crm_xml_add(action_xml, PCMK__XA_MODE, XML_TAG_CIB); } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { if (pcmk__str_eq(action->task, CRM_OP_MAINTENANCE_NODES, pcmk__str_none)) { needs_maintenance_info = true; } action_xml = create_xml_node(parent, XML_GRAPH_TAG_PSEUDO_EVENT); needs_node_info = false; } else { action_xml = create_xml_node(parent, XML_GRAPH_TAG_RSC_OP); } crm_xml_add_int(action_xml, XML_ATTR_ID, action->id); crm_xml_add(action_xml, XML_LRM_ATTR_TASK, action->task); if ((action->rsc != NULL) && (action->rsc->clone_name != NULL)) { char *clone_key = NULL; guint interval_ms; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } clone_key = clone_op_key(action, interval_ms); crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, clone_key); crm_xml_add(action_xml, "internal_" XML_LRM_ATTR_TASK_KEY, action->uuid); free(clone_key); } else { crm_xml_add(action_xml, XML_LRM_ATTR_TASK_KEY, action->uuid); } if (needs_node_info && (action->node != NULL)) { add_node_details(action, action_xml); g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET), strdup(action->node->details->uname)); g_hash_table_insert(action->meta, strdup(XML_LRM_ATTR_TARGET_UUID), strdup(action->node->details->id)); } if (skip_details) { return; } if ((action->rsc != NULL) && !pcmk_is_set(action->flags, pe_action_pseudo)) { // This is a real resource action, so add resource details add_resource_details(action, action_xml); } /* List any attributes in effect */ add_action_attributes(action, action_xml); /* List any nodes this action is expected to make down */ if (needs_node_info && (action->node != NULL)) { add_downed_nodes(action_xml, action, data_set); } if (needs_maintenance_info) { add_maintenance_nodes(action_xml, data_set); } } /*! * \internal * \brief Check whether an action should be added to the transition graph * * \param[in] action Action to check * * \return true if action should be added to graph, otherwise false */ static bool should_add_action_to_graph(pe_action_t *action) { if (!pcmk_is_set(action->flags, pe_action_runnable)) { crm_trace("Ignoring action %s (%d): unrunnable", action->uuid, action->id); return false; } if (pcmk_is_set(action->flags, pe_action_optional) && !pcmk_is_set(action->flags, pe_action_print_always)) { crm_trace("Ignoring action %s (%d): optional", action->uuid, action->id); return false; } /* Actions for unmanaged resources should be excluded from the graph, * with the exception of monitors and cancellation of recurring monitors. */ if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pe_rsc_managed) && !pcmk__str_eq(action->task, RSC_STATUS, pcmk__str_none)) { const char *interval_ms_s; /* A cancellation of a recurring monitor will get here because the task * is cancel rather than monitor, but the interval can still be used to * recognize it. The interval has been normalized to milliseconds by * this point, so a string comparison is sufficient. */ interval_ms_s = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); if (pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches)) { crm_trace("Ignoring action %s (%d): for unmanaged resource (%s)", action->uuid, action->id, action->rsc->id); return false; } } /* Always add pseudo-actions, fence actions, and shutdown actions (already * determined to be required and runnable by this point) */ if (pcmk_is_set(action->flags, pe_action_pseudo) || pcmk__strcase_any_of(action->task, CRM_OP_FENCE, CRM_OP_SHUTDOWN, NULL)) { return true; } if (action->node == NULL) { pe_err("Skipping action %s (%d) " "because it was not allocated to a node (bug?)", action->uuid, action->id); pcmk__log_action("Unallocated", action, false); return false; } if (pcmk_is_set(action->flags, pe_action_dc)) { crm_trace("Action %s (%d) should be dumped: " "can run on DC instead of %s", action->uuid, action->id, pe__node_name(action->node)); } else if (pe__is_guest_node(action->node) && !action->node->details->remote_requires_reset) { crm_trace("Action %s (%d) should be dumped: " "assuming will be runnable on guest %s", action->uuid, action->id, pe__node_name(action->node)); } else if (!action->node->details->online) { pe_err("Skipping action %s (%d) " "because it was scheduled for offline node (bug?)", action->uuid, action->id); pcmk__log_action("Offline node", action, false); return false; } else if (action->node->details->unclean) { pe_err("Skipping action %s (%d) " "because it was scheduled for unclean node (bug?)", action->uuid, action->id); pcmk__log_action("Unclean node", action, false); return false; } return true; } /*! * \internal * \brief Check whether an ordering's flags can change an action * * \param[in] ordering Ordering to check * * \return true if ordering has flags that can change an action, false otherwise */ static bool ordering_can_change_actions(pe_action_wrapper_t *ordering) { return pcmk_any_flags_set(ordering->type, ~(pe_order_implies_first_printed |pe_order_implies_then_printed |pe_order_optional)); } /*! * \internal * \brief Check whether an action input should be in the transition graph * * \param[in] action Action to check * \param[in,out] input Action input to check * * \return true if input should be in graph, false otherwise * \note This function may not only check an input, but disable it under certian * circumstances (load or anti-colocation orderings that are not needed). */ static bool should_add_input_to_graph(pe_action_t *action, pe_action_wrapper_t *input) { if (input->state == pe_link_dumped) { return true; } if (input->type == pe_order_none) { crm_trace("Ignoring %s (%d) input %s (%d): " "ordering disabled", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && !ordering_can_change_actions(input)) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (!pcmk_is_set(input->action->flags, pe_action_runnable) && pcmk_is_set(input->type, pe_order_one_or_more)) { crm_trace("Ignoring %s (%d) input %s (%d): " "one-or-more and input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_implies_first_migratable) && !pcmk_is_set(input->action->flags, pe_action_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "implies input migratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (pcmk_is_set(input->type, pe_order_apply_first_non_migratable) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable)) { crm_trace("Ignoring %s (%d) input %s (%d): " "only if input unmigratable but input unrunnable", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if ((input->type == pe_order_optional) && pcmk_is_set(input->action->flags, pe_action_migrate_runnable) && pcmk__ends_with(input->action->uuid, "_stop_0")) { crm_trace("Ignoring %s (%d) input %s (%d): " "optional but stop in migration", action->uuid, action->id, input->action->uuid, input->action->id); return false; } else if (input->type == pe_order_load) { pe_node_t *input_node = input->action->node; // load orderings are relevant only if actions are for same node if (action->rsc && pcmk__str_eq(action->task, RSC_MIGRATE, pcmk__str_casei)) { pe_node_t *allocated = action->rsc->allocated_to; /* For load_stopped -> migrate_to orderings, we care about where it * has been allocated to, not where it will be executed. */ if ((input_node == NULL) || (allocated == NULL) || (input_node->details != allocated->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, (allocated? allocated->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } } else if ((input_node == NULL) || (action->node == NULL) || (input_node->details != action->node->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, (action->node? action->node->details->uname : ""), (input_node? input_node->details->uname : "")); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "load ordering input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->type == pe_order_anti_colocation) { if (input->action->node && action->node && (input->action->node->details != action->node->details)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation node mismatch %s vs %s", action->uuid, action->id, input->action->uuid, input->action->id, pe__node_name(action->node), pe__node_name(input->action->node)); input->type = pe_order_none; return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional)) { crm_trace("Ignoring %s (%d) input %s (%d): " "anti-colocation input optional", action->uuid, action->id, input->action->uuid, input->action->id); input->type = pe_order_none; return false; } } else if (input->action->rsc && input->action->rsc != action->rsc && pcmk_is_set(input->action->rsc->flags, pe_rsc_failed) && !pcmk_is_set(input->action->rsc->flags, pe_rsc_managed) && pcmk__ends_with(input->action->uuid, "_stop_0") && action->rsc && pe_rsc_is_clone(action->rsc)) { crm_warn("Ignoring requirement that %s complete before %s:" " unmanaged failed resources cannot prevent clone shutdown", input->action->uuid, action->uuid); return false; } else if (pcmk_is_set(input->action->flags, pe_action_optional) && !pcmk_any_flags_set(input->action->flags, pe_action_print_always|pe_action_dumped) && !should_add_action_to_graph(input->action)) { crm_trace("Ignoring %s (%d) input %s (%d): " "input optional", action->uuid, action->id, input->action->uuid, input->action->id); return false; } crm_trace("%s (%d) input %s %s (%d) on %s should be dumped: %s %s %#.6x", action->uuid, action->id, action_type_str(input->action->flags), input->action->uuid, input->action->id, action_node_str(input->action), action_runnable_str(input->action->flags), action_optional_str(input->action->flags), input->type); return true; } /*! * \internal * \brief Check whether an ordering creates an ordering loop * * \param[in] init_action "First" action in ordering * \param[in] action Callers should always set this the same as * \p init_action (this function may use a different * value for recursive calls) * \param[in] input Action wrapper for "then" action in ordering * * \return true if the ordering creates a loop, otherwise false */ bool pcmk__graph_has_loop(pe_action_t *init_action, pe_action_t *action, pe_action_wrapper_t *input) { bool has_loop = false; if (pcmk_is_set(input->action->flags, pe_action_tracking)) { crm_trace("Breaking tracking loop: %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); return false; } // Don't need to check inputs that won't be used if (!should_add_input_to_graph(action, input)) { return false; } if (input->action == init_action) { crm_debug("Input loop found in %s@%s ->...-> %s@%s", action->uuid, action->node? action->node->details->uname : "", init_action->uuid, init_action->node? init_action->node->details->uname : ""); return true; } pe__set_action_flags(input->action, pe_action_tracking); crm_trace("Checking inputs of action %s@%s input %s@%s (%#.6x)" "for graph loop with %s@%s ", action->uuid, action->node? action->node->details->uname : "", input->action->uuid, input->action->node? input->action->node->details->uname : "", input->type, init_action->uuid, init_action->node? init_action->node->details->uname : ""); // Recursively check input itself for loops for (GList *iter = input->action->actions_before; iter != NULL; iter = iter->next) { if (pcmk__graph_has_loop(init_action, input->action, (pe_action_wrapper_t *) iter->data)) { // Recursive call already logged a debug message has_loop = true; break; } } pe__clear_action_flags(input->action, pe_action_tracking); if (!has_loop) { crm_trace("No input loop found in %s@%s -> %s@%s (%#.6x)", input->action->uuid, input->action->node? input->action->node->details->uname : "", action->uuid, action->node? action->node->details->uname : "", input->type); } return has_loop; } /*! * \internal * \brief Create a synapse XML element for a transition graph * * \param[in] action Action that synapse is for * \param[in] data_set Cluster working set containing graph * * \return Newly added XML element for new graph synapse */ static xmlNode * create_graph_synapse(pe_action_t *action, pe_working_set_t *data_set) { int synapse_priority = 0; xmlNode *syn = create_xml_node(data_set->graph, "synapse"); crm_xml_add_int(syn, XML_ATTR_ID, data_set->num_synapse); data_set->num_synapse++; if (action->rsc != NULL) { synapse_priority = action->rsc->priority; } if (action->priority > synapse_priority) { synapse_priority = action->priority; } if (synapse_priority > 0) { crm_xml_add_int(syn, XML_CIB_ATTR_PRIORITY, synapse_priority); } return syn; } /*! * \internal * \brief Add an action to the transition graph XML if appropriate * * \param[in] data Action to possibly add * \param[in] user_data Cluster working set * * \note This will de-duplicate the action inputs, meaning that the * pe_action_wrapper_t:type flags can no longer be relied on to retain * their original settings. That means this MUST be called after * pcmk__apply_orderings() is complete, and nothing after this should rely * on those type flags. (For example, some code looks for type equal to * some flag rather than whether the flag is set, and some code looks for * particular combinations of flags -- such code must be done before * pcmk__create_graph().) */ static void add_action_to_graph(gpointer data, gpointer user_data) { pe_action_t *action = (pe_action_t *) data; pe_working_set_t *data_set = (pe_working_set_t *) user_data; xmlNode *syn = NULL; xmlNode *set = NULL; xmlNode *in = NULL; /* If we haven't already, de-duplicate inputs (even if we won't be adding * the action to the graph, so that crm_simulate's dot graphs don't have * duplicates). */ if (!pcmk_is_set(action->flags, pe_action_dedup)) { pcmk__deduplicate_action_inputs(action); pe__set_action_flags(action, pe_action_dedup); } if (pcmk_is_set(action->flags, pe_action_dumped) // Already added, or || !should_add_action_to_graph(action)) { // shouldn't be added return; } pe__set_action_flags(action, pe_action_dumped); crm_trace("Adding action %d (%s%s%s) to graph", action->id, action->uuid, ((action->node == NULL)? "" : " on "), ((action->node == NULL)? "" : action->node->details->uname)); syn = create_graph_synapse(action, data_set); set = create_xml_node(syn, "action_set"); in = create_xml_node(syn, "inputs"); create_graph_action(set, action, false, data_set); for (GList *lpc = action->actions_before; lpc != NULL; lpc = lpc->next) { pe_action_wrapper_t *input = (pe_action_wrapper_t *) lpc->data; if (should_add_input_to_graph(action, input)) { xmlNode *input_xml = create_xml_node(in, "trigger"); input->state = pe_link_dumped; create_graph_action(input_xml, input->action, true, data_set); } } } static int transition_id = -1; /*! * \internal * \brief Log a message after calculating a transition * * \param[in] filename Where transition input is stored */ void pcmk__log_transition_summary(const char *filename) { if (was_processing_error) { crm_err("Calculated transition %d (with errors)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else if (was_processing_warning) { crm_warn("Calculated transition %d (with warnings)%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } else { crm_notice("Calculated transition %d%s%s", transition_id, (filename == NULL)? "" : ", saving inputs in ", (filename == NULL)? "" : filename); } if (crm_config_error) { crm_notice("Configuration errors found during scheduler processing," " please run \"crm_verify -L\" to identify issues"); } } /*! * \internal * \brief Add a resource's actions to the transition graph * * \param[in] rsc Resource whose actions should be added */ void pcmk__add_rsc_actions_to_graph(pe_resource_t *rsc) { GList *iter = NULL; CRM_ASSERT(rsc != NULL); pe_rsc_trace(rsc, "Adding actions for %s to graph", rsc->id); // First add the resource's own actions g_list_foreach(rsc->actions, add_action_to_graph, rsc->cluster); // Then recursively add its children's actions (appropriate to variant) for (iter = rsc->children; iter != NULL; iter = iter->next) { pe_resource_t *child_rsc = (pe_resource_t *) iter->data; child_rsc->cmds->add_actions_to_graph(child_rsc); } } /*! * \internal * \brief Create a transition graph with all cluster actions needed * * \param[in] data_set Cluster working set */ void pcmk__create_graph(pe_working_set_t *data_set) { GList *iter = NULL; const char *value = NULL; long long limit = 0LL; transition_id++; crm_trace("Creating transition graph %d", transition_id); data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH); value = pe_pref(data_set->config_hash, "cluster-delay"); crm_xml_add(data_set->graph, "cluster-delay", value); value = pe_pref(data_set->config_hash, "stonith-timeout"); crm_xml_add(data_set->graph, "stonith-timeout", value); crm_xml_add(data_set->graph, "failed-stop-offset", "INFINITY"); if (pcmk_is_set(data_set->flags, pe_flag_start_failure_fatal)) { crm_xml_add(data_set->graph, "failed-start-offset", "INFINITY"); } else { crm_xml_add(data_set->graph, "failed-start-offset", "1"); } value = pe_pref(data_set->config_hash, "batch-limit"); crm_xml_add(data_set->graph, "batch-limit", value); crm_xml_add_int(data_set->graph, "transition_id", transition_id); value = pe_pref(data_set->config_hash, "migration-limit"); if ((pcmk__scan_ll(value, &limit, 0LL) == pcmk_rc_ok) && (limit > 0)) { crm_xml_add(data_set->graph, "migration-limit", value); } if (data_set->recheck_by > 0) { char *recheck_epoch = NULL; recheck_epoch = crm_strdup_printf("%llu", (long long) data_set->recheck_by); crm_xml_add(data_set->graph, "recheck-by", recheck_epoch); free(recheck_epoch); } /* The following code will de-duplicate action inputs, so nothing past this * should rely on the action input type flags retaining their original * values. */ // Add resource actions to graph for (iter = data_set->resources; iter != NULL; iter = iter->next) { pe_resource_t *rsc = (pe_resource_t *) iter->data; pe_rsc_trace(rsc, "Processing actions for %s", rsc->id); rsc->cmds->add_actions_to_graph(rsc); } // Add pseudo-action for list of nodes with maintenance state update add_maintenance_update(data_set); // Add non-resource (node) actions for (iter = data_set->actions; iter != NULL; iter = iter->next) { pe_action_t *action = (pe_action_t *) iter->data; if ((action->rsc != NULL) && (action->node != NULL) && action->node->details->shutdown && !pcmk_is_set(action->rsc->flags, pe_rsc_maintenance) && !pcmk_any_flags_set(action->flags, pe_action_optional|pe_action_runnable) && pcmk__str_eq(action->task, RSC_STOP, pcmk__str_none)) { /* Eventually we should just ignore the 'fence' case, but for now * it's the best way to detect (in CTS) when CIB resource updates * are being lost. */ if (pcmk_is_set(data_set->flags, pe_flag_have_quorum) || (data_set->no_quorum_policy == no_quorum_ignore)) { crm_crit("Cannot %s %s because of %s:%s%s (%s)", action->node->details->unclean? "fence" : "shut down", pe__node_name(action->node), action->rsc->id, pcmk_is_set(action->rsc->flags, pe_rsc_managed)? " blocked" : " unmanaged", pcmk_is_set(action->rsc->flags, pe_rsc_failed)? " failed" : "", action->uuid); } } add_action_to_graph((gpointer) action, (gpointer) data_set); } crm_log_xml_trace(data_set->graph, "graph"); } diff --git a/lib/pacemaker/pcmk_injections.c b/lib/pacemaker/pcmk_injections.c index 46b87c6b77..2a2efbcdac 100644 --- a/lib/pacemaker/pcmk_injections.c +++ b/lib/pacemaker/pcmk_injections.c @@ -1,751 +1,751 @@ /* * Copyright 2009-2022 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 #include #include "libpacemaker_private.h" bool pcmk__simulate_node_config = false; #define XPATH_NODE_CONFIG "//" XML_CIB_TAG_NODE "[@uname='%s']" #define XPATH_NODE_STATE "//" XML_CIB_TAG_STATE "[@uname='%s']" #define XPATH_RSC_HISTORY XPATH_NODE_STATE "//" \ XML_LRM_TAG_RESOURCE "[@id='%s']" /*! * \internal * \brief Inject a fictitious transient node attribute into scheduler input * * \param[in] out Output object for displaying error messages * \param[in] cib_node node_state XML to inject attribute into * \param[in] name Transient node attribute name to inject * \param[in] value Transient node attribute value to inject */ static void inject_transient_attr(pcmk__output_t *out, 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); } /*! * \internal * \brief Inject a fictitious fail count into a scheduler input * * \param[in] out Output object for displaying error messages * \param[in] cib_node Node state XML to inject into * \param[in] resource ID of resource for fail count to inject * \param[in] task Action name for fail count to inject * \param[in] interval_ms Action interval (in milliseconds) for fail count * \param[in] rc Action result for fail count to inject (if 0, or 7 * when interval_ms is 0, nothing will be injected) */ void pcmk__inject_failcount(pcmk__output_t *out, 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(out, cib_node, name, "value++"); free(name); name = pcmk__lastfailure_name(resource, task, interval_ms); inject_transient_attr(out, cib_node, name, now); free(name); free(now); } } /*! * \internal * \brief Create a CIB configuration entry for a fictitious node * * \param[in] cib_conn CIB object to use * \param[in] node Node name to use */ static void create_node_entry(cib_t *cib_conn, const char *node) { int rc = pcmk_ok; char *xpath = crm_strdup_printf(XPATH_NODE_CONFIG, node); rc = cib_conn->cmds->query(cib_conn, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); if (rc == -ENXIO) { // Only add if not already existing 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); } /*! * \internal * \brief Synthesize a fake executor event for an action * * \param[in] cib_resource XML for any existing resource action history * \param[in] task Name of action to synthesize * \param[in] interval_ms Interval of action to synthesize * \param[in] outcome Result of action to synthesize * * \return Newly allocated executor event * \note It is the caller's responsibility to free the result with * lrmd_free_event(). */ 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); lrmd__set_result(op, outcome, PCMK_EXEC_DONE, "Simulated action result"); op->params = NULL; // Not needed for simulation purposes op->t_run = (unsigned int) time(NULL); op->t_rcchange = op->t_run; // Use a call ID higher than any existing history entries 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; } /*! * \internal * \brief Inject a fictitious resource history entry into a scheduler input * * \param[in] cib_resource Resource history XML to inject entry into * \param[in] op Action result to inject * \param[in] target_rc Expected result for action to inject * * \return XML of injected resource history entry */ xmlNode * pcmk__inject_action_result(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); } /*! * \internal * \brief Inject a fictitious node into a scheduler input * * \param[in] cib_conn Scheduler input CIB to inject node into * \param[in] node Name of node to inject * \param[in] uuid UUID of node to inject * * \return XML of node_state entry for new node * \note If the global pcmk__simulate_node_config has been set to true, a * node entry in the configuration section will be added, as well as a * node state entry in the status section. */ xmlNode * pcmk__inject_node(cib_t *cib_conn, const char *node, const char *uuid) { int rc = pcmk_ok; xmlNode *cib_object = NULL; char *xpath = crm_strdup_printf(XPATH_NODE_STATE, node); if (pcmk__simulate_node_config) { 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 != NULL) && (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_ID, 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=%d)", node, rc); } free(xpath); CRM_ASSERT(rc == pcmk_ok); return cib_object; } /*! * \internal * \brief Inject a fictitious node state change into a scheduler input * * \param[in] cib_conn Scheduler input CIB to inject into * \param[in] node Name of node to inject change for * \param[in] up If true, change state to online, otherwise offline * * \return XML of changed (or added) node state entry */ xmlNode * pcmk__inject_node_state_change(cib_t *cib_conn, const char *node, bool up) { xmlNode *cib_node = pcmk__inject_node(cib_conn, node, NULL); if (up) { pcmk__xe_set_props(cib_node, XML_NODE_IN_CLUSTER, XML_BOOLEAN_YES, XML_NODE_IS_PEER, ONLINESTATUS, XML_NODE_JOIN_STATE, CRMD_JOINSTATE_MEMBER, XML_NODE_EXPECTED, CRMD_JOINSTATE_MEMBER, NULL); } else { pcmk__xe_set_props(cib_node, XML_NODE_IN_CLUSTER, XML_BOOLEAN_NO, XML_NODE_IS_PEER, OFFLINESTATUS, XML_NODE_JOIN_STATE, CRMD_JOINSTATE_DOWN, XML_NODE_EXPECTED, CRMD_JOINSTATE_DOWN, NULL); } crm_xml_add(cib_node, XML_ATTR_ORIGIN, crm_system_name); return cib_node; } /*! * \internal * \brief Check whether a node has history for a given resource * * \param[in] cib_node Node state XML to check * \param[in] resource Resource name to check for * * \return Resource's lrm_resource XML entry beneath \p cib_node if found, * otherwise NULL */ static xmlNode * find_resource_xml(xmlNode *cib_node, const char *resource) { const char *node = crm_element_value(cib_node, XML_ATTR_UNAME); char *xpath = crm_strdup_printf(XPATH_RSC_HISTORY, node, resource); xmlNode *match = get_xpath_object(xpath, cib_node, LOG_TRACE); free(xpath); return match; } /*! * \internal * \brief Inject a resource history element into a scheduler input * * \param[in] out Output object for displaying error messages * \param[in] cib_node Node state XML to inject resource history entry into * \param[in] resource ID (in configuration) of resource to inject * \param[in] lrm_name ID of resource as used in history (e.g. clone instance) * \param[in] rclass Resource agent class of resource to inject * \param[in] rtype Resource agent type of resource to inject * \param[in] rprovider Resource agent provider of resource to inject * * \return XML of injected resource history element * \note If a history element already exists under either \p resource or * \p lrm_name, this will return it rather than injecting a new one. */ xmlNode * pcmk__inject_resource_history(pcmk__output_t *out, xmlNode *cib_node, const char *resource, const char *lrm_name, const char *rclass, const char *rtype, const char *rprovider) { xmlNode *lrm = NULL; xmlNode *container = NULL; xmlNode *cib_resource = 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) != 0) { cib_resource = find_resource_xml(cib_node, lrm_name); if (cib_resource != NULL) { return cib_resource; } } if ((rclass == NULL) || (rtype == NULL)) { // @TODO query configuration for class, provider, type 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)) { // @TODO query configuration for provider out->err(out, "Please specify the provider for resource %s", resource); return NULL; } crm_info("Injecting new resource %s into node state '%s'", lrm_name, ID(cib_node)); 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; } static int find_ticket_state(pcmk__output_t *out, cib_t *the_cib, const char *ticket_id, xmlNode **ticket_state_xml) { int rc = pcmk_ok; xmlNode *xml_search = NULL; GString *xpath = g_string_sized_new(256); CRM_ASSERT(ticket_state_xml != NULL); *ticket_state_xml = NULL; g_string_append(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS "/" XML_CIB_TAG_TICKETS); if (ticket_id) { pcmk__g_strcat(xpath, "/" XML_CIB_TAG_TICKET_STATE "[@" XML_ATTR_ID "=\"", ticket_id, "\"]", NULL); } rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search, cib_sync_call|cib_scope_local|cib_xpath); g_string_free(xpath, TRUE); if (rc != pcmk_ok) { return rc; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search) && (ticket_id != NULL)) { out->err(out, "Multiple ticket_states match ticket_id=%s", ticket_id); } *ticket_state_xml = xml_search; return rc; } /*! * \internal * \brief Inject a ticket attribute into ticket state * * \param[in] out Output object for displaying error messages * \param[in] ticket_id Ticket whose state should be changed * \param[in] attr_name Ticket attribute name to inject * \param[in] attr_value Boolean value of ticket attribute to inject * \param[in] cib CIB object to use * * \return Standard Pacemaker return code */ static int set_ticket_state_attr(pcmk__output_t *out, const char *ticket_id, const char *attr_name, bool attr_value, cib_t *cib) { int rc = pcmk_rc_ok; xmlNode *xml_top = NULL; xmlNode *ticket_state_xml = NULL; // Check for an existing ticket state entry rc = find_ticket_state(out, cib, ticket_id, &ticket_state_xml); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { // Ticket state found, use it crm_debug("Injecting attribute into existing ticket state %s", ticket_id); xml_top = ticket_state_xml; } else if (rc == ENXIO) { // No ticket state, create it 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); } else { // Error return rc; } // Add the attribute to the ticket state pcmk__xe_set_bool_attr(ticket_state_xml, attr_name, attr_value); crm_log_xml_debug(xml_top, "Update"); // Commit the change to the CIB rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, xml_top, cib_sync_call|cib_scope_local); rc = pcmk_legacy2rc(rc); free_xml(xml_top); return rc; } /*! * \internal * \brief Inject a fictitious action into the cluster * * \param[in] out Output object for displaying error messages * \param[in] spec Action specification to inject * \param[in] cib CIB object for scheduler input * \param[in] data_set Cluster working set */ static void inject_action(pcmk__output_t *out, const char *spec, cib_t *cib, pe_working_set_t *data_set) { int rc; 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; xmlNode *cib_op = NULL; xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; pe_resource_t *rsc = NULL; lrmd_event_data_t *op = 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); goto done; } 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); goto done; } 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 = pcmk__inject_node(cib, node, NULL); CRM_ASSERT(cib_node != NULL); pcmk__inject_failcount(out, cib_node, resource, task, interval_ms, outcome); cib_resource = pcmk__inject_resource_history(out, 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 = pcmk__inject_action_result(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); done: free(task); free(node); free(key); } /*! * \internal * \brief Inject fictitious scheduler inputs * * \param[in] data_set Cluster working set * \param[in] cib CIB object for scheduler input to modify * \param[in] injections Injections to apply */ void pcmk__inject_scheduler_input(pe_working_set_t *data_set, cib_t *cib, pcmk_injections_t *injections) { int rc = pcmk_ok; GList *iter = NULL; xmlNode *cib_node = NULL; pcmk__output_t *out = data_set->priv; out->message(out, "inject-modify-config", injections->quorum, injections->watchdog); if (injections->quorum != NULL) { 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 != NULL) { rc = cib__update_node_attr(out, cib, cib_sync_call|cib_scope_local, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, XML_ATTR_HAVE_WATCHDOG, injections->watchdog, NULL, NULL); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->node_up; iter != NULL; iter = iter->next) { char *node = (char *) iter->data; out->message(out, "inject-modify-node", "Online", node); cib_node = pcmk__inject_node_state_change(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 (iter = injections->node_down; iter != NULL; iter = iter->next) { char *node = (char *) iter->data; char *xpath = NULL; out->message(out, "inject-modify-node", "Offline", node); cib_node = pcmk__inject_node_state_change(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); xpath = crm_strdup_printf("//node_state[@uname='%s']/%s", node, XML_CIB_TAG_LRM); cib->cmds->remove(cib, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); free(xpath); xpath = crm_strdup_printf("//node_state[@uname='%s']/%s", node, XML_TAG_TRANSIENT_NODEATTRS); cib->cmds->remove(cib, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); free(xpath); } for (iter = injections->node_fail; iter != NULL; iter = iter->next) { char *node = (char *) iter->data; out->message(out, "inject-modify-node", "Failing", node); cib_node = pcmk__inject_node_state_change(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 (iter = injections->ticket_grant; iter != NULL; iter = iter->next) { char *ticket_id = (char *) iter->data; out->message(out, "inject-modify-ticket", "Granting", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "granted", true, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_revoke; iter != NULL; iter = iter->next) { char *ticket_id = (char *) iter->data; out->message(out, "inject-modify-ticket", "Revoking", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "granted", false, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_standby; iter != NULL; iter = iter->next) { char *ticket_id = (char *) iter->data; out->message(out, "inject-modify-ticket", "Standby", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "standby", true, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_activate; iter != NULL; iter = iter->next) { char *ticket_id = (char *) iter->data; out->message(out, "inject-modify-ticket", "Activating", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "standby", false, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->op_inject; iter != NULL; iter = iter->next) { inject_action(out, (char *) iter->data, cib, data_set); } if (!out->is_quiet(out)) { out->end_list(out); } } 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/pacemaker/pcmk_output.c b/lib/pacemaker/pcmk_output.c index ab7a81b2bd..9ce13f247f 100644 --- a/lib/pacemaker/pcmk_output.c +++ b/lib/pacemaker/pcmk_output.c @@ -1,2315 +1,2315 @@ /* * Copyright 2019-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include static char * colocations_header(pe_resource_t *rsc, pcmk__colocation_t *cons, bool dependents) { char *retval = NULL; if (cons->primary_role > RSC_ROLE_STARTED) { retval = crm_strdup_printf("%s (score=%s, %s role=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), (dependents? "needs" : "with"), role2text(cons->primary_role), cons->id); } else { retval = crm_strdup_printf("%s (score=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), cons->id); } return retval; } static void colocations_xml_node(pcmk__output_t *out, pe_resource_t *rsc, pcmk__colocation_t *cons) { xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_DEPEND, "id", cons->id, "rsc", cons->dependent->id, "with-rsc", cons->primary->id, "score", pcmk_readable_score(cons->score), NULL); if (cons->node_attribute) { xmlSetProp(node, (pcmkXmlStr) "node-attribute", (pcmkXmlStr) cons->node_attribute); } if (cons->dependent_role != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "rsc-role", (pcmkXmlStr) role2text(cons->dependent_role)); } if (cons->primary_role != RSC_ROLE_UNKNOWN) { xmlSetProp(node, (pcmkXmlStr) "with-rsc-role", (pcmkXmlStr) role2text(cons->primary_role)); } } static int do_locations_list_xml(pcmk__output_t *out, pe_resource_t *rsc, bool add_header) { GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; if (add_header) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "locations"); } pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_LOCATION, "node", node->details->uname, "rsc", rsc->id, "id", cons->id, "score", pcmk_readable_score(node->weight), NULL); } } if (add_header) { PCMK__OUTPUT_LIST_FOOTER(out, rc); } return rc; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pe_resource_t *", "pe_node_t *", "pe_node_t *", "pe_action_t *", "pe_action_t *") static int rsc_action_item(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *origin = va_arg(args, pe_node_t *); pe_node_t *destination = va_arg(args, pe_node_t *); pe_action_t *action = va_arg(args, pe_action_t *); pe_action_t *source = va_arg(args, pe_action_t *); int len = 0; char *reason = NULL; char *details = NULL; bool same_host = false; bool same_role = false; bool need_role = false; static int rsc_width = 5; static int detail_width = 5; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if(source == NULL) { source = action; } len = strlen(rsc->id); if(len > rsc_width) { rsc_width = len + 2; } if ((rsc->role > RSC_ROLE_STARTED) || (rsc->next_role > RSC_ROLE_UNPROMOTED)) { need_role = true; } if(origin != NULL && destination != NULL && origin->details == destination->details) { same_host = true; } if(rsc->role == rsc->next_role) { same_role = true; } if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(destination)); } else if (origin == NULL) { /* Starting a resource */ details = crm_strdup_printf("%s", pe__node_name(destination)); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (destination == NULL) { /* Stopping a resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role) { /* Moving a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", pe__node_name(origin), pe__node_name(destination), role2text(rsc->role)); } else if (same_role) { /* Moving a normal resource */ details = crm_strdup_printf("%s -> %s", pe__node_name(origin), pe__node_name(destination)); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(origin)); } else { /* Moving and promoting/demoting */ details = crm_strdup_printf("%s %s -> %s %s", role2text(rsc->role), pe__node_name(origin), role2text(rsc->next_role), pe__node_name(destination)); } len = strlen(details); if(len > detail_width) { detail_width = len; } if(source->reason && !pcmk_is_set(action->flags, pe_action_runnable)) { reason = crm_strdup_printf("due to %s (blocked)", source->reason); } else if(source->reason) { reason = crm_strdup_printf("due to %s", source->reason); } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { reason = strdup("blocked"); } out->list_item(out, NULL, "%-8s %-*s ( %*s )%s%s", change, rsc_width, rsc->id, detail_width, details, reason ? " " : "", reason ? reason : ""); free(details); free(reason); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pe_resource_t *", "pe_node_t *", "pe_node_t *", "pe_action_t *", "pe_action_t *") static int rsc_action_item_xml(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *origin = va_arg(args, pe_node_t *); pe_node_t *destination = va_arg(args, pe_node_t *); pe_action_t *action = va_arg(args, pe_action_t *); pe_action_t *source = va_arg(args, pe_action_t *); char *change_str = NULL; bool same_host = false; bool same_role = false; bool need_role = false; xmlNode *xml = NULL; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if (source == NULL) { source = action; } if ((rsc->role > RSC_ROLE_STARTED) || (rsc->next_role > RSC_ROLE_UNPROMOTED)) { need_role = true; } if(origin != NULL && destination != NULL && origin->details == destination->details) { same_host = true; } if(rsc->role == rsc->next_role) { same_role = true; } change_str = g_ascii_strdown(change, -1); xml = pcmk__output_create_xml_node(out, "rsc_action", "action", change_str, "resource", rsc->id, NULL); g_free(change_str); if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } else if (origin == NULL) { /* Starting a resource */ crm_xml_add(xml, "node", destination->details->uname); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "node", origin->details->uname, NULL); } else if (destination == NULL) { /* Stopping a resource */ crm_xml_add(xml, "node", origin->details->uname); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, NULL); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ crm_xml_add(xml, "source", origin->details->uname); } else if (need_role && same_role) { /* Moving a promotable clone instance */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, "role", role2text(rsc->role), NULL); } else if (same_role) { /* Moving a normal resource */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, NULL); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "source", origin->details->uname, NULL); } else { /* Moving and promoting/demoting */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } if (source->reason && !pcmk_is_set(action->flags, pe_action_runnable)) { pcmk__xe_set_props(xml, "reason", source->reason, "blocked", "true", NULL); } else if(source->reason) { crm_xml_add(xml, "reason", source->reason); } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { pcmk__xe_set_bool_attr(xml, "blocked", true); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "bool") static int rsc_is_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources %s is colocated with", rsc->id); if (pcmk_is_set(cons->primary->flags, pe_rsc_allocating)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->primary->id, cons->id); continue; } hdr = colocations_header(cons->primary, cons, false); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->primary); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pe_resource_t *", "bool") static int rsc_is_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->primary->flags, pe_rsc_allocating)) { colocations_xml_node(out, cons->primary, cons); continue; } colocations_xml_node(out, cons->primary, cons); do_locations_list_xml(out, cons->primary, false); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } } return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "bool") static int rscs_colocated_with_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources colocated with %s", rsc->id); if (pcmk_is_set(cons->dependent->flags, pe_rsc_allocating)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->dependent->id, cons->id); continue; } hdr = colocations_header(cons->dependent, cons, true); out->list_item(out, NULL, "%s", hdr); free(hdr); /* Empty list header just for indentation of information about this resource. */ out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->dependent); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pe_resource_t *", "bool") static int rscs_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pe_rsc_allocating)) { return rc; } pe__set_resource_flags(rsc, pe_rsc_allocating); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->dependent->flags, pe_rsc_allocating)) { colocations_xml_node(out, cons->dependent, cons); continue; } colocations_xml_node(out, cons->dependent, cons); do_locations_list_xml(out, cons->dependent, false); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } } return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pe__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->node_list_rh; lpc2 != NULL; lpc2 = lpc2->next) { pe_node_t *node = (pe_node_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Locations"); out->list_item(out, NULL, "Node %s (score=%s, id=%s, rsc=%s)", pe__node_name(node), pcmk_readable_score(node->weight), cons->id, rsc->id); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("locations-list", "pe_resource_t *") static int locations_list_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); return do_locations_list_xml(out, rsc, true); } PCMK__OUTPUT_ARGS("locations-and-colocations", "pe_resource_t *", "pe_working_set_t *", "bool", "bool") static int locations_and_colocations(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(data_set); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } out->message(out, "locations-list", rsc); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("locations-and-colocations", "pe_resource_t *", "pe_working_set_t *", "bool", "bool") static int locations_and_colocations_xml(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(data_set); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } pcmk__output_xml_create_parent(out, "constraints", NULL); do_locations_list_xml(out, rsc, false); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(data_set, pe_rsc_allocating); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); pcmk__output_xml_pop_parent(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health(pcmk__output_t *out, va_list args) { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); return out->info(out, "Controller on %s in state %s: %s", pcmk__s(host_from, "unknown node"), pcmk__s(fsa_state, "unknown"), pcmk__s(result, "unknown result")); } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from G_GNUC_UNUSED = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result G_GNUC_UNUSED = va_arg(args, const char *); if (fsa_state != NULL) { pcmk__formatted_printf(out, "%s\n", fsa_state); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); pcmk__output_create_xml_node(out, pcmk__s(sys_from, ""), "node_name", pcmk__s(host_from, ""), "state", pcmk__s(fsa_state, ""), "result", pcmk__s(result, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "long long") static int pacemakerd_health(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = (time_t) va_arg(args, long long); char *last_updated_s = NULL; int rc = pcmk_rc_ok; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } rc = out->info(out, "Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); free(last_updated_s); return rc; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "long long") static int pacemakerd_health_html(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = (time_t) va_arg(args, long long); char *last_updated_s = NULL; char *msg = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } msg = crm_strdup_printf("Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); pcmk__output_create_html_node(out, "li", NULL, NULL, msg); free(msg); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "long long") static int pacemakerd_health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return pacemakerd_health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated G_GNUC_UNUSED = (time_t) va_arg(args, long long); if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } pcmk__formatted_printf(out, "%s\n", state_s); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "long long") static int pacemakerd_health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = (time_t) va_arg(args, long long); char *last_updated_s = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } pcmk__output_create_xml_node(out, "pacemakerd", "sys_from", sys_from, "state", state_s, "last_updated", last_updated_s, NULL); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_default(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); out->list_item(out, NULL, "Testing %s ... %.2f secs", xml_file, (end - start) / (float) CLOCKS_PER_SEC); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_xml(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); char *duration = pcmk__ftoa((end - start) / (float) CLOCKS_PER_SEC); pcmk__output_create_xml_node(out, "timing", "file", xml_file, "duration", duration, NULL); free(duration); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); return out->info(out, "Designated Controller is: %s", pcmk__s(dc, "not yet elected")); } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return dc(out, args); } else { const char *dc = va_arg(args, const char *); if (dc != NULL) { pcmk__formatted_printf(out, "%s\n", pcmk__s(dc, "")); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_xml(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); pcmk__output_create_xml_node(out, "dc", "node_name", pcmk__s(dc, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export = va_arg(args, int); if (bash_export) { return out->info(out, "export %s=%s", pcmk__s(name, ""), pcmk__s(id, "")); } else { return out->info(out, "%s node: %s (%s)", type ? type : "cluster", pcmk__s(name, ""), pcmk__s(id, "")); } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return crmadmin_node(out, args); } else { const char *type G_GNUC_UNUSED = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id G_GNUC_UNUSED = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__formatted_printf(out, "%s\n", pcmk__s(name, "")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_xml(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__output_create_xml_node(out, "node", "type", type ? type : "cluster", "name", pcmk__s(name, ""), "id", pcmk__s(id, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("digests", "pe_resource_t *", "pe_node_t *", "const char *", "guint", "op_digest_cache_t *") static int digests_text(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 *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); op_digest_cache_t *digests = va_arg(args, op_digest_cache_t *); char *action_desc = NULL; const char *rsc_desc = "unknown resource"; const char *node_desc = "unknown node"; if (interval_ms != 0) { action_desc = crm_strdup_printf("%ums-interval %s action", interval_ms, ((task == NULL)? "unknown" : task)); } else if (pcmk__str_eq(task, "monitor", pcmk__str_none)) { action_desc = strdup("probe action"); } else { action_desc = crm_strdup_printf("%s action", ((task == NULL)? "unknown" : task)); } if ((rsc != NULL) && (rsc->id != NULL)) { rsc_desc = rsc->id; } if ((node != NULL) && (node->details->uname != NULL)) { node_desc = node->details->uname; } out->begin_list(out, NULL, NULL, "Digests for %s %s on %s", rsc_desc, action_desc, node_desc); free(action_desc); if (digests == NULL) { out->list_item(out, NULL, "none"); out->end_list(out); return pcmk_rc_ok; } if (digests->digest_all_calc != NULL) { out->list_item(out, NULL, "%s (all parameters)", digests->digest_all_calc); } if (digests->digest_secure_calc != NULL) { out->list_item(out, NULL, "%s (non-private parameters)", digests->digest_secure_calc); } if (digests->digest_restart_calc != NULL) { out->list_item(out, NULL, "%s (non-reloadable parameters)", digests->digest_restart_calc); } out->end_list(out); return pcmk_rc_ok; } static void add_digest_xml(xmlNode *parent, const char *type, const char *digest, xmlNode *digest_source) { if (digest != NULL) { xmlNodePtr digest_xml = create_xml_node(parent, "digest"); crm_xml_add(digest_xml, "type", ((type == NULL)? "unspecified" : type)); crm_xml_add(digest_xml, "hash", digest); if (digest_source != NULL) { add_node_copy(digest_xml, digest_source); } } } PCMK__OUTPUT_ARGS("digests", "pe_resource_t *", "pe_node_t *", "const char *", "guint", "op_digest_cache_t *") static int digests_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 *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); op_digest_cache_t *digests = va_arg(args, op_digest_cache_t *); char *interval_s = crm_strdup_printf("%ums", interval_ms); xmlNode *xml = NULL; xml = pcmk__output_create_xml_node(out, "digests", "resource", pcmk__s(rsc->id, ""), "node", pcmk__s(node->details->uname, ""), "task", pcmk__s(task, ""), "interval", interval_s, NULL); free(interval_s); if (digests != NULL) { add_digest_xml(xml, "all", digests->digest_all_calc, digests->params_all); add_digest_xml(xml, "nonprivate", digests->digest_secure_calc, digests->params_secure); add_digest_xml(xml, "nonreloadable", digests->digest_restart_calc, digests->params_restart); } return pcmk_rc_ok; } #define STOP_SANITY_ASSERT(lineno) do { \ if(current && current->details->unclean) { \ /* It will be a pseudo op */ \ } else if(stop == NULL) { \ crm_err("%s:%d: No stop action exists for %s", \ __func__, lineno, rsc->id); \ CRM_ASSERT(stop != NULL); \ } else if (pcmk_is_set(stop->flags, pe_action_optional)) { \ crm_err("%s:%d: Action %s is still optional", \ __func__, lineno, stop->uuid); \ CRM_ASSERT(!pcmk_is_set(stop->flags, pe_action_optional)); \ } \ } while(0) PCMK__OUTPUT_ARGS("rsc-action", "pe_resource_t *", "pe_node_t *", "pe_node_t *") static int rsc_action_default(pcmk__output_t *out, va_list args) { pe_resource_t *rsc = va_arg(args, pe_resource_t *); pe_node_t *current = va_arg(args, pe_node_t *); pe_node_t *next = va_arg(args, pe_node_t *); GList *possible_matches = NULL; char *key = NULL; int rc = pcmk_rc_no_output; bool moving = false; pe_node_t *start_node = NULL; pe_action_t *start = NULL; pe_action_t *stop = NULL; pe_action_t *promote = NULL; pe_action_t *demote = NULL; if (!pcmk_is_set(rsc->flags, pe_rsc_managed) || (current == NULL && next == NULL)) { pe_rsc_info(rsc, "Leave %s\t(%s%s)", rsc->id, role2text(rsc->role), !pcmk_is_set(rsc->flags, pe_rsc_managed)? " unmanaged" : ""); return rc; } moving = (current != NULL) && (next != NULL) && (current->details != next->details); possible_matches = pe__resource_actions(rsc, next, RSC_START, false); if (possible_matches) { start = possible_matches->data; g_list_free(possible_matches); } if ((start == NULL) || !pcmk_is_set(start->flags, pe_action_runnable)) { start_node = NULL; } else { start_node = current; } possible_matches = pe__resource_actions(rsc, start_node, RSC_STOP, false); if (possible_matches) { stop = possible_matches->data; g_list_free(possible_matches); } else if (pcmk_is_set(rsc->flags, pe_rsc_stop_unexpected)) { /* The resource is multiply active with multiple-active set to * stop_unexpected, and not stopping on its current node, but it should * be stopping elsewhere. */ possible_matches = pe__resource_actions(rsc, NULL, RSC_STOP, false); if (possible_matches != NULL) { stop = possible_matches->data; g_list_free(possible_matches); } } possible_matches = pe__resource_actions(rsc, next, RSC_PROMOTE, false); if (possible_matches) { promote = possible_matches->data; g_list_free(possible_matches); } possible_matches = pe__resource_actions(rsc, next, RSC_DEMOTE, false); if (possible_matches) { demote = possible_matches->data; g_list_free(possible_matches); } if (rsc->role == rsc->next_role) { pe_action_t *migrate_op = NULL; CRM_CHECK(next != NULL, return rc); possible_matches = pe__resource_actions(rsc, next, RSC_MIGRATED, false); if (possible_matches) { migrate_op = possible_matches->data; } if ((migrate_op != NULL) && (current != NULL) && pcmk_is_set(migrate_op->flags, pe_action_runnable)) { rc = out->message(out, "rsc-action-item", "Migrate", rsc, current, next, start, NULL); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if (start == NULL || pcmk_is_set(start->flags, pe_action_optional)) { if ((demote != NULL) && (promote != NULL) && !pcmk_is_set(demote->flags, pe_action_optional) && !pcmk_is_set(promote->flags, pe_action_optional)) { rc = out->message(out, "rsc-action-item", "Re-promote", rsc, current, next, promote, demote); } else { pe_rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id, role2text(rsc->role), pe__node_name(next)); } } else if (!pcmk_is_set(start->flags, pe_action_runnable)) { rc = out->message(out, "rsc-action-item", "Stop", rsc, current, NULL, stop, (stop && stop->reason)? stop : start); STOP_SANITY_ASSERT(__LINE__); } else if (moving && current) { rc = out->message(out, "rsc-action-item", pcmk_is_set(rsc->flags, pe_rsc_failed)? "Recover" : "Move", rsc, current, next, stop, NULL); } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { rc = out->message(out, "rsc-action-item", "Recover", rsc, current, NULL, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); /* STOP_SANITY_ASSERT(__LINE__); False positive for migrate-fail-7 */ } g_list_free(possible_matches); return rc; } if(stop && (rsc->next_role == RSC_ROLE_STOPPED || (start && !pcmk_is_set(start->flags, pe_action_runnable)))) { GList *gIter = NULL; key = stop_key(rsc); for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; pe_action_t *stop_op = NULL; possible_matches = find_actions(rsc->actions, key, node); if (possible_matches) { stop_op = possible_matches->data; g_list_free(possible_matches); } if (stop_op && (stop_op->flags & pe_action_runnable)) { STOP_SANITY_ASSERT(__LINE__); } if (out->message(out, "rsc-action-item", "Stop", rsc, node, NULL, stop_op, (stop_op && stop_op->reason)? stop_op : start) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } free(key); } else if ((stop != NULL) && pcmk_all_flags_set(rsc->flags, pe_rsc_failed|pe_rsc_stop)) { /* 'stop' may be NULL if the failure was ignored */ rc = out->message(out, "rsc-action-item", "Recover", rsc, current, next, stop, start); STOP_SANITY_ASSERT(__LINE__); } else if (moving) { rc = out->message(out, "rsc-action-item", "Move", rsc, current, next, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (pcmk_is_set(rsc->flags, pe_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if (stop != NULL && !pcmk_is_set(stop->flags, pe_action_optional)) { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (rsc->role == RSC_ROLE_PROMOTED) { CRM_LOG_ASSERT(current != NULL); rc = out->message(out, "rsc-action-item", "Demote", rsc, current, next, demote, NULL); } else if (rsc->next_role == RSC_ROLE_PROMOTED) { CRM_LOG_ASSERT(next); rc = out->message(out, "rsc-action-item", "Promote", rsc, current, next, promote, NULL); } else if (rsc->role == RSC_ROLE_STOPPED && rsc->next_role > RSC_ROLE_STOPPED) { rc = out->message(out, "rsc-action-item", "Start", rsc, current, next, start, NULL); } return rc; } PCMK__OUTPUT_ARGS("node-action", "char *", "char *", "char *") static int node_action(pcmk__output_t *out, va_list args) { char *task = va_arg(args, char *); char *node_name = va_arg(args, char *); char *reason = va_arg(args, char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { out->list_item(out, NULL, "%s %s '%s'", task, node_name, reason); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-action", "char *", "char *", "char *") static int node_action_xml(pcmk__output_t *out, va_list args) { char *task = va_arg(args, char *); char *node_name = va_arg(args, char *); char *reason = va_arg(args, char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { pcmk__output_create_xml_node(out, "node_action", "task", task, "node", node_name, "reason", reason, NULL); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "const char *", "const char *", "bool", "bool") static int node_info_default(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); return out->info(out, "Node %" PRIu32 ": %s " "(uuid=%s, state=%s, have_quorum=%s, is_remote=%s)", node_id, pcmk__s(node_name, "unknown"), pcmk__s(uuid, "unknown"), pcmk__s(state, "unknown"), pcmk__btoa(have_quorum), pcmk__btoa(is_remote)); } PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "bool", "bool", "const char *", "const char *") static int node_info_xml(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); char *id_s = crm_strdup_printf("%" PRIu32, node_id); pcmk__output_create_xml_node(out, "node-info", "nodeid", id_s, XML_ATTR_UNAME, node_name, - XML_ATTR_UUID, uuid, + XML_ATTR_ID, uuid, XML_NODE_IS_PEER, state, XML_ATTR_HAVE_QUORUM, pcmk__btoa(have_quorum), XML_NODE_IS_REMOTE, pcmk__btoa(is_remote), NULL); free(id_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if(rsc) { out->list_item(out, NULL, "Cluster action: %s for %s on %s", task, ID(rsc), node); } else { out->list_item(out, NULL, "Cluster action: %s on %s", task, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "cluster_action", "task", task, "node", node, NULL); if (rsc) { crm_xml_add(xml_node, "id", ID(rsc)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "char *", "const char *") static int inject_fencing_action(pcmk__output_t *out, va_list args) { char *target = va_arg(args, char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Fencing %s (%s)", target, op); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "char *", "const char *") static int inject_fencing_action_xml(pcmk__output_t *out, va_list args) { char *target = va_arg(args, char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "fencing_action", "target", target, "op", op, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); out->list_item(out, NULL, "Injecting attribute %s=%s into %s '%s'", name, value, node_path, ID(cib_node)); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); pcmk__output_create_xml_node(out, "inject_attr", "name", name, "value", value, "node_path", node_path, "cib_node", ID(cib_node), NULL); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Injecting %s into the configuration", spec); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec_xml(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "inject_spec", "spec", spec, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "char *", "char *") static int inject_modify_config(pcmk__output_t *out, va_list args) { char *quorum = va_arg(args, char *); char *watchdog = va_arg(args, char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->begin_list(out, NULL, NULL, "Performing Requested Modifications"); if (quorum) { out->list_item(out, NULL, "Setting quorum: %s", quorum); } if (watchdog) { out->list_item(out, NULL, "Setting watchdog: %s", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "char *", "char *") static int inject_modify_config_xml(pcmk__output_t *out, va_list args) { char *quorum = va_arg(args, char *); char *watchdog = va_arg(args, char *); xmlNodePtr node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node = pcmk__output_xml_create_parent(out, "modifications", NULL); if (quorum) { crm_xml_add(node, "quorum", quorum); } if (watchdog) { crm_xml_add(node, "watchdog", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "char *") static int inject_modify_node(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); char *node = va_arg(args, char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Online", pcmk__str_none)) { out->list_item(out, NULL, "Bringing node %s online", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Offline", pcmk__str_none)) { out->list_item(out, NULL, "Taking node %s offline", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Failing", pcmk__str_none)) { out->list_item(out, NULL, "Failing node %s", node); return pcmk_rc_ok; } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "char *") static int inject_modify_node_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); char *node = va_arg(args, char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_node", "action", action, "node", node, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "char *") static int inject_modify_ticket(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); char *ticket = va_arg(args, char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Standby", pcmk__str_none)) { out->list_item(out, NULL, "Making ticket %s standby", ticket); } else { out->list_item(out, NULL, "%s ticket %s", action, ticket); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "char *") static int inject_modify_ticket_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); char *ticket = va_arg(args, char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_ticket", "action", action, "ticket", ticket, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Pseudo action: %s%s%s", task, node ? " on " : "", node ? node : ""); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "pseudo_action", "task", task, NULL); if (node) { crm_xml_add(xml_node, "node", node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "char *", "guint") static int inject_rsc_action(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); char *node = va_arg(args, char *); guint interval_ms = va_arg(args, guint); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (interval_ms) { out->list_item(out, NULL, "Resource action: %-15s %s=%u on %s", rsc, operation, interval_ms, node); } else { out->list_item(out, NULL, "Resource action: %-15s %s on %s", rsc, operation, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "char *", "guint") static int inject_rsc_action_xml(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); char *node = va_arg(args, char *); guint interval_ms = va_arg(args, guint); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "rsc_action", "resource", rsc, "op", operation, "node", node, NULL); if (interval_ms) { char *interval_s = pcmk__itoa(interval_ms); crm_xml_add(xml_node, "interval", interval_s); free(interval_s); } return pcmk_rc_ok; } #define CHECK_RC(retcode, retval) \ if (retval == pcmk_rc_ok) { \ retcode = pcmk_rc_ok; \ } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") int pcmk__cluster_status_text(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); int rc = pcmk_rc_no_output; bool already_printed_failure = false; CHECK_RC(rc, out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts)); if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { CHECK_RC(rc, out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { CHECK_RC(rc, out->message(out, "resource-list", data_set, show_opts, true, unames, resources, rc == pcmk_rc_ok)); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { CHECK_RC(rc, out->message(out, "node-attribute-list", data_set, show_opts, rc == pcmk_rc_ok, unames, resources)); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { CHECK_RC(rc, out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, rc == pcmk_rc_ok)); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { CHECK_RC(rc, out->message(out, "failed-action-list", data_set, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); already_printed_failure = true; } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { CHECK_RC(rc, out->message(out, "ticket-list", data_set, rc == pcmk_rc_ok)); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { CHECK_RC(rc, out->message(out, "ban-list", data_set, prefix, resources, show_opts, rc == pcmk_rc_ok)); } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { CHECK_RC(rc, out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } } return rc; } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_xml(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts); /*** NODES ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes)) { out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { /* XML output always displays full details. */ uint32_t full_show_opts = show_opts & ~pcmk_show_brief; out->message(out, "resource-list", data_set, full_show_opts, false, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", data_set, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { out->message(out, "failed-action-list", data_set, unames, resources, show_opts, false); } /* Print stonith history */ if (pcmk_is_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { out->message(out, "full-fencing-list", history_rc, stonith_history, unames, section_opts, show_opts, false); } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", data_set, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", data_set, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-status", "pe_working_set_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_html(pcmk__output_t *out, va_list args) { pe_working_set_t *data_set = va_arg(args, pe_working_set_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); bool already_printed_failure = false; out->message(out, "cluster-summary", data_set, pcmkd_state, section_opts, show_opts); /*** NODE LIST ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { out->message(out, "node-list", data_set->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { out->message(out, "resource-list", data_set, show_opts, true, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", data_set, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations | pcmk_section_failcounts)) { out->message(out, "node-summary", data_set, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && xml_has_children(data_set->failed)) { out->message(out, "failed-action-list", data_set, unames, resources, show_opts, false); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, false); } } else { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "fencing-list", hp, unames, section_opts, show_opts, false); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, false); } } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", data_set, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", data_set, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_default(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); GString *s = g_string_sized_new(50); if (!pcmk__str_empty(scope)) { pcmk__g_strcat(s, "scope=\"", scope, "\" ", NULL); } if (!pcmk__str_empty(instance)) { pcmk__g_strcat(s, "id=\"", instance, "\" ", NULL); } pcmk__g_strcat(s, "name=\"", pcmk__s(name, ""), "\" ", NULL); if (!pcmk__str_empty(host)) { pcmk__g_strcat(s, "host=\"", host, "\" ", NULL); } pcmk__g_strcat(s, "value=\"", pcmk__s(value, ""), "\"", NULL); out->info(out, "%s", s->str); g_string_free(s, TRUE); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_xml(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, "attribute", "name", name, "value", value ? value : "", NULL); if (!pcmk__str_empty(scope)) { crm_xml_add(node, "scope", scope); } if (!pcmk__str_empty(instance)) { crm_xml_add(node, "id", instance); } if (!pcmk__str_empty(host)) { crm_xml_add(node, "host", host); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_default(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); switch (result) { case pcmk_rc_within_range: return out->info(out, "Rule %s is still in effect", rule_id); case pcmk_rc_ok: return out->info(out, "Rule %s satisfies conditions", rule_id); case pcmk_rc_after_range: return out->info(out, "Rule %s is expired", rule_id); case pcmk_rc_before_range: return out->info(out, "Rule %s has not yet taken effect", rule_id); case pcmk_rc_op_unsatisfied: return out->info(out, "Rule %s does not satisfy conditions", rule_id); default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_xml(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); char *rc_str = pcmk__itoa(pcmk_rc2exitc(result)); pcmk__output_create_xml_node(out, "rule-check", "rule-id", rule_id, "rc", rc_str, NULL); free(rc_str); switch (result) { case pcmk_rc_within_range: case pcmk_rc_ok: case pcmk_rc_after_range: case pcmk_rc_before_range: case pcmk_rc_op_unsatisfied: return pcmk_rc_ok; default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_none(pcmk__output_t *out, va_list args) { return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_text(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); static int code_width = 0; if (out->is_quiet(out)) { /* If out->is_quiet(), don't print the code. Print name and/or desc in a * compact format for text output, or print nothing at all for none-type * output. */ if ((name != NULL) && (desc != NULL)) { pcmk__formatted_printf(out, "%s - %s\n", name, desc); } else if ((name != NULL) || (desc != NULL)) { pcmk__formatted_printf(out, "%s\n", ((name != NULL)? name : desc)); } return pcmk_rc_ok; } /* Get length of longest (most negative) standard Pacemaker return code * This should be longer than all the values of any other type of return * code. */ if (code_width == 0) { long long most_negative = pcmk_rc_error - (long long) pcmk__n_rc + 1; code_width = (int) snprintf(NULL, 0, "%lld", most_negative); } if ((name != NULL) && (desc != NULL)) { static int name_width = 0; if (name_width == 0) { // Get length of longest standard Pacemaker return code name for (int lpc = 0; lpc < pcmk__n_rc; lpc++) { int len = (int) strlen(pcmk_rc_name(pcmk_rc_error - lpc)); name_width = QB_MAX(name_width, len); } } return out->info(out, "% *d: %-*s %s", code_width, code, name_width, name, desc); } if ((name != NULL) || (desc != NULL)) { return out->info(out, "% *d: %s", code_width, code, ((name != NULL)? name : desc)); } return out->info(out, "% *d", code_width, code); } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_xml(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); char *code_str = pcmk__itoa(code); pcmk__output_create_xml_node(out, "result-code", "code", code_str, XML_ATTR_NAME, name, XML_ATTR_DESC, desc, NULL); free(code_str); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "attribute", "default", attribute_default }, { "attribute", "xml", attribute_xml }, { "cluster-status", "default", pcmk__cluster_status_text }, { "cluster-status", "html", cluster_status_html }, { "cluster-status", "xml", cluster_status_xml }, { "crmadmin-node", "default", crmadmin_node }, { "crmadmin-node", "text", crmadmin_node_text }, { "crmadmin-node", "xml", crmadmin_node_xml }, { "dc", "default", dc }, { "dc", "text", dc_text }, { "dc", "xml", dc_xml }, { "digests", "default", digests_text }, { "digests", "xml", digests_xml }, { "health", "default", health }, { "health", "text", health_text }, { "health", "xml", health_xml }, { "inject-attr", "default", inject_attr }, { "inject-attr", "xml", inject_attr_xml }, { "inject-cluster-action", "default", inject_cluster_action }, { "inject-cluster-action", "xml", inject_cluster_action_xml }, { "inject-fencing-action", "default", inject_fencing_action }, { "inject-fencing-action", "xml", inject_fencing_action_xml }, { "inject-modify-config", "default", inject_modify_config }, { "inject-modify-config", "xml", inject_modify_config_xml }, { "inject-modify-node", "default", inject_modify_node }, { "inject-modify-node", "xml", inject_modify_node_xml }, { "inject-modify-ticket", "default", inject_modify_ticket }, { "inject-modify-ticket", "xml", inject_modify_ticket_xml }, { "inject-pseudo-action", "default", inject_pseudo_action }, { "inject-pseudo-action", "xml", inject_pseudo_action_xml }, { "inject-rsc-action", "default", inject_rsc_action }, { "inject-rsc-action", "xml", inject_rsc_action_xml }, { "inject-spec", "default", inject_spec }, { "inject-spec", "xml", inject_spec_xml }, { "locations-list", "default", locations_list }, { "locations-list", "xml", locations_list_xml }, { "node-action", "default", node_action }, { "node-action", "xml", node_action_xml }, { "node-info", "default", node_info_default }, { "node-info", "xml", node_info_xml }, { "pacemakerd-health", "default", pacemakerd_health }, { "pacemakerd-health", "html", pacemakerd_health_html }, { "pacemakerd-health", "text", pacemakerd_health_text }, { "pacemakerd-health", "xml", pacemakerd_health_xml }, { "profile", "default", profile_default, }, { "profile", "xml", profile_xml }, { "result-code", "none", result_code_none }, { "result-code", "text", result_code_text }, { "result-code", "xml", result_code_xml }, { "rsc-action", "default", rsc_action_default }, { "rsc-action-item", "default", rsc_action_item }, { "rsc-action-item", "xml", rsc_action_item_xml }, { "rsc-is-colocated-with-list", "default", rsc_is_colocated_with_list }, { "rsc-is-colocated-with-list", "xml", rsc_is_colocated_with_list_xml }, { "rscs-colocated-with-list", "default", rscs_colocated_with_list }, { "rscs-colocated-with-list", "xml", rscs_colocated_with_list_xml }, { "rule-check", "default", rule_check_default }, { "rule-check", "xml", rule_check_xml }, { "locations-and-colocations", "default", locations_and_colocations }, { "locations-and-colocations", "xml", locations_and_colocations_xml }, { NULL, NULL, NULL } }; void pcmk__register_lib_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); }