diff --git a/daemons/controld/controld_te_actions.c b/daemons/controld/controld_te_actions.c index 26bfd41fd7..35b20dec5b 100644 --- a/daemons/controld/controld_te_actions.c +++ b/daemons/controld/controld_te_actions.c @@ -1,755 +1,748 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include char *te_uuid = NULL; GHashTable *te_targets = NULL; void send_rsc_command(crm_action_t * action); static void te_update_job_count(crm_action_t * action, int offset); static void te_start_action_timer(crm_graph_t * graph, crm_action_t * action) { action->timer = calloc(1, sizeof(crm_action_timer_t)); action->timer->timeout = action->timeout; action->timer->action = action; action->timer->source_id = g_timeout_add(action->timer->timeout + graph->network_delay, action_timer_callback, (void *)action->timer); CRM_ASSERT(action->timer->source_id != 0); } static gboolean te_pseudo_action(crm_graph_t * graph, crm_action_t * pseudo) { const char *task = crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK); /* send to peers as well? */ if (safe_str_eq(task, CRM_OP_MAINTENANCE_NODES)) { 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 (safe_str_eq(fsa_our_uname, node->uname)) { 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); update_graph(graph, pseudo); trigger_graph(); return TRUE; } void send_stonith_update(crm_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, __FUNCTION__); /* we have to mark whether or not remote nodes have already been fenced */ if (peer->flags & crm_remote_node) { time_t now = time(NULL); char *now_s = crm_itoa(now); 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); rc = fsa_cib_conn->cmds->update(fsa_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 */ /* fsa_cib_conn->cmds->bump_epoch(fsa_cib_conn, cib_quorum_override|cib_scope_local); */ erase_status_tag(peer->uname, XML_CIB_TAG_LRM, cib_scope_local); erase_status_tag(peer->uname, XML_TAG_TRANSIENT_NODEATTRS, cib_scope_local); free_xml(node_state); return; } static gboolean te_fence_node(crm_graph_t * graph, crm_action_t * action) { int rc = 0; const char *id = NULL; const char *uuid = NULL; const char *target = NULL; const char *type = NULL; gboolean invalid_action = FALSE; enum stonith_call_options options = st_opt_none; id = ID(action->xml); target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); type = crm_meta_value(action->params, "stonith_action"); 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 FALSE; } crm_notice("Requesting fencing (%s) of node %s " CRM_XS " action=%s timeout=%d", type, target, id, transition_graph->stonith_timeout); /* Passing NULL means block until we can connect... */ te_connect_stonith(NULL); if (crmd_join_phase_count(crm_join_confirmed) == 1) { options |= st_opt_allow_suicide; } rc = stonith_api->cmds->fence(stonith_api, options, target, type, transition_graph->stonith_timeout / 1000, 0); stonith_api->cmds->register_callback(stonith_api, rc, transition_graph->stonith_timeout / 1000, st_opt_timeout_updates, generate_transition_key(transition_graph->id, action->id, 0, te_uuid), "tengine_stonith_callback", tengine_stonith_callback); return TRUE; } static int get_target_rc(crm_action_t * action) { const char *target_rc_s = crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC); if (target_rc_s != NULL) { return crm_parse_int(target_rc_s, "0"); } return 0; } static gboolean te_crm_command(crm_graph_t * graph, crm_action_t * action) { char *counter = NULL; xmlNode *cmd = NULL; gboolean is_local = FALSE; const char *id = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; id = ID(action->xml); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command (id=%s) %s: no node", crm_str(id), crm_str(task)); return FALSE); crm_info("Executing crm-event (%s): %s on %s%s%s", crm_str(id), crm_str(task), on_node, is_local ? " (local)" : "", no_wait ? " - no waiting" : ""); if (safe_str_eq(router_node, fsa_our_uname)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } if (is_local && safe_str_eq(task, CRM_OP_SHUTDOWN)) { /* defer until everything else completes */ crm_info("crm-event (%s) is a local shutdown", crm_str(id)); graph->completion_action = tg_shutdown; graph->abort_reason = "local shutdown"; te_action_confirmed(action); update_graph(graph, action); trigger_graph(); return TRUE; } else if (safe_str_eq(task, CRM_OP_SHUTDOWN)) { crm_node_t *peer = crm_get_peer(0, router_node); crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN); } cmd = create_request(task, action->xml, router_node, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(cmd, XML_ATTR_TRANSITION_KEY, counter); rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_crmd, cmd, TRUE); free(counter); free_xml(cmd); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { te_action_confirmed(action); update_graph(graph, action); trigger_graph(); } else { if (action->timeout <= 0) { crm_err("Action %d: %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_start_action_timer(graph, action); } return TRUE; } -gboolean -cib_action_update(crm_action_t * action, int status, int op_rc) +void +controld_record_action_timeout(crm_action_t *action) { lrmd_event_data_t *op = NULL; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *xml_op = NULL; xmlNode *action_rsc = NULL; int rc = pcmk_ok; const char *rsc_id = NULL; - const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); 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); - if (status == PCMK_LRM_OP_PENDING) { - crm_debug("%s %d: Recording pending operation %s on %s", - crm_element_name(action->xml), action->id, task_uuid, target); - } else { - crm_warn("%s %d: %s on %s timed out", - crm_element_name(action->xml), action->id, task_uuid, target); - } + crm_warn("%s %d: %s on %s timed out", + crm_element_name(action->xml), action->id, task_uuid, target); action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if (action_rsc == NULL) { - return FALSE; + return; } rsc_id = ID(action_rsc); - CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); - return FALSE); + CRM_CHECK(rsc_id != NULL, + crm_log_xml_err(action->xml, "Bad:action"); return); /* update the CIB */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, XML_ATTR_UUID, target_uuid); crm_xml_add(state, XML_ATTR_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, XML_ATTR_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, XML_ATTR_ID, rsc_id); crm_copy_xml_element(action_rsc, rsc, XML_ATTR_TYPE); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_CLASS); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_PROVIDER); - op = convert_graph_action(NULL, action, status, op_rc); + /* If the executor gets a timeout while waiting for the action to complete, + * that will be reported via the usual callback. This timeout means that we + * didn't hear from the executor or the controller that relayed the action + * to the executor. + * + * @TODO Using PCMK_OCF_UNKNOWN_ERROR instead of PCMK_OCF_TIMEOUT is one way + * to distinguish those situations, but perhaps PCMK_OCF_TIMEOUT would be + * preferable anyway. + */ + op = convert_graph_action(NULL, action, PCMK_LRM_OP_TIMEOUT, + PCMK_OCF_UNKNOWN_ERROR); op->call_id = -1; op->user_data = generate_transition_key(transition_graph->id, action->id, target_rc, te_uuid); xml_op = create_operation_update(rsc, op, CRM_FEATURE_SET, target_rc, target, __FUNCTION__, LOG_INFO); lrmd_free_event(op); - crm_trace("Updating CIB with \"%s\" (%s): %s %s on %s", - status < 0 ? "new action" : XML_ATTR_TIMEOUT, - crm_element_name(action->xml), crm_str(task), rsc_id, target); - crm_log_xml_trace(xml_op, "Op"); + crm_log_xml_trace(xml_op, "Action timeout"); rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, state, call_options); - - crm_trace("Updating CIB with %s action %d: %s on %s (call_id=%d)", - services_lrm_status_str(status), action->id, task_uuid, target, rc); - fsa_register_cib_callback(rc, FALSE, NULL, cib_action_updated); free_xml(state); + crm_trace("Sent CIB update (call ID %d) for timeout of action %d (%s on %s)", + rc, action->id, task_uuid, target); action->sent_update = TRUE; - - if (rc < pcmk_ok) { - return FALSE; - } - - return TRUE; } static gboolean te_rsc_command(crm_graph_t * graph, crm_action_t * action) { /* never overwrite stop actions in the CIB with * anything other than completed results * * Writing pending stops makes it look like the * resource is running again */ xmlNode *cmd = NULL; xmlNode *rsc_op = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; gboolean is_local = FALSE; char *counter = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; const char *task_uuid = NULL; CRM_ASSERT(action != NULL); CRM_ASSERT(action->xml != NULL); action->executed = FALSE; on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), crm_str(task)); return FALSE); rsc_op = action->xml; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); router_node = crm_element_value(rsc_op, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (safe_str_eq(router_node, fsa_our_uname)) { 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 = __FUNCTION__, }; do_lrm_invoke(A_LRM_INVOKE, C_FSA_INTERNAL, fsa_state, I_NULL, &msg); } else { rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_lrmd, cmd, TRUE); } free(counter); free_xml(cmd); action->executed = TRUE; if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { crm_info("Action %d confirmed - no wait", action->id); action->confirmed = TRUE; /* Just mark confirmed. * Don't bump the job count only to immediately decrement it */ update_graph(transition_graph, action); trigger_graph(); } else if (action->confirmed == TRUE) { crm_debug("Action %d: %s %s on %s(timeout %dms) was already confirmed.", action->id, task, task_uuid, on_node, action->timeout); } else { if (action->timeout <= 0) { crm_err("Action %d: %s %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, task_uuid, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_update_job_count(action, 1); te_start_action_timer(graph, action); } return TRUE; } struct te_peer_s { char *name; int jobs; int migrate_jobs; }; static void te_peer_free(gpointer p) { struct te_peer_s *peer = p; free(peer->name); free(peer); } void te_reset_job_counts(void) { GHashTableIter iter; struct te_peer_s *peer = NULL; if(te_targets == NULL) { te_targets = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, te_peer_free); } g_hash_table_iter_init(&iter, te_targets); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & peer)) { peer->jobs = 0; peer->migrate_jobs = 0; } } static void te_update_job_count_on(const char *target, int offset, bool migrate) { struct te_peer_s *r = NULL; if(target == NULL || te_targets == NULL) { return; } r = g_hash_table_lookup(te_targets, target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } r->jobs += offset; if(migrate) { r->migrate_jobs += offset; } crm_trace("jobs[%s] = %d", target, r->jobs); } static void te_update_job_count(crm_action_t * action, int offset) { const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (action->type != action_type_rsc || target == NULL) { /* No limit on these */ return; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED))) { const char *t1 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); const char *t2 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); te_update_job_count_on(t1, offset, TRUE); te_update_job_count_on(t2, offset, TRUE); return; } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } te_update_job_count_on(target, offset, FALSE); } static gboolean te_should_perform_action_on(crm_graph_t * graph, crm_action_t * action, const char *target) { int limit = 0; struct te_peer_s *r = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *id = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); if(target == NULL) { /* No limit on these */ return TRUE; } else if(te_targets == NULL) { return FALSE; } r = g_hash_table_lookup(te_targets, target); limit = throttle_get_job_limit(target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } if(limit <= r->jobs) { crm_trace("Peer %s is over their job limit of %d (%d): deferring %s", target, limit, r->jobs, id); return FALSE; } else if(graph->migration_limit > 0 && r->migrate_jobs >= graph->migration_limit) { if (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) { crm_trace("Peer %s is over their migration job limit of %d (%d): deferring %s", target, graph->migration_limit, r->migrate_jobs, id); return FALSE; } } crm_trace("Peer %s has not hit their limit yet. current jobs = %d limit= %d limit", target, r->jobs, limit); return TRUE; } static gboolean te_should_perform_action(crm_graph_t * graph, crm_action_t * action) { const char *target = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != action_type_rsc) { /* No limit on these */ return TRUE; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED))) { target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); if(te_should_perform_action_on(graph, action, target) == FALSE) { return FALSE; } target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } return te_should_perform_action_on(graph, action, target); } void te_action_confirmed(crm_action_t * action) { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (action->confirmed == FALSE && action->type == action_type_rsc && target != NULL) { te_update_job_count(action, -1); } action->confirmed = TRUE; } crm_graph_functions_t te_graph_fns = { te_pseudo_action, te_rsc_command, te_crm_command, te_fence_node, te_should_perform_action, }; void notify_crmd(crm_graph_t * graph) { const char *type = "unknown"; enum crmd_fsa_input event = I_NULL; crm_debug("Processing transition completion in state %s", fsa_state2string(fsa_state)); if (graph->complete == FALSE) { CRM_CHECK(graph->complete,); graph->complete = TRUE; } switch (graph->completion_action) { case tg_stop: type = "stop"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_done: type = "done"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_restart: type = "restart"; if (fsa_state == S_TRANSITION_ENGINE) { if (transition_timer->period_ms > 0) { crm_timer_stop(transition_timer); crm_timer_start(transition_timer); } else { event = I_PE_CALC; } } else if (fsa_state == S_POLICY_ENGINE) { register_fsa_action(A_PE_INVOKE); } break; case tg_shutdown: type = "shutdown"; if (is_set(fsa_input_register, R_SHUTDOWN)) { event = I_STOP; } else { crm_err("We didn't ask to be shut down, yet the scheduler is telling us to"); event = I_TERMINATE; } } crm_debug("Transition %d status: %s - %s", graph->id, type, crm_str(graph->abort_reason)); graph->abort_reason = NULL; graph->completion_action = tg_done; clear_bit(fsa_input_register, R_IN_TRANSITION); if (event != I_NULL) { register_fsa_input(C_FSA_INTERNAL, event, NULL); } else if (fsa_source) { mainloop_set_trigger(fsa_source); } } diff --git a/daemons/controld/controld_te_callbacks.c b/daemons/controld/controld_te_callbacks.c index ce1edb7375..709141499f 100644 --- a/daemons/controld/controld_te_callbacks.c +++ b/daemons/controld/controld_te_callbacks.c @@ -1,951 +1,951 @@ /* * Copyright 2004-2018 Andrew Beekhof * * 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 /* For ONLINESTATUS etc */ void te_update_confirm(const char *event, xmlNode * msg); extern char *te_uuid; gboolean shuttingdown = FALSE; crm_graph_t *transition_graph; crm_trigger_t *transition_trigger = NULL; static unsigned long int stonith_max_attempts = 10; /* #define rsc_op_template "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_CIB_TAG_STATE"[@uname='%s']"//"XML_LRM_TAG_RSC_OP"[@id='%s]" */ #define rsc_op_template "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_LRM_TAG_RSC_OP"[@id='%s']" static const char * get_node_id(xmlNode * rsc_op) { xmlNode *node = rsc_op; while (node != NULL && safe_str_neq(XML_CIB_TAG_STATE, TYPE(node))) { node = node->parent; } CRM_CHECK(node != NULL, return NULL); return ID(node); } void update_stonith_max_attempts(const char* value) { if (safe_str_eq(value, CRM_INFINITY_S)) { stonith_max_attempts = CRM_SCORE_INFINITY; } else { stonith_max_attempts = crm_int_helper(value, NULL); } } static void te_update_diff_v1(const char *event, xmlNode *diff) { int lpc, max; xmlXPathObject *xpathObj = NULL; CRM_CHECK(diff != NULL, return); xml_log_patchset(LOG_TRACE, __FUNCTION__, diff); if (cib_config_changed(NULL, NULL, &diff)) { abort_transition(INFINITY, tg_restart, "Non-status change", diff); goto bail; /* configuration changed */ } /* Tickets Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: update", aborted); goto bail; } freeXpathObject(xpathObj); /* Tickets Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* Transient Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_TRANSIENT_NODEATTRS "//" XML_CIB_TAG_NVPAIR); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *attr = getXpathResult(xpathObj, lpc); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = NULL; if (safe_str_eq(CRM_OP_PROBED, name)) { value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); } if (crm_is_true(value) == FALSE) { abort_transition(INFINITY, tg_restart, "Transient attribute: update", attr); crm_log_xml_trace(attr, "Abort"); goto bail; } } freeXpathObject(xpathObj); /* Transient Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_TRANSIENT_NODEATTRS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Transient attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if (transition_graph->pending == 0) { xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RESOURCE); max = numXpathResults(xpathObj); if (max > 1) { crm_debug("Ignoring resource operation updates due to history refresh of %d resources", max); crm_log_xml_trace(diff, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); goto bail; } freeXpathObject(xpathObj); } /* Process operation updates */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); if (max > 0) { int lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } } freeXpathObject(xpathObj); /* Detect deleted (as opposed to replaced or added) actions - eg. crm_resource -C */ xpathObj = xpath_search(diff, "//" XML_TAG_DIFF_REMOVED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { int path_max = 0; const char *op_id = NULL; char *rsc_op_xpath = NULL; xmlXPathObject *op_match = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; op_id = ID(match); path_max = strlen(rsc_op_template) + strlen(op_id) + 1; rsc_op_xpath = calloc(1, path_max); snprintf(rsc_op_xpath, path_max, rsc_op_template, op_id); op_match = xpath_search(diff, rsc_op_xpath); if (numXpathResults(op_match) == 0) { /* Prevent false positives by matching cancelations too */ const char *node = get_node_id(match); crm_action_t *cancelled = get_cancel_action(op_id, node); if (cancelled == NULL) { crm_debug("No match for deleted action %s (%s on %s)", rsc_op_xpath, op_id, node); abort_transition(INFINITY, tg_restart, "Resource op removal", match); freeXpathObject(op_match); free(rsc_op_xpath); goto bail; } else { crm_debug("Deleted lrm_rsc_op %s on %s was for graph event %d", op_id, node, cancelled->id); } } freeXpathObject(op_match); free(rsc_op_xpath); } bail: freeXpathObject(xpathObj); } static void process_lrm_resource_diff(xmlNode *lrm_resource, const char *node) { for (xmlNode *rsc_op = __xml_first_child(lrm_resource); rsc_op != NULL; rsc_op = __xml_next(rsc_op)) { process_graph_event(rsc_op, node); } } static void process_resource_updates(const char *node, xmlNode *xml, xmlNode *change, const char *op, const char *xpath) { xmlNode *rsc = NULL; if (xml == NULL) { return; } else if (strcmp((const char*)xml->name, XML_CIB_TAG_LRM) == 0) { xml = first_named_child(xml, XML_LRM_TAG_RESOURCES); crm_trace("Got %p in %s", xml, XML_CIB_TAG_LRM); } CRM_ASSERT(strcmp((const char*)xml->name, XML_LRM_TAG_RESOURCES) == 0); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if ((transition_graph->pending == 0) && xml->children && xml->children->next) { crm_log_xml_trace(change, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); return; } for (rsc = __xml_first_child(xml); rsc != NULL; rsc = __xml_next(rsc)) { crm_trace("Processing %s", ID(rsc)); process_lrm_resource_diff(rsc, node); } } #define NODE_PATT "/lrm[@id=" static char *get_node_from_xpath(const char *xpath) { char *nodeid = NULL; char *tmp = strstr(xpath, NODE_PATT); if(tmp) { tmp += strlen(NODE_PATT); tmp += 1; nodeid = strdup(tmp); tmp = strstr(nodeid, "\'"); CRM_ASSERT(tmp); tmp[0] = 0; } return nodeid; } static char *extract_node_uuid(const char *xpath) { char *mutable_path = strdup(xpath); char *node_uuid = NULL; char *search = NULL; char *match = NULL; match = strstr(mutable_path, "node_state[@id=\'"); if (match == NULL) { free(mutable_path); return NULL; } match += strlen("node_state[@id=\'"); search = strchr(match, '\''); if (search == NULL) { free(mutable_path); return NULL; } search[0] = 0; node_uuid = strdup(match); free(mutable_path); return node_uuid; } static void abort_unless_down(const char *xpath, const char *op, xmlNode *change, const char *reason) { char *node_uuid = NULL; crm_action_t *down = NULL; if(safe_str_neq(op, "delete")) { abort_transition(INFINITY, tg_restart, reason, change); return; } node_uuid = extract_node_uuid(xpath); if(node_uuid == NULL) { crm_err("Could not extract node ID from %s", xpath); abort_transition(INFINITY, tg_restart, reason, change); return; } down = match_down_event(node_uuid); if (down == NULL) { crm_trace("Not expecting %s to be down (%s)", node_uuid, xpath); abort_transition(INFINITY, tg_restart, reason, change); } else { crm_trace("Expecting changes to %s (%s)", node_uuid, xpath); } free(node_uuid); } static void process_op_deletion(const char *xpath, xmlNode *change) { char *mutable_key = strdup(xpath); char *key; char *node_uuid; crm_action_t *cancel = NULL; // Extract the part of xpath between last pair of single quotes key = strrchr(mutable_key, '\''); if (key != NULL) { *key = '\0'; key = strrchr(mutable_key, '\''); } if (key == NULL) { crm_warn("Ignoring malformed CIB update (resource deletion of %s)", xpath); free(mutable_key); return; } ++key; node_uuid = extract_node_uuid(xpath); cancel = get_cancel_action(key, node_uuid); if (cancel) { crm_info("Cancellation of %s on %s confirmed (%d)", key, node_uuid, cancel->id); stop_te_timer(cancel->timer); te_action_confirmed(cancel); update_graph(transition_graph, cancel); trigger_graph(); } else { abort_transition(INFINITY, tg_restart, "Resource operation removal", change); } free(mutable_key); free(node_uuid); } static void process_delete_diff(const char *xpath, const char *op, xmlNode *change) { if (strstr(xpath, "/" XML_LRM_TAG_RSC_OP "[")) { process_op_deletion(xpath, change); } else if (strstr(xpath, "/" XML_CIB_TAG_LRM "[")) { abort_unless_down(xpath, op, change, "Resource state removal"); } else if (strstr(xpath, "/" XML_CIB_TAG_STATE "[")) { abort_unless_down(xpath, op, change, "Node state removal"); } else { crm_trace("Ignoring delete of %s", xpath); } } static void process_node_state_diff(xmlNode *state, xmlNode *change, const char *op, const char *xpath) { xmlNode *lrm = first_named_child(state, XML_CIB_TAG_LRM); process_resource_updates(ID(state), lrm, change, op, xpath); } static void process_status_diff(xmlNode *status, xmlNode *change, const char *op, const char *xpath) { for (xmlNode *state = __xml_first_child(status); state != NULL; state = __xml_next(state)) { process_node_state_diff(state, change, op, xpath); } } static void process_cib_diff(xmlNode *cib, xmlNode *change, const char *op, const char *xpath) { xmlNode *status = first_named_child(cib, XML_CIB_TAG_STATUS); xmlNode *config = first_named_child(cib, XML_CIB_TAG_CONFIGURATION); if (status) { process_status_diff(status, change, op, xpath); } if (config) { abort_transition(INFINITY, tg_restart, "Non-status-only change", change); } } static void te_update_diff_v2(xmlNode *diff) { crm_log_xml_trace(diff, "Patch:Raw"); for (xmlNode *change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) { xmlNode *match = NULL; const char *name = NULL; const char *xpath = crm_element_value(change, XML_DIFF_PATH); // Possible ops: create, modify, delete, move const char *op = crm_element_value(change, XML_DIFF_OP); // Ignore uninteresting updates if (op == NULL) { continue; } else if (xpath == NULL) { crm_trace("Ignoring %s change for version field", op); continue; } else if (strcmp(op, "move") == 0) { crm_trace("Ignoring move change at %s", xpath); continue; } // Find the result of create/modify ops if (strcmp(op, "create") == 0) { match = change->children; } else if (strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } else if (strcmp(op, "delete") != 0) { crm_warn("Ignoring malformed CIB update (%s operation on %s is unrecognized)", op, xpath); continue; } if (match) { if (match->type == XML_COMMENT_NODE) { crm_trace("Ignoring %s operation for comment at %s", op, xpath); continue; } name = (const char *)match->name; } crm_trace("Handling %s operation for %s%s%s", op, (xpath? xpath : "CIB"), (name? " matched by " : ""), (name? name : "")); if (strstr(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION)) { abort_transition(INFINITY, tg_restart, "Configuration change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_CIB_TAG_TICKETS) || safe_str_eq(name, XML_CIB_TAG_TICKETS)) { abort_transition(INFINITY, tg_restart, "Ticket attribute change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_TAG_TRANSIENT_NODEATTRS "[") || safe_str_eq(name, XML_TAG_TRANSIENT_NODEATTRS)) { abort_unless_down(xpath, op, change, "Transient attribute change"); break; // Won't be packaged with operation results we may be waiting for } else if (strcmp(op, "delete") == 0) { process_delete_diff(xpath, op, change); } else if (name == NULL) { crm_warn("Ignoring malformed CIB update (%s at %s has no result)", op, xpath); } else if (strcmp(name, XML_TAG_CIB) == 0) { process_cib_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATUS) == 0) { process_status_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATE) == 0) { process_node_state_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_LRM) == 0) { process_resource_updates(ID(match), match, change, op, xpath); } else if (strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = get_node_from_xpath(xpath); process_resource_updates(local_node, match, change, op, xpath); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = get_node_from_xpath(xpath); process_lrm_resource_diff(match, local_node); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = get_node_from_xpath(xpath); process_graph_event(match, local_node); free(local_node); } else { crm_warn("Ignoring malformed CIB update (%s at %s has unrecognized result %s)", op, xpath, name); } } } void te_update_diff(const char *event, xmlNode * msg) { xmlNode *diff = NULL; const char *op = NULL; int rc = -EINVAL; int format = 1; int p_add[] = { 0, 0, 0 }; int p_del[] = { 0, 0, 0 }; CRM_CHECK(msg != NULL, return); crm_element_value_int(msg, F_CIB_RC, &rc); if (transition_graph == NULL) { crm_trace("No graph"); return; } else if (rc < pcmk_ok) { crm_trace("Filter rc=%d (%s)", rc, pcmk_strerror(rc)); return; } else if (transition_graph->complete && fsa_state != S_IDLE && fsa_state != S_TRANSITION_ENGINE && fsa_state != S_POLICY_ENGINE) { crm_trace("Filter state=%s, complete=%d", fsa_state2string(fsa_state), transition_graph->complete); return; } op = crm_element_value(msg, F_CIB_OPERATION); diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); xml_patch_versions(diff, p_add, p_del); crm_debug("Processing (%s) diff: %d.%d.%d -> %d.%d.%d (%s)", op, p_del[0], p_del[1], p_del[2], p_add[0], p_add[1], p_add[2], fsa_state2string(fsa_state)); crm_element_value_int(diff, "format", &format); switch (format) { case 1: te_update_diff_v1(event, diff); break; case 2: te_update_diff_v2(diff); break; default: crm_warn("Ignoring malformed CIB update (unknown patch format %d)", format); } } gboolean process_te_message(xmlNode * msg, xmlNode * xml_data) { const char *from = crm_element_value(msg, F_ORIG); 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 *ref = crm_element_value(msg, F_CRM_REFERENCE); const char *op = crm_element_value(msg, F_CRM_TASK); const char *type = crm_element_value(msg, F_CRM_MSG_TYPE); crm_trace("Processing %s (%s) message", op, ref); crm_log_xml_trace(msg, "ipc"); if (op == NULL) { /* error */ } else if (sys_to == NULL || strcasecmp(sys_to, CRM_SYSTEM_TENGINE) != 0) { crm_trace("Bad sys-to %s", crm_str(sys_to)); return FALSE; } else if (safe_str_eq(op, CRM_OP_INVOKE_LRM) && safe_str_eq(sys_from, CRM_SYSTEM_LRMD) /* && safe_str_eq(type, XML_ATTR_RESPONSE) */ ) { xmlXPathObject *xpathObj = NULL; crm_log_xml_trace(msg, "Processing (N)ACK"); crm_debug("Processing (N)ACK %s from %s", crm_element_value(msg, F_CRM_REFERENCE), from); xpathObj = xpath_search(xml_data, "//" XML_LRM_TAG_RSC_OP); if (numXpathResults(xpathObj)) { int lpc = 0, max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } freeXpathObject(xpathObj); } else { crm_log_xml_err(msg, "Invalid (N)ACK"); freeXpathObject(xpathObj); return FALSE; } } else { crm_err("Unknown command: %s::%s from %s", type, op, sys_from); } crm_trace("finished processing message"); return TRUE; } GHashTable *stonith_failures = NULL; struct st_fail_rec { int count; }; 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; } } } void st_fail_count_increment(const char *target) { struct st_fail_rec *rec = NULL; if (stonith_failures == NULL) { stonith_failures = crm_str_table_new(); } 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); } } /*! * \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) */ void abort_for_stonith_failure(enum transition_action abort_action, const char *target, xmlNode *reason) { /* If stonith repeatedly fails, we eventually give up on starting a new * transition for that reason. */ if ((abort_action != tg_stop) && too_many_st_failures(target)) { abort_action = tg_stop; } abort_transition(INFINITY, abort_action, "Stonith failed", reason); } void tengine_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data) { char *uuid = NULL; int target_rc = -1; int stonith_id = -1; int transition_id = -1; crm_action_t *action = NULL; int call_id = data->call_id; int rc = data->rc; char *userdata = data->userdata; CRM_CHECK(userdata != NULL, return); crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata, pcmk_strerror(rc), rc); if (AM_I_DC == FALSE) { return; } /* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */ /* op->call_id, op->optype, op->node_name, op->op_result, */ /* (char *)op->node_list, op->private_data); */ /* filter out old STONITH actions */ CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, &target_rc), crm_err("Invalid event detected"); goto bail; ); if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid) || transition_graph->id != transition_id) { crm_info("Ignoring STONITH action initiated outside of the current transition"); goto bail; } action = get_action(stonith_id, FALSE); if (action == NULL) { crm_err("Stonith action not matched"); goto bail; } stop_te_timer(action->timer); if (rc == pcmk_ok) { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); 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("Stonith operation %d for %s passed", call_id, target); if (action->confirmed == FALSE) { te_action_confirmed(action); if (safe_str_eq("on", op)) { const char *value = NULL; char *now = crm_itoa(time(NULL)); update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE); free(now); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL); update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE); value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE); update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE); } else if (action->sent_update == FALSE) { send_stonith_update(action, target, uuid); action->sent_update = TRUE; } } st_fail_count_reset(target); } else { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); enum transition_action abort_action = tg_restart; action->failed = TRUE; crm_notice("Stonith operation %d for %s failed (%s): aborting transition.", call_id, target, pcmk_strerror(rc)); /* 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 (rc == -ENODEV) { crm_warn("No devices found in cluster to fence %s, giving up", target); abort_action = tg_stop; } /* Increment the fail count now, so abort_for_stonith_failure() can * check it. Non-DC nodes will increment it in tengine_stonith_notify(). */ st_fail_count_increment(target); abort_for_stonith_failure(abort_action, target, NULL); } update_graph(transition_graph, action); trigger_graph(); bail: free(userdata); free(uuid); return; } 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, tg_shutdown, "CIB update failed", NULL); } else { crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data); } } void cib_action_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc < pcmk_ok) { crm_err("Update %d FAILED: %s", call_id, pcmk_strerror(rc)); } } gboolean action_timer_callback(gpointer data) { crm_action_timer_t *timer = NULL; CRM_CHECK(data != NULL, return FALSE); timer = (crm_action_timer_t *) data; stop_te_timer(timer); crm_warn("Timer popped (timeout=%d, abort_level=%d, complete=%s)", timer->timeout, transition_graph->abort_priority, transition_graph->complete ? "true" : "false"); CRM_CHECK(timer->action != NULL, return FALSE); if (transition_graph->complete) { crm_warn("Ignoring timeout while not in transition"); } else { /* fail the action */ gboolean send_update = TRUE; const char *task = crm_element_value(timer->action->xml, XML_LRM_ATTR_TASK); print_action(LOG_ERR, "Aborting transition, action lost: ", timer->action); timer->action->failed = TRUE; te_action_confirmed(timer->action); abort_transition(INFINITY, tg_restart, "Action lost", NULL); update_graph(transition_graph, timer->action); trigger_graph(); if (timer->action->type != action_type_rsc) { send_update = FALSE; } else if (safe_str_eq(task, RSC_CANCEL)) { /* we don't need to update the CIB with these */ send_update = FALSE; } if (send_update) { - cib_action_update(timer->action, PCMK_LRM_OP_TIMEOUT, PCMK_OCF_UNKNOWN_ERROR); + controld_record_action_timeout(timer->action); } } return FALSE; } diff --git a/daemons/controld/controld_transition.h b/daemons/controld/controld_transition.h index 43c8c9fd35..563030642b 100644 --- a/daemons/controld/controld_transition.h +++ b/daemons/controld/controld_transition.h @@ -1,83 +1,83 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef TENGINE__H # define TENGINE__H # include # include # include # include extern stonith_t *stonith_api; extern void send_stonith_update(crm_action_t * stonith_action, const char *target, const char *uuid); /* stonith cleanup list */ void add_stonith_cleanup(const char *target); void remove_stonith_cleanup(const char *target); void purge_stonith_cleanup(void); void execute_stonith_cleanup(void); /* tengine */ extern crm_action_t *match_down_event(const char *target); extern crm_action_t *get_cancel_action(const char *id, const char *node); -extern gboolean cib_action_update(crm_action_t * action, int status, int op_rc); +void controld_record_action_timeout(crm_action_t *action); extern gboolean fail_incompletable_actions(crm_graph_t * graph, const char *down_node); void process_graph_event(xmlNode *event, const char *event_node); /* utils */ extern crm_action_t *get_action(int id, gboolean confirmed); extern gboolean stop_te_timer(crm_action_timer_t * timer); extern const char *get_rsc_state(const char *task, enum op_status status); /* unpack */ extern gboolean process_te_message(xmlNode * msg, xmlNode * xml_data); extern crm_graph_t *transition_graph; extern crm_trigger_t *transition_trigger; extern char *te_uuid; extern void notify_crmd(crm_graph_t * graph); void cib_fencing_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data); void cib_action_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data); gboolean action_timer_callback(gpointer data); gboolean te_graph_trigger(gpointer user_data); void te_update_diff(const char *event, xmlNode *msg); void tengine_stonith_callback(stonith_t *stonith, stonith_callback_data_t *data); void update_stonith_max_attempts(const char* value); extern void trigger_graph_processing(const char *fn, int line); void abort_after_delay(int abort_priority, enum transition_action abort_action, const char *abort_text, guint delay_ms); extern void abort_transition_graph(int abort_priority, enum transition_action abort_action, const char *abort_text, xmlNode * reason, const char *fn, int line); # define trigger_graph() trigger_graph_processing(__FUNCTION__, __LINE__) # define abort_transition(pri, action, text, reason) \ abort_transition_graph(pri, action, text, reason,__FUNCTION__,__LINE__); extern gboolean te_connect_stonith(gpointer user_data); extern crm_trigger_t *transition_trigger; extern crm_trigger_t *stonith_reconnect; extern char *failed_stop_offset; extern char *failed_start_offset; extern int active_timeout; extern int stonith_op_active; void te_action_confirmed(crm_action_t * action); void te_reset_job_counts(void); #endif