diff --git a/daemons/controld/controld_control.c b/daemons/controld/controld_control.c index 188de50110..cfae5b058a 100644 --- a/daemons/controld/controld_control.c +++ b/daemons/controld/controld_control.c @@ -1,838 +1,846 @@ /* * 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 qb_ipcs_service_t *ipcs = NULL; #if SUPPORT_COROSYNC extern gboolean crm_connect_corosync(crm_cluster_t * cluster); #endif void crm_shutdown(int nsig); gboolean crm_read_options(gpointer user_data); gboolean fsa_has_quorum = FALSE; crm_trigger_t *fsa_source = NULL; crm_trigger_t *config_read = NULL; bool no_quorum_suicide_escalation = FALSE; bool controld_shutdown_lock_enabled = false; /* A_HA_CONNECT */ void do_ha_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) { gboolean registered = FALSE; static crm_cluster_t *cluster = NULL; if (cluster == NULL) { cluster = calloc(1, sizeof(crm_cluster_t)); } if (action & A_HA_DISCONNECT) { crm_cluster_disconnect(cluster); crm_info("Disconnected from the cluster"); controld_set_fsa_input_flags(R_HA_DISCONNECTED); } if (action & A_HA_CONNECT) { crm_set_status_callback(&peer_update_callback); crm_set_autoreap(FALSE); if (is_corosync_cluster()) { #if SUPPORT_COROSYNC registered = crm_connect_corosync(cluster); #endif } if (registered == TRUE) { controld_election_init(cluster->uname); fsa_our_uname = cluster->uname; fsa_our_uuid = cluster->uuid; if(cluster->uuid == NULL) { crm_err("Could not obtain local uuid"); registered = FALSE; } } if (registered == FALSE) { controld_set_fsa_input_flags(R_HA_DISCONNECTED); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } populate_cib_nodes(node_update_none, __func__); controld_clear_fsa_input_flags(R_HA_DISCONNECTED); crm_info("Connected to the cluster"); } if (action & ~(A_HA_CONNECT | A_HA_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __func__); } } /* A_SHUTDOWN */ void do_shutdown(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) { /* just in case */ controld_set_fsa_input_flags(R_SHUTDOWN); controld_disconnect_fencer(FALSE); } /* A_SHUTDOWN_REQ */ void do_shutdown_req(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 *msg = NULL; controld_set_fsa_input_flags(R_SHUTDOWN); //controld_set_fsa_input_flags(R_STAYDOWN); crm_info("Sending shutdown request to all peers (DC is %s)", (fsa_our_dc? fsa_our_dc : "not set")); msg = create_request(CRM_OP_SHUTDOWN_REQ, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } free_xml(msg); } extern char *max_generation_from; extern xmlNode *max_generation_xml; extern GHashTable *resource_history; extern GHashTable *voted; +extern pcmk__output_t *logger_out; void crmd_fast_exit(crm_exit_t exit_code) { if (pcmk_is_set(fsa_input_register, R_STAYDOWN)) { crm_warn("Inhibiting respawn "CRM_XS" remapping exit code %d to %d", exit_code, CRM_EX_FATAL); exit_code = CRM_EX_FATAL; } else if ((exit_code == CRM_EX_OK) && pcmk_is_set(fsa_input_register, R_IN_RECOVERY)) { crm_err("Could not recover from internal error"); exit_code = CRM_EX_ERROR; } + + if (logger_out != NULL) { + logger_out->finish(logger_out, exit_code, true, NULL); + pcmk__output_free(logger_out); + logger_out = NULL; + } + crm_exit(exit_code); } crm_exit_t crmd_exit(crm_exit_t exit_code) { GList *gIter = NULL; GMainLoop *mloop = crmd_mainloop; static bool in_progress = FALSE; if (in_progress && (exit_code == CRM_EX_OK)) { crm_debug("Exit is already in progress"); return exit_code; } else if(in_progress) { crm_notice("Error during shutdown process, exiting now with status %d (%s)", exit_code, crm_exit_str(exit_code)); crm_write_blackbox(SIGTRAP, NULL); crmd_fast_exit(exit_code); } in_progress = TRUE; crm_trace("Preparing to exit with status %d (%s)", exit_code, crm_exit_str(exit_code)); /* Suppress secondary errors resulting from us disconnecting everything */ controld_set_fsa_input_flags(R_HA_DISCONNECTED); /* Close all IPC servers and clients to ensure any and all shared memory files are cleaned up */ if(ipcs) { crm_trace("Closing IPC server"); mainloop_del_ipc_server(ipcs); ipcs = NULL; } controld_close_attrd_ipc(); controld_shutdown_schedulerd_ipc(); controld_disconnect_fencer(TRUE); if ((exit_code == CRM_EX_OK) && (crmd_mainloop == NULL)) { crm_debug("No mainloop detected"); exit_code = CRM_EX_ERROR; } /* On an error, just get out. * * Otherwise, make the effort to have mainloop exit gracefully so * that it (mostly) cleans up after itself and valgrind has less * to report on - allowing real errors stand out */ if (exit_code != CRM_EX_OK) { crm_notice("Forcing immediate exit with status %d (%s)", exit_code, crm_exit_str(exit_code)); crm_write_blackbox(SIGTRAP, NULL); crmd_fast_exit(exit_code); } /* Clean up as much memory as possible for valgrind */ for (gIter = fsa_message_queue; gIter != NULL; gIter = gIter->next) { fsa_data_t *fsa_data = gIter->data; crm_info("Dropping %s: [ state=%s cause=%s origin=%s ]", fsa_input2string(fsa_data->fsa_input), fsa_state2string(fsa_state), fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin); delete_fsa_input(fsa_data); } controld_clear_fsa_input_flags(R_MEMBERSHIP); g_list_free(fsa_message_queue); fsa_message_queue = NULL; metadata_cache_fini(); controld_election_fini(); /* Tear down the CIB manager connection, but don't free it yet -- it could * be used when we drain the mainloop later. */ controld_disconnect_cib_manager(); verify_stopped(fsa_state, LOG_WARNING); controld_clear_fsa_input_flags(R_LRM_CONNECTED); lrm_state_destroy_all(); /* This basically will not work, since mainloop has a reference to it */ mainloop_destroy_trigger(fsa_source); fsa_source = NULL; mainloop_destroy_trigger(config_read); config_read = NULL; mainloop_destroy_trigger(transition_trigger); transition_trigger = NULL; pcmk__client_cleanup(); crm_peer_destroy(); controld_free_fsa_timers(); te_cleanup_stonith_history_sync(NULL, TRUE); controld_free_sched_timer(); free(fsa_our_dc_version); fsa_our_dc_version = NULL; free(fsa_our_uname); fsa_our_uname = NULL; free(fsa_our_uuid); fsa_our_uuid = NULL; free(fsa_our_dc); fsa_our_dc = NULL; free(fsa_cluster_name); fsa_cluster_name = NULL; free(te_uuid); te_uuid = NULL; free(failed_stop_offset); failed_stop_offset = NULL; free(failed_start_offset); failed_start_offset = NULL; free(max_generation_from); max_generation_from = NULL; free_xml(max_generation_xml); max_generation_xml = NULL; mainloop_destroy_signal(SIGPIPE); mainloop_destroy_signal(SIGUSR1); mainloop_destroy_signal(SIGTERM); mainloop_destroy_signal(SIGTRAP); /* leave SIGCHLD engaged as we might still want to drain some service-actions */ if (mloop) { GMainContext *ctx = g_main_loop_get_context(crmd_mainloop); /* Don't re-enter this block */ crmd_mainloop = NULL; /* no signals on final draining anymore */ mainloop_destroy_signal(SIGCHLD); crm_trace("Draining mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx)); { int lpc = 0; while((g_main_context_pending(ctx) && lpc < 10)) { lpc++; crm_trace("Iteration %d", lpc); g_main_context_dispatch(ctx); } } crm_trace("Closing mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx)); g_main_loop_quit(mloop); /* Won't do anything yet, since we're inside it now */ g_main_loop_unref(mloop); } else { mainloop_destroy_signal(SIGCHLD); } cib_delete(fsa_cib_conn); fsa_cib_conn = NULL; throttle_fini(); /* Graceful */ crm_trace("Done preparing for exit with status %d (%s)", exit_code, crm_exit_str(exit_code)); return exit_code; } /* A_EXIT_0, A_EXIT_1 */ void do_exit(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_exit_t exit_code = CRM_EX_OK; int log_level = LOG_INFO; const char *exit_type = "gracefully"; if (action & A_EXIT_1) { log_level = LOG_ERR; exit_type = "forcefully"; exit_code = CRM_EX_ERROR; } verify_stopped(cur_state, LOG_ERR); do_crm_log(log_level, "Performing %s - %s exiting the controller", fsa_action2string(action), exit_type); crm_info("[%s] stopped (%d)", crm_system_name, exit_code); crmd_exit(exit_code); } static void sigpipe_ignore(int nsig) { return; } /* A_STARTUP */ void do_startup(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("Registering Signal Handlers"); mainloop_add_signal(SIGTERM, crm_shutdown); mainloop_add_signal(SIGPIPE, sigpipe_ignore); fsa_source = mainloop_add_trigger(G_PRIORITY_HIGH, crm_fsa_trigger, NULL); config_read = mainloop_add_trigger(G_PRIORITY_HIGH, crm_read_options, NULL); transition_trigger = mainloop_add_trigger(G_PRIORITY_LOW, te_graph_trigger, NULL); crm_debug("Creating CIB manager and executor objects"); fsa_cib_conn = cib_new(); lrm_state_init_local(); if (controld_init_fsa_timers() == FALSE) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } // \return libqb error code (0 on success, -errno on error) static int32_t accept_controller_client(qb_ipcs_connection_t *c, uid_t uid, gid_t gid) { crm_trace("Accepting new IPC client connection"); if (pcmk__new_client(c, uid, gid) == NULL) { return -EIO; } return 0; } // \return libqb error code (0 on success, -errno on error) static int32_t dispatch_controller_ipc(qb_ipcs_connection_t * c, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; pcmk__client_t *client = pcmk__find_client(c); xmlNode *msg = pcmk__client_data2xml(client, data, &id, &flags); if (msg == NULL) { pcmk__ipc_send_ack(client, id, flags, "ack", CRM_EX_PROTOCOL); return 0; } pcmk__ipc_send_ack(client, id, flags, "ack", CRM_EX_INDETERMINATE); CRM_ASSERT(client->user != NULL); pcmk__update_acl_user(msg, F_CRM_USER, client->user); crm_xml_add(msg, F_CRM_SYS_FROM, client->id); if (controld_authorize_ipc_message(msg, client, NULL)) { crm_trace("Processing IPC message from client %s", pcmk__client_name(client)); route_message(C_IPC_MESSAGE, msg); } trigger_fsa(); free_xml(msg); return 0; } static int32_t crmd_ipc_closed(qb_ipcs_connection_t * c) { pcmk__client_t *client = pcmk__find_client(c); if (client) { crm_trace("Disconnecting %sregistered client %s (%p/%p)", (client->userdata? "" : "un"), pcmk__client_name(client), c, client); free(client->userdata); pcmk__free_client(client); trigger_fsa(); } return 0; } static void crmd_ipc_destroy(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); crmd_ipc_closed(c); } /* A_STOP */ void do_stop(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_trace("Closing IPC server"); mainloop_del_ipc_server(ipcs); ipcs = NULL; register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL); } /* A_STARTED */ void do_started(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) { static struct qb_ipcs_service_handlers crmd_callbacks = { .connection_accept = accept_controller_client, .connection_created = NULL, .msg_process = dispatch_controller_ipc, .connection_closed = crmd_ipc_closed, .connection_destroyed = crmd_ipc_destroy }; if (cur_state != S_STARTING) { crm_err("Start cancelled... %s", fsa_state2string(cur_state)); return; } else if (!pcmk_is_set(fsa_input_register, R_MEMBERSHIP)) { crm_info("Delaying start, no membership data (%.16llx)", R_MEMBERSHIP); crmd_fsa_stall(TRUE); return; } else if (!pcmk_is_set(fsa_input_register, R_LRM_CONNECTED)) { crm_info("Delaying start, not connected to executor (%.16llx)", R_LRM_CONNECTED); crmd_fsa_stall(TRUE); return; } else if (!pcmk_is_set(fsa_input_register, R_CIB_CONNECTED)) { crm_info("Delaying start, CIB not connected (%.16llx)", R_CIB_CONNECTED); crmd_fsa_stall(TRUE); return; } else if (!pcmk_is_set(fsa_input_register, R_READ_CONFIG)) { crm_info("Delaying start, Config not read (%.16llx)", R_READ_CONFIG); crmd_fsa_stall(TRUE); return; } else if (!pcmk_is_set(fsa_input_register, R_PEER_DATA)) { crm_info("Delaying start, No peer data (%.16llx)", R_PEER_DATA); crmd_fsa_stall(TRUE); return; } crm_debug("Init server comms"); ipcs = pcmk__serve_controld_ipc(&crmd_callbacks); if (ipcs == NULL) { crm_err("Failed to create IPC server: shutting down and inhibiting respawn"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } else { crm_notice("Pacemaker controller successfully started and accepting connections"); } controld_trigger_fencer_connect(); controld_clear_fsa_input_flags(R_STARTING); register_fsa_input(msg_data->fsa_cause, I_PENDING, NULL); } /* A_RECOVER */ void do_recover(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) { controld_set_fsa_input_flags(R_IN_RECOVERY); crm_warn("Fast-tracking shutdown in response to errors"); register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL); } static pcmk__cluster_option_t crmd_opts[] = { /* name, old name, type, allowed values, * default value, validator, * short description, * long description */ { "dc-version", NULL, "string", NULL, "none", NULL, "Pacemaker version on cluster node elected Designated Controller (DC)", "Includes a hash which identifies the exact changeset the code was " "built from. Used for diagnostic purposes." }, { "cluster-infrastructure", NULL, "string", NULL, "corosync", NULL, "The messaging stack on which Pacemaker is currently running", "Used for informational and diagnostic purposes." }, { "cluster-name", NULL, "string", NULL, NULL, NULL, "An arbitrary name for the cluster", "This optional value is mostly for users' convenience as desired " "in administration, but may also be used in Pacemaker " "configuration rules via the #cluster-name node attribute, and " "by higher-level tools and resource agents." }, { XML_CONFIG_ATTR_DC_DEADTIME, NULL, "time", NULL, "20s", pcmk__valid_interval_spec, "How long to wait for a response from other nodes during start-up", "The optimal value will depend on the speed and load of your network " "and the type of switches used." }, { XML_CONFIG_ATTR_RECHECK, NULL, "time", N_("Zero disables polling, while positive values are an interval in seconds" "(unless other units are specified, for example \"5min\")"), "15min", pcmk__valid_interval_spec, "Polling interval to recheck cluster state and evaluate rules " "with date specifications", "Pacemaker is primarily event-driven, and looks ahead to know when to " "recheck cluster state for failure timeouts and most time-based " "rules. However, it will also recheck the cluster after this " "amount of inactivity, to evaluate rules with date specifications " "and serve as a fail-safe for certain types of scheduler bugs." }, { "load-threshold", NULL, "percentage", NULL, "80%", pcmk__valid_percentage, "Maximum amount of system load that should be used by cluster nodes", "The cluster will slow down its recovery process when the amount of " "system resources used (currently CPU) approaches this limit", }, { "node-action-limit", NULL, "integer", NULL, "0", pcmk__valid_number, "Maximum number of jobs that can be scheduled per node " "(defaults to 2x cores)" }, { XML_CONFIG_ATTR_FENCE_REACTION, NULL, "string", NULL, "stop", NULL, "How a cluster node should react if notified of its own fencing", "A cluster node may receive notification of its own fencing if fencing " "is misconfigured, or if fabric fencing is in use that doesn't cut " "cluster communication. Allowed values are \"stop\" to attempt to " "immediately stop Pacemaker and stay stopped, or \"panic\" to attempt " "to immediately reboot the local node, falling back to stop on failure." }, { XML_CONFIG_ATTR_ELECTION_FAIL, NULL, "time", NULL, "2min", pcmk__valid_interval_spec, "*** Advanced Use Only ***", "Declare an election failed if it is not decided within this much " "time. If you need to adjust this value, it probably indicates " "the presence of a bug." }, { XML_CONFIG_ATTR_FORCE_QUIT, NULL, "time", NULL, "20min", pcmk__valid_interval_spec, "*** Advanced Use Only ***", "Exit immediately if shutdown does not complete within this much " "time. If you need to adjust this value, it probably indicates " "the presence of a bug." }, { "join-integration-timeout", "crmd-integration-timeout", "time", NULL, "3min", pcmk__valid_interval_spec, "*** Advanced Use Only ***", "If you need to adjust this value, it probably indicates " "the presence of a bug." }, { "join-finalization-timeout", "crmd-finalization-timeout", "time", NULL, "30min", pcmk__valid_interval_spec, "*** Advanced Use Only ***", "If you need to adjust this value, it probably indicates " "the presence of a bug." }, { "transition-delay", "crmd-transition-delay", "time", NULL, "0s", pcmk__valid_interval_spec, "*** Advanced Use Only *** Enabling this option will slow down " "cluster recovery under all conditions", "Delay cluster recovery for this much time to allow for additional " "events to occur. Useful if your configuration is sensitive to " "the order in which ping updates arrive." }, { "stonith-watchdog-timeout", NULL, "time", NULL, "0", controld_verify_stonith_watchdog_timeout, "How long to wait before we can assume nodes are safely down " "when watchdog-based self-fencing via SBD is in use", "If nonzero, along with `have-watchdog=true` automatically set by the " "cluster, when fencing is required, watchdog-based self-fencing " "will be performed via SBD without requiring a fencing resource " "explicitly configured. " "If `stonith-watchdog-timeout` is set to a positive value, unseen " "nodes are assumed to self-fence within this much time. +WARNING:+ " "It must be ensured that this value is larger than the " "`SBD_WATCHDOG_TIMEOUT` environment variable on all nodes. " "Pacemaker verifies the settings individually on all nodes and " "prevents startup or shuts down if configured wrongly on the fly. " "It's strongly recommended that `SBD_WATCHDOG_TIMEOUT` is set to " "the same value on all nodes. " "If `stonith-watchdog-timeout` is set to a negative value, and " "`SBD_WATCHDOG_TIMEOUT` is set, twice that value will be used. " "+WARNING:+ In this case, it's essential (currently not verified by " "Pacemaker) that `SBD_WATCHDOG_TIMEOUT` is set to the same value on " "all nodes." }, { "stonith-max-attempts", NULL, "integer", NULL, "10", pcmk__valid_positive_number, "How many times fencing can fail before it will no longer be " "immediately re-attempted on a target" }, // Already documented in libpe_status (other values must be kept identical) { "no-quorum-policy", NULL, "select", "stop, freeze, ignore, demote, suicide", "stop", pcmk__valid_quorum, NULL, NULL }, { XML_CONFIG_ATTR_SHUTDOWN_LOCK, NULL, "boolean", NULL, "false", pcmk__valid_boolean, NULL, NULL }, }; void crmd_metadata(void) { char *s = pcmk__format_option_metadata("pacemaker-controld", "Pacemaker controller options", "Cluster options used by Pacemaker's " "controller", crmd_opts, PCMK__NELEM(crmd_opts)); printf("%s", s); free(s); } static void verify_crmd_options(GHashTable * options) { pcmk__validate_cluster_options(options, crmd_opts, PCMK__NELEM(crmd_opts)); } static const char * crmd_pref(GHashTable * options, const char *name) { return pcmk__cluster_option(options, crmd_opts, PCMK__NELEM(crmd_opts), name); } static void config_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { const char *value = NULL; GHashTable *config_hash = NULL; crm_time_t *now = crm_time_new(NULL); xmlNode *crmconfig = NULL; xmlNode *alerts = NULL; if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Local CIB query resulted in an error: %s", pcmk_strerror(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); if (rc == -EACCES || rc == -pcmk_err_schema_validation) { crm_err("The cluster is mis-configured - shutting down and staying down"); controld_set_fsa_input_flags(R_STAYDOWN); } goto bail; } crmconfig = output; if ((crmconfig) && (crm_element_name(crmconfig)) && (strcmp(crm_element_name(crmconfig), XML_CIB_TAG_CRMCONFIG) != 0)) { crmconfig = first_named_child(crmconfig, XML_CIB_TAG_CRMCONFIG); } if (!crmconfig) { fsa_data_t *msg_data = NULL; crm_err("Local CIB query for " XML_CIB_TAG_CRMCONFIG " section failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); goto bail; } crm_debug("Call %d : Parsing CIB options", call_id); config_hash = pcmk__strkey_table(free, free); pe_unpack_nvpairs(crmconfig, crmconfig, XML_CIB_TAG_PROPSET, NULL, config_hash, CIB_OPTIONS_FIRST, FALSE, now, NULL); verify_crmd_options(config_hash); value = crmd_pref(config_hash, XML_CONFIG_ATTR_DC_DEADTIME); election_trigger->period_ms = crm_parse_interval_spec(value); value = crmd_pref(config_hash, "node-action-limit"); /* Also checks migration-limit */ throttle_update_job_max(value); value = crmd_pref(config_hash, "load-threshold"); if(value) { throttle_set_load_target(strtof(value, NULL) / 100.0); } value = crmd_pref(config_hash, "no-quorum-policy"); if (pcmk__str_eq(value, "suicide", pcmk__str_casei) && pcmk__locate_sbd()) { no_quorum_suicide_escalation = TRUE; } set_fence_reaction(crmd_pref(config_hash, XML_CONFIG_ATTR_FENCE_REACTION)); value = crmd_pref(config_hash,"stonith-max-attempts"); update_stonith_max_attempts(value); value = crmd_pref(config_hash, XML_CONFIG_ATTR_FORCE_QUIT); shutdown_escalation_timer->period_ms = crm_parse_interval_spec(value); crm_debug("Shutdown escalation occurs if DC has not responded to request in %ums", shutdown_escalation_timer->period_ms); value = crmd_pref(config_hash, XML_CONFIG_ATTR_ELECTION_FAIL); controld_set_election_period(value); value = crmd_pref(config_hash, XML_CONFIG_ATTR_RECHECK); recheck_interval_ms = crm_parse_interval_spec(value); crm_debug("Re-run scheduler after %dms of inactivity", recheck_interval_ms); value = crmd_pref(config_hash, "transition-delay"); transition_timer->period_ms = crm_parse_interval_spec(value); value = crmd_pref(config_hash, "join-integration-timeout"); integration_timer->period_ms = crm_parse_interval_spec(value); value = crmd_pref(config_hash, "join-finalization-timeout"); finalization_timer->period_ms = crm_parse_interval_spec(value); value = crmd_pref(config_hash, XML_CONFIG_ATTR_SHUTDOWN_LOCK); controld_shutdown_lock_enabled = crm_is_true(value); free(fsa_cluster_name); fsa_cluster_name = NULL; value = g_hash_table_lookup(config_hash, "cluster-name"); if (value) { fsa_cluster_name = strdup(value); } alerts = first_named_child(output, XML_CIB_TAG_ALERTS); crmd_unpack_alerts(alerts); controld_set_fsa_input_flags(R_READ_CONFIG); crm_trace("Triggering FSA: %s", __func__); mainloop_set_trigger(fsa_source); g_hash_table_destroy(config_hash); bail: crm_time_free(now); } gboolean crm_read_options(gpointer user_data) { int call_id = fsa_cib_conn->cmds->query(fsa_cib_conn, "//" XML_CIB_TAG_CRMCONFIG " | //" XML_CIB_TAG_ALERTS, NULL, cib_xpath | cib_scope_local); fsa_register_cib_callback(call_id, FALSE, NULL, config_query_callback); crm_trace("Querying the CIB... call %d", call_id); return TRUE; } /* A_READCONFIG */ void do_read_config(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) { throttle_init(); mainloop_set_trigger(config_read); } void crm_shutdown(int nsig) { if ((crmd_mainloop == NULL) || !g_main_loop_is_running(crmd_mainloop)) { crmd_exit(CRM_EX_OK); return; } if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { crm_err("Escalating shutdown"); register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL); return; } controld_set_fsa_input_flags(R_SHUTDOWN); register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); if (shutdown_escalation_timer->period_ms == 0) { const char *value = crmd_pref(NULL, XML_CONFIG_ATTR_FORCE_QUIT); shutdown_escalation_timer->period_ms = crm_parse_interval_spec(value); } crm_notice("Initiating controller shutdown sequence " CRM_XS " limit=%ums", shutdown_escalation_timer->period_ms); controld_start_timer(shutdown_escalation_timer); } diff --git a/daemons/controld/controld_election.c b/daemons/controld/controld_election.c index 8c073c8a4f..4d73f3cc9b 100644 --- a/daemons/controld/controld_election.c +++ b/daemons/controld/controld_election.c @@ -1,275 +1,280 @@ /* - * Copyright 2004-2021 the Pacemaker project contributors + * 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 +extern pcmk__output_t *logger_out; + static election_t *fsa_election = NULL; static gboolean election_win_cb(gpointer data) { register_fsa_input(C_FSA_INTERNAL, I_ELECTION_DC, NULL); return FALSE; } void controld_election_init(const char *uname) { fsa_election = election_init("DC", uname, 60000 /*60s*/, election_win_cb); } void controld_remove_voter(const char *uname) { election_remove(fsa_election, uname); if (pcmk__str_eq(uname, fsa_our_dc, pcmk__str_casei)) { /* Clear any election dampening in effect. Otherwise, if the lost DC had * just won, an immediate new election could fizzle out with no new DC. */ election_clear_dampening(fsa_election); } } void controld_election_fini() { election_fini(fsa_election); fsa_election = NULL; } void controld_set_election_period(const char *value) { election_timeout_set_period(fsa_election, crm_parse_interval_spec(value)); } void controld_stop_election_timer() { election_timeout_stop(fsa_election); } /* A_ELECTION_VOTE */ void do_election_vote(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) { gboolean not_voting = FALSE; /* don't vote if we're in one of these states or wanting to shut down */ switch (cur_state) { case S_STARTING: case S_RECOVERY: case S_STOPPING: case S_TERMINATE: crm_warn("Not voting in election, we're in state %s", fsa_state2string(cur_state)); not_voting = TRUE; break; case S_ELECTION: case S_INTEGRATION: case S_RELEASE_DC: break; default: crm_err("Broken? Voting in state %s", fsa_state2string(cur_state)); break; } if (not_voting == FALSE) { if (pcmk_is_set(fsa_input_register, R_STARTING)) { not_voting = TRUE; } } if (not_voting) { if (AM_I_DC) { register_fsa_input(C_FSA_INTERNAL, I_RELEASE_DC, NULL); } else { register_fsa_input(C_FSA_INTERNAL, I_PENDING, NULL); } return; } election_vote(fsa_election); return; } void do_election_check(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) { if (fsa_state == S_ELECTION) { election_check(fsa_election); } else { crm_debug("Ignoring election check because we are not in an election"); } } /* A_ELECTION_COUNT */ void do_election_count_vote(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) { enum election_result rc = 0; ha_msg_input_t *vote = fsa_typed_data(fsa_dt_ha_msg); if(crm_peer_cache == NULL) { if (!pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { crm_err("Internal error, no peer cache"); } return; } rc = election_count_vote(fsa_election, vote->msg, cur_state != S_STARTING); switch(rc) { case election_start: election_reset(fsa_election); register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL); break; case election_lost: update_dc(NULL); if (fsa_input_register & R_THE_DC) { register_fsa_input(C_FSA_INTERNAL, I_RELEASE_DC, NULL); fsa_cib_conn->cmds->set_slave(fsa_cib_conn, cib_scope_local); } else if (cur_state != S_STARTING) { register_fsa_input(C_FSA_INTERNAL, I_PENDING, NULL); } break; default: crm_trace("Election message resulted in state %d", rc); } } static void feature_update_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_notice("Feature update failed: %s "CRM_XS" rc=%d", pcmk_strerror(rc), rc); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } /* A_DC_TAKEOVER */ void do_dc_takeover(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 rc = pcmk_ok; xmlNode *cib = NULL; const char *cluster_type = name_for_cluster_type(get_cluster_type()); pid_t watchdog = pcmk__locate_sbd(); crm_info("Taking over DC status for this partition"); controld_set_fsa_input_flags(R_THE_DC); execute_stonith_cleanup(); election_reset(fsa_election); controld_set_fsa_input_flags(R_JOIN_OK|R_INVOKE_PE); fsa_cib_conn->cmds->set_master(fsa_cib_conn, cib_scope_local); cib = create_xml_node(NULL, XML_TAG_CIB); crm_xml_add(cib, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); fsa_cib_update(XML_TAG_CIB, cib, cib_quorum_override, rc, NULL); fsa_register_cib_callback(rc, FALSE, NULL, feature_update_callback); - update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, - NULL, NULL, NULL, XML_ATTR_HAVE_WATCHDOG, - pcmk__btoa(watchdog), FALSE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, + NULL, NULL, NULL, NULL, XML_ATTR_HAVE_WATCHDOG, + pcmk__btoa(watchdog), NULL, NULL); - update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, - "dc-version", PACEMAKER_VERSION "-" BUILD_VERSION, FALSE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, + NULL, NULL, NULL, NULL, "dc-version", + PACEMAKER_VERSION "-" BUILD_VERSION, NULL, NULL); - update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, - "cluster-infrastructure", cluster_type, FALSE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, + NULL, NULL, NULL, NULL, "cluster-infrastructure", + cluster_type, NULL, NULL); #if SUPPORT_COROSYNC if (fsa_cluster_name == NULL && is_corosync_cluster()) { char *cluster_name = pcmk__corosync_cluster_name(); if (cluster_name) { - update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, - "cluster-name", cluster_name, FALSE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_none, + XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, + "cluster-name", cluster_name, NULL, NULL); } free(cluster_name); } #endif mainloop_set_trigger(config_read); free_xml(cib); } /* A_DC_RELEASE */ void do_dc_release(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) { if (action & A_DC_RELEASE) { crm_debug("Releasing the role of DC"); controld_clear_fsa_input_flags(R_THE_DC); controld_expect_sched_reply(NULL); } else if (action & A_DC_RELEASED) { crm_info("DC role released"); #if 0 if (are there errors) { /* we can't stay up if not healthy */ /* or perhaps I_ERROR and go to S_RECOVER? */ result = I_SHUTDOWN; } #endif if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { xmlNode *update = NULL; crm_node_t *node = crm_get_peer(0, fsa_our_uname); pcmk__update_peer_expected(__func__, node, CRMD_JOINSTATE_DOWN); update = create_node_state_update(node, node_update_expected, NULL, __func__); /* Don't need a based response because controld will stop. */ fsa_cib_anon_update_discard_reply(XML_CIB_TAG_STATUS, update); free_xml(update); } register_fsa_input(C_FSA_INTERNAL, I_RELEASE_SUCCESS, NULL); } else { crm_err("Unknown DC action %s", fsa_action2string(action)); } crm_trace("Am I still the DC? %s", AM_I_DC ? XML_BOOLEAN_YES : XML_BOOLEAN_NO); } diff --git a/daemons/controld/controld_execd.c b/daemons/controld/controld_execd.c index 57774b3899..fa411a6d9b 100644 --- a/daemons/controld/controld_execd.c +++ b/daemons/controld/controld_execd.c @@ -1,2924 +1,2925 @@ /* * 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; }; +extern pcmk__output_t *logger_out; + static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id); static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list); static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data); static lrmd_event_data_t *construct_op(lrm_state_t * lrm_state, xmlNode * rsc_op, const char *rsc_id, const char *operation); static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg); static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level); static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time); static void lrm_connection_destroy(void) { if (pcmk_is_set(fsa_input_register, R_LRM_CONNECTED)) { crm_crit("Connection to executor failed"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); controld_clear_fsa_input_flags(R_LRM_CONNECTED); } else { crm_info("Disconnected from executor"); } } static char * make_stop_id(const char *rsc, int call_id) { return crm_strdup_printf("%s:%d", rsc, call_id); } static void copy_instance_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") == NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } static void copy_meta_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") != NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } /*! * \internal * \brief Remove a recurring operation from a resource's history * * \param[in,out] history Resource history to modify * \param[in] op Operation to remove * * \return TRUE if the operation was found and removed, FALSE otherwise */ static gboolean history_remove_recurring_op(rsc_history_t *history, const lrmd_event_data_t *op) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_event_data_t *existing = iter->data; if ((op->interval_ms == existing->interval_ms) && pcmk__str_eq(op->rsc_id, existing->rsc_id, pcmk__str_none) && pcmk__str_eq(op->op_type, existing->op_type, pcmk__str_casei)) { history->recurring_op_list = g_list_delete_link(history->recurring_op_list, iter); lrmd_free_event(existing); return TRUE; } } return FALSE; } /*! * \internal * \brief Free all recurring operations in resource history * * \param[in,out] history Resource history to modify */ static void history_free_recurring_ops(rsc_history_t *history) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_free_event(iter->data); } g_list_free(history->recurring_op_list); history->recurring_op_list = NULL; } /*! * \internal * \brief Free resource history * * \param[in,out] history Resource history to free */ void history_free(gpointer data) { rsc_history_t *history = (rsc_history_t*)data; if (history->stop_params) { g_hash_table_destroy(history->stop_params); } /* Don't need to free history->rsc.id because it's set to history->id */ free(history->rsc.type); free(history->rsc.standard); free(history->rsc.provider); lrmd_free_event(history->failed); lrmd_free_event(history->last); free(history->id); history_free_recurring_ops(history); free(history); } static void update_history_cache(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { int target_rc = 0; rsc_history_t *entry = NULL; if (op->rsc_deleted) { crm_debug("Purged history for '%s' after %s", op->rsc_id, op->op_type); controld_delete_resource_history(op->rsc_id, lrm_state->node_name, NULL, crmd_cib_smart_opt()); return; } if (pcmk__str_eq(op->op_type, RSC_NOTIFY, pcmk__str_casei)) { return; } crm_debug("Updating history for '%s' with %s op", op->rsc_id, op->op_type); entry = g_hash_table_lookup(lrm_state->resource_history, op->rsc_id); if (entry == NULL && rsc) { entry = calloc(1, sizeof(rsc_history_t)); entry->id = strdup(op->rsc_id); g_hash_table_insert(lrm_state->resource_history, entry->id, entry); entry->rsc.id = entry->id; entry->rsc.type = strdup(rsc->type); entry->rsc.standard = strdup(rsc->standard); 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(lrm_state_t *lrm_state, ha_msg_input_t *input, const char *rsc_id, lrmd_rsc_info_t *rsc, const char *task, const char *ack_host, const char *ack_sys) { lrmd_event_data_t *op = construct_op(lrm_state, input->xml, rsc_id, task); lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(ack_host, ack_sys, rsc, op, rsc_id); lrmd_free_event(op); } static inline const char * op_node_name(lrmd_event_data_t *op) { return op->remote_nodename? op->remote_nodename : fsa_our_uname; } void lrm_op_callback(lrmd_event_data_t * op) { CRM_CHECK(op != NULL, return); switch (op->type) { case lrmd_event_disconnect: if (op->remote_nodename == NULL) { /* If this is the local executor IPC connection, set the right * bits in the controller when the connection goes down. */ lrm_connection_destroy(); } break; case lrmd_event_exec_complete: { lrm_state_t *lrm_state = lrm_state_find(op_node_name(op)); CRM_ASSERT(lrm_state != NULL); process_lrm_event(lrm_state, op, NULL, NULL); } break; default: break; } } static void try_local_executor_connect(long long action, fsa_data_t *msg_data, lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; crm_debug("Connecting to the local executor"); // If we can connect, great rc = controld_connect_local_executor(lrm_state); if (rc == pcmk_rc_ok) { controld_set_fsa_input_flags(R_LRM_CONNECTED); crm_info("Connection to the local executor established"); return; } // Otherwise, if we can try again, set a timer to do so if (lrm_state->num_lrm_register_fails < MAX_LRM_REG_FAILS) { crm_warn("Failed to connect to the local executor %d time%s " "(%d max): %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), MAX_LRM_REG_FAILS, pcmk_rc_str(rc)); controld_start_timer(wait_timer); crmd_fsa_stall(FALSE); return; } // Otherwise give up crm_err("Failed to connect to the executor the max allowed " "%d time%s: %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), pcmk_rc_str(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } /* A_LRM_CONNECT */ void do_lrm_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { /* This only pertains to local executor connections. Remote connections are * handled as resources within the scheduler. Connecting and disconnecting * from remote executor instances is handled differently. */ lrm_state_t *lrm_state = NULL; if(fsa_our_uname == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(fsa_our_uname); if (lrm_state == NULL) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (action & A_LRM_DISCONNECT) { if (lrm_state_verify_stopped(lrm_state, cur_state, LOG_INFO) == FALSE) { if (action == A_LRM_DISCONNECT) { crmd_fsa_stall(FALSE); return; } } controld_clear_fsa_input_flags(R_LRM_CONNECTED); crm_info("Disconnecting from the executor"); lrm_state_disconnect(lrm_state); lrm_state_reset_tables(lrm_state, FALSE); crm_notice("Disconnected from the executor"); } if (action & A_LRM_CONNECT) { try_local_executor_connect(action, msg_data, lrm_state); } if (action & ~(A_LRM_CONNECT | A_LRM_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __func__); } } static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level) { int counter = 0; gboolean rc = TRUE; const char *when = "lrm disconnect"; GHashTableIter gIter; const char *key = NULL; rsc_history_t *entry = NULL; active_op_t *pending = NULL; crm_debug("Checking for active resources before exit"); if (cur_state == S_TERMINATE) { log_level = LOG_ERR; when = "shutdown"; } else if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { when = "shutdown... waiting"; } if (lrm_state->pending_ops && lrm_state_is_connected(lrm_state) == TRUE) { guint removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_actions, lrm_state); guint nremaining = g_hash_table_size(lrm_state->pending_ops); if (removed || nremaining) { crm_notice("Stopped %u recurring operation%s at %s (%u remaining)", removed, pcmk__plural_s(removed), when, nremaining); } } if (lrm_state->pending_ops) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, NULL, (void **)&pending)) { /* Ignore recurring actions in the shutdown calculations */ if (pending->interval_ms == 0) { counter++; } } } if (counter > 0) { do_crm_log(log_level, "%d pending executor operation%s at %s", counter, pcmk__plural_s(counter), when); if ((cur_state == S_TERMINATE) || !pcmk_is_set(fsa_input_register, R_SENT_RSC_STOP)) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, (gpointer*)&key, (gpointer*)&pending)) { do_crm_log(log_level, "Pending action: %s (%s)", key, pending->op_key); } } else { rc = FALSE; } return rc; } if (lrm_state->resource_history == NULL) { return rc; } if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { /* At this point we're not waiting, we're just shutting down */ when = "shutdown"; } counter = 0; g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (gpointer*)&entry)) { if (is_rsc_active(lrm_state, entry->id) == FALSE) { continue; } counter++; if (log_level == LOG_ERR) { crm_info("Found %s active at %s", entry->id, when); } else { crm_trace("Found %s active at %s", entry->id, when); } if (lrm_state->pending_ops) { GHashTableIter hIter; g_hash_table_iter_init(&hIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&hIter, (gpointer*)&key, (gpointer*)&pending)) { if (pcmk__str_eq(entry->id, pending->rsc_id, pcmk__str_none)) { crm_notice("%sction %s (%s) incomplete at %s", pending->interval_ms == 0 ? "A" : "Recurring a", key, pending->op_key, when); } } } } if (counter) { crm_err("%d resource%s active at %s", counter, (counter == 1)? " was" : "s were", when); } return rc; } /*! * \internal * \brief Build XML and string of parameters meeting some criteria, for digest * * \param[in] op Executor event with parameter table to use * \param[in] metadata Parsed meta-data for executed resource agent * \param[in] param_type Flag used for selection criteria * \param[out] result Will be set to newly created XML with selected * parameters as attributes * * \return Newly allocated space-separated string of parameter names * \note Selection criteria varies by param_type: for the restart digest, we * want parameters that are *not* marked reloadable (OCF 1.1) or that * *are* marked unique (pre-1.1), for both string and XML results; for the * secure digest, we want parameters that *are* marked private for the * string, but parameters that are *not* marked private for the XML. * \note It is the caller's responsibility to free the string return value with * free() and the XML result with free_xml(). */ static char * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, enum ra_param_flags_e param_type, xmlNode **result) { char *list = NULL; size_t len = 0; *result = create_xml_node(NULL, XML_TAG_PARAMS); for (GList *iter = metadata->ra_params; iter != NULL; iter = iter->next) { struct ra_param_s *param = (struct ra_param_s *) iter->data; bool accept_for_list = false; bool accept_for_xml = false; switch (param_type) { case ra_param_reloadable: accept_for_list = !pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_unique: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_private: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = !accept_for_list; break; } if (accept_for_list) { crm_trace("Attr %s is %s", param->rap_name, ra_param_flag2text(param_type)); if (list == NULL) { // We will later search for " WORD ", so start list with a space pcmk__add_word(&list, &len, " "); } pcmk__add_word(&list, &len, param->rap_name); } else { crm_trace("Rejecting %s for %s", param->rap_name, ra_param_flag2text(param_type)); } if (accept_for_xml) { const char *v = g_hash_table_lookup(op->params, param->rap_name); if (v != NULL) { crm_trace("Adding attr %s=%s to the xml result", param->rap_name, v); crm_xml_add(*result, param->rap_name, v); } } } if (list != NULL) { // We will later search for " WORD ", so end list with a space pcmk__add_word(&list, &len, " "); } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *restart = NULL; CRM_LOG_ASSERT(op->params != NULL); if (op->interval_ms > 0) { /* monitors are not reloadable */ return; } if (pcmk_is_set(metadata->ra_flags, ra_supports_reload_agent)) { // Add parameters not marked reloadable to the "op-force-restart" list list = build_parameter_list(op, metadata, ra_param_reloadable, &restart); } else if (pcmk_is_set(metadata->ra_flags, ra_supports_legacy_reload)) { /* @COMPAT pre-OCF-1.1 resource agents * * Before OCF 1.1, Pacemaker abused "unique=0" to indicate * reloadability. Add any parameters with unique="1" to the * "op-force-restart" list. */ list = build_parameter_list(op, metadata, ra_param_unique, &restart); } else { // Resource does not support agent reloads return; } digest = calculate_operation_digest(restart, version); /* Add "op-force-restart" and "op-restart-digest" to indicate the resource supports reload, * no matter if it actually supports any parameters with unique="1"). */ crm_xml_add(update, XML_LRM_ATTR_OP_RESTART, list? list: ""); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(restart, "restart digest source"); free_xml(restart); free(digest); free(list); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *secure = NULL; CRM_LOG_ASSERT(op->params != NULL); /* * To keep XML_LRM_ATTR_OP_SECURE short, we want it to contain the * secure parameters but XML_LRM_ATTR_SECURE_DIGEST to be based on * the insecure ones */ list = build_parameter_list(op, metadata, ra_param_private, &secure); if (list != NULL) { digest = calculate_operation_digest(secure, version); crm_xml_add(update, XML_LRM_ATTR_OP_SECURE, list); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(secure, "secure digest source"); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); free(list); } static gboolean build_operation_update(xmlNode * parent, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *node_name, const char *src) { int target_rc = 0; xmlNode *xml_op = NULL; struct ra_metadata_s *metadata = NULL; const char *caller_version = NULL; lrm_state_t *lrm_state = NULL; uint32_t metadata_source = controld_metadata_from_agent; if (op == NULL) { return FALSE; } target_rc = rsc_op_expected_rc(op); /* there is a small risk in formerly mixed clusters that it will * be sub-optimal. * * however with our upgrade policy, the update we send should * still be completely supported anyway */ caller_version = g_hash_table_lookup(op->params, XML_ATTR_CRM_VERSION); CRM_LOG_ASSERT(caller_version != NULL); if(caller_version == NULL) { caller_version = CRM_FEATURE_SET; } xml_op = pcmk__create_history_xml(parent, op, caller_version, target_rc, fsa_our_uname, 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; } /* Getting meta-data from cache is OK unless this is a successful start * action -- always refresh from the agent for those, in case the * resource agent was updated. * * @TODO Only refresh the meta-data after starts if the agent actually * changed (using something like inotify, or a hash or modification time of * the agent executable). */ if ((op->op_status != PCMK_EXEC_DONE) || (op->rc != target_rc) || !pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { metadata_source |= controld_metadata_from_cache; } metadata = controld_get_rsc_metadata(lrm_state, rsc, metadata_source); if (metadata == NULL) { return TRUE; } #if ENABLE_VERSIONED_ATTRS crm_xml_add(xml_op, XML_ATTR_RA_VERSION, metadata->ra_version); #endif crm_trace("Including additional digests for %s:%s:%s", rsc->standard, rsc->provider, rsc->type); append_restart_list(op, metadata, xml_op, caller_version); append_secure_list(op, metadata, xml_op, caller_version); return TRUE; } static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id) { rsc_history_t *entry = NULL; entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->last == NULL) { return FALSE; } crm_trace("Processing %s: %s.%d=%d", rsc_id, entry->last->op_type, entry->last->interval_ms, entry->last->rc); if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_STOP, pcmk__str_casei)) { return FALSE; } else if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { // A stricter check is too complex ... leave that to the scheduler return FALSE; } else if (entry->last->rc == PCMK_OCF_NOT_RUNNING) { return FALSE; } else if ((entry->last->interval_ms == 0) && (entry->last->rc == PCMK_OCF_NOT_CONFIGURED)) { /* Badly configured resources can't be reliably stopped */ return FALSE; } return TRUE; } static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list) { GHashTableIter iter; rsc_history_t *entry = NULL; g_hash_table_iter_init(&iter, lrm_state->resource_history); while (g_hash_table_iter_next(&iter, NULL, (void **)&entry)) { GList *gIter = NULL; xmlNode *xml_rsc = create_xml_node(rsc_list, XML_LRM_TAG_RESOURCE); crm_xml_add(xml_rsc, XML_ATTR_ID, entry->id); crm_xml_add(xml_rsc, XML_ATTR_TYPE, entry->rsc.type); crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, entry->rsc.standard); crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, entry->rsc.provider); if (entry->last && entry->last->params) { const char *container = g_hash_table_lookup(entry->last->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); if (container) { crm_trace("Resource %s is a part of container resource %s", entry->id, container); crm_xml_add(xml_rsc, XML_RSC_ATTR_CONTAINER, container); } } build_operation_update(xml_rsc, &(entry->rsc), entry->failed, lrm_state->node_name, __func__); build_operation_update(xml_rsc, &(entry->rsc), entry->last, lrm_state->node_name, __func__); for (gIter = entry->recurring_op_list; gIter != NULL; gIter = gIter->next) { build_operation_update(xml_rsc, &(entry->rsc), gIter->data, lrm_state->node_name, __func__); } } return FALSE; } static xmlNode * do_lrm_query_internal(lrm_state_t *lrm_state, int update_flags) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; peer = crm_get_peer_full(0, lrm_state->node_name, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return NULL); xml_state = create_node_state_update(peer, update_flags, NULL, __func__); if (xml_state == NULL) { return NULL; } xml_data = create_xml_node(xml_state, XML_CIB_TAG_LRM); crm_xml_add(xml_data, XML_ATTR_ID, peer->uuid); rsc_list = create_xml_node(xml_data, XML_LRM_TAG_RESOURCES); /* Build a list of active (not always running) resources */ build_active_RAs(lrm_state, rsc_list); crm_log_xml_trace(xml_state, "Current executor state"); return xml_state; } xmlNode * controld_query_executor_state(const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (!lrm_state) { crm_err("Could not find executor state for node %s", node_name); return NULL; } return do_lrm_query_internal(lrm_state, node_update_cluster|node_update_peer); } /*! * \internal * \brief Map standard Pacemaker return code to operation status and OCF code * * \param[out] event Executor event whose status and return code should be set * \param[in] rc Standard Pacemaker return code */ void controld_rc2event(lrmd_event_data_t *event, int rc) { /* This is called for cleanup requests from controller peers/clients, not * for resource actions, so no exit reason is needed. */ switch (rc) { case pcmk_rc_ok: 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); - update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, - NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, - FALSE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_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_gIter, int rc, const char *user_name) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_gIter) { g_hash_table_iter_remove(rsc_gIter); } else { g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); } controld_delete_resource_history(rsc_id_copy, lrm_state->node_name, user_name, crmd_cib_smart_opt()); g_hash_table_foreach_remove(lrm_state->pending_ops, lrm_remove_deleted_op, rsc_id_copy); free(rsc_id_copy); } if (input) { notify_deleted(lrm_state, input, rsc_id, rc); } event.rc = rc; event.rsc = rsc_id; event.lrm_state = lrm_state; g_hash_table_foreach_remove(lrm_state->deletion_ops, lrm_remove_deleted_rsc, &event); } /*! * \internal * \brief Erase an LRM history entry from the CIB, given the operation data * * \param[in] lrm_state LRM state of the desired node * \param[in] op Operation whose history should be deleted */ static void erase_lrm_history_by_op(lrm_state_t *lrm_state, lrmd_event_data_t *op) { xmlNode *xml_top = NULL; CRM_CHECK(op != NULL, return); xml_top = create_xml_node(NULL, XML_LRM_TAG_RSC_OP); crm_xml_add_int(xml_top, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add(xml_top, XML_ATTR_TRANSITION_KEY, op->user_data); if (op->interval_ms > 0) { char *op_id = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); /* Avoid deleting last_failure too (if it was a result of this recurring op failing) */ crm_xml_add(xml_top, XML_ATTR_ID, op_id); free(op_id); } crm_debug("Erasing resource operation history for " PCMK__OP_FMT " (call=%d)", op->rsc_id, op->op_type, op->interval_ms, op->call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, XML_CIB_TAG_STATUS, xml_top, cib_quorum_override); crm_log_xml_trace(xml_top, "op:cancel"); free_xml(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" XML_ATTR_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP /* ... and also by operation key */ #define XPATH_HISTORY_ID XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s']" /* ... and also by operation key and operation call ID */ #define XPATH_HISTORY_CALL XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_CALLID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_TASK_KEY "='%s']" /*! * \internal * \brief Erase an LRM history entry from the CIB, given operation identifiers * * \param[in] lrm_state LRM state of the node to clear history for * \param[in] rsc_id Name of resource to clear history for * \param[in] key Operation key of operation to clear history for * \param[in] orig_op If specified, delete only if it has this original op * \param[in] call_id If specified, delete entry only if it has this call ID */ static void erase_lrm_history_by_id(lrm_state_t *lrm_state, const char *rsc_id, const char *key, const char *orig_op, int call_id) { char *op_xpath = NULL; CRM_CHECK((rsc_id != NULL) && (key != NULL), return); if (call_id > 0) { op_xpath = crm_strdup_printf(XPATH_HISTORY_CALL, lrm_state->node_name, rsc_id, key, call_id); } else if (orig_op) { op_xpath = crm_strdup_printf(XPATH_HISTORY_ORIG, lrm_state->node_name, rsc_id, key, orig_op); } else { op_xpath = crm_strdup_printf(XPATH_HISTORY_ID, lrm_state->node_name, rsc_id, key); } crm_debug("Erasing resource operation history for %s on %s (call=%d)", key, rsc_id, call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, op_xpath, NULL, cib_quorum_override | cib_xpath); free(op_xpath); } static inline gboolean last_failed_matches_op(rsc_history_t *entry, const char *op, guint interval_ms) { if (entry == NULL) { return FALSE; } if (op == NULL) { return TRUE; } return (pcmk__str_eq(op, entry->failed->op_type, pcmk__str_casei) && (interval_ms == entry->failed->interval_ms)); } /*! * \internal * \brief Clear a resource's last failure * * Erase a resource's last failure on a particular node from both the * LRM resource history in the CIB, and the resource history remembered * for the LRM state. * * \param[in] rsc_id Resource name * \param[in] node_name Node name * \param[in] operation If specified, only clear if matching this operation * \param[in] interval_ms If operation is specified, it has this interval */ void lrm_clear_last_failure(const char *rsc_id, const char *node_name, const char *operation, guint interval_ms) { char *op_key = NULL; char *orig_op_key = NULL; lrm_state_t *lrm_state = NULL; lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { return; } /* Erase from CIB */ op_key = pcmk__op_key(rsc_id, "last_failure", 0); if (operation) { orig_op_key = pcmk__op_key(rsc_id, operation, interval_ms); } erase_lrm_history_by_id(lrm_state, rsc_id, op_key, orig_op_key, 0); free(op_key); free(orig_op_key); /* Remove from memory */ if (lrm_state->resource_history) { rsc_history_t *entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (last_failed_matches_op(entry, operation, interval_ms)) { lrmd_free_event(entry->failed); entry->failed = NULL; } } } /* Returns: gboolean - cancellation is in progress */ static gboolean cancel_op(lrm_state_t * lrm_state, const char *rsc_id, const char *key, int op, gboolean remove) { int rc = pcmk_ok; char *local_key = NULL; active_op_t *pending = NULL; CRM_CHECK(op != 0, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); if (key == NULL) { local_key = make_stop_id(rsc_id, op); key = local_key; } pending = g_hash_table_lookup(lrm_state->pending_ops, key); if (pending) { if (remove && !pcmk_is_set(pending->flags, active_op_remove)) { controld_set_active_op_flags(pending, active_op_remove); crm_debug("Scheduling %s for removal", key); } if (pcmk_is_set(pending->flags, active_op_cancelled)) { crm_debug("Operation %s already cancelled", key); free(local_key); return FALSE; } controld_set_active_op_flags(pending, active_op_cancelled); } else { crm_info("No pending op found for %s", key); free(local_key); return FALSE; } crm_debug("Cancelling op %d for %s (%s)", op, rsc_id, key); rc = lrm_state_cancel(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms); if (rc == pcmk_ok) { crm_debug("Op %d for %s (%s): cancelled", op, rsc_id, key); free(local_key); return TRUE; } crm_debug("Op %d for %s (%s): Nothing to cancel", op, rsc_id, key); /* The caller needs to make sure the entry is * removed from the pending_ops list * * Usually by returning TRUE inside the worker function * supplied to g_hash_table_foreach_remove() * * Not removing the entry from pending_ops will block * the node from shutting down */ free(local_key); return FALSE; } struct cancel_data { gboolean done; gboolean remove; const char *key; lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean cancel_action_by_key(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct cancel_data *data = user_data; active_op_t *op = value; if (pcmk__str_eq(op->op_key, data->key, pcmk__str_none)) { data->done = TRUE; remove = !cancel_op(data->lrm_state, data->rsc->id, key, op->call_id, data->remove); } return remove; } static gboolean cancel_op_key(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *key, gboolean remove) { guint removed = 0; struct cancel_data data; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(key != NULL, return FALSE); data.key = key; data.rsc = rsc; data.done = FALSE; data.remove = remove; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->pending_ops, cancel_action_by_key, &data); crm_trace("Removed %u op cache entries, new size: %u", removed, g_hash_table_size(lrm_state->pending_ops)); return data.done; } /*! * \internal * \brief Retrieve resource information from LRM * * \param[in] lrm_state LRM connection to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource with LRM if not already * \param[out] rsc_info Where to store resource information obtained from LRM * * \retval pcmk_ok Success (and rsc_info holds newly allocated result) * \retval -EINVAL Required information is missing from arguments * \retval -ENOTCONN No active connection to LRM * \retval -ENODEV Resource not found * \retval -errno Error communicating with executor when registering resource * * \note Caller is responsible for freeing result on success. */ static int get_lrm_resource(lrm_state_t *lrm_state, xmlNode *rsc_xml, gboolean do_create, lrmd_rsc_info_t **rsc_info) { const char *id = ID(rsc_xml); CRM_CHECK(lrm_state && rsc_xml && rsc_info, return -EINVAL); CRM_CHECK(id, return -EINVAL); if (lrm_state_is_connected(lrm_state) == FALSE) { return -ENOTCONN; } crm_trace("Retrieving resource information for %s from the executor", id); *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); // If resource isn't known by ID, try clone name, if provided if (!*rsc_info) { const char *long_id = crm_element_value(rsc_xml, XML_ATTR_ID_LONG); if (long_id) { *rsc_info = lrm_state_get_rsc_info(lrm_state, long_id, 0); } } if ((*rsc_info == NULL) && do_create) { const char *class = crm_element_value(rsc_xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(rsc_xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(rsc_xml, XML_ATTR_TYPE); int rc; crm_trace("Registering resource %s with the executor", id); rc = lrm_state_register_rsc(lrm_state, id, class, provider, type, lrmd_opt_drop_recurring); if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Could not register resource %s with the executor on %s: %s " CRM_XS " rc=%d", id, lrm_state->node_name, pcmk_strerror(rc), rc); /* Register this as an internal error if this involves the local * executor. Otherwise, we're likely dealing with an unresponsive * remote node, which is not an FSA failure. */ if (lrm_state_is_local(lrm_state) == TRUE) { register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } return rc; } *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); } return *rsc_info? pcmk_ok : -ENODEV; } static void delete_resource(lrm_state_t * lrm_state, const char *id, lrmd_rsc_info_t * rsc, GHashTableIter * gIter, const char *sys, const char *user, ha_msg_input_t * request, gboolean unregister) { int rc = pcmk_ok; crm_info("Removing resource %s from executor for %s%s%s", id, sys, (user? " as " : ""), (user? user : "")); if (rsc && unregister) { rc = lrm_state_unregister_rsc(lrm_state, id, 0); } if (rc == pcmk_ok) { crm_trace("Resource %s deleted from executor", id); } else if (rc == -EINPROGRESS) { crm_info("Deletion of resource '%s' from executor is pending", id); if (request) { struct pending_deletion_op_s *op = NULL; char *ref = crm_element_value_copy(request->msg, XML_ATTR_REFERENCE); op = calloc(1, sizeof(struct pending_deletion_op_s)); op->rsc = strdup(rsc->id); op->input = copy_ha_msg_input(request); g_hash_table_insert(lrm_state->deletion_ops, ref, op); } return; } else { crm_warn("Could not delete '%s' from executor for %s%s%s: %s " CRM_XS " rc=%d", id, sys, (user? " as " : ""), (user? user : ""), pcmk_strerror(rc), rc); } delete_rsc_entry(lrm_state, request, id, gIter, rc, user); } static int get_fake_call_id(lrm_state_t *lrm_state, const char *rsc_id) { int call_id = 999999999; rsc_history_t *entry = NULL; if(lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* Make sure the call id is greater than the last successful operation, * otherwise the failure will not result in a possible recovery of the resource * as it could appear the failure occurred before the successful start */ if (entry) { call_id = entry->last_callid + 1; } if (call_id < 0) { call_id = 1; } return call_id; } static void fake_op_status(lrm_state_t *lrm_state, lrmd_event_data_t *op, int op_status, enum ocf_exitcode op_exitcode, 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) { GHashTableIter gIter; rsc_history_t *entry = NULL; crm_info("Clearing resource history on node %s", lrm_state->node_name); g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { /* only unregister the resource during a reprobe if it is not a remote connection * resource. otherwise unregistering the connection will terminate remote-node * membership */ gboolean unregister = TRUE; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state) { /* when forcing a reprobe, make sure to clear remote node before * clearing the remote node's connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE); } unregister = FALSE; } delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, user_name, NULL, unregister); } /* Now delete the copy in the CIB */ controld_delete_node_state(lrm_state->node_name, controld_section_lrm, cib_scope_local); // @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] lrm_state Executor connection that action is for * \param[in] action Action XML from request * \param[in] rc Desired return code to use * \param[in] op_status Desired operation status to use * \param[in] exit_reason Human-friendly detail, if error */ static void synthesize_lrmd_failure(lrm_state_t *lrm_state, xmlNode *action, int op_status, enum ocf_exitcode rc, 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 * * \param[in] xml LRM operation data XML * * \return LRM operation target node name (local node or Pacemaker Remote node) */ static const char * lrm_op_target(xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = fsa_our_uname; } return target; } static void fail_lrm_resource(xmlNode *xml, lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { lrmd_event_data_t *op = NULL; lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(xml, XML_CIB_TAG_RESOURCE, TRUE); CRM_CHECK(xml_rsc != NULL, return); /* The executor simply executes operations and reports the results, without * any concept of success or failure, so to fail a resource, we must fake * what a failure looks like. * * To do this, we create a fake executor operation event for the resource, * and pass that event to the executor client callback so it will be * processed as if it came from the executor. */ op = construct_op(lrm_state, xml, ID(xml_rsc), "asyncmon"); 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_refresh_op(lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { int rc = pcmk_ok; xmlNode *fragment = do_lrm_query_internal(lrm_state, node_update_all); fsa_cib_update(XML_CIB_TAG_STATUS, fragment, cib_quorum_override, rc, user_name); crm_info("Forced a local resource history refresh: call=%d", rc); if (!pcmk__str_eq(CRM_SYSTEM_CRMD, from_sys, pcmk__str_casei)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, fragment, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing refresh from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } free_xml(fragment); } static void handle_query_op(xmlNode *msg, lrm_state_t *lrm_state) { xmlNode *data = do_lrm_query_internal(lrm_state, node_update_all); xmlNode *reply = create_reply(msg, data); if (relay_message(reply, TRUE) == FALSE) { crm_err("Unable to route reply"); crm_log_xml_err(reply, "reply"); } free_xml(reply); free_xml(data); } static void handle_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node); if (!pcmk__strcase_any_of(from_sys, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, NULL, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing re-probe from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } } static bool do_lrm_cancel(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_host, const char *from_sys) { char *op_key = NULL; char *meta_key = NULL; int call = 0; const char *call_id = NULL; const char *op_task = NULL; guint interval_ms = 0; gboolean in_progress = FALSE; xmlNode *params = find_xml_node(input->xml, XML_TAG_ATTRS, TRUE); CRM_CHECK(params != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_TASK); op_task = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(op_task != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); if (crm_element_value_ms(params, meta_key, &interval_ms) != pcmk_ok) { free(meta_key); return FALSE; } free(meta_key); op_key = pcmk__op_key(rsc->id, op_task, interval_ms); meta_key = crm_meta_name(XML_LRM_ATTR_CALLID); call_id = crm_element_value(params, meta_key); free(meta_key); crm_debug("Scheduler requested op %s (call=%s) be cancelled", op_key, (call_id? call_id : "NA")); pcmk__scan_min_int(call_id, &call, 0); if (call == 0) { // Normal case when the scheduler cancels a recurring op in_progress = cancel_op_key(lrm_state, rsc, op_key, TRUE); } else { // Normal case when the scheduler cancels an orphan op in_progress = cancel_op(lrm_state, rsc->id, NULL, call, TRUE); } // Acknowledge cancellation operation if for a remote connection resource if (!in_progress || is_remote_lrmd_ra(NULL, NULL, rsc->id)) { char *op_id = make_stop_id(rsc->id, call); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) == FALSE) { crm_info("Nothing known about operation %d for %s", call, op_key); } erase_lrm_history_by_id(lrm_state, rsc->id, op_key, NULL, call); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); /* needed at least for cancellation of a remote operation */ g_hash_table_remove(lrm_state->pending_ops, op_id); free(op_id); } else { /* No ack is needed since abcdaa8, but peers with older versions * in a rolling upgrade need one. We didn't bump the feature set * at that commit, so we can only compare against the previous * CRM version (3.0.8). If any peers have feature set 3.0.9 but * not abcdaa8, they will time out waiting for the ack (no * released versions of Pacemaker are affected). */ const char *peer_version = crm_element_value(params, XML_ATTR_CRM_VERSION); if (compare_version(peer_version, "3.0.8") <= 0) { crm_info("Sending compatibility ack for %s cancellation to %s (CRM version %s)", op_key, from_host, peer_version); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); } } free(op_key); return TRUE; } static void do_lrm_delete(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_sys, const char *from_host, bool crm_rsc_delete, const char *user_name) { gboolean unregister = TRUE; int cib_rc = controld_delete_resource_history(rsc->id, lrm_state->node_name, user_name, cib_dryrun|cib_sync_call); if (cib_rc != pcmk_rc_ok) { lrmd_event_data_t *op = NULL; op = construct_op(lrm_state, input->xml, rsc->id, CRMD_ACTION_DELETE); /* These are resource clean-ups, not actions, so no exit reason is * needed. */ lrmd__set_result(op, pcmk_rc2ocf(cib_rc), PCMK_EXEC_ERROR, NULL); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc->id); lrmd_free_event(op); return; } if (crm_rsc_delete && is_remote_lrmd_ra(NULL, NULL, rsc->id)) { unregister = FALSE; } delete_resource(lrm_state, rsc->id, rsc, NULL, from_sys, user_name, input, unregister); } /* A_LRM_INVOKE */ void do_lrm_invoke(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { lrm_state_t *lrm_state = NULL; const char *crm_op = NULL; const char *from_sys = NULL; const char *from_host = NULL; const char *operation = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *user_name = NULL; const char *target_node = NULL; gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; target_node = lrm_op_target(input->xml); is_remote_node = !pcmk__str_eq(target_node, fsa_our_uname, pcmk__str_casei); lrm_state = lrm_state_find(target_node); if ((lrm_state == NULL) && is_remote_node) { crm_err("Failing action because local node has never had connection to remote node %s", target_node); synthesize_lrmd_failure(NULL, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Local node has no connection to remote"); return; } CRM_ASSERT(lrm_state != NULL); user_name = pcmk__update_acl_user(input->msg, F_CRM_USER, NULL); crm_op = crm_element_value(input->msg, F_CRM_TASK); from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } crm_trace("Executor %s command from %s as user %s", crm_op, from_sys, user_name); if (pcmk__str_eq(crm_op, CRM_OP_LRM_DELETE, pcmk__str_casei)) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { crm_rsc_delete = TRUE; // from crm_resource } operation = CRMD_ACTION_DELETE; } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_FAIL, pcmk__str_casei)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); return; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } if (pcmk__str_eq(crm_op, CRM_OP_LRM_REFRESH, pcmk__str_casei)) { handle_refresh_op(lrm_state, user_name, from_host, from_sys); } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_QUERY, pcmk__str_casei)) { handle_query_op(input->msg, lrm_state); // @COMPAT DCs <1.1.14 in a rolling upgrade might schedule this op } else if (pcmk__str_eq(operation, CRM_OP_PROBED, pcmk__str_casei)) { update_attrd(lrm_state->node_name, CRM_OP_PROBED, XML_BOOLEAN_TRUE, user_name, is_remote_node); } else if (pcmk__str_eq(crm_op, CRM_OP_REPROBE, pcmk__str_casei) || pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_casei)) { handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node); } else if (operation != NULL) { lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(input->xml, XML_CIB_TAG_RESOURCE, TRUE); gboolean create_rsc = !pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei); int rc; // We can't return anything meaningful without a resource ID CRM_CHECK(xml_rsc && ID(xml_rsc), return); rc = get_lrm_resource(lrm_state, xml_rsc, create_rsc, &rsc); if (rc == -ENOTCONN) { synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Not connected to remote executor"); return; } else if ((rc < 0) && !create_rsc) { /* Delete of malformed or nonexistent resource * (deleting something that does not exist is a success) */ crm_notice("Not registering resource '%s' for a %s event " CRM_XS " get-rc=%d (%s) transition-key=%s", ID(xml_rsc), operation, rc, pcmk_strerror(rc), ID(input->xml)); delete_rsc_entry(lrm_state, input, ID(xml_rsc), NULL, pcmk_ok, user_name); return; } else if (rc == -EINVAL) { // Resource operation on malformed resource crm_err("Invalid resource definition for %s", ID(xml_rsc)); crm_log_xml_warn(input->msg, "invalid resource"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, PCMK_OCF_NOT_CONFIGURED, // fatal error "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_casei)) { if (!do_lrm_cancel(input, lrm_state, rsc, from_host, from_sys)) { crm_log_xml_warn(input->xml, "Bad command"); } } else if (pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else if (pcmk__str_any_of(operation, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { /* Pre-2.1.0 DCs will schedule reload actions only, and 2.1.0+ DCs * will schedule reload-agent actions only. In either case, we need * to map that to whatever the resource agent actually supports. * Default to the OCF 1.1 name. */ struct ra_metadata_s *md = NULL; const char *reload_name = CRMD_ACTION_RELOAD_AGENT; md = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if ((md != NULL) && pcmk_is_set(md->ra_flags, ra_supports_legacy_reload)) { reload_name = CRMD_ACTION_RELOAD; } do_lrm_rsc_op(lrm_state, rsc, reload_name, input->xml); } else { do_lrm_rsc_op(lrm_state, rsc, operation, input->xml); } lrmd_free_rsc_info(rsc); } else { crm_err("Cannot perform operation %s of unknown type", crm_str(crm_op)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } #if ENABLE_VERSIONED_ATTRS static void resolve_versioned_parameters(lrm_state_t *lrm_state, const char *rsc_id, const xmlNode *rsc_op, GHashTable *params) { /* Resource info *should* already be cached, so we don't get * executor call */ lrmd_rsc_info_t *rsc = lrm_state_get_rsc_info(lrm_state, rsc_id, 0); struct ra_metadata_s *metadata; metadata = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if (metadata) { xmlNode *versioned_attrs = NULL; GHashTable *hash = NULL; char *key = NULL; char *value = NULL; GHashTableIter iter; versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_META); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_replace(params, crm_meta_name(key), strdup(value)); if (pcmk__str_eq(key, XML_ATTR_TIMEOUT, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->timeout, 0); } else if (pcmk__str_eq(key, XML_OP_ATTR_START_DELAY, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->start_delay, 0); } } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_RSC_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); } lrmd_free_rsc_info(rsc); } #endif static lrmd_event_data_t * construct_op(lrm_state_t *lrm_state, xmlNode *rsc_op, const char *rsc_id, const char *operation) { lrmd_event_data_t *op = NULL; const char *op_delay = NULL; const char *op_timeout = NULL; GHashTable *params = NULL; xmlNode *primitive = NULL; const char *class = NULL; const char *transition = NULL; CRM_ASSERT(rsc_id && operation); op = lrmd_new_event(rsc_id, operation, 0); op->type = lrmd_event_exec_complete; op->timeout = 0; op->start_delay = 0; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); if (rsc_op == NULL) { CRM_LOG_ASSERT(pcmk__str_eq(CRMD_ACTION_STOP, operation, pcmk__str_casei)); op->user_data = NULL; /* the stop_all_resources() case * by definition there is no DC (or they'd be shutting * us down). * So we should put our version here. */ op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); crm_trace("Constructed %s op for %s", operation, rsc_id); return op; } params = xml2list(rsc_op); g_hash_table_remove(params, CRM_META "_op_target_rc"); op_delay = crm_meta_value(params, XML_OP_ATTR_START_DELAY); pcmk__scan_min_int(op_delay, &op->start_delay, 0); op_timeout = crm_meta_value(params, XML_ATTR_TIMEOUT); pcmk__scan_min_int(op_timeout, &op->timeout, 0); if (pcmk__guint_from_hash(params, CRM_META "_" XML_LRM_ATTR_INTERVAL_MS, 0, &(op->interval_ms)) != pcmk_rc_ok) { op->interval_ms = 0; } /* Use pcmk_monitor_timeout instead of meta timeout for stonith recurring monitor, if set */ primitive = find_xml_node(rsc_op, XML_CIB_TAG_RESOURCE, FALSE); class = crm_element_value(primitive, XML_AGENT_ATTR_CLASS); if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_fence_params) && pcmk__str_eq(operation, CRMD_ACTION_STATUS, pcmk__str_casei) && (op->interval_ms > 0)) { op_timeout = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (op_timeout != NULL) { op->timeout = crm_get_msec(op_timeout); } } #if ENABLE_VERSIONED_ATTRS if (lrm_state && !is_remote_lrmd_ra(NULL, NULL, rsc_id) && !pcmk__strcase_any_of(op_type, CRMD_ACTION_METADATA, CRMD_ACTION_DELETE, NULL)) { resolve_versioned_parameters(lrm_state, rsc_id, rsc_op, params); } #endif if (!pcmk__str_eq(operation, RSC_STOP, pcmk__str_casei)) { op->params = params; } else { rsc_history_t *entry = NULL; if (lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* If we do not have stop parameters cached, use * whatever we are given */ if (!entry || !entry->stop_params) { op->params = params; } else { /* Copy the cached parameter list so that we stop the resource * with the old attributes, not the new ones */ op->params = pcmk__strkey_table(free, free); g_hash_table_foreach(params, copy_meta_keys, op->params); g_hash_table_foreach(entry->stop_params, copy_instance_keys, op->params); g_hash_table_destroy(params); params = NULL; } } /* sanity */ if (op->timeout <= 0) { op->timeout = op->interval_ms; } if (op->start_delay < 0) { op->start_delay = 0; } transition = crm_element_value(rsc_op, XML_ATTR_TRANSITION_KEY); CRM_CHECK(transition != NULL, return op); op->user_data = strdup(transition); if (op->interval_ms != 0) { if (pcmk__strcase_any_of(operation, CRMD_ACTION_START, CRMD_ACTION_STOP, NULL)) { crm_err("Start and Stop actions cannot have an interval: %u", op->interval_ms); op->interval_ms = 0; } } crm_trace("Constructed %s op for %s: interval=%u", operation, rsc_id, op->interval_ms); return op; } /*! * \internal * \brief Send a (synthesized) event result * * Reply with a synthesized event result directly, as opposed to going through * the executor. * * \param[in] to_host Host to send result to * \param[in] to_sys IPC name to send result to (NULL for transition engine) * \param[in] rsc Type information about resource the result is for * \param[in] op Event with result to send * \param[in] rsc_id ID of resource the result is for */ void controld_ack_event_directly(const char *to_host, const char *to_sys, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, const char *rsc_id) { xmlNode *reply = NULL; xmlNode *update, *iter; crm_node_t *peer = NULL; CRM_CHECK(op != NULL, return); if (op->rsc_id == NULL) { CRM_ASSERT(rsc_id != NULL); op->rsc_id = strdup(rsc_id); } if (to_sys == NULL) { to_sys = CRM_SYSTEM_TENGINE; } peer = crm_get_peer(0, fsa_our_uname); update = create_node_state_update(peer, node_update_none, NULL, __func__); iter = create_xml_node(update, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, fsa_our_uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, fsa_our_uname, __func__); reply = create_request(CRM_OP_INVOKE_LRM, update, to_host, to_sys, CRM_SYSTEM_LRMD, NULL); crm_log_xml_trace(update, "[direct ACK]"); crm_debug("ACK'ing resource op " PCMK__OP_FMT " from %s: %s", op->rsc_id, op->op_type, op->interval_ms, op->user_data, crm_element_value(reply, XML_ATTR_REFERENCE)); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(update); free_xml(reply); } gboolean verify_stopped(enum crmd_fsa_state cur_state, int log_level) { gboolean res = TRUE; GList *lrm_state_list = lrm_state_get_list(); GList *state_entry; for (state_entry = lrm_state_list; state_entry != NULL; state_entry = state_entry->next) { lrm_state_t *lrm_state = state_entry->data; if (!lrm_state_verify_stopped(lrm_state, cur_state, log_level)) { /* keep iterating through all even when false is returned */ res = FALSE; } } controld_set_fsa_input_flags(R_SENT_RSC_STOP); g_list_free(lrm_state_list); lrm_state_list = NULL; return res; } struct stop_recurring_action_s { lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean stop_recurring_action_by_rsc(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct stop_recurring_action_s *event = user_data; active_op_t *op = value; if ((op->interval_ms != 0) && pcmk__str_eq(op->rsc_id, event->rsc->id, pcmk__str_none)) { crm_debug("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (char*)key); remove = !cancel_op(event->lrm_state, event->rsc->id, key, op->call_id, FALSE); } return remove; } static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; lrm_state_t *lrm_state = user_data; active_op_t *op = value; if (op->interval_ms != 0) { crm_info("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (const char *) key); remove = !cancel_op(lrm_state, op->rsc_id, key, op->call_id, FALSE); } return remove; } static void record_pending_op(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op) { const char *record_pending = NULL; CRM_CHECK(node_name != NULL, return); CRM_CHECK(rsc != NULL, return); CRM_CHECK(op != NULL, return); // Never record certain operation types as pending if ((op->op_type == NULL) || (op->params == NULL) || !controld_action_is_recordable(op->op_type)) { return; } // defaults to true record_pending = crm_meta_value(op->params, XML_OP_ATTR_PENDING); if (record_pending && !crm_is_true(record_pending)) { return; } op->call_id = -1; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); op->t_run = time(NULL); op->t_rcchange = op->t_run; /* write a "pending" entry to the CIB, inhibit notification */ crm_debug("Recording pending op " PCMK__OP_FMT " on %s in the CIB", op->rsc_id, op->op_type, op->interval_ms, node_name); do_update_resource(node_name, rsc, op, 0); } static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg) { int rc; int call_id = 0; char *op_id = NULL; lrmd_event_data_t *op = NULL; lrmd_key_value_t *params = NULL; fsa_data_t *msg_data = NULL; const char *transition = NULL; gboolean stop_recurring = FALSE; const char *nack_reason = NULL; CRM_CHECK(rsc != NULL, return); CRM_CHECK(operation != NULL, return); if (msg != NULL) { transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (transition == NULL) { crm_log_xml_err(msg, "Missing transition number"); } } op = construct_op(lrm_state, msg, rsc->id, operation); CRM_CHECK(op != NULL, return); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) && (op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_MIGRATE) == 0) { /* pcmk remote connections are a special use case. * We never ever want to stop monitoring a connection resource until * the entire migration has completed. If the connection is unexpectedly * severed, even during a migration, this is an event we must detect.*/ stop_recurring = FALSE; } else if ((op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_STATUS) != 0 && strcmp(operation, CRMD_ACTION_NOTIFY) != 0) { /* stop any previous monitor operations before changing the resource state */ stop_recurring = TRUE; } if (stop_recurring == TRUE) { guint removed = 0; struct stop_recurring_action_s data; data.rsc = rsc; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_action_by_rsc, &data); if (removed) { crm_debug("Stopped %u recurring operation%s in preparation for " PCMK__OP_FMT, removed, pcmk__plural_s(removed), rsc->id, operation, op->interval_ms); } } /* now do the op */ crm_notice("Requesting local execution of %s operation for %s on %s " CRM_XS " transition_key=%s op_key=" PCMK__OP_FMT, crm_action_str(op->op_type, op->interval_ms), rsc->id, lrm_state->node_name, transition, rsc->id, operation, op->interval_ms); if (pcmk_is_set(fsa_input_register, R_SHUTDOWN) && pcmk__str_eq(operation, RSC_START, pcmk__str_casei)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); nack_reason = "Not attempting start due to shutdown in progress"; } else if (fsa_state != S_NOT_DC && fsa_state != S_POLICY_ENGINE /* Recalculating */ && fsa_state != S_TRANSITION_ENGINE && !pcmk__str_eq(operation, CRMD_ACTION_STOP, pcmk__str_casei)) { nack_reason = "Controller cannot attempt actions at this time"; } if (nack_reason != NULL) { crm_notice("Discarding attempt to perform action %s on %s in state %s (shutdown=%s)", operation, rsc->id, fsa_state2string(fsa_state), pcmk__btoa(pcmk_is_set(fsa_input_register, R_SHUTDOWN))); 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); } int last_resource_update = 0; static void cib_rsc_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { switch (rc) { case pcmk_ok: case -pcmk_err_diff_failed: case -pcmk_err_diff_resync: crm_trace("Resource update %d complete: rc=%d", call_id, rc); break; default: crm_warn("Resource update %d failed: (rc=%d) %s", call_id, rc, pcmk_strerror(rc)); } if (call_id == last_resource_update) { last_resource_update = 0; trigger_fsa(); } } /* Only successful stops, and probes that found the resource inactive, get locks * recorded in the history. This ensures the resource stays locked to the node * until it is active there again after the node comes back up. */ static bool should_preserve_lock(lrmd_event_data_t *op) { if (!controld_shutdown_lock_enabled) { return false; } if (!strcmp(op->op_type, RSC_STOP) && (op->rc == PCMK_OCF_OK)) { return true; } if (!strcmp(op->op_type, RSC_STATUS) && (op->rc == PCMK_OCF_NOT_RUNNING)) { return true; } return false; } static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time) { /* */ int rc = pcmk_ok; xmlNode *update, *iter = NULL; int call_opt = crmd_cib_smart_opt(); const char *uuid = NULL; CRM_CHECK(op != NULL, return 0); iter = create_xml_node(iter, XML_CIB_TAG_STATUS); update = iter; iter = create_xml_node(iter, XML_CIB_TAG_STATE); if (pcmk__str_eq(node_name, fsa_our_uname, pcmk__str_casei)) { uuid = fsa_our_uuid; } else { /* remote nodes uuid and uname are equal */ uuid = node_name; 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_UNAME, node_name); crm_xml_add(iter, XML_ATTR_ORIGIN, __func__); iter = create_xml_node(iter, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, node_name, __func__); if (rsc) { const char *container = NULL; crm_xml_add(iter, XML_ATTR_TYPE, rsc->type); crm_xml_add(iter, XML_AGENT_ATTR_CLASS, rsc->standard); crm_xml_add(iter, XML_AGENT_ATTR_PROVIDER, rsc->provider); if (lock_time != 0) { /* Actions on a locked resource should either preserve the lock by * recording it with the action result, or clear it. */ if (!should_preserve_lock(op)) { lock_time = 0; } crm_xml_add_ll(iter, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) lock_time); } if (op->params) { container = g_hash_table_lookup(op->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); } if (container) { crm_trace("Resource %s is a part of container resource %s", op->rsc_id, container); crm_xml_add(iter, XML_RSC_ATTR_CONTAINER, container); } } else { crm_warn("Resource %s no longer exists in the executor", op->rsc_id); controld_ack_event_directly(NULL, NULL, rsc, op, op->rsc_id); goto cleanup; } crm_log_xml_trace(update, __func__); /* make it an asynchronous call and be done with it * * Best case: * the resource state will be discovered during * the next signup or election. * * Bad case: * we are shutting down and there is no DC at the time, * but then why were we shutting down then anyway? * (probably because of an internal error) * * Worst case: * we get shot for having resources "running" that really weren't * * the alternative however means blocking here for too long, which * isn't acceptable */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, rc, NULL); if (rc > 0) { last_resource_update = rc; } done: /* the return code is a call number, not an error code */ crm_trace("Sent resource state update message: %d for %s=%u on %s", rc, op->op_type, op->interval_ms, op->rsc_id); fsa_register_cib_callback(rc, FALSE, NULL, cib_rsc_callback); cleanup: free_xml(update); return rc; } void do_lrm_event(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input cur_input, fsa_data_t * msg_data) { CRM_CHECK(FALSE, return); } static char * unescape_newlines(const char *string) { char *pch = NULL; char *ret = NULL; static const char *escaped_newline = "\\n"; if (!string) { return NULL; } ret = strdup(string); pch = strstr(ret, escaped_newline); while (pch != NULL) { /* Replace newline escape pattern with actual newline (and a space so we * don't have to shuffle the rest of the buffer) */ pch[0] = '\n'; pch[1] = ' '; pch = strstr(pch, escaped_newline); } return ret; } static bool did_lrm_rsc_op_fail(lrm_state_t *lrm_state, const char * rsc_id, const char * op_type, guint interval_ms) { rsc_history_t *entry = NULL; CRM_CHECK(lrm_state != NULL, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); CRM_CHECK(op_type != NULL, return FALSE); entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->failed == NULL) { return FALSE; } if (pcmk__str_eq(entry->failed->rsc_id, rsc_id, pcmk__str_none) && pcmk__str_eq(entry->failed->op_type, op_type, pcmk__str_casei) && entry->failed->interval_ms == interval_ms) { return TRUE; } return FALSE; } /*! * \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(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 g_string_printf(str, "Result of %s operation for %s", crm_action_str(op->op_type, op->interval_ms), op->rsc_id); if (node_name != NULL) { g_string_append_printf(str, " on %s", node_name); } switch (op->op_status) { case PCMK_EXEC_DONE: log_level = LOG_NOTICE; g_string_append_printf(str, ": %s", services_ocf_exitcode_str(op->rc)); break; case PCMK_EXEC_TIMEOUT: g_string_append_printf(str, ": %s after %s", pcmk_exec_status_str(op->op_status), pcmk__readable_interval(op->timeout)); break; case PCMK_EXEC_CANCELLED: log_level = LOG_INFO; // Fall through default: g_string_append_printf(str, ": %s", pcmk_exec_status_str(op->op_status)); } if ((op->exit_reason != NULL) && ((op->op_status != PCMK_EXEC_DONE) || (op->rc != PCMK_OCF_OK))) { g_string_append_printf(str, " (%s)", op->exit_reason); } 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, xmlNode *action_xml) { char *op_id = NULL; char *op_key = NULL; int update_id = 0; gboolean remove = FALSE; gboolean removed = FALSE; bool need_direct_ack = FALSE; lrmd_rsc_info_t *rsc = NULL; const char *node_name = NULL; CRM_CHECK(op != NULL, return); CRM_CHECK(op->rsc_id != NULL, return); // Remap new status codes for older DCs if (compare_version(fsa_our_dc_version, "3.2.0") < 0) { switch (op->op_status) { case PCMK_EXEC_NOT_CONNECTED: 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(lrm_state, op); } /* If the recurring operation had failed, the lrm_rsc_op is recorded as * "last_failure" which won't get erased from the cib given the logic on * purpose in erase_lrm_history_by_op(). So that the cancel action won't * have a chance to get confirmed by DC with process_op_deletion(). * Cluster transition would get stuck waiting for the remaining action * timer to time out. * * Directly acknowledge the cancel operation in this case. */ if (did_lrm_rsc_op_fail(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms)) { need_direct_ack = TRUE; } } else if (op->rsc_deleted) { /* This recurring operation was cancelled (but not by us, and the * executor does not have resource information, likely due to resource * cleanup, refresh, or removal) and pending. */ crm_debug("Recurring op %s was cancelled due to resource deletion", op_key); need_direct_ack = TRUE; } else { /* This recurring operation was cancelled (but not by us, likely by the * executor before stopping the resource) and pending. We don't need to * do anything special. */ } if (need_direct_ack) { controld_ack_event_directly(NULL, NULL, NULL, op, op->rsc_id); } if(remove == FALSE) { /* The caller will do this afterwards, but keep the logging consistent */ removed = TRUE; } else if (lrm_state && ((op->interval_ms == 0) || (op->op_status == PCMK_EXEC_CANCELLED))) { gboolean found = g_hash_table_remove(lrm_state->pending_ops, op_id); if (op->interval_ms != 0) { removed = TRUE; } else if (found) { removed = TRUE; crm_trace("Op %s (call=%d, stop-id=%s, remaining=%u): Confirmed", op_key, op->call_id, op_id, g_hash_table_size(lrm_state->pending_ops)); } } 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); metadata_cache_update(lrm_state->metadata_cache, rsc, metadata); free(metadata); } } if (op->rsc_deleted) { crm_info("Deletion of resource '%s' complete after %s", op->rsc_id, op_key); if (lrm_state) { delete_rsc_entry(lrm_state, NULL, op->rsc_id, NULL, pcmk_ok, NULL); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ mainloop_set_trigger(fsa_source); if (lrm_state && rsc) { update_history_cache(lrm_state, rsc, op); } lrmd_free_rsc_info(rsc); free(op_key); free(op_id); } diff --git a/daemons/controld/controld_join_client.c b/daemons/controld/controld_join_client.c index f84baff845..6485856da1 100644 --- a/daemons/controld/controld_join_client.c +++ b/daemons/controld/controld_join_client.c @@ -1,311 +1,315 @@ /* - * Copyright 2004-2020 the Pacemaker project contributors + * 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 +extern pcmk__output_t *logger_out; + int reannounce_count = 0; void join_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data); extern ha_msg_input_t *copy_ha_msg_input(ha_msg_input_t * orig); /*! * \internal * \brief Remember if DC is shutting down as we join * * If we're joining while the current DC is shutting down, update its expected * state, so we don't fence it if we become the new DC. (We weren't a peer * when it broadcast its shutdown request.) * * \param[in] msg A join message from the DC */ static void update_dc_expected(xmlNode *msg) { if (fsa_our_dc && pcmk__xe_attr_is_true(msg, F_CRM_DC_LEAVING)) { crm_node_t *dc_node = crm_get_peer(0, fsa_our_dc); pcmk__update_peer_expected(__func__, dc_node, CRMD_JOINSTATE_DOWN); } } /* A_CL_JOIN_QUERY */ /* is there a DC out there? */ void do_cl_join_query(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 *req = create_request(CRM_OP_JOIN_ANNOUNCE, NULL, NULL, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); sleep(1); // Give the cluster layer time to propagate to the DC update_dc(NULL); /* Unset any existing value so that the result is not discarded */ crm_debug("Querying for a DC"); send_cluster_message(NULL, crm_msg_crmd, req, FALSE); free_xml(req); } /* A_CL_JOIN_ANNOUNCE */ /* this is kind of a workaround for the fact that we may not be around or * are otherwise unable to reply when the DC sends out A_DC_JOIN_OFFER_ALL */ void do_cl_join_announce(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) { /* don't announce if we're in one of these states */ if (cur_state != S_PENDING) { crm_warn("Not announcing cluster join because in state %s", fsa_state2string(cur_state)); return; } if (AM_I_OPERATIONAL) { /* send as a broadcast */ xmlNode *req = create_request(CRM_OP_JOIN_ANNOUNCE, NULL, NULL, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); crm_debug("Announcing availability"); update_dc(NULL); send_cluster_message(NULL, crm_msg_crmd, req, FALSE); free_xml(req); } else { /* Delay announce until we have finished local startup */ crm_warn("Delaying announce of cluster join until local startup is complete"); return; } } static int query_call_id = 0; /* A_CL_JOIN_REQUEST */ /* aka. accept the welcome offer */ void do_cl_join_offer_respond(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); const char *welcome_from; const char *join_id; CRM_CHECK(input != NULL, return); #if 0 if (we are sick) { log error; /* save the request for later? */ return; } #endif welcome_from = crm_element_value(input->msg, F_CRM_HOST_FROM); join_id = crm_element_value(input->msg, F_CRM_JOIN_ID); crm_trace("Accepting cluster join offer from node %s "CRM_XS" join-%s", welcome_from, crm_element_value(input->msg, F_CRM_JOIN_ID)); /* we only ever want the last one */ if (query_call_id > 0) { crm_trace("Cancelling previous join query: %d", query_call_id); remove_cib_op_callback(query_call_id, FALSE); query_call_id = 0; } if (update_dc(input->msg) == FALSE) { crm_warn("Discarding cluster join offer from node %s (expected %s)", welcome_from, fsa_our_dc); return; } update_dc_expected(input->msg); query_call_id = fsa_cib_conn->cmds->query(fsa_cib_conn, NULL, NULL, cib_scope_local | cib_no_children); fsa_register_cib_callback(query_call_id, FALSE, strdup(join_id), join_query_callback); crm_trace("Registered join query callback: %d", query_call_id); controld_set_fsa_action_flags(A_DC_TIMER_STOP); trigger_fsa(); } void join_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *join_id = user_data; xmlNode *generation = create_xml_node(NULL, XML_CIB_TAG_GENERATION_TUPPLE); CRM_LOG_ASSERT(join_id != NULL); if (query_call_id != call_id) { crm_trace("Query %d superseded", call_id); goto done; } query_call_id = 0; if(rc != pcmk_ok || output == NULL) { crm_err("Could not retrieve version details for join-%s: %s (%d)", join_id, pcmk_strerror(rc), rc); register_fsa_error_adv(C_FSA_INTERNAL, I_ERROR, NULL, NULL, __func__); } else if (fsa_our_dc == NULL) { crm_debug("Membership is in flux, not continuing join-%s", join_id); } else { xmlNode *reply = NULL; crm_debug("Respond to join offer join-%s from %s", join_id, fsa_our_dc); copy_in_properties(generation, output); reply = create_request(CRM_OP_JOIN_REQUEST, generation, fsa_our_dc, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); crm_xml_add(reply, F_CRM_JOIN_ID, join_id); crm_xml_add(reply, XML_ATTR_CRM_VERSION, CRM_FEATURE_SET); send_cluster_message(crm_get_peer(0, fsa_our_dc), crm_msg_crmd, reply, TRUE); free_xml(reply); } done: free_xml(generation); } static void set_join_state(const char * start_state) { if (pcmk__str_eq(start_state, "standby", pcmk__str_casei)) { crm_notice("Forcing node %s to join in %s state per configured environment", fsa_our_uname, start_state); - update_attr_delegate(fsa_cib_conn, cib_sync_call, XML_CIB_TAG_NODES, fsa_our_uuid, - NULL, NULL, NULL, "standby", "on", TRUE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_sync_call, + XML_CIB_TAG_NODES, fsa_our_uuid, NULL, NULL, + NULL, "standby", "on", NULL, NULL); } else if (pcmk__str_eq(start_state, "online", pcmk__str_casei)) { crm_notice("Forcing node %s to join in %s state per configured environment", fsa_our_uname, start_state); - update_attr_delegate(fsa_cib_conn, cib_sync_call, XML_CIB_TAG_NODES, fsa_our_uuid, - NULL, NULL, NULL, "standby", "off", TRUE, NULL, NULL); + cib__update_node_attr(logger_out, fsa_cib_conn, cib_sync_call, + XML_CIB_TAG_NODES, fsa_our_uuid, NULL, NULL, + NULL, "standby", "off", NULL, NULL); } else if (pcmk__str_eq(start_state, "default", pcmk__str_casei)) { crm_debug("Not forcing a starting state on node %s", fsa_our_uname); } else { crm_warn("Unrecognized start state '%s', using 'default' (%s)", start_state, fsa_our_uname); } } /* A_CL_JOIN_RESULT */ /* aka. this is notification that we have (or have not) been accepted */ void do_cl_join_finalize_respond(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 *tmp1 = NULL; gboolean was_nack = TRUE; static gboolean first_join = TRUE; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *start_state = pcmk__env_option(PCMK__ENV_NODE_START_STATE); int join_id = -1; const char *op = crm_element_value(input->msg, F_CRM_TASK); const char *welcome_from = crm_element_value(input->msg, F_CRM_HOST_FROM); if (!pcmk__str_eq(op, CRM_OP_JOIN_ACKNAK, pcmk__str_casei)) { crm_trace("Ignoring op=%s message", op); return; } /* calculate if it was an ack or a nack */ if (pcmk__xe_attr_is_true(input->msg, CRM_OP_JOIN_ACKNAK)) { was_nack = FALSE; } crm_element_value_int(input->msg, F_CRM_JOIN_ID, &join_id); if (was_nack) { crm_err("Shutting down because cluster join with leader %s failed " CRM_XS" join-%d NACK'd", welcome_from, join_id); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (AM_I_DC == FALSE && pcmk__str_eq(welcome_from, fsa_our_uname, pcmk__str_casei)) { crm_warn("Discarding our own welcome - we're no longer the DC"); return; } if (update_dc(input->msg) == FALSE) { crm_warn("Discarding %s from node %s (expected from %s)", op, welcome_from, fsa_our_dc); return; } update_dc_expected(input->msg); /* send our status section to the DC */ tmp1 = controld_query_executor_state(fsa_our_uname); if (tmp1 != NULL) { xmlNode *reply = create_request(CRM_OP_JOIN_CONFIRM, tmp1, fsa_our_dc, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); crm_xml_add_int(reply, F_CRM_JOIN_ID, join_id); crm_debug("Confirming join-%d: sending local operation history to %s", join_id, fsa_our_dc); /* * If this is the node's first join since the controller started on it, * set its initial state (standby or member) according to the user's * preference. * * We do not clear the LRM history here. Even if the DC failed to do it * when we last left, removing them here creates a race condition if the * controller is being recovered. Instead of a list of active resources * from the executor, we may end up with a blank status section. If we * are _NOT_ lucky, we will probe for the "wrong" instance of anonymous * clones and end up with multiple active instances on the machine. */ if (first_join && !pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { first_join = FALSE; if (start_state) { set_join_state(start_state); } } send_cluster_message(crm_get_peer(0, fsa_our_dc), crm_msg_crmd, reply, TRUE); free_xml(reply); if (AM_I_DC == FALSE) { register_fsa_input_adv(cause, I_NOT_DC, NULL, A_NOTHING, TRUE, __func__); } free_xml(tmp1); } else { crm_err("Could not confirm join-%d with %s: Local operation history failed", join_id, fsa_our_dc); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } } diff --git a/daemons/controld/pacemaker-controld.c b/daemons/controld/pacemaker-controld.c index 83e629dc27..686f35d8b1 100644 --- a/daemons/controld/pacemaker-controld.c +++ b/daemons/controld/pacemaker-controld.c @@ -1,176 +1,186 @@ /* - * Copyright 2004-2020 the Pacemaker project contributors + * 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 #include #include #define OPTARGS "hV" void usage(const char *cmd, int exit_status); _Noreturn void crmd_init(void); void crmd_hamsg_callback(const xmlNode * msg, void *private_data); extern void init_dotfile(void); GMainLoop *crmd_mainloop = NULL; +pcmk__output_t *logger_out = NULL; + static pcmk__cli_option_t long_options[] = { // long option, argument type, storage, short option, description, flags { "help", no_argument, NULL, '?', "\tThis text", pcmk__option_default }, { "verbose", no_argument, NULL, 'V', "\tIncrease debug output", pcmk__option_default }, { 0, 0, 0, 0 } }; int main(int argc, char **argv) { int flag; int index = 0; int argerr = 0; crm_ipc_t *old_instance = NULL; crmd_mainloop = g_main_loop_new(NULL, FALSE); crm_log_preinit(NULL, argc, argv); pcmk__set_cli_options(NULL, "[options]", long_options, "daemon for coordinating a Pacemaker cluster's " "response to events"); while (1) { flag = pcmk__next_cli_option(argc, argv, &index, NULL); if (flag == -1) break; switch (flag) { case 'V': crm_bump_log_level(argc, argv); break; case 'h': /* Help message */ pcmk__cli_help(flag, CRM_EX_OK); break; default: ++argerr; break; } } if (argc - optind == 1 && pcmk__str_eq("metadata", argv[optind], pcmk__str_casei)) { crmd_metadata(); return CRM_EX_OK; } else if (argc - optind == 1 && pcmk__str_eq("version", argv[optind], pcmk__str_casei)) { fprintf(stdout, "CRM Version: %s (%s)\n", PACEMAKER_VERSION, BUILD_VERSION); return CRM_EX_OK; } crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); if (optind > argc) { ++argerr; } if (argerr) { pcmk__cli_help('?', CRM_EX_USAGE); } crm_notice("Starting Pacemaker controller"); old_instance = crm_ipc_new(CRM_SYSTEM_CRMD, 0); if (crm_ipc_connect(old_instance)) { /* IPC end-point already up */ crm_ipc_close(old_instance); crm_ipc_destroy(old_instance); crm_err("pacemaker-controld is already active, aborting startup"); crm_exit(CRM_EX_OK); } else { /* not up or not authentic, we'll proceed either way */ crm_ipc_destroy(old_instance); old_instance = NULL; } if (pcmk__daemon_can_write(PE_STATE_DIR, NULL) == FALSE) { crm_err("Terminating due to bad permissions on " PE_STATE_DIR); fprintf(stderr, "ERROR: Bad permissions on " PE_STATE_DIR " (see logs for details)\n"); fflush(stderr); return CRM_EX_FATAL; } else if (pcmk__daemon_can_write(CRM_CONFIG_DIR, NULL) == FALSE) { crm_err("Terminating due to bad permissions on " CRM_CONFIG_DIR); fprintf(stderr, "ERROR: Bad permissions on " CRM_CONFIG_DIR " (see logs for details)\n"); fflush(stderr); return CRM_EX_FATAL; } + logger_out = pcmk__new_logger(); + if (logger_out == NULL) { + return CRM_EX_FATAL; + } + + pcmk__output_set_log_level(logger_out, LOG_TRACE); + crmd_init(); return 0; // not reachable } static void log_deprecation_warnings(void) { // Add deprecations here as needed } void crmd_init(void) { crm_exit_t exit_code = CRM_EX_OK; enum crmd_fsa_state state; log_deprecation_warnings(); fsa_state = S_STARTING; fsa_input_register = 0; /* zero out the regester */ init_dotfile(); register_fsa_input(C_STARTUP, I_STARTUP, NULL); crm_peer_init(); state = s_crmd_fsa(C_STARTUP); if (state == S_PENDING || state == S_STARTING) { /* Create the mainloop and run it... */ crm_trace("Starting %s's mainloop", crm_system_name); g_main_loop_run(crmd_mainloop); if (pcmk_is_set(fsa_input_register, R_STAYDOWN)) { crm_info("Inhibiting automated respawn"); exit_code = CRM_EX_FATAL; } } else { crm_err("Startup of %s failed. Current state: %s", crm_system_name, fsa_state2string(state)); exit_code = CRM_EX_ERROR; } crm_info("%s[%lu] exiting with status %d (%s)", crm_system_name, (unsigned long) getpid(), exit_code, crm_exit_str(exit_code)); + crmd_fast_exit(exit_code); } diff --git a/lib/pacemaker/pcmk_injections.c b/lib/pacemaker/pcmk_injections.c index 1aa90a5a0b..895003435e 100644 --- a/lib/pacemaker/pcmk_injections.c +++ b/lib/pacemaker/pcmk_injections.c @@ -1,757 +1,758 @@ /* * 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_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; } #define XPATH_MAX 1024 static int find_ticket_state(pcmk__output_t *out, cib_t *the_cib, const char *ticket_id, xmlNode **ticket_state_xml) { int offset = 0; int rc = pcmk_ok; xmlNode *xml_search = NULL; char *xpath_string = NULL; CRM_ASSERT(ticket_state_xml != NULL); *ticket_state_xml = NULL; xpath_string = calloc(1, XPATH_MAX); offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "%s", "/cib/status/tickets"); if (ticket_id) { offset += snprintf(xpath_string + offset, XPATH_MAX - offset, "/%s[@id=\"%s\"]", XML_CIB_TAG_TICKET_STATE, ticket_id); } CRM_LOG_ASSERT(offset > 0); rc = the_cib->cmds->query(the_cib, xpath_string, &xml_search, cib_sync_call|cib_scope_local|cib_xpath); if (rc != pcmk_ok) { goto bail; } crm_log_xml_debug(xml_search, "Match"); if (xml_has_children(xml_search)) { if (ticket_id) { out->err(out, "Multiple ticket_states match ticket_id=%s", ticket_id); } *ticket_state_xml = xml_search; } else { *ticket_state_xml = xml_search; } bail: free(xpath_string); return rc; } /*! * \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 = update_attr_delegate(cib, cib_sync_call|cib_scope_local, - XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, - XML_ATTR_HAVE_WATCHDOG, injections->watchdog, - FALSE, NULL, NULL); - CRM_ASSERT(rc == pcmk_ok); + 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); }