diff --git a/daemons/controld/Makefile.am b/daemons/controld/Makefile.am index 17c33424d3..858e1bb390 100644 --- a/daemons/controld/Makefile.am +++ b/daemons/controld/Makefile.am @@ -1,77 +1,80 @@ # -# Copyright 2004-2018 Andrew Beekhof +# Copyright 2018-2019 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 $(top_srcdir)/Makefile.common halibdir = $(CRM_DAEMON_DIR) halib_PROGRAMS = pacemaker-controld noinst_HEADERS = controld_alerts.h \ controld_callbacks.h \ controld_fsa.h \ controld_lrm.h \ controld_matrix.h \ controld_membership.h \ controld_messages.h \ controld_metadata.h \ controld_throttle.h \ controld_transition.h \ controld_utils.h \ pacemaker-controld.h pacemaker_controld_CFLAGS = $(CFLAGS_HARDENED_EXE) pacemaker_controld_LDFLAGS = $(LDFLAGS_HARDENED_EXE) pacemaker_controld_LDADD = $(top_builddir)/lib/fencing/libstonithd.la \ $(top_builddir)/lib/pacemaker/libpacemaker.la \ $(top_builddir)/lib/pengine/libpe_rules.la \ $(top_builddir)/lib/cib/libcib.la \ $(top_builddir)/lib/cluster/libcrmcluster.la \ $(top_builddir)/lib/common/libcrmcommon.la \ $(top_builddir)/lib/services/libcrmservice.la \ $(top_builddir)/lib/lrmd/liblrmd.la \ $(CLUSTERLIBS) pacemaker_controld_SOURCES = pacemaker-controld.c \ controld_attrd.c \ controld_callbacks.c \ controld_based.c \ controld_control.c \ controld_corosync.c \ controld_election.c \ controld_execd.c \ controld_execd_state.c \ + controld_fencing.c \ controld_fsa.c \ controld_join_client.c \ controld_join_dc.c \ controld_membership.c \ controld_messages.c \ controld_metadata.c \ controld_remote_ra.c \ controld_schedulerd.c \ controld_te_actions.c \ controld_te_callbacks.c \ controld_te_events.c \ controld_te_utils.c \ controld_throttle.c \ controld_transition.c \ controld_utils.c if BUILD_XML_HELP man7_MANS = pacemaker-controld.7 endif CLEANFILES = $(man7_MANS) if BUILD_LEGACY_LINKS install-exec-hook: cd $(DESTDIR)$(CRM_DAEMON_DIR) && rm -f crmd && $(LN_S) pacemaker-controld crmd uninstall-hook: cd $(DESTDIR)$(CRM_DAEMON_DIR) && rm -f crmd endif diff --git a/daemons/controld/controld_callbacks.c b/daemons/controld/controld_callbacks.c index a188263109..06ffb9d1b9 100644 --- a/daemons/controld/controld_callbacks.c +++ b/daemons/controld/controld_callbacks.c @@ -1,339 +1,340 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include /* From join_dc... */ extern gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); void crmd_ha_msg_filter(xmlNode * msg) { if (AM_I_DC) { const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM); if (safe_str_eq(sys_from, CRM_SYSTEM_DC)) { const char *from = crm_element_value(msg, F_ORIG); if (safe_str_neq(from, fsa_our_uname)) { int level = LOG_INFO; const char *op = crm_element_value(msg, F_CRM_TASK); /* make sure the election happens NOW */ if (fsa_state != S_ELECTION) { ha_msg_input_t new_input; level = LOG_WARNING; new_input.msg = msg; register_fsa_error_adv(C_FSA_INTERNAL, I_ELECTION, NULL, &new_input, __FUNCTION__); } do_crm_log(level, "Another DC detected: %s (op=%s)", from, op); goto done; } } } else { const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); if (safe_str_eq(sys_to, CRM_SYSTEM_DC)) { return; } } /* crm_log_xml_trace("HA[inbound]", msg); */ route_message(C_HA_MESSAGE, msg); done: trigger_fsa(fsa_source); } /*! * \internal * \brief Check whether a node is online * * \param[in] node Node to check * * \retval -1 if completely dead * \retval 0 if partially alive * \retval 1 if completely alive */ static int node_alive(const crm_node_t *node) { if (is_set(node->flags, crm_remote_node)) { // Pacemaker Remote nodes can't be partially alive return safe_str_eq(node->state, CRM_NODE_MEMBER)? 1: -1; } else if (crm_is_peer_active(node)) { // Completely up cluster node: both cluster member and peer return 1; } else if (is_not_set(node->processes, crm_get_cluster_proc()) && safe_str_neq(node->state, CRM_NODE_MEMBER)) { // Completely down cluster node: neither cluster member nor peer return -1; } // Partially up cluster node: only cluster member or only peer return 0; } #define state_text(state) ((state)? (const char *)(state) : "in unknown state") void peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data) { uint32_t old = 0; bool appeared = FALSE; bool is_remote = is_set(node->flags, crm_remote_node); /* The controller waits to receive some information from the membership * layer before declaring itself operational. If this is being called for a * cluster node, indicate that we have it. */ if (!is_remote) { set_bit(fsa_input_register, R_PEER_DATA); } if (node->uname == NULL) { return; } switch (type) { case crm_status_uname: /* If we've never seen the node, then it also won't be in the status section */ crm_info("%s node %s is now %s", (is_remote? "Remote" : "Cluster"), node->uname, state_text(node->state)); return; case crm_status_nstate: /* This callback should not be called unless the state actually * changed, but here's a failsafe just in case. */ CRM_CHECK(safe_str_neq(data, node->state), return); crm_info("%s node %s is now %s (was %s)", (is_remote? "Remote" : "Cluster"), node->uname, state_text(node->state), state_text(data)); if (safe_str_eq(CRM_NODE_MEMBER, node->state)) { appeared = TRUE; if (!is_remote) { remove_stonith_cleanup(node->uname); } } else { controld_remove_voter(node->uname); } crmd_alert_node_event(node); break; case crm_status_processes: CRM_CHECK(data != NULL, return); old = *(const uint32_t *)data; appeared = is_set(node->processes, crm_get_cluster_proc()); crm_info("Node %s is %s a peer " CRM_XS " DC=%s old=0x%07x new=0x%07x", node->uname, (appeared? "now" : "no longer"), (AM_I_DC? "true" : (fsa_our_dc? fsa_our_dc : "")), old, node->processes); if (is_not_set((node->processes ^ old), crm_get_cluster_proc())) { /* Peer status did not change. This should not be possible, * since we don't track process flags other than peer status. */ crm_trace("Process flag 0x%7x did not change from 0x%7x to 0x%7x", crm_get_cluster_proc(), old, node->processes); return; } if (!appeared) { controld_remove_voter(node->uname); } if (is_not_set(fsa_input_register, R_CIB_CONNECTED)) { crm_trace("Ignoring peer status change because not connected to CIB"); return; } else if (fsa_state == S_STOPPING) { crm_trace("Ignoring peer status change because stopping"); return; } if (safe_str_eq(node->uname, fsa_our_uname) && !appeared) { /* Did we get evicted? */ crm_notice("Our peer connection failed"); register_fsa_input(C_CRMD_STATUS_CALLBACK, I_ERROR, NULL); } else if (safe_str_eq(node->uname, fsa_our_dc) && crm_is_peer_active(node) == FALSE) { /* Did the DC leave us? */ crm_notice("Our peer on the DC (%s) is dead", fsa_our_dc); register_fsa_input(C_CRMD_STATUS_CALLBACK, I_ELECTION, NULL); /* @COMPAT DC < 1.1.13: If a DC shuts down normally, we don't * want to fence it. Newer DCs will send their shutdown request * to all peers, who will update the DC's expected state to * down, thus avoiding fencing. We can safely erase the DC's * transient attributes when it leaves in that case. However, * the only way to avoid fencing older DCs is to leave the * transient attributes intact until it rejoins. */ if (compare_version(fsa_our_dc_version, "3.0.9") > 0) { erase_status_tag(node->uname, XML_TAG_TRANSIENT_NODEATTRS, cib_scope_local); } } else if(AM_I_DC) { if (appeared) { te_trigger_stonith_history_sync(); } else { erase_status_tag(node->uname, XML_TAG_TRANSIENT_NODEATTRS, cib_scope_local); } } break; } if (AM_I_DC) { xmlNode *update = NULL; int flags = node_update_peer; int alive = node_alive(node); crm_action_t *down = match_down_event(node->uuid); crm_trace("Alive=%d, appeared=%d, down=%d", alive, appeared, (down? down->id : -1)); if (appeared && (alive > 0)) { register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } if (down) { const char *task = crm_element_value(down->xml, XML_LRM_ATTR_TASK); if (safe_str_eq(task, CRM_OP_FENCE)) { /* tengine_stonith_callback() confirms fence actions */ crm_trace("Updating CIB %s fencer reported fencing of %s complete", (down->confirmed? "after" : "before"), node->uname); } else if (!appeared && safe_str_eq(task, CRM_OP_SHUTDOWN)) { // Shutdown actions are immediately confirmed (i.e. no_wait) if (!is_remote) { flags |= node_update_join | node_update_expected; crmd_peer_down(node, FALSE); check_join_state(fsa_state, __FUNCTION__); } if (alive >= 0) { crm_info("%s of peer %s is in progress " CRM_XS " action=%d", task, node->uname, down->id); } else { crm_notice("%s of peer %s is complete " CRM_XS " action=%d", task, node->uname, down->id); update_graph(transition_graph, down); trigger_graph(); } } else { crm_trace("Node %s is %s, was expected to %s (op %d)", node->uname, ((alive > 0)? "alive" : ((alive < 0)? "dead" : "partially alive")), task, down->id); } } else if (appeared == FALSE) { crm_warn("Stonith/shutdown of node %s was not expected", node->uname); if (!is_remote) { crm_update_peer_join(__FUNCTION__, node, crm_join_none); check_join_state(fsa_state, __FUNCTION__); } abort_transition(INFINITY, tg_restart, "Node failure", NULL); fail_incompletable_actions(transition_graph, node->uuid); } else { crm_trace("Node %s came up, was not expected to be down", node->uname); } if (is_remote) { /* A pacemaker_remote node won't have its cluster status updated * in the CIB by membership-layer callbacks, so do it here. */ flags |= node_update_cluster; /* Trigger resource placement on newly integrated nodes */ if (appeared) { abort_transition(INFINITY, tg_restart, "pacemaker_remote node integrated", NULL); } } /* Update the CIB node state */ update = create_node_state_update(node, flags, NULL, __FUNCTION__); if (update == NULL) { crm_debug("Node state update not yet possible for %s", node->uname); } else { fsa_cib_anon_update(XML_CIB_TAG_STATUS, update); } free_xml(update); } trigger_fsa(fsa_source); } void crmd_cib_connection_destroy(gpointer user_data) { CRM_CHECK(user_data == fsa_cib_conn,;); crm_trace("Invoked"); trigger_fsa(fsa_source); fsa_cib_conn->state = cib_disconnected; if (is_set(fsa_input_register, R_CIB_CONNECTED) == FALSE) { crm_info("Connection to the CIB manager terminated"); return; } // @TODO This should trigger a reconnect, not a shutdown crm_crit("Lost connection to the CIB manager, shutting down"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); clear_bit(fsa_input_register, R_CIB_CONNECTED); return; } gboolean crm_fsa_trigger(gpointer user_data) { crm_trace("Invoked (queue len: %d)", g_list_length(fsa_message_queue)); s_crmd_fsa(C_FSA_INTERNAL); crm_trace("Exited (queue len: %d)", g_list_length(fsa_message_queue)); return TRUE; } diff --git a/daemons/controld/controld_control.c b/daemons/controld/controld_control.c index 89b5b5dc41..e99d605988 100644 --- a/daemons/controld/controld_control.c +++ b/daemons/controld/controld_control.c @@ -1,899 +1,878 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include 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; /* 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"); set_bit(fsa_input_register, 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) { set_bit(fsa_input_register, R_HA_DISCONNECTED); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } populate_cib_nodes(node_update_none, __FUNCTION__); clear_bit(fsa_input_register, 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), __FUNCTION__); } } /* 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 */ set_bit(fsa_input_register, R_SHUTDOWN); - - if (stonith_api) { - /* Prevent it from coming up again */ - clear_bit(fsa_input_register, R_ST_REQUIRED); - - crm_info("Disconnecting from fencer"); - stonith_api->cmds->disconnect(stonith_api); - } + 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; set_bit(fsa_input_register, R_SHUTDOWN); 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); /* set_bit(fsa_input_register, R_STAYDOWN); */ 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 char *te_client_id; void crmd_fast_exit(crm_exit_t exit_code) { if (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) && is_set(fsa_input_register, R_IN_RECOVERY)) { crm_err("Could not recover from internal error"); exit_code = CRM_EX_ERROR; } crm_exit(exit_code); } crm_exit_t crmd_exit(crm_exit_t exit_code) { GListPtr 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 */ set_bit(fsa_input_register, 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(); pe_subsystem_free(); - - if(stonith_api) { - crm_trace("Disconnecting fencing API"); - clear_bit(fsa_input_register, R_ST_REQUIRED); - stonith_api->cmds->free(stonith_api); stonith_api = NULL; - } + 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); } clear_bit(fsa_input_register, 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. */ cib_free_callbacks(fsa_cib_conn); fsa_cib_conn->cmds->signoff(fsa_cib_conn); verify_stopped(fsa_state, LOG_WARNING); clear_bit(fsa_input_register, 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(stonith_reconnect); stonith_reconnect = NULL; mainloop_destroy_trigger(transition_trigger); transition_trigger = NULL; crm_client_cleanup(); crm_peer_destroy(); crm_timer_stop(transition_timer); crm_timer_stop(integration_timer); crm_timer_stop(finalization_timer); crm_timer_stop(election_trigger); crm_timer_stop(shutdown_escalation_timer); crm_timer_stop(wait_timer); crm_timer_stop(recheck_timer); free(transition_timer); transition_timer = NULL; free(integration_timer); integration_timer = NULL; free(finalization_timer); finalization_timer = NULL; free(election_trigger); election_trigger = NULL; free(shutdown_escalation_timer); shutdown_escalation_timer = NULL; free(wait_timer); wait_timer = NULL; free(recheck_timer); recheck_timer = NULL; 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(te_client_id); te_client_id = NULL; free(fsa_pe_ref); fsa_pe_ref = 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) { int was_error = 0; 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(); /* set up the timers */ transition_timer = calloc(1, sizeof(fsa_timer_t)); integration_timer = calloc(1, sizeof(fsa_timer_t)); finalization_timer = calloc(1, sizeof(fsa_timer_t)); election_trigger = calloc(1, sizeof(fsa_timer_t)); shutdown_escalation_timer = calloc(1, sizeof(fsa_timer_t)); wait_timer = calloc(1, sizeof(fsa_timer_t)); recheck_timer = calloc(1, sizeof(fsa_timer_t)); if (election_trigger != NULL) { election_trigger->source_id = 0; election_trigger->period_ms = -1; election_trigger->fsa_input = I_DC_TIMEOUT; election_trigger->callback = crm_timer_popped; election_trigger->repeat = FALSE; } else { was_error = TRUE; } if (transition_timer != NULL) { transition_timer->source_id = 0; transition_timer->period_ms = -1; transition_timer->fsa_input = I_PE_CALC; transition_timer->callback = crm_timer_popped; transition_timer->repeat = FALSE; } else { was_error = TRUE; } if (integration_timer != NULL) { integration_timer->source_id = 0; integration_timer->period_ms = -1; integration_timer->fsa_input = I_INTEGRATED; integration_timer->callback = crm_timer_popped; integration_timer->repeat = FALSE; } else { was_error = TRUE; } if (finalization_timer != NULL) { finalization_timer->source_id = 0; finalization_timer->period_ms = -1; finalization_timer->fsa_input = I_FINALIZED; finalization_timer->callback = crm_timer_popped; finalization_timer->repeat = FALSE; /* for possible enabling... a bug in the join protocol left * a slave in S_PENDING while we think it's in S_NOT_DC * * raising I_FINALIZED put us into a transition loop which is * never resolved. * in this loop we continually send probes which the node * NACK's because it's in S_PENDING * * if we have nodes where the cluster layer is active but the * CRM is not... then this will be handled in the * integration phase */ finalization_timer->fsa_input = I_ELECTION; } else { was_error = TRUE; } if (shutdown_escalation_timer != NULL) { shutdown_escalation_timer->source_id = 0; shutdown_escalation_timer->period_ms = -1; shutdown_escalation_timer->fsa_input = I_STOP; shutdown_escalation_timer->callback = crm_timer_popped; shutdown_escalation_timer->repeat = FALSE; } else { was_error = TRUE; } if (wait_timer != NULL) { wait_timer->source_id = 0; wait_timer->period_ms = 2000; wait_timer->fsa_input = I_NULL; wait_timer->callback = crm_timer_popped; wait_timer->repeat = FALSE; } else { was_error = TRUE; } if (recheck_timer != NULL) { recheck_timer->source_id = 0; recheck_timer->period_ms = -1; recheck_timer->fsa_input = I_PE_CALC; recheck_timer->callback = crm_timer_popped; recheck_timer->repeat = FALSE; } else { was_error = TRUE; } if (was_error) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } static int32_t crmd_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { crm_trace("Connection %p", c); if (crm_client_new(c, uid, gid) == NULL) { return -EIO; } return 0; } static void crmd_ipc_created(qb_ipcs_connection_t * c) { crm_trace("Connection %p", c); } static int32_t crmd_ipc_dispatch(qb_ipcs_connection_t * c, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; crm_client_t *client = crm_client_get(c); xmlNode *msg = crm_ipcs_recv(client, data, size, &id, &flags); crm_trace("Invoked: %s", crm_client_name(client)); crm_ipcs_send_ack(client, id, flags, "ack", __FUNCTION__, __LINE__); if (msg == NULL) { return 0; } #if ENABLE_ACL CRM_ASSERT(client->user != NULL); crm_acl_get_set_user(msg, F_CRM_USER, client->user); #endif crm_trace("Processing msg from %s", crm_client_name(client)); crm_log_xml_trace(msg, "controller[inbound]"); crm_xml_add(msg, F_CRM_SYS_FROM, client->id); if (crmd_authorize_message(msg, client, NULL)) { route_message(C_IPC_MESSAGE, msg); } trigger_fsa(fsa_source); free_xml(msg); return 0; } static int32_t crmd_ipc_closed(qb_ipcs_connection_t * c) { crm_client_t *client = crm_client_get(c); if (client) { crm_trace("Disconnecting %sregistered client %s (%p/%p)", (client->userdata? "" : "un"), crm_client_name(client), c, client); free(client->userdata); crm_client_destroy(client); trigger_fsa(fsa_source); } 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 = crmd_ipc_accept, .connection_created = crmd_ipc_created, .msg_process = crmd_ipc_dispatch, .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 (is_set(fsa_input_register, R_MEMBERSHIP) == FALSE) { crm_info("Delaying start, no membership data (%.16llx)", R_MEMBERSHIP); crmd_fsa_stall(TRUE); return; } else if (is_set(fsa_input_register, R_LRM_CONNECTED) == FALSE) { crm_info("Delaying start, not connected to executor (%.16llx)", R_LRM_CONNECTED); crmd_fsa_stall(TRUE); return; } else if (is_set(fsa_input_register, R_CIB_CONNECTED) == FALSE) { crm_info("Delaying start, CIB not connected (%.16llx)", R_CIB_CONNECTED); crmd_fsa_stall(TRUE); return; } else if (is_set(fsa_input_register, R_READ_CONFIG) == FALSE) { crm_info("Delaying start, Config not read (%.16llx)", R_READ_CONFIG); crmd_fsa_stall(TRUE); return; } else if (is_set(fsa_input_register, R_PEER_DATA) == FALSE) { crm_info("Delaying start, No peer data (%.16llx)", R_PEER_DATA); crmd_fsa_stall(TRUE); return; } crm_debug("Init server comms"); ipcs = crmd_ipc_server_init(&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); } - - if (stonith_reconnect == NULL) { - int dummy; - - stonith_reconnect = mainloop_add_trigger(G_PRIORITY_LOW, te_connect_stonith, &dummy); - } - set_bit(fsa_input_register, R_ST_REQUIRED); - mainloop_set_trigger(stonith_reconnect); + controld_trigger_fencer_connect(); crm_notice("Pacemaker controller successfully started and accepting connections"); clear_bit(fsa_input_register, 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) { set_bit(fsa_input_register, R_IN_RECOVERY); crm_warn("Fast-tracking shutdown in response to errors"); register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL); } /* *INDENT-OFF* */ static pe_cluster_option crmd_opts[] = { /* name, old-name, validate, values, default, short description, long description */ { "dc-version", NULL, "string", NULL, "none", NULL, "Version of Pacemaker on the cluster's DC.", "Includes the hash which identifies the exact changeset it 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." }, { XML_CONFIG_ATTR_DC_DEADTIME, NULL, "time", NULL, "20s", &check_time, "How long to wait for a response from other nodes during startup.", "The \"correct\" value will depend on the speed/load of your network and the type of switches used." }, { XML_CONFIG_ATTR_RECHECK, NULL, "time", "Zero disables polling. Positive values are an interval in seconds (unless other SI units are specified. eg. 5min)", "15min", &check_timer, "Polling interval for time based changes to options, resource parameters and constraints.", "The Cluster is primarily event driven, however the configuration can have elements that change based on time." " To ensure these changes take effect, we can optionally poll the cluster's status for changes." }, { "load-threshold", NULL, "percentage", NULL, "80%", &check_utilization, "The maximum amount of system resources that should be used by nodes in the cluster", "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", &check_number, "The maximum number of jobs that can be scheduled per node. Defaults to 2x cores"}, { XML_CONFIG_ATTR_ELECTION_FAIL, NULL, "time", NULL, "2min", &check_timer, "*** Advanced Use Only ***.", "If need to adjust this value, it probably indicates the presence of a bug." }, { XML_CONFIG_ATTR_FORCE_QUIT, NULL, "time", NULL, "20min", &check_timer, "*** Advanced Use Only ***.", "If need to adjust this value, it probably indicates the presence of a bug." }, { "join-integration-timeout", "crmd-integration-timeout", "time", NULL, "3min", &check_timer, "*** Advanced Use Only ***", "If need to adjust this value, it probably indicates the presence of a bug" }, { "join-finalization-timeout", "crmd-finalization-timeout", "time", NULL, "30min", &check_timer, "*** 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", &check_timer, "*** Advanced Use Only *** Enabling this option will slow down cluster recovery under all conditions", "Delay cluster recovery for the configured interval to allow for additional/related events to occur.\n" "Useful if your configuration is sensitive to the order in which ping updates arrive." }, { "stonith-watchdog-timeout", NULL, "time", NULL, NULL, &check_sbd_timeout, "How long to wait before we can assume nodes are safely down", NULL }, { "stonith-max-attempts",NULL,"integer",NULL,"10",&check_positive_number, "How many times stonith can fail before it will no longer be attempted on a target" }, { "no-quorum-policy", NULL, "enum", "stop, freeze, ignore, suicide", "stop", &check_quorum, NULL, NULL }, }; /* *INDENT-ON* */ void crmd_metadata(void) { config_metadata("pacemaker-controld", "1.0", "controller properties", "Cluster properties used by Pacemaker's controller," " formerly known as crmd", crmd_opts, DIMOF(crmd_opts)); } static void verify_crmd_options(GHashTable * options) { verify_all_options(options, crmd_opts, DIMOF(crmd_opts)); } static const char * crmd_pref(GHashTable * options, const char *name) { return get_cluster_pref(options, crmd_opts, DIMOF(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"); set_bit(fsa_input_register, 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 = crm_str_table_new(); unpack_instance_attributes(crmconfig, crmconfig, XML_CIB_TAG_PROPSET, NULL, config_hash, CIB_OPTIONS_FIRST, FALSE, now); verify_crmd_options(config_hash); value = crmd_pref(config_hash, XML_CONFIG_ATTR_DC_DEADTIME); election_trigger->period_ms = crm_get_msec(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 (safe_str_eq(value, "suicide") && pcmk_locate_sbd()) { no_quorum_suicide_escalation = TRUE; } 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_get_msec(value); /* How long to declare an election over - even if not everyone voted */ crm_debug("Shutdown escalation occurs after: %dms", 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_timer->period_ms = crm_get_msec(value); crm_debug("Checking for expired actions every %dms", recheck_timer->period_ms); value = crmd_pref(config_hash, "transition-delay"); transition_timer->period_ms = crm_get_msec(value); value = crmd_pref(config_hash, "join-integration-timeout"); integration_timer->period_ms = crm_get_msec(value); value = crmd_pref(config_hash, "join-finalization-timeout"); finalization_timer->period_ms = crm_get_msec(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); set_bit(fsa_input_register, R_READ_CONFIG); crm_trace("Triggering FSA: %s", __FUNCTION__); 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)) { if (is_set(fsa_input_register, R_SHUTDOWN)) { crm_err("Escalating the shutdown"); register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL); } else { set_bit(fsa_input_register, R_SHUTDOWN); register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); if (shutdown_escalation_timer->period_ms < 1) { const char *value = crmd_pref(NULL, XML_CONFIG_ATTR_FORCE_QUIT); int msec = crm_get_msec(value); crm_debug("Using default shutdown escalation: %dms", msec); shutdown_escalation_timer->period_ms = msec; } /* can't rely on this... */ crm_notice("Shutting down cluster resource manager " CRM_XS " limit=%dms", shutdown_escalation_timer->period_ms); crm_timer_start(shutdown_escalation_timer); } } else { crm_info("exit from shutdown"); crmd_exit(CRM_EX_OK); } } diff --git a/daemons/controld/controld_election.c b/daemons/controld/controld_election.c index 5d6858c093..9e49c7bae5 100644 --- a/daemons/controld/controld_election.c +++ b/daemons/controld/controld_election.c @@ -1,281 +1,282 @@ /* - * Copyright 2004-2019 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include +#include #include #include #include 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 (safe_str_eq(uname, fsa_our_dc)) { /* 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_stop_election_timeout() { election_timeout_stop(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_get_msec(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 (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(is_not_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"); set_bit(fsa_input_register, R_THE_DC); execute_stonith_cleanup(); election_reset(fsa_election); set_bit(fsa_input_register, R_JOIN_OK); set_bit(fsa_input_register, 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, watchdog?"true":"false", FALSE, 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); update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "cluster-infrastructure", cluster_type, FALSE, NULL, NULL); #if SUPPORT_COROSYNC if (fsa_cluster_name == NULL && is_corosync_cluster()) { char *cluster_name = 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); } 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"); clear_bit(fsa_input_register, R_THE_DC); } 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 (is_set(fsa_input_register, R_SHUTDOWN)) { xmlNode *update = NULL; crm_node_t *node = crm_get_peer(0, fsa_our_uname); crm_update_peer_expected(__FUNCTION__, node, CRMD_JOINSTATE_DOWN); update = create_node_state_update(node, node_update_expected, NULL, __FUNCTION__); fsa_cib_anon_update(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_fencing.c b/daemons/controld/controld_fencing.c new file mode 100644 index 0000000000..92336e970c --- /dev/null +++ b/daemons/controld/controld_fencing.c @@ -0,0 +1,879 @@ +/* + * Copyright 2004-2019 the Pacemaker project contributors + * + * 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 + +#ifdef HAVE_SYS_REBOOT_H +# include +# include +#endif + +/* + * stonith failure counting + * + * We don't want to get stuck in a permanent fencing loop. Keep track of the + * number of fencing failures for each target node, and the most we'll restart a + * transition for. + */ + +struct st_fail_rec { + int count; +}; + +static unsigned long int stonith_max_attempts = 10; +static GHashTable *stonith_failures = NULL; + +void +update_stonith_max_attempts(const char *value) +{ + if (safe_str_eq(value, CRM_INFINITY_S)) { + stonith_max_attempts = CRM_SCORE_INFINITY; + } else { + stonith_max_attempts = crm_int_helper(value, NULL); + } +} + +static gboolean +too_many_st_failures(const char *target) +{ + GHashTableIter iter; + const char *key = NULL; + struct st_fail_rec *value = NULL; + + if (stonith_failures == NULL) { + return FALSE; + } + + if (target == NULL) { + g_hash_table_iter_init(&iter, stonith_failures); + while (g_hash_table_iter_next(&iter, (gpointer *) &key, + (gpointer *) &value)) { + + if (value->count >= stonith_max_attempts) { + target = (const char*)key; + goto too_many; + } + } + } else { + value = g_hash_table_lookup(stonith_failures, target); + if ((value != NULL) && (value->count >= stonith_max_attempts)) { + goto too_many; + } + } + return FALSE; + +too_many: + crm_warn("Too many failures (%d) to fence %s, giving up", + value->count, target); + return TRUE; +} + +/*! + * \internal + * \brief Reset a stonith fail count + * + * \param[in] target Name of node to reset, or NULL for all + */ +void +st_fail_count_reset(const char *target) +{ + if (stonith_failures == NULL) { + return; + } + + if (target) { + struct st_fail_rec *rec = NULL; + + rec = g_hash_table_lookup(stonith_failures, target); + if (rec) { + rec->count = 0; + } + } else { + GHashTableIter iter; + const char *key = NULL; + struct st_fail_rec *rec = NULL; + + g_hash_table_iter_init(&iter, stonith_failures); + while (g_hash_table_iter_next(&iter, (gpointer *) &key, + (gpointer *) &rec)) { + rec->count = 0; + } + } +} + +static void +st_fail_count_increment(const char *target) +{ + struct st_fail_rec *rec = NULL; + + if (stonith_failures == NULL) { + stonith_failures = crm_str_table_new(); + } + + rec = g_hash_table_lookup(stonith_failures, target); + if (rec) { + rec->count++; + } else { + rec = malloc(sizeof(struct st_fail_rec)); + if(rec == NULL) { + return; + } + + rec->count = 1; + g_hash_table_insert(stonith_failures, strdup(target), rec); + } +} + +/* end stonith fail count functions */ + + +static void +cib_fencing_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, + void *user_data) +{ + if (rc < pcmk_ok) { + crm_err("Fencing update %d for %s: failed - %s (%d)", + call_id, (char *)user_data, pcmk_strerror(rc), rc); + crm_log_xml_warn(msg, "Failed update"); + abort_transition(INFINITY, tg_shutdown, "CIB update failed", NULL); + + } else { + crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data); + } +} + +static void +send_stonith_update(crm_action_t *action, const char *target, const char *uuid) +{ + int rc = pcmk_ok; + crm_node_t *peer = NULL; + + /* We (usually) rely on the membership layer to do node_update_cluster, + * and the peer status callback to do node_update_peer, because the node + * might have already rejoined before we get the stonith result here. + */ + int flags = node_update_join | node_update_expected; + + /* zero out the node-status & remove all LRM status info */ + xmlNode *node_state = NULL; + + CRM_CHECK(target != NULL, return); + CRM_CHECK(uuid != NULL, return); + + /* Make sure the membership and join caches are accurate */ + peer = crm_get_peer_full(0, target, CRM_GET_PEER_ANY); + + CRM_CHECK(peer != NULL, return); + + if (peer->state == NULL) { + /* Usually, we rely on the membership layer to update the cluster state + * in the CIB. However, if the node has never been seen, do it here, so + * the node is not considered unclean. + */ + flags |= node_update_cluster; + } + + if (peer->uuid == NULL) { + crm_info("Recording uuid '%s' for node '%s'", uuid, target); + peer->uuid = strdup(uuid); + } + + crmd_peer_down(peer, TRUE); + + /* Generate a node state update for the CIB */ + node_state = create_node_state_update(peer, flags, NULL, __FUNCTION__); + + /* we have to mark whether or not remote nodes have already been fenced */ + if (peer->flags & crm_remote_node) { + time_t now = time(NULL); + char *now_s = crm_itoa(now); + crm_xml_add(node_state, XML_NODE_IS_FENCED, now_s); + free(now_s); + } + + /* Force our known ID */ + crm_xml_add(node_state, XML_ATTR_UUID, uuid); + + rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, node_state, + cib_quorum_override | cib_scope_local | cib_can_create); + + /* Delay processing the trigger until the update completes */ + crm_debug("Sending fencing update %d for %s", rc, target); + fsa_register_cib_callback(rc, FALSE, strdup(target), cib_fencing_updated); + + /* Make sure it sticks */ + /* fsa_cib_conn->cmds->bump_epoch(fsa_cib_conn, cib_quorum_override|cib_scope_local); */ + + erase_status_tag(peer->uname, XML_CIB_TAG_LRM, cib_scope_local); + erase_status_tag(peer->uname, XML_TAG_TRANSIENT_NODEATTRS, cib_scope_local); + + free_xml(node_state); + return; +} + +/*! + * \internal + * \brief Abort transition due to stonith failure + * + * \param[in] abort_action Whether to restart or stop transition + * \param[in] target Don't restart if this (NULL for any) has too many failures + * \param[in] reason Log this stonith action XML as abort reason (or NULL) + */ +static void +abort_for_stonith_failure(enum transition_action abort_action, + const char *target, xmlNode *reason) +{ + /* If stonith repeatedly fails, we eventually give up on starting a new + * transition for that reason. + */ + if ((abort_action != tg_stop) && too_many_st_failures(target)) { + abort_action = tg_stop; + } + abort_transition(INFINITY, abort_action, "Stonith failed", reason); +} + + +/* + * stonith cleanup list + * + * If the DC is shot, proper notifications might not go out. + * The stonith cleanup list allows the cluster to (re-)send + * notifications once a new DC is elected. + */ + +static GListPtr stonith_cleanup_list = NULL; + +/*! + * \internal + * \brief Add a node to the stonith cleanup list + * + * \param[in] target Name of node to add + */ +void +add_stonith_cleanup(const char *target) { + stonith_cleanup_list = g_list_append(stonith_cleanup_list, strdup(target)); +} + +/*! + * \internal + * \brief Remove a node from the stonith cleanup list + * + * \param[in] Name of node to remove + */ +void +remove_stonith_cleanup(const char *target) +{ + GListPtr iter = stonith_cleanup_list; + + while (iter != NULL) { + GListPtr tmp = iter; + char *iter_name = tmp->data; + + iter = iter->next; + if (safe_str_eq(target, iter_name)) { + crm_trace("Removing %s from the cleanup list", iter_name); + stonith_cleanup_list = g_list_delete_link(stonith_cleanup_list, tmp); + free(iter_name); + } + } +} + +/*! + * \internal + * \brief Purge all entries from the stonith cleanup list + */ +void +purge_stonith_cleanup() +{ + if (stonith_cleanup_list) { + GListPtr iter = NULL; + + for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { + char *target = iter->data; + + crm_info("Purging %s from stonith cleanup list", target); + free(target); + } + g_list_free(stonith_cleanup_list); + stonith_cleanup_list = NULL; + } +} + +/*! + * \internal + * \brief Send stonith updates for all entries in cleanup list, then purge it + */ +void +execute_stonith_cleanup() +{ + GListPtr iter; + + for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { + char *target = iter->data; + crm_node_t *target_node = crm_get_peer(0, target); + const char *uuid = crm_peer_uuid(target_node); + + crm_notice("Marking %s, target of a previous stonith action, as clean", target); + send_stonith_update(NULL, target, uuid); + free(target); + } + g_list_free(stonith_cleanup_list); + stonith_cleanup_list = NULL; +} + +/* end stonith cleanup list functions */ + + +/* stonith API client + * + * Functions that need to interact directly with the fencer via its API + */ + +static stonith_t *stonith_api = NULL; +static crm_trigger_t *stonith_reconnect = NULL; +static char *te_client_id = NULL; + +static gboolean +fail_incompletable_stonith(crm_graph_t *graph) +{ + GListPtr lpc = NULL; + const char *task = NULL; + xmlNode *last_action = NULL; + + if (graph == NULL) { + return FALSE; + } + + for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { + GListPtr lpc2 = NULL; + synapse_t *synapse = (synapse_t *) lpc->data; + + if (synapse->confirmed) { + continue; + } + + for (lpc2 = synapse->actions; lpc2 != NULL; lpc2 = lpc2->next) { + crm_action_t *action = (crm_action_t *) lpc2->data; + + if (action->type != action_type_crm || action->confirmed) { + continue; + } + + task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); + if (task && safe_str_eq(task, CRM_OP_FENCE)) { + action->failed = TRUE; + last_action = action->xml; + update_graph(graph, action); + crm_notice("Failing action %d (%s): fencer terminated", + action->id, ID(action->xml)); + } + } + } + + if (last_action != NULL) { + crm_warn("Fencer failure resulted in unrunnable actions"); + abort_for_stonith_failure(tg_restart, NULL, last_action); + return TRUE; + } + + return FALSE; +} + +static void +tengine_stonith_connection_destroy(stonith_t *st, stonith_event_t *e) +{ + if (is_set(fsa_input_register, R_ST_REQUIRED)) { + crm_crit("Fencing daemon connection failed"); + mainloop_set_trigger(stonith_reconnect); + + } else { + crm_info("Fencing daemon disconnected"); + } + + if (stonith_api) { + stonith_api->state = stonith_disconnected; + } + + if (AM_I_DC) { + fail_incompletable_stonith(transition_graph); + trigger_graph(); + } +} + +static void +tengine_stonith_notify(stonith_t *st, stonith_event_t *st_event) +{ + if (te_client_id == NULL) { + te_client_id = crm_strdup_printf("%s.%lu", crm_system_name, + (unsigned long) getpid()); + } + + if (st_event == NULL) { + crm_err("Notify data not found"); + return; + } + + crmd_alert_fencing_op(st_event); + + if ((st_event->result == pcmk_ok) && safe_str_eq("on", st_event->action)) { + crm_notice("%s was successfully unfenced by %s (at the request of %s)", + st_event->target, + st_event->executioner? st_event->executioner : "", + st_event->origin); + /* TODO: Hook up st_event->device */ + return; + + } else if (safe_str_eq("on", st_event->action)) { + crm_err("Unfencing of %s by %s failed: %s (%d)", + st_event->target, + st_event->executioner? st_event->executioner : "", + pcmk_strerror(st_event->result), st_event->result); + return; + + } else if ((st_event->result == pcmk_ok) + && crm_str_eq(st_event->target, fsa_our_uname, TRUE)) { + + crm_crit("We were allegedly just fenced by %s for %s!", + st_event->executioner? st_event->executioner : "", + st_event->origin); /* Dumps blackbox if enabled */ + + qb_log_fini(); /* Try to get the above log message to disk - somehow */ + + /* Get out ASAP and do not come back up. + * + * Triggering a reboot is also not the worst idea either since + * the rest of the cluster thinks we're safely down + */ + +#ifdef RB_HALT_SYSTEM + reboot(RB_HALT_SYSTEM); +#endif + + /* + * If reboot() fails or is not supported, coming back up will + * probably lead to a situation where the other nodes set our + * status to 'lost' because of the fencing callback and will + * discard subsequent election votes with: + * + * Election 87 (current: 5171, owner: 103): Processed vote from east-03 (Peer is not part of our cluster) + * + * So just stay dead, something is seriously messed up anyway. + * + */ + exit(CRM_EX_FATAL); // None of our wrappers since we already called qb_log_fini() + return; + } + + /* Update the count of stonith failures for this target, in case we become + * DC later. The current DC has already updated its fail count in + * tengine_stonith_callback(). + */ + if (!AM_I_DC && safe_str_eq(st_event->operation, T_STONITH_NOTIFY_FENCE)) { + if (st_event->result == pcmk_ok) { + st_fail_count_reset(st_event->target); + } else { + st_fail_count_increment(st_event->target); + } + } + + crm_notice("Peer %s was%s terminated (%s) by %s on behalf of %s: %s " + CRM_XS " initiator=%s ref=%s", + st_event->target, st_event->result == pcmk_ok ? "" : " not", + st_event->action, + st_event->executioner ? st_event->executioner : "", + (st_event->client_origin? st_event->client_origin : ""), + pcmk_strerror(st_event->result), + st_event->origin, st_event->id); + + if (st_event->result == pcmk_ok) { + crm_node_t *peer = crm_find_known_peer_full(0, st_event->target, CRM_GET_PEER_ANY); + const char *uuid = NULL; + gboolean we_are_executioner = safe_str_eq(st_event->executioner, fsa_our_uname); + + if (peer == NULL) { + return; + } + + uuid = crm_peer_uuid(peer); + + crm_trace("target=%s dc=%s", st_event->target, fsa_our_dc); + if(AM_I_DC) { + /* The DC always sends updates */ + send_stonith_update(NULL, st_event->target, uuid); + + /* @TODO Ideally, at this point, we'd check whether the fenced node + * hosted any guest nodes, and call remote_node_down() for them. + * Unfortunately, the controller doesn't have a simple, reliable way + * to map hosts to guests. It might be possible to track this in the + * peer cache via crm_remote_peer_cache_refresh(). For now, we rely + * on the PE creating fence pseudo-events for the guests. + */ + + if (st_event->client_origin + && safe_str_neq(st_event->client_origin, te_client_id)) { + + /* Abort the current transition graph if it wasn't us + * that invoked stonith to fence someone + */ + crm_info("External fencing operation from %s fenced %s", st_event->client_origin, st_event->target); + abort_transition(INFINITY, tg_restart, "External Fencing Operation", NULL); + } + + /* Assume it was our leader if we don't currently have one */ + } else if (((fsa_our_dc == NULL) || safe_str_eq(fsa_our_dc, st_event->target)) + && is_not_set(peer->flags, crm_remote_node)) { + + crm_notice("Target %s our leader %s (recorded: %s)", + fsa_our_dc ? "was" : "may have been", st_event->target, + fsa_our_dc ? fsa_our_dc : ""); + + /* Given the CIB resyncing that occurs around elections, + * have one node update the CIB now and, if the new DC is different, + * have them do so too after the election + */ + if (we_are_executioner) { + send_stonith_update(NULL, st_event->target, uuid); + } + add_stonith_cleanup(st_event->target); + } + + /* If the target is a remote node, and we host its connection, + * immediately fail all monitors so it can be recovered quickly. + * The connection won't necessarily drop when a remote node is fenced, + * so the failure might not otherwise be detected until the next poke. + */ + if (is_set(peer->flags, crm_remote_node)) { + remote_ra_fail(st_event->target); + } + + crmd_peer_down(peer, TRUE); + } +} + +/*! + * \brief Connect to fencer + * + * \param[in] user_data If NULL, retry failures now, otherwise retry in main loop + * + * \return TRUE + * \note If user_data is NULL, this will wait 2s between attempts, for up to + * 30 attempts, meaning the controller could be blocked as long as 58s. + */ +static gboolean +te_connect_stonith(gpointer user_data) +{ + int rc = pcmk_ok; + + if (stonith_api == NULL) { + stonith_api = stonith_api_new(); + } + + if (stonith_api->state != stonith_disconnected) { + crm_trace("Already connected to fencer, no need to retry"); + return TRUE; + } + + if (user_data == NULL) { + // Blocking (retry failures now until successful) + rc = stonith_api_connect_retry(stonith_api, crm_system_name, 30); + if (rc != pcmk_ok) { + crm_err("Could not connect to fencer in 30 attempts: %s " + CRM_XS " rc=%d", pcmk_strerror(rc), rc); + } + } else { + // Non-blocking (retry failures later in main loop) + rc = stonith_api->cmds->connect(stonith_api, crm_system_name, NULL); + if (rc != pcmk_ok) { + if (is_set(fsa_input_register, R_ST_REQUIRED)) { + crm_err("Fencer connection failed (will retry): %s " + CRM_XS " rc=%d", pcmk_strerror(rc), rc); + mainloop_set_trigger(stonith_reconnect); + } else { + crm_info("Fencer connection failed (ignoring because no longer required): %s " + CRM_XS " rc=%d", pcmk_strerror(rc), rc); + } + return TRUE; + } + } + + if (rc == pcmk_ok) { + stonith_api->cmds->register_notification(stonith_api, + T_STONITH_NOTIFY_DISCONNECT, + tengine_stonith_connection_destroy); + stonith_api->cmds->register_notification(stonith_api, + T_STONITH_NOTIFY_FENCE, + tengine_stonith_notify); + } + return TRUE; +} + +/*! + \internal + \brief Schedule fencer connection attempt in main loop +*/ +void +controld_trigger_fencer_connect() +{ + if (stonith_reconnect == NULL) { + stonith_reconnect = mainloop_add_trigger(G_PRIORITY_LOW, + te_connect_stonith, + GINT_TO_POINTER(TRUE)); + } + set_bit(fsa_input_register, R_ST_REQUIRED); + mainloop_set_trigger(stonith_reconnect); +} + +void +controld_disconnect_fencer(bool destroy) +{ + if (stonith_api) { + // Prevent fencer connection from coming up again + clear_bit(fsa_input_register, R_ST_REQUIRED); + + stonith_api->cmds->disconnect(stonith_api); + } + if (destroy) { + if (stonith_api) { + stonith_api->cmds->free(stonith_api); + stonith_api = NULL; + } + if (stonith_reconnect) { + mainloop_destroy_trigger(stonith_reconnect); + stonith_reconnect = NULL; + } + if (te_client_id) { + free(te_client_id); + te_client_id = NULL; + } + } +} + +static gboolean +do_stonith_history_sync(gpointer user_data) +{ + if (stonith_api && (stonith_api->state != stonith_disconnected)) { + stonith_history_t *history = NULL; + + stonith_api->cmds->history(stonith_api, + st_opt_sync_call | st_opt_broadcast, + NULL, &history, 5); + stonith_history_free(history); + return TRUE; + } else { + crm_info("Skip triggering stonith history-sync as stonith is disconnected"); + return FALSE; + } +} + +static void +tengine_stonith_callback(stonith_t *stonith, stonith_callback_data_t *data) +{ + char *uuid = NULL; + int stonith_id = -1; + int transition_id = -1; + crm_action_t *action = NULL; + int call_id = data->call_id; + int rc = data->rc; + char *userdata = data->userdata; + + CRM_CHECK(userdata != NULL, return); + crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata, + pcmk_strerror(rc), rc); + + if (AM_I_DC == FALSE) { + return; + } + + /* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */ + /* op->call_id, op->optype, op->node_name, op->op_result, */ + /* (char *)op->node_list, op->private_data); */ + + /* filter out old STONITH actions */ + CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, NULL), + goto bail); + + if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid) + || transition_graph->id != transition_id) { + crm_info("Ignoring STONITH action initiated outside of the current transition"); + goto bail; + } + + action = get_action(stonith_id, FALSE); + if (action == NULL) { + crm_err("Stonith action not matched"); + goto bail; + } + + stop_te_timer(action->timer); + if (rc == pcmk_ok) { + const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); + const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); + const char *op = crm_meta_value(action->params, "stonith_action"); + + crm_info("Stonith operation %d for %s passed", call_id, target); + if (action->confirmed == FALSE) { + te_action_confirmed(action); + if (safe_str_eq("on", op)) { + const char *value = NULL; + char *now = crm_itoa(time(NULL)); + + update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE); + free(now); + + value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL); + update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE); + + value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE); + update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE); + + } else if (action->sent_update == FALSE) { + send_stonith_update(action, target, uuid); + action->sent_update = TRUE; + } + } + st_fail_count_reset(target); + + } else { + const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); + enum transition_action abort_action = tg_restart; + + action->failed = TRUE; + crm_notice("Stonith operation %d for %s failed (%s): aborting transition.", + call_id, target, pcmk_strerror(rc)); + + /* If no fence devices were available, there's no use in immediately + * checking again, so don't start a new transition in that case. + */ + if (rc == -ENODEV) { + crm_warn("No devices found in cluster to fence %s, giving up", + target); + abort_action = tg_stop; + } + + /* Increment the fail count now, so abort_for_stonith_failure() can + * check it. Non-DC nodes will increment it in tengine_stonith_notify(). + */ + st_fail_count_increment(target); + abort_for_stonith_failure(abort_action, target, NULL); + } + + update_graph(transition_graph, action); + trigger_graph(); + + bail: + free(userdata); + free(uuid); + return; +} + +gboolean +te_fence_node(crm_graph_t *graph, crm_action_t *action) +{ + int rc = 0; + const char *id = NULL; + const char *uuid = NULL; + const char *target = NULL; + const char *type = NULL; + gboolean invalid_action = FALSE; + enum stonith_call_options options = st_opt_none; + + id = ID(action->xml); + target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); + uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); + type = crm_meta_value(action->params, "stonith_action"); + + CRM_CHECK(id != NULL, invalid_action = TRUE); + CRM_CHECK(uuid != NULL, invalid_action = TRUE); + CRM_CHECK(type != NULL, invalid_action = TRUE); + CRM_CHECK(target != NULL, invalid_action = TRUE); + + if (invalid_action) { + crm_log_xml_warn(action->xml, "BadAction"); + return FALSE; + } + + crm_notice("Requesting fencing (%s) of node %s " + CRM_XS " action=%s timeout=%d", + type, target, id, transition_graph->stonith_timeout); + + /* Passing NULL means block until we can connect... */ + te_connect_stonith(NULL); + + if (crmd_join_phase_count(crm_join_confirmed) == 1) { + options |= st_opt_allow_suicide; + } + + rc = stonith_api->cmds->fence(stonith_api, options, target, type, + transition_graph->stonith_timeout / 1000, 0); + + stonith_api->cmds->register_callback(stonith_api, rc, transition_graph->stonith_timeout / 1000, + st_opt_timeout_updates, + generate_transition_key(transition_graph->id, action->id, + 0, te_uuid), + "tengine_stonith_callback", tengine_stonith_callback); + + return TRUE; +} + +/* end stonith API client functions */ + + +/* + * stonith history synchronization + * + * Each node's fencer keeps track of a cluster-wide fencing history. When a node + * joins or leaves, we need to synchronize the history across all nodes. + */ + +static crm_trigger_t *stonith_history_sync_trigger = NULL; +static mainloop_timer_t *stonith_history_sync_timer = NULL; + +static gboolean +stonith_history_sync_set_trigger(gpointer user_data) +{ + mainloop_set_trigger(stonith_history_sync_trigger); + return FALSE; +} + +void +te_trigger_stonith_history_sync(void) +{ + /* trigger a sync in 5s to give more nodes the + * chance to show up so that we don't create + * unnecessary stonith-history-sync traffic + */ + + /* as we are finally checking the stonith-connection + * in do_stonith_history_sync we should be fine + * leaving stonith_history_sync_time & stonith_history_sync_trigger + * around + */ + if (stonith_history_sync_trigger == NULL) { + stonith_history_sync_trigger = + mainloop_add_trigger(G_PRIORITY_LOW, + do_stonith_history_sync, NULL); + } + + if(stonith_history_sync_timer == NULL) { + stonith_history_sync_timer = + mainloop_timer_add("history_sync", 5000, + FALSE, stonith_history_sync_set_trigger, + NULL); + } + crm_info("Fence history will be synchronized cluster-wide within 5 seconds"); + mainloop_timer_start(stonith_history_sync_timer); +} + +/* end stonith history synchronization functions */ diff --git a/daemons/controld/controld_fencing.h b/daemons/controld/controld_fencing.h new file mode 100644 index 0000000000..3ef537f6da --- /dev/null +++ b/daemons/controld/controld_fencing.h @@ -0,0 +1,34 @@ +/* + * Copyright 2004-2019 the Pacemaker project contributors + * + * The version control history for this file may have further details. + * + * This source code is licensed under the GNU Lesser General Public License + * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. + */ + +#ifndef CONTROLD_FENCING__H +# define CONTROLD_FENCING__H + +#include // bool +#include // crm_graph_t, crm_action_t + +// stonith fail counts +void st_fail_count_reset(const char * target); +void update_stonith_max_attempts(const char* value); + +// stonith API client +void controld_trigger_fencer_connect(void); +void controld_disconnect_fencer(bool destroy); +gboolean te_fence_node(crm_graph_t *graph, crm_action_t *action); + +// stonith cleanup list +void add_stonith_cleanup(const char *target); +void remove_stonith_cleanup(const char *target); +void purge_stonith_cleanup(void); +void execute_stonith_cleanup(void); + +// stonith history synchronization +void te_trigger_stonith_history_sync(void); + +#endif diff --git a/daemons/controld/controld_fsa.c b/daemons/controld/controld_fsa.c index 9eca530de1..dc1937fc81 100644 --- a/daemons/controld/controld_fsa.c +++ b/daemons/controld/controld_fsa.c @@ -1,660 +1,661 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include char *fsa_our_dc = NULL; cib_t *fsa_cib_conn = NULL; char *fsa_our_dc_version = NULL; char *fsa_our_uuid = NULL; char *fsa_our_uname = NULL; char *fsa_cluster_name = NULL; fsa_timer_t *wait_timer = NULL; // How long to wait before retrying a cib or executor connection fsa_timer_t *recheck_timer = NULL; // Periodically re-run scheduler to handle time-based actions fsa_timer_t *election_trigger = NULL; /* How long to wait at startup, or after an election, for the DC to make contact */ fsa_timer_t *transition_timer = NULL; /* How long to delay the start of a new transition with the expectation something else might happen too */ fsa_timer_t *integration_timer = NULL; fsa_timer_t *finalization_timer = NULL; fsa_timer_t *shutdown_escalation_timer = NULL; /* How long to wait for the DC to stop all resources and give us the all-clear to shut down */ gboolean do_fsa_stall = FALSE; long long fsa_input_register = 0; long long fsa_actions = A_NOTHING; enum crmd_fsa_state fsa_state = S_STARTING; extern uint highest_born_on; extern uint num_join_invites; extern void initialize_join(gboolean before); #define DOT_PREFIX "actions:trace: " #define do_dot_log(fmt, args...) crm_trace( fmt, ##args) long long do_state_transition(long long actions, enum crmd_fsa_state cur_state, enum crmd_fsa_state next_state, fsa_data_t * msg_data); void s_crmd_fsa_actions(fsa_data_t * fsa_data); void log_fsa_input(fsa_data_t * stored_msg); void init_dotfile(void); void init_dotfile(void) { do_dot_log(DOT_PREFIX "digraph \"g\" {"); do_dot_log(DOT_PREFIX " size = \"30,30\""); do_dot_log(DOT_PREFIX " graph ["); do_dot_log(DOT_PREFIX " fontsize = \"12\""); do_dot_log(DOT_PREFIX " fontname = \"Times-Roman\""); do_dot_log(DOT_PREFIX " fontcolor = \"black\""); do_dot_log(DOT_PREFIX " bb = \"0,0,398.922306,478.927856\""); do_dot_log(DOT_PREFIX " color = \"black\""); do_dot_log(DOT_PREFIX " ]"); do_dot_log(DOT_PREFIX " node ["); do_dot_log(DOT_PREFIX " fontsize = \"12\""); do_dot_log(DOT_PREFIX " fontname = \"Times-Roman\""); do_dot_log(DOT_PREFIX " fontcolor = \"black\""); do_dot_log(DOT_PREFIX " shape = \"ellipse\""); do_dot_log(DOT_PREFIX " color = \"black\""); do_dot_log(DOT_PREFIX " ]"); do_dot_log(DOT_PREFIX " edge ["); do_dot_log(DOT_PREFIX " fontsize = \"12\""); do_dot_log(DOT_PREFIX " fontname = \"Times-Roman\""); do_dot_log(DOT_PREFIX " fontcolor = \"black\""); do_dot_log(DOT_PREFIX " color = \"black\""); do_dot_log(DOT_PREFIX " ]"); do_dot_log(DOT_PREFIX "// special nodes"); do_dot_log(DOT_PREFIX " \"S_PENDING\" "); do_dot_log(DOT_PREFIX " ["); do_dot_log(DOT_PREFIX " color = \"blue\""); do_dot_log(DOT_PREFIX " fontcolor = \"blue\""); do_dot_log(DOT_PREFIX " ]"); do_dot_log(DOT_PREFIX " \"S_TERMINATE\" "); do_dot_log(DOT_PREFIX " ["); do_dot_log(DOT_PREFIX " color = \"red\""); do_dot_log(DOT_PREFIX " fontcolor = \"red\""); do_dot_log(DOT_PREFIX " ]"); do_dot_log(DOT_PREFIX "// DC only nodes"); do_dot_log(DOT_PREFIX " \"S_INTEGRATION\" [ fontcolor = \"green\" ]"); do_dot_log(DOT_PREFIX " \"S_POLICY_ENGINE\" [ fontcolor = \"green\" ]"); do_dot_log(DOT_PREFIX " \"S_TRANSITION_ENGINE\" [ fontcolor = \"green\" ]"); do_dot_log(DOT_PREFIX " \"S_RELEASE_DC\" [ fontcolor = \"green\" ]"); do_dot_log(DOT_PREFIX " \"S_IDLE\" [ fontcolor = \"green\" ]"); } static void do_fsa_action(fsa_data_t * fsa_data, long long an_action, void (*function) (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)) { fsa_actions &= ~an_action; crm_trace(DOT_PREFIX "\t// %s", fsa_action2string(an_action)); function(an_action, fsa_data->fsa_cause, fsa_state, fsa_data->fsa_input, fsa_data); } static long long startup_actions = A_STARTUP | A_CIB_START | A_LRM_CONNECT | A_HA_CONNECT | A_READCONFIG | A_STARTED | A_CL_JOIN_QUERY; // A_LOG, A_WARN, A_ERROR void do_log(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) { unsigned log_type = LOG_TRACE; if (action & A_LOG) { log_type = LOG_INFO; } else if (action & A_WARN) { log_type = LOG_WARNING; } else if (action & A_ERROR) { log_type = LOG_ERR; } do_crm_log(log_type, "Input %s received in state %s from %s", fsa_input2string(msg_data->fsa_input), fsa_state2string(cur_state), msg_data->origin); if (msg_data->data_type == fsa_dt_ha_msg) { ha_msg_input_t *input = fsa_typed_data(msg_data->data_type); crm_log_xml_debug(input->msg, __FUNCTION__); } else if (msg_data->data_type == fsa_dt_xml) { xmlNode *input = fsa_typed_data(msg_data->data_type); crm_log_xml_debug(input, __FUNCTION__); } else if (msg_data->data_type == fsa_dt_lrm) { lrmd_event_data_t *input = fsa_typed_data(msg_data->data_type); do_crm_log(log_type, "Resource %s: Call ID %d returned %d (%d)." " New status if rc=0: %s", input->rsc_id, input->call_id, input->rc, input->op_status, (char *)input->user_data); } } enum crmd_fsa_state s_crmd_fsa(enum crmd_fsa_cause cause) { fsa_data_t *fsa_data = NULL; long long register_copy = fsa_input_register; long long new_actions = A_NOTHING; enum crmd_fsa_state last_state; crm_trace("FSA invoked with Cause: %s\tState: %s", fsa_cause2string(cause), fsa_state2string(fsa_state)); fsa_dump_actions(fsa_actions, "Initial"); do_fsa_stall = FALSE; if (is_message() == FALSE && fsa_actions != A_NOTHING) { /* fake the first message so we can get into the loop */ fsa_data = calloc(1, sizeof(fsa_data_t)); fsa_data->fsa_input = I_NULL; fsa_data->fsa_cause = C_FSA_INTERNAL; fsa_data->origin = __FUNCTION__; fsa_data->data_type = fsa_dt_none; fsa_message_queue = g_list_append(fsa_message_queue, fsa_data); fsa_data = NULL; } while (is_message() && do_fsa_stall == FALSE) { crm_trace("Checking messages (%d remaining)", g_list_length(fsa_message_queue)); fsa_data = get_message(); if(fsa_data == NULL) { continue; } log_fsa_input(fsa_data); /* add any actions back to the queue */ fsa_actions |= fsa_data->actions; fsa_dump_actions(fsa_data->actions, "Restored actions"); /* get the next batch of actions */ new_actions = crmd_fsa_actions[fsa_data->fsa_input][fsa_state]; fsa_actions |= new_actions; fsa_dump_actions(new_actions, "New actions"); if (fsa_data->fsa_input != I_NULL && fsa_data->fsa_input != I_ROUTER) { crm_debug("Processing %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); } /* logging : *before* the state is changed */ if (is_set(fsa_actions, A_ERROR)) { do_fsa_action(fsa_data, A_ERROR, do_log); } if (is_set(fsa_actions, A_WARN)) { do_fsa_action(fsa_data, A_WARN, do_log); } if (is_set(fsa_actions, A_LOG)) { do_fsa_action(fsa_data, A_LOG, do_log); } /* update state variables */ last_state = fsa_state; fsa_state = crmd_fsa_state[fsa_data->fsa_input][fsa_state]; /* * Remove certain actions during shutdown */ if (fsa_state == S_STOPPING || ((fsa_input_register & R_SHUTDOWN) == R_SHUTDOWN)) { clear_bit(fsa_actions, startup_actions); } /* * Hook for change of state. * Allows actions to be added or removed when entering a state */ if (last_state != fsa_state) { fsa_actions = do_state_transition(fsa_actions, last_state, fsa_state, fsa_data); } else { do_dot_log(DOT_PREFIX "\t// FSA input: State=%s \tCause=%s" " \tInput=%s \tOrigin=%s() \tid=%d", fsa_state2string(fsa_state), fsa_cause2string(fsa_data->fsa_cause), fsa_input2string(fsa_data->fsa_input), fsa_data->origin, fsa_data->id); } /* start doing things... */ s_crmd_fsa_actions(fsa_data); delete_fsa_input(fsa_data); fsa_data = NULL; } if (g_list_length(fsa_message_queue) > 0 || fsa_actions != A_NOTHING || do_fsa_stall) { crm_debug("Exiting the FSA: queue=%d, fsa_actions=0x%llx, stalled=%s", g_list_length(fsa_message_queue), fsa_actions, do_fsa_stall ? "true" : "false"); } else { crm_trace("Exiting the FSA"); } /* cleanup inputs? */ if (register_copy != fsa_input_register) { long long same = register_copy & fsa_input_register; fsa_dump_inputs(LOG_DEBUG, "Added", fsa_input_register ^ same); fsa_dump_inputs(LOG_DEBUG, "Removed", register_copy ^ same); } fsa_dump_actions(fsa_actions, "Remaining"); fsa_dump_queue(LOG_DEBUG); return fsa_state; } void s_crmd_fsa_actions(fsa_data_t * fsa_data) { /* * Process actions in order of priority but do only one * action at a time to avoid complicating the ordering. */ CRM_CHECK(fsa_data != NULL, return); while (fsa_actions != A_NOTHING && do_fsa_stall == FALSE) { /* regular action processing in order of action priority * * Make sure all actions that connect to required systems * are performed first */ if (fsa_actions & A_ERROR) { do_fsa_action(fsa_data, A_ERROR, do_log); } else if (fsa_actions & A_WARN) { do_fsa_action(fsa_data, A_WARN, do_log); } else if (fsa_actions & A_LOG) { do_fsa_action(fsa_data, A_LOG, do_log); /* get out of here NOW! before anything worse happens */ } else if (fsa_actions & A_EXIT_1) { do_fsa_action(fsa_data, A_EXIT_1, do_exit); /* sub-system restart */ } else if ((fsa_actions & O_LRM_RECONNECT) == O_LRM_RECONNECT) { do_fsa_action(fsa_data, O_LRM_RECONNECT, do_lrm_control); } else if ((fsa_actions & O_CIB_RESTART) == O_CIB_RESTART) { do_fsa_action(fsa_data, O_CIB_RESTART, do_cib_control); } else if ((fsa_actions & O_PE_RESTART) == O_PE_RESTART) { do_fsa_action(fsa_data, O_PE_RESTART, do_pe_control); } else if ((fsa_actions & O_TE_RESTART) == O_TE_RESTART) { do_fsa_action(fsa_data, O_TE_RESTART, do_te_control); /* essential start tasks */ } else if (fsa_actions & A_STARTUP) { do_fsa_action(fsa_data, A_STARTUP, do_startup); } else if (fsa_actions & A_CIB_START) { do_fsa_action(fsa_data, A_CIB_START, do_cib_control); } else if (fsa_actions & A_HA_CONNECT) { do_fsa_action(fsa_data, A_HA_CONNECT, do_ha_control); } else if (fsa_actions & A_READCONFIG) { do_fsa_action(fsa_data, A_READCONFIG, do_read_config); /* sub-system start/connect */ } else if (fsa_actions & A_LRM_CONNECT) { do_fsa_action(fsa_data, A_LRM_CONNECT, do_lrm_control); } else if (fsa_actions & A_TE_START) { do_fsa_action(fsa_data, A_TE_START, do_te_control); } else if (fsa_actions & A_PE_START) { do_fsa_action(fsa_data, A_PE_START, do_pe_control); /* Timers */ /* else if(fsa_actions & O_DC_TIMER_RESTART) { do_fsa_action(fsa_data, O_DC_TIMER_RESTART, do_timer_control) */ ; } else if (fsa_actions & A_DC_TIMER_STOP) { do_fsa_action(fsa_data, A_DC_TIMER_STOP, do_timer_control); } else if (fsa_actions & A_INTEGRATE_TIMER_STOP) { do_fsa_action(fsa_data, A_INTEGRATE_TIMER_STOP, do_timer_control); } else if (fsa_actions & A_INTEGRATE_TIMER_START) { do_fsa_action(fsa_data, A_INTEGRATE_TIMER_START, do_timer_control); } else if (fsa_actions & A_FINALIZE_TIMER_STOP) { do_fsa_action(fsa_data, A_FINALIZE_TIMER_STOP, do_timer_control); } else if (fsa_actions & A_FINALIZE_TIMER_START) { do_fsa_action(fsa_data, A_FINALIZE_TIMER_START, do_timer_control); /* * Highest priority actions */ } else if (fsa_actions & A_MSG_ROUTE) { do_fsa_action(fsa_data, A_MSG_ROUTE, do_msg_route); } else if (fsa_actions & A_RECOVER) { do_fsa_action(fsa_data, A_RECOVER, do_recover); } else if (fsa_actions & A_CL_JOIN_RESULT) { do_fsa_action(fsa_data, A_CL_JOIN_RESULT, do_cl_join_finalize_respond); } else if (fsa_actions & A_CL_JOIN_REQUEST) { do_fsa_action(fsa_data, A_CL_JOIN_REQUEST, do_cl_join_offer_respond); } else if (fsa_actions & A_SHUTDOWN_REQ) { do_fsa_action(fsa_data, A_SHUTDOWN_REQ, do_shutdown_req); } else if (fsa_actions & A_ELECTION_VOTE) { do_fsa_action(fsa_data, A_ELECTION_VOTE, do_election_vote); } else if (fsa_actions & A_ELECTION_COUNT) { do_fsa_action(fsa_data, A_ELECTION_COUNT, do_election_count_vote); } else if (fsa_actions & A_LRM_EVENT) { do_fsa_action(fsa_data, A_LRM_EVENT, do_lrm_event); /* * High priority actions */ } else if (fsa_actions & A_STARTED) { do_fsa_action(fsa_data, A_STARTED, do_started); } else if (fsa_actions & A_CL_JOIN_QUERY) { do_fsa_action(fsa_data, A_CL_JOIN_QUERY, do_cl_join_query); } else if (fsa_actions & A_DC_TIMER_START) { do_fsa_action(fsa_data, A_DC_TIMER_START, do_timer_control); /* * Medium priority actions * - Membership */ } else if (fsa_actions & A_DC_TAKEOVER) { do_fsa_action(fsa_data, A_DC_TAKEOVER, do_dc_takeover); } else if (fsa_actions & A_DC_RELEASE) { do_fsa_action(fsa_data, A_DC_RELEASE, do_dc_release); } else if (fsa_actions & A_DC_JOIN_FINAL) { do_fsa_action(fsa_data, A_DC_JOIN_FINAL, do_dc_join_final); } else if (fsa_actions & A_ELECTION_CHECK) { do_fsa_action(fsa_data, A_ELECTION_CHECK, do_election_check); } else if (fsa_actions & A_ELECTION_START) { do_fsa_action(fsa_data, A_ELECTION_START, do_election_vote); } else if (fsa_actions & A_DC_JOIN_OFFER_ALL) { do_fsa_action(fsa_data, A_DC_JOIN_OFFER_ALL, do_dc_join_offer_all); } else if (fsa_actions & A_DC_JOIN_OFFER_ONE) { do_fsa_action(fsa_data, A_DC_JOIN_OFFER_ONE, do_dc_join_offer_one); } else if (fsa_actions & A_DC_JOIN_PROCESS_REQ) { do_fsa_action(fsa_data, A_DC_JOIN_PROCESS_REQ, do_dc_join_filter_offer); } else if (fsa_actions & A_DC_JOIN_PROCESS_ACK) { do_fsa_action(fsa_data, A_DC_JOIN_PROCESS_ACK, do_dc_join_ack); } else if (fsa_actions & A_DC_JOIN_FINALIZE) { do_fsa_action(fsa_data, A_DC_JOIN_FINALIZE, do_dc_join_finalize); } else if (fsa_actions & A_CL_JOIN_ANNOUNCE) { do_fsa_action(fsa_data, A_CL_JOIN_ANNOUNCE, do_cl_join_announce); /* * Low(er) priority actions * Make sure the CIB is always updated before invoking the * PE, and the PE before the TE */ } else if (fsa_actions & A_TE_HALT) { do_fsa_action(fsa_data, A_TE_HALT, do_te_invoke); } else if (fsa_actions & A_TE_CANCEL) { do_fsa_action(fsa_data, A_TE_CANCEL, do_te_invoke); } else if (fsa_actions & A_LRM_INVOKE) { do_fsa_action(fsa_data, A_LRM_INVOKE, do_lrm_invoke); } else if (fsa_actions & A_PE_INVOKE) { do_fsa_action(fsa_data, A_PE_INVOKE, do_pe_invoke); } else if (fsa_actions & A_TE_INVOKE) { do_fsa_action(fsa_data, A_TE_INVOKE, do_te_invoke); /* Shutdown actions */ } else if (fsa_actions & A_DC_RELEASED) { do_fsa_action(fsa_data, A_DC_RELEASED, do_dc_release); } else if (fsa_actions & A_PE_STOP) { do_fsa_action(fsa_data, A_PE_STOP, do_pe_control); } else if (fsa_actions & A_TE_STOP) { do_fsa_action(fsa_data, A_TE_STOP, do_te_control); } else if (fsa_actions & A_SHUTDOWN) { do_fsa_action(fsa_data, A_SHUTDOWN, do_shutdown); } else if (fsa_actions & A_LRM_DISCONNECT) { do_fsa_action(fsa_data, A_LRM_DISCONNECT, do_lrm_control); } else if (fsa_actions & A_HA_DISCONNECT) { do_fsa_action(fsa_data, A_HA_DISCONNECT, do_ha_control); } else if (fsa_actions & A_CIB_STOP) { do_fsa_action(fsa_data, A_CIB_STOP, do_cib_control); } else if (fsa_actions & A_STOP) { do_fsa_action(fsa_data, A_STOP, do_stop); /* exit gracefully */ } else if (fsa_actions & A_EXIT_0) { do_fsa_action(fsa_data, A_EXIT_0, do_exit); /* Error checking and reporting */ } else { crm_err("Action %s not supported "CRM_XS" 0x%llx", fsa_action2string(fsa_actions), fsa_actions); register_fsa_error_adv(C_FSA_INTERNAL, I_ERROR, fsa_data, NULL, __FUNCTION__); } } } void log_fsa_input(fsa_data_t * stored_msg) { CRM_ASSERT(stored_msg); crm_trace("Processing queued input %d", stored_msg->id); if (stored_msg->fsa_cause == C_LRM_OP_CALLBACK) { crm_trace("FSA processing LRM callback from %s", stored_msg->origin); } else if (stored_msg->data == NULL) { crm_trace("FSA processing input from %s", stored_msg->origin); } else { ha_msg_input_t *ha_input = fsa_typed_data_adv(stored_msg, fsa_dt_ha_msg, __FUNCTION__); crm_trace("FSA processing XML message from %s", stored_msg->origin); crm_log_xml_trace(ha_input->xml, "FSA message data"); } } long long do_state_transition(long long actions, enum crmd_fsa_state cur_state, enum crmd_fsa_state next_state, fsa_data_t * msg_data) { int level = LOG_INFO; long long tmp = actions; gboolean clear_recovery_bit = TRUE; enum crmd_fsa_cause cause = msg_data->fsa_cause; enum crmd_fsa_input current_input = msg_data->fsa_input; const char *state_from = fsa_state2string(cur_state); const char *state_to = fsa_state2string(next_state); const char *input = fsa_input2string(current_input); CRM_LOG_ASSERT(cur_state != next_state); do_dot_log(DOT_PREFIX "\t%s -> %s [ label=%s cause=%s origin=%s ]", state_from, state_to, input, fsa_cause2string(cause), msg_data->origin); if (cur_state == S_IDLE || next_state == S_IDLE) { level = LOG_NOTICE; } else if (cur_state == S_NOT_DC || next_state == S_NOT_DC) { level = LOG_NOTICE; } else if (cur_state == S_ELECTION) { level = LOG_NOTICE; } else if (cur_state == S_STARTING) { level = LOG_NOTICE; } else if (next_state == S_RECOVERY) { level = LOG_WARNING; } do_crm_log(level, "State transition %s -> %s " CRM_XS " input=%s cause=%s origin=%s", state_from, state_to, input, fsa_cause2string(cause), msg_data->origin); if (next_state != S_ELECTION && cur_state != S_RELEASE_DC) { controld_stop_election_timer(); } #if 0 if ((fsa_input_register & R_SHUTDOWN)) { set_bit(tmp, A_DC_TIMER_STOP); } #endif if (next_state == S_INTEGRATION) { set_bit(tmp, A_INTEGRATE_TIMER_START); } else { set_bit(tmp, A_INTEGRATE_TIMER_STOP); } if (next_state == S_FINALIZE_JOIN) { set_bit(tmp, A_FINALIZE_TIMER_START); } else { set_bit(tmp, A_FINALIZE_TIMER_STOP); } if (next_state != S_PENDING) { set_bit(tmp, A_DC_TIMER_STOP); } if (next_state != S_ELECTION) { highest_born_on = 0; } if (next_state != S_IDLE) { crm_timer_stop(recheck_timer); } if (cur_state == S_FINALIZE_JOIN && next_state == S_POLICY_ENGINE) { populate_cib_nodes(node_update_quick|node_update_all, __FUNCTION__); } switch (next_state) { case S_PENDING: fsa_cib_conn->cmds->set_slave(fsa_cib_conn, cib_scope_local); /* fall through */ case S_ELECTION: crm_trace("Resetting our DC to NULL on transition to %s", fsa_state2string(next_state)); update_dc(NULL); break; case S_NOT_DC: election_trigger->counter = 0; purge_stonith_cleanup(); if (is_set(fsa_input_register, R_SHUTDOWN)) { crm_info("(Re)Issuing shutdown request now" " that we have a new DC"); set_bit(tmp, A_SHUTDOWN_REQ); } CRM_LOG_ASSERT(fsa_our_dc != NULL); if (fsa_our_dc == NULL) { crm_err("Reached S_NOT_DC without a DC" " being recorded"); } break; case S_RECOVERY: clear_recovery_bit = FALSE; break; case S_FINALIZE_JOIN: CRM_LOG_ASSERT(AM_I_DC); if (cause == C_TIMER_POPPED) { crm_warn("Progressed to state %s after %s", fsa_state2string(next_state), fsa_cause2string(cause)); } if (crmd_join_phase_count(crm_join_welcomed) > 0) { crm_warn("%u cluster nodes failed to respond" " to the join offer.", crmd_join_phase_count(crm_join_welcomed)); crmd_join_phase_log(LOG_NOTICE); } else { crm_debug("All %d cluster nodes responded to the join offer.", crmd_join_phase_count(crm_join_integrated)); } break; case S_POLICY_ENGINE: election_trigger->counter = 0; CRM_LOG_ASSERT(AM_I_DC); if (cause == C_TIMER_POPPED) { crm_info("Progressed to state %s after %s", fsa_state2string(next_state), fsa_cause2string(cause)); } if (crmd_join_phase_count(crm_join_finalized) > 0) { crm_err("%u cluster nodes failed to confirm their join.", crmd_join_phase_count(crm_join_finalized)); crmd_join_phase_log(LOG_NOTICE); } else if (crmd_join_phase_count(crm_join_confirmed) == crm_active_peers()) { crm_debug("All %u cluster nodes are" " eligible to run resources.", crm_active_peers()); } else if (crmd_join_phase_count(crm_join_confirmed) > crm_active_peers()) { crm_err("We have more confirmed nodes than our membership does: %d vs. %d", crmd_join_phase_count(crm_join_confirmed), crm_active_peers()); register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL); } else if (saved_ccm_membership_id != crm_peer_seq) { crm_info("Membership changed: %llu -> %llu - join restart", saved_ccm_membership_id, crm_peer_seq); register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } else { crm_warn("Only %u of %u cluster " "nodes are eligible to run resources - continue %d", crmd_join_phase_count(crm_join_confirmed), crm_active_peers(), crmd_join_phase_count(crm_join_welcomed)); } /* initialize_join(FALSE); */ break; case S_STOPPING: case S_TERMINATE: /* possibly redundant */ set_bit(fsa_input_register, R_SHUTDOWN); break; case S_IDLE: CRM_LOG_ASSERT(AM_I_DC); if (is_set(fsa_input_register, R_SHUTDOWN)) { crm_info("(Re)Issuing shutdown request now" " that we are the DC"); set_bit(tmp, A_SHUTDOWN_REQ); } if (recheck_timer->period_ms > 0) { crm_debug("Starting %s", get_timer_desc(recheck_timer)); crm_timer_start(recheck_timer); } break; default: break; } if (clear_recovery_bit && next_state != S_PENDING) { tmp &= ~A_RECOVER; } else if (clear_recovery_bit == FALSE) { tmp |= A_RECOVER; } if (tmp != actions) { /* fsa_dump_actions(actions ^ tmp, "New actions"); */ actions = tmp; } return actions; } diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c index 2ebc203e17..8f37cbf079 100644 --- a/daemons/controld/controld_messages.c +++ b/daemons/controld/controld_messages.c @@ -1,1173 +1,1174 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include GListPtr fsa_message_queue = NULL; extern void crm_shutdown(int nsig); void handle_response(xmlNode * stored_msg); enum crmd_fsa_input handle_request(xmlNode * stored_msg, enum crmd_fsa_cause cause); enum crmd_fsa_input handle_shutdown_request(xmlNode * stored_msg); #define ROUTER_RESULT(x) crm_trace("Router result: %s", x) /* debug only, can wrap all it likes */ int last_data_id = 0; void register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, fsa_data_t * cur_data, void *new_data, const char *raised_from) { /* save the current actions if any */ if (fsa_actions != A_NOTHING) { register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL, I_NULL, cur_data ? cur_data->data : NULL, fsa_actions, TRUE, __FUNCTION__); } /* reset the action list */ crm_info("Resetting the current action list"); fsa_dump_actions(fsa_actions, "Drop"); fsa_actions = A_NOTHING; /* register the error */ register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from); } int register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, void *data, long long with_actions, gboolean prepend, const char *raised_from) { unsigned old_len = g_list_length(fsa_message_queue); fsa_data_t *fsa_data = NULL; if (raised_from == NULL) { raised_from = ""; } if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) { /* no point doing anything */ crm_err("Cannot add entry to queue: no input and no action"); return 0; } if (input == I_WAIT_FOR_EVENT) { do_fsa_stall = TRUE; crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d", raised_from, fsa_cause2string(cause), data, old_len); if (old_len > 0) { fsa_dump_queue(LOG_TRACE); prepend = FALSE; } if (data == NULL) { fsa_actions |= with_actions; fsa_dump_actions(with_actions, "Restored"); return 0; } /* Store everything in the new event and reset fsa_actions */ with_actions |= fsa_actions; fsa_actions = A_NOTHING; } last_data_id++; crm_trace("%s %s FSA input %d (%s) (cause=%s) %s data", raised_from, prepend ? "prepended" : "appended", last_data_id, fsa_input2string(input), fsa_cause2string(cause), data ? "with" : "without"); fsa_data = calloc(1, sizeof(fsa_data_t)); fsa_data->id = last_data_id; fsa_data->fsa_input = input; fsa_data->fsa_cause = cause; fsa_data->origin = raised_from; fsa_data->data = NULL; fsa_data->data_type = fsa_dt_none; fsa_data->actions = with_actions; if (with_actions != A_NOTHING) { crm_trace("Adding actions %.16llx to input", with_actions); } if (data != NULL) { switch (cause) { case C_FSA_INTERNAL: case C_CRMD_STATUS_CALLBACK: case C_IPC_MESSAGE: case C_HA_MESSAGE: crm_trace("Copying %s data from %s as a HA msg", fsa_cause2string(cause), raised_from); CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL, crm_err("Bogus data from %s", raised_from)); fsa_data->data = copy_ha_msg_input(data); fsa_data->data_type = fsa_dt_ha_msg; break; case C_LRM_OP_CALLBACK: crm_trace("Copying %s data from %s as lrmd_event_data_t", fsa_cause2string(cause), raised_from); fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data); fsa_data->data_type = fsa_dt_lrm; break; case C_TIMER_POPPED: case C_SHUTDOWN: case C_UNKNOWN: case C_STARTUP: crm_err("Copying %s data (from %s)" " not yet implemented", fsa_cause2string(cause), raised_from); crmd_exit(CRM_EX_SOFTWARE); break; } crm_trace("%s data copied", fsa_cause2string(fsa_data->fsa_cause)); } /* make sure to free it properly later */ if (prepend) { crm_trace("Prepending input"); fsa_message_queue = g_list_prepend(fsa_message_queue, fsa_data); } else { fsa_message_queue = g_list_append(fsa_message_queue, fsa_data); } crm_trace("Queue len: %d", g_list_length(fsa_message_queue)); /* fsa_dump_queue(LOG_TRACE); */ if (old_len == g_list_length(fsa_message_queue)) { crm_err("Couldn't add message to the queue"); } if (fsa_source && input != I_WAIT_FOR_EVENT) { crm_trace("Triggering FSA: %s", __FUNCTION__); mainloop_set_trigger(fsa_source); } return last_data_id; } void fsa_dump_queue(int log_level) { int offset = 0; GListPtr lpc = NULL; for (lpc = fsa_message_queue; lpc != NULL; lpc = lpc->next) { fsa_data_t *data = (fsa_data_t *) lpc->data; do_crm_log_unlikely(log_level, "queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)", offset++, data->id, fsa_input2string(data->fsa_input), data->origin, data->data, data->data_type, fsa_cause2string(data->fsa_cause)); } } ha_msg_input_t * copy_ha_msg_input(ha_msg_input_t * orig) { ha_msg_input_t *copy = NULL; xmlNodePtr data = NULL; if (orig != NULL) { crm_trace("Copy msg"); data = copy_xml(orig->msg); } else { crm_trace("No message to copy"); } copy = new_ha_msg_input(data); if (orig && orig->msg != NULL) { CRM_CHECK(copy->msg != NULL, crm_err("copy failed")); } return copy; } void delete_fsa_input(fsa_data_t * fsa_data) { lrmd_event_data_t *op = NULL; xmlNode *foo = NULL; if (fsa_data == NULL) { return; } crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause)); if (fsa_data->data != NULL) { switch (fsa_data->data_type) { case fsa_dt_ha_msg: delete_ha_msg_input(fsa_data->data); break; case fsa_dt_xml: foo = fsa_data->data; free_xml(foo); break; case fsa_dt_lrm: op = (lrmd_event_data_t *) fsa_data->data; lrmd_free_event(op); break; case fsa_dt_none: if (fsa_data->data != NULL) { crm_err("Don't know how to free %s data from %s", fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin); crmd_exit(CRM_EX_SOFTWARE); } break; } crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause)); } free(fsa_data); } /* returns the next message */ fsa_data_t * get_message(void) { fsa_data_t *message = g_list_nth_data(fsa_message_queue, 0); fsa_message_queue = g_list_remove(fsa_message_queue, message); crm_trace("Processing input %d", message->id); return message; } /* returns the current head of the FIFO queue */ gboolean is_message(void) { return (g_list_length(fsa_message_queue) > 0); } void * fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller) { void *ret_val = NULL; if (fsa_data == NULL) { crm_err("%s: No FSA data available", caller); } else if (fsa_data->data == NULL) { crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin); } else if (fsa_data->data_type != a_type) { crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s", caller, fsa_data->data_type, a_type, fsa_data->origin); CRM_ASSERT(fsa_data->data_type == a_type); } else { ret_val = fsa_data->data; } return ret_val; } /* A_MSG_ROUTE */ void do_msg_route(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); route_message(msg_data->fsa_cause, input->msg); } void route_message(enum crmd_fsa_cause cause, xmlNode * input) { ha_msg_input_t fsa_input; enum crmd_fsa_input result = I_NULL; fsa_input.msg = input; CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return); /* try passing the buck first */ if (relay_message(input, cause == C_IPC_MESSAGE)) { return; } /* handle locally */ result = handle_message(input, cause); /* done or process later? */ switch (result) { case I_NULL: case I_CIB_OP: case I_ROUTER: case I_NODE_JOIN: case I_JOIN_REQUEST: case I_JOIN_RESULT: break; default: /* Defering local processing of message */ register_fsa_input_later(cause, result, &fsa_input); return; } if (result != I_NULL) { /* add to the front of the queue */ register_fsa_input(cause, result, &fsa_input); } } gboolean relay_message(xmlNode * msg, gboolean originated_locally) { int dest = 1; int is_for_dc = 0; int is_for_dcib = 0; int is_for_te = 0; int is_for_crm = 0; int is_for_cib = 0; int is_local = 0; gboolean processing_complete = FALSE; const char *host_to = crm_element_value(msg, F_CRM_HOST_TO); const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM); const char *type = crm_element_value(msg, F_TYPE); const char *task = crm_element_value(msg, F_CRM_TASK); const char *msg_error = NULL; crm_trace("Routing message %s", crm_element_value(msg, XML_ATTR_REFERENCE)); if (msg == NULL) { msg_error = "Cannot route empty message"; } else if (safe_str_eq(task, CRM_OP_HELLO)) { /* quietly ignore */ processing_complete = TRUE; } else if (safe_str_neq(type, T_CRM)) { msg_error = "Bad message type"; } else if (sys_to == NULL) { msg_error = "Bad message destination: no subsystem"; } if (msg_error != NULL) { processing_complete = TRUE; crm_err("%s", msg_error); crm_log_xml_warn(msg, "bad msg"); } if (processing_complete) { return TRUE; } processing_complete = TRUE; is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0); is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0); is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0); is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0); is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0); is_local = 0; if (host_to == NULL || strlen(host_to) == 0) { if (is_for_dc || is_for_te) { is_local = 0; } else if (is_for_crm) { if (safe_str_eq(task, CRM_OP_NODE_INFO)) { /* Node info requests do not specify a host, which is normally * treated as "all hosts", because the whole point is that the * client doesn't know the local node name. Always handle these * requests locally. */ is_local = 1; } else { is_local = !originated_locally; } } else { is_local = 1; } } else if (safe_str_eq(fsa_our_uname, host_to)) { is_local = 1; } if (is_for_dc || is_for_dcib || is_for_te) { if (AM_I_DC && is_for_te) { ROUTER_RESULT("Message result: Local relay"); send_msg_via_ipc(msg, sys_to); } else if (AM_I_DC) { ROUTER_RESULT("Message result: DC/controller process"); processing_complete = FALSE; /* more to be done by caller */ } else if (originated_locally && safe_str_neq(sys_from, CRM_SYSTEM_PENGINE) && safe_str_neq(sys_from, CRM_SYSTEM_TENGINE)) { /* Neither the TE nor the scheduler should be sending messages * to DCs on other nodes. By definition, if we are no longer the DC, * then the scheduler's or TE's data should be discarded. */ #if SUPPORT_COROSYNC if (is_corosync_cluster()) { dest = text2msg_type(sys_to); } #endif ROUTER_RESULT("Message result: External relay to DC"); send_cluster_message(host_to ? crm_get_peer(0, host_to) : NULL, dest, msg, TRUE); } else { /* discard */ ROUTER_RESULT("Message result: Discard, not DC"); } } else if (is_local && (is_for_crm || is_for_cib)) { ROUTER_RESULT("Message result: controller process"); processing_complete = FALSE; /* more to be done by caller */ } else if (is_local) { ROUTER_RESULT("Message result: Local relay"); send_msg_via_ipc(msg, sys_to); } else { crm_node_t *node_to = NULL; #if SUPPORT_COROSYNC if (is_corosync_cluster()) { dest = text2msg_type(sys_to); if (dest == crm_msg_none || dest > crm_msg_stonith_ng) { dest = crm_msg_crmd; } } #endif if (host_to) { node_to = crm_find_peer(0, host_to); if (node_to == NULL) { crm_err("Cannot route message to unknown node %s", host_to); return TRUE; } } ROUTER_RESULT("Message result: External relay"); send_cluster_message(host_to ? node_to : NULL, dest, msg, TRUE); } return processing_complete; } static gboolean process_hello_message(xmlNode * hello, char **client_name, char **major_version, char **minor_version) { const char *local_client_name; const char *local_major_version; const char *local_minor_version; *client_name = NULL; *major_version = NULL; *minor_version = NULL; if (hello == NULL) { return FALSE; } local_client_name = crm_element_value(hello, "client_name"); local_major_version = crm_element_value(hello, "major_version"); local_minor_version = crm_element_value(hello, "minor_version"); if (local_client_name == NULL || strlen(local_client_name) == 0) { crm_err("Hello message was not valid (field %s not found)", "client name"); return FALSE; } else if (local_major_version == NULL || strlen(local_major_version) == 0) { crm_err("Hello message was not valid (field %s not found)", "major version"); return FALSE; } else if (local_minor_version == NULL || strlen(local_minor_version) == 0) { crm_err("Hello message was not valid (field %s not found)", "minor version"); return FALSE; } *client_name = strdup(local_client_name); *major_version = strdup(local_major_version); *minor_version = strdup(local_minor_version); crm_trace("Hello message ok"); return TRUE; } gboolean crmd_authorize_message(xmlNode * client_msg, crm_client_t * curr_client, const char *proxy_session) { char *client_name = NULL; char *major_version = NULL; char *minor_version = NULL; gboolean auth_result = FALSE; xmlNode *xml = NULL; const char *op = crm_element_value(client_msg, F_CRM_TASK); const char *uuid = curr_client ? curr_client->id : proxy_session; if (uuid == NULL) { crm_warn("Message [%s] not authorized", crm_element_value(client_msg, XML_ATTR_REFERENCE)); return FALSE; } else if (safe_str_neq(CRM_OP_HELLO, op)) { return TRUE; } xml = get_message_xml(client_msg, F_CRM_DATA); auth_result = process_hello_message(xml, &client_name, &major_version, &minor_version); if (auth_result == TRUE) { if (client_name == NULL) { crm_err("Bad client details (client_name=%s, uuid=%s)", crm_str(client_name), uuid); auth_result = FALSE; } } if (auth_result == TRUE) { /* check version */ int mav = atoi(major_version); int miv = atoi(minor_version); crm_trace("Checking client version number"); if (mav < 0 || miv < 0) { crm_err("Client version (%d:%d) is not acceptable", mav, miv); auth_result = FALSE; } } if (auth_result == TRUE) { crm_trace("Accepted client %s", client_name); if (curr_client) { curr_client->userdata = strdup(client_name); } crm_trace("Triggering FSA: %s", __FUNCTION__); mainloop_set_trigger(fsa_source); } else { crm_warn("Rejected client logon request"); if (curr_client) { qb_ipcs_disconnect(curr_client->ipcs); } } free(minor_version); free(major_version); free(client_name); /* hello messages should never be processed further */ return FALSE; } enum crmd_fsa_input handle_message(xmlNode * msg, enum crmd_fsa_cause cause) { const char *type = NULL; CRM_CHECK(msg != NULL, return I_NULL); type = crm_element_value(msg, F_CRM_MSG_TYPE); if (crm_str_eq(type, XML_ATTR_REQUEST, TRUE)) { return handle_request(msg, cause); } else if (crm_str_eq(type, XML_ATTR_RESPONSE, TRUE)) { handle_response(msg); return I_NULL; } crm_err("Unknown message type: %s", type); return I_NULL; } static enum crmd_fsa_input handle_failcount_op(xmlNode * stored_msg) { const char *rsc = NULL; const char *uname = NULL; const char *op = NULL; const char *interval_ms_s = NULL; char *interval_spec = NULL; guint interval_ms = 0; gboolean is_remote_node = FALSE; xmlNode *xml_op = get_message_xml(stored_msg, F_CRM_DATA); if (xml_op) { xmlNode *xml_rsc = first_named_child(xml_op, XML_CIB_TAG_RESOURCE); xmlNode *xml_attrs = first_named_child(xml_op, XML_TAG_ATTRS); if (xml_rsc) { rsc = ID(xml_rsc); } if (xml_attrs) { op = crm_element_value(xml_attrs, CRM_META "_" XML_RSC_ATTR_CLEAR_OP); interval_ms_s = crm_element_value(xml_attrs, CRM_META "_" XML_RSC_ATTR_CLEAR_INTERVAL); interval_ms = crm_parse_ms(interval_ms_s); } } uname = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if ((rsc == NULL) || (uname == NULL)) { crm_log_xml_warn(stored_msg, "invalid failcount op"); return I_NULL; } if (crm_element_value(xml_op, XML_LRM_ATTR_ROUTER_NODE)) { is_remote_node = TRUE; } if (interval_ms) { interval_spec = crm_strdup_printf("%ums", interval_ms); } update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node); free(interval_spec); lrm_clear_last_failure(rsc, uname, op, interval_ms); return I_NULL; } /*! * \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_remote_state(xmlNode *msg) { const char *remote_uname = ID(msg); const char *remote_is_up = crm_element_value(msg, XML_NODE_IN_CLUSTER); crm_node_t *remote_peer; CRM_CHECK(remote_uname && remote_is_up, return I_NULL); remote_peer = crm_remote_peer_get(remote_uname); CRM_CHECK(remote_peer, return I_NULL); crm_update_peer_state(__FUNCTION__, remote_peer, crm_is_true(remote_is_up)? CRM_NODE_MEMBER : CRM_NODE_LOST, 0); return I_NULL; } /*! * \brief Handle a CRM_OP_PING message * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_ping(xmlNode *msg) { const char *value = NULL; xmlNode *ping = NULL; // Build reply ping = create_xml_node(NULL, XML_CRM_TAG_PING); value = crm_element_value(msg, F_CRM_SYS_TO); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value); // Add controller state value = fsa_state2string(fsa_state); crm_xml_add(ping, XML_PING_ATTR_CRMDSTATE, value); crm_notice("Current ping state: %s", value); // CTS needs this // Add controller health // @TODO maybe do some checks to determine meaningful status crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); // Send reply msg = create_reply(msg, ping); free_xml(ping); if (msg) { (void) relay_message(msg, TRUE); free_xml(msg); } // Nothing further to do return I_NULL; } /*! * \brief Handle a CRM_OP_NODE_INFO request * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_info_request(xmlNode *msg) { const char *value = NULL; crm_node_t *node = NULL; int node_id = 0; xmlNode *reply = NULL; // Build reply reply = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(reply, XML_PING_ATTR_SYSFROM, CRM_SYSTEM_CRMD); // Add whether current partition has quorum crm_xml_add_boolean(reply, XML_ATTR_HAVE_QUORUM, fsa_has_quorum); // Check whether client requested node info by ID and/or name crm_element_value_int(msg, XML_ATTR_ID, &node_id); if (node_id < 0) { node_id = 0; } value = crm_element_value(msg, XML_ATTR_UNAME); // Default to local node if none given if ((node_id == 0) && (value == NULL)) { value = fsa_our_uname; } node = crm_find_peer_full(node_id, value, CRM_GET_PEER_ANY); if (node) { crm_xml_add_int(reply, XML_ATTR_ID, node->id); crm_xml_add(reply, XML_ATTR_UUID, node->uuid); crm_xml_add(reply, XML_ATTR_UNAME, node->uname); crm_xml_add(reply, XML_NODE_IS_PEER, node->state); crm_xml_add_boolean(reply, XML_NODE_IS_REMOTE, node->flags & crm_remote_node); } // Send reply msg = create_reply(msg, reply); free_xml(reply); if (msg) { (void) relay_message(msg, TRUE); free_xml(msg); } // Nothing further to do return I_NULL; } static void verify_feature_set(xmlNode *msg) { const char *dc_version = crm_element_value(msg, XML_ATTR_CRM_VERSION); if (dc_version == NULL) { /* All we really know is that the DC feature set is older than 3.1.0, * but that's also all that really matters. */ dc_version = "3.0.14"; } if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) { crm_trace("Local feature set (%s) is compatible with DC's (%s)", CRM_FEATURE_SET, dc_version); } else { crm_err("Local feature set (%s) is incompatible with DC's (%s)", CRM_FEATURE_SET, dc_version); // Nothing is likely to improve without administrator involvement set_bit(fsa_input_register, R_STAYDOWN); crmd_exit(CRM_EX_FATAL); } } enum crmd_fsa_input handle_request(xmlNode * stored_msg, enum crmd_fsa_cause cause) { xmlNode *msg = NULL; const char *op = crm_element_value(stored_msg, F_CRM_TASK); /* Optimize this for the DC - it has the most to do */ if (op == NULL) { crm_log_xml_err(stored_msg, "Bad message"); return I_NULL; } if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { const char *from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_node_t *node = crm_find_peer(0, from); crm_update_peer_expected(__FUNCTION__, node, CRMD_JOINSTATE_DOWN); if(AM_I_DC == FALSE) { return I_NULL; /* Done */ } } /*========== DC-Only Actions ==========*/ if (AM_I_DC) { if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) { return I_NODE_JOIN; } else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) { return I_JOIN_REQUEST; } else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) { return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); gboolean dc_match = safe_str_eq(host_from, fsa_our_dc); if (is_set(fsa_input_register, R_SHUTDOWN)) { crm_info("Shutting ourselves down (DC)"); return I_STOP; } else if (dc_match) { crm_err("We didn't ask to be shut down, yet our" " TE is telling us to. Better get out now!"); return I_TERMINATE; } else if (fsa_state != S_STOPPING) { crm_err("Another node is asking us to shutdown" " but we think we're ok."); return I_ELECTION; } } else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { /* a slave wants to shut down */ /* create cib fragment and add to message */ return handle_shutdown_request(stored_msg); } else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) { /* a remote connection host is letting us know the node state */ return handle_remote_state(stored_msg); } } /*========== common actions ==========*/ if (strcmp(op, CRM_OP_NOVOTE) == 0) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__); } else if (strcmp(op, CRM_OP_THROTTLE) == 0) { throttle_update(stored_msg); if (AM_I_DC && transition_graph != NULL) { if (transition_graph->complete == FALSE) { crm_debug("The throttle changed. Trigger a graph."); trigger_graph(); } } return I_NULL; } else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) { return handle_failcount_op(stored_msg); } else if (strcmp(op, CRM_OP_VOTE) == 0) { /* count the vote and decide what to do after that */ ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__); /* Sometimes we _must_ go into S_ELECTION */ if (fsa_state == S_HALT) { crm_debug("Forcing an election from S_HALT"); return I_ELECTION; #if 0 } else if (AM_I_DC) { /* This is the old way of doing things but what is gained? */ return I_ELECTION; #endif } } else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) { verify_feature_set(stored_msg); crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_OFFER; } else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) { crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_LRM_DELETE) == 0 || strcmp(op, CRM_OP_LRM_FAIL) == 0 || strcmp(op, CRM_OP_LRM_REFRESH) == 0 || strcmp(op, CRM_OP_REPROBE) == 0) { crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else if (strcmp(op, CRM_OP_NOOP) == 0) { return I_NULL; } else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) { crm_shutdown(SIGTERM); /*return I_SHUTDOWN; */ return I_NULL; } else if (strcmp(op, CRM_OP_PING) == 0) { return handle_ping(stored_msg); } else if (strcmp(op, CRM_OP_NODE_INFO) == 0) { return handle_node_info_request(stored_msg); } else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) { int id = 0; const char *name = NULL; crm_element_value_int(stored_msg, XML_ATTR_ID, &id); name = crm_element_value(stored_msg, XML_ATTR_UNAME); if(cause == C_IPC_MESSAGE) { msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) { crm_err("Could not instruct peers to remove references to node %s/%u", name, id); } else { crm_notice("Instructing peers to remove references to node %s/%u", name, id); } free_xml(msg); } else { reap_crm_member(id, name); /* If we're forgetting this node, also forget any failures to fence * it, so we don't carry that over to any node added later with the * same name. */ st_fail_count_reset(name); } } else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) { xmlNode *xml = get_message_xml(stored_msg, F_CRM_DATA); remote_ra_process_maintenance_nodes(xml); /*========== (NOT_DC)-Only Actions ==========*/ } else if (AM_I_DC == FALSE && strcmp(op, CRM_OP_SHUTDOWN) == 0) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); gboolean dc_match = safe_str_eq(host_from, fsa_our_dc); if (dc_match || fsa_our_dc == NULL) { if (is_set(fsa_input_register, R_SHUTDOWN) == FALSE) { crm_err("We didn't ask to be shut down, yet our DC is telling us to."); set_bit(fsa_input_register, R_STAYDOWN); return I_STOP; } crm_info("Shutting down"); return I_STOP; } else { crm_warn("Discarding %s op from %s", op, host_from); } } else { crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node"); crm_log_xml_err(stored_msg, "Unexpected"); } return I_NULL; } void handle_response(xmlNode * stored_msg) { const char *op = crm_element_value(stored_msg, F_CRM_TASK); if (op == NULL) { crm_log_xml_err(stored_msg, "Bad message"); } else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) { // Check whether scheduler answer been superseded by subsequent request const char *msg_ref = crm_element_value(stored_msg, XML_ATTR_REFERENCE); if (msg_ref == NULL) { crm_err("%s - Ignoring calculation with no reference", op); } else if (safe_str_eq(msg_ref, fsa_pe_ref)) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input); crm_trace("Completed: %s...", fsa_pe_ref); } else { crm_info("%s calculation %s is obsolete", op, msg_ref); } } else if (strcmp(op, CRM_OP_VOTE) == 0 || strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) { } else { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_err("Unexpected response (op=%s, src=%s) sent to the %s", op, host_from, AM_I_DC ? "DC" : "controller"); } } enum crmd_fsa_input handle_shutdown_request(xmlNode * stored_msg) { /* handle here to avoid potential version issues * where the shutdown message/procedure may have * been changed in later versions. * * This way the DC is always in control of the shutdown */ char *now_s = NULL; time_t now = time(NULL); const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { /* we're shutting down and the DC */ host_from = fsa_our_uname; } crm_info("Creating shutdown request for %s (state=%s)", host_from, fsa_state2string(fsa_state)); crm_log_xml_trace(stored_msg, "message"); now_s = crm_itoa(now); update_attrd(host_from, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, FALSE); free(now_s); /* will be picked up by the TE as long as its running */ return I_NULL; } /* msg is deleted by the time this returns */ extern gboolean process_te_message(xmlNode * msg, xmlNode * xml_data); gboolean send_msg_via_ipc(xmlNode * msg, const char *sys) { gboolean send_ok = TRUE; crm_client_t *client_channel = crm_client_get_by_id(sys); if (crm_element_value(msg, F_CRM_HOST_FROM) == NULL) { crm_xml_add(msg, F_CRM_HOST_FROM, fsa_our_uname); } if (client_channel != NULL) { /* Transient clients such as crmadmin */ send_ok = crm_ipcs_send(client_channel, 0, msg, crm_ipc_server_event); } else if (sys != NULL && strcmp(sys, CRM_SYSTEM_TENGINE) == 0) { xmlNode *data = get_message_xml(msg, F_CRM_DATA); process_te_message(msg, data); } else if (sys != NULL && strcmp(sys, CRM_SYSTEM_LRMD) == 0) { fsa_data_t fsa_data; ha_msg_input_t fsa_input; fsa_input.msg = msg; fsa_input.xml = get_message_xml(msg, F_CRM_DATA); fsa_data.id = 0; fsa_data.actions = 0; fsa_data.data = &fsa_input; fsa_data.fsa_input = I_MESSAGE; fsa_data.fsa_cause = C_IPC_MESSAGE; fsa_data.origin = __FUNCTION__; fsa_data.data_type = fsa_dt_ha_msg; #ifdef FSA_TRACE crm_trace("Invoking action A_LRM_INVOKE (%.16llx)", A_LRM_INVOKE); #endif do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, fsa_state, I_MESSAGE, &fsa_data); } else if (sys != NULL && crmd_is_proxy_session(sys)) { crmd_proxy_send(sys, msg); } else { crm_debug("Unknown Sub-system (%s)... discarding message.", crm_str(sys)); send_ok = FALSE; } return send_ok; } ha_msg_input_t * new_ha_msg_input(xmlNode * orig) { ha_msg_input_t *input_copy = NULL; input_copy = calloc(1, sizeof(ha_msg_input_t)); input_copy->msg = orig; input_copy->xml = get_message_xml(input_copy->msg, F_CRM_DATA); return input_copy; } void delete_ha_msg_input(ha_msg_input_t * orig) { if (orig == NULL) { return; } free_xml(orig->msg); free(orig); } /*! * \internal * \brief Notify the DC of a remote node state change * * \param[in] node_name Node's name * \param[in] node_up TRUE if node is up, FALSE if down */ void send_remote_state_message(const char *node_name, gboolean node_up) { /* If we don't have a DC, or the message fails, we have a failsafe: * the DC will eventually pick up the change via the CIB node state. * The message allows it to happen sooner if possible. */ if (fsa_our_dc) { xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, fsa_our_dc, CRM_SYSTEM_DC, CRM_SYSTEM_CRMD, NULL); crm_info("Notifying DC %s of pacemaker_remote node %s %s", fsa_our_dc, node_name, (node_up? "coming up" : "going down")); crm_xml_add(msg, XML_ATTR_ID, node_name); crm_xml_add_boolean(msg, XML_NODE_IN_CLUSTER, node_up); send_cluster_message(crm_get_peer(0, fsa_our_dc), crm_msg_crmd, msg, TRUE); free_xml(msg); } else { crm_debug("No DC to notify of pacemaker_remote node %s %s", node_name, (node_up? "coming up" : "going down")); } } diff --git a/daemons/controld/controld_te_actions.c b/daemons/controld/controld_te_actions.c index c95c6c7e90..2f61556dee 100644 --- a/daemons/controld/controld_te_actions.c +++ b/daemons/controld/controld_te_actions.c @@ -1,748 +1,631 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * 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 char *te_uuid = NULL; GHashTable *te_targets = NULL; void send_rsc_command(crm_action_t * action); static void te_update_job_count(crm_action_t * action, int offset); static void te_start_action_timer(crm_graph_t * graph, crm_action_t * action) { action->timer = calloc(1, sizeof(crm_action_timer_t)); action->timer->timeout = action->timeout; action->timer->action = action; action->timer->source_id = g_timeout_add(action->timer->timeout + graph->network_delay, action_timer_callback, (void *)action->timer); CRM_ASSERT(action->timer->source_id != 0); } static gboolean te_pseudo_action(crm_graph_t * graph, crm_action_t * pseudo) { const char *task = crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK); /* send to peers as well? */ if (safe_str_eq(task, CRM_OP_MAINTENANCE_NODES)) { GHashTableIter iter; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *cmd = NULL; if (safe_str_eq(fsa_our_uname, node->uname)) { continue; } cmd = create_request(task, pseudo->xml, node->uname, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); send_cluster_message(node, crm_msg_crmd, cmd, FALSE); free_xml(cmd); } remote_ra_process_maintenance_nodes(pseudo->xml); } else { /* Check action for Pacemaker Remote node side effects */ remote_ra_process_pseudo(pseudo->xml); } crm_debug("Pseudo-action %d (%s) fired and confirmed", pseudo->id, crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK_KEY)); te_action_confirmed(pseudo); update_graph(graph, pseudo); trigger_graph(); return TRUE; } -void -send_stonith_update(crm_action_t * action, const char *target, const char *uuid) -{ - int rc = pcmk_ok; - crm_node_t *peer = NULL; - - /* We (usually) rely on the membership layer to do node_update_cluster, - * and the peer status callback to do node_update_peer, because the node - * might have already rejoined before we get the stonith result here. - */ - int flags = node_update_join | node_update_expected; - - /* zero out the node-status & remove all LRM status info */ - xmlNode *node_state = NULL; - - CRM_CHECK(target != NULL, return); - CRM_CHECK(uuid != NULL, return); - - /* Make sure the membership and join caches are accurate */ - peer = crm_get_peer_full(0, target, CRM_GET_PEER_ANY); - - CRM_CHECK(peer != NULL, return); - - if (peer->state == NULL) { - /* Usually, we rely on the membership layer to update the cluster state - * in the CIB. However, if the node has never been seen, do it here, so - * the node is not considered unclean. - */ - flags |= node_update_cluster; - } - - if (peer->uuid == NULL) { - crm_info("Recording uuid '%s' for node '%s'", uuid, target); - peer->uuid = strdup(uuid); - } - - crmd_peer_down(peer, TRUE); - - /* Generate a node state update for the CIB */ - node_state = create_node_state_update(peer, flags, NULL, __FUNCTION__); - - /* we have to mark whether or not remote nodes have already been fenced */ - if (peer->flags & crm_remote_node) { - time_t now = time(NULL); - char *now_s = crm_itoa(now); - crm_xml_add(node_state, XML_NODE_IS_FENCED, now_s); - free(now_s); - } - - /* Force our known ID */ - crm_xml_add(node_state, XML_ATTR_UUID, uuid); - - rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, node_state, - cib_quorum_override | cib_scope_local | cib_can_create); - - /* Delay processing the trigger until the update completes */ - crm_debug("Sending fencing update %d for %s", rc, target); - fsa_register_cib_callback(rc, FALSE, strdup(target), cib_fencing_updated); - - /* Make sure it sticks */ - /* fsa_cib_conn->cmds->bump_epoch(fsa_cib_conn, cib_quorum_override|cib_scope_local); */ - - erase_status_tag(peer->uname, XML_CIB_TAG_LRM, cib_scope_local); - erase_status_tag(peer->uname, XML_TAG_TRANSIENT_NODEATTRS, cib_scope_local); - - free_xml(node_state); - return; -} - -static gboolean -te_fence_node(crm_graph_t * graph, crm_action_t * action) -{ - int rc = 0; - const char *id = NULL; - const char *uuid = NULL; - const char *target = NULL; - const char *type = NULL; - gboolean invalid_action = FALSE; - enum stonith_call_options options = st_opt_none; - - id = ID(action->xml); - target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); - uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); - type = crm_meta_value(action->params, "stonith_action"); - - CRM_CHECK(id != NULL, invalid_action = TRUE); - CRM_CHECK(uuid != NULL, invalid_action = TRUE); - CRM_CHECK(type != NULL, invalid_action = TRUE); - CRM_CHECK(target != NULL, invalid_action = TRUE); - - if (invalid_action) { - crm_log_xml_warn(action->xml, "BadAction"); - return FALSE; - } - - crm_notice("Requesting fencing (%s) of node %s " - CRM_XS " action=%s timeout=%d", - type, target, id, transition_graph->stonith_timeout); - - /* Passing NULL means block until we can connect... */ - te_connect_stonith(NULL); - - if (crmd_join_phase_count(crm_join_confirmed) == 1) { - options |= st_opt_allow_suicide; - } - - rc = stonith_api->cmds->fence(stonith_api, options, target, type, - transition_graph->stonith_timeout / 1000, 0); - - stonith_api->cmds->register_callback(stonith_api, rc, transition_graph->stonith_timeout / 1000, - st_opt_timeout_updates, - generate_transition_key(transition_graph->id, action->id, - 0, te_uuid), - "tengine_stonith_callback", tengine_stonith_callback); - - return TRUE; -} - static int get_target_rc(crm_action_t * action) { const char *target_rc_s = crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC); if (target_rc_s != NULL) { return crm_parse_int(target_rc_s, "0"); } return 0; } static gboolean te_crm_command(crm_graph_t * graph, crm_action_t * action) { char *counter = NULL; xmlNode *cmd = NULL; gboolean is_local = FALSE; const char *id = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; id = ID(action->xml); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command (id=%s) %s: no node", crm_str(id), crm_str(task)); return FALSE); if (safe_str_eq(router_node, fsa_our_uname)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_info("Executing crm-event (%s)%s%s: %s on %s", crm_str(id), (is_local? " locally" : ""), (no_wait? " without waiting" : ""), crm_str(task), on_node); if (is_local && safe_str_eq(task, CRM_OP_SHUTDOWN)) { /* defer until everything else completes */ crm_info("crm-event (%s) is a local shutdown", crm_str(id)); graph->completion_action = tg_shutdown; graph->abort_reason = "local shutdown"; te_action_confirmed(action); update_graph(graph, action); trigger_graph(); return TRUE; } else if (safe_str_eq(task, CRM_OP_SHUTDOWN)) { crm_node_t *peer = crm_get_peer(0, router_node); crm_update_peer_expected(__FUNCTION__, peer, CRMD_JOINSTATE_DOWN); } cmd = create_request(task, action->xml, router_node, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(cmd, XML_ATTR_TRANSITION_KEY, counter); rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_crmd, cmd, TRUE); free(counter); free_xml(cmd); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { te_action_confirmed(action); update_graph(graph, action); trigger_graph(); } else { if (action->timeout <= 0) { crm_err("Action %d: %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_start_action_timer(graph, action); } return TRUE; } void controld_record_action_timeout(crm_action_t *action) { lrmd_event_data_t *op = NULL; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *xml_op = NULL; xmlNode *action_rsc = NULL; int rc = pcmk_ok; const char *rsc_id = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); const char *target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); int call_options = cib_quorum_override | cib_scope_local; int target_rc = get_target_rc(action); crm_warn("%s %d: %s on %s timed out", crm_element_name(action->xml), action->id, task_uuid, target); action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if (action_rsc == NULL) { return; } rsc_id = ID(action_rsc); CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); return); /* update the CIB */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, XML_ATTR_UUID, target_uuid); crm_xml_add(state, XML_ATTR_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, XML_ATTR_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, XML_ATTR_ID, rsc_id); crm_copy_xml_element(action_rsc, rsc, XML_ATTR_TYPE); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_CLASS); crm_copy_xml_element(action_rsc, rsc, XML_AGENT_ATTR_PROVIDER); /* If the executor gets a timeout while waiting for the action to complete, * that will be reported via the usual callback. This timeout means that we * didn't hear from the executor or the controller that relayed the action * to the executor. * * @TODO Using PCMK_OCF_UNKNOWN_ERROR instead of PCMK_OCF_TIMEOUT is one way * to distinguish those situations, but perhaps PCMK_OCF_TIMEOUT would be * preferable anyway. */ op = convert_graph_action(NULL, action, PCMK_LRM_OP_TIMEOUT, PCMK_OCF_UNKNOWN_ERROR); op->call_id = -1; op->user_data = generate_transition_key(transition_graph->id, action->id, target_rc, te_uuid); xml_op = create_operation_update(rsc, op, CRM_FEATURE_SET, target_rc, target, __FUNCTION__, LOG_INFO); lrmd_free_event(op); crm_log_xml_trace(xml_op, "Action timeout"); rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, state, call_options); fsa_register_cib_callback(rc, FALSE, NULL, cib_action_updated); free_xml(state); crm_trace("Sent CIB update (call ID %d) for timeout of action %d (%s on %s)", rc, action->id, task_uuid, target); action->sent_update = TRUE; } static gboolean te_rsc_command(crm_graph_t * graph, crm_action_t * action) { /* never overwrite stop actions in the CIB with * anything other than completed results * * Writing pending stops makes it look like the * resource is running again */ xmlNode *cmd = NULL; xmlNode *rsc_op = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; gboolean is_local = FALSE; char *counter = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; const char *task_uuid = NULL; CRM_ASSERT(action != NULL); CRM_ASSERT(action->xml != NULL); action->executed = FALSE; on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(on_node != NULL && strlen(on_node) != 0, crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), crm_str(task)); return FALSE); rsc_op = action->xml; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); router_node = crm_element_value(rsc_op, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } counter = generate_transition_key(transition_graph->id, action->id, get_target_rc(action), te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (safe_str_eq(router_node, fsa_our_uname)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_notice("Initiating %s operation %s%s on %s%s "CRM_XS" action %d", task, task_uuid, (is_local? " locally" : ""), on_node, (no_wait? " without waiting" : ""), action->id); cmd = create_request(CRM_OP_INVOKE_LRM, rsc_op, router_node, CRM_SYSTEM_LRMD, CRM_SYSTEM_TENGINE, NULL); if (is_local) { /* shortcut local resource commands */ ha_msg_input_t data = { .msg = cmd, .xml = rsc_op, }; fsa_data_t msg = { .id = 0, .data = &data, .data_type = fsa_dt_ha_msg, .fsa_input = I_NULL, .fsa_cause = C_FSA_INTERNAL, .actions = A_LRM_INVOKE, .origin = __FUNCTION__, }; do_lrm_invoke(A_LRM_INVOKE, C_FSA_INTERNAL, fsa_state, I_NULL, &msg); } else { rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_lrmd, cmd, TRUE); } free(counter); free_xml(cmd); action->executed = TRUE; if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return FALSE; } else if (no_wait) { crm_info("Action %d confirmed - no wait", action->id); action->confirmed = TRUE; /* Just mark confirmed. * Don't bump the job count only to immediately decrement it */ update_graph(transition_graph, action); trigger_graph(); } else if (action->confirmed == TRUE) { crm_debug("Action %d: %s %s on %s(timeout %dms) was already confirmed.", action->id, task, task_uuid, on_node, action->timeout); } else { if (action->timeout <= 0) { crm_err("Action %d: %s %s on %s had an invalid timeout (%dms). Using %dms instead", action->id, task, task_uuid, on_node, action->timeout, graph->network_delay); action->timeout = graph->network_delay; } te_update_job_count(action, 1); te_start_action_timer(graph, action); } return TRUE; } struct te_peer_s { char *name; int jobs; int migrate_jobs; }; static void te_peer_free(gpointer p) { struct te_peer_s *peer = p; free(peer->name); free(peer); } void te_reset_job_counts(void) { GHashTableIter iter; struct te_peer_s *peer = NULL; if(te_targets == NULL) { te_targets = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, te_peer_free); } g_hash_table_iter_init(&iter, te_targets); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & peer)) { peer->jobs = 0; peer->migrate_jobs = 0; } } static void te_update_job_count_on(const char *target, int offset, bool migrate) { struct te_peer_s *r = NULL; if(target == NULL || te_targets == NULL) { return; } r = g_hash_table_lookup(te_targets, target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } r->jobs += offset; if(migrate) { r->migrate_jobs += offset; } crm_trace("jobs[%s] = %d", target, r->jobs); } static void te_update_job_count(crm_action_t * action, int offset) { const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (action->type != action_type_rsc || target == NULL) { /* No limit on these */ return; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED))) { const char *t1 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); const char *t2 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); te_update_job_count_on(t1, offset, TRUE); te_update_job_count_on(t2, offset, TRUE); return; } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } te_update_job_count_on(target, offset, FALSE); } static gboolean te_should_perform_action_on(crm_graph_t * graph, crm_action_t * action, const char *target) { int limit = 0; struct te_peer_s *r = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *id = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); if(target == NULL) { /* No limit on these */ return TRUE; } else if(te_targets == NULL) { return FALSE; } r = g_hash_table_lookup(te_targets, target); limit = throttle_get_job_limit(target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } if(limit <= r->jobs) { crm_trace("Peer %s is over their job limit of %d (%d): deferring %s", target, limit, r->jobs, id); return FALSE; } else if(graph->migration_limit > 0 && r->migrate_jobs >= graph->migration_limit) { if (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) { crm_trace("Peer %s is over their migration job limit of %d (%d): deferring %s", target, graph->migration_limit, r->migrate_jobs, id); return FALSE; } } crm_trace("Peer %s has not hit their limit yet. current jobs = %d limit= %d limit", target, r->jobs, limit); return TRUE; } static gboolean te_should_perform_action(crm_graph_t * graph, crm_action_t * action) { const char *target = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != action_type_rsc) { /* No limit on these */ return TRUE; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && (safe_str_eq(task, CRMD_ACTION_MIGRATE) || safe_str_eq(task, CRMD_ACTION_MIGRATED))) { target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); if(te_should_perform_action_on(graph, action, target) == FALSE) { return FALSE; } target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } return te_should_perform_action_on(graph, action, target); } void te_action_confirmed(crm_action_t * action) { const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if (action->confirmed == FALSE && action->type == action_type_rsc && target != NULL) { te_update_job_count(action, -1); } action->confirmed = TRUE; } crm_graph_functions_t te_graph_fns = { te_pseudo_action, te_rsc_command, te_crm_command, te_fence_node, te_should_perform_action, }; void notify_crmd(crm_graph_t * graph) { const char *type = "unknown"; enum crmd_fsa_input event = I_NULL; crm_debug("Processing transition completion in state %s", fsa_state2string(fsa_state)); if (graph->complete == FALSE) { CRM_CHECK(graph->complete,); graph->complete = TRUE; } switch (graph->completion_action) { case tg_stop: type = "stop"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_done: type = "done"; if (fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case tg_restart: type = "restart"; if (fsa_state == S_TRANSITION_ENGINE) { if (transition_timer->period_ms > 0) { crm_timer_stop(transition_timer); crm_timer_start(transition_timer); } else { event = I_PE_CALC; } } else if (fsa_state == S_POLICY_ENGINE) { register_fsa_action(A_PE_INVOKE); } break; case tg_shutdown: type = "shutdown"; if (is_set(fsa_input_register, R_SHUTDOWN)) { event = I_STOP; } else { crm_err("We didn't ask to be shut down, yet the scheduler is telling us to"); event = I_TERMINATE; } } crm_debug("Transition %d status: %s - %s", graph->id, type, crm_str(graph->abort_reason)); graph->abort_reason = NULL; graph->completion_action = tg_done; clear_bit(fsa_input_register, R_IN_TRANSITION); if (event != I_NULL) { register_fsa_input(C_FSA_INTERNAL, event, NULL); } else if (fsa_source) { mainloop_set_trigger(fsa_source); } } diff --git a/daemons/controld/controld_te_callbacks.c b/daemons/controld/controld_te_callbacks.c index 22b5f4bae9..1ab703f766 100644 --- a/daemons/controld/controld_te_callbacks.c +++ b/daemons/controld/controld_te_callbacks.c @@ -1,942 +1,701 @@ /* * Copyright 2004-2019 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 /* For ONLINESTATUS etc */ void te_update_confirm(const char *event, xmlNode * msg); extern char *te_uuid; gboolean shuttingdown = FALSE; crm_graph_t *transition_graph; crm_trigger_t *transition_trigger = NULL; -static unsigned long int stonith_max_attempts = 10; - /* #define RSC_OP_TEMPLATE "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_CIB_TAG_STATE"[@uname='%s']"//"XML_LRM_TAG_RSC_OP"[@id='%s]" */ #define RSC_OP_TEMPLATE "//"XML_TAG_DIFF_ADDED"//"XML_TAG_CIB"//"XML_LRM_TAG_RSC_OP"[@id='%s']" -void -update_stonith_max_attempts(const char* value) -{ - if (safe_str_eq(value, CRM_INFINITY_S)) { - stonith_max_attempts = CRM_SCORE_INFINITY; - } - else { - stonith_max_attempts = crm_int_helper(value, NULL); - } -} static void te_update_diff_v1(const char *event, xmlNode *diff) { int lpc, max; xmlXPathObject *xpathObj = NULL; CRM_CHECK(diff != NULL, return); xml_log_patchset(LOG_TRACE, __FUNCTION__, diff); if (cib_config_changed(NULL, NULL, &diff)) { abort_transition(INFINITY, tg_restart, "Non-status change", diff); goto bail; /* configuration changed */ } /* Tickets Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: update", aborted); goto bail; } freeXpathObject(xpathObj); /* Tickets Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_TICKETS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Ticket attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* Transient Attributes - Added/Updated */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_TRANSIENT_NODEATTRS "//" XML_CIB_TAG_NVPAIR); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *attr = getXpathResult(xpathObj, lpc); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = NULL; if (safe_str_eq(CRM_OP_PROBED, name)) { value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); } if (crm_is_true(value) == FALSE) { abort_transition(INFINITY, tg_restart, "Transient attribute: update", attr); crm_log_xml_trace(attr, "Abort"); goto bail; } } freeXpathObject(xpathObj); /* Transient Attributes - Removed */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_TRANSIENT_NODEATTRS); if (numXpathResults(xpathObj) > 0) { xmlNode *aborted = getXpathResult(xpathObj, 0); abort_transition(INFINITY, tg_restart, "Transient attribute: removal", aborted); goto bail; } freeXpathObject(xpathObj); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if (transition_graph->pending == 0) { xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RESOURCE); max = numXpathResults(xpathObj); if (max > 1) { crm_debug("Ignoring resource operation updates due to history refresh of %d resources", max); crm_log_xml_trace(diff, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); goto bail; } freeXpathObject(xpathObj); } /* Process operation updates */ xpathObj = xpath_search(diff, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); if (max > 0) { int lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } } freeXpathObject(xpathObj); /* Detect deleted (as opposed to replaced or added) actions - eg. crm_resource -C */ xpathObj = xpath_search(diff, "//" XML_TAG_DIFF_REMOVED "//" XML_LRM_TAG_RSC_OP); max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { int path_max = 0; const char *op_id = NULL; char *rsc_op_xpath = NULL; xmlXPathObject *op_match = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; op_id = ID(match); path_max = strlen(RSC_OP_TEMPLATE) + strlen(op_id) + 1; rsc_op_xpath = calloc(1, path_max); snprintf(rsc_op_xpath, path_max, RSC_OP_TEMPLATE, op_id); op_match = xpath_search(diff, rsc_op_xpath); if (numXpathResults(op_match) == 0) { /* Prevent false positives by matching cancelations too */ const char *node = get_node_id(match); crm_action_t *cancelled = get_cancel_action(op_id, node); if (cancelled == NULL) { crm_debug("No match for deleted action %s (%s on %s)", rsc_op_xpath, op_id, node); abort_transition(INFINITY, tg_restart, "Resource op removal", match); freeXpathObject(op_match); free(rsc_op_xpath); goto bail; } else { crm_debug("Deleted lrm_rsc_op %s on %s was for graph event %d", op_id, node, cancelled->id); } } freeXpathObject(op_match); free(rsc_op_xpath); } bail: freeXpathObject(xpathObj); } static void process_lrm_resource_diff(xmlNode *lrm_resource, const char *node) { for (xmlNode *rsc_op = __xml_first_child(lrm_resource); rsc_op != NULL; rsc_op = __xml_next(rsc_op)) { process_graph_event(rsc_op, node); } } static void process_resource_updates(const char *node, xmlNode *xml, xmlNode *change, const char *op, const char *xpath) { xmlNode *rsc = NULL; if (xml == NULL) { return; } else if (strcmp((const char*)xml->name, XML_CIB_TAG_LRM) == 0) { xml = first_named_child(xml, XML_LRM_TAG_RESOURCES); crm_trace("Got %p in %s", xml, XML_CIB_TAG_LRM); } CRM_ASSERT(strcmp((const char*)xml->name, XML_LRM_TAG_RESOURCES) == 0); /* * Updates by, or in response to, TE actions will never contain updates * for more than one resource at a time, so such updates indicate an * LRM refresh. * * In that case, start a new transition rather than check each result * individually, which can result in _huge_ speedups in large clusters. * * Unfortunately, we can only do so when there are no pending actions. * Otherwise, we could mistakenly throw away those results here, and * the cluster will stall waiting for them and time out the operation. */ if ((transition_graph->pending == 0) && xml->children && xml->children->next) { crm_log_xml_trace(change, "lrm-refresh"); abort_transition(INFINITY, tg_restart, "History refresh", NULL); return; } for (rsc = __xml_first_child(xml); rsc != NULL; rsc = __xml_next(rsc)) { crm_trace("Processing %s", ID(rsc)); process_lrm_resource_diff(rsc, node); } } #define NODE_PATT "/lrm[@id=" static char *get_node_from_xpath(const char *xpath) { char *nodeid = NULL; char *tmp = strstr(xpath, NODE_PATT); if(tmp) { tmp += strlen(NODE_PATT); tmp += 1; nodeid = strdup(tmp); tmp = strstr(nodeid, "\'"); CRM_ASSERT(tmp); tmp[0] = 0; } return nodeid; } static char *extract_node_uuid(const char *xpath) { char *mutable_path = strdup(xpath); char *node_uuid = NULL; char *search = NULL; char *match = NULL; match = strstr(mutable_path, "node_state[@id=\'"); if (match == NULL) { free(mutable_path); return NULL; } match += strlen("node_state[@id=\'"); search = strchr(match, '\''); if (search == NULL) { free(mutable_path); return NULL; } search[0] = 0; node_uuid = strdup(match); free(mutable_path); return node_uuid; } static void abort_unless_down(const char *xpath, const char *op, xmlNode *change, const char *reason) { char *node_uuid = NULL; crm_action_t *down = NULL; if(safe_str_neq(op, "delete")) { abort_transition(INFINITY, tg_restart, reason, change); return; } node_uuid = extract_node_uuid(xpath); if(node_uuid == NULL) { crm_err("Could not extract node ID from %s", xpath); abort_transition(INFINITY, tg_restart, reason, change); return; } down = match_down_event(node_uuid); if (down == NULL) { crm_trace("Not expecting %s to be down (%s)", node_uuid, xpath); abort_transition(INFINITY, tg_restart, reason, change); } else { crm_trace("Expecting changes to %s (%s)", node_uuid, xpath); } free(node_uuid); } static void process_op_deletion(const char *xpath, xmlNode *change) { char *mutable_key = strdup(xpath); char *key; char *node_uuid; crm_action_t *cancel = NULL; // Extract the part of xpath between last pair of single quotes key = strrchr(mutable_key, '\''); if (key != NULL) { *key = '\0'; key = strrchr(mutable_key, '\''); } if (key == NULL) { crm_warn("Ignoring malformed CIB update (resource deletion of %s)", xpath); free(mutable_key); return; } ++key; node_uuid = extract_node_uuid(xpath); cancel = get_cancel_action(key, node_uuid); if (cancel) { confirm_cancel_action(cancel); } else { abort_transition(INFINITY, tg_restart, "Resource operation removal", change); } free(mutable_key); free(node_uuid); } static void process_delete_diff(const char *xpath, const char *op, xmlNode *change) { if (strstr(xpath, "/" XML_LRM_TAG_RSC_OP "[")) { process_op_deletion(xpath, change); } else if (strstr(xpath, "/" XML_CIB_TAG_LRM "[")) { abort_unless_down(xpath, op, change, "Resource state removal"); } else if (strstr(xpath, "/" XML_CIB_TAG_STATE "[")) { abort_unless_down(xpath, op, change, "Node state removal"); } else { crm_trace("Ignoring delete of %s", xpath); } } static void process_node_state_diff(xmlNode *state, xmlNode *change, const char *op, const char *xpath) { xmlNode *lrm = first_named_child(state, XML_CIB_TAG_LRM); process_resource_updates(ID(state), lrm, change, op, xpath); } static void process_status_diff(xmlNode *status, xmlNode *change, const char *op, const char *xpath) { for (xmlNode *state = __xml_first_child(status); state != NULL; state = __xml_next(state)) { process_node_state_diff(state, change, op, xpath); } } static void process_cib_diff(xmlNode *cib, xmlNode *change, const char *op, const char *xpath) { xmlNode *status = first_named_child(cib, XML_CIB_TAG_STATUS); xmlNode *config = first_named_child(cib, XML_CIB_TAG_CONFIGURATION); if (status) { process_status_diff(status, change, op, xpath); } if (config) { abort_transition(INFINITY, tg_restart, "Non-status-only change", change); } } static void te_update_diff_v2(xmlNode *diff) { crm_log_xml_trace(diff, "Patch:Raw"); for (xmlNode *change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) { xmlNode *match = NULL; const char *name = NULL; const char *xpath = crm_element_value(change, XML_DIFF_PATH); // Possible ops: create, modify, delete, move const char *op = crm_element_value(change, XML_DIFF_OP); // Ignore uninteresting updates if (op == NULL) { continue; } else if (xpath == NULL) { crm_trace("Ignoring %s change for version field", op); continue; } else if (strcmp(op, "move") == 0) { crm_trace("Ignoring move change at %s", xpath); continue; } // Find the result of create/modify ops if (strcmp(op, "create") == 0) { match = change->children; } else if (strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } else if (strcmp(op, "delete") != 0) { crm_warn("Ignoring malformed CIB update (%s operation on %s is unrecognized)", op, xpath); continue; } if (match) { if (match->type == XML_COMMENT_NODE) { crm_trace("Ignoring %s operation for comment at %s", op, xpath); continue; } name = (const char *)match->name; } crm_trace("Handling %s operation for %s%s%s", op, (xpath? xpath : "CIB"), (name? " matched by " : ""), (name? name : "")); if (strstr(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION)) { abort_transition(INFINITY, tg_restart, "Configuration change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_CIB_TAG_TICKETS) || safe_str_eq(name, XML_CIB_TAG_TICKETS)) { abort_transition(INFINITY, tg_restart, "Ticket attribute change", change); break; // Won't be packaged with operation results we may be waiting for } else if (strstr(xpath, "/" XML_TAG_TRANSIENT_NODEATTRS "[") || safe_str_eq(name, XML_TAG_TRANSIENT_NODEATTRS)) { abort_unless_down(xpath, op, change, "Transient attribute change"); break; // Won't be packaged with operation results we may be waiting for } else if (strcmp(op, "delete") == 0) { process_delete_diff(xpath, op, change); } else if (name == NULL) { crm_warn("Ignoring malformed CIB update (%s at %s has no result)", op, xpath); } else if (strcmp(name, XML_TAG_CIB) == 0) { process_cib_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATUS) == 0) { process_status_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_STATE) == 0) { process_node_state_diff(match, change, op, xpath); } else if (strcmp(name, XML_CIB_TAG_LRM) == 0) { process_resource_updates(ID(match), match, change, op, xpath); } else if (strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = get_node_from_xpath(xpath); process_resource_updates(local_node, match, change, op, xpath); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = get_node_from_xpath(xpath); process_lrm_resource_diff(match, local_node); free(local_node); } else if (strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = get_node_from_xpath(xpath); process_graph_event(match, local_node); free(local_node); } else { crm_warn("Ignoring malformed CIB update (%s at %s has unrecognized result %s)", op, xpath, name); } } } void te_update_diff(const char *event, xmlNode * msg) { xmlNode *diff = NULL; const char *op = NULL; int rc = -EINVAL; int format = 1; int p_add[] = { 0, 0, 0 }; int p_del[] = { 0, 0, 0 }; CRM_CHECK(msg != NULL, return); crm_element_value_int(msg, F_CIB_RC, &rc); if (transition_graph == NULL) { crm_trace("No graph"); return; } else if (rc < pcmk_ok) { crm_trace("Filter rc=%d (%s)", rc, pcmk_strerror(rc)); return; } else if (transition_graph->complete && fsa_state != S_IDLE && fsa_state != S_TRANSITION_ENGINE && fsa_state != S_POLICY_ENGINE) { crm_trace("Filter state=%s, complete=%d", fsa_state2string(fsa_state), transition_graph->complete); return; } op = crm_element_value(msg, F_CIB_OPERATION); diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); xml_patch_versions(diff, p_add, p_del); crm_debug("Processing (%s) diff: %d.%d.%d -> %d.%d.%d (%s)", op, p_del[0], p_del[1], p_del[2], p_add[0], p_add[1], p_add[2], fsa_state2string(fsa_state)); crm_element_value_int(diff, "format", &format); switch (format) { case 1: te_update_diff_v1(event, diff); break; case 2: te_update_diff_v2(diff); break; default: crm_warn("Ignoring malformed CIB update (unknown patch format %d)", format); } } gboolean process_te_message(xmlNode * msg, xmlNode * xml_data) { const char *from = crm_element_value(msg, F_ORIG); const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO); const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM); const char *ref = crm_element_value(msg, F_CRM_REFERENCE); const char *op = crm_element_value(msg, F_CRM_TASK); const char *type = crm_element_value(msg, F_CRM_MSG_TYPE); crm_trace("Processing %s (%s) message", op, ref); crm_log_xml_trace(msg, "ipc"); if (op == NULL) { /* error */ } else if (sys_to == NULL || strcasecmp(sys_to, CRM_SYSTEM_TENGINE) != 0) { crm_trace("Bad sys-to %s", crm_str(sys_to)); return FALSE; } else if (safe_str_eq(op, CRM_OP_INVOKE_LRM) && safe_str_eq(sys_from, CRM_SYSTEM_LRMD) /* && safe_str_eq(type, XML_ATTR_RESPONSE) */ ) { xmlXPathObject *xpathObj = NULL; crm_log_xml_trace(msg, "Processing (N)ACK"); crm_debug("Processing (N)ACK %s from %s", crm_element_value(msg, F_CRM_REFERENCE), from); xpathObj = xpath_search(xml_data, "//" XML_LRM_TAG_RSC_OP); if (numXpathResults(xpathObj)) { int lpc = 0, max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); const char *node = get_node_id(rsc_op); process_graph_event(rsc_op, node); } freeXpathObject(xpathObj); } else { crm_log_xml_err(msg, "Invalid (N)ACK"); freeXpathObject(xpathObj); return FALSE; } } else { crm_err("Unknown command: %s::%s from %s", type, op, sys_from); } crm_trace("finished processing message"); return TRUE; } -GHashTable *stonith_failures = NULL; -struct st_fail_rec { - int count; -}; - -static gboolean -too_many_st_failures(const char *target) -{ - GHashTableIter iter; - const char *key = NULL; - struct st_fail_rec *value = NULL; - - if (stonith_failures == NULL) { - return FALSE; - } - - if (target == NULL) { - g_hash_table_iter_init(&iter, stonith_failures); - while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { - if (value->count >= stonith_max_attempts) { - target = (const char*)key; - goto too_many; - } - } - } else { - value = g_hash_table_lookup(stonith_failures, target); - if ((value != NULL) && (value->count >= stonith_max_attempts)) { - goto too_many; - } - } - return FALSE; - -too_many: - crm_warn("Too many failures (%d) to fence %s, giving up", - value->count, target); - return TRUE; -} - -/*! - * \internal - * \brief Reset a stonith fail count - * - * \param[in] target Name of node to reset, or NULL for all - */ -void -st_fail_count_reset(const char *target) -{ - if (stonith_failures == NULL) { - return; - } - - if (target) { - struct st_fail_rec *rec = NULL; - - rec = g_hash_table_lookup(stonith_failures, target); - if (rec) { - rec->count = 0; - } - } else { - GHashTableIter iter; - const char *key = NULL; - struct st_fail_rec *rec = NULL; - - g_hash_table_iter_init(&iter, stonith_failures); - while (g_hash_table_iter_next(&iter, (gpointer *) &key, - (gpointer *) &rec)) { - rec->count = 0; - } - } -} - -void -st_fail_count_increment(const char *target) -{ - struct st_fail_rec *rec = NULL; - - if (stonith_failures == NULL) { - stonith_failures = crm_str_table_new(); - } - - rec = g_hash_table_lookup(stonith_failures, target); - if (rec) { - rec->count++; - } else { - rec = malloc(sizeof(struct st_fail_rec)); - if(rec == NULL) { - return; - } - - rec->count = 1; - g_hash_table_insert(stonith_failures, strdup(target), rec); - } -} - -/*! - * \internal - * \brief Abort transition due to stonith failure - * - * \param[in] abort_action Whether to restart or stop transition - * \param[in] target Don't restart if this (NULL for any) has too many failures - * \param[in] reason Log this stonith action XML as abort reason (or NULL) - */ -void -abort_for_stonith_failure(enum transition_action abort_action, - const char *target, xmlNode *reason) -{ - /* If stonith repeatedly fails, we eventually give up on starting a new - * transition for that reason. - */ - if ((abort_action != tg_stop) && too_many_st_failures(target)) { - abort_action = tg_stop; - } - abort_transition(INFINITY, abort_action, "Stonith failed", reason); -} - -void -tengine_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data) -{ - char *uuid = NULL; - int stonith_id = -1; - int transition_id = -1; - crm_action_t *action = NULL; - int call_id = data->call_id; - int rc = data->rc; - char *userdata = data->userdata; - - CRM_CHECK(userdata != NULL, return); - crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata, - pcmk_strerror(rc), rc); - - if (AM_I_DC == FALSE) { - return; - } - - /* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */ - /* op->call_id, op->optype, op->node_name, op->op_result, */ - /* (char *)op->node_list, op->private_data); */ - - /* filter out old STONITH actions */ - CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, NULL), - goto bail); - - if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid) - || transition_graph->id != transition_id) { - crm_info("Ignoring STONITH action initiated outside of the current transition"); - goto bail; - } - - action = get_action(stonith_id, FALSE); - if (action == NULL) { - crm_err("Stonith action not matched"); - goto bail; - } - - stop_te_timer(action->timer); - if (rc == pcmk_ok) { - const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); - const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); - const char *op = crm_meta_value(action->params, "stonith_action"); - - crm_info("Stonith operation %d for %s passed", call_id, target); - if (action->confirmed == FALSE) { - te_action_confirmed(action); - if (safe_str_eq("on", op)) { - const char *value = NULL; - char *now = crm_itoa(time(NULL)); - - update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE); - free(now); - - value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL); - update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE); - - value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE); - update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE); - - } else if (action->sent_update == FALSE) { - send_stonith_update(action, target, uuid); - action->sent_update = TRUE; - } - } - st_fail_count_reset(target); - - } else { - const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); - enum transition_action abort_action = tg_restart; - - action->failed = TRUE; - crm_notice("Stonith operation %d for %s failed (%s): aborting transition.", - call_id, target, pcmk_strerror(rc)); - - /* If no fence devices were available, there's no use in immediately - * checking again, so don't start a new transition in that case. - */ - if (rc == -ENODEV) { - crm_warn("No devices found in cluster to fence %s, giving up", - target); - abort_action = tg_stop; - } - - /* Increment the fail count now, so abort_for_stonith_failure() can - * check it. Non-DC nodes will increment it in tengine_stonith_notify(). - */ - st_fail_count_increment(target); - abort_for_stonith_failure(abort_action, target, NULL); - } - - update_graph(transition_graph, action); - trigger_graph(); - - bail: - free(userdata); - free(uuid); - return; -} - -void -cib_fencing_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) -{ - if (rc < pcmk_ok) { - crm_err("Fencing update %d for %s: failed - %s (%d)", - call_id, (char *)user_data, pcmk_strerror(rc), rc); - crm_log_xml_warn(msg, "Failed update"); - abort_transition(INFINITY, tg_shutdown, "CIB update failed", NULL); - - } else { - crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data); - } -} - void cib_action_updated(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc < pcmk_ok) { crm_err("Update %d FAILED: %s", call_id, pcmk_strerror(rc)); } } /*! * \brief Handle a timeout in node-to-node communication * * \param[in] data Pointer to action timer * * \return FALSE (indicating that source should be not be re-added) */ gboolean action_timer_callback(gpointer data) { crm_action_timer_t *timer = NULL; const char *task = NULL; const char *on_node = NULL; const char *via_node = NULL; CRM_CHECK(data != NULL, return FALSE); timer = (crm_action_timer_t *) data; stop_te_timer(timer); CRM_CHECK(timer->action != NULL, return FALSE); task = crm_element_value(timer->action->xml, XML_LRM_ATTR_TASK); on_node = crm_element_value(timer->action->xml, XML_LRM_ATTR_TARGET); via_node = crm_element_value(timer->action->xml, XML_LRM_ATTR_ROUTER_NODE); if (transition_graph->complete) { crm_notice("Node %s did not send %s result (via %s) within %dms " "(ignoring because transition not in progress)", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), timer->timeout); } else { /* fail the action */ crm_err("Node %s did not send %s result (via %s) within %dms " "(action timeout plus cluster-delay)", (on_node? on_node : ""), (task? task : "unknown action"), (via_node? via_node : "controller"), timer->timeout); print_action(LOG_ERR, "Aborting transition, action lost: ", timer->action); timer->action->failed = TRUE; te_action_confirmed(timer->action); abort_transition(INFINITY, tg_restart, "Action lost", NULL); update_graph(transition_graph, timer->action); trigger_graph(); // Record timeout in the CIB if appropriate if ((timer->action->type == action_type_rsc) && controld_action_is_recordable(task)) { controld_record_action_timeout(timer->action); } } return FALSE; } diff --git a/daemons/controld/controld_te_utils.c b/daemons/controld/controld_te_utils.c index 5606ed654a..149624479c 100644 --- a/daemons/controld/controld_te_utils.c +++ b/daemons/controld/controld_te_utils.c @@ -1,718 +1,298 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * 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 - -crm_trigger_t *stonith_reconnect = NULL; -static crm_trigger_t *stonith_history_sync_trigger = NULL; -static mainloop_timer_t *stonith_history_sync_timer = NULL; - -/* - * stonith cleanup list - * - * If the DC is shot, proper notifications might not go out. - * The stonith cleanup list allows the cluster to (re-)send - * notifications once a new DC is elected. - */ - -static GListPtr stonith_cleanup_list = NULL; - -/*! - * \internal - * \brief Add a node to the stonith cleanup list - * - * \param[in] target Name of node to add - */ -void -add_stonith_cleanup(const char *target) { - stonith_cleanup_list = g_list_append(stonith_cleanup_list, strdup(target)); -} - -/*! - * \internal - * \brief Remove a node from the stonith cleanup list - * - * \param[in] Name of node to remove - */ -void -remove_stonith_cleanup(const char *target) -{ - GListPtr iter = stonith_cleanup_list; - - while (iter != NULL) { - GListPtr tmp = iter; - char *iter_name = tmp->data; - - iter = iter->next; - if (safe_str_eq(target, iter_name)) { - crm_trace("Removing %s from the cleanup list", iter_name); - stonith_cleanup_list = g_list_delete_link(stonith_cleanup_list, tmp); - free(iter_name); - } - } -} - -/*! - * \internal - * \brief Purge all entries from the stonith cleanup list - */ -void -purge_stonith_cleanup() -{ - if (stonith_cleanup_list) { - GListPtr iter = NULL; - - for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { - char *target = iter->data; - - crm_info("Purging %s from stonith cleanup list", target); - free(target); - } - g_list_free(stonith_cleanup_list); - stonith_cleanup_list = NULL; - } -} - -/*! - * \internal - * \brief Send stonith updates for all entries in cleanup list, then purge it - */ -void -execute_stonith_cleanup() -{ - GListPtr iter; - - for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) { - char *target = iter->data; - crm_node_t *target_node = crm_get_peer(0, target); - const char *uuid = crm_peer_uuid(target_node); - - crm_notice("Marking %s, target of a previous stonith action, as clean", target); - send_stonith_update(NULL, target, uuid); - free(target); - } - g_list_free(stonith_cleanup_list); - stonith_cleanup_list = NULL; -} - -/* end stonith cleanup list functions */ - -static gboolean -fail_incompletable_stonith(crm_graph_t * graph) -{ - GListPtr lpc = NULL; - const char *task = NULL; - xmlNode *last_action = NULL; - - if (graph == NULL) { - return FALSE; - } - - for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) { - GListPtr lpc2 = NULL; - synapse_t *synapse = (synapse_t *) lpc->data; - - if (synapse->confirmed) { - continue; - } - - for (lpc2 = synapse->actions; lpc2 != NULL; lpc2 = lpc2->next) { - crm_action_t *action = (crm_action_t *) lpc2->data; - - if (action->type != action_type_crm || action->confirmed) { - continue; - } - - task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); - if (task && safe_str_eq(task, CRM_OP_FENCE)) { - action->failed = TRUE; - last_action = action->xml; - update_graph(graph, action); - crm_notice("Failing action %d (%s): fencer terminated", - action->id, ID(action->xml)); - } - } - } - - if (last_action != NULL) { - crm_warn("Fencer failure resulted in unrunnable actions"); - abort_for_stonith_failure(tg_restart, NULL, last_action); - return TRUE; - } - - return FALSE; -} - -static void -tengine_stonith_connection_destroy(stonith_t * st, stonith_event_t * e) -{ - if (is_set(fsa_input_register, R_ST_REQUIRED)) { - crm_crit("Fencing daemon connection failed"); - mainloop_set_trigger(stonith_reconnect); - - } else { - crm_info("Fencing daemon disconnected"); - } - - /* cbchan will be garbage at this point, arrange for it to be reset */ - if(stonith_api) { - stonith_api->state = stonith_disconnected; - } - - if (AM_I_DC) { - fail_incompletable_stonith(transition_graph); - trigger_graph(); - } -} - -char *te_client_id = NULL; - -#ifdef HAVE_SYS_REBOOT_H -# include -# include -#endif - -static void -tengine_stonith_notify(stonith_t * st, stonith_event_t * st_event) -{ - if(te_client_id == NULL) { - te_client_id = crm_strdup_printf("%s.%lu", crm_system_name, - (unsigned long) getpid()); - } - - if (st_event == NULL) { - crm_err("Notify data not found"); - return; - } - - crmd_alert_fencing_op(st_event); - - if (st_event->result == pcmk_ok && safe_str_eq("on", st_event->action)) { - crm_notice("%s was successfully unfenced by %s (at the request of %s)", - st_event->target, st_event->executioner ? st_event->executioner : "", st_event->origin); - /* TODO: Hook up st_event->device */ - return; - - } else if (safe_str_eq("on", st_event->action)) { - crm_err("Unfencing of %s by %s failed: %s (%d)", - st_event->target, st_event->executioner ? st_event->executioner : "", - pcmk_strerror(st_event->result), st_event->result); - return; - - } else if (st_event->result == pcmk_ok && crm_str_eq(st_event->target, fsa_our_uname, TRUE)) { - crm_crit("We were allegedly just fenced by %s for %s!", - st_event->executioner ? st_event->executioner : "", st_event->origin); /* Dumps blackbox if enabled */ - - qb_log_fini(); /* Try to get the above log message to disk - somehow */ - - /* Get out ASAP and do not come back up. - * - * Triggering a reboot is also not the worst idea either since - * the rest of the cluster thinks we're safely down - */ - -#ifdef RB_HALT_SYSTEM - reboot(RB_HALT_SYSTEM); -#endif - - /* - * If reboot() fails or is not supported, coming back up will - * probably lead to a situation where the other nodes set our - * status to 'lost' because of the fencing callback and will - * discard subsequent election votes with: - * - * Election 87 (current: 5171, owner: 103): Processed vote from east-03 (Peer is not part of our cluster) - * - * So just stay dead, something is seriously messed up anyway. - * - */ - exit(CRM_EX_FATAL); // None of our wrappers since we already called qb_log_fini() - return; - } - - /* Update the count of stonith failures for this target, in case we become - * DC later. The current DC has already updated its fail count in - * tengine_stonith_callback(). - */ - if (!AM_I_DC && safe_str_eq(st_event->operation, T_STONITH_NOTIFY_FENCE)) { - if (st_event->result == pcmk_ok) { - st_fail_count_reset(st_event->target); - } else { - st_fail_count_increment(st_event->target); - } - } - - crm_notice("Peer %s was%s terminated (%s) by %s on behalf of %s: %s " - CRM_XS " initiator=%s ref=%s", - st_event->target, st_event->result == pcmk_ok ? "" : " not", - st_event->action, - st_event->executioner ? st_event->executioner : "", - (st_event->client_origin? st_event->client_origin : ""), - pcmk_strerror(st_event->result), - st_event->origin, st_event->id); - - if (st_event->result == pcmk_ok) { - crm_node_t *peer = crm_find_known_peer_full(0, st_event->target, CRM_GET_PEER_ANY); - const char *uuid = NULL; - gboolean we_are_executioner = safe_str_eq(st_event->executioner, fsa_our_uname); - - if (peer == NULL) { - return; - } - - uuid = crm_peer_uuid(peer); - - crm_trace("target=%s dc=%s", st_event->target, fsa_our_dc); - if(AM_I_DC) { - /* The DC always sends updates */ - send_stonith_update(NULL, st_event->target, uuid); - - /* @TODO Ideally, at this point, we'd check whether the fenced node - * hosted any guest nodes, and call remote_node_down() for them. - * Unfortunately, the controller doesn't have a simple, reliable way - * to map hosts to guests. It might be possible to track this in the - * peer cache via crm_remote_peer_cache_refresh(). For now, we rely - * on the PE creating fence pseudo-events for the guests. - */ - - if (st_event->client_origin && safe_str_neq(st_event->client_origin, te_client_id)) { - - /* Abort the current transition graph if it wasn't us - * that invoked stonith to fence someone - */ - crm_info("External fencing operation from %s fenced %s", st_event->client_origin, st_event->target); - abort_transition(INFINITY, tg_restart, "External Fencing Operation", NULL); - } - - /* Assume it was our leader if we don't currently have one */ - } else if (((fsa_our_dc == NULL) || safe_str_eq(fsa_our_dc, st_event->target)) - && !is_set(peer->flags, crm_remote_node)) { - - crm_notice("Target %s our leader %s (recorded: %s)", - fsa_our_dc ? "was" : "may have been", st_event->target, - fsa_our_dc ? fsa_our_dc : ""); - - /* Given the CIB resyncing that occurs around elections, - * have one node update the CIB now and, if the new DC is different, - * have them do so too after the election - */ - if (we_are_executioner) { - send_stonith_update(NULL, st_event->target, uuid); - } - add_stonith_cleanup(st_event->target); - } - - /* If the target is a remote node, and we host its connection, - * immediately fail all monitors so it can be recovered quickly. - * The connection won't necessarily drop when a remote node is fenced, - * so the failure might not otherwise be detected until the next poke. - */ - if (is_set(peer->flags, crm_remote_node)) { - remote_ra_fail(st_event->target); - } - - crmd_peer_down(peer, TRUE); - } -} - -static gboolean -do_stonith_history_sync(gpointer user_data) -{ - if (stonith_api && (stonith_api->state != stonith_disconnected)) { - stonith_history_t *history = NULL; - - stonith_api->cmds->history(stonith_api, - st_opt_sync_call | st_opt_broadcast, - NULL, &history, 5); - stonith_history_free(history); - return TRUE; - } else { - crm_info("Skip triggering stonith history-sync as stonith is disconnected"); - return FALSE; - } -} - -static gboolean -stonith_history_sync_set_trigger(gpointer user_data) -{ - mainloop_set_trigger(stonith_history_sync_trigger); - return FALSE; -} - -void -te_trigger_stonith_history_sync(void) -{ - /* trigger a sync in 5s to give more nodes the - * chance to show up so that we don't create - * unnecessary stonith-history-sync traffic - */ - - /* as we are finally checking the stonith-connection - * in do_stonith_history_sync we should be fine - * leaving stonith_history_sync_time & stonith_history_sync_trigger - * around - */ - if (stonith_history_sync_trigger == NULL) { - stonith_history_sync_trigger = - mainloop_add_trigger(G_PRIORITY_LOW, - do_stonith_history_sync, NULL); - } - - if(stonith_history_sync_timer == NULL) { - stonith_history_sync_timer = - mainloop_timer_add("history_sync", 5000, - FALSE, stonith_history_sync_set_trigger, - NULL); - } - crm_info("Fence history will be synchronized cluster-wide within 5 seconds"); - mainloop_timer_start(stonith_history_sync_timer); -} - -gboolean -te_connect_stonith(gpointer user_data) -{ - int lpc = 0; - int rc = pcmk_ok; - - if (stonith_api == NULL) { - stonith_api = stonith_api_new(); - } - - if (stonith_api->state != stonith_disconnected) { - crm_trace("Still connected"); - return TRUE; - } - - for (lpc = 0; lpc < 30; lpc++) { - crm_debug("Attempting connection to fencing daemon..."); - - sleep(1); - rc = stonith_api->cmds->connect(stonith_api, crm_system_name, NULL); - - if (rc == pcmk_ok) { - break; - } - - if (user_data != NULL) { - if (is_set(fsa_input_register, R_ST_REQUIRED)) { - crm_err("Sign-in failed: triggered a retry"); - mainloop_set_trigger(stonith_reconnect); - } else { - crm_info("Sign-in failed, but no longer required"); - } - return TRUE; - } - - crm_err("Sign-in failed: pausing and trying again in 2s..."); - sleep(1); - } - - CRM_CHECK(rc == pcmk_ok, return TRUE); /* If not, we failed 30 times... just get out */ - stonith_api->cmds->register_notification(stonith_api, T_STONITH_NOTIFY_DISCONNECT, - tengine_stonith_connection_destroy); - - stonith_api->cmds->register_notification(stonith_api, T_STONITH_NOTIFY_FENCE, - tengine_stonith_notify); - - crm_trace("Connected"); - return TRUE; -} gboolean stop_te_timer(crm_action_timer_t * timer) { if (timer == NULL) { return FALSE; } if (timer->source_id != 0) { crm_trace("Stopping action timer"); g_source_remove(timer->source_id); timer->source_id = 0; } else { crm_trace("Action timer was already stopped"); return FALSE; } return TRUE; } gboolean te_graph_trigger(gpointer user_data) { enum transition_status graph_rc = -1; if (transition_graph == NULL) { crm_debug("Nothing to do"); return TRUE; } crm_trace("Invoking graph %d in state %s", transition_graph->id, fsa_state2string(fsa_state)); switch (fsa_state) { case S_STARTING: case S_PENDING: case S_NOT_DC: case S_HALT: case S_ILLEGAL: case S_STOPPING: case S_TERMINATE: return TRUE; break; default: break; } if (transition_graph->complete == FALSE) { int limit = transition_graph->batch_limit; transition_graph->batch_limit = throttle_get_total_job_limit(limit); graph_rc = run_graph(transition_graph); transition_graph->batch_limit = limit; /* Restore the configured value */ /* significant overhead... */ /* print_graph(LOG_TRACE, transition_graph); */ if (graph_rc == transition_active) { crm_trace("Transition not yet complete"); return TRUE; } else if (graph_rc == transition_pending) { crm_trace("Transition not yet complete - no actions fired"); return TRUE; } if (graph_rc != transition_complete) { crm_warn("Transition failed: %s", transition_status(graph_rc)); print_graph(LOG_NOTICE, transition_graph); } } crm_debug("Transition %d is now complete", transition_graph->id); transition_graph->complete = TRUE; notify_crmd(transition_graph); return TRUE; } void trigger_graph_processing(const char *fn, int line) { crm_trace("%s:%d - Triggered graph processing", fn, line); mainloop_set_trigger(transition_trigger); } static struct abort_timer_s { bool aborted; guint id; int priority; enum transition_action action; const char *text; } abort_timer = { 0, }; static gboolean abort_timer_popped(gpointer data) { if (AM_I_DC && (abort_timer.aborted == FALSE)) { abort_transition(abort_timer.priority, abort_timer.action, abort_timer.text, NULL); } abort_timer.id = 0; return FALSE; // do not immediately reschedule timer } /*! * \internal * \brief Abort transition after delay, if not already aborted in that time * * \param[in] abort_text Must be literal string */ void abort_after_delay(int abort_priority, enum transition_action abort_action, const char *abort_text, guint delay_ms) { if (abort_timer.id) { // Timer already in progress, stop and reschedule g_source_remove(abort_timer.id); } abort_timer.aborted = FALSE; abort_timer.priority = abort_priority; abort_timer.action = abort_action; abort_timer.text = abort_text; abort_timer.id = g_timeout_add(delay_ms, abort_timer_popped, NULL); } void abort_transition_graph(int abort_priority, enum transition_action abort_action, const char *abort_text, xmlNode * reason, const char *fn, int line) { int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; int level = LOG_INFO; xmlNode *diff = NULL; xmlNode *change = NULL; CRM_CHECK(transition_graph != NULL, return); switch (fsa_state) { case S_STARTING: case S_PENDING: case S_NOT_DC: case S_HALT: case S_ILLEGAL: case S_STOPPING: case S_TERMINATE: crm_info("Abort %s suppressed: state=%s (complete=%d)", abort_text, fsa_state2string(fsa_state), transition_graph->complete); return; default: break; } abort_timer.aborted = TRUE; /* Make sure any queued calculations are discarded ASAP */ free(fsa_pe_ref); fsa_pe_ref = NULL; if (transition_graph->complete == FALSE) { if(update_abort_priority(transition_graph, abort_priority, abort_action, abort_text)) { level = LOG_NOTICE; } } if(reason) { xmlNode *search = NULL; for(search = reason; search; search = search->parent) { if (safe_str_eq(XML_TAG_DIFF, TYPE(search))) { diff = search; break; } } if(diff) { xml_patch_versions(diff, add, del); for(search = reason; search; search = search->parent) { if (safe_str_eq(XML_DIFF_CHANGE, TYPE(search))) { change = search; break; } } } } if(reason == NULL) { do_crm_log(level, "Transition %d aborted: %s "CRM_XS" source=%s:%d complete=%s", transition_graph->id, abort_text, fn, line, (transition_graph->complete? "true" : "false")); } else if(change == NULL) { char *local_path = xml_get_path(reason); do_crm_log(level, "Transition %d aborted by %s.%s: %s " CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s", transition_graph->id, TYPE(reason), ID(reason), abort_text, add[0], add[1], add[2], fn, line, local_path, (transition_graph->complete? "true" : "false")); free(local_path); } else { const char *kind = NULL; const char *op = crm_element_value(change, XML_DIFF_OP); const char *path = crm_element_value(change, XML_DIFF_PATH); if(change == reason) { if(strcmp(op, "create") == 0) { reason = reason->children; } else if(strcmp(op, "modify") == 0) { reason = first_named_child(reason, XML_DIFF_RESULT); if(reason) { reason = reason->children; } } } kind = TYPE(reason); if(strcmp(op, "delete") == 0) { const char *shortpath = strrchr(path, '/'); do_crm_log(level, "Transition %d aborted by deletion of %s: %s " CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s", transition_graph->id, (shortpath? (shortpath + 1) : path), abort_text, add[0], add[1], add[2], fn, line, path, (transition_graph->complete? "true" : "false")); } else if (safe_str_eq(XML_CIB_TAG_NVPAIR, kind)) { do_crm_log(level, "Transition %d aborted by %s doing %s %s=%s: %s " CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s", transition_graph->id, crm_element_value(reason, XML_ATTR_ID), op, crm_element_value(reason, XML_NVPAIR_ATTR_NAME), crm_element_value(reason, XML_NVPAIR_ATTR_VALUE), abort_text, add[0], add[1], add[2], fn, line, path, (transition_graph->complete? "true" : "false")); } else if (safe_str_eq(XML_LRM_TAG_RSC_OP, kind)) { const char *magic = crm_element_value(reason, XML_ATTR_TRANSITION_MAGIC); do_crm_log(level, "Transition %d aborted by operation %s '%s' on %s: %s " CRM_XS " magic=%s cib=%d.%d.%d source=%s:%d complete=%s", transition_graph->id, crm_element_value(reason, XML_LRM_ATTR_TASK_KEY), op, crm_element_value(reason, XML_LRM_ATTR_TARGET), abort_text, magic, add[0], add[1], add[2], fn, line, (transition_graph->complete? "true" : "false")); } else if (safe_str_eq(XML_CIB_TAG_STATE, kind) || safe_str_eq(XML_CIB_TAG_NODE, kind)) { const char *uname = crm_peer_uname(ID(reason)); do_crm_log(level, "Transition %d aborted by %s '%s' on %s: %s " CRM_XS " cib=%d.%d.%d source=%s:%d complete=%s", transition_graph->id, kind, op, (uname? uname : ID(reason)), abort_text, add[0], add[1], add[2], fn, line, (transition_graph->complete? "true" : "false")); } else { const char *id = ID(reason); do_crm_log(level, "Transition %d aborted by %s.%s '%s': %s " CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s", transition_graph->id, TYPE(reason), (id? id : ""), (op? op : "change"), abort_text, add[0], add[1], add[2], fn, line, path, (transition_graph->complete? "true" : "false")); } } if (transition_graph->complete) { if (transition_timer->period_ms > 0) { crm_timer_stop(transition_timer); crm_timer_start(transition_timer); } else { register_fsa_input(C_FSA_INTERNAL, I_PE_CALC, NULL); } return; } mainloop_set_trigger(transition_trigger); } diff --git a/daemons/controld/controld_transition.c b/daemons/controld/controld_transition.c index 5f164abd86..b942ab4cd9 100644 --- a/daemons/controld/controld_transition.c +++ b/daemons/controld/controld_transition.c @@ -1,215 +1,214 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include extern crm_graph_functions_t te_graph_fns; -stonith_t *stonith_api = NULL; static void global_cib_callback(const xmlNode * msg, int callid, int rc, xmlNode * output) { } static crm_graph_t * create_blank_graph(void) { crm_graph_t *a_graph = unpack_graph(NULL, NULL); a_graph->complete = TRUE; a_graph->abort_reason = "DC Takeover"; a_graph->completion_action = tg_restart; return a_graph; } /* A_TE_START, A_TE_STOP, O_TE_RESTART */ void do_te_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 init_ok = TRUE; if (action & A_TE_STOP) { if (transition_graph) { destroy_graph(transition_graph); transition_graph = NULL; } if (fsa_cib_conn) { fsa_cib_conn->cmds->del_notify_callback(fsa_cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff); } clear_bit(fsa_input_register, R_TE_CONNECTED); crm_info("Transitioner is now inactive"); } if ((action & A_TE_START) == 0) { return; } else if (is_set(fsa_input_register, R_TE_CONNECTED)) { crm_debug("The transitioner is already active"); return; } else if ((action & A_TE_START) && cur_state == S_STOPPING) { crm_info("Ignoring request to start the transitioner while shutting down"); return; } if (te_uuid == NULL) { te_uuid = crm_generate_uuid(); crm_info("Registering TE UUID: %s", te_uuid); } if (fsa_cib_conn == NULL) { crm_err("Could not set CIB callbacks"); init_ok = FALSE; } else { if (fsa_cib_conn->cmds->add_notify_callback(fsa_cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff) != pcmk_ok) { crm_err("Could not set CIB notification callback"); init_ok = FALSE; } if (fsa_cib_conn->cmds->set_op_callback(fsa_cib_conn, global_cib_callback) != pcmk_ok) { crm_err("Could not set CIB global callback"); init_ok = FALSE; } } if (init_ok) { set_graph_functions(&te_graph_fns); if (transition_graph) { destroy_graph(transition_graph); } /* create a blank one */ crm_debug("Transitioner is now active"); transition_graph = create_blank_graph(); set_bit(fsa_input_register, R_TE_CONNECTED); } } /* A_TE_INVOKE, A_TE_CANCEL */ void do_te_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) { if (AM_I_DC == FALSE || (fsa_state != S_TRANSITION_ENGINE && (action & A_TE_INVOKE))) { crm_notice("No need to invoke the TE (%s) in state %s", fsa_action2string(action), fsa_state2string(fsa_state)); return; } if (action & A_TE_CANCEL) { crm_debug("Cancelling the transition: %s", transition_graph->complete ? "inactive" : "active"); abort_transition(INFINITY, tg_restart, "Peer Cancelled", NULL); if (transition_graph->complete == FALSE) { crmd_fsa_stall(FALSE); } } else if (action & A_TE_HALT) { crm_debug("Halting the transition: %s", transition_graph->complete ? "inactive" : "active"); abort_transition(INFINITY, tg_stop, "Peer Halt", NULL); if (transition_graph->complete == FALSE) { crmd_fsa_stall(FALSE); } } else if (action & A_TE_INVOKE) { const char *value = NULL; xmlNode *graph_data = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *ref = crm_element_value(input->msg, XML_ATTR_REFERENCE); const char *graph_file = crm_element_value(input->msg, F_CRM_TGRAPH); const char *graph_input = crm_element_value(input->msg, F_CRM_TGRAPH_INPUT); if (graph_file == NULL && input->xml == NULL) { crm_log_xml_err(input->msg, "Bad command"); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); return; } if (transition_graph->complete == FALSE) { crm_info("Another transition is already active"); abort_transition(INFINITY, tg_restart, "Transition Active", NULL); return; } if (fsa_pe_ref == NULL || safe_str_neq(fsa_pe_ref, ref)) { crm_info("Transition is redundant: %s vs. %s", crm_str(fsa_pe_ref), crm_str(ref)); abort_transition(INFINITY, tg_restart, "Transition Redundant", NULL); } graph_data = input->xml; if (graph_data == NULL && graph_file != NULL) { graph_data = filename2xml(graph_file); } if (is_timer_started(transition_timer)) { crm_debug("The transitioner wait for a transition timer"); return; } CRM_CHECK(graph_data != NULL, crm_err("Input raised by %s is invalid", msg_data->origin); crm_log_xml_err(input->msg, "Bad command"); return); destroy_graph(transition_graph); transition_graph = unpack_graph(graph_data, graph_input); if (transition_graph == NULL) { CRM_CHECK(transition_graph != NULL,); transition_graph = create_blank_graph(); return; } crm_info("Processing graph %d (ref=%s) derived from %s", transition_graph->id, ref, graph_input); te_reset_job_counts(); value = crm_element_value(graph_data, "failed-stop-offset"); if (value) { free(failed_stop_offset); failed_stop_offset = strdup(value); } value = crm_element_value(graph_data, "failed-start-offset"); if (value) { free(failed_start_offset); failed_start_offset = strdup(value); } trigger_graph(); print_graph(LOG_TRACE, transition_graph); if (graph_data != input->xml) { free_xml(graph_data); } } } diff --git a/daemons/controld/controld_transition.h b/daemons/controld/controld_transition.h index a162f9947b..f31ac2d9a6 100644 --- a/daemons/controld/controld_transition.h +++ b/daemons/controld/controld_transition.h @@ -1,84 +1,65 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef TENGINE__H # define TENGINE__H # include # include # include # include -extern stonith_t *stonith_api; -extern void send_stonith_update(crm_action_t * stonith_action, const char *target, - const char *uuid); - -/* stonith cleanup list */ -void add_stonith_cleanup(const char *target); -void remove_stonith_cleanup(const char *target); -void purge_stonith_cleanup(void); -void execute_stonith_cleanup(void); /* tengine */ extern crm_action_t *match_down_event(const char *target); extern crm_action_t *get_cancel_action(const char *id, const char *node); void confirm_cancel_action(crm_action_t *cancel); void controld_record_action_timeout(crm_action_t *action); extern gboolean fail_incompletable_actions(crm_graph_t * graph, const char *down_node); void process_graph_event(xmlNode *event, const char *event_node); /* utils */ extern crm_action_t *get_action(int id, gboolean confirmed); extern gboolean stop_te_timer(crm_action_timer_t * timer); extern const char *get_rsc_state(const char *task, enum op_status status); /* unpack */ extern gboolean process_te_message(xmlNode * msg, xmlNode * xml_data); extern crm_graph_t *transition_graph; extern crm_trigger_t *transition_trigger; extern char *te_uuid; extern void notify_crmd(crm_graph_t * graph); -void cib_fencing_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, - void *user_data); void cib_action_updated(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data); gboolean action_timer_callback(gpointer data); gboolean te_graph_trigger(gpointer user_data); void te_update_diff(const char *event, xmlNode *msg); -void tengine_stonith_callback(stonith_t *stonith, - stonith_callback_data_t *data); -void update_stonith_max_attempts(const char* value); extern void trigger_graph_processing(const char *fn, int line); void abort_after_delay(int abort_priority, enum transition_action abort_action, const char *abort_text, guint delay_ms); extern void abort_transition_graph(int abort_priority, enum transition_action abort_action, const char *abort_text, xmlNode * reason, const char *fn, int line); # define trigger_graph() trigger_graph_processing(__FUNCTION__, __LINE__) # define abort_transition(pri, action, text, reason) \ abort_transition_graph(pri, action, text, reason,__FUNCTION__,__LINE__); -extern gboolean te_connect_stonith(gpointer user_data); - -extern void te_trigger_stonith_history_sync(void); - extern crm_trigger_t *transition_trigger; -extern crm_trigger_t *stonith_reconnect; extern char *failed_stop_offset; extern char *failed_start_offset; void te_action_confirmed(crm_action_t * action); void te_reset_job_counts(void); #endif diff --git a/daemons/controld/controld_utils.h b/daemons/controld/controld_utils.h index 68992f5694..8b80e3c141 100644 --- a/daemons/controld/controld_utils.h +++ b/daemons/controld/controld_utils.h @@ -1,110 +1,106 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef CRMD_UTILS__H # define CRMD_UTILS__H # include # include # include # include /* For CIB_OP_MODIFY */ # include "controld_fsa.h" // For fsa_cib_conn # include "controld_alerts.h" # define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx" # define fsa_cib_update(section, data, options, call_id, user_name) \ if(fsa_cib_conn != NULL) { \ call_id = cib_internal_op( \ fsa_cib_conn, CIB_OP_MODIFY, NULL, section, data, \ NULL, options, user_name); \ \ } else { \ crm_err("No CIB manager connection available"); \ } static inline void fsa_cib_anon_update(const char *section, xmlNode *data) { if (fsa_cib_conn == NULL) { crm_err("No CIB connection available"); } else { int opts = cib_scope_local | cib_quorum_override | cib_can_create; fsa_cib_conn->cmds->modify(fsa_cib_conn, section, data, opts); } } extern gboolean fsa_has_quorum; extern int last_peer_update; extern int last_resource_update; enum node_update_flags { node_update_none = 0x0000, node_update_quick = 0x0001, node_update_cluster = 0x0010, node_update_peer = 0x0020, node_update_join = 0x0040, node_update_expected = 0x0100, node_update_all = node_update_cluster|node_update_peer|node_update_join|node_update_expected, }; gboolean crm_timer_stop(fsa_timer_t * timer); gboolean crm_timer_start(fsa_timer_t * timer); gboolean crm_timer_popped(gpointer data); gboolean is_timer_started(fsa_timer_t * timer); crm_exit_t crmd_exit(crm_exit_t exit_code); _Noreturn void crmd_fast_exit(crm_exit_t exit_code); void pe_subsystem_free(void); void fsa_dump_actions(long long action, const char *text); void fsa_dump_inputs(int log_level, const char *text, long long input_register); gboolean update_dc(xmlNode * msg); void crm_update_peer_join(const char *source, crm_node_t * node, enum crm_join_phase phase); xmlNode *create_node_state_update(crm_node_t *node, int flags, xmlNode *parent, const char *source); void populate_cib_nodes(enum node_update_flags flags, const char *source); void crm_update_quorum(gboolean quorum, gboolean force_update); void erase_status_tag(const char *uname, const char *tag, int options); void controld_close_attrd_ipc(void); void update_attrd(const char *host, const char *name, const char *value, const char *user_name, gboolean is_remote_node); void update_attrd_remote_node_removed(const char *host, const char *user_name); void update_attrd_clear_failures(const char *host, const char *rsc, const char *op, const char *interval_spec, gboolean is_remote_node); int crmd_join_phase_count(enum crm_join_phase phase); void crmd_join_phase_log(int level); const char *get_timer_desc(fsa_timer_t * timer); -void st_fail_count_reset(const char * target); -void st_fail_count_increment(const char *target); -void abort_for_stonith_failure(enum transition_action abort_action, - const char *target, xmlNode *reason); void crmd_peer_down(crm_node_t *peer, bool full); unsigned int cib_op_timeout(void); bool feature_set_compatible(const char *dc_version, const char *join_version); bool controld_action_is_recordable(const char *action); const char *get_node_id(xmlNode *lrm_rsc_op); /* Convenience macro for registering a CIB callback * (assumes that data can be freed with free()) */ # define fsa_register_cib_callback(id, flag, data, fn) do { \ CRM_ASSERT(fsa_cib_conn); \ fsa_cib_conn->cmds->register_callback_full( \ fsa_cib_conn, id, cib_op_timeout(), \ flag, data, #fn, fn, free); \ } while(0) #endif diff --git a/daemons/execd/pacemaker-execd.c b/daemons/execd/pacemaker-execd.c index 21bb0edde8..e2fdfcad71 100644 --- a/daemons/execd/pacemaker-execd.c +++ b/daemons/execd/pacemaker-execd.c @@ -1,630 +1,622 @@ /* * Copyright 2012-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include "pacemaker-execd.h" #if defined(HAVE_GNUTLS_GNUTLS_H) && defined(SUPPORT_REMOTE) # define ENABLE_PCMK_REMOTE // Hidden in liblrmd extern int lrmd_tls_send_msg(crm_remote_t *session, xmlNode *msg, uint32_t id, const char *msg_type); #endif static GMainLoop *mainloop = NULL; static qb_ipcs_service_t *ipcs = NULL; static stonith_t *stonith_api = NULL; int lrmd_call_id = 0; #ifdef ENABLE_PCMK_REMOTE /* whether shutdown request has been sent */ static sig_atomic_t shutting_down = FALSE; /* timer for waiting for acknowledgment of shutdown request */ static guint shutdown_ack_timer = 0; static gboolean lrmd_exit(gpointer data); #endif static void stonith_connection_destroy_cb(stonith_t * st, stonith_event_t * e) { stonith_api->state = stonith_disconnected; crm_err("Connection to fencer lost"); stonith_connection_failed(); } stonith_t * get_stonith_connection(void) { if (stonith_api && stonith_api->state == stonith_disconnected) { stonith_api_delete(stonith_api); stonith_api = NULL; } - if (!stonith_api) { - int rc = 0; - int tries = 10; + if (stonith_api == NULL) { + int rc = pcmk_ok; stonith_api = stonith_api_new(); - do { - rc = stonith_api->cmds->connect(stonith_api, "pacemaker-execd", NULL); - if (rc == pcmk_ok) { - stonith_api->cmds->register_notification(stonith_api, - T_STONITH_NOTIFY_DISCONNECT, - stonith_connection_destroy_cb); - break; - } - sleep(1); - tries--; - } while (tries); - - if (rc) { - crm_err("Unable to connect to stonith daemon to execute command. error: %s", - pcmk_strerror(rc)); + rc = stonith_api_connect_retry(stonith_api, crm_system_name, 10); + if (rc != pcmk_ok) { + crm_err("Could not connect to fencer in 10 attempts: %s " + CRM_XS " rc=%d", pcmk_strerror(rc), rc); stonith_api_delete(stonith_api); stonith_api = NULL; + } else { + stonith_api->cmds->register_notification(stonith_api, + T_STONITH_NOTIFY_DISCONNECT, + stonith_connection_destroy_cb); } } return stonith_api; } static int32_t lrmd_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid) { crm_trace("Connection %p", c); if (crm_client_new(c, uid, gid) == NULL) { return -EIO; } return 0; } static void lrmd_ipc_created(qb_ipcs_connection_t * c) { crm_client_t *new_client = crm_client_get(c); crm_trace("Connection %p", c); CRM_ASSERT(new_client != NULL); /* Now that the connection is offically established, alert * the other clients a new connection exists. */ notify_of_new_client(new_client); } static int32_t lrmd_ipc_dispatch(qb_ipcs_connection_t * c, void *data, size_t size) { uint32_t id = 0; uint32_t flags = 0; crm_client_t *client = crm_client_get(c); xmlNode *request = crm_ipcs_recv(client, data, size, &id, &flags); CRM_CHECK(client != NULL, crm_err("Invalid client"); return FALSE); CRM_CHECK(client->id != NULL, crm_err("Invalid client: %p", client); return FALSE); CRM_CHECK(flags & crm_ipc_client_response, crm_err("Invalid client request: %p", client); return FALSE); if (!request) { return 0; } if (!client->name) { const char *value = crm_element_value(request, F_LRMD_CLIENTNAME); if (value == NULL) { client->name = crm_itoa(crm_ipcs_client_pid(c)); } else { client->name = strdup(value); } } lrmd_call_id++; if (lrmd_call_id < 1) { lrmd_call_id = 1; } crm_xml_add(request, F_LRMD_CLIENTID, client->id); crm_xml_add(request, F_LRMD_CLIENTNAME, client->name); crm_xml_add_int(request, F_LRMD_CALLID, lrmd_call_id); process_lrmd_message(client, id, request); free_xml(request); return 0; } /*! * \internal * \brief Free a client connection, and exit if appropriate * * \param[in] client Client connection to free */ void lrmd_client_destroy(crm_client_t *client) { crm_client_destroy(client); #ifdef ENABLE_PCMK_REMOTE /* If we were waiting to shut down, we can now safely do so * if there are no more proxied IPC providers */ if (shutting_down && (ipc_proxy_get_provider() == NULL)) { lrmd_exit(NULL); } #endif } static int32_t lrmd_ipc_closed(qb_ipcs_connection_t * c) { crm_client_t *client = crm_client_get(c); if (client == NULL) { return 0; } crm_trace("Connection %p", c); client_disconnect_cleanup(client->id); #ifdef ENABLE_PCMK_REMOTE ipc_proxy_remove_provider(client); #endif lrmd_client_destroy(client); return 0; } static void lrmd_ipc_destroy(qb_ipcs_connection_t * c) { lrmd_ipc_closed(c); crm_trace("Connection %p", c); } static struct qb_ipcs_service_handlers lrmd_ipc_callbacks = { .connection_accept = lrmd_ipc_accept, .connection_created = lrmd_ipc_created, .msg_process = lrmd_ipc_dispatch, .connection_closed = lrmd_ipc_closed, .connection_destroyed = lrmd_ipc_destroy }; int lrmd_server_send_reply(crm_client_t * client, uint32_t id, xmlNode * reply) { crm_trace("Sending reply (%d) to client (%s)", id, client->id); switch (client->kind) { case CRM_CLIENT_IPC: return crm_ipcs_send(client, id, reply, FALSE); #ifdef ENABLE_PCMK_REMOTE case CRM_CLIENT_TLS: return lrmd_tls_send_msg(client->remote, reply, id, "reply"); #endif default: crm_err("Could not send reply: unknown client type %d", client->kind); } return -ENOTCONN; } int lrmd_server_send_notify(crm_client_t * client, xmlNode * msg) { crm_trace("Sending notification to client (%s)", client->id); switch (client->kind) { case CRM_CLIENT_IPC: if (client->ipcs == NULL) { crm_trace("Could not notify local client: disconnected"); return -ENOTCONN; } return crm_ipcs_send(client, 0, msg, crm_ipc_server_event); #ifdef ENABLE_PCMK_REMOTE case CRM_CLIENT_TLS: if (client->remote == NULL) { crm_trace("Could not notify remote client: disconnected"); return -ENOTCONN; } return lrmd_tls_send_msg(client->remote, msg, 0, "notify"); #endif default: crm_err("Could not notify client: unknown type %d", client->kind); } return -ENOTCONN; } /*! * \internal * \brief Clean up and exit immediately * * \param[in] data Ignored * * \return Doesn't return * \note This can be used as a timer callback. */ static gboolean lrmd_exit(gpointer data) { crm_info("Terminating with %d clients", crm_hash_table_size(client_connections)); if (stonith_api) { stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_DISCONNECT); stonith_api->cmds->disconnect(stonith_api); stonith_api_delete(stonith_api); } if (ipcs) { mainloop_del_ipc_server(ipcs); } #ifdef ENABLE_PCMK_REMOTE lrmd_tls_server_destroy(); ipc_proxy_cleanup(); #endif crm_client_cleanup(); g_hash_table_destroy(rsc_list); if (mainloop) { lrmd_drain_alerts(mainloop); } crm_exit(CRM_EX_OK); return FALSE; } /*! * \internal * \brief Request cluster shutdown if appropriate, otherwise exit immediately * * \param[in] nsig Signal that caused invocation (ignored) */ static void lrmd_shutdown(int nsig) { #ifdef ENABLE_PCMK_REMOTE crm_client_t *ipc_proxy = ipc_proxy_get_provider(); /* If there are active proxied IPC providers, then we may be running * resources, so notify the cluster that we wish to shut down. */ if (ipc_proxy) { if (shutting_down) { crm_notice("Waiting for cluster to stop resources before exiting"); return; } crm_info("Sending shutdown request to cluster"); if (ipc_proxy_shutdown_req(ipc_proxy) < 0) { crm_crit("Shutdown request failed, exiting immediately"); } else { /* We requested a shutdown. Now, we need to wait for an * acknowledgement from the proxy host (which ensures the proxy host * supports shutdown requests), then wait for all proxy hosts to * disconnect (which ensures that all resources have been stopped). */ shutting_down = TRUE; /* Stop accepting new proxy connections */ lrmd_tls_server_destroy(); /* Older controller versions will never acknowledge our request, so * set a fairly short timeout to exit quickly in that case. If we * get the ack, we'll defuse this timer. */ shutdown_ack_timer = g_timeout_add_seconds(20, lrmd_exit, NULL); /* Currently, we let the OS kill us if the clients don't disconnect * in a reasonable time. We could instead set a long timer here * (shorter than what the OS is likely to use) and exit immediately * if it pops. */ return; } } #endif lrmd_exit(NULL); } /*! * \internal * \brief Defuse short exit timer if shutting down */ void handle_shutdown_ack() { #ifdef ENABLE_PCMK_REMOTE if (shutting_down) { crm_info("Received shutdown ack"); if (shutdown_ack_timer > 0) { g_source_remove(shutdown_ack_timer); shutdown_ack_timer = 0; } return; } #endif crm_debug("Ignoring unexpected shutdown ack"); } /*! * \internal * \brief Make short exit timer fire immediately */ void handle_shutdown_nack() { #ifdef ENABLE_PCMK_REMOTE if (shutting_down) { crm_info("Received shutdown nack"); if (shutdown_ack_timer > 0) { g_source_remove(shutdown_ack_timer); shutdown_ack_timer = g_timeout_add(0, lrmd_exit, NULL); } return; } #endif crm_debug("Ignoring unexpected shutdown nack"); } static pid_t main_pid = 0; static void sigdone(void) { exit(CRM_EX_OK); } static void sigreap(void) { pid_t pid = 0; int status; do { /* * Opinions seem to differ as to what to put here: * -1, any child process * 0, any child process whose process group ID is equal to that of the calling process */ pid = waitpid(-1, &status, WNOHANG); if(pid == main_pid) { /* Exit when pacemaker-remote exits and use the same return code */ if (WIFEXITED(status)) { exit(WEXITSTATUS(status)); } exit(CRM_EX_ERROR); } } while (pid > 0); } static struct { int sig; void (*handler)(void); } sigmap[] = { { SIGCHLD, sigreap }, { SIGINT, sigdone }, }; static void spawn_pidone(int argc, char **argv, char **envp) { sigset_t set; if (getpid() != 1) { return; } sigfillset(&set); sigprocmask(SIG_BLOCK, &set, 0); main_pid = fork(); switch (main_pid) { case 0: sigprocmask(SIG_UNBLOCK, &set, NULL); setsid(); setpgid(0, 0); /* Child remains as pacemaker-remoted */ return; case -1: perror("fork"); } /* Parent becomes the reaper of zombie processes */ /* Safe to initialize logging now if needed */ #ifdef HAVE___PROGNAME /* Differentiate ourselves in the 'ps' output */ { char *p; int i, maxlen; char *LastArgv = NULL; const char *name = "pcmk-init"; for(i = 0; i < argc; i++) { if(!i || (LastArgv + 1 == argv[i])) LastArgv = argv[i] + strlen(argv[i]); } for(i = 0; envp[i] != NULL; i++) { if((LastArgv + 1) == envp[i]) { LastArgv = envp[i] + strlen(envp[i]); } } maxlen = (LastArgv - argv[0]) - 2; i = strlen(name); /* We can overwrite individual argv[] arguments */ snprintf(argv[0], maxlen, "%s", name); /* Now zero out everything else */ p = &argv[0][i]; while(p < LastArgv) *p++ = '\0'; argv[1] = NULL; } #endif /* HAVE___PROGNAME */ while (1) { int sig; size_t i; sigwait(&set, &sig); for (i = 0; i < DIMOF(sigmap); i++) { if (sigmap[i].sig == sig) { sigmap[i].handler(); break; } } } } /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"version", 0, 0, '$', "\tVersion information" }, {"verbose", 0, 0, 'V', "\tIncrease debug output"}, {"logfile", 1, 0, 'l', "\tSend logs to the additional named logfile"}, #ifdef ENABLE_PCMK_REMOTE {"port", 1, 0, 'p', "\tPort to listen on"}, #endif {0, 0, 0, 0} }; /* *INDENT-ON* */ #ifdef ENABLE_PCMK_REMOTE # define EXECD_TYPE "remote" #else # define EXECD_TYPE "local" #endif int main(int argc, char **argv, char **envp) { int flag = 0; int index = 0; int bump_log_num = 0; const char *option = NULL; /* If necessary, create PID1 now before any FDs are opened */ spawn_pidone(argc, argv, envp); #ifndef ENABLE_PCMK_REMOTE crm_log_preinit("pacemaker-execd", argc, argv); crm_set_options(NULL, "[options]", long_options, "Resource agent executor daemon for cluster nodes"); #else crm_log_preinit("pacemaker-remoted", argc, argv); crm_set_options(NULL, "[options]", long_options, "Resource agent executor daemon for Pacemaker Remote nodes"); #endif while (1) { flag = crm_get_option(argc, argv, &index); if (flag == -1) { break; } switch (flag) { case 'l': crm_add_logfile(optarg); break; case 'p': setenv("PCMK_remote_port", optarg, 1); break; case 'V': bump_log_num++; break; case '?': case '$': crm_help(flag, CRM_EX_OK); break; default: crm_help('?', CRM_EX_USAGE); break; } } crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE); while (bump_log_num > 0) { crm_bump_log_level(argc, argv); bump_log_num--; } option = daemon_option("logfacility"); if(option && safe_str_neq(option, "none")) { setenv("HA_LOGFACILITY", option, 1); /* Used by the ocf_log/ha_log OCF macro */ } option = daemon_option("logfile"); if(option && safe_str_neq(option, "none")) { setenv("HA_LOGFILE", option, 1); /* Used by the ocf_log/ha_log OCF macro */ if (daemon_option_enabled(crm_system_name, "debug")) { setenv("HA_DEBUGLOG", option, 1); /* Used by the ocf_log/ha_debug OCF macro */ } } crm_notice("Starting Pacemaker " EXECD_TYPE " executor"); /* The presence of this variable allegedly controls whether child * processes like httpd will try and use Systemd's sd_notify * API */ unsetenv("NOTIFY_SOCKET"); /* Used by RAs - Leave owned by root */ crm_build_path(CRM_RSCTMP_DIR, 0755); rsc_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_rsc); ipcs = mainloop_add_ipc_server(CRM_SYSTEM_LRMD, QB_IPC_SHM, &lrmd_ipc_callbacks); if (ipcs == NULL) { crm_err("Failed to create IPC server: shutting down and inhibiting respawn"); crm_exit(CRM_EX_FATAL); } #ifdef ENABLE_PCMK_REMOTE if (lrmd_init_remote_tls_server() < 0) { crm_err("Failed to create TLS listener: shutting down and staying down"); crm_exit(CRM_EX_FATAL); } ipc_proxy_init(); #endif mainloop_add_signal(SIGTERM, lrmd_shutdown); mainloop = g_main_loop_new(NULL, FALSE); crm_notice("Pacemaker " EXECD_TYPE " executor successfully started and accepting connections"); g_main_loop_run(mainloop); /* should never get here */ lrmd_exit(NULL); return CRM_EX_OK; } diff --git a/daemons/fenced/cts-fence-helper.c b/daemons/fenced/cts-fence-helper.c index c5ce1ab01a..4552fc1178 100644 --- a/daemons/fenced/cts-fence-helper.c +++ b/daemons/fenced/cts-fence-helper.c @@ -1,658 +1,654 @@ /* - * Copyright 2009-2018 Andrew Beekhof + * Copyright 2009-2019 the Pacemaker project contributors * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static GMainLoop *mainloop = NULL; static crm_trigger_t *trig = NULL; static int mainloop_iter = 0; static int callback_rc = 0; typedef void (*mainloop_test_iteration_cb) (int check_event); #define MAINLOOP_DEFAULT_TIMEOUT 2 #define mainloop_test_done(pass) \ if (pass) { \ crm_info("SUCCESS - %s", __FUNCTION__); \ mainloop_iter++; \ mainloop_set_trigger(trig); \ } else { \ crm_err("FAILURE = %s async_callback %d", __FUNCTION__, callback_rc); \ crm_exit(CRM_EX_ERROR); \ } \ callback_rc = 0; \ /* *INDENT-OFF* */ enum test_modes { test_standard = 0, // test using a specific developer environment test_passive, // watch notifications only test_api_sanity, // sanity-test stonith client API using fence_dummy test_api_mainloop, // sanity-test mainloop code with async responses }; static struct crm_option long_options[] = { {"verbose", 0, 0, 'V'}, {"version", 0, 0, '$'}, {"help", 0, 0, '?'}, {"passive", 0, 0, 'p'}, {"api_test", 0, 0, 't'}, {"mainloop_api_test", 0, 0, 'm'}, {0, 0, 0, 0} }; /* *INDENT-ON* */ static stonith_t *st = NULL; static struct pollfd pollfd; static int st_opts = st_opt_sync_call; static int expected_notifications = 0; static int verbose = 0; static void dispatch_helper(int timeout) { int rc; crm_debug("Looking for notification"); pollfd.events = POLLIN; while (true) { rc = poll(&pollfd, 1, timeout); /* wait 10 minutes, -1 forever */ if (rc > 0) { if (!stonith_dispatch(st)) { break; } } else { break; } } } static void st_callback(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { crm_exit(CRM_EX_DISCONNECT); } crm_notice("Operation %s requested by %s %s for peer %s. %s reported: %s (ref=%s)", e->operation, e->origin, e->result == pcmk_ok ? "completed" : "failed", e->target, e->executioner ? e->executioner : "", pcmk_strerror(e->result), e->id); if (expected_notifications) { expected_notifications--; } } static void st_global_callback(stonith_t * stonith, stonith_callback_data_t * data) { crm_notice("Call id %d completed with rc %d", data->call_id, data->rc); } static void passive_test(void) { int rc = 0; rc = st->cmds->connect(st, crm_system_name, &pollfd.fd); - crm_debug("Connect: %d", rc); - + if (rc != pcmk_ok) { + stonith_api_delete(st); + crm_exit(CRM_EX_DISCONNECT); + } st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, st_callback); st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, st_callback); st->cmds->register_notification(st, STONITH_OP_DEVICE_ADD, st_callback); st->cmds->register_notification(st, STONITH_OP_DEVICE_DEL, st_callback); st->cmds->register_callback(st, 0, 120, st_opt_timeout_updates, NULL, "st_global_callback", st_global_callback); dispatch_helper(600 * 1000); } #define single_test(cmd, str, num_notifications, expected_rc) \ { \ int rc = 0; \ rc = cmd; \ expected_notifications = 0; \ if (num_notifications) { \ expected_notifications = num_notifications; \ dispatch_helper(500); \ } \ if (rc != expected_rc) { \ crm_err("FAILURE - expected rc %d != %d(%s) for cmd - %s", expected_rc, rc, pcmk_strerror(rc), str); \ crm_exit(CRM_EX_ERROR); \ } else if (expected_notifications) { \ crm_err("FAILURE - expected %d notifications, got only %d for cmd - %s", \ num_notifications, num_notifications - expected_notifications, str); \ crm_exit(CRM_EX_ERROR); \ } else { \ if (verbose) { \ crm_info("SUCCESS - %s: %d", str, rc); \ } else { \ crm_debug("SUCCESS - %s: %d", str, rc); \ } \ } \ }\ static void run_fence_failure_test(void) { stonith_key_value_t *params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "false_1_node1=1,2 false_1_node2=3,4"); params = stonith_key_value_add(params, "mode", "fail"); single_test(st-> cmds->register_device(st, st_opts, "test-id1", "stonith-ng", "fence_dummy", params), "Register device1 for failure test", 1, 0); single_test(st->cmds->fence(st, st_opts, "false_1_node2", "off", 3, 0), "Fence failure results off", 1, -pcmk_err_generic); single_test(st->cmds->fence(st, st_opts, "false_1_node2", "reboot", 3, 0), "Fence failure results reboot", 1, -pcmk_err_generic); single_test(st->cmds->remove_device(st, st_opts, "test-id1"), "Remove device1 for failure test", 1, 0); stonith_key_value_freeall(params, 1, 1); } static void run_fence_failure_rollover_test(void) { stonith_key_value_t *params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "false_1_node1=1,2 false_1_node2=3,4"); params = stonith_key_value_add(params, "mode", "fail"); single_test(st-> cmds->register_device(st, st_opts, "test-id1", "stonith-ng", "fence_dummy", params), "Register device1 for rollover test", 1, 0); stonith_key_value_freeall(params, 1, 1); params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "false_1_node1=1,2 false_1_node2=3,4"); params = stonith_key_value_add(params, "mode", "pass"); single_test(st-> cmds->register_device(st, st_opts, "test-id2", "stonith-ng", "fence_dummy", params), "Register device2 for rollover test", 1, 0); single_test(st->cmds->fence(st, st_opts, "false_1_node2", "off", 3, 0), "Fence rollover results off", 1, 0); /* Expect -ENODEV because fence_dummy requires 'on' to be executed on target */ single_test(st->cmds->fence(st, st_opts, "false_1_node2", "on", 3, 0), "Fence rollover results on", 1, -ENODEV); single_test(st->cmds->remove_device(st, st_opts, "test-id1"), "Remove device1 for rollover tests", 1, 0); single_test(st->cmds->remove_device(st, st_opts, "test-id2"), "Remove device2 for rollover tests", 1, 0); stonith_key_value_freeall(params, 1, 1); } static void run_standard_test(void) { stonith_key_value_t *params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "false_1_node1=1,2 false_1_node2=3,4"); params = stonith_key_value_add(params, "mode", "pass"); params = stonith_key_value_add(params, "mock_dynamic_hosts", "false_1_node1 false_1_node2"); single_test(st-> cmds->register_device(st, st_opts, "test-id", "stonith-ng", "fence_dummy", params), "Register", 1, 0); stonith_key_value_freeall(params, 1, 1); params = NULL; single_test(st->cmds->list(st, st_opts, "test-id", NULL, 1), "list", 1, 0); single_test(st->cmds->monitor(st, st_opts, "test-id", 1), "Monitor", 1, 0); single_test(st->cmds->status(st, st_opts, "test-id", "false_1_node2", 1), "Status false_1_node2", 1, 0); single_test(st->cmds->status(st, st_opts, "test-id", "false_1_node1", 1), "Status false_1_node1", 1, 0); single_test(st->cmds->fence(st, st_opts, "unknown-host", "off", 1, 0), "Fence unknown-host (expected failure)", 0, -ENODEV); single_test(st->cmds->fence(st, st_opts, "false_1_node1", "off", 1, 0), "Fence false_1_node1", 1, 0); /* Expect -ENODEV because fence_dummy requires 'on' to be executed on target */ single_test(st->cmds->fence(st, st_opts, "false_1_node1", "on", 1, 0), "Unfence false_1_node1", 1, -ENODEV); /* Confirm that an invalid level index is rejected */ single_test(st->cmds->register_level(st, st_opts, "node1", 999, params), "Attempt to register an invalid level index", 0, -EINVAL); single_test(st->cmds->remove_device(st, st_opts, "test-id"), "Remove test-id", 1, 0); stonith_key_value_freeall(params, 1, 1); } static void sanity_tests(void) { int rc = 0; rc = st->cmds->connect(st, crm_system_name, &pollfd.fd); - crm_debug("Connect: %d", rc); - + if (rc != pcmk_ok) { + stonith_api_delete(st); + crm_exit(CRM_EX_DISCONNECT); + } st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, st_callback); st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, st_callback); st->cmds->register_notification(st, STONITH_OP_DEVICE_ADD, st_callback); st->cmds->register_notification(st, STONITH_OP_DEVICE_DEL, st_callback); st->cmds->register_callback(st, 0, 120, st_opt_timeout_updates, NULL, "st_global_callback", st_global_callback); crm_info("Starting API Sanity Tests"); run_standard_test(); run_fence_failure_test(); run_fence_failure_rollover_test(); crm_info("Sanity Tests Passed"); } static void standard_dev_test(void) { int rc = 0; char *tmp = NULL; stonith_key_value_t *params = NULL; rc = st->cmds->connect(st, crm_system_name, &pollfd.fd); - crm_debug("Connect: %d", rc); + if (rc != pcmk_ok) { + stonith_api_delete(st); + crm_exit(CRM_EX_DISCONNECT); + } params = stonith_key_value_add(params, "pcmk_host_map", "some-host=pcmk-7 true_1_node1=3,4"); rc = st->cmds->register_device(st, st_opts, "test-id", "stonith-ng", "fence_xvm", params); crm_debug("Register: %d", rc); rc = st->cmds->list(st, st_opts, "test-id", &tmp, 10); crm_debug("List: %d output: %s", rc, tmp ? tmp : ""); rc = st->cmds->monitor(st, st_opts, "test-id", 10); crm_debug("Monitor: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "false_1_node2", 10); crm_debug("Status false_1_node2: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "false_1_node1", 10); crm_debug("Status false_1_node1: %d", rc); rc = st->cmds->fence(st, st_opts, "unknown-host", "off", 60, 0); crm_debug("Fence unknown-host: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "false_1_node1", 10); crm_debug("Status false_1_node1: %d", rc); rc = st->cmds->fence(st, st_opts, "false_1_node1", "off", 60, 0); crm_debug("Fence false_1_node1: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "false_1_node1", 10); crm_debug("Status false_1_node1: %d", rc); rc = st->cmds->fence(st, st_opts, "false_1_node1", "on", 10, 0); crm_debug("Unfence false_1_node1: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "false_1_node1", 10); crm_debug("Status false_1_node1: %d", rc); rc = st->cmds->fence(st, st_opts, "some-host", "off", 10, 0); crm_debug("Fence alias: %d", rc); rc = st->cmds->status(st, st_opts, "test-id", "some-host", 10); crm_debug("Status alias: %d", rc); rc = st->cmds->fence(st, st_opts, "false_1_node1", "on", 10, 0); crm_debug("Unfence false_1_node1: %d", rc); rc = st->cmds->remove_device(st, st_opts, "test-id"); crm_debug("Remove test-id: %d", rc); stonith_key_value_freeall(params, 1, 1); } static void iterate_mainloop_tests(gboolean event_ready); static void mainloop_callback(stonith_t * stonith, stonith_callback_data_t * data) { callback_rc = data->rc; iterate_mainloop_tests(TRUE); } static int register_callback_helper(int callid) { return st->cmds->register_callback(st, callid, MAINLOOP_DEFAULT_TIMEOUT, st_opt_timeout_updates, NULL, "callback", mainloop_callback); } static void test_async_fence_pass(int check_event) { int rc = 0; if (check_event) { if (callback_rc != 0) { mainloop_test_done(FALSE); } else { mainloop_test_done(TRUE); } return; } rc = st->cmds->fence(st, 0, "true_1_node1", "off", MAINLOOP_DEFAULT_TIMEOUT, 0); if (rc < 0) { crm_err("fence failed with rc %d", rc); mainloop_test_done(FALSE); } register_callback_helper(rc); /* wait for event */ } #define CUSTOM_TIMEOUT_ADDITION 10 static void test_async_fence_custom_timeout(int check_event) { int rc = 0; static time_t begin = 0; if (check_event) { uint32_t diff = (time(NULL) - begin); if (callback_rc != -ETIME) { mainloop_test_done(FALSE); } else if (diff < CUSTOM_TIMEOUT_ADDITION + MAINLOOP_DEFAULT_TIMEOUT) { crm_err ("Custom timeout test failed, callback expiration should be updated to %d, actual timeout was %d", CUSTOM_TIMEOUT_ADDITION + MAINLOOP_DEFAULT_TIMEOUT, diff); mainloop_test_done(FALSE); } else { mainloop_test_done(TRUE); } return; } begin = time(NULL); rc = st->cmds->fence(st, 0, "custom_timeout_node1", "off", MAINLOOP_DEFAULT_TIMEOUT, 0); if (rc < 0) { crm_err("fence failed with rc %d", rc); mainloop_test_done(FALSE); } register_callback_helper(rc); /* wait for event */ } static void test_async_fence_timeout(int check_event) { int rc = 0; if (check_event) { if (callback_rc != -ENODEV) { mainloop_test_done(FALSE); } else { mainloop_test_done(TRUE); } return; } rc = st->cmds->fence(st, 0, "false_1_node2", "off", MAINLOOP_DEFAULT_TIMEOUT, 0); if (rc < 0) { crm_err("fence failed with rc %d", rc); mainloop_test_done(FALSE); } register_callback_helper(rc); /* wait for event */ } static void test_async_monitor(int check_event) { int rc = 0; if (check_event) { if (callback_rc) { mainloop_test_done(FALSE); } else { mainloop_test_done(TRUE); } return; } rc = st->cmds->monitor(st, 0, "false_1", MAINLOOP_DEFAULT_TIMEOUT); if (rc < 0) { crm_err("monitor failed with rc %d", rc); mainloop_test_done(FALSE); } register_callback_helper(rc); /* wait for event */ } static void test_register_async_devices(int check_event) { char buf[16] = { 0, }; stonith_key_value_t *params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "false_1_node1=1,2"); params = stonith_key_value_add(params, "mode", "fail"); st->cmds->register_device(st, st_opts, "false_1", "stonith-ng", "fence_dummy", params); stonith_key_value_freeall(params, 1, 1); params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "true_1_node1=1,2"); params = stonith_key_value_add(params, "mode", "pass"); st->cmds->register_device(st, st_opts, "true_1", "stonith-ng", "fence_dummy", params); stonith_key_value_freeall(params, 1, 1); params = NULL; params = stonith_key_value_add(params, "pcmk_host_map", "custom_timeout_node1=1,2"); params = stonith_key_value_add(params, "mode", "fail"); params = stonith_key_value_add(params, "delay", "1000"); snprintf(buf, sizeof(buf) - 1, "%d", MAINLOOP_DEFAULT_TIMEOUT + CUSTOM_TIMEOUT_ADDITION); params = stonith_key_value_add(params, "pcmk_off_timeout", buf); st->cmds->register_device(st, st_opts, "false_custom_timeout", "stonith-ng", "fence_dummy", params); stonith_key_value_freeall(params, 1, 1); mainloop_test_done(TRUE); } static void try_mainloop_connect(int check_event) { - int tries = 10; - int i = 0; - int rc = 0; + int rc = stonith_api_connect_retry(st, crm_system_name, 10); - for (i = 0; i < tries; i++) { - rc = st->cmds->connect(st, crm_system_name, NULL); - - if (!rc) { - crm_info("stonith client connection established"); - mainloop_test_done(TRUE); - return; - } else { - crm_info("stonith client connection failed"); - } - sleep(1); + if (rc == pcmk_ok) { + mainloop_test_done(TRUE); + return; } - crm_err("API CONNECTION FAILURE"); mainloop_test_done(FALSE); } static void iterate_mainloop_tests(gboolean event_ready) { static mainloop_test_iteration_cb callbacks[] = { try_mainloop_connect, test_register_async_devices, test_async_monitor, test_async_fence_pass, test_async_fence_timeout, test_async_fence_custom_timeout, }; if (mainloop_iter == (sizeof(callbacks) / sizeof(mainloop_test_iteration_cb))) { /* all tests ran, everything passed */ crm_info("ALL MAINLOOP TESTS PASSED!"); crm_exit(CRM_EX_OK); } callbacks[mainloop_iter] (event_ready); } static gboolean trigger_iterate_mainloop_tests(gpointer user_data) { iterate_mainloop_tests(FALSE); return TRUE; } static void test_shutdown(int nsig) { int rc = 0; if (st) { rc = st->cmds->disconnect(st); crm_info("Disconnect: %d", rc); crm_debug("Destroy"); stonith_api_delete(st); } if (rc) { crm_exit(CRM_EX_ERROR); } } static void mainloop_tests(void) { trig = mainloop_add_trigger(G_PRIORITY_HIGH, trigger_iterate_mainloop_tests, NULL); mainloop_set_trigger(trig); mainloop_add_signal(SIGTERM, test_shutdown); crm_info("Starting"); mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); } int main(int argc, char **argv) { int argerr = 0; int flag; int option_index = 0; enum test_modes mode = test_standard; crm_log_cli_init("cts-fence-helper"); crm_set_options(NULL, "mode [options]", long_options, "Provides a summary of cluster's current state." "\n\nOutputs varying levels of detail in a number of different formats.\n"); while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) { break; } switch (flag) { case 'V': verbose = 1; break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; case 'p': mode = test_passive; break; case 't': mode = test_api_sanity; break; case 'm': mode = test_api_mainloop; break; default: ++argerr; break; } } crm_log_init(NULL, LOG_INFO, TRUE, (verbose? TRUE : FALSE), argc, argv, FALSE); if (optind > argc) { ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } crm_debug("Create"); st = stonith_api_new(); switch (mode) { case test_standard: standard_dev_test(); break; case test_passive: passive_test(); break; case test_api_sanity: sanity_tests(); break; case test_api_mainloop: mainloop_tests(); break; } test_shutdown(0); return CRM_EX_OK; } diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index eba83cc6d6..1be2508598 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,2771 +1,2771 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if SUPPORT_CIBSECRETS # include #endif #include GHashTable *device_list = NULL; GHashTable *topology = NULL; GList *cmd_list = NULL; struct device_search_s { /* target of fence action */ char *host; /* requested fence action */ char *action; /* timeout to use if a device is queried dynamically for possible targets */ int per_device_timeout; /* number of registered fencing devices at time of request */ int replies_needed; /* number of device replies received so far */ int replies_received; /* whether the target is eligible to perform requested action (or off) */ bool allow_suicide; /* private data to pass to search callback function */ void *user_data; /* function to call when all replies have been received */ void (*callback) (GList * devices, void *user_data); /* devices capable of performing requested action (or off if remapping) */ GListPtr capable; }; static gboolean stonith_device_dispatch(gpointer user_data); static void st_child_done(GPid pid, int rc, const char *output, gpointer user_data); static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer, const char *client_id); static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence); typedef struct async_command_s { int id; int pid; int fd_stdout; int options; int default_timeout; /* seconds */ int timeout; /* seconds */ int start_delay; /* milliseconds */ int delay_id; char *op; char *origin; char *client; char *client_name; char *remote_op_id; char *victim; uint32_t victim_nodeid; char *action; char *device; char *mode; GListPtr device_list; GListPtr device_next; void *internal_user_data; void (*done_cb) (GPid pid, int rc, const char *output, gpointer user_data); guint timer_sigterm; guint timer_sigkill; /*! If the operation timed out, this is the last signal * we sent to the process to get it to terminate */ int last_timeout_signo; stonith_device_t *active_on; stonith_device_t *activating_on; } async_command_t; static xmlNode *stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc); static gboolean is_action_required(const char *action, stonith_device_t *device) { return device && device->automatic_unfencing && safe_str_eq(action, "on"); } static int get_action_delay_max(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_max_ms = 0; if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) { return 0; } value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_MAX); if (value) { delay_max_ms = crm_get_msec(value); } return delay_max_ms; } static int get_action_delay_base(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_base_ms = 0; if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) { return 0; } value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_BASE); if (value) { delay_base_ms = crm_get_msec(value); } return delay_base_ms; } /*! * \internal * \brief Override STONITH timeout with pcmk_*_timeout if available * * \param[in] device STONITH device to use * \param[in] action STONITH action name * \param[in] default_timeout Timeout to use if device does not have * a pcmk_*_timeout parameter for action * * \return Value of pcmk_(action)_timeout if available, otherwise default_timeout * \note For consistency, it would be nice if reboot/off/on timeouts could be * set the same way as start/stop/monitor timeouts, i.e. with an * entry in the fencing resource configuration. However that * is insufficient because fencing devices may be registered directly via * the fencer's register_device() API instead of going through the CIB * (e.g. stonith_admin uses it for its -R option, and the executor uses it * to ensure a device is registered when a command is issued). As device * properties, pcmk_*_timeout parameters can be grabbed by the fencer when * the device is registered, whether by CIB change or API call. */ static int get_action_timeout(stonith_device_t * device, const char *action, int default_timeout) { if (action && device && device->params) { char buffer[64] = { 0, }; const char *value = NULL; /* If "reboot" was requested but the device does not support it, * we will remap to "off", so check timeout for "off" instead */ if (safe_str_eq(action, "reboot") && is_not_set(device->flags, st_device_supports_reboot)) { crm_trace("%s doesn't support reboot, using timeout for off instead", device->id); action = "off"; } /* If the device config specified an action-specific timeout, use it */ snprintf(buffer, sizeof(buffer), "pcmk_%s_timeout", action); value = g_hash_table_lookup(device->params, buffer); if (value) { return atoi(value); } } return default_timeout; } static void free_async_command(async_command_t * cmd) { if (!cmd) { return; } if (cmd->delay_id) { g_source_remove(cmd->delay_id); } cmd_list = g_list_remove(cmd_list, cmd); g_list_free_full(cmd->device_list, free); free(cmd->device); free(cmd->action); free(cmd->victim); free(cmd->remote_op_id); free(cmd->client); free(cmd->client_name); free(cmd->origin); free(cmd->mode); free(cmd->op); free(cmd); } static async_command_t * create_async_command(xmlNode * msg) { async_command_t *cmd = NULL; xmlNode *op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR); const char *action = crm_element_value(op, F_STONITH_ACTION); CRM_CHECK(action != NULL, crm_log_xml_warn(msg, "NoAction"); return NULL); crm_log_xml_trace(msg, "Command"); cmd = calloc(1, sizeof(async_command_t)); crm_element_value_int(msg, F_STONITH_CALLID, &(cmd->id)); crm_element_value_int(msg, F_STONITH_CALLOPTS, &(cmd->options)); crm_element_value_int(msg, F_STONITH_TIMEOUT, &(cmd->default_timeout)); cmd->timeout = cmd->default_timeout; cmd->origin = crm_element_value_copy(msg, F_ORIG); cmd->remote_op_id = crm_element_value_copy(msg, F_STONITH_REMOTE_OP_ID); cmd->client = crm_element_value_copy(msg, F_STONITH_CLIENTID); cmd->client_name = crm_element_value_copy(msg, F_STONITH_CLIENTNAME); cmd->op = crm_element_value_copy(msg, F_STONITH_OPERATION); cmd->action = strdup(action); cmd->victim = crm_element_value_copy(op, F_STONITH_TARGET); cmd->mode = crm_element_value_copy(op, F_STONITH_MODE); cmd->device = crm_element_value_copy(op, F_STONITH_DEVICE); CRM_CHECK(cmd->op != NULL, crm_log_xml_warn(msg, "NoOp"); free_async_command(cmd); return NULL); CRM_CHECK(cmd->client != NULL, crm_log_xml_warn(msg, "NoClient")); cmd->done_cb = st_child_done; cmd_list = g_list_append(cmd_list, cmd); return cmd; } static int get_action_limit(stonith_device_t * device) { const char *value = NULL; int action_limit = 1; value = g_hash_table_lookup(device->params, STONITH_ATTR_ACTION_LIMIT); if (value) { action_limit = crm_parse_int(value, "1"); if (action_limit == 0) { /* pcmk_action_limit should not be 0. Enforce it to be 1. */ action_limit = 1; } } return action_limit; } static int get_active_cmds(stonith_device_t * device) { int counter = 0; GListPtr gIter = NULL; GListPtr gIterNext = NULL; CRM_CHECK(device != NULL, return 0); for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) { async_command_t *cmd = gIter->data; gIterNext = gIter->next; if (cmd->active_on == device) { counter++; } } return counter; } static void fork_cb(GPid pid, gpointer user_data) { async_command_t *cmd = (async_command_t *) user_data; stonith_device_t * device = /* in case of a retry we've done the move from activating_on to active_on already */ cmd->activating_on?cmd->activating_on:cmd->active_on; CRM_ASSERT(device); crm_debug("Operation %s%s%s on %s now running with pid=%d, timeout=%ds", cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", device->id, pid, cmd->timeout); cmd->active_on = device; cmd->activating_on = NULL; } static gboolean stonith_device_execute(stonith_device_t * device) { int exec_rc = 0; const char *action_str = NULL; async_command_t *cmd = NULL; stonith_action_t *action = NULL; int active_cmds = 0; int action_limit = 0; CRM_CHECK(device != NULL, return FALSE); active_cmds = get_active_cmds(device); action_limit = get_action_limit(device); if (action_limit > -1 && active_cmds >= action_limit) { crm_trace("%s is over its action limit of %d (%u active action%s)", device->id, action_limit, active_cmds, active_cmds > 1 ? "s" : ""); return TRUE; } if (device->pending_ops) { GList *first = device->pending_ops; cmd = first->data; if (cmd && cmd->delay_id) { crm_trace ("Operation %s%s%s on %s was asked to run too early, waiting for start_delay timeout of %dms", cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", device->id, cmd->start_delay); return TRUE; } device->pending_ops = g_list_remove_link(device->pending_ops, first); g_list_free_1(first); } if (cmd == NULL) { crm_trace("Nothing further to do for %s", device->id); return TRUE; } if(safe_str_eq(device->agent, STONITH_WATCHDOG_AGENT)) { if(safe_str_eq(cmd->action, "reboot")) { pcmk_panic(__FUNCTION__); return TRUE; } else if(safe_str_eq(cmd->action, "off")) { pcmk_panic(__FUNCTION__); return TRUE; } else { crm_info("Faking success for %s watchdog operation", cmd->action); cmd->done_cb(0, 0, NULL, cmd); return TRUE; } } #if SUPPORT_CIBSECRETS if (replace_secret_params(device->id, device->params) < 0) { /* replacing secrets failed! */ if (safe_str_eq(cmd->action,"stop")) { /* don't fail on stop! */ crm_info("proceeding with the stop operation for %s", device->id); } else { crm_err("failed to get secrets for %s, " "considering resource not configured", device->id); exec_rc = PCMK_OCF_NOT_CONFIGURED; cmd->done_cb(0, exec_rc, NULL, cmd); return TRUE; } } #endif action_str = cmd->action; if (safe_str_eq(cmd->action, "reboot") && is_not_set(device->flags, st_device_supports_reboot)) { crm_warn("Agent '%s' does not advertise support for 'reboot', performing 'off' action instead", device->agent); action_str = "off"; } action = stonith_action_create(device->agent, action_str, cmd->victim, cmd->victim_nodeid, cmd->timeout, device->params, device->aliases); /* for async exec, exec_rc is negative for early error exit otherwise handling of success/errors is done via callbacks */ cmd->activating_on = device; exec_rc = stonith_action_execute_async(action, (void *)cmd, cmd->done_cb, fork_cb); if (exec_rc < 0) { crm_warn("Operation %s%s%s on %s failed: %s (%d)", cmd->action, cmd->victim ? " for node " : "", cmd->victim ? cmd->victim : "", device->id, pcmk_strerror(exec_rc), exec_rc); cmd->activating_on = NULL; cmd->done_cb(0, exec_rc, NULL, cmd); } return TRUE; } static gboolean stonith_device_dispatch(gpointer user_data) { return stonith_device_execute(user_data); } static gboolean start_delay_helper(gpointer data) { async_command_t *cmd = data; stonith_device_t *device = NULL; cmd->delay_id = 0; device = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; if (device) { mainloop_set_trigger(device->work); } return FALSE; } static void schedule_stonith_command(async_command_t * cmd, stonith_device_t * device) { int delay_max = 0; int delay_base = 0; CRM_CHECK(cmd != NULL, return); CRM_CHECK(device != NULL, return); if (cmd->device) { free(cmd->device); } if (device->include_nodeid && cmd->victim) { crm_node_t *node = crm_get_peer(0, cmd->victim); cmd->victim_nodeid = node->id; } cmd->device = strdup(device->id); cmd->timeout = get_action_timeout(device, cmd->action, cmd->default_timeout); if (cmd->remote_op_id) { crm_debug("Scheduling %s on %s for remote peer %s with op id (%s) (timeout=%ds)", cmd->action, device->id, cmd->origin, cmd->remote_op_id, cmd->timeout); } else { crm_debug("Scheduling %s on %s for %s (timeout=%ds)", cmd->action, device->id, cmd->client, cmd->timeout); } device->pending_ops = g_list_append(device->pending_ops, cmd); mainloop_set_trigger(device->work); delay_max = get_action_delay_max(device, cmd->action); delay_base = get_action_delay_base(device, cmd->action); if (delay_max == 0) { delay_max = delay_base; } if (delay_max < delay_base) { crm_warn("Base-delay (%dms) is larger than max-delay (%dms) " "for %s on %s - limiting to max-delay", delay_base, delay_max, cmd->action, device->id); delay_base = delay_max; } if (delay_max > 0) { // coverity[dont_call] We're not using rand() for security cmd->start_delay = ((delay_max != delay_base)?(rand() % (delay_max - delay_base)):0) + delay_base; crm_notice("Delaying %s on %s for %dms (timeout=%ds, base=%dms, " "max=%dms)", cmd->action, device->id, cmd->start_delay, cmd->timeout, delay_base, delay_max); cmd->delay_id = g_timeout_add(cmd->start_delay, start_delay_helper, cmd); } } static void free_device(gpointer data) { GListPtr gIter = NULL; stonith_device_t *device = data; g_hash_table_destroy(device->params); g_hash_table_destroy(device->aliases); for (gIter = device->pending_ops; gIter != NULL; gIter = gIter->next) { async_command_t *cmd = gIter->data; crm_warn("Removal of device '%s' purged operation %s", device->id, cmd->action); cmd->done_cb(0, -ENODEV, NULL, cmd); } g_list_free(device->pending_ops); g_list_free_full(device->targets, free); mainloop_destroy_trigger(device->work); free_xml(device->agent_metadata); free(device->namespace); free(device->on_target_actions); free(device->agent); free(device->id); free(device); } void free_device_list() { if (device_list != NULL) { g_hash_table_destroy(device_list); device_list = NULL; } } void init_device_list() { if (device_list == NULL) { device_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_device); } } static GHashTable * build_port_aliases(const char *hostmap, GListPtr * targets) { char *name = NULL; int last = 0, lpc = 0, max = 0, added = 0; GHashTable *aliases = crm_strcase_table_new(); if (hostmap == NULL) { return aliases; } max = strlen(hostmap); for (; lpc <= max; lpc++) { switch (hostmap[lpc]) { /* Assignment chars */ case '=': case ':': if (lpc > last) { free(name); name = calloc(1, 1 + lpc - last); memcpy(name, hostmap + last, lpc - last); } last = lpc + 1; break; /* Delimeter chars */ /* case ',': Potentially used to specify multiple ports */ case 0: case ';': case ' ': case '\t': if (name) { char *value = NULL; value = calloc(1, 1 + lpc - last); memcpy(value, hostmap + last, lpc - last); crm_debug("Adding alias '%s'='%s'", name, value); g_hash_table_replace(aliases, name, value); if (targets) { *targets = g_list_append(*targets, strdup(value)); } value = NULL; name = NULL; added++; } else if (lpc > last) { crm_debug("Parse error at offset %d near '%s'", lpc - last, hostmap + last); } last = lpc + 1; break; } if (hostmap[lpc] == 0) { break; } } if (added == 0) { crm_info("No host mappings detected in '%s'", hostmap); } free(name); return aliases; } static void parse_host_line(const char *line, int max, GListPtr * output) { int lpc = 0; int last = 0; if (max <= 0) { return; } /* Check for any complaints about additional parameters that the device doesn't understand */ if (strstr(line, "invalid") || strstr(line, "variable")) { crm_debug("Skipping: %s", line); return; } crm_trace("Processing %d bytes: [%s]", max, line); /* Skip initial whitespace */ for (lpc = 0; lpc <= max && isspace(line[lpc]); lpc++) { last = lpc + 1; } /* Now the actual content */ for (lpc = 0; lpc <= max; lpc++) { gboolean a_space = isspace(line[lpc]); if (a_space && lpc < max && isspace(line[lpc + 1])) { /* fast-forward to the end of the spaces */ } else if (a_space || line[lpc] == ',' || line[lpc] == ';' || line[lpc] == 0) { int rc = 1; char *entry = NULL; if (lpc != last) { entry = calloc(1, 1 + lpc - last); rc = sscanf(line + last, "%[a-zA-Z0-9_-.]", entry); } if (entry == NULL) { /* Skip */ } else if (rc != 1) { crm_warn("Could not parse (%d %d): %s", last, lpc, line + last); } else if (safe_str_neq(entry, "on") && safe_str_neq(entry, "off")) { crm_trace("Adding '%s'", entry); *output = g_list_append(*output, entry); entry = NULL; } free(entry); last = lpc + 1; } } } static GListPtr parse_host_list(const char *hosts) { int lpc = 0; int max = 0; int last = 0; GListPtr output = NULL; if (hosts == NULL) { return output; } max = strlen(hosts); for (lpc = 0; lpc <= max; lpc++) { if (hosts[lpc] == '\n' || hosts[lpc] == 0) { int len = lpc - last; if(len > 1) { char *line = strndup(hosts + last, len); line[len] = 0; /* Because it might be '\n' */ parse_host_line(line, len, &output); free(line); } last = lpc + 1; } } crm_trace("Parsed %d entries from '%s'", g_list_length(output), hosts); return output; } GHashTable *metadata_cache = NULL; void free_metadata_cache() { if (metadata_cache != NULL) { g_hash_table_destroy(metadata_cache); metadata_cache = NULL; } } static void init_metadata_cache() { if (metadata_cache == NULL) { metadata_cache = crm_str_table_new(); } } static xmlNode * get_agent_metadata(const char *agent) { xmlNode *xml = NULL; char *buffer = NULL; init_metadata_cache(); buffer = g_hash_table_lookup(metadata_cache, agent); if(safe_str_eq(agent, STONITH_WATCHDOG_AGENT)) { return NULL; } else if(buffer == NULL) { stonith_t *st = stonith_api_new(); int rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, 10); stonith_api_delete(st); if (rc || !buffer) { crm_err("Could not retrieve metadata for fencing agent %s", agent); return NULL; } g_hash_table_replace(metadata_cache, strdup(agent), buffer); } xml = string2xml(buffer); return xml; } static gboolean is_nodeid_required(xmlNode * xml) { xmlXPathObjectPtr xpath = NULL; if (stand_alone) { return FALSE; } if (!xml) { return FALSE; } xpath = xpath_search(xml, "//parameter[@name='nodeid']"); if (numXpathResults(xpath) <= 0) { freeXpathObject(xpath); return FALSE; } freeXpathObject(xpath); return TRUE; } #define MAX_ACTION_LEN 256 static char * add_action(char *actions, const char *action) { int offset = 0; if (actions == NULL) { actions = calloc(1, MAX_ACTION_LEN); } else { offset = strlen(actions); } if (offset > 0) { offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, " "); } offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, "%s", action); return actions; } static void read_action_metadata(stonith_device_t *device) { xmlXPathObjectPtr xpath = NULL; int max = 0; int lpc = 0; if (device->agent_metadata == NULL) { return; } xpath = xpath_search(device->agent_metadata, "//action"); max = numXpathResults(xpath); if (max <= 0) { freeXpathObject(xpath); return; } for (lpc = 0; lpc < max; lpc++) { const char *on_target = NULL; const char *action = NULL; xmlNode *match = getXpathResult(xpath, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; on_target = crm_element_value(match, "on_target"); action = crm_element_value(match, "name"); if(safe_str_eq(action, "list")) { set_bit(device->flags, st_device_supports_list); } else if(safe_str_eq(action, "status")) { set_bit(device->flags, st_device_supports_status); } else if(safe_str_eq(action, "reboot")) { set_bit(device->flags, st_device_supports_reboot); } else if (safe_str_eq(action, "on")) { /* "automatic" means the cluster will unfence node when it joins */ const char *automatic = crm_element_value(match, "automatic"); /* "required" is a deprecated synonym for "automatic" */ const char *required = crm_element_value(match, "required"); if (crm_is_true(automatic) || crm_is_true(required)) { device->automatic_unfencing = TRUE; } } if (action && crm_is_true(on_target)) { device->on_target_actions = add_action(device->on_target_actions, action); } } freeXpathObject(xpath); } /*! * \internal * \brief Set a pcmk_*_action parameter if not already set * * \param[in,out] params Device parameters * \param[in] action Name of action * \param[in] value Value to use if action is not already set */ static void map_action(GHashTable *params, const char *action, const char *value) { char *key = crm_strdup_printf("pcmk_%s_action", action); if (g_hash_table_lookup(params, key)) { crm_warn("Ignoring %s='%s', see %s instead", STONITH_ATTR_ACTION_OP, value, key); free(key); } else { crm_warn("Mapping %s='%s' to %s='%s'", STONITH_ATTR_ACTION_OP, value, key, value); g_hash_table_insert(params, key, strdup(value)); } } /*! * \internal * \brief Create device parameter table from XML * * \param[in] name Device name (used for logging only) * \param[in,out] params Device parameters */ static GHashTable * xml2device_params(const char *name, xmlNode *dev) { GHashTable *params = xml2list(dev); const char *value; /* Action should never be specified in the device configuration, * but we support it for users who are familiar with other software * that worked that way. */ value = g_hash_table_lookup(params, STONITH_ATTR_ACTION_OP); if (value != NULL) { crm_warn("%s has '%s' parameter, which should never be specified in configuration", name, STONITH_ATTR_ACTION_OP); if (*value == '\0') { crm_warn("Ignoring empty '%s' parameter", STONITH_ATTR_ACTION_OP); } else if (strcmp(value, "reboot") == 0) { crm_warn("Ignoring %s='reboot' (see stonith-action cluster property instead)", STONITH_ATTR_ACTION_OP); } else if (strcmp(value, "off") == 0) { map_action(params, "reboot", value); } else { map_action(params, "off", value); map_action(params, "reboot", value); } g_hash_table_remove(params, STONITH_ATTR_ACTION_OP); } return params; } static stonith_device_t * build_device_from_xml(xmlNode * msg) { const char *value; xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR); stonith_device_t *device = NULL; char *agent = crm_element_value_copy(dev, "agent"); CRM_CHECK(agent != NULL, return device); device = calloc(1, sizeof(stonith_device_t)); CRM_CHECK(device != NULL, {free(agent); return device;}); device->id = crm_element_value_copy(dev, XML_ATTR_ID); device->agent = agent; device->namespace = crm_element_value_copy(dev, "namespace"); device->params = xml2device_params(device->id, dev); value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTLIST); if (value) { device->targets = parse_host_list(value); } value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTMAP); device->aliases = build_port_aliases(value, &(device->targets)); device->agent_metadata = get_agent_metadata(device->agent); read_action_metadata(device); value = g_hash_table_lookup(device->params, "nodeid"); if (!value) { device->include_nodeid = is_nodeid_required(device->agent_metadata); } value = crm_element_value(dev, "rsc_provides"); if (safe_str_eq(value, "unfencing")) { device->automatic_unfencing = TRUE; } if (is_action_required("on", device)) { crm_info("The fencing device '%s' requires unfencing", device->id); } if (device->on_target_actions) { crm_info("The fencing device '%s' requires actions (%s) to be executed on the target node", device->id, device->on_target_actions); } device->work = mainloop_add_trigger(G_PRIORITY_HIGH, stonith_device_dispatch, device); /* TODO: Hook up priority */ return device; } static const char * target_list_type(stonith_device_t * dev) { const char *check_type = NULL; check_type = g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTCHECK); if (check_type == NULL) { if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTLIST)) { check_type = "static-list"; } else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP)) { check_type = "static-list"; } else if(is_set(dev->flags, st_device_supports_list)){ check_type = "dynamic-list"; } else if(is_set(dev->flags, st_device_supports_status)){ check_type = "status"; } else { check_type = "none"; } } return check_type; } -void +static void schedule_internal_command(const char *origin, stonith_device_t * device, const char *action, const char *victim, int timeout, void *internal_user_data, void (*done_cb) (GPid pid, int rc, const char *output, gpointer user_data)) { async_command_t *cmd = NULL; cmd = calloc(1, sizeof(async_command_t)); cmd->id = -1; cmd->default_timeout = timeout ? timeout : 60; cmd->timeout = cmd->default_timeout; cmd->action = strdup(action); cmd->victim = victim ? strdup(victim) : NULL; cmd->device = strdup(device->id); cmd->origin = strdup(origin); cmd->client = strdup(crm_system_name); cmd->client_name = strdup(crm_system_name); cmd->internal_user_data = internal_user_data; cmd->done_cb = done_cb; /* cmd, not internal_user_data, is passed to 'done_cb' as the userdata */ schedule_stonith_command(cmd, device); } gboolean string_in_list(GListPtr list, const char *item) { int lpc = 0; int max = g_list_length(list); for (lpc = 0; lpc < max; lpc++) { const char *value = g_list_nth_data(list, lpc); if (safe_str_eq(item, value)) { return TRUE; } else { crm_trace("%d: '%s' != '%s'", lpc, item, value); } } return FALSE; } static void status_search_cb(GPid pid, int rc, const char *output, gpointer user_data) { async_command_t *cmd = user_data; struct device_search_s *search = cmd->internal_user_data; stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; gboolean can = FALSE; free_async_command(cmd); if (!dev) { search_devices_record_result(search, NULL, FALSE); return; } mainloop_set_trigger(dev->work); if (rc == 1 /* unknown */ ) { crm_trace("Host %s is not known by %s", search->host, dev->id); } else if (rc == 0 /* active */ || rc == 2 /* inactive */ ) { crm_trace("Host %s is known by %s", search->host, dev->id); can = TRUE; } else { crm_notice("Unknown result when testing if %s can fence %s: rc=%d", dev->id, search->host, rc); } search_devices_record_result(search, dev->id, can); } static void dynamic_list_search_cb(GPid pid, int rc, const char *output, gpointer user_data) { async_command_t *cmd = user_data; struct device_search_s *search = cmd->internal_user_data; stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL; gboolean can_fence = FALSE; free_async_command(cmd); /* Host/alias must be in the list output to be eligible to be fenced * * Will cause problems if down'd nodes aren't listed or (for virtual nodes) * if the guest is still listed despite being moved to another machine */ if (!dev) { search_devices_record_result(search, NULL, FALSE); return; } mainloop_set_trigger(dev->work); /* If we successfully got the targets earlier, don't disable. */ if (rc != 0 && !dev->targets) { crm_notice("Disabling port list queries for %s (%d): %s", dev->id, rc, output); /* Fall back to status */ g_hash_table_replace(dev->params, strdup(STONITH_ATTR_HOSTCHECK), strdup("status")); g_list_free_full(dev->targets, free); dev->targets = NULL; } else if (!rc) { crm_info("Refreshing port list for %s", dev->id); g_list_free_full(dev->targets, free); dev->targets = parse_host_list(output); dev->targets_age = time(NULL); } if (dev->targets) { const char *alias = g_hash_table_lookup(dev->aliases, search->host); if (!alias) { alias = search->host; } if (string_in_list(dev->targets, alias)) { can_fence = TRUE; } } search_devices_record_result(search, dev->id, can_fence); } /*! * \internal * \brief Returns true if any key in first is not in second or second has a different value for key */ static int device_params_diff(GHashTable *first, GHashTable *second) { char *key = NULL; char *value = NULL; GHashTableIter gIter; g_hash_table_iter_init(&gIter, first); while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&value)) { if(strstr(key, "CRM_meta") == key) { continue; } else if(strcmp(key, "crm_feature_set") == 0) { continue; } else { char *other_value = g_hash_table_lookup(second, key); if (!other_value || safe_str_neq(other_value, value)) { crm_trace("Different value for %s: %s != %s", key, other_value, value); return 1; } } } return 0; } /*! * \internal * \brief Checks to see if an identical device already exists in the device_list */ static stonith_device_t * device_has_duplicate(stonith_device_t * device) { stonith_device_t *dup = g_hash_table_lookup(device_list, device->id); if (!dup) { crm_trace("No match for %s", device->id); return NULL; } else if (safe_str_neq(dup->agent, device->agent)) { crm_trace("Different agent: %s != %s", dup->agent, device->agent); return NULL; } /* Use calculate_operation_digest() here? */ if (device_params_diff(device->params, dup->params) || device_params_diff(dup->params, device->params)) { return NULL; } crm_trace("Match"); return dup; } int stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib) { stonith_device_t *dup = NULL; stonith_device_t *device = build_device_from_xml(msg); CRM_CHECK(device != NULL, return -ENOMEM); dup = device_has_duplicate(device); if (dup) { crm_debug("Device '%s' already existed in device list (%d active devices)", device->id, g_hash_table_size(device_list)); free_device(device); device = dup; } else { stonith_device_t *old = g_hash_table_lookup(device_list, device->id); if (from_cib && old && old->api_registered) { /* If the cib is writing over an entry that is shared with a stonith client, * copy any pending ops that currently exist on the old entry to the new one. * Otherwise the pending ops will be reported as failures */ crm_info("Overwriting an existing entry for %s from the cib", device->id); device->pending_ops = old->pending_ops; device->api_registered = TRUE; old->pending_ops = NULL; if (device->pending_ops) { mainloop_set_trigger(device->work); } } g_hash_table_replace(device_list, device->id, device); crm_notice("Added '%s' to the device list (%d active devices)", device->id, g_hash_table_size(device_list)); } if (desc) { *desc = device->id; } if (from_cib) { device->cib_registered = TRUE; } else { device->api_registered = TRUE; } return pcmk_ok; } int stonith_device_remove(const char *id, gboolean from_cib) { stonith_device_t *device = g_hash_table_lookup(device_list, id); if (!device) { crm_info("Device '%s' not found (%d active devices)", id, g_hash_table_size(device_list)); return pcmk_ok; } if (from_cib) { device->cib_registered = FALSE; } else { device->verified = FALSE; device->api_registered = FALSE; } if (!device->cib_registered && !device->api_registered) { g_hash_table_remove(device_list, id); crm_info("Removed '%s' from the device list (%d active devices)", id, g_hash_table_size(device_list)); } else { crm_trace("Not removing '%s' from the device list (%d active devices) " "- still %s%s_registered", id, g_hash_table_size(device_list), device->cib_registered?"cib":"", device->api_registered?"api":""); } return pcmk_ok; } /*! * \internal * \brief Return the number of stonith levels registered for a node * * \param[in] tp Node's topology table entry * * \return Number of non-NULL levels in topology entry * \note This function is used only for log messages. */ static int count_active_levels(stonith_topology_t * tp) { int lpc = 0; int count = 0; for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) { if (tp->levels[lpc] != NULL) { count++; } } return count; } static void free_topology_entry(gpointer data) { stonith_topology_t *tp = data; int lpc = 0; for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) { if (tp->levels[lpc] != NULL) { g_list_free_full(tp->levels[lpc], free); } } free(tp->target); free(tp->target_value); free(tp->target_pattern); free(tp->target_attribute); free(tp); } void free_topology_list() { if (topology != NULL) { g_hash_table_destroy(topology); topology = NULL; } } void init_topology_list() { if (topology == NULL) { topology = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_topology_entry); } } char *stonith_level_key(xmlNode *level, int mode) { if(mode == -1) { mode = stonith_level_kind(level); } switch(mode) { case 0: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET); case 1: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN); case 2: { const char *name = crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE); const char *value = crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE); if(name && value) { return crm_strdup_printf("%s=%s", name, value); } } default: return crm_strdup_printf("Unknown-%d-%s", mode, ID(level)); } } int stonith_level_kind(xmlNode * level) { int mode = 0; const char *target = crm_element_value(level, XML_ATTR_STONITH_TARGET); if(target == NULL) { mode++; target = crm_element_value(level, XML_ATTR_STONITH_TARGET_PATTERN); } if(stand_alone == FALSE && target == NULL) { mode++; if(crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE) == NULL) { mode++; } else if(crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE) == NULL) { mode++; } } return mode; } static stonith_key_value_t * parse_device_list(const char *devices) { int lpc = 0; int max = 0; int last = 0; stonith_key_value_t *output = NULL; if (devices == NULL) { return output; } max = strlen(devices); for (lpc = 0; lpc <= max; lpc++) { if (devices[lpc] == ',' || devices[lpc] == 0) { char *line = strndup(devices + last, lpc - last); output = stonith_key_value_add(output, NULL, line); free(line); last = lpc + 1; } } return output; } /*! * \internal * \brief Register a STONITH level for a target * * Given an XML request specifying the target name, level index, and device IDs * for the level, this will create an entry for the target in the global topology * table if one does not already exist, then append the specified device IDs to * the entry's device list for the specified level. * * \param[in] msg XML request for STONITH level registration * \param[out] desc If not NULL, will be set to string representation ("TARGET[LEVEL]") * * \return pcmk_ok on success, -EINVAL if XML does not specify valid level index */ int stonith_level_register(xmlNode *msg, char **desc) { int id = 0; xmlNode *level; int mode; char *target; stonith_topology_t *tp; stonith_key_value_t *dIter = NULL; stonith_key_value_t *devices = NULL; /* Allow the XML here to point to the level tag directly, or wrapped in * another tag. If directly, don't search by xpath, because it might give * multiple hits (e.g. if the XML is the CIB). */ if (safe_str_eq(TYPE(msg), XML_TAG_FENCING_LEVEL)) { level = msg; } else { level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR); } CRM_CHECK(level != NULL, return -EINVAL); mode = stonith_level_kind(level); target = stonith_level_key(level, mode); crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id); if (desc) { *desc = crm_strdup_printf("%s[%d]", target, id); } /* Sanity-check arguments */ if (mode >= 3 || (id <= 0) || (id >= ST_LEVEL_MAX)) { crm_trace("Could not add %s[%d] (%d) to the topology (%d active entries)", target, id, mode, g_hash_table_size(topology)); free(target); crm_log_xml_err(level, "Bad topology"); return -EINVAL; } /* Find or create topology table entry */ tp = g_hash_table_lookup(topology, target); if (tp == NULL) { tp = calloc(1, sizeof(stonith_topology_t)); tp->kind = mode; tp->target = target; tp->target_value = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_VALUE); tp->target_pattern = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN); tp->target_attribute = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE); g_hash_table_replace(topology, tp->target, tp); crm_trace("Added %s (%d) to the topology (%d active entries)", target, mode, g_hash_table_size(topology)); } else { free(target); } if (tp->levels[id] != NULL) { crm_info("Adding to the existing %s[%d] topology entry", tp->target, id); } devices = parse_device_list(crm_element_value(level, XML_ATTR_STONITH_DEVICES)); for (dIter = devices; dIter; dIter = dIter->next) { const char *device = dIter->value; crm_trace("Adding device '%s' for %s[%d]", device, tp->target, id); tp->levels[id] = g_list_append(tp->levels[id], strdup(device)); } stonith_key_value_freeall(devices, 1, 1); crm_info("Target %s has %d active fencing levels", tp->target, count_active_levels(tp)); return pcmk_ok; } int stonith_level_remove(xmlNode *msg, char **desc) { int id = 0; stonith_topology_t *tp; char *target; /* Unlike additions, removal requests should always have one level tag */ xmlNode *level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR); CRM_CHECK(level != NULL, return -EINVAL); target = stonith_level_key(level, -1); crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id); if (desc) { *desc = crm_strdup_printf("%s[%d]", target, id); } /* Sanity-check arguments */ if (id >= ST_LEVEL_MAX) { free(target); return -EINVAL; } tp = g_hash_table_lookup(topology, target); if (tp == NULL) { crm_info("Topology for %s not found (%d active entries)", target, g_hash_table_size(topology)); } else if (id == 0 && g_hash_table_remove(topology, target)) { crm_info("Removed all %s related entries from the topology (%d active entries)", target, g_hash_table_size(topology)); } else if (id > 0 && tp->levels[id] != NULL) { g_list_free_full(tp->levels[id], free); tp->levels[id] = NULL; crm_info("Removed level '%d' from topology for %s (%d active levels remaining)", id, target, count_active_levels(tp)); } free(target); return pcmk_ok; } static int stonith_device_action(xmlNode * msg, char **output) { int rc = pcmk_ok; xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR); const char *id = crm_element_value(dev, F_STONITH_DEVICE); async_command_t *cmd = NULL; stonith_device_t *device = NULL; if (id) { crm_trace("Looking for '%s'", id); device = g_hash_table_lookup(device_list, id); } if (device && device->api_registered == FALSE) { rc = -ENODEV; } else if (device) { cmd = create_async_command(msg); if (cmd == NULL) { return -EPROTO; } schedule_stonith_command(cmd, device); rc = -EINPROGRESS; } else { crm_info("Device %s not found", id ? id : ""); rc = -ENODEV; } return rc; } static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence) { search->replies_received++; if (can_fence && device) { search->capable = g_list_append(search->capable, strdup(device)); } if (search->replies_needed == search->replies_received) { crm_debug("Finished Search. %d devices can perform action (%s) on node %s", g_list_length(search->capable), search->action ? search->action : "", search->host ? search->host : ""); search->callback(search->capable, search->user_data); free(search->host); free(search->action); free(search); } } /*! * \internal * \brief Check whether the local host is allowed to execute a fencing action * * \param[in] device Fence device to check * \param[in] action Fence action to check * \param[in] target Hostname of fence target * \param[in] allow_suicide Whether self-fencing is allowed for this operation * * \return TRUE if local host is allowed to execute action, FALSE otherwise */ static gboolean localhost_is_eligible(const stonith_device_t *device, const char *action, const char *target, gboolean allow_suicide) { gboolean localhost_is_target = safe_str_eq(target, stonith_our_uname); if (device && action && device->on_target_actions && strstr(device->on_target_actions, action)) { if (!localhost_is_target) { crm_trace("%s operation with %s can only be executed for localhost not %s", action, device->id, target); return FALSE; } } else if (localhost_is_target && !allow_suicide) { crm_trace("%s operation does not support self-fencing", action); return FALSE; } return TRUE; } static void can_fence_host_with_device(stonith_device_t * dev, struct device_search_s *search) { gboolean can = FALSE; const char *check_type = NULL; const char *host = search->host; const char *alias = NULL; CRM_LOG_ASSERT(dev != NULL); if (dev == NULL) { goto search_report_results; } else if (host == NULL) { can = TRUE; goto search_report_results; } /* Short-circuit query if this host is not allowed to perform the action */ if (safe_str_eq(search->action, "reboot")) { /* A "reboot" *might* get remapped to "off" then "on", so short-circuit * only if all three are disallowed. If only one or two are disallowed, * we'll report that with the results. We never allow suicide for * remapped "on" operations because the host is off at that point. */ if (!localhost_is_eligible(dev, "reboot", host, search->allow_suicide) && !localhost_is_eligible(dev, "off", host, search->allow_suicide) && !localhost_is_eligible(dev, "on", host, FALSE)) { goto search_report_results; } } else if (!localhost_is_eligible(dev, search->action, host, search->allow_suicide)) { goto search_report_results; } alias = g_hash_table_lookup(dev->aliases, host); if (alias == NULL) { alias = host; } check_type = target_list_type(dev); if (safe_str_eq(check_type, "none")) { can = TRUE; } else if (safe_str_eq(check_type, "static-list")) { /* Presence in the hostmap is sufficient * Only use if all hosts on which the device can be active can always fence all listed hosts */ if (string_in_list(dev->targets, host)) { can = TRUE; } else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP) && g_hash_table_lookup(dev->aliases, host)) { can = TRUE; } } else if (safe_str_eq(check_type, "dynamic-list")) { time_t now = time(NULL); if (dev->targets == NULL || dev->targets_age + 60 < now) { crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)", check_type, dev->id, search->host, search->action); schedule_internal_command(__FUNCTION__, dev, "list", NULL, search->per_device_timeout, search, dynamic_list_search_cb); /* we'll respond to this search request async in the cb */ return; } if (string_in_list(dev->targets, alias)) { can = TRUE; } } else if (safe_str_eq(check_type, "status")) { crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)", check_type, dev->id, search->host, search->action); schedule_internal_command(__FUNCTION__, dev, "status", search->host, search->per_device_timeout, search, status_search_cb); /* we'll respond to this search request async in the cb */ return; } else { crm_err("Invalid value for " STONITH_ATTR_HOSTCHECK ": %s", check_type); check_type = "Invalid " STONITH_ATTR_HOSTCHECK; } if (safe_str_eq(host, alias)) { crm_notice("%s is%s eligible to fence (%s) %s: %s", dev->id, (can? "" : " not"), search->action, host, check_type); } else { crm_notice("%s is%s eligible to fence (%s) %s (aka. '%s'): %s", dev->id, (can? "" : " not"), search->action, host, alias, check_type); } search_report_results: search_devices_record_result(search, dev ? dev->id : NULL, can); } static void search_devices(gpointer key, gpointer value, gpointer user_data) { stonith_device_t *dev = value; struct device_search_s *search = user_data; can_fence_host_with_device(dev, search); } #define DEFAULT_QUERY_TIMEOUT 20 static void get_capable_devices(const char *host, const char *action, int timeout, bool suicide, void *user_data, void (*callback) (GList * devices, void *user_data)) { struct device_search_s *search; int per_device_timeout = DEFAULT_QUERY_TIMEOUT; int devices_needing_async_query = 0; char *key = NULL; const char *check_type = NULL; GHashTableIter gIter; stonith_device_t *device = NULL; if (!g_hash_table_size(device_list)) { callback(NULL, user_data); return; } search = calloc(1, sizeof(struct device_search_s)); if (!search) { callback(NULL, user_data); return; } g_hash_table_iter_init(&gIter, device_list); while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&device)) { check_type = target_list_type(device); if (safe_str_eq(check_type, "status") || safe_str_eq(check_type, "dynamic-list")) { devices_needing_async_query++; } } /* If we have devices that require an async event in order to know what * nodes they can fence, we have to give the events a timeout. The total * query timeout is divided among those events. */ if (devices_needing_async_query) { per_device_timeout = timeout / devices_needing_async_query; if (!per_device_timeout) { crm_err("STONITH timeout %ds is too low; using %ds, but consider raising to at least %ds", timeout, DEFAULT_QUERY_TIMEOUT, DEFAULT_QUERY_TIMEOUT * devices_needing_async_query); per_device_timeout = DEFAULT_QUERY_TIMEOUT; } else if (per_device_timeout < DEFAULT_QUERY_TIMEOUT) { crm_notice("STONITH timeout %ds is low for the current configuration;" " consider raising to at least %ds", timeout, DEFAULT_QUERY_TIMEOUT * devices_needing_async_query); } } search->host = host ? strdup(host) : NULL; search->action = action ? strdup(action) : NULL; search->per_device_timeout = per_device_timeout; /* We are guaranteed this many replies. Even if a device gets * unregistered some how during the async search, we will get * the correct number of replies. */ search->replies_needed = g_hash_table_size(device_list); search->allow_suicide = suicide; search->callback = callback; search->user_data = user_data; /* kick off the search */ crm_debug("Searching through %d devices to see what is capable of action (%s) for target %s", search->replies_needed, search->action ? search->action : "", search->host ? search->host : ""); g_hash_table_foreach(device_list, search_devices, search); } struct st_query_data { xmlNode *reply; char *remote_peer; char *client_id; char *target; char *action; int call_options; }; /*! * \internal * \brief Add action-specific attributes to query reply XML * * \param[in,out] xml XML to add attributes to * \param[in] action Fence action * \param[in] device Fence device */ static void add_action_specific_attributes(xmlNode *xml, const char *action, stonith_device_t *device) { int action_specific_timeout; int delay_max; int delay_base; CRM_CHECK(xml && action && device, return); if (is_action_required(action, device)) { crm_trace("Action %s is required on %s", action, device->id); crm_xml_add_int(xml, F_STONITH_DEVICE_REQUIRED, 1); } action_specific_timeout = get_action_timeout(device, action, 0); if (action_specific_timeout) { crm_trace("Action %s has timeout %dms on %s", action, action_specific_timeout, device->id); crm_xml_add_int(xml, F_STONITH_ACTION_TIMEOUT, action_specific_timeout); } delay_max = get_action_delay_max(device, action); if (delay_max > 0) { crm_trace("Action %s has maximum random delay %dms on %s", action, delay_max, device->id); crm_xml_add_int(xml, F_STONITH_DELAY_MAX, delay_max / 1000); } delay_base = get_action_delay_base(device, action); if (delay_base > 0) { crm_xml_add_int(xml, F_STONITH_DELAY_BASE, delay_base / 1000); } if ((delay_max > 0) && (delay_base == 0)) { crm_trace("Action %s has maximum random delay %dms on %s", action, delay_max, device->id); } else if ((delay_max == 0) && (delay_base > 0)) { crm_trace("Action %s has a static delay of %dms on %s", action, delay_base, device->id); } else if ((delay_max > 0) && (delay_base > 0)) { crm_trace("Action %s has a minimum delay of %dms and a randomly chosen " "maximum delay of %dms on %s", action, delay_base, delay_max, device->id); } } /*! * \internal * \brief Add "disallowed" attribute to query reply XML if appropriate * * \param[in,out] xml XML to add attribute to * \param[in] action Fence action * \param[in] device Fence device * \param[in] target Fence target * \param[in] allow_suicide Whether self-fencing is allowed */ static void add_disallowed(xmlNode *xml, const char *action, stonith_device_t *device, const char *target, gboolean allow_suicide) { if (!localhost_is_eligible(device, action, target, allow_suicide)) { crm_trace("Action %s on %s is disallowed for local host", action, device->id); crm_xml_add(xml, F_STONITH_ACTION_DISALLOWED, XML_BOOLEAN_TRUE); } } /*! * \internal * \brief Add child element with action-specific values to query reply XML * * \param[in,out] xml XML to add attribute to * \param[in] action Fence action * \param[in] device Fence device * \param[in] target Fence target * \param[in] allow_suicide Whether self-fencing is allowed */ static void add_action_reply(xmlNode *xml, const char *action, stonith_device_t *device, const char *target, gboolean allow_suicide) { xmlNode *child = create_xml_node(xml, F_STONITH_ACTION); crm_xml_add(child, XML_ATTR_ID, action); add_action_specific_attributes(child, action, device); add_disallowed(child, action, device, target, allow_suicide); } static void stonith_query_capable_device_cb(GList * devices, void *user_data) { struct st_query_data *query = user_data; int available_devices = 0; xmlNode *dev = NULL; xmlNode *list = NULL; GListPtr lpc = NULL; /* Pack the results into XML */ list = create_xml_node(NULL, __FUNCTION__); crm_xml_add(list, F_STONITH_TARGET, query->target); for (lpc = devices; lpc != NULL; lpc = lpc->next) { stonith_device_t *device = g_hash_table_lookup(device_list, lpc->data); const char *action = query->action; if (!device) { /* It is possible the device got unregistered while * determining who can fence the target */ continue; } available_devices++; dev = create_xml_node(list, F_STONITH_DEVICE); crm_xml_add(dev, XML_ATTR_ID, device->id); crm_xml_add(dev, "namespace", device->namespace); crm_xml_add(dev, "agent", device->agent); crm_xml_add_int(dev, F_STONITH_DEVICE_VERIFIED, device->verified); /* If the originating fencer wants to reboot the node, and we have a * capable device that doesn't support "reboot", remap to "off" instead. */ if (is_not_set(device->flags, st_device_supports_reboot) && safe_str_eq(query->action, "reboot")) { crm_trace("%s doesn't support reboot, using values for off instead", device->id); action = "off"; } /* Add action-specific values if available */ add_action_specific_attributes(dev, action, device); if (safe_str_eq(query->action, "reboot")) { /* A "reboot" *might* get remapped to "off" then "on", so after * sending the "reboot"-specific values in the main element, we add * sub-elements for "off" and "on" values. * * We short-circuited earlier if "reboot", "off" and "on" are all * disallowed for the local host. However if only one or two are * disallowed, we send back the results and mark which ones are * disallowed. If "reboot" is disallowed, this might cause problems * with older fencer versions, which won't check for it. Older * versions will ignore "off" and "on", so they are not a problem. */ add_disallowed(dev, action, device, query->target, is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "off", device, query->target, is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "on", device, query->target, FALSE); } /* A query without a target wants device parameters */ if (query->target == NULL) { xmlNode *attrs = create_xml_node(dev, XML_TAG_ATTRS); g_hash_table_foreach(device->params, hash2field, attrs); } } crm_xml_add_int(list, F_STONITH_AVAILABLE_DEVICES, available_devices); if (query->target) { crm_debug("Found %d matching devices for '%s'", available_devices, query->target); } else { crm_debug("%d devices installed", available_devices); } if (list != NULL) { crm_log_xml_trace(list, "Add query results"); add_message_xml(query->reply, F_STONITH_CALLDATA, list); } stonith_send_reply(query->reply, query->call_options, query->remote_peer, query->client_id); free_xml(query->reply); free(query->remote_peer); free(query->client_id); free(query->target); free(query->action); free(query); free_xml(list); g_list_free_full(devices, free); } static void stonith_query(xmlNode * msg, const char *remote_peer, const char *client_id, int call_options) { struct st_query_data *query = NULL; const char *action = NULL; const char *target = NULL; int timeout = 0; xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_TRACE); crm_element_value_int(msg, F_STONITH_TIMEOUT, &timeout); if (dev) { const char *device = crm_element_value(dev, F_STONITH_DEVICE); target = crm_element_value(dev, F_STONITH_TARGET); action = crm_element_value(dev, F_STONITH_ACTION); if (device && safe_str_eq(device, "manual_ack")) { /* No query or reply necessary */ return; } } crm_log_xml_debug(msg, "Query"); query = calloc(1, sizeof(struct st_query_data)); query->reply = stonith_construct_reply(msg, NULL, NULL, pcmk_ok); query->remote_peer = remote_peer ? strdup(remote_peer) : NULL; query->client_id = client_id ? strdup(client_id) : NULL; query->target = target ? strdup(target) : NULL; query->action = action ? strdup(action) : NULL; query->call_options = call_options; get_capable_devices(target, action, timeout, is_set(call_options, st_opt_allow_suicide), query, stonith_query_capable_device_cb); } #define ST_LOG_OUTPUT_MAX 512 static void log_operation(async_command_t * cmd, int rc, int pid, const char *next, const char *output) { if (rc == 0) { next = NULL; } if (cmd->victim != NULL) { do_crm_log(rc == 0 ? LOG_NOTICE : LOG_ERR, "Operation '%s' [%d] (call %d from %s) for host '%s' with device '%s' returned: %d (%s)%s%s", cmd->action, pid, cmd->id, cmd->client_name, cmd->victim, cmd->device, rc, pcmk_strerror(rc), (next? ", retrying with " : ""), (next ? next : "")); } else { do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE, "Operation '%s' [%d] for device '%s' returned: %d (%s)%s%s", cmd->action, pid, cmd->device, rc, pcmk_strerror(rc), (next? ", retrying with " : ""), (next ? next : "")); } if (output) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s:%d", cmd->device, pid); crm_log_output(rc == 0 ? LOG_DEBUG : LOG_WARNING, prefix, output); free(prefix); } } static void stonith_send_async_reply(async_command_t * cmd, const char *output, int rc, GPid pid) { xmlNode *reply = NULL; gboolean bcast = FALSE; reply = stonith_construct_async_reply(cmd, output, NULL, rc); if (safe_str_eq(cmd->action, "metadata")) { /* Too verbose to log */ crm_trace("Metadata query for %s", cmd->device); output = NULL; } else if (crm_str_eq(cmd->action, "monitor", TRUE) || crm_str_eq(cmd->action, "list", TRUE) || crm_str_eq(cmd->action, "status", TRUE)) { crm_trace("Never broadcast %s replies", cmd->action); } else if (!stand_alone && safe_str_eq(cmd->origin, cmd->victim) && safe_str_neq(cmd->action, "on")) { crm_trace("Broadcast %s reply for %s", cmd->action, cmd->victim); crm_xml_add(reply, F_SUBTYPE, "broadcast"); bcast = TRUE; } log_operation(cmd, rc, pid, NULL, output); crm_log_xml_trace(reply, "Reply"); if (bcast) { crm_xml_add(reply, F_STONITH_OPERATION, T_STONITH_NOTIFY); send_cluster_message(NULL, crm_msg_stonith_ng, reply, FALSE); } else if (cmd->origin) { crm_trace("Directed reply to %s", cmd->origin); send_cluster_message(crm_get_peer(0, cmd->origin), crm_msg_stonith_ng, reply, FALSE); } else { crm_trace("Directed local %ssync reply to %s", (cmd->options & st_opt_sync_call) ? "" : "a-", cmd->client_name); do_local_reply(reply, cmd->client, cmd->options & st_opt_sync_call, FALSE); } if (stand_alone) { /* Do notification with a clean data object */ xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); crm_xml_add_int(notify_data, F_STONITH_RC, rc); crm_xml_add(notify_data, F_STONITH_TARGET, cmd->victim); crm_xml_add(notify_data, F_STONITH_OPERATION, cmd->op); crm_xml_add(notify_data, F_STONITH_DELEGATE, "localhost"); crm_xml_add(notify_data, F_STONITH_DEVICE, cmd->device); crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id); crm_xml_add(notify_data, F_STONITH_ORIGIN, cmd->client); do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data); do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); } free_xml(reply); } static void cancel_stonith_command(async_command_t * cmd) { stonith_device_t *device; CRM_CHECK(cmd != NULL, return); if (!cmd->device) { return; } device = g_hash_table_lookup(device_list, cmd->device); if (device) { crm_trace("Cancel scheduled %s on %s", cmd->action, device->id); device->pending_ops = g_list_remove(device->pending_ops, cmd); } } static void st_child_done(GPid pid, int rc, const char *output, gpointer user_data) { stonith_device_t *device = NULL; stonith_device_t *next_device = NULL; async_command_t *cmd = user_data; GListPtr gIter = NULL; GListPtr gIterNext = NULL; CRM_CHECK(cmd != NULL, return); cmd->active_on = NULL; /* The device is ready to do something else now */ device = g_hash_table_lookup(device_list, cmd->device); if (device) { if (rc == pcmk_ok && (safe_str_eq(cmd->action, "list") || safe_str_eq(cmd->action, "monitor") || safe_str_eq(cmd->action, "status"))) { device->verified = TRUE; } mainloop_set_trigger(device->work); } crm_debug("Operation '%s' on '%s' completed with rc=%d (%d remaining)", cmd->action, cmd->device, rc, g_list_length(cmd->device_next)); if (rc == 0) { GListPtr iter; /* see if there are any required devices left to execute for this op */ for (iter = cmd->device_next; iter != NULL; iter = iter->next) { next_device = g_hash_table_lookup(device_list, iter->data); if (next_device != NULL && is_action_required(cmd->action, next_device)) { cmd->device_next = iter->next; break; } next_device = NULL; } } else if (rc != 0 && cmd->device_next && (is_action_required(cmd->action, device) == FALSE)) { /* if this device didn't work out, see if there are any others we can try. * if the failed device was 'required', we can't pick another device. */ next_device = g_hash_table_lookup(device_list, cmd->device_next->data); cmd->device_next = cmd->device_next->next; } /* this operation requires more fencing, hooray! */ if (next_device) { log_operation(cmd, rc, pid, next_device->id, output); schedule_stonith_command(cmd, next_device); /* Prevent cmd from being freed */ cmd = NULL; goto done; } stonith_send_async_reply(cmd, output, rc, pid); if (rc != 0) { goto done; } /* Check to see if any operations are scheduled to do the exact * same thing that just completed. If so, rather than * performing the same fencing operation twice, return the result * of this operation for all pending commands it matches. */ for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) { async_command_t *cmd_other = gIter->data; gIterNext = gIter->next; if (cmd == cmd_other) { continue; } /* A pending scheduled command matches the command that just finished if. * 1. The client connections are different. * 2. The node victim is the same. * 3. The fencing action is the same. * 4. The device scheduled to execute the action is the same. */ if (safe_str_eq(cmd->client, cmd_other->client) || safe_str_neq(cmd->victim, cmd_other->victim) || safe_str_neq(cmd->action, cmd_other->action) || safe_str_neq(cmd->device, cmd_other->device)) { continue; } /* Duplicate merging will do the right thing for either type of remapped * reboot. If the executing fencer remapped an unsupported reboot to * off, then cmd->action will be reboot and will be merged with any * other reboot requests. If the originating fencer remapped a * topology reboot to off then on, we will get here once with * cmd->action "off" and once with "on", and they will be merged * separately with similar requests. */ crm_notice ("Merging stonith action %s for node %s originating from client %s with identical stonith request from client %s", cmd_other->action, cmd_other->victim, cmd_other->client_name, cmd->client_name); cmd_list = g_list_remove_link(cmd_list, gIter); stonith_send_async_reply(cmd_other, output, rc, pid); cancel_stonith_command(cmd_other); free_async_command(cmd_other); g_list_free_1(gIter); } done: free_async_command(cmd); } static gint sort_device_priority(gconstpointer a, gconstpointer b) { const stonith_device_t *dev_a = a; const stonith_device_t *dev_b = b; if (dev_a->priority > dev_b->priority) { return -1; } else if (dev_a->priority < dev_b->priority) { return 1; } return 0; } static void stonith_fence_get_devices_cb(GList * devices, void *user_data) { async_command_t *cmd = user_data; stonith_device_t *device = NULL; crm_info("Found %d matching devices for '%s'", g_list_length(devices), cmd->victim); if (g_list_length(devices) > 0) { /* Order based on priority */ devices = g_list_sort(devices, sort_device_priority); device = g_hash_table_lookup(device_list, devices->data); if (device) { cmd->device_list = devices; cmd->device_next = devices->next; devices = NULL; /* list owned by cmd now */ } } /* we have a device, schedule it for fencing. */ if (device) { schedule_stonith_command(cmd, device); /* in progress */ return; } /* no device found! */ stonith_send_async_reply(cmd, NULL, -ENODEV, 0); free_async_command(cmd); g_list_free_full(devices, free); } static int stonith_fence(xmlNode * msg) { const char *device_id = NULL; stonith_device_t *device = NULL; async_command_t *cmd = create_async_command(msg); xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR); if (cmd == NULL) { return -EPROTO; } device_id = crm_element_value(dev, F_STONITH_DEVICE); if (device_id) { device = g_hash_table_lookup(device_list, device_id); if (device == NULL) { crm_err("Requested device '%s' is not available", device_id); return -ENODEV; } schedule_stonith_command(cmd, device); } else { const char *host = crm_element_value(dev, F_STONITH_TARGET); if (cmd->options & st_opt_cs_nodeid) { int nodeid = crm_atoi(host, NULL); crm_node_t *node = crm_find_known_peer_full(nodeid, NULL, CRM_GET_PEER_ANY); if (node) { host = node->uname; } } /* If we get to here, then self-fencing is implicitly allowed */ get_capable_devices(host, cmd->action, cmd->default_timeout, TRUE, cmd, stonith_fence_get_devices_cb); } return -EINPROGRESS; } xmlNode * stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data, int rc) { int lpc = 0; xmlNode *reply = NULL; const char *name = NULL; const char *value = NULL; const char *names[] = { F_STONITH_OPERATION, F_STONITH_CALLID, F_STONITH_CLIENTID, F_STONITH_CLIENTNAME, F_STONITH_REMOTE_OP_ID, F_STONITH_CALLOPTS }; crm_trace("Creating a basic reply"); reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __FUNCTION__); crm_xml_add(reply, F_TYPE, T_STONITH_NG); crm_xml_add(reply, "st_output", output); crm_xml_add_int(reply, F_STONITH_RC, rc); CRM_CHECK(request != NULL, crm_warn("Can't create a sane reply"); return reply); for (lpc = 0; lpc < DIMOF(names); lpc++) { name = names[lpc]; value = crm_element_value(request, name); crm_xml_add(reply, name, value); } if (data != NULL) { crm_trace("Attaching reply output"); add_message_xml(reply, F_STONITH_CALLDATA, data); } return reply; } static xmlNode * stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc) { xmlNode *reply = NULL; crm_trace("Creating a basic reply"); reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __FUNCTION__); crm_xml_add(reply, F_TYPE, T_STONITH_NG); crm_xml_add(reply, F_STONITH_OPERATION, cmd->op); crm_xml_add(reply, F_STONITH_DEVICE, cmd->device); crm_xml_add(reply, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id); crm_xml_add(reply, F_STONITH_CLIENTID, cmd->client); crm_xml_add(reply, F_STONITH_CLIENTNAME, cmd->client_name); crm_xml_add(reply, F_STONITH_TARGET, cmd->victim); crm_xml_add(reply, F_STONITH_ACTION, cmd->op); crm_xml_add(reply, F_STONITH_ORIGIN, cmd->origin); crm_xml_add_int(reply, F_STONITH_CALLID, cmd->id); crm_xml_add_int(reply, F_STONITH_CALLOPTS, cmd->options); crm_xml_add_int(reply, F_STONITH_RC, rc); crm_xml_add(reply, "st_output", output); if (data != NULL) { crm_info("Attaching reply output"); add_message_xml(reply, F_STONITH_CALLDATA, data); } return reply; } bool fencing_peer_active(crm_node_t *peer) { if (peer == NULL) { return FALSE; } else if (peer->uname == NULL) { return FALSE; } else if (is_set(peer->processes, crm_get_cluster_proc())) { return TRUE; } return FALSE; } /*! * \internal * \brief Determine if we need to use an alternate node to * fence the target. If so return that node's uname * * \retval NULL, no alternate host * \retval uname, uname of alternate host to use */ static const char * check_alternate_host(const char *target) { const char *alternate_host = NULL; crm_trace("Checking if we (%s) can fence %s", stonith_our_uname, target); if (find_topology_for_host(target) && safe_str_eq(target, stonith_our_uname)) { GHashTableIter gIter; crm_node_t *entry = NULL; g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { crm_trace("Checking for %s.%d != %s", entry->uname, entry->id, target); if (fencing_peer_active(entry) && safe_str_neq(entry->uname, target)) { alternate_host = entry->uname; break; } } if (alternate_host == NULL) { crm_err("No alternate host available to handle complex self fencing request"); g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { crm_notice("Peer[%d] %s", entry->id, entry->uname); } } } return alternate_host; } static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer, const char *client_id) { if (remote_peer) { send_cluster_message(crm_get_peer(0, remote_peer), crm_msg_stonith_ng, reply, FALSE); } else { do_local_reply(reply, client_id, is_set(call_options, st_opt_sync_call), remote_peer != NULL); } } static int handle_request(crm_client_t * client, uint32_t id, uint32_t flags, xmlNode * request, const char *remote_peer) { int call_options = 0; int rc = -EOPNOTSUPP; xmlNode *data = NULL; xmlNode *reply = NULL; char *output = NULL; const char *op = crm_element_value(request, F_STONITH_OPERATION); const char *client_id = crm_element_value(request, F_STONITH_CLIENTID); crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (crm_str_eq(op, CRM_OP_REGISTER, TRUE)) { xmlNode *reply = create_xml_node(NULL, "reply"); CRM_ASSERT(client); crm_xml_add(reply, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(reply, F_STONITH_CLIENTID, client->id); crm_ipcs_send(client, id, reply, flags); client->request_id = 0; free_xml(reply); return 0; } else if (crm_str_eq(op, STONITH_OP_EXEC, TRUE)) { rc = stonith_device_action(request, &output); } else if (crm_str_eq(op, STONITH_OP_TIMEOUT_UPDATE, TRUE)) { const char *call_id = crm_element_value(request, F_STONITH_CALLID); const char *client_id = crm_element_value(request, F_STONITH_CLIENTID); int op_timeout = 0; crm_element_value_int(request, F_STONITH_TIMEOUT, &op_timeout); do_stonith_async_timeout_update(client_id, call_id, op_timeout); return 0; } else if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) { if (remote_peer) { create_remote_stonith_op(client_id, request, TRUE); /* Record it for the future notification */ } stonith_query(request, remote_peer, client_id, call_options); return 0; } else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) { const char *flag_name = NULL; CRM_ASSERT(client); flag_name = crm_element_value(request, F_STONITH_NOTIFY_ACTIVATE); if (flag_name) { crm_debug("Setting %s callbacks for %s (%s): ON", flag_name, client->name, client->id); client->options |= get_stonith_flag(flag_name); } flag_name = crm_element_value(request, F_STONITH_NOTIFY_DEACTIVATE); if (flag_name) { crm_debug("Setting %s callbacks for %s (%s): off", flag_name, client->name, client->id); client->options |= get_stonith_flag(flag_name); } if (flags & crm_ipc_client_response) { crm_ipcs_send_ack(client, id, flags, "ack", __FUNCTION__, __LINE__); } return 0; } else if (crm_str_eq(op, STONITH_OP_RELAY, TRUE)) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); crm_notice("Peer %s has received a forwarded fencing request from %s to fence (%s) peer %s", stonith_our_uname, client ? client->name : remote_peer, crm_element_value(dev, F_STONITH_ACTION), crm_element_value(dev, F_STONITH_TARGET)); if (initiate_remote_stonith_op(NULL, request, FALSE) != NULL) { rc = -EINPROGRESS; } } else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) { if (remote_peer || stand_alone) { rc = stonith_fence(request); } else if (call_options & st_opt_manual_ack) { remote_fencing_op_t *rop = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); const char *target = crm_element_value(dev, F_STONITH_TARGET); crm_notice("Received manual confirmation that %s is fenced", target); rop = initiate_remote_stonith_op(client, request, TRUE); rc = stonith_manual_ack(request, rop); } else { const char *alternate_host = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); const char *target = crm_element_value(dev, F_STONITH_TARGET); const char *action = crm_element_value(dev, F_STONITH_ACTION); const char *device = crm_element_value(dev, F_STONITH_DEVICE); if (client) { int tolerance = 0; crm_notice("Client %s.%.8s wants to fence (%s) '%s' with device '%s'", client->name, client->id, action, target, device ? device : "(any)"); crm_element_value_int(dev, F_STONITH_TOLERANCE, &tolerance); if (stonith_check_fence_tolerance(tolerance, target, action)) { rc = 0; goto done; } } else { crm_notice("Peer %s wants to fence (%s) '%s' with device '%s'", remote_peer, action, target, device ? device : "(any)"); } alternate_host = check_alternate_host(target); if (alternate_host && client) { const char *client_id = NULL; crm_notice("Forwarding complex self fencing request to peer %s", alternate_host); if (client->id) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } /* Create a record of it, otherwise call_id will be 0 if we need to notify of failures */ create_remote_stonith_op(client_id, request, FALSE); crm_xml_add(request, F_STONITH_OPERATION, STONITH_OP_RELAY); crm_xml_add(request, F_STONITH_CLIENTID, client->id); send_cluster_message(crm_get_peer(0, alternate_host), crm_msg_stonith_ng, request, FALSE); rc = -EINPROGRESS; } else if (initiate_remote_stonith_op(client, request, FALSE) != NULL) { rc = -EINPROGRESS; } } } else if (crm_str_eq(op, STONITH_OP_FENCE_HISTORY, TRUE)) { rc = stonith_fence_history(request, &data, remote_peer, call_options); if (call_options & st_opt_discard_reply) { /* we don't expect answers to the broadcast * we might have sent out */ free_xml(data); return pcmk_ok; } } else if (crm_str_eq(op, STONITH_OP_DEVICE_ADD, TRUE)) { const char *device_id = NULL; rc = stonith_device_register(request, &device_id, FALSE); do_stonith_notify_device(call_options, op, rc, device_id); } else if (crm_str_eq(op, STONITH_OP_DEVICE_DEL, TRUE)) { xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request, LOG_ERR); const char *device_id = crm_element_value(dev, XML_ATTR_ID); rc = stonith_device_remove(device_id, FALSE); do_stonith_notify_device(call_options, op, rc, device_id); } else if (crm_str_eq(op, STONITH_OP_LEVEL_ADD, TRUE)) { char *device_id = NULL; rc = stonith_level_register(request, &device_id); do_stonith_notify_level(call_options, op, rc, device_id); free(device_id); } else if (crm_str_eq(op, STONITH_OP_LEVEL_DEL, TRUE)) { char *device_id = NULL; rc = stonith_level_remove(request, &device_id); do_stonith_notify_level(call_options, op, rc, device_id); } else if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) { int node_id = 0; const char *name = NULL; crm_element_value_int(request, XML_ATTR_ID, &node_id); name = crm_element_value(request, XML_ATTR_UNAME); reap_crm_member(node_id, name); return pcmk_ok; } else { crm_err("Unknown %s from %s", op, client ? client->name : remote_peer); crm_log_xml_warn(request, "UnknownOp"); } done: /* Always reply unless the request is in process still. * If in progress, a reply will happen async after the request * processing is finished */ if (rc != -EINPROGRESS) { crm_trace("Reply handling: %p %u %u %d %d %s", client, client?client->request_id:0, id, is_set(call_options, st_opt_sync_call), call_options, crm_element_value(request, F_STONITH_CALLOPTS)); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } reply = stonith_construct_reply(request, output, data, rc); stonith_send_reply(reply, call_options, remote_peer, client_id); } free(output); free_xml(data); free_xml(reply); return rc; } static void handle_reply(crm_client_t * client, xmlNode * request, const char *remote_peer) { const char *op = crm_element_value(request, F_STONITH_OPERATION); if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) { process_remote_stonith_query(request); } else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) { process_remote_stonith_exec(request); } else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) { /* Reply to a complex fencing op */ process_remote_stonith_exec(request); } else { crm_err("Unknown %s reply from %s", op, client ? client->name : remote_peer); crm_log_xml_warn(request, "UnknownOp"); } } void stonith_command(crm_client_t * client, uint32_t id, uint32_t flags, xmlNode * request, const char *remote_peer) { int call_options = 0; int rc = 0; gboolean is_reply = FALSE; /* Copy op for reporting. The original might get freed by handle_reply() * before we use it in crm_debug(): * handle_reply() * |- process_remote_stonith_exec() * |-- remote_op_done() * |--- handle_local_reply_and_notify() * |---- crm_xml_add(...F_STONITH_OPERATION...) * |--- free_xml(op->request) */ char *op = crm_element_value_copy(request, F_STONITH_OPERATION); if (get_xpath_object("//" T_STONITH_REPLY, request, LOG_TRACE)) { is_reply = TRUE; } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_debug("Processing %s%s %u from %s (%16x)", op, is_reply ? " reply" : "", id, client ? client->name : remote_peer, call_options); if (is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (is_reply) { handle_reply(client, request, remote_peer); } else { rc = handle_request(client, id, flags, request, remote_peer); } crm_debug("Processed %s%s from %s: %s (%d)", op, is_reply ? " reply" : "", client ? client->name : remote_peer, rc > 0 ? "" : pcmk_strerror(rc), rc); free(op); } diff --git a/daemons/fenced/pacemaker-fenced.h b/daemons/fenced/pacemaker-fenced.h index 3194e358cf..3a2edbbc4a 100644 --- a/daemons/fenced/pacemaker-fenced.h +++ b/daemons/fenced/pacemaker-fenced.h @@ -1,264 +1,254 @@ /* - * Copyright 2009-2018 Andrew Beekhof + * Copyright 2009-2019 the Pacemaker project contributors * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include /*! * \internal * \brief Check to see if target was fenced in the last few seconds. * \param tolerance, The number of seconds to look back in time * \param target, The node to search for * \param action, The action we want to match. * * \retval FALSE, not match * \retval TRUE, fencing operation took place in the last 'tolerance' number of seconds. */ gboolean stonith_check_fence_tolerance(int tolerance, const char *target, const char *action); enum st_device_flags { st_device_supports_list = 0x0001, st_device_supports_status = 0x0002, st_device_supports_reboot = 0x0004, }; typedef struct stonith_device_s { char *id; char *agent; char *namespace; /*! list of actions that must execute on the target node. Used for unfencing */ char *on_target_actions; GListPtr targets; time_t targets_age; gboolean has_attr_map; /* should nodeid parameter for victim be included in agent arguments */ gboolean include_nodeid; /* whether the cluster should automatically unfence nodes with the device */ gboolean automatic_unfencing; guint priority; enum st_device_flags flags; GHashTable *params; GHashTable *aliases; GList *pending_ops; crm_trigger_t *work; xmlNode *agent_metadata; /*! A verified device is one that has contacted the * agent successfully to perform a monitor operation */ gboolean verified; gboolean cib_registered; gboolean api_registered; } stonith_device_t; /* These values are used to index certain arrays by "phase". Usually an * operation has only one "phase", so phase is always zero. However, some * reboots are remapped to "off" then "on", in which case "reboot" will be * phase 0, "off" will be phase 1 and "on" will be phase 2. */ enum st_remap_phase { st_phase_requested = 0, st_phase_off = 1, st_phase_on = 2, st_phase_max = 3 }; typedef struct remote_fencing_op_s { /* The unique id associated with this operation */ char *id; /*! The node this operation will fence */ char *target; /*! The fencing action to perform on the target. (reboot, on, off) */ char *action; /*! When was the fencing action recorded (seconds since epoch) */ time_t created; /*! Marks if the final notifications have been sent to local stonith clients. */ gboolean notify_sent; /*! The number of query replies received */ guint replies; /*! The number of query replies expected */ guint replies_expected; /*! Does this node own control of this operation */ gboolean owner; /*! After query is complete, This the high level timer that expires the entire operation */ guint op_timer_total; /*! This timer expires the current fencing request. Many fencing * requests may exist in a single operation */ guint op_timer_one; /*! This timer expires the query request sent out to determine * what nodes are contain what devices, and who those devices can fence */ guint query_timer; /*! This is the default timeout to use for each fencing device if no * custom timeout is received in the query. */ gint base_timeout; /*! This is the calculated total timeout an operation can take before * expiring. This is calculated by adding together all the timeout * values associated with the devices this fencing operation may call */ gint total_timeout; /*! Delegate is the node being asked to perform a fencing action * on behalf of the node that owns the remote operation. Some operations * will involve multiple delegates. This value represents the final delegate * that is used. */ char *delegate; /*! The point at which the remote operation completed */ time_t completed; /*! The stonith_call_options associated with this remote operation */ long long call_options; /*! The current state of the remote operation. This indicates * what stage the op is in, query, exec, done, duplicate, failed. */ enum op_state state; /*! The node that owns the remote operation */ char *originator; /*! The local client id that initiated the fencing request */ char *client_id; /*! The client's call_id that initiated the fencing request */ int client_callid; /*! The name of client that initiated the fencing request */ char *client_name; /*! List of the received query results for all the nodes in the cpg group */ GListPtr query_results; /*! The original request that initiated the remote stonith operation */ xmlNode *request; /*! The current topology level being executed */ guint level; /*! The current operation phase being executed */ enum st_remap_phase phase; /*! Devices with automatic unfencing (always run if "on" requested, never if remapped) */ GListPtr automatic_list; /*! List of all devices at the currently executing topology level */ GListPtr devices_list; /*! Current entry in the topology device list */ GListPtr devices; /*! List of duplicate operations attached to this operation. Once this operation * completes, the duplicate operations will be closed out as well. */ GListPtr duplicates; } remote_fencing_op_t; enum st_callback_flags { st_callback_unknown = 0x0000, st_callback_notify_fence = 0x0001, st_callback_device_add = 0x0004, st_callback_device_del = 0x0010, st_callback_notify_history = 0x0020 }; /* * Complex fencing requirements are specified via fencing topologies. * A topology consists of levels; each level is a list of fencing devices. * Topologies are stored in a hash table by node name. When a node needs to be * fenced, if it has an entry in the topology table, the levels are tried * sequentially, and the devices in each level are tried sequentially. * Fencing is considered successful as soon as any level succeeds; * a level is considered successful if all its devices succeed. * Essentially, all devices at a given level are "and-ed" and the * levels are "or-ed". * * This structure is used for the topology table entries. * Topology levels start from 1, so levels[0] is unused and always NULL. */ typedef struct stonith_topology_s { int kind; /*! Node name regex or attribute name=value for which topology applies */ char *target; char *target_value; char *target_pattern; char *target_attribute; /*! Names of fencing devices at each topology level */ GListPtr levels[ST_LEVEL_MAX]; } stonith_topology_t; void init_device_list(void); void free_device_list(void); void init_topology_list(void); void free_topology_list(void); void free_stonith_remote_op_list(void); void init_stonith_remote_op_hash_table(GHashTable **table); void free_metadata_cache(void); long long get_stonith_flag(const char *name); void stonith_command(crm_client_t * client, uint32_t id, uint32_t flags, xmlNode * op_request, const char *remote_peer); int stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib); int stonith_device_remove(const char *id, gboolean from_cib); char *stonith_level_key(xmlNode * msg, int mode); int stonith_level_kind(xmlNode * msg); int stonith_level_register(xmlNode * msg, char **desc); int stonith_level_remove(xmlNode * msg, char **desc); stonith_topology_t *find_topology_for_host(const char *host); void do_local_reply(xmlNode * notify_src, const char *client_id, gboolean sync_reply, gboolean from_peer); xmlNode *stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data, int rc); void do_stonith_async_timeout_update(const char *client, const char *call_id, int timeout); void do_stonith_notify(int options, const char *type, int result, xmlNode * data); void do_stonith_notify_device(int options, const char *op, int rc, const char *desc); void do_stonith_notify_level(int options, const char *op, int rc, const char *desc); remote_fencing_op_t *initiate_remote_stonith_op(crm_client_t * client, xmlNode * request, gboolean manual_ack); int process_remote_stonith_exec(xmlNode * msg); int process_remote_stonith_query(xmlNode * msg); void *create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer); int stonith_fence_history(xmlNode *msg, xmlNode **output, const char *remote_peer, int options); void stonith_fence_history_trim(void); bool fencing_peer_active(crm_node_t *peer); int stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op); gboolean string_in_list(GListPtr list, const char *item); gboolean node_has_attr(const char *node, const char *name, const char *value); -void -schedule_internal_command(const char *origin, - stonith_device_t * device, - const char *action, - const char *victim, - int timeout, - void *internal_user_data, - void (*done_cb) (GPid pid, int rc, const char *output, - gpointer user_data)); - extern char *stonith_our_uname; extern gboolean stand_alone; extern GHashTable *device_list; extern GHashTable *topology; extern long stonith_watchdog_timeout_ms; extern GHashTable *stonith_remote_op_list; diff --git a/include/crm/stonith-ng.h b/include/crm/stonith-ng.h index b7365a9700..b6407324df 100644 --- a/include/crm/stonith-ng.h +++ b/include/crm/stonith-ng.h @@ -1,550 +1,554 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifdef __cplusplus extern "C" { #endif /** * \file * \brief Fencing aka. STONITH * \ingroup fencing */ #ifndef STONITH_NG__H # define STONITH_NG__H # include # include # include // bool # include // uint32_t # include // time_t # define T_STONITH_NOTIFY_DISCONNECT "st_notify_disconnect" # define T_STONITH_NOTIFY_FENCE "st_notify_fence" # define T_STONITH_NOTIFY_HISTORY "st_notify_history" /* *INDENT-OFF* */ enum stonith_state { stonith_connected_command, stonith_connected_query, stonith_disconnected, }; enum stonith_call_options { st_opt_none = 0x00000000, st_opt_verbose = 0x00000001, st_opt_allow_suicide = 0x00000002, st_opt_manual_ack = 0x00000008, st_opt_discard_reply = 0x00000010, /* st_opt_all_replies = 0x00000020, */ st_opt_topology = 0x00000040, st_opt_scope_local = 0x00000100, st_opt_cs_nodeid = 0x00000200, st_opt_sync_call = 0x00001000, /*! Allow the timeout period for a callback to be adjusted * based on the time the server reports the operation will take. */ st_opt_timeout_updates = 0x00002000, /*! Only report back if operation is a success in callback */ st_opt_report_only_success = 0x00004000, /* used where ever apropriate - e.g. cleanup of history */ st_opt_cleanup = 0x000080000, /* used where ever apropriate - e.g. send out a history query to all nodes */ st_opt_broadcast = 0x000100000, }; /*! Order matters here, do not change values */ enum op_state { st_query, st_exec, st_done, st_duplicate, st_failed, }; // Supported fence agent interface standards enum stonith_namespace { st_namespace_invalid, st_namespace_any, st_namespace_internal, // Implemented internally by Pacemaker /* Neither of these projects are active any longer, but the fence agent * interfaces they created are still in use and supported by Pacemaker. */ st_namespace_rhcs, // Red Hat Cluster Suite compatible st_namespace_lha, // Linux-HA compatible }; enum stonith_namespace stonith_text2namespace(const char *namespace_s); const char *stonith_namespace2text(enum stonith_namespace st_namespace); enum stonith_namespace stonith_get_namespace(const char *agent, const char *namespace_s); typedef struct stonith_key_value_s { char *key; char *value; struct stonith_key_value_s *next; } stonith_key_value_t; typedef struct stonith_history_s { char *target; char *action; char *origin; char *delegate; char *client; int state; time_t completed; struct stonith_history_s *next; } stonith_history_t; typedef struct stonith_s stonith_t; typedef struct stonith_event_s { char *id; char *type; char *message; char *operation; int result; char *origin; char *target; char *action; char *executioner; char *device; /*! The name of the client that initiated the action. */ char *client_origin; } stonith_event_t; typedef struct stonith_callback_data_s { int rc; int call_id; void *userdata; } stonith_callback_data_t; typedef struct stonith_api_operations_s { /*! * \brief Destroy the stonith api structure. */ int (*free) (stonith_t *st); /*! * \brief Connect to the local stonith daemon. * * \retval 0, success * \retval negative error code on failure */ int (*connect) (stonith_t *st, const char *name, int *stonith_fd); /*! * \brief Disconnect from the local stonith daemon. * * \retval 0, success * \retval negative error code on failure */ int (*disconnect)(stonith_t *st); /*! * \brief Remove a registered stonith device with the local stonith daemon. * * \note Synchronous, guaranteed to occur in daemon before function returns. * * \retval 0, success * \retval negative error code on failure */ int (*remove_device)( stonith_t *st, int options, const char *name); /*! * \brief Register a stonith device with the local stonith daemon. * * \note Synchronous, guaranteed to occur in daemon before function returns. * * \retval 0, success * \retval negative error code on failure */ int (*register_device)( stonith_t *st, int options, const char *id, const char *provider, const char *agent, stonith_key_value_t *params); /*! * \brief Remove a fencing level for a specific node. * * \note This feature is not available when stonith is in standalone mode. * * \retval 0, success * \retval negative error code on failure */ int (*remove_level)( stonith_t *st, int options, const char *node, int level); /*! * \brief Register a fencing level containing the fencing devices to be used * at that level for a specific node. * * \note This feature is not available when stonith is in standalone mode. * * \retval 0, success * \retval negative error code on failure */ int (*register_level)( stonith_t *st, int options, const char *node, int level, stonith_key_value_t *device_list); /*! * \brief Get the metadata documentation for a resource. * * \note Value is returned in output. Output must be freed when set. * * \retval 0 success * \retval negative error code on failure */ int (*metadata)(stonith_t *st, int options, const char *device, const char *provider, char **output, int timeout); /*! * \brief Retrieve a list of installed stonith agents * * \note if provider is not provided, all known agents will be returned * \note list must be freed using stonith_key_value_freeall() * \note call_options parameter is not used, it is reserved for future use. * * \retval num items in list on success * \retval negative error code on failure */ int (*list_agents)(stonith_t *stonith, int call_options, const char *provider, stonith_key_value_t **devices, int timeout); /*! * \brief Retrieve string listing hosts and port assignments from a local stonith device. * * \retval 0 on success * \retval negative error code on failure */ int (*list)(stonith_t *st, int options, const char *id, char **list_output, int timeout); /*! * \brief Check to see if a local stonith device is reachable * * \retval 0 on success * \retval negative error code on failure */ int (*monitor)(stonith_t *st, int options, const char *id, int timeout); /*! * \brief Check to see if a local stonith device's port is reachable * * \retval 0 on success * \retval negative error code on failure */ int (*status)(stonith_t *st, int options, const char *id, const char *port, int timeout); /*! * \brief Retrieve a list of registered stonith devices. * * \note If node is provided, only devices that can fence the node id * will be returned. * * \retval num items in list on success * \retval negative error code on failure */ int (*query)(stonith_t *st, int options, const char *node, stonith_key_value_t **devices, int timeout); /*! * \brief Issue a fencing action against a node. * * \note Possible actions are, 'on', 'off', and 'reboot'. * * \param st, stonith connection * \param options, call options * \param node, The target node to fence * \param action, The fencing action to take * \param timeout, The default per device timeout to use with each device * capable of fencing the target. * * \retval 0 success * \retval negative error code on failure. */ int (*fence)(stonith_t *st, int options, const char *node, const char *action, int timeout, int tolerance); /*! * \brief Manually confirm that a node is down. * * \retval 0 success * \retval negative error code on failure. */ int (*confirm)(stonith_t *st, int options, const char *node); /*! * \brief Retrieve a list of fencing operations that have occurred for a specific node. * * \note History is not available in standalone mode. * * \retval 0 success * \retval negative error code on failure. */ int (*history)(stonith_t *st, int options, const char *node, stonith_history_t **output, int timeout); int (*register_notification)( stonith_t *st, const char *event, void (*notify)(stonith_t *st, stonith_event_t *e)); int (*remove_notification)(stonith_t *st, const char *event); /*! * \brief Register a callback to receive the result of an asynchronous call * * \param[in] call_id The call ID to register callback for * \param[in] timeout Default time to wait until callback expires * \param[in] options Bitmask of \c stonith_call_options (respects * \c st_opt_timeout_updates and * \c st_opt_report_only_success) * \param[in] userdata Pointer that will be given to callback * \param[in] callback_name Unique name to identify callback * \param[in] callback The callback function to register * * \return \c TRUE on success, \c FALSE if call_id is negative, -errno otherwise * * \todo This function should return \c pcmk_ok on success, and \c call_id * when negative, but that would break backward compatibility. */ int (*register_callback)(stonith_t *st, int call_id, int timeout, int options, void *userdata, const char *callback_name, void (*callback)(stonith_t *st, stonith_callback_data_t *data)); /*! * \brief Remove a registered callback for a given call id. */ int (*remove_callback)(stonith_t *st, int call_id, bool all_callbacks); /*! * \brief Remove fencing level for specific node, node regex or attribute * * \param[in] st Fencer connection to use * \param[in] options Bitmask of stonith_call_options to pass to the fencer * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to remove * * \return 0 on success, negative error code otherwise * * \note This feature is not available when stonith is in standalone mode. * The caller should set only one of node, pattern or attr/value. */ int (*remove_level_full)(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level); /*! * \brief Register fencing level for specific node, node regex or attribute * * \param[in] st Fencer connection to use * \param[in] options Bitmask of stonith_call_options to pass to fencer * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to add * \param[in] device_list Devices to use in level * * \return 0 on success, negative error code otherwise * * \note This feature is not available when stonith is in standalone mode. * The caller should set only one of node, pattern or attr/value. */ int (*register_level_full)(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list); /*! * \brief Validate an arbitrary stonith device configuration * * \param[in] st Stonithd connection to use * \param[in] call_options Bitmask of stonith_call_options to use with fencer * \param[in] rsc_id ID used to replace CIB secrets in params * \param[in] namespace_s Namespace of fence agent to validate (optional) * \param[in] agent Fence agent to validate * \param[in] params Configuration parameters to pass to fence agent * \param[in] timeout Fail if no response within this many seconds * \param[out] output If non-NULL, where to store any agent output * \param[out] error_output If non-NULL, where to store agent error output * * \return pcmk_ok if validation succeeds, -errno otherwise * * \note If pcmk_ok is returned, the caller is responsible for freeing * the output (if requested). */ int (*validate)(stonith_t *st, int call_options, const char *rsc_id, const char *namespace_s, const char *agent, stonith_key_value_t *params, int timeout, char **output, char **error_output); } stonith_api_operations_t; struct stonith_s { enum stonith_state state; int call_id; int call_timeout; void *st_private; stonith_api_operations_t *cmds; }; /* *INDENT-ON* */ /* Core functions */ stonith_t *stonith_api_new(void); void stonith_api_delete(stonith_t * st); void stonith_dump_pending_callbacks(stonith_t * st); // deprecated (use stonith_get_namespace() instead) const char *get_stonith_provider(const char *agent, const char *provider); bool stonith_dispatch(stonith_t * st); stonith_key_value_t *stonith_key_value_add(stonith_key_value_t * kvp, const char *key, const char *value); void stonith_key_value_freeall(stonith_key_value_t * kvp, int keys, int values); void stonith_history_free(stonith_history_t *history); +// Convenience functions +int stonith_api_connect_retry(stonith_t *st, const char *name, + int max_attempts); + /* Basic helpers that allows nodes to be fenced and the history to be * queried without mainloop or the caller understanding the full API * * At least one of nodeid and uname are required */ int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off); time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress); /* * Helpers for using the above functions without install-time dependencies * * Usage: * #include * * To turn a node off by corosync nodeid: * stonith_api_kick_helper(nodeid, 120, 1); * * To check the last fence date/time (also by nodeid): * last = stonith_api_time_helper(nodeid, 0); * * To check if fencing is in progress: * if(stonith_api_time_helper(nodeid, 1) > 0) { ... } * * eg. #include #include #include int main(int argc, char ** argv) { int rc = 0; int nodeid = 102; rc = stonith_api_time_helper(nodeid, 0); printf("%d last fenced at %s\n", nodeid, ctime(rc)); rc = stonith_api_kick_helper(nodeid, 120, 1); printf("%d fence result: %d\n", nodeid, rc); rc = stonith_api_time_helper(nodeid, 0); printf("%d last fenced at %s\n", nodeid, ctime(rc)); return 0; } */ # define STONITH_LIBRARY "libstonithd.so.26" typedef int (*st_api_kick_fn) (int nodeid, const char *uname, int timeout, bool off); typedef time_t (*st_api_time_fn) (int nodeid, const char *uname, bool in_progress); static inline int stonith_api_kick_helper(uint32_t nodeid, int timeout, bool off) { static void *st_library = NULL; static st_api_kick_fn st_kick_fn; if (st_library == NULL) { st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY); } if (st_library && st_kick_fn == NULL) { st_kick_fn = (st_api_kick_fn) dlsym(st_library, "stonith_api_kick"); } if (st_kick_fn == NULL) { #ifdef ELIBACC return -ELIBACC; #else return -ENOSYS; #endif } return (*st_kick_fn) (nodeid, NULL, timeout, off); } static inline time_t stonith_api_time_helper(uint32_t nodeid, bool in_progress) { static void *st_library = NULL; static st_api_time_fn st_time_fn; if (st_library == NULL) { st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY); } if (st_library && st_time_fn == NULL) { st_time_fn = (st_api_time_fn) dlsym(st_library, "stonith_api_time"); } if (st_time_fn == NULL) { return 0; } return (*st_time_fn) (nodeid, NULL, in_progress); } /** * Does the given agent describe a stonith resource that can exist? * * \param[in] agent What is the name of the agent? * \param[in] timeout Timeout to use when querying. If 0 is given, * use a default of 120. * * \return A boolean */ bool stonith_agent_exists(const char *agent, int timeout); /*! * \brief Turn stonith action into a more readable string. * * \param action Stonith action */ const char *stonith_action_str(const char *action); #ifdef __cplusplus } #endif #endif diff --git a/lib/fencing/st_client.c b/lib/fencing/st_client.c index 270ef8d196..d949fe1615 100644 --- a/lib/fencing/st_client.c +++ b/lib/fencing/st_client.c @@ -1,2281 +1,2306 @@ /* - * Copyright 2004-2018 Andrew Beekhof + * Copyright 2004-2019 the Pacemaker project contributors * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if SUPPORT_CIBSECRETS # include #endif CRM_TRACE_INIT_DATA(stonith); struct stonith_action_s { /*! user defined data */ char *agent; char *action; char *victim; GHashTable *args; int timeout; int async; void *userdata; void (*done_cb) (GPid pid, gint status, const char *output, gpointer user_data); void (*fork_cb) (GPid pid, gpointer user_data); svc_action_t *svc_action; /*! internal timing information */ time_t initial_start_time; int tries; int remaining_timeout; int max_retries; /* device output data */ GPid pid; int rc; char *output; char *error; }; typedef struct stonith_private_s { char *token; crm_ipc_t *ipc; mainloop_io_t *source; GHashTable *stonith_op_callback_table; GList *notify_list; void (*op_callback) (stonith_t * st, stonith_callback_data_t * data); } stonith_private_t; typedef struct stonith_notify_client_s { const char *event; const char *obj_id; /* implement one day */ const char *obj_type; /* implement one day */ void (*notify) (stonith_t * st, stonith_event_t * e); } stonith_notify_client_t; typedef struct stonith_callback_client_s { void (*callback) (stonith_t * st, stonith_callback_data_t * data); const char *id; void *user_data; gboolean only_success; gboolean allow_timeout_updates; struct timer_rec_s *timer; } stonith_callback_client_t; struct notify_blob_s { stonith_t *stonith; xmlNode *xml; }; struct timer_rec_s { int call_id; int timeout; guint ref; stonith_t *stonith; }; typedef int (*stonith_op_t) (const char *, int, const char *, xmlNode *, xmlNode *, xmlNode *, xmlNode **, xmlNode **); bool stonith_dispatch(stonith_t * st); xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options); int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout); static void stonith_connection_destroy(gpointer user_data); static void stonith_send_notification(gpointer data, gpointer user_data); static int internal_stonith_action_execute(stonith_action_t * action); static void log_action(stonith_action_t *action, pid_t pid); /*! * \brief Get agent namespace by name * * \param[in] namespace_s Name of namespace as string * * \return Namespace as enum value */ enum stonith_namespace stonith_text2namespace(const char *namespace_s) { if ((namespace_s == NULL) || !strcmp(namespace_s, "any")) { return st_namespace_any; } else if (!strcmp(namespace_s, "redhat") || !strcmp(namespace_s, "stonith-ng")) { return st_namespace_rhcs; } else if (!strcmp(namespace_s, "internal")) { return st_namespace_internal; } else if (!strcmp(namespace_s, "heartbeat")) { return st_namespace_lha; } return st_namespace_invalid; } /*! * \brief Get agent namespace name * * \param[in] namespace Namespace as enum value * * \return Namespace name as string */ const char * stonith_namespace2text(enum stonith_namespace st_namespace) { switch (st_namespace) { case st_namespace_any: return "any"; case st_namespace_rhcs: return "stonith-ng"; case st_namespace_internal: return "internal"; case st_namespace_lha: return "heartbeat"; default: break; } return "unsupported"; } /*! * \brief Determine namespace of a fence agent * * \param[in] agent Fence agent type * \param[in] namespace_s Name of agent namespace as string, if known * * \return Namespace of specified agent, as enum value */ enum stonith_namespace stonith_get_namespace(const char *agent, const char *namespace_s) { if (safe_str_eq(namespace_s, "internal")) { return st_namespace_internal; } if (stonith__agent_is_rhcs(agent)) { return st_namespace_rhcs; } #if HAVE_STONITH_STONITH_H if (stonith__agent_is_lha(agent)) { return st_namespace_lha; } #endif crm_err("Unknown fence agent: %s", agent); return st_namespace_invalid; } static void log_action(stonith_action_t *action, pid_t pid) { if (action->output) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s[%d] stdout:", action->agent, pid); crm_log_output(LOG_TRACE, prefix, action->output); free(prefix); } if (action->error) { /* Logging the whole string confuses syslog when the string is xml */ char *prefix = crm_strdup_printf("%s[%d] stderr:", action->agent, pid); crm_log_output(LOG_WARNING, prefix, action->error); free(prefix); } } static void stonith_connection_destroy(gpointer user_data) { stonith_t *stonith = user_data; stonith_private_t *native = NULL; struct notify_blob_s blob; crm_trace("Sending destroyed notification"); blob.stonith = stonith; blob.xml = create_xml_node(NULL, "notify"); native = stonith->st_private; native->ipc = NULL; native->source = NULL; stonith->state = stonith_disconnected; crm_xml_add(blob.xml, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(blob.xml, F_SUBTYPE, T_STONITH_NOTIFY_DISCONNECT); g_list_foreach(native->notify_list, stonith_send_notification, &blob); free_xml(blob.xml); } xmlNode * create_device_registration_xml(const char *id, enum stonith_namespace namespace, const char *agent, stonith_key_value_t *params, const char *rsc_provides) { xmlNode *data = create_xml_node(NULL, F_STONITH_DEVICE); xmlNode *args = create_xml_node(data, XML_TAG_ATTRS); #if HAVE_STONITH_STONITH_H if (namespace == st_namespace_any) { namespace = stonith_get_namespace(agent, NULL); } if (namespace == st_namespace_lha) { hash2field((gpointer) "plugin", (gpointer) agent, args); agent = "fence_legacy"; } #endif crm_xml_add(data, XML_ATTR_ID, id); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, "agent", agent); if ((namespace != st_namespace_any) && (namespace != st_namespace_invalid)) { crm_xml_add(data, "namespace", stonith_namespace2text(namespace)); } if (rsc_provides) { crm_xml_add(data, "rsc_provides", rsc_provides); } for (; params; params = params->next) { hash2field((gpointer) params->key, (gpointer) params->value, args); } return data; } static int stonith_api_register_device(stonith_t * st, int call_options, const char *id, const char *namespace, const char *agent, stonith_key_value_t * params) { int rc = 0; xmlNode *data = NULL; data = create_device_registration_xml(id, stonith_text2namespace(namespace), agent, params, NULL); rc = stonith_send_command(st, STONITH_OP_DEVICE_ADD, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_device(stonith_t * st, int call_options, const char *name) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, XML_ATTR_ID, name); rc = stonith_send_command(st, STONITH_OP_DEVICE_DEL, data, NULL, call_options, 0); free_xml(data); return rc; } static int stonith_api_remove_level_full(stonith_t *st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level) { int rc = 0; xmlNode *data = NULL; CRM_CHECK(node || pattern || (attr && value), return -EINVAL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); rc = stonith_send_command(st, STONITH_OP_LEVEL_DEL, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_remove_level(stonith_t * st, int options, const char *node, int level) { return stonith_api_remove_level_full(st, options, node, NULL, NULL, NULL, level); } /*! * \internal * \brief Create XML for fence topology level registration request * * \param[in] node If not NULL, target level by this node name * \param[in] pattern If not NULL, target by node name using this regex * \param[in] attr If not NULL, target by this node attribute * \param[in] value If not NULL, target by this node attribute value * \param[in] level Index number of level to register * \param[in] device_list List of devices in level * * \return Newly allocated XML tree on success, NULL otherwise * * \note The caller should set only one of node, pattern or attr/value. */ xmlNode * create_level_registration_xml(const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list) { int len = 0; char *list = NULL; xmlNode *data; CRM_CHECK(node || pattern || (attr && value), return NULL); data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); CRM_CHECK(data, return NULL); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add_int(data, XML_ATTR_ID, level); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); if (node) { crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); } else if (pattern) { crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern); } else { crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr); crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value); } for (; device_list; device_list = device_list->next) { int adding = strlen(device_list->value); if(list) { adding++; /* +1 space */ } crm_trace("Adding %s (%dc) at offset %d", device_list->value, adding, len); list = realloc_safe(list, len + adding + 1); /* +1 EOS */ if (list == NULL) { crm_perror(LOG_CRIT, "Could not create device list"); free_xml(data); return NULL; } sprintf(list + len, "%s%s", len?",":"", device_list->value); len += adding; } crm_xml_add(data, XML_ATTR_STONITH_DEVICES, list); free(list); return data; } static int stonith_api_register_level_full(stonith_t * st, int options, const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list) { int rc = 0; xmlNode *data = create_level_registration_xml(node, pattern, attr, value, level, device_list); CRM_CHECK(data != NULL, return -EINVAL); rc = stonith_send_command(st, STONITH_OP_LEVEL_ADD, data, NULL, options, 0); free_xml(data); return rc; } static int stonith_api_register_level(stonith_t * st, int options, const char *node, int level, stonith_key_value_t * device_list) { return stonith_api_register_level_full(st, options, node, NULL, NULL, NULL, level, device_list); } static void append_arg(const char *key, const char *value, GHashTable **args) { CRM_CHECK(key != NULL, return); CRM_CHECK(value != NULL, return); CRM_CHECK(args != NULL, return); if (strstr(key, "pcmk_")) { return; } else if (strstr(key, CRM_META)) { return; } else if (safe_str_eq(key, "crm_feature_set")) { return; } if (!*args) { *args = crm_str_table_new(); } CRM_CHECK(*args != NULL, return); crm_trace("Appending: %s=%s", key, value); g_hash_table_replace(*args, strdup(key), strdup(value)); } static void append_config_arg(gpointer key, gpointer value, gpointer user_data) { /* The fencer will filter action out when it registers the device, * but ignore it here just in case any other library callers * fail to do so. */ if (safe_str_neq(key, STONITH_ATTR_ACTION_OP)) { append_arg(key, value, user_data); return; } } static GHashTable * make_args(const char *agent, const char *action, const char *victim, uint32_t victim_nodeid, GHashTable * device_args, GHashTable * port_map) { char buffer[512]; GHashTable *arg_list = NULL; const char *value = NULL; CRM_CHECK(action != NULL, return NULL); snprintf(buffer, sizeof(buffer), "pcmk_%s_action", action); if (device_args) { value = g_hash_table_lookup(device_args, buffer); } if (value) { crm_info("Substituting action '%s' for requested operation '%s'", value, action); action = value; } append_arg(STONITH_ATTR_ACTION_OP, action, &arg_list); if (victim && device_args) { const char *alias = victim; const char *param = g_hash_table_lookup(device_args, STONITH_ATTR_HOSTARG); if (port_map && g_hash_table_lookup(port_map, victim)) { alias = g_hash_table_lookup(port_map, victim); } /* Always supply the node's name, too: * https://github.com/ClusterLabs/fence-agents/blob/master/doc/FenceAgentAPI.md */ append_arg("nodename", victim, &arg_list); if (victim_nodeid) { char nodeid_str[33] = { 0, }; if (snprintf(nodeid_str, 33, "%u", (unsigned int)victim_nodeid)) { crm_info("For stonith action (%s) for victim %s, adding nodeid (%s) to parameters", action, victim, nodeid_str); append_arg("nodeid", nodeid_str, &arg_list); } } /* Check if we need to supply the victim in any other form */ if(safe_str_eq(agent, "fence_legacy")) { value = agent; } else if (param == NULL) { param = "port"; value = g_hash_table_lookup(device_args, param); } else if (safe_str_eq(param, "none")) { value = param; /* Nothing more to do */ } else { value = g_hash_table_lookup(device_args, param); } /* Don't overwrite explictly set values for $param */ if (value == NULL || safe_str_eq(value, "dynamic")) { crm_debug("Performing %s action for node '%s' as '%s=%s'", action, victim, param, alias); append_arg(param, alias, &arg_list); } } if (device_args) { g_hash_table_foreach(device_args, append_config_arg, &arg_list); } return arg_list; } /*! * \internal * \brief Free all memory used by a stonith action * * \param[in,out] action Action to free */ void stonith__destroy_action(stonith_action_t *action) { if (action) { free(action->agent); if (action->args) { g_hash_table_destroy(action->args); } free(action->action); free(action->victim); if (action->svc_action) { services_action_free(action->svc_action); } free(action->output); free(action->error); free(action); } } /*! * \internal * \brief Get the result of an executed stonith action * * \param[in,out] action Executed action * \param[out] rc Where to store result code (or NULL) * \param[out] output Where to store standard output (or NULL) * \param[out] error_output Where to store standard error output (or NULL) * * \note If output or error_output is not NULL, the caller is responsible for * freeing the memory. */ void stonith__action_result(stonith_action_t *action, int *rc, char **output, char **error_output) { if (rc) { *rc = pcmk_ok; } if (output) { *output = NULL; } if (error_output) { *error_output = NULL; } if (action != NULL) { if (rc) { *rc = action->rc; } if (output && action->output) { *output = action->output; action->output = NULL; // hand off memory management to caller } if (error_output && action->error) { *error_output = action->error; action->error = NULL; // hand off memory management to caller } } } #define FAILURE_MAX_RETRIES 2 stonith_action_t * stonith_action_create(const char *agent, const char *_action, const char *victim, uint32_t victim_nodeid, int timeout, GHashTable * device_args, GHashTable * port_map) { stonith_action_t *action; action = calloc(1, sizeof(stonith_action_t)); action->args = make_args(agent, _action, victim, victim_nodeid, device_args, port_map); crm_debug("Preparing '%s' action for %s using agent %s", _action, (victim? victim : "no target"), agent); action->agent = strdup(agent); action->action = strdup(_action); if (victim) { action->victim = strdup(victim); } action->timeout = action->remaining_timeout = timeout; action->max_retries = FAILURE_MAX_RETRIES; if (device_args) { char buffer[512]; const char *value = NULL; snprintf(buffer, sizeof(buffer), "pcmk_%s_retries", _action); value = g_hash_table_lookup(device_args, buffer); if (value) { action->max_retries = atoi(value); } } return action; } static gboolean update_remaining_timeout(stonith_action_t * action) { int diff = time(NULL) - action->initial_start_time; if (action->tries >= action->max_retries) { crm_info("Attempted to execute agent %s (%s) the maximum number of times (%d) allowed", action->agent, action->action, action->max_retries); action->remaining_timeout = 0; } else if ((action->rc != -ETIME) && diff < (action->timeout * 0.7)) { /* only set remaining timeout period if there is 30% * or greater of the original timeout period left */ action->remaining_timeout = action->timeout - diff; } else { action->remaining_timeout = 0; } return action->remaining_timeout ? TRUE : FALSE; } static int svc_action_to_errno(svc_action_t *svc_action) { int rv = pcmk_ok; if (svc_action->rc > 0) { /* Try to provide a useful error code based on the fence agent's * error output. */ if (svc_action->rc == PCMK_OCF_TIMEOUT) { rv = -ETIME; } else if (svc_action->stderr_data == NULL) { rv = -ENODATA; } else if (strstr(svc_action->stderr_data, "imed out")) { /* Some agents have their own internal timeouts */ rv = -ETIME; } else if (strstr(svc_action->stderr_data, "Unrecognised action")) { rv = -EOPNOTSUPP; } else { rv = -pcmk_err_generic; } } return rv; } static void stonith_action_async_done(svc_action_t *svc_action) { stonith_action_t *action = (stonith_action_t *) svc_action->cb_data; action->rc = svc_action_to_errno(svc_action); action->output = svc_action->stdout_data; svc_action->stdout_data = NULL; action->error = svc_action->stderr_data; svc_action->stderr_data = NULL; svc_action->params = NULL; crm_debug("Child process %d performing action '%s' exited with rc %d", action->pid, action->action, svc_action->rc); log_action(action, action->pid); if (action->rc != pcmk_ok && update_remaining_timeout(action)) { int rc = internal_stonith_action_execute(action); if (rc == pcmk_ok) { return; } } if (action->done_cb) { action->done_cb(action->pid, action->rc, action->output, action->userdata); } action->svc_action = NULL; // don't remove our caller stonith__destroy_action(action); } static void stonith_action_async_forked(svc_action_t *svc_action) { stonith_action_t *action = (stonith_action_t *) svc_action->cb_data; action->pid = svc_action->pid; action->svc_action = svc_action; if (action->fork_cb) { (action->fork_cb) (svc_action->pid, action->userdata); } crm_trace("Child process %d performing action '%s' successfully forked", action->pid, action->action); } static int internal_stonith_action_execute(stonith_action_t * action) { int rc = -EPROTO; int is_retry = 0; svc_action_t *svc_action = NULL; static int stonith_sequence = 0; char *buffer = NULL; if (!action->tries) { action->initial_start_time = time(NULL); } action->tries++; if (action->tries > 1) { crm_info("Attempt %d to execute %s (%s). remaining timeout is %d", action->tries, action->agent, action->action, action->remaining_timeout); is_retry = 1; } if (action->args == NULL || action->agent == NULL) goto fail; buffer = crm_strdup_printf(RH_STONITH_DIR "/%s", basename(action->agent)); svc_action = services_action_create_generic(buffer, NULL); free(buffer); svc_action->timeout = 1000 * action->remaining_timeout; svc_action->standard = strdup(PCMK_RESOURCE_CLASS_STONITH); svc_action->id = crm_strdup_printf("%s_%s_%d", basename(action->agent), action->action, action->tries); svc_action->agent = strdup(action->agent); svc_action->sequence = stonith_sequence++; svc_action->params = action->args; svc_action->cb_data = (void *) action; /* keep retries from executing out of control and free previous results */ if (is_retry) { free(action->output); action->output = NULL; free(action->error); action->error = NULL; sleep(1); } if (action->async) { /* async */ if(services_action_async_fork_notify(svc_action, &stonith_action_async_done, &stonith_action_async_forked) == FALSE) { services_action_free(svc_action); svc_action = NULL; } else { rc = 0; } } else { /* sync */ if (services_action_sync(svc_action)) { rc = 0; action->rc = svc_action_to_errno(svc_action); action->output = svc_action->stdout_data; svc_action->stdout_data = NULL; action->error = svc_action->stderr_data; svc_action->stderr_data = NULL; } else { action->rc = -ECONNABORTED; rc = action->rc; } svc_action->params = NULL; services_action_free(svc_action); } fail: return rc; } /*! * \internal * \brief Kick off execution of an async stonith action * * \param[in,out] action Action to be executed * \param[in,out] userdata Datapointer to be passed to callbacks * \param[in] done Callback to notify action has failed/succeeded * \param[in] fork_callback Callback to notify successful fork of child * * \return pcmk_ok if ownership of action has been taken, -errno otherwise */ int stonith_action_execute_async(stonith_action_t * action, void *userdata, void (*done) (GPid pid, int rc, const char *output, gpointer user_data), void (*fork_cb) (GPid pid, gpointer user_data)) { if (!action) { - return -1; + return -EINVAL; } action->userdata = userdata; action->done_cb = done; action->fork_cb = fork_cb; action->async = 1; return internal_stonith_action_execute(action); } /*! * \internal * \brief Execute a stonith action * * \param[in,out] action Action to execute * * \return pcmk_ok on success, -errno otherwise */ int stonith__execute(stonith_action_t *action) { int rc = pcmk_ok; CRM_CHECK(action != NULL, return -EINVAL); // Keep trying until success, max retries, or timeout do { rc = internal_stonith_action_execute(action); } while ((rc != pcmk_ok) && update_remaining_timeout(action)); return rc; } static int stonith_api_device_list(stonith_t * stonith, int call_options, const char *namespace, stonith_key_value_t ** devices, int timeout) { int count = 0; enum stonith_namespace ns = stonith_text2namespace(namespace); if (devices == NULL) { crm_err("Parameter error: stonith_api_device_list"); return -EFAULT; } #if HAVE_STONITH_STONITH_H // Include Linux-HA agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_lha)) { count += stonith__list_lha_agents(devices); } #endif // Include Red Hat agents if requested if ((ns == st_namespace_any) || (ns == st_namespace_rhcs)) { count += stonith__list_rhcs_agents(devices); } return count; } static int stonith_api_device_metadata(stonith_t * stonith, int call_options, const char *agent, const char *namespace, char **output, int timeout) { /* By executing meta-data directly, we can get it from stonith_admin when * the cluster is not running, which is important for higher-level tools. */ enum stonith_namespace ns = stonith_get_namespace(agent, namespace); crm_trace("Looking up metadata for %s agent %s", stonith_namespace2text(ns), agent); switch (ns) { case st_namespace_rhcs: return stonith__rhcs_metadata(agent, timeout, output); #if HAVE_STONITH_STONITH_H case st_namespace_lha: return stonith__lha_metadata(agent, timeout, output); #endif default: errno = EINVAL; crm_perror(LOG_ERR, "Agent %s not found or does not support meta-data", agent); break; } return -EINVAL; } static int stonith_api_query(stonith_t * stonith, int call_options, const char *target, stonith_key_value_t ** devices, int timeout) { int rc = 0, lpc = 0, max = 0; xmlNode *data = NULL; xmlNode *output = NULL; xmlXPathObjectPtr xpathObj = NULL; CRM_CHECK(devices != NULL, return -EINVAL); data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, target); crm_xml_add(data, F_STONITH_ACTION, "off"); rc = stonith_send_command(stonith, STONITH_OP_QUERY, data, &output, call_options, timeout); if (rc < 0) { return rc; } xpathObj = xpath_search(output, "//@agent"); if (xpathObj) { max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { xmlChar *match_path = xmlGetNodePath(match); crm_info("%s[%d] = %s", "//@agent", lpc, match_path); free(match_path); *devices = stonith_key_value_add(*devices, NULL, crm_element_value(match, XML_ATTR_ID)); } } freeXpathObject(xpathObj); } free_xml(output); free_xml(data); return max; } static int stonith_api_call(stonith_t * stonith, int call_options, const char *id, const char *action, const char *victim, int timeout, xmlNode ** output) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, F_STONITH_DEVICE); crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__); crm_xml_add(data, F_STONITH_DEVICE, id); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add(data, F_STONITH_TARGET, victim); rc = stonith_send_command(stonith, STONITH_OP_EXEC, data, output, call_options, timeout); free_xml(data); return rc; } static int stonith_api_list(stonith_t * stonith, int call_options, const char *id, char **list_info, int timeout) { int rc; xmlNode *output = NULL; rc = stonith_api_call(stonith, call_options, id, "list", NULL, timeout, &output); if (output && list_info) { const char *list_str; list_str = crm_element_value(output, "st_output"); if (list_str) { *list_info = strdup(list_str); } } if (output) { free_xml(output); } return rc; } static int stonith_api_monitor(stonith_t * stonith, int call_options, const char *id, int timeout) { return stonith_api_call(stonith, call_options, id, "monitor", NULL, timeout, NULL); } static int stonith_api_status(stonith_t * stonith, int call_options, const char *id, const char *port, int timeout) { return stonith_api_call(stonith, call_options, id, "status", port, timeout, NULL); } static int stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, node); crm_xml_add(data, F_STONITH_ACTION, action); crm_xml_add_int(data, F_STONITH_TIMEOUT, timeout); crm_xml_add_int(data, F_STONITH_TOLERANCE, tolerance); rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout); free_xml(data); return rc; } static int stonith_api_confirm(stonith_t * stonith, int call_options, const char *target) { return stonith_api_fence(stonith, call_options | st_opt_manual_ack, target, "off", 0, 0); } static int stonith_api_history(stonith_t * stonith, int call_options, const char *node, stonith_history_t ** history, int timeout) { int rc = 0; xmlNode *data = NULL; xmlNode *output = NULL; stonith_history_t *last = NULL; *history = NULL; if (node) { data = create_xml_node(NULL, __FUNCTION__); crm_xml_add(data, F_STONITH_TARGET, node); } rc = stonith_send_command(stonith, STONITH_OP_FENCE_HISTORY, data, &output, call_options | st_opt_sync_call, timeout); free_xml(data); if (rc == 0) { xmlNode *op = NULL; xmlNode *reply = get_xpath_object("//" F_STONITH_HISTORY_LIST, output, LOG_TRACE); for (op = __xml_first_child(reply); op != NULL; op = __xml_next(op)) { stonith_history_t *kvp; int completed; kvp = calloc(1, sizeof(stonith_history_t)); kvp->target = crm_element_value_copy(op, F_STONITH_TARGET); kvp->action = crm_element_value_copy(op, F_STONITH_ACTION); kvp->origin = crm_element_value_copy(op, F_STONITH_ORIGIN); kvp->delegate = crm_element_value_copy(op, F_STONITH_DELEGATE); kvp->client = crm_element_value_copy(op, F_STONITH_CLIENTNAME); crm_element_value_int(op, F_STONITH_DATE, &completed); kvp->completed = (time_t) completed; crm_element_value_int(op, F_STONITH_STATE, &kvp->state); if (last) { last->next = kvp; } else { *history = kvp; } last = kvp; } } free_xml(output); return rc; } void stonith_history_free(stonith_history_t *history) { stonith_history_t *hp, *hp_old; for (hp = history; hp; hp_old = hp, hp = hp->next, free(hp_old)) { free(hp->target); free(hp->action); free(hp->origin); free(hp->delegate); free(hp->client); } } /*! * \brief Deprecated (use stonith_get_namespace() instead) */ const char * get_stonith_provider(const char *agent, const char *provider) { return stonith_namespace2text(stonith_get_namespace(agent, provider)); } static gint stonithlib_GCompareFunc(gconstpointer a, gconstpointer b) { int rc = 0; const stonith_notify_client_t *a_client = a; const stonith_notify_client_t *b_client = b; CRM_CHECK(a_client->event != NULL && b_client->event != NULL, return 0); rc = strcmp(a_client->event, b_client->event); if (rc == 0) { if (a_client->notify == NULL || b_client->notify == NULL) { return 0; } else if (a_client->notify == b_client->notify) { return 0; } else if (((long)a_client->notify) < ((long)b_client->notify)) { crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return -1; } crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return 1; } return rc; } xmlNode * stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options) { xmlNode *op_msg = create_xml_node(NULL, "stonith_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "stonith_command"); crm_xml_add(op_msg, F_TYPE, T_STONITH_NG); crm_xml_add(op_msg, F_STONITH_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_STONITH_OPERATION, op); crm_xml_add_int(op_msg, F_STONITH_CALLID, call_id); crm_trace("Sending call options: %.8lx, %d", (long)call_options, call_options); crm_xml_add_int(op_msg, F_STONITH_CALLOPTS, call_options); if (data != NULL) { add_message_xml(op_msg, F_STONITH_CALLDATA, data); } return op_msg; } static void stonith_destroy_op_callback(gpointer data) { stonith_callback_client_t *blob = data; if (blob->timer && blob->timer->ref > 0) { g_source_remove(blob->timer->ref); } free(blob->timer); free(blob); } static int stonith_api_signoff(stonith_t * stonith) { stonith_private_t *native = stonith->st_private; crm_debug("Disconnecting from the fencer"); if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } free(native->token); native->token = NULL; stonith->state = stonith_disconnected; return pcmk_ok; } static int stonith_api_del_callback(stonith_t * stonith, int call_id, bool all_callbacks) { stonith_private_t *private = stonith->st_private; if (all_callbacks) { private->op_callback = NULL; g_hash_table_destroy(private->stonith_op_callback_table); private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal, NULL, stonith_destroy_op_callback); } else if (call_id == 0) { private->op_callback = NULL; } else { g_hash_table_remove(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); } return pcmk_ok; } static void invoke_callback(stonith_t * st, int call_id, int rc, void *userdata, void (*callback) (stonith_t * st, stonith_callback_data_t * data)) { stonith_callback_data_t data = { 0, }; data.call_id = call_id; data.rc = rc; data.userdata = userdata; callback(st, &data); } static void stonith_perform_callback(stonith_t * stonith, xmlNode * msg, int call_id, int rc) { stonith_private_t *private = NULL; stonith_callback_client_t *blob = NULL; stonith_callback_client_t local_blob; CRM_CHECK(stonith != NULL, return); CRM_CHECK(stonith->st_private != NULL, return); private = stonith->st_private; local_blob.id = NULL; local_blob.callback = NULL; local_blob.user_data = NULL; local_blob.only_success = FALSE; if (msg != NULL) { crm_element_value_int(msg, F_STONITH_RC, &rc); crm_element_value_int(msg, F_STONITH_CALLID, &call_id); } CRM_CHECK(call_id > 0, crm_log_xml_err(msg, "Bad result")); blob = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); if (blob != NULL) { local_blob = *blob; blob = NULL; stonith_api_del_callback(stonith, call_id, FALSE); } else { crm_trace("No callback found for call %d", call_id); local_blob.callback = NULL; } if (local_blob.callback != NULL && (rc == pcmk_ok || local_blob.only_success == FALSE)) { crm_trace("Invoking callback %s for call %d", crm_str(local_blob.id), call_id); invoke_callback(stonith, call_id, rc, local_blob.user_data, local_blob.callback); } else if (private->op_callback == NULL && rc != pcmk_ok) { crm_warn("Fencing command failed: %s", pcmk_strerror(rc)); crm_log_xml_debug(msg, "Failed fence update"); } if (private->op_callback != NULL) { crm_trace("Invoking global callback for call %d", call_id); invoke_callback(stonith, call_id, rc, NULL, private->op_callback); } crm_trace("OP callback activated."); } static gboolean stonith_async_timeout_handler(gpointer data) { struct timer_rec_s *timer = data; crm_err("Async call %d timed out after %dms", timer->call_id, timer->timeout); stonith_perform_callback(timer->stonith, NULL, timer->call_id, -ETIME); /* Always return TRUE, never remove the handler * We do that in stonith_del_callback() */ return TRUE; } static void set_callback_timeout(stonith_callback_client_t * callback, stonith_t * stonith, int call_id, int timeout) { struct timer_rec_s *async_timer = callback->timer; if (timeout <= 0) { return; } if (!async_timer) { async_timer = calloc(1, sizeof(struct timer_rec_s)); callback->timer = async_timer; } async_timer->stonith = stonith; async_timer->call_id = call_id; /* Allow a fair bit of grace to allow the server to tell us of a timeout * This is only a fallback */ async_timer->timeout = (timeout + 60) * 1000; if (async_timer->ref) { g_source_remove(async_timer->ref); } async_timer->ref = g_timeout_add(async_timer->timeout, stonith_async_timeout_handler, async_timer); } static void update_callback_timeout(int call_id, int timeout, stonith_t * st) { stonith_callback_client_t *callback = NULL; stonith_private_t *private = st->st_private; callback = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id)); if (!callback || !callback->allow_timeout_updates) { return; } set_callback_timeout(callback, st, call_id, timeout); } static int stonith_dispatch_internal(const char *buffer, ssize_t length, gpointer userdata) { const char *type = NULL; struct notify_blob_s blob; stonith_t *st = userdata; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; blob.stonith = st; blob.xml = string2xml(buffer); if (blob.xml == NULL) { crm_warn("Received malformed message from fencer: %s", buffer); return 0; } /* do callbacks */ type = crm_element_value(blob.xml, F_TYPE); crm_trace("Activating %s callbacks...", type); if (safe_str_eq(type, T_STONITH_NG)) { stonith_perform_callback(st, blob.xml, 0, 0); } else if (safe_str_eq(type, T_STONITH_NOTIFY)) { g_list_foreach(private->notify_list, stonith_send_notification, &blob); } else if (safe_str_eq(type, T_STONITH_TIMEOUT_VALUE)) { int call_id = 0; int timeout = 0; crm_element_value_int(blob.xml, F_STONITH_TIMEOUT, &timeout); crm_element_value_int(blob.xml, F_STONITH_CALLID, &call_id); update_callback_timeout(call_id, timeout, st); } else { crm_err("Unknown message type: %s", type); crm_log_xml_warn(blob.xml, "BadReply"); } free_xml(blob.xml); return 1; } static int stonith_api_signon(stonith_t * stonith, const char *name, int *stonith_fd) { int rc = pcmk_ok; - stonith_private_t *native = stonith->st_private; + stonith_private_t *native = NULL; + const char *display_name = name? name : "client"; static struct ipc_client_callbacks st_callbacks = { .dispatch = stonith_dispatch_internal, .destroy = stonith_connection_destroy }; - crm_trace("Connecting command channel"); + CRM_CHECK(stonith != NULL, return -EINVAL); + + native = stonith->st_private; + CRM_ASSERT(native != NULL); + + crm_debug("Attempting fencer connection by %s with%s mainloop", + display_name, (stonith_fd? "out" : "")); stonith->state = stonith_connected_command; if (stonith_fd) { /* No mainloop */ native->ipc = crm_ipc_new("stonith-ng", 0); if (native->ipc && crm_ipc_connect(native->ipc)) { *stonith_fd = crm_ipc_get_fd(native->ipc); } else if (native->ipc) { - crm_perror(LOG_ERR, "Connection to fencer failed"); - rc = -ENOTCONN; + crm_ipc_close(native->ipc); + crm_ipc_destroy(native->ipc); + native->ipc = NULL; } } else { /* With mainloop */ native->source = mainloop_add_ipc_client("stonith-ng", G_PRIORITY_MEDIUM, 0, stonith, &st_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { - crm_debug("Could not connect to the Stonith API"); rc = -ENOTCONN; - } - - if (rc == pcmk_ok) { + } else { xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "stonith_command"); crm_xml_add(hello, F_TYPE, T_STONITH_NG); crm_xml_add(hello, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_STONITH_CLIENTNAME, name); rc = crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply); if (rc < 0) { - crm_perror(LOG_DEBUG, "Couldn't complete registration with the fencing API: %d", rc); + crm_debug("Couldn't register with the fencer: %s " + CRM_XS " rc=%d", pcmk_strerror(rc), rc); rc = -ECOMM; } else if (reply == NULL) { - crm_err("Did not receive registration reply"); + crm_debug("Couldn't register with the fencer: no reply"); rc = -EPROTO; } else { const char *msg_type = crm_element_value(reply, F_STONITH_OPERATION); const char *tmp_ticket = crm_element_value(reply, F_STONITH_CLIENTID); if (safe_str_neq(msg_type, CRM_OP_REGISTER)) { - crm_err("Invalid registration message: %s", msg_type); - crm_log_xml_err(reply, "Bad reply"); + crm_debug("Couldn't register with the fencer: invalid reply type '%s'", + (msg_type? msg_type : "(missing)")); + crm_log_xml_debug(reply, "Invalid fencer reply"); rc = -EPROTO; } else if (tmp_ticket == NULL) { - crm_err("No registration token provided"); - crm_log_xml_err(reply, "Bad reply"); + crm_debug("Couldn't register with the fencer: no token in reply"); + crm_log_xml_debug(reply, "Invalid fencer reply"); rc = -EPROTO; } else { - crm_trace("Obtained registration token: %s", tmp_ticket); native->token = strdup(tmp_ticket); +#if HAVE_MSGFROMIPC_TIMEOUT + stonith->call_timeout = MAX_IPC_DELAY; +#endif + crm_debug("Connection to fencer by %s succeeded (registration token: %s)", + display_name, native->token); rc = pcmk_ok; } } free_xml(reply); free_xml(hello); } - if (rc == pcmk_ok) { -#if HAVE_MSGFROMIPC_TIMEOUT - stonith->call_timeout = MAX_IPC_DELAY; -#endif - crm_debug("Connection to fencer successful"); - return pcmk_ok; + if (rc != pcmk_ok) { + crm_debug("Connection attempt to fencer by %s failed: %s " + CRM_XS " rc=%d", display_name, pcmk_strerror(rc), rc); + stonith->cmds->disconnect(stonith); } - - crm_debug("Connection to fencer failed: %s", pcmk_strerror(rc)); - stonith->cmds->disconnect(stonith); return rc; } static int stonith_set_notification(stonith_t * stonith, const char *callback, int enabled) { int rc = pcmk_ok; xmlNode *notify_msg = create_xml_node(NULL, __FUNCTION__); stonith_private_t *native = stonith->st_private; if (stonith->state != stonith_disconnected) { crm_xml_add(notify_msg, F_STONITH_OPERATION, T_STONITH_NOTIFY); if (enabled) { crm_xml_add(notify_msg, F_STONITH_NOTIFY_ACTIVATE, callback); } else { crm_xml_add(notify_msg, F_STONITH_NOTIFY_DEACTIVATE, callback); } rc = crm_ipc_send(native->ipc, notify_msg, crm_ipc_client_response, -1, NULL); if (rc < 0) { crm_perror(LOG_DEBUG, "Couldn't register for fencing notifications: %d", rc); rc = -ECOMM; } else { rc = pcmk_ok; } } free_xml(notify_msg); return rc; } static int stonith_api_add_notification(stonith_t * stonith, const char *event, void (*callback) (stonith_t * stonith, stonith_event_t * e)) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = NULL; private = stonith->st_private; crm_trace("Adding callback for %s events (%d)", event, g_list_length(private->notify_list)); new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = callback; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); if (list_item != NULL) { crm_warn("Callback already present"); free(new_client); return -ENOTUNIQ; } else { private->notify_list = g_list_append(private->notify_list, new_client); stonith_set_notification(stonith, event, 1); crm_trace("Callback added (%d)", g_list_length(private->notify_list)); } return pcmk_ok; } static int stonith_api_del_notification(stonith_t * stonith, const char *event) { GList *list_item = NULL; stonith_notify_client_t *new_client = NULL; stonith_private_t *private = NULL; crm_debug("Removing callback for %s events", event); private = stonith->st_private; new_client = calloc(1, sizeof(stonith_notify_client_t)); new_client->event = event; new_client->notify = NULL; list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc); stonith_set_notification(stonith, event, 0); if (list_item != NULL) { stonith_notify_client_t *list_client = list_item->data; private->notify_list = g_list_remove(private->notify_list, list_client); free(list_client); crm_trace("Removed callback"); } else { crm_trace("Callback not present"); } free(new_client); return pcmk_ok; } static int stonith_api_add_callback(stonith_t * stonith, int call_id, int timeout, int options, void *user_data, const char *callback_name, void (*callback) (stonith_t * st, stonith_callback_data_t * data)) { stonith_callback_client_t *blob = NULL; stonith_private_t *private = NULL; CRM_CHECK(stonith != NULL, return -EINVAL); CRM_CHECK(stonith->st_private != NULL, return -EINVAL); private = stonith->st_private; if (call_id == 0) { private->op_callback = callback; } else if (call_id < 0) { if (!(options & st_opt_report_only_success)) { crm_trace("Call failed, calling %s: %s", callback_name, pcmk_strerror(call_id)); invoke_callback(stonith, call_id, call_id, user_data, callback); } else { crm_warn("Fencer call failed: %s", pcmk_strerror(call_id)); } return FALSE; } blob = calloc(1, sizeof(stonith_callback_client_t)); blob->id = callback_name; blob->only_success = (options & st_opt_report_only_success) ? TRUE : FALSE; blob->user_data = user_data; blob->callback = callback; blob->allow_timeout_updates = (options & st_opt_timeout_updates) ? TRUE : FALSE; if (timeout > 0) { set_callback_timeout(blob, stonith, call_id, timeout); } g_hash_table_insert(private->stonith_op_callback_table, GINT_TO_POINTER(call_id), blob); crm_trace("Added callback to %s for call %d", callback_name, call_id); return TRUE; } static void stonith_dump_pending_op(gpointer key, gpointer value, gpointer user_data) { int call = GPOINTER_TO_INT(key); stonith_callback_client_t *blob = value; crm_debug("Call %d (%s): pending", call, crm_str(blob->id)); } void stonith_dump_pending_callbacks(stonith_t * stonith) { stonith_private_t *private = stonith->st_private; if (private->stonith_op_callback_table == NULL) { return; } return g_hash_table_foreach(private->stonith_op_callback_table, stonith_dump_pending_op, NULL); } /* */ static stonith_event_t * xml_to_event(xmlNode * msg) { stonith_event_t *event = calloc(1, sizeof(stonith_event_t)); const char *ntype = crm_element_value(msg, F_SUBTYPE); char *data_addr = crm_strdup_printf("//%s", ntype); xmlNode *data = get_xpath_object(data_addr, msg, LOG_DEBUG); crm_log_xml_trace(msg, "stonith_notify"); crm_element_value_int(msg, F_STONITH_RC, &(event->result)); if (safe_str_eq(ntype, T_STONITH_NOTIFY_FENCE)) { event->operation = crm_element_value_copy(msg, F_STONITH_OPERATION); if (data) { event->origin = crm_element_value_copy(data, F_STONITH_ORIGIN); event->action = crm_element_value_copy(data, F_STONITH_ACTION); event->target = crm_element_value_copy(data, F_STONITH_TARGET); event->executioner = crm_element_value_copy(data, F_STONITH_DELEGATE); event->id = crm_element_value_copy(data, F_STONITH_REMOTE_OP_ID); event->client_origin = crm_element_value_copy(data, F_STONITH_CLIENTNAME); event->device = crm_element_value_copy(data, F_STONITH_DEVICE); } else { crm_err("No data for %s event", ntype); crm_log_xml_notice(msg, "BadEvent"); } } free(data_addr); return event; } static void event_free(stonith_event_t * event) { free(event->id); free(event->type); free(event->message); free(event->operation); free(event->origin); free(event->action); free(event->target); free(event->executioner); free(event->device); free(event->client_origin); free(event); } static void stonith_send_notification(gpointer data, gpointer user_data) { struct notify_blob_s *blob = user_data; stonith_notify_client_t *entry = data; stonith_event_t *st_event = NULL; const char *event = NULL; if (blob->xml == NULL) { crm_warn("Skipping callback - NULL message"); return; } event = crm_element_value(blob->xml, F_SUBTYPE); if (entry == NULL) { crm_warn("Skipping callback - NULL callback client"); return; } else if (entry->notify == NULL) { crm_warn("Skipping callback - NULL callback"); return; } else if (safe_str_neq(entry->event, event)) { crm_trace("Skipping callback - event mismatch %p/%s vs. %s", entry, entry->event, event); return; } st_event = xml_to_event(blob->xml); crm_trace("Invoking callback for %p/%s event...", entry, event); entry->notify(blob->stonith, st_event); crm_trace("Callback invoked..."); event_free(st_event); } int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout) { int rc = 0; int reply_id = -1; enum crm_ipc_flags ipc_flags = crm_ipc_flags_none; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; stonith_private_t *native = stonith->st_private; if (stonith->state == stonith_disconnected) { return -ENOTCONN; } if (output_data != NULL) { *output_data = NULL; } if (op == NULL) { crm_err("No operation specified"); return -EINVAL; } if (call_options & st_opt_sync_call) { ipc_flags |= crm_ipc_client_response; } stonith->call_id++; /* prevent call_id from being negative (or zero) and conflicting * with the stonith_errors enum * use 2 because we use it as (stonith->call_id - 1) below */ if (stonith->call_id < 1) { stonith->call_id = 1; } CRM_CHECK(native->token != NULL,; ); op_msg = stonith_create_op(stonith->call_id, native->token, op, data, call_options); if (op_msg == NULL) { return -EINVAL; } crm_xml_add_int(op_msg, F_STONITH_TIMEOUT, timeout); crm_trace("Sending %s message to fencer with timeout %ds", op, timeout); rc = crm_ipc_send(native->ipc, op_msg, ipc_flags, 1000 * (timeout + 60), &op_reply); free_xml(op_msg); if (rc < 0) { crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%ds): %d", op, timeout, rc); rc = -ECOMM; goto done; } crm_log_xml_trace(op_reply, "Reply"); if (!(call_options & st_opt_sync_call)) { crm_trace("Async call %d, returning", stonith->call_id); CRM_CHECK(stonith->call_id != 0, return -EPROTO); free_xml(op_reply); return stonith->call_id; } rc = pcmk_ok; crm_element_value_int(op_reply, F_STONITH_CALLID, &reply_id); if (reply_id == stonith->call_id) { crm_trace("Synchronous reply %d received", reply_id); if (crm_element_value_int(op_reply, F_STONITH_RC, &rc) != 0) { rc = -ENOMSG; } if ((call_options & st_opt_discard_reply) || output_data == NULL) { crm_trace("Discarding reply"); } else { *output_data = op_reply; op_reply = NULL; /* Prevent subsequent free */ } } else if (reply_id <= 0) { crm_err("Received bad reply: No id set"); crm_log_xml_err(op_reply, "Bad reply"); free_xml(op_reply); rc = -ENOMSG; } else { crm_err("Received bad reply: %d (wanted %d)", reply_id, stonith->call_id); crm_log_xml_err(op_reply, "Old reply"); free_xml(op_reply); rc = -ENOMSG; } done: if (crm_ipc_connected(native->ipc) == FALSE) { crm_err("Fencer disconnected"); stonith->state = stonith_disconnected; } free_xml(op_reply); return rc; } /* Not used with mainloop */ bool stonith_dispatch(stonith_t * st) { gboolean stay_connected = TRUE; stonith_private_t *private = NULL; CRM_ASSERT(st != NULL); private = st->st_private; while (crm_ipc_ready(private->ipc)) { if (crm_ipc_read(private->ipc) > 0) { const char *msg = crm_ipc_buffer(private->ipc); stonith_dispatch_internal(msg, strlen(msg), st); } if (crm_ipc_connected(private->ipc) == FALSE) { crm_err("Connection closed"); stay_connected = FALSE; } } return stay_connected; } static int stonith_api_free(stonith_t * stonith) { int rc = pcmk_ok; crm_trace("Destroying %p", stonith); if (stonith->state != stonith_disconnected) { crm_trace("Disconnecting %p first", stonith); rc = stonith->cmds->disconnect(stonith); } if (stonith->state == stonith_disconnected) { stonith_private_t *private = stonith->st_private; crm_trace("Removing %d callbacks", g_hash_table_size(private->stonith_op_callback_table)); g_hash_table_destroy(private->stonith_op_callback_table); crm_trace("Destroying %d notification clients", g_list_length(private->notify_list)); g_list_free_full(private->notify_list, free); free(stonith->st_private); free(stonith->cmds); free(stonith); } else { crm_err("Not free'ing active connection: %s (%d)", pcmk_strerror(rc), rc); } return rc; } void stonith_api_delete(stonith_t * stonith) { crm_trace("Destroying %p", stonith); if(stonith) { stonith->cmds->free(stonith); } } static int stonith_api_validate(stonith_t *st, int call_options, const char *rsc_id, const char *namespace_s, const char *agent, stonith_key_value_t *params, int timeout, char **output, char **error_output) { /* Validation should be done directly via the agent, so we can get it from * stonith_admin when the cluster is not running, which is important for * higher-level tools. */ int rc = pcmk_ok; /* Use a dummy node name in case the agent requires a target. We assume the * actual target doesn't matter for validation purposes (if in practice, * that is incorrect, we will need to allow the caller to pass the target). */ const char *target = "node1"; GHashTable *params_table = crm_str_table_new(); // Convert parameter list to a hash table for (; params; params = params->next) { // Strip out Pacemaker-implemented parameters if (!crm_starts_with(params->key, "pcmk_") && strcmp(params->key, "provides") && strcmp(params->key, "stonith-timeout")) { g_hash_table_insert(params_table, strdup(params->key), strdup(params->value)); } } #if SUPPORT_CIBSECRETS rc = replace_secret_params(rsc_id, params_table); if (rc < 0) { crm_warn("Could not replace secret parameters for validation of %s: %s", agent, pcmk_strerror(rc)); } #endif if (output) { *output = NULL; } if (error_output) { *error_output = NULL; } switch (stonith_get_namespace(agent, namespace_s)) { case st_namespace_rhcs: rc = stonith__rhcs_validate(st, call_options, target, agent, params_table, timeout, output, error_output); break; #if HAVE_STONITH_STONITH_H case st_namespace_lha: rc = stonith__lha_validate(st, call_options, target, agent, params_table, timeout, output, error_output); break; #endif default: rc = -EINVAL; errno = EINVAL; crm_perror(LOG_ERR, "Agent %s not found or does not support validation", agent); break; } g_hash_table_destroy(params_table); return rc; } stonith_t * stonith_api_new(void) { stonith_t *new_stonith = NULL; stonith_private_t *private = NULL; new_stonith = calloc(1, sizeof(stonith_t)); private = calloc(1, sizeof(stonith_private_t)); new_stonith->st_private = private; private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal, NULL, stonith_destroy_op_callback); private->notify_list = NULL; new_stonith->call_id = 1; new_stonith->state = stonith_disconnected; new_stonith->cmds = calloc(1, sizeof(stonith_api_operations_t)); /* *INDENT-OFF* */ new_stonith->cmds->free = stonith_api_free; new_stonith->cmds->connect = stonith_api_signon; new_stonith->cmds->disconnect = stonith_api_signoff; new_stonith->cmds->list = stonith_api_list; new_stonith->cmds->monitor = stonith_api_monitor; new_stonith->cmds->status = stonith_api_status; new_stonith->cmds->fence = stonith_api_fence; new_stonith->cmds->confirm = stonith_api_confirm; new_stonith->cmds->history = stonith_api_history; new_stonith->cmds->list_agents = stonith_api_device_list; new_stonith->cmds->metadata = stonith_api_device_metadata; new_stonith->cmds->query = stonith_api_query; new_stonith->cmds->remove_device = stonith_api_remove_device; new_stonith->cmds->register_device = stonith_api_register_device; new_stonith->cmds->remove_level = stonith_api_remove_level; new_stonith->cmds->remove_level_full = stonith_api_remove_level_full; new_stonith->cmds->register_level = stonith_api_register_level; new_stonith->cmds->register_level_full = stonith_api_register_level_full; new_stonith->cmds->remove_callback = stonith_api_del_callback; new_stonith->cmds->register_callback = stonith_api_add_callback; new_stonith->cmds->remove_notification = stonith_api_del_notification; new_stonith->cmds->register_notification = stonith_api_add_notification; new_stonith->cmds->validate = stonith_api_validate; /* *INDENT-ON* */ return new_stonith; } +/*! + * \brief Make a blocking connection attempt to the fencer + * + * \param[in,out] st Fencer API object + * \param[in] name Client name to use with fencer + * \param[in] max_attempts Return error if this many attempts fail + * + * \return pcmk_ok on success, result of last attempt otherwise + */ +int +stonith_api_connect_retry(stonith_t *st, const char *name, int max_attempts) +{ + int rc = -EINVAL; // if max_attempts is not positive + + for (int attempt = 1; attempt <= max_attempts; attempt++) { + rc = st->cmds->connect(st, name, NULL); + if (rc == pcmk_ok) { + return pcmk_ok; + } else if (attempt < max_attempts) { + crm_notice("Fencer connection attempt %d of %d failed (retrying in 2s): %s " + CRM_XS " rc=%d", + attempt, max_attempts, pcmk_strerror(rc), rc); + sleep(2); + } + } + crm_notice("Could not connect to fencer: %s " CRM_XS " rc=%d", + pcmk_strerror(rc), rc); + return rc; +} + stonith_key_value_t * stonith_key_value_add(stonith_key_value_t * head, const char *key, const char *value) { stonith_key_value_t *p, *end; p = calloc(1, sizeof(stonith_key_value_t)); if (key) { p->key = strdup(key); } if (value) { p->value = strdup(value); } end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void stonith_key_value_freeall(stonith_key_value_t * head, int keys, int values) { stonith_key_value_t *p; while (head) { p = head->next; if (keys) { free(head->key); } if (values) { free(head->value); } free(head); head = p; } } #define api_log_open() openlog("stonith-api", LOG_CONS | LOG_NDELAY | LOG_PID, LOG_DAEMON) #define api_log(level, fmt, args...) syslog(level, "%s: "fmt, __FUNCTION__, args) int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off) { - char *name = NULL; - const char *action = "reboot"; - - int rc = -EPROTO; - stonith_t *st = NULL; - enum stonith_call_options opts = st_opt_sync_call | st_opt_allow_suicide; + int rc = pcmk_ok; + stonith_t *st = stonith_api_new(); + const char *action = off? "off" : "reboot"; api_log_open(); - st = stonith_api_new(); - if (st) { - rc = st->cmds->connect(st, "stonith-api", NULL); - if(rc != pcmk_ok) { - api_log(LOG_ERR, "Connection failed, could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); - } + if (st == NULL) { + api_log(LOG_ERR, "API initialization failed, could not kick (%s) node %u/%s", + action, nodeid, uname); + return -EPROTO; } - if (uname != NULL) { - name = strdup(uname); - - } else if (nodeid > 0) { - opts |= st_opt_cs_nodeid; - name = crm_itoa(nodeid); - } - - if (off) { - action = "off"; - } - - if (rc == pcmk_ok) { + rc = st->cmds->connect(st, "stonith-api", NULL); + if (rc != pcmk_ok) { + api_log(LOG_ERR, "Connection failed, could not kick (%s) node %u/%s : %s (%d)", + action, nodeid, uname, pcmk_strerror(rc), rc); + } else { + char *name = NULL; + enum stonith_call_options opts = st_opt_sync_call | st_opt_allow_suicide; + + if (uname != NULL) { + name = strdup(uname); + } else if (nodeid > 0) { + opts |= st_opt_cs_nodeid; + name = crm_itoa(nodeid); + } rc = st->cmds->fence(st, opts, name, action, timeout, 0); - if(rc != pcmk_ok) { - api_log(LOG_ERR, "Could not kick (%s) node %u/%s : %s (%d)", action, nodeid, uname, pcmk_strerror(rc), rc); + free(name); + + if (rc != pcmk_ok) { + api_log(LOG_ERR, "Could not kick (%s) node %u/%s : %s (%d)", + action, nodeid, uname, pcmk_strerror(rc), rc); } else { - api_log(LOG_NOTICE, "Node %u/%s kicked: %s ", nodeid, uname, action); + api_log(LOG_NOTICE, "Node %u/%s kicked: %s", nodeid, uname, action); } } - if (st) { - st->cmds->disconnect(st); - stonith_api_delete(st); - } - - free(name); + stonith_api_delete(st); return rc; } time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress) { - int rc = 0; - char *name = NULL; - + int rc = pcmk_ok; time_t when = 0; - stonith_t *st = NULL; + stonith_t *st = stonith_api_new(); stonith_history_t *history = NULL, *hp = NULL; - enum stonith_call_options opts = st_opt_sync_call; - - st = stonith_api_new(); - if (st) { - rc = st->cmds->connect(st, "stonith-api", NULL); - if(rc != pcmk_ok) { - api_log(LOG_NOTICE, "Connection failed: %s (%d)", pcmk_strerror(rc), rc); - } - } - - if (uname != NULL) { - name = strdup(uname); - } else if (nodeid > 0) { - opts |= st_opt_cs_nodeid; - name = crm_itoa(nodeid); + if (st == NULL) { + api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: " + "API initialization failed", nodeid, uname); + return when; } - if (st && rc == pcmk_ok) { + rc = st->cmds->connect(st, "stonith-api", NULL); + if (rc != pcmk_ok) { + api_log(LOG_NOTICE, "Connection failed: %s (%d)", pcmk_strerror(rc), rc); + } else { int entries = 0; int progress = 0; int completed = 0; - + char *name = NULL; + enum stonith_call_options opts = st_opt_sync_call; + + if (uname != NULL) { + name = strdup(uname); + } else if (nodeid > 0) { + opts |= st_opt_cs_nodeid; + name = crm_itoa(nodeid); + } rc = st->cmds->history(st, opts, name, &history, 120); + free(name); for (hp = history; hp; hp = hp->next) { entries++; if (in_progress) { progress++; if (hp->state != st_done && hp->state != st_failed) { when = time(NULL); } } else if (hp->state == st_done) { completed++; if (hp->completed > when) { when = hp->completed; } } } stonith_history_free(history); if(rc == pcmk_ok) { api_log(LOG_INFO, "Found %d entries for %u/%s: %d in progress, %d completed", entries, nodeid, uname, progress, completed); } else { api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: %s (%d)", nodeid, uname, pcmk_strerror(rc), rc); } } - if (st) { - st->cmds->disconnect(st); - stonith_api_delete(st); - } + stonith_api_delete(st); if(when) { api_log(LOG_INFO, "Node %u/%s last kicked at: %ld", nodeid, uname, (long int)when); } - free(name); return when; } bool stonith_agent_exists(const char *agent, int timeout) { stonith_t *st = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; bool rc = FALSE; if (agent == NULL) { return rc; } st = stonith_api_new(); st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout == 0 ? 120 : timeout); for (dIter = devices; dIter != NULL; dIter = dIter->next) { if (crm_str_eq(dIter->value, agent, TRUE)) { rc = TRUE; break; } } stonith_key_value_freeall(devices, 1, 1); stonith_api_delete(st); return rc; } const char * stonith_action_str(const char *action) { if (action == NULL) { return "fencing"; } else if (!strcmp(action, "on")) { return "unfencing"; } else if (!strcmp(action, "off")) { return "turning off"; } else { return action; } } diff --git a/lib/pengine/common.c b/lib/pengine/common.c index 9513633d39..3a283b4af9 100644 --- a/lib/pengine/common.c +++ b/lib/pengine/common.c @@ -1,468 +1,474 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include gboolean was_processing_error = FALSE; gboolean was_processing_warning = FALSE; static gboolean check_health(const char *value) { if (safe_str_eq(value, "none")) { return TRUE; } else if (safe_str_eq(value, "custom")) { return TRUE; } else if (safe_str_eq(value, "only-green")) { return TRUE; } else if (safe_str_eq(value, "progressive")) { return TRUE; } else if (safe_str_eq(value, "migrate-on-red")) { return TRUE; } return FALSE; } static gboolean check_stonith_action(const char *value) { if (safe_str_eq(value, "reboot")) { return TRUE; } else if (safe_str_eq(value, "poweroff")) { return TRUE; } else if (safe_str_eq(value, "off")) { return TRUE; } return FALSE; } static gboolean check_placement_strategy(const char *value) { if (safe_str_eq(value, "default")) { return TRUE; } else if (safe_str_eq(value, "utilization")) { return TRUE; } else if (safe_str_eq(value, "minimal")) { return TRUE; } else if (safe_str_eq(value, "balanced")) { return TRUE; } return FALSE; } /* *INDENT-OFF* */ static pe_cluster_option pe_opts[] = { /* name, old-name, validate, default, description */ { "no-quorum-policy", NULL, "enum", "stop, freeze, ignore, suicide", "stop", &check_quorum, "What to do when the cluster does not have quorum", NULL }, { "symmetric-cluster", NULL, "boolean", NULL, "true", &check_boolean, "All resources can run anywhere by default", NULL }, { "maintenance-mode", NULL, "boolean", NULL, "false", &check_boolean, "Should the cluster monitor resources and start/stop them as required", NULL }, { "start-failure-is-fatal", NULL, "boolean", NULL, "true", &check_boolean, "Always treat start failures as fatal", "When set to TRUE, the cluster will immediately ban a resource from a node if it fails to start there. When FALSE, the cluster will instead check the resource's fail count against its migration-threshold." }, { "enable-startup-probes", NULL, "boolean", NULL, "true", &check_boolean, "Should the cluster check for active resources during startup", NULL }, /* Stonith Options */ { "stonith-enabled", NULL, "boolean", NULL, "true", &check_boolean, "Failed nodes are STONITH'd", NULL }, { "stonith-action", NULL, "enum", "reboot, off, poweroff", "reboot", &check_stonith_action, "Action to send to STONITH device ('poweroff' is a deprecated alias for 'off')", NULL }, { "stonith-timeout", NULL, "time", NULL, "60s", &check_timer, "How long to wait for the STONITH action (reboot,on,off) to complete", NULL }, { XML_ATTR_HAVE_WATCHDOG, NULL, "boolean", NULL, "false", &check_boolean, "Enable watchdog integration", "Set automatically by the cluster if SBD is detected. User configured values are ignored." }, - { "concurrent-fencing", NULL, "boolean", NULL, "false", &check_boolean, + { "concurrent-fencing", NULL, "boolean", NULL, +#ifdef DEFAULT_CONCURRENT_FENCING_TRUE + "true", +#else + "false", +#endif + &check_boolean, "Allow performing fencing operations in parallel", NULL }, { "startup-fencing", NULL, "boolean", NULL, "true", &check_boolean, "STONITH unseen nodes", "Advanced Use Only! Not using the default is very unsafe!" }, /* Timeouts etc */ { "cluster-delay", NULL, "time", NULL, "60s", &check_time, "Round trip delay over the network (excluding action execution)", "The \"correct\" value will depend on the speed and load of your network and cluster nodes." }, { "batch-limit", NULL, "integer", NULL, "0", &check_number, "The number of jobs that the TE is allowed to execute in parallel", "The \"correct\" value will depend on the speed and load of your network and cluster nodes." }, { "migration-limit", NULL, "integer", NULL, "-1", &check_number, "The number of migration jobs that the TE is allowed to execute in parallel on a node"}, /* Orphans and stopping */ { "stop-all-resources", NULL, "boolean", NULL, "false", &check_boolean, "Should the cluster stop all active resources (except those needed for fencing)", NULL }, { "stop-orphan-resources", NULL, "boolean", NULL, "true", &check_boolean, "Should deleted resources be stopped", NULL }, { "stop-orphan-actions", NULL, "boolean", NULL, "true", &check_boolean, "Should deleted actions be cancelled", NULL }, { "remove-after-stop", NULL, "boolean", NULL, "false", &check_boolean, "Remove resources from the executor after they are stopped", "Always set this to false. Other values are, at best, poorly tested and potentially dangerous." }, /* { "", "", , "0", "", NULL }, */ /* Storing inputs */ { "pe-error-series-max", NULL, "integer", NULL, "-1", &check_number, "The number of scheduler inputs resulting in ERRORs to save", "Zero to disable, -1 to store unlimited" }, { "pe-warn-series-max", NULL, "integer", NULL, "5000", &check_number, "The number of scheduler inputs resulting in WARNINGs to save", "Zero to disable, -1 to store unlimited" }, { "pe-input-series-max", NULL, "integer", NULL, "4000", &check_number, "The number of other scheduler inputs to save", "Zero to disable, -1 to store unlimited" }, /* Node health */ { "node-health-strategy", NULL, "enum", "none, migrate-on-red, only-green, progressive, custom", "none", &check_health, "The strategy combining node attributes to determine overall node health.", "Requires external entities to create node attributes (named with the prefix '#health') with values: 'red', 'yellow' or 'green'."}, { "node-health-base", NULL, "integer", NULL, "0", &check_number, "The base score assigned to a node", "Only used when node-health-strategy is set to progressive." }, { "node-health-green", NULL, "integer", NULL, "0", &check_number, "The score 'green' translates to in rsc_location constraints", "Only used when node-health-strategy is set to custom or progressive." }, { "node-health-yellow", NULL, "integer", NULL, "0", &check_number, "The score 'yellow' translates to in rsc_location constraints", "Only used when node-health-strategy is set to custom or progressive." }, { "node-health-red", NULL, "integer", NULL, "-INFINITY", &check_number, "The score 'red' translates to in rsc_location constraints", "Only used when node-health-strategy is set to custom or progressive." }, /*Placement Strategy*/ { "placement-strategy", NULL, "enum", "default, utilization, minimal, balanced", "default", &check_placement_strategy, "The strategy to determine resource placement", NULL}, }; /* *INDENT-ON* */ void pe_metadata(void) { config_metadata("pacemaker-schedulerd", "1.0", "scheduler properties", "Cluster properties used by Pacemaker's scheduler," " formerly known as pengine", pe_opts, DIMOF(pe_opts)); } void verify_pe_options(GHashTable * options) { verify_all_options(options, pe_opts, DIMOF(pe_opts)); } const char * pe_pref(GHashTable * options, const char *name) { return get_cluster_pref(options, pe_opts, DIMOF(pe_opts), name); } const char * fail2text(enum action_fail_response fail) { const char *result = ""; switch (fail) { case action_fail_ignore: result = "ignore"; break; case action_fail_block: result = "block"; break; case action_fail_recover: result = "recover"; break; case action_fail_migrate: result = "migrate"; break; case action_fail_stop: result = "stop"; break; case action_fail_fence: result = "fence"; break; case action_fail_standby: result = "standby"; break; case action_fail_restart_container: result = "restart-container"; break; case action_fail_reset_remote: result = "reset-remote"; break; } return result; } enum action_tasks text2task(const char *task) { if (safe_str_eq(task, CRMD_ACTION_STOP)) { return stop_rsc; } else if (safe_str_eq(task, CRMD_ACTION_STOPPED)) { return stopped_rsc; } else if (safe_str_eq(task, CRMD_ACTION_START)) { return start_rsc; } else if (safe_str_eq(task, CRMD_ACTION_STARTED)) { return started_rsc; } else if (safe_str_eq(task, CRM_OP_SHUTDOWN)) { return shutdown_crm; } else if (safe_str_eq(task, CRM_OP_FENCE)) { return stonith_node; } else if (safe_str_eq(task, CRMD_ACTION_STATUS)) { return monitor_rsc; } else if (safe_str_eq(task, CRMD_ACTION_NOTIFY)) { return action_notify; } else if (safe_str_eq(task, CRMD_ACTION_NOTIFIED)) { return action_notified; } else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) { return action_promote; } else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) { return action_demote; } else if (safe_str_eq(task, CRMD_ACTION_PROMOTED)) { return action_promoted; } else if (safe_str_eq(task, CRMD_ACTION_DEMOTED)) { return action_demoted; } #if SUPPORT_TRACING if (safe_str_eq(task, CRMD_ACTION_CANCEL)) { return no_action; } else if (safe_str_eq(task, CRMD_ACTION_DELETE)) { return no_action; } else if (safe_str_eq(task, CRMD_ACTION_STATUS)) { return no_action; } else if (safe_str_eq(task, CRM_OP_PROBED)) { return no_action; } else if (safe_str_eq(task, CRM_OP_LRM_REFRESH)) { return no_action; } else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) { return no_action; } else if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) { return no_action; } crm_trace("Unsupported action: %s", task); #endif return no_action; } const char * task2text(enum action_tasks task) { const char *result = ""; switch (task) { case no_action: result = "no_action"; break; case stop_rsc: result = CRMD_ACTION_STOP; break; case stopped_rsc: result = CRMD_ACTION_STOPPED; break; case start_rsc: result = CRMD_ACTION_START; break; case started_rsc: result = CRMD_ACTION_STARTED; break; case shutdown_crm: result = CRM_OP_SHUTDOWN; break; case stonith_node: result = CRM_OP_FENCE; break; case monitor_rsc: result = CRMD_ACTION_STATUS; break; case action_notify: result = CRMD_ACTION_NOTIFY; break; case action_notified: result = CRMD_ACTION_NOTIFIED; break; case action_promote: result = CRMD_ACTION_PROMOTE; break; case action_promoted: result = CRMD_ACTION_PROMOTED; break; case action_demote: result = CRMD_ACTION_DEMOTE; break; case action_demoted: result = CRMD_ACTION_DEMOTED; break; } return result; } const char * role2text(enum rsc_role_e role) { switch (role) { case RSC_ROLE_UNKNOWN: return RSC_ROLE_UNKNOWN_S; case RSC_ROLE_STOPPED: return RSC_ROLE_STOPPED_S; case RSC_ROLE_STARTED: return RSC_ROLE_STARTED_S; case RSC_ROLE_SLAVE: return RSC_ROLE_SLAVE_S; case RSC_ROLE_MASTER: return RSC_ROLE_MASTER_S; } CRM_CHECK(role >= RSC_ROLE_UNKNOWN, return RSC_ROLE_UNKNOWN_S); CRM_CHECK(role < RSC_ROLE_MAX, return RSC_ROLE_UNKNOWN_S); // coverity[dead_error_line] return RSC_ROLE_UNKNOWN_S; } enum rsc_role_e text2role(const char *role) { CRM_ASSERT(role != NULL); if (safe_str_eq(role, RSC_ROLE_STOPPED_S)) { return RSC_ROLE_STOPPED; } else if (safe_str_eq(role, RSC_ROLE_STARTED_S)) { return RSC_ROLE_STARTED; } else if (safe_str_eq(role, RSC_ROLE_SLAVE_S)) { return RSC_ROLE_SLAVE; } else if (safe_str_eq(role, RSC_ROLE_MASTER_S)) { return RSC_ROLE_MASTER; } else if (safe_str_eq(role, RSC_ROLE_UNKNOWN_S)) { return RSC_ROLE_UNKNOWN; } crm_err("Unknown role: %s", role); return RSC_ROLE_UNKNOWN; } int merge_weights(int w1, int w2) { int result = w1 + w2; if (w1 <= -INFINITY || w2 <= -INFINITY) { if (w1 >= INFINITY || w2 >= INFINITY) { crm_trace("-INFINITY + INFINITY == -INFINITY"); } return -INFINITY; } else if (w1 >= INFINITY || w2 >= INFINITY) { return INFINITY; } /* detect wrap-around */ if (result > 0) { if (w1 <= 0 && w2 < 0) { result = -INFINITY; } } else if (w1 > 0 && w2 > 0) { result = INFINITY; } /* detect +/- INFINITY */ if (result >= INFINITY) { result = INFINITY; } else if (result <= -INFINITY) { result = -INFINITY; } crm_trace("%d + %d = %d", w1, w2, result); return result; } void add_hash_param(GHashTable * hash, const char *name, const char *value) { CRM_CHECK(hash != NULL, return); crm_trace("adding: name=%s value=%s", crm_str(name), crm_str(value)); if (name == NULL || value == NULL) { return; } else if (safe_str_eq(value, "#default")) { return; } else if (g_hash_table_lookup(hash, name) == NULL) { g_hash_table_insert(hash, strdup(name), strdup(value)); } } const char * pe_node_attribute_calculated(const pe_node_t *node, const char *name, const resource_t *rsc) { const char *source; if(node == NULL) { return NULL; } else if(rsc == NULL) { return g_hash_table_lookup(node->details->attrs, name); } source = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET); if(source == NULL || safe_str_eq("host", source) == FALSE) { return g_hash_table_lookup(node->details->attrs, name); } /* Use attributes set for the containers location * instead of for the container itself * * Useful when the container is using the host's local * storage */ CRM_ASSERT(node->details->remote_rsc); CRM_ASSERT(node->details->remote_rsc->container); if(node->details->remote_rsc->container->running_on) { pe_node_t *host = node->details->remote_rsc->container->running_on->data; pe_rsc_trace(rsc, "%s: Looking for %s on the container host %s", rsc->id, name, host->details->uname); return g_hash_table_lookup(host->details->attrs, name); } pe_rsc_trace(rsc, "%s: Not looking for %s on the container host: %s is inactive", rsc->id, name, node->details->remote_rsc->container->id); return NULL; } const char * pe_node_attribute_raw(pe_node_t *node, const char *name) { if(node == NULL) { return NULL; } return g_hash_table_lookup(node->details->attrs, name); } diff --git a/lib/pengine/status.c b/lib/pengine/status.c index 3ccfac4e6c..a8b094775f 100644 --- a/lib/pengine/status.c +++ b/lib/pengine/status.c @@ -1,425 +1,428 @@ /* * Copyright 2004-2018 Andrew Beekhof * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include /*! * \brief Create a new working set * * \return New, initialized working set on success, else NULL (and set errno) * \note Only pe_working_set_t objects created with this function (as opposed * to statically declared or directly allocated) should be used with the * functions in this library, to allow for future extensions to the * data type. The caller is responsible for freeing the memory with * pe_free_working_set() when the instance is no longer needed. */ pe_working_set_t * pe_new_working_set() { pe_working_set_t *data_set = calloc(1, sizeof(pe_working_set_t)); if (data_set != NULL) { set_working_set_defaults(data_set); } return data_set; } /*! * \brief Free a working set * * \param[in] data_set Working set to free */ void pe_free_working_set(pe_working_set_t *data_set) { if (data_set != NULL) { pe_reset_working_set(data_set); free(data_set); } } /* * Unpack everything * At the end you'll have: * - A list of nodes * - A list of resources (each with any dependencies on other resources) * - A list of constraints between resources and nodes * - A list of constraints between start/stop actions * - A list of nodes that need to be stonith'd * - A list of nodes that need to be shutdown * - A list of the possible stop/start actions (without dependencies) */ gboolean cluster_status(pe_working_set_t * data_set) { xmlNode *config = get_xpath_object("//"XML_CIB_TAG_CRMCONFIG, data_set->input, LOG_TRACE); xmlNode *cib_nodes = get_xpath_object("//"XML_CIB_TAG_NODES, data_set->input, LOG_TRACE); xmlNode *cib_resources = get_xpath_object("//"XML_CIB_TAG_RESOURCES, data_set->input, LOG_TRACE); xmlNode *cib_status = get_xpath_object("//"XML_CIB_TAG_STATUS, data_set->input, LOG_TRACE); xmlNode *cib_tags = get_xpath_object("//"XML_CIB_TAG_TAGS, data_set->input, LOG_TRACE); const char *value = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM); crm_trace("Beginning unpack"); /* reset remaining global variables */ data_set->failed = create_xml_node(NULL, "failed-ops"); if (data_set->input == NULL) { return FALSE; } if (data_set->now == NULL) { data_set->now = crm_time_new(NULL); } if (data_set->dc_uuid == NULL) { data_set->dc_uuid = crm_element_value_copy(data_set->input, XML_ATTR_DC_UUID); } clear_bit(data_set->flags, pe_flag_have_quorum); if (crm_is_true(value)) { set_bit(data_set->flags, pe_flag_have_quorum); } data_set->op_defaults = get_xpath_object("//"XML_CIB_TAG_OPCONFIG, data_set->input, LOG_TRACE); data_set->rsc_defaults = get_xpath_object("//"XML_CIB_TAG_RSCCONFIG, data_set->input, LOG_TRACE); unpack_config(config, data_set); if (is_not_set(data_set->flags, pe_flag_quick_location) && is_not_set(data_set->flags, pe_flag_have_quorum) && data_set->no_quorum_policy != no_quorum_ignore) { crm_warn("Fencing and resource management disabled due to lack of quorum"); } unpack_nodes(cib_nodes, data_set); if(is_not_set(data_set->flags, pe_flag_quick_location)) { unpack_remote_nodes(cib_resources, data_set); } unpack_resources(cib_resources, data_set); unpack_tags(cib_tags, data_set); if(is_not_set(data_set->flags, pe_flag_quick_location)) { unpack_status(cib_status, data_set); } set_bit(data_set->flags, pe_flag_have_status); return TRUE; } /*! * \internal * \brief Free a list of pe_resource_t * * \param[in] resources List to free * * \note When a working set's resource list is freed, that includes the original * storage for the uname and id of any Pacemaker Remote nodes in the * working set's node list, so take care not to use those afterward. * \todo Refactor pe_node_t to strdup() the node name. */ static void pe_free_resources(GListPtr resources) { resource_t *rsc = NULL; GListPtr iterator = resources; while (iterator != NULL) { rsc = (resource_t *) iterator->data; iterator = iterator->next; rsc->fns->free(rsc); } if (resources != NULL) { g_list_free(resources); } } static void pe_free_actions(GListPtr actions) { GListPtr iterator = actions; while (iterator != NULL) { pe_free_action(iterator->data); iterator = iterator->next; } if (actions != NULL) { g_list_free(actions); } } static void pe_free_nodes(GListPtr nodes) { for (GList *iterator = nodes; iterator != NULL; iterator = iterator->next) { pe_node_t *node = (pe_node_t *) iterator->data; // Shouldn't be possible, but to be safe ... if (node == NULL) { continue; } if (node->details == NULL) { free(node); continue; } /* This is called after pe_free_resources(), which means that we can't * use node->details->uname for Pacemaker Remote nodes. */ crm_trace("Freeing node %s", (pe__is_guest_or_remote_node(node)? "(guest or remote)" : node->details->uname)); if (node->details->attrs != NULL) { g_hash_table_destroy(node->details->attrs); } if (node->details->utilization != NULL) { g_hash_table_destroy(node->details->utilization); } if (node->details->digest_cache != NULL) { g_hash_table_destroy(node->details->digest_cache); } g_list_free(node->details->running_rsc); g_list_free(node->details->allocated_rsc); free(node->details); free(node); } if (nodes != NULL) { g_list_free(nodes); } } static void pe__free_ordering(GListPtr constraints) { GListPtr iterator = constraints; while (iterator != NULL) { pe__ordering_t *order = iterator->data; iterator = iterator->next; free(order->lh_action_task); free(order->rh_action_task); free(order); } if (constraints != NULL) { g_list_free(constraints); } } static void pe__free_location(GListPtr constraints) { GListPtr iterator = constraints; while (iterator != NULL) { pe__location_t *cons = iterator->data; iterator = iterator->next; g_list_free_full(cons->node_list_rh, free); free(cons->id); free(cons); } if (constraints != NULL) { g_list_free(constraints); } } /*! * \brief Reset working set to default state without freeing it or constraints * * \param[in,out] data_set Working set to reset * * \deprecated This function is deprecated as part of the API; * pe_reset_working_set() should be used instead. */ void cleanup_calculations(pe_working_set_t * data_set) { if (data_set == NULL) { return; } clear_bit(data_set->flags, pe_flag_have_status); if (data_set->config_hash != NULL) { g_hash_table_destroy(data_set->config_hash); } if (data_set->singletons != NULL) { g_hash_table_destroy(data_set->singletons); } if (data_set->tickets) { g_hash_table_destroy(data_set->tickets); } if (data_set->template_rsc_sets) { g_hash_table_destroy(data_set->template_rsc_sets); } if (data_set->tags) { g_hash_table_destroy(data_set->tags); } free(data_set->dc_uuid); crm_trace("deleting resources"); pe_free_resources(data_set->resources); crm_trace("deleting actions"); pe_free_actions(data_set->actions); crm_trace("deleting nodes"); pe_free_nodes(data_set->nodes); pe__free_param_checks(data_set); g_list_free(data_set->stop_needed); free_xml(data_set->graph); crm_time_free(data_set->now); free_xml(data_set->input); free_xml(data_set->failed); set_working_set_defaults(data_set); CRM_CHECK(data_set->ordering_constraints == NULL,; ); CRM_CHECK(data_set->placement_constraints == NULL,; ); } /*! * \brief Reset a working set to default state without freeing it * * \param[in,out] data_set Working set to reset */ void pe_reset_working_set(pe_working_set_t *data_set) { if (data_set == NULL) { return; } crm_trace("Deleting %d ordering constraints", g_list_length(data_set->ordering_constraints)); pe__free_ordering(data_set->ordering_constraints); data_set->ordering_constraints = NULL; crm_trace("Deleting %d location constraints", g_list_length(data_set->placement_constraints)); pe__free_location(data_set->placement_constraints); data_set->placement_constraints = NULL; crm_trace("Deleting %d colocation constraints", g_list_length(data_set->colocation_constraints)); g_list_free_full(data_set->colocation_constraints, free); data_set->colocation_constraints = NULL; crm_trace("Deleting %d ticket constraints", g_list_length(data_set->ticket_constraints)); g_list_free_full(data_set->ticket_constraints, free); data_set->ticket_constraints = NULL; cleanup_calculations(data_set); } void set_working_set_defaults(pe_working_set_t * data_set) { memset(data_set, 0, sizeof(pe_working_set_t)); data_set->order_id = 1; data_set->action_id = 1; data_set->no_quorum_policy = no_quorum_freeze; data_set->flags = 0x0ULL; set_bit(data_set->flags, pe_flag_stop_rsc_orphans); set_bit(data_set->flags, pe_flag_symmetric_cluster); set_bit(data_set->flags, pe_flag_stop_action_orphans); +#ifdef DEFAULT_CONCURRENT_FENCING_TRUE + set_bit(data_set->flags, pe_flag_concurrent_fencing); +#endif } resource_t * pe_find_resource(GListPtr rsc_list, const char *id) { return pe_find_resource_with_flags(rsc_list, id, pe_find_renamed); } resource_t * pe_find_resource_with_flags(GListPtr rsc_list, const char *id, enum pe_find flags) { GListPtr rIter = NULL; for (rIter = rsc_list; id && rIter; rIter = rIter->next) { resource_t *parent = rIter->data; resource_t *match = parent->fns->find_rsc(parent, id, NULL, flags); if (match != NULL) { return match; } } crm_trace("No match for %s", id); return NULL; } node_t * pe_find_node_any(GListPtr nodes, const char *id, const char *uname) { node_t *match = pe_find_node_id(nodes, id); if (match) { return match; } crm_trace("Looking up %s via its uname instead", uname); return pe_find_node(nodes, uname); } node_t * pe_find_node_id(GListPtr nodes, const char *id) { GListPtr gIter = nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node && safe_str_eq(node->details->id, id)) { return node; } } /* error */ return NULL; } node_t * pe_find_node(GListPtr nodes, const char *uname) { GListPtr gIter = nodes; for (; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node && safe_str_eq(node->details->uname, uname)) { return node; } } /* error */ return NULL; } diff --git a/tools/crm_mon.c b/tools/crm_mon.c index e101b626b7..bed079611c 100644 --- a/tools/crm_mon.c +++ b/tools/crm_mon.c @@ -1,4415 +1,4414 @@ /* * Copyright 2004-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* crm_ends_with_ext */ #include #include #include #include #include #include #include #include #include <../lib/pengine/unpack.h> #include #include static void clean_up_connections(void); static crm_exit_t clean_up(crm_exit_t exit_code); static void crm_diff_update(const char *event, xmlNode * msg); static gboolean mon_refresh_display(gpointer user_data); static int cib_connect(gboolean full); static void mon_st_callback_event(stonith_t * st, stonith_event_t * e); static void mon_st_callback_display(stonith_t * st, stonith_event_t * e); static void kick_refresh(gboolean data_updated); static char *get_node_display_name(node_t *node); /* * Definitions indicating which items to print */ #define mon_show_times (0x0001U) #define mon_show_stack (0x0002U) #define mon_show_dc (0x0004U) #define mon_show_count (0x0008U) #define mon_show_nodes (0x0010U) #define mon_show_resources (0x0020U) #define mon_show_attributes (0x0040U) #define mon_show_failcounts (0x0080U) #define mon_show_operations (0x0100U) #define mon_show_tickets (0x0200U) #define mon_show_bans (0x0400U) #define mon_show_fence_history (0x0800U) #define mon_show_headers (mon_show_times | mon_show_stack | mon_show_dc \ | mon_show_count) #define mon_show_default (mon_show_headers | mon_show_nodes \ | mon_show_resources) #define mon_show_all (mon_show_default | mon_show_attributes \ | mon_show_failcounts | mon_show_operations \ | mon_show_tickets | mon_show_bans \ | mon_show_fence_history) static unsigned int show = mon_show_default; /* * Definitions indicating how to output */ enum mon_output_format_e { mon_output_none, mon_output_monitor, mon_output_plain, mon_output_console, mon_output_xml, mon_output_html, mon_output_cgi } output_format = mon_output_console; static char *output_filename = NULL; /* if sending output to a file, its name */ /* other globals */ static char *pid_file = NULL; static gboolean group_by_node = FALSE; static gboolean inactive_resources = FALSE; static int reconnect_msec = 5000; static gboolean daemonize = FALSE; static GMainLoop *mainloop = NULL; static guint timer_id = 0; static mainloop_timer_t *refresh_timer = NULL; static pe_working_set_t *mon_data_set = NULL; static GList *attr_list = NULL; static const char *external_agent = NULL; static const char *external_recipient = NULL; static cib_t *cib = NULL; static stonith_t *st = NULL; static xmlNode *current_cib = NULL; static gboolean one_shot = FALSE; static gboolean has_warnings = FALSE; static gboolean print_timing = FALSE; static gboolean watch_fencing = FALSE; static gboolean fence_history = FALSE; static gboolean fence_full_history = FALSE; static gboolean fence_connect = FALSE; static int fence_history_level = 1; static gboolean print_brief = FALSE; static gboolean print_pending = TRUE; static gboolean print_clone_detail = FALSE; #if CURSES_ENABLED static gboolean curses_console_initialized = FALSE; #endif /* FIXME allow, detect, and correctly interpret glob pattern or regex? */ const char *print_neg_location_prefix = ""; /* Never display node attributes whose name starts with one of these prefixes */ #define FILTER_STR { CRM_FAIL_COUNT_PREFIX, CRM_LAST_FAILURE_PREFIX, \ "shutdown", "terminate", "standby", "probe_complete", \ "#", NULL } long last_refresh = 0; crm_trigger_t *refresh_trigger = NULL; /* Define exit codes for monitoring-compatible output * For nagios plugins, the possibilities are * OK=0, WARN=1, CRIT=2, and UNKNOWN=3 */ #define MON_STATUS_WARN CRM_EX_ERROR #define MON_STATUS_CRIT CRM_EX_INVALID_PARAM #define MON_STATUS_UNKNOWN CRM_EX_UNIMPLEMENT_FEATURE /* Convenience macro for prettifying output (e.g. "node" vs "nodes") */ #define s_if_plural(i) (((i) == 1)? "" : "s") #if CURSES_ENABLED # define print_dot() if (output_format == mon_output_console) { \ printw("."); \ clrtoeol(); \ refresh(); \ } else { \ fprintf(stdout, "."); \ } #else # define print_dot() fprintf(stdout, "."); #endif #if CURSES_ENABLED # define print_as(fmt, args...) if (output_format == mon_output_console) { \ printw(fmt, ##args); \ clrtoeol(); \ refresh(); \ } else { \ fprintf(stdout, fmt, ##args); \ } #else # define print_as(fmt, args...) fprintf(stdout, fmt, ##args); #endif static void blank_screen(void) { #if CURSES_ENABLED int lpc = 0; for (lpc = 0; lpc < LINES; lpc++) { move(lpc, 0); clrtoeol(); } move(0, 0); refresh(); #endif } static gboolean mon_timer_popped(gpointer data) { int rc = pcmk_ok; #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif if (timer_id > 0) { g_source_remove(timer_id); timer_id = 0; } print_as("Reconnecting...\n"); rc = cib_connect(TRUE); if (rc != pcmk_ok) { timer_id = g_timeout_add(reconnect_msec, mon_timer_popped, NULL); } return FALSE; } static void mon_cib_connection_destroy(gpointer user_data) { print_as("Connection to the cluster-daemons terminated\n"); if (refresh_timer != NULL) { /* we'll trigger a refresh after reconnect */ mainloop_timer_stop(refresh_timer); } if (timer_id) { /* we'll trigger a new reconnect-timeout at the end */ g_source_remove(timer_id); timer_id = 0; } if (st) { /* the client API won't properly reconnect notifications * if they are still in the table - so remove them */ st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT); st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE); st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY); if (st->state != stonith_disconnected) { st->cmds->disconnect(st); } } if (cib) { cib->cmds->signoff(cib); timer_id = g_timeout_add(reconnect_msec, mon_timer_popped, NULL); } return; } /* * Mainloop signal handler. */ static void mon_shutdown(int nsig) { clean_up(CRM_EX_OK); } #if CURSES_ENABLED static sighandler_t ncurses_winch_handler; static void mon_winresize(int nsig) { static int not_done; int lines = 0, cols = 0; if (!not_done++) { if (ncurses_winch_handler) /* the original ncurses WINCH signal handler does the * magic of retrieving the new window size; * otherwise, we'd have to use ioctl or tgetent */ (*ncurses_winch_handler) (SIGWINCH); getmaxyx(stdscr, lines, cols); resizeterm(lines, cols); mainloop_set_trigger(refresh_trigger); } not_done--; } #endif static int cib_connect(gboolean full) { int rc = pcmk_ok; static gboolean need_pass = TRUE; CRM_CHECK(cib != NULL, return -EINVAL); if (getenv("CIB_passwd") != NULL) { need_pass = FALSE; } if ((fence_connect) && (st == NULL)) { st = stonith_api_new(); } if ((fence_connect) && (st->state == stonith_disconnected)) { - crm_trace("Connecting to stonith"); rc = st->cmds->connect(st, crm_system_name, NULL); if (rc == pcmk_ok) { crm_trace("Setting up stonith callbacks"); if (watch_fencing) { st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, mon_st_callback_event); st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, mon_st_callback_event); } else { st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT, mon_st_callback_display); st->cmds->register_notification(st, T_STONITH_NOTIFY_HISTORY, mon_st_callback_display); } } } if (cib->state != cib_connected_query && cib->state != cib_connected_command) { crm_trace("Connecting to the CIB"); if ((output_format == mon_output_console) && need_pass && (cib->variant == cib_remote)) { need_pass = FALSE; print_as("Password:"); } rc = cib->cmds->signon(cib, crm_system_name, cib_query); if (rc != pcmk_ok) { return rc; } rc = cib->cmds->query(cib, NULL, ¤t_cib, cib_scope_local | cib_sync_call); if (rc == pcmk_ok) { mon_refresh_display(NULL); } if (rc == pcmk_ok && full) { if (rc == pcmk_ok) { rc = cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy); if (rc == -EPROTONOSUPPORT) { print_as ("Notification setup not supported, won't be able to reconnect after failure"); if (output_format == mon_output_console) { sleep(2); } rc = pcmk_ok; } } if (rc == pcmk_ok) { cib->cmds->del_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update); rc = cib->cmds->add_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update); } if (rc != pcmk_ok) { print_as("Notification setup failed, could not monitor CIB actions"); if (output_format == mon_output_console) { sleep(2); } clean_up_connections(); } } } return rc; } /* *INDENT-OFF* */ static struct crm_option long_options[] = { /* Top-level Options */ {"help", 0, 0, '?', "\tThis text"}, {"version", 0, 0, '$', "\tVersion information" }, {"verbose", 0, 0, 'V', "\tIncrease debug output"}, {"quiet", 0, 0, 'Q', "\tDisplay only essential output" }, {"-spacer-", 1, 0, '-', "\nModes (mutually exclusive):"}, {"as-html", 1, 0, 'h', "\tWrite cluster status to the named html file"}, {"as-xml", 0, 0, 'X', "\t\tWrite cluster status as xml to stdout. This will enable one-shot mode."}, {"web-cgi", 0, 0, 'w', "\t\tWeb mode with output suitable for CGI (preselected when run as *.cgi)"}, {"simple-status", 0, 0, 's', "\tDisplay the cluster status once as a simple one line output (suitable for nagios)"}, {"-spacer-", 1, 0, '-', "\nDisplay Options:"}, {"group-by-node", 0, 0, 'n', "\tGroup resources by node" }, {"inactive", 0, 0, 'r', "\t\tDisplay inactive resources" }, {"failcounts", 0, 0, 'f', "\tDisplay resource fail counts"}, {"operations", 0, 0, 'o', "\tDisplay resource operation history" }, {"timing-details", 0, 0, 't', "\tDisplay resource operation history with timing details" }, {"tickets", 0, 0, 'c', "\t\tDisplay cluster tickets"}, {"watch-fencing", 0, 0, 'W', "\tListen for fencing events. For use with --external-agent"}, {"fence-history", 2, 0, 'm', "Show fence history\n" "\t\t\t\t\t0=off, 1=failures and pending (default without option),\n" "\t\t\t\t\t2=add successes (default without value for option),\n" "\t\t\t\t\t3=show full history without reduction to most recent of each flavor"}, {"neg-locations", 2, 0, 'L', "Display negative location constraints [optionally filtered by id prefix]"}, {"show-node-attributes", 0, 0, 'A', "Display node attributes" }, {"hide-headers", 0, 0, 'D', "\tHide all headers" }, {"show-detail", 0, 0, 'R', "\tShow more details (node IDs, individual clone instances)" }, {"brief", 0, 0, 'b', "\t\tBrief output" }, {"pending", 0, 0, 'j', "\t\tDisplay pending state if 'record-pending' is enabled", pcmk_option_hidden}, {"-spacer-", 1, 0, '-', "\nAdditional Options:"}, {"interval", 1, 0, 'i', "\tUpdate frequency in seconds" }, {"one-shot", 0, 0, '1', "\t\tDisplay the cluster status once on the console and exit"}, {"disable-ncurses",0, 0, 'N', "\tDisable the use of ncurses", !CURSES_ENABLED}, {"daemonize", 0, 0, 'd', "\tRun in the background as a daemon"}, {"pid-file", 1, 0, 'p', "\t(Advanced) Daemon pid file location"}, {"external-agent", 1, 0, 'E', "A program to run when resource operations take place."}, {"external-recipient",1, 0, 'e', "A recipient for your program (assuming you want the program to send something to someone)."}, {"xml-file", 1, 0, 'x', NULL, pcmk_option_hidden}, {"-spacer-", 1, 0, '-', "\nExamples:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', "Display the cluster status on the console with updates as they occur:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Display the cluster status on the console just once then exit:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon -1", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Display your cluster status, group resources by node, and include inactive resources in the list:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --group-by-node --inactive", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Start crm_mon as a background daemon and have it write the cluster status to an HTML file:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --daemonize --as-html /path/to/docroot/filename.html", pcmk_option_example}, {"-spacer-", 1, 0, '-', "Start crm_mon and export the current cluster status as xml to stdout, then exit.:", pcmk_option_paragraph}, {"-spacer-", 1, 0, '-', " crm_mon --as-xml", pcmk_option_example}, {NULL, 0, 0, 0} }; /* *INDENT-ON* */ #if CURSES_ENABLED static const char * get_option_desc(char c) { int lpc; for (lpc = 0; long_options[lpc].name != NULL; lpc++) { if (long_options[lpc].name[0] == '-') continue; if (long_options[lpc].val == c) { static char *buf = NULL; const char *rv; char *nl; /* chop off tabs and cut at newline */ free(buf); /* free string from last usage */ buf = strdup(long_options[lpc].desc); rv = buf; /* make a copy to keep buf pointer unaltered for freeing when we come by next time. Like this the result stays valid until the next call. */ while(isspace(rv[0])) { rv++; } nl = strchr(rv, '\n'); if (nl) { *nl = '\0'; } return rv; } } return NULL; } #define print_option_help(option, condition) \ print_as("%c %c: \t%s\n", ((condition)? '*': ' '), option, get_option_desc(option)); static gboolean detect_user_input(GIOChannel *channel, GIOCondition condition, gpointer unused) { int c; gboolean config_mode = FALSE; while (1) { /* Get user input */ c = getchar(); switch (c) { case 'm': if (!fence_history_level) { fence_history = TRUE; fence_connect = TRUE; if (st == NULL) { mon_cib_connection_destroy(NULL); } } show ^= mon_show_fence_history; break; case 'c': show ^= mon_show_tickets; break; case 'f': show ^= mon_show_failcounts; break; case 'n': group_by_node = ! group_by_node; break; case 'o': show ^= mon_show_operations; if ((show & mon_show_operations) == 0) { print_timing = 0; } break; case 'r': inactive_resources = ! inactive_resources; break; case 'R': print_clone_detail = ! print_clone_detail; break; case 't': print_timing = ! print_timing; if (print_timing) { show |= mon_show_operations; } break; case 'A': show ^= mon_show_attributes; break; case 'L': show ^= mon_show_bans; break; case 'D': /* If any header is shown, clear them all, otherwise set them all */ if (show & mon_show_headers) { show &= ~mon_show_headers; } else { show |= mon_show_headers; } break; case 'b': print_brief = ! print_brief; break; case 'j': print_pending = ! print_pending; break; case '?': config_mode = TRUE; break; default: goto refresh; } if (!config_mode) goto refresh; blank_screen(); print_as("Display option change mode\n"); print_as("\n"); print_option_help('c', show & mon_show_tickets); print_option_help('f', show & mon_show_failcounts); print_option_help('n', group_by_node); print_option_help('o', show & mon_show_operations); print_option_help('r', inactive_resources); print_option_help('t', print_timing); print_option_help('A', show & mon_show_attributes); print_option_help('L', show & mon_show_bans); print_option_help('D', (show & mon_show_headers) == 0); print_option_help('R', print_clone_detail); print_option_help('b', print_brief); print_option_help('j', print_pending); print_option_help('m', (show & mon_show_fence_history)); print_as("\n"); print_as("Toggle fields via field letter, type any other key to return"); } refresh: mon_refresh_display(NULL); return TRUE; } #endif // Basically crm_signal_handler(SIGCHLD, SIG_IGN) plus the SA_NOCLDWAIT flag static void avoid_zombies() { struct sigaction sa; memset(&sa, 0, sizeof(struct sigaction)); if (sigemptyset(&sa.sa_mask) < 0) { crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno)); return; } sa.sa_handler = SIG_IGN; sa.sa_flags = SA_RESTART|SA_NOCLDWAIT; if (sigaction(SIGCHLD, &sa, NULL) < 0) { crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno)); } } int main(int argc, char **argv) { int flag; int argerr = 0; int option_index = 0; int rc = pcmk_ok; pid_file = strdup("/tmp/ClusterMon.pid"); crm_log_cli_init("crm_mon"); crm_set_options(NULL, "mode [options]", long_options, "Provides a summary of cluster's current state." "\n\nOutputs varying levels of detail in a number of different formats.\n"); // Avoid needing to wait for subprocesses forked for -E/--external-agent avoid_zombies(); if (crm_ends_with_ext(argv[0], ".cgi") == TRUE) { output_format = mon_output_cgi; one_shot = TRUE; } /* to enable stonith-connection when called via some application like pcs * set environment-variable FENCE_HISTORY to desired level * so you don't have to modify this application */ /* fence_history_level = crm_atoi(getenv("FENCE_HISTORY"), "0"); */ while (1) { flag = crm_get_option(argc, argv, &option_index); if (flag == -1) break; switch (flag) { case 'V': crm_bump_log_level(argc, argv); break; case 'Q': show &= ~mon_show_times; break; case 'i': reconnect_msec = crm_get_msec(optarg); break; case 'n': group_by_node = TRUE; break; case 'r': inactive_resources = TRUE; break; case 'W': watch_fencing = TRUE; fence_connect = TRUE; break; case 'm': fence_history_level = crm_atoi(optarg, "2"); break; case 'd': daemonize = TRUE; break; case 't': print_timing = TRUE; show |= mon_show_operations; break; case 'o': show |= mon_show_operations; break; case 'f': show |= mon_show_failcounts; break; case 'A': show |= mon_show_attributes; break; case 'L': show |= mon_show_bans; print_neg_location_prefix = optarg? optarg : ""; break; case 'D': show &= ~mon_show_headers; break; case 'b': print_brief = TRUE; break; case 'j': print_pending = TRUE; break; case 'R': print_clone_detail = TRUE; break; case 'c': show |= mon_show_tickets; break; case 'p': free(pid_file); if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } pid_file = strdup(optarg); break; case 'x': if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } setenv("CIB_file", optarg, 1); one_shot = TRUE; break; case 'h': if(optarg == NULL) { crm_help(flag, CRM_EX_USAGE); } argerr += (output_format != mon_output_console); output_format = mon_output_html; output_filename = strdup(optarg); umask(S_IWGRP | S_IWOTH); break; case 'X': argerr += (output_format != mon_output_console); output_format = mon_output_xml; one_shot = TRUE; break; case 'w': /* do not allow argv[0] and argv[1...] redundancy */ argerr += (output_format != mon_output_console); output_format = mon_output_cgi; one_shot = TRUE; break; case 's': argerr += (output_format != mon_output_console); output_format = mon_output_monitor; one_shot = TRUE; break; case 'E': external_agent = optarg; break; case 'e': external_recipient = optarg; break; case '1': one_shot = TRUE; break; case 'N': if (output_format == mon_output_console) { output_format = mon_output_plain; } break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; default: printf("Argument code 0%o (%c) is not (?yet?) supported\n", flag, flag); ++argerr; break; } } if (watch_fencing) { /* don't moan as fence_history_level == 1 is default */ fence_history_level = 0; } /* create the cib-object early to be able to do further * decisions based on the cib-source */ cib = cib_new(); if (cib == NULL) { rc = -EINVAL; } else { switch (cib->variant) { case cib_native: /* cib & fencing - everything available */ break; case cib_file: /* Don't try to connect to fencing as we * either don't have a running cluster or * the fencing-information would possibly * not match the cib data from a file. * As we don't expect cib-updates coming * in enforce one-shot. */ fence_history_level = 0; one_shot = TRUE; break; case cib_remote: /* updates coming in but no fencing */ fence_history_level = 0; break; case cib_undefined: case cib_database: default: /* something is odd */ rc = -EINVAL; crm_err("Invalid cib-source"); break; } } switch (fence_history_level) { case 3: fence_full_history = TRUE; /* fall through to next lower level */ case 2: show |= mon_show_fence_history; /* fall through to next lower level */ case 1: fence_history = TRUE; fence_connect = TRUE; break; default: break; } /* Extra sanity checks when in CGI mode */ if (output_format == mon_output_cgi) { argerr += (optind < argc); argerr += (output_filename != NULL); argerr += ((cib) && (cib->variant == cib_file)); argerr += (external_agent != NULL); argerr += (daemonize == TRUE); /* paranoia */ } else if (optind < argc) { printf("non-option ARGV-elements: "); while (optind < argc) printf("%s ", argv[optind++]); printf("\n"); } if (argerr) { return clean_up(CRM_EX_USAGE); } /* XML output always prints everything */ if (output_format == mon_output_xml) { show = mon_show_all; print_timing = TRUE; } if (one_shot) { if (output_format == mon_output_console) { output_format = mon_output_plain; } } else if (daemonize) { if ((output_format == mon_output_console) || (output_format == mon_output_plain)) { output_format = mon_output_none; } crm_enable_stderr(FALSE); if ((output_format != mon_output_html) && !external_agent) { printf ("Looks like you forgot to specify one or more of: " "--as-html, --external-agent\n"); return clean_up(CRM_EX_USAGE); } if (cib) { /* to be on the safe side don't have cib-object around * when we are forking */ cib_delete(cib); cib = NULL; crm_make_daemon(crm_system_name, TRUE, pid_file); cib = cib_new(); if (cib == NULL) { rc = -EINVAL; } /* otherwise assume we've got the same cib-object we've just destroyed * in our parent */ } } else if (output_format == mon_output_console) { #if CURSES_ENABLED initscr(); cbreak(); noecho(); crm_enable_stderr(FALSE); curses_console_initialized = TRUE; #else one_shot = TRUE; output_format = mon_output_plain; printf("Defaulting to one-shot mode\n"); printf("You need to have curses available at compile time to enable console mode\n"); #endif } crm_info("Starting %s", crm_system_name); if (cib) { do { if (!one_shot) { print_as("Waiting until cluster is available on this node ...\n"); } rc = cib_connect(!one_shot); if (one_shot) { break; } else if (rc != pcmk_ok) { sleep(reconnect_msec / 1000); #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif } else { if (output_format == mon_output_html) { print_as("Writing html to %s ...\n", output_filename); } } } while (rc == -ENOTCONN); } if (rc != pcmk_ok) { if (output_format == mon_output_monitor) { printf("CLUSTER CRIT: Connection to cluster failed: %s\n", pcmk_strerror(rc)); return clean_up(MON_STATUS_CRIT); } else { if (rc == -ENOTCONN) { print_as("\nError: cluster is not available on this node\n"); } else { print_as("\nConnection to cluster failed: %s\n", pcmk_strerror(rc)); } } if (output_format == mon_output_console) { sleep(2); } return clean_up(crm_errno2exit(rc)); } if (one_shot) { return clean_up(CRM_EX_OK); } mainloop = g_main_loop_new(NULL, FALSE); mainloop_add_signal(SIGTERM, mon_shutdown); mainloop_add_signal(SIGINT, mon_shutdown); #if CURSES_ENABLED if (output_format == mon_output_console) { ncurses_winch_handler = crm_signal_handler(SIGWINCH, mon_winresize); if (ncurses_winch_handler == SIG_DFL || ncurses_winch_handler == SIG_IGN || ncurses_winch_handler == SIG_ERR) ncurses_winch_handler = NULL; g_io_add_watch(g_io_channel_unix_new(STDIN_FILENO), G_IO_IN, detect_user_input, NULL); } #endif refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_display, NULL); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); crm_info("Exiting %s", crm_system_name); return clean_up(CRM_EX_OK); } #define mon_warn(fmt...) do { \ if (!has_warnings) { \ print_as("CLUSTER WARN:"); \ } else { \ print_as(","); \ } \ print_as(fmt); \ has_warnings = TRUE; \ } while(0) static int count_resources(pe_working_set_t * data_set, resource_t * rsc) { int count = 0; GListPtr gIter = NULL; if (rsc == NULL) { gIter = data_set->resources; } else if (rsc->children) { gIter = rsc->children; } else { return is_not_set(rsc->flags, pe_rsc_orphan); } for (; gIter != NULL; gIter = gIter->next) { count += count_resources(data_set, gIter->data); } return count; } /*! * \internal * \brief Print one-line status suitable for use with monitoring software * * \param[in] data_set Working set of CIB state * \param[in] history List of stonith actions * * \note This function's output (and the return code when the program exits) * should conform to https://www.monitoring-plugins.org/doc/guidelines.html */ static void print_simple_status(pe_working_set_t * data_set, stonith_history_t *history) { GListPtr gIter = NULL; int nodes_online = 0; int nodes_standby = 0; int nodes_maintenance = 0; if (data_set->dc_node == NULL) { mon_warn(" No DC"); } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; if (node->details->standby && node->details->online) { nodes_standby++; } else if (node->details->maintenance && node->details->online) { nodes_maintenance++; } else if (node->details->online) { nodes_online++; } else { mon_warn(" offline node: %s", node->details->uname); } } if (!has_warnings) { int nresources = count_resources(data_set, NULL); print_as("CLUSTER OK: %d node%s online", nodes_online, s_if_plural(nodes_online)); if (nodes_standby > 0) { print_as(", %d standby node%s", nodes_standby, s_if_plural(nodes_standby)); } if (nodes_maintenance > 0) { print_as(", %d maintenance node%s", nodes_maintenance, s_if_plural(nodes_maintenance)); } print_as(", %d resource%s configured", nresources, s_if_plural(nresources)); } print_as("\n"); } /*! * \internal * \brief Print a [name]=[value][units] pair, optionally using time string * * \param[in] stream File stream to display output to * \param[in] name Name to display * \param[in] value Value to display (or NULL to convert time instead) * \param[in] units Units to display (or NULL for no units) * \param[in] epoch_time Epoch time to convert if value is NULL */ static void print_nvpair(FILE *stream, const char *name, const char *value, const char *units, time_t epoch_time) { /* print name= */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" %s=", name); break; case mon_output_html: case mon_output_cgi: case mon_output_xml: fprintf(stream, " %s=", name); break; default: break; } /* If we have a value (and optionally units), print it */ if (value) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as("%s%s", value, (units? units : "")); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "%s%s", value, (units? units : "")); break; case mon_output_xml: fprintf(stream, "\"%s%s\"", value, (units? units : "")); break; default: break; } /* Otherwise print user-friendly time string */ } else { static char empty_str[] = ""; char *c, *date_str = asctime(localtime(&epoch_time)); for (c = (date_str != NULL) ? date_str : empty_str; *c != '\0'; ++c) { if (*c == '\n') { *c = '\0'; break; } } switch (output_format) { case mon_output_plain: case mon_output_console: print_as("'%s'", date_str); break; case mon_output_html: case mon_output_cgi: case mon_output_xml: fprintf(stream, "\"%s\"", date_str); break; default: break; } } } /*! * \internal * \brief Print whatever is needed to start a node section * * \param[in] stream File stream to display output to * \param[in] node Node to print */ static void print_node_start(FILE *stream, node_t *node) { char *node_name; switch (output_format) { case mon_output_plain: case mon_output_console: node_name = get_node_display_name(node); print_as("* Node %s:\n", node_name); free(node_name); break; case mon_output_html: case mon_output_cgi: node_name = get_node_display_name(node); fprintf(stream, "

Node: %s

\n
    \n", node_name); free(node_name); break; case mon_output_xml: fprintf(stream, " \n", node->details->uname); break; default: break; } } /*! * \internal * \brief Print whatever is needed to end a node section * * \param[in] stream File stream to display output to */ static void print_node_end(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
\n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print resources section heading appropriate to options * * \param[in] stream File stream to display output to */ static void print_resources_heading(FILE *stream) { const char *heading; if (group_by_node) { /* Active resources have already been printed by node */ heading = (inactive_resources? "Inactive resources" : NULL); } else if (inactive_resources) { heading = "Full list of resources"; } else { heading = "Active resources"; } /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n%s:\n\n", heading); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
\n

%s

\n", heading); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print whatever resource section closing is appropriate * * \param[in] stream File stream to display output to */ static void print_resources_closing(FILE *stream, gboolean printed_heading) { const char *heading; /* What type of resources we did or did not display */ if (group_by_node) { heading = "inactive "; } else if (inactive_resources) { heading = ""; } else { heading = "active "; } switch (output_format) { case mon_output_plain: case mon_output_console: if (!printed_heading) { print_as("\nNo %sresources\n\n", heading); } break; case mon_output_html: case mon_output_cgi: if (!printed_heading) { fprintf(stream, "
\n

No %sresources

\n", heading); } break; case mon_output_xml: fprintf(stream, " %s\n", (printed_heading? "" : "")); break; default: break; } } /*! * \internal * \brief Print whatever resource section(s) are appropriate * * \param[in] stream File stream to display output to * \param[in] data_set Cluster state to display * \param[in] print_opts Bitmask of pe_print_options */ static void print_resources(FILE *stream, pe_working_set_t *data_set, int print_opts) { GListPtr rsc_iter; const char *prefix = NULL; gboolean printed_heading = FALSE; gboolean brief_output = print_brief; /* If we already showed active resources by node, and * we're not showing inactive resources, we have nothing to do */ if (group_by_node && !inactive_resources) { return; } /* XML uses an indent, and ignores brief option for resources */ if (output_format == mon_output_xml) { prefix = " "; brief_output = FALSE; } /* If we haven't already printed resources grouped by node, * and brief output was requested, print resource summary */ if (brief_output && !group_by_node) { print_resources_heading(stream); printed_heading = TRUE; print_rscs_brief(data_set->resources, NULL, print_opts, stream, inactive_resources); } /* For each resource, display it if appropriate */ for (rsc_iter = data_set->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { resource_t *rsc = (resource_t *) rsc_iter->data; /* Complex resources may have some sub-resources active and some inactive */ gboolean is_active = rsc->fns->active(rsc, TRUE); gboolean partially_active = rsc->fns->active(rsc, FALSE); /* Skip inactive orphans (deleted but still in CIB) */ if (is_set(rsc->flags, pe_rsc_orphan) && !is_active) { continue; /* Skip active resources if we already displayed them by node */ } else if (group_by_node) { if (is_active) { continue; } /* Skip primitives already counted in a brief summary */ } else if (brief_output && (rsc->variant == pe_native)) { continue; /* Skip resources that aren't at least partially active, * unless we're displaying inactive resources */ } else if (!partially_active && !inactive_resources) { continue; } /* Print this resource */ if (printed_heading == FALSE) { print_resources_heading(stream); printed_heading = TRUE; } rsc->fns->print(rsc, prefix, print_opts, stream); } print_resources_closing(stream, printed_heading); } /*! * \internal * \brief Print heading for resource history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node that ran this resource * \param[in] rsc Resource to print * \param[in] rsc_id ID of resource to print * \param[in] all Whether to print every resource or just failed ones */ static void print_rsc_history_start(FILE *stream, pe_working_set_t *data_set, node_t *node, resource_t *rsc, const char *rsc_id, gboolean all) { time_t last_failure = 0; int failcount = rsc? pe_get_failcount(node, rsc, &last_failure, pe_fc_default, NULL, data_set) : 0; if (!all && !failcount && (last_failure <= 0)) { return; } /* Print resource ID */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" %s:", rsc_id); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • %s:", rsc_id); break; case mon_output_xml: fprintf(stream, " 0)) { /* Print migration threshold */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" migration-threshold=%d", rsc->migration_threshold); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " migration-threshold=%d", rsc->migration_threshold); break; case mon_output_xml: fprintf(stream, " orphan=\"false\" migration-threshold=\"%d\"", rsc->migration_threshold); break; default: break; } /* Print fail count if any */ if (failcount > 0) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" " CRM_FAIL_COUNT_PREFIX "=%d", failcount); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " " CRM_FAIL_COUNT_PREFIX "=%d", failcount); break; case mon_output_xml: fprintf(stream, " " CRM_FAIL_COUNT_PREFIX "=\"%d\"", failcount); break; default: break; } } /* Print last failure time if any */ if (last_failure > 0) { print_nvpair(stream, CRM_LAST_FAILURE_PREFIX, NULL, NULL, last_failure); } } /* End the heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "\n
      \n"); break; case mon_output_xml: fprintf(stream, ">\n"); break; default: break; } } /*! * \internal * \brief Print closing for resource history * * \param[in] stream File stream to display output to */ static void print_rsc_history_end(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n
    • \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print operation history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node this operation is for * \param[in] xml_op Root of XML tree describing this operation * \param[in] task Task parsed from this operation's XML * \param[in] interval_ms_s Interval parsed from this operation's XML * \param[in] rc Return code parsed from this operation's XML */ static void print_op_history(FILE *stream, pe_working_set_t *data_set, node_t *node, xmlNode *xml_op, const char *task, const char *interval_ms_s, int rc) { const char *value = NULL; const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); /* Begin the operation description */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" + (%s) %s:", call, task); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • (%s) %s:", call, task); break; case mon_output_xml: fprintf(stream, " 0) { print_nvpair(stream, attr, NULL, NULL, int_value); } } attr = XML_RSC_OP_LAST_RUN; value = crm_element_value(xml_op, attr); if (value) { int_value = crm_parse_int(value, NULL); if (int_value > 0) { print_nvpair(stream, attr, NULL, NULL, int_value); } } attr = XML_RSC_OP_T_EXEC; value = crm_element_value(xml_op, attr); if (value) { print_nvpair(stream, attr, value, "ms", 0); } attr = XML_RSC_OP_T_QUEUE; value = crm_element_value(xml_op, attr); if (value) { print_nvpair(stream, attr, value, "ms", 0); } } /* End the operation description */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" rc=%d (%s)\n", rc, services_ocf_exitcode_str(rc)); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " rc=%d (%s)
    • \n", rc, services_ocf_exitcode_str(rc)); break; case mon_output_xml: fprintf(stream, " rc=\"%d\" rc_text=\"%s\" />\n", rc, services_ocf_exitcode_str(rc)); break; default: break; } } /*! * \internal * \brief Print resource operation/failure history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node Node that ran this resource * \param[in] rsc_entry Root of XML tree describing resource status * \param[in] operations Whether to print operations or just failcounts */ static void print_rsc_history(FILE *stream, pe_working_set_t *data_set, node_t *node, xmlNode *rsc_entry, gboolean operations) { GListPtr gIter = NULL; GListPtr op_list = NULL; gboolean printed = FALSE; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); resource_t *rsc = pe_find_resource(data_set->resources, rsc_id); xmlNode *rsc_op = NULL; /* If we're not showing operations, just print the resource failure summary */ if (operations == FALSE) { print_rsc_history_start(stream, data_set, node, rsc, rsc_id, FALSE); print_rsc_history_end(stream); return; } /* Create a list of this resource's operations */ for (rsc_op = __xml_first_child(rsc_entry); rsc_op != NULL; rsc_op = __xml_next(rsc_op)) { if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) { op_list = g_list_append(op_list, rsc_op); } } op_list = g_list_sort(op_list, sort_op_by_callid); /* Print each operation */ for (gIter = op_list; gIter != NULL; gIter = gIter->next) { xmlNode *xml_op = (xmlNode *) gIter->data; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(xml_op, XML_LRM_ATTR_RC); int rc = crm_parse_int(op_rc, "0"); /* Display 0-interval monitors as "probe" */ if (safe_str_eq(task, CRMD_ACTION_STATUS) && ((interval_ms_s == NULL) || safe_str_eq(interval_ms_s, "0"))) { task = "probe"; } /* Ignore notifies and some probes */ if (safe_str_eq(task, CRMD_ACTION_NOTIFY) || (safe_str_eq(task, "probe") && (rc == 7))) { continue; } /* If this is the first printed operation, print heading for resource */ if (printed == FALSE) { printed = TRUE; print_rsc_history_start(stream, data_set, node, rsc, rsc_id, TRUE); } /* Print the operation */ print_op_history(stream, data_set, node, xml_op, task, interval_ms_s, rc); } /* Free the list we created (no need to free the individual items) */ g_list_free(op_list); /* If we printed anything, close the resource */ if (printed) { print_rsc_history_end(stream); } } /*! * \internal * \brief Print node operation/failure history * * \param[in] stream File stream to display output to * \param[in] data_set Current state of CIB * \param[in] node_state Root of XML tree describing node status * \param[in] operations Whether to print operations or just failcounts */ static void print_node_history(FILE *stream, pe_working_set_t *data_set, xmlNode *node_state, gboolean operations) { node_t *node = pe_find_node_id(data_set->nodes, ID(node_state)); xmlNode *lrm_rsc = NULL; xmlNode *rsc_entry = NULL; if (node && node->details && node->details->online) { print_node_start(stream, node); lrm_rsc = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE); /* Print history of each of the node's resources */ for (rsc_entry = __xml_first_child(lrm_rsc); rsc_entry != NULL; rsc_entry = __xml_next(rsc_entry)) { if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) { print_rsc_history(stream, data_set, node, rsc_entry, operations); } } print_node_end(stream); } } /*! * \internal * \brief Print extended information about an attribute if appropriate * * \param[in] data_set Working set of CIB state * * \return TRUE if extended information was printed, FALSE otherwise * \note Currently, extended information is only supported for ping/pingd * resources, for which a message will be printed if connectivity is lost * or degraded. */ static gboolean print_attr_msg(FILE *stream, node_t * node, GListPtr rsc_list, const char *attrname, const char *attrvalue) { GListPtr gIter = NULL; for (gIter = rsc_list; gIter != NULL; gIter = gIter->next) { resource_t *rsc = (resource_t *) gIter->data; const char *type = g_hash_table_lookup(rsc->meta, "type"); if (rsc->children != NULL) { if (print_attr_msg(stream, node, rsc->children, attrname, attrvalue)) { return TRUE; } } if (safe_str_eq(type, "ping") || safe_str_eq(type, "pingd")) { const char *name = g_hash_table_lookup(rsc->parameters, "name"); if (name == NULL) { name = "pingd"; } /* To identify the resource with the attribute name. */ if (safe_str_eq(name, attrname)) { int host_list_num = 0; int expected_score = 0; int value = crm_parse_int(attrvalue, "0"); const char *hosts = g_hash_table_lookup(rsc->parameters, "host_list"); const char *multiplier = g_hash_table_lookup(rsc->parameters, "multiplier"); if(hosts) { char **host_list = g_strsplit(hosts, " ", 0); host_list_num = g_strv_length(host_list); g_strfreev(host_list); } /* pingd multiplier is the same as the default value. */ expected_score = host_list_num * crm_parse_int(multiplier, "1"); switch (output_format) { case mon_output_plain: case mon_output_console: if (value <= 0) { print_as("\t: Connectivity is lost"); } else if (value < expected_score) { print_as("\t: Connectivity is degraded (Expected=%d)", expected_score); } break; case mon_output_html: case mon_output_cgi: if (value <= 0) { fprintf(stream, " (connectivity is lost)"); } else if (value < expected_score) { fprintf(stream, " (connectivity is degraded -- expected %d)", expected_score); } break; case mon_output_xml: fprintf(stream, " expected=\"%d\"", expected_score); break; default: break; } return TRUE; } } } return FALSE; } static int compare_attribute(gconstpointer a, gconstpointer b) { int rc; rc = strcmp((const char *)a, (const char *)b); return rc; } static void create_attr_list(gpointer name, gpointer value, gpointer data) { int i; const char *filt_str[] = FILTER_STR; CRM_CHECK(name != NULL, return); /* filtering automatic attributes */ for (i = 0; filt_str[i] != NULL; i++) { if (g_str_has_prefix(name, filt_str[i])) { return; } } attr_list = g_list_insert_sorted(attr_list, name, compare_attribute); } /* structure for passing multiple user data to g_list_foreach() */ struct mon_attr_data { FILE *stream; node_t *node; }; static void print_node_attribute(gpointer name, gpointer user_data) { const char *value = NULL; struct mon_attr_data *data = (struct mon_attr_data *) user_data; value = pe_node_attribute_raw(data->node, name); /* Print attribute name and value */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as(" + %-32s\t: %-10s", (char *)name, value); break; case mon_output_html: case mon_output_cgi: fprintf(data->stream, "
  • %s: %s", (char *)name, value); break; case mon_output_xml: fprintf(data->stream, " stream, data->node, data->node->details->running_rsc, name, value); /* Close out the attribute */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(data->stream, "
    • \n"); break; case mon_output_xml: fprintf(data->stream, " />\n"); break; default: break; } } static void print_node_summary(FILE *stream, pe_working_set_t * data_set, gboolean operations) { xmlNode *node_state = NULL; xmlNode *cib_status = get_object_root(XML_CIB_TAG_STATUS, data_set->input); /* Print heading */ switch (output_format) { case mon_output_plain: case mon_output_console: if (operations) { print_as("\nOperations:\n"); } else { print_as("\nMigration Summary:\n"); } break; case mon_output_html: case mon_output_cgi: if (operations) { fprintf(stream, "
    \n

    Operations

    \n"); } else { fprintf(stream, "
    \n

    Migration Summary

    \n"); } break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each node in the CIB status */ for (node_state = __xml_first_child(cib_status); node_state != NULL; node_state = __xml_next(node_state)) { if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) { print_node_history(stream, data_set, node_state, operations); } } /* Close section */ switch (output_format) { case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } static void print_ticket(gpointer name, gpointer value, gpointer data) { ticket_t *ticket = (ticket_t *) value; FILE *stream = (FILE *) data; switch (output_format) { case mon_output_plain: case mon_output_console: print_as("* %s:\t%s%s", ticket->id, (ticket->granted? "granted" : "revoked"), (ticket->standby? " [standby]" : "")); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
  • %s: %s%s", ticket->id, (ticket->granted? "granted" : "revoked"), (ticket->standby? " [standby]" : "")); break; case mon_output_xml: fprintf(stream, " id, (ticket->granted? "granted" : "revoked"), (ticket->standby? "true" : "false")); break; default: break; } if (ticket->last_granted > -1) { print_nvpair(stdout, "last-granted", NULL, NULL, ticket->last_granted); } switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    • \n"); break; case mon_output_xml: fprintf(stream, " />\n"); break; default: break; } } static void print_cluster_tickets(FILE *stream, pe_working_set_t * data_set) { /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nTickets:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Tickets

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each ticket */ g_hash_table_foreach(data_set->tickets, print_ticket, stream); /* Close section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Return human-friendly string representing node name * * The returned string will be in the format * uname[@hostUname] [(nodeID)] * "@hostUname" will be printed if the node is a guest node. * "(nodeID)" will be printed if the node ID is different from the node uname, * and detailed output has been requested. * * \param[in] node Node to represent * \return Newly allocated string with representation of node name * \note It is the caller's responsibility to free the result with free(). */ static char * get_node_display_name(node_t *node) { char *node_name; const char *node_host = NULL; const char *node_id = NULL; int name_len; CRM_ASSERT((node != NULL) && (node->details != NULL) && (node->details->uname != NULL)); /* Host is displayed only if this is a guest node */ if (pe__is_guest_node(node)) { node_t *host_node = pe__current_node(node->details->remote_rsc); if (host_node && host_node->details) { node_host = host_node->details->uname; } if (node_host == NULL) { node_host = ""; /* so we at least get "uname@" to indicate guest */ } } /* Node ID is displayed if different from uname and detail is requested */ if (print_clone_detail && safe_str_neq(node->details->uname, node->details->id)) { node_id = node->details->id; } /* Determine name length */ name_len = strlen(node->details->uname) + 1; if (node_host) { name_len += strlen(node_host) + 1; /* "@node_host" */ } if (node_id) { name_len += strlen(node_id) + 3; /* + " (node_id)" */ } /* Allocate and populate display name */ node_name = malloc(name_len); CRM_ASSERT(node_name != NULL); strcpy(node_name, node->details->uname); if (node_host) { strcat(node_name, "@"); strcat(node_name, node_host); } if (node_id) { strcat(node_name, " ("); strcat(node_name, node_id); strcat(node_name, ")"); } return node_name; } /*! * \internal * \brief Print a negative location constraint * * \param[in] stream File stream to display output to * \param[in] node Node affected by constraint * \param[in] location Constraint to print */ static void print_ban(FILE *stream, pe_node_t *node, pe__location_t *location) { char *node_name = NULL; switch (output_format) { case mon_output_plain: case mon_output_console: node_name = get_node_display_name(node); print_as(" %s\tprevents %s from running %son %s\n", location->id, location->rsc_lh->id, ((location->role_filter == RSC_ROLE_MASTER)? "as Master " : ""), node_name); break; case mon_output_html: case mon_output_cgi: node_name = get_node_display_name(node); fprintf(stream, "
  • %s prevents %s from running %son %s
    • \n", location->id, location->rsc_lh->id, ((location->role_filter == RSC_ROLE_MASTER)? "as Master " : ""), node_name); break; case mon_output_xml: fprintf(stream, " \n", location->id, location->rsc_lh->id, node->details->uname, node->weight, ((location->role_filter == RSC_ROLE_MASTER)? "true" : "false")); break; default: break; } free(node_name); } /*! * \internal * \brief Print section for negative location constraints * * \param[in] stream File stream to display output to * \param[in] data_set Working set corresponding to CIB status to display */ static void print_neg_locations(FILE *stream, pe_working_set_t *data_set) { GListPtr gIter, gIter2; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nNegative Location Constraints:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Negative Location Constraints

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each ban */ for (gIter = data_set->placement_constraints; gIter != NULL; gIter = gIter->next) { pe__location_t *location = gIter->data; if (!g_str_has_prefix(location->id, print_neg_location_prefix)) continue; for (gIter2 = location->node_list_rh; gIter2 != NULL; gIter2 = gIter2->next) { node_t *node = (node_t *) gIter2->data; if (node->weight < 0) { print_ban(stream, node, location); } } } /* Close section */ switch (output_format) { case mon_output_cgi: case mon_output_html: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } static void crm_mon_get_parameters(resource_t *rsc, pe_working_set_t * data_set) { get_rsc_attributes(rsc->parameters, rsc, NULL, data_set); crm_trace("Beekhof: unpacked params for %s (%d)", rsc->id, g_hash_table_size(rsc->parameters)); if(rsc->children) { GListPtr gIter = NULL; for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) { crm_mon_get_parameters(gIter->data, data_set); } } } /*! * \internal * \brief Print node attributes section * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_node_attributes(FILE *stream, pe_working_set_t *data_set) { GListPtr gIter = NULL; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nNode Attributes:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Node Attributes

    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Unpack all resource parameters (it would be more efficient to do this * only when needed for the first time in print_attr_msg()) */ for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { crm_mon_get_parameters(gIter->data, data_set); } /* Display each node's attributes */ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { struct mon_attr_data data; data.stream = stream; data.node = (node_t *) gIter->data; if (data.node && data.node->details && data.node->details->online) { print_node_start(stream, data.node); g_hash_table_foreach(data.node->details->attrs, create_attr_list, NULL); g_list_foreach(attr_list, print_node_attribute, &data); g_list_free(attr_list); attr_list = NULL; print_node_end(stream); } } /* Print section footer */ switch (output_format) { case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Return resource display options corresponding to command-line choices * * \return Bitmask of pe_print_options suitable for resource print functions */ static int get_resource_display_options(void) { int print_opts; /* Determine basic output format */ switch (output_format) { case mon_output_console: print_opts = pe_print_ncurses; break; case mon_output_html: case mon_output_cgi: print_opts = pe_print_html; break; case mon_output_xml: print_opts = pe_print_xml; break; default: print_opts = pe_print_printf; break; } /* Add optional display elements */ if (print_pending) { print_opts |= pe_print_pending; } if (print_clone_detail) { print_opts |= pe_print_clone_details|pe_print_implicit; } if (!inactive_resources) { print_opts |= pe_print_clone_active; } if (print_brief) { print_opts |= pe_print_brief; } return print_opts; } /*! * \internal * \brief Print header for cluster summary if needed * * \param[in] stream File stream to display output to */ static void print_cluster_summary_header(FILE *stream) { switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "

    Cluster Summary

    \n

    \n"); break; case mon_output_xml: fprintf(stream, "

    \n"); break; default: break; } } /*! * \internal * \brief Print footer for cluster summary if needed * * \param[in] stream File stream to display output to */ static void print_cluster_summary_footer(FILE *stream) { switch (output_format) { case mon_output_cgi: case mon_output_html: fprintf(stream, "

    \n"); break; case mon_output_xml: fprintf(stream, "     \n"); break; default: break; } } /*! * \internal * \brief Print times the display was last updated and CIB last changed * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_times(FILE *stream, pe_working_set_t *data_set) { const char *last_written = crm_element_value(data_set->input, XML_CIB_ATTR_WRITTEN); const char *user = crm_element_value(data_set->input, XML_ATTR_UPDATE_USER); const char *client = crm_element_value(data_set->input, XML_ATTR_UPDATE_CLIENT); const char *origin = crm_element_value(data_set->input, XML_ATTR_UPDATE_ORIG); switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Last updated: %s", crm_now_string()); print_as((user || client || origin)? "\n" : "\t\t"); print_as("Last change: %s", last_written ? last_written : ""); if (user) { print_as(" by %s", user); } if (client) { print_as(" via %s", client); } if (origin) { print_as(" on %s", origin); } print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Last updated: %s
    \n", crm_now_string()); fprintf(stream, " Last change: %s", last_written ? last_written : ""); if (user) { fprintf(stream, " by %s", user); } if (client) { fprintf(stream, " via %s", client); } if (origin) { fprintf(stream, " on %s", origin); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n", crm_now_string()); fprintf(stream, " \n", last_written ? last_written : "", user ? user : "", client ? client : "", origin ? origin : ""); break; default: break; } } /*! * \internal * \brief Print cluster stack * * \param[in] stream File stream to display output to * \param[in] stack_s Stack name */ static void print_cluster_stack(FILE *stream, const char *stack_s) { switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Stack: %s\n", stack_s); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Stack: %s
    \n", stack_s); break; case mon_output_xml: fprintf(stream, " \n", stack_s); break; default: break; } } /*! * \internal * \brief Print current DC and its version * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_dc(FILE *stream, pe_working_set_t *data_set) { node_t *dc = data_set->dc_node; xmlNode *dc_version = get_xpath_object("//nvpair[@name='dc-version']", data_set->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, XML_NVPAIR_ATTR_VALUE) : NULL; const char *quorum = crm_element_value(data_set->input, XML_ATTR_HAVE_QUORUM); char *dc_name = dc? get_node_display_name(dc) : NULL; switch (output_format) { case mon_output_plain: case mon_output_console: print_as("Current DC: "); if (dc) { print_as("%s (version %s) - partition %s quorum\n", dc_name, (dc_version_s? dc_version_s : "unknown"), (crm_is_true(quorum) ? "with" : "WITHOUT")); } else { print_as("NONE\n"); } break; case mon_output_html: case mon_output_cgi: fprintf(stream, " Current DC: "); if (dc) { fprintf(stream, "%s (version %s) - partition %s quorum", dc_name, (dc_version_s? dc_version_s : "unknown"), (crm_is_true(quorum)? "with" : "WITHOUT")); } else { fprintf(stream, "NONE"); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " details->uname, dc->details->id, (crm_is_true(quorum) ? "true" : "false")); } else { fprintf(stream, "present=\"false\""); } fprintf(stream, " />\n"); break; default: break; } free(dc_name); } /*! * \internal * \brief Print counts of configured nodes and resources * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state * \param[in] stack_s Stack name */ static void print_cluster_counts(FILE *stream, pe_working_set_t *data_set, const char *stack_s) { int nnodes = g_list_length(data_set->nodes); int nresources = count_resources(data_set, NULL); switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\n%d node%s configured\n", nnodes, s_if_plural(nnodes)); print_as("%d resource%s configured", nresources, s_if_plural(nresources)); if(data_set->disabled_resources || data_set->blocked_resources) { print_as(" ("); if (data_set->disabled_resources) { print_as("%d DISABLED", data_set->disabled_resources); } if (data_set->disabled_resources && data_set->blocked_resources) { print_as(", "); } if (data_set->blocked_resources) { print_as("%d BLOCKED from starting due to failure", data_set->blocked_resources); } print_as(")"); } print_as("\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, " %d node%s configured
    \n", nnodes, s_if_plural(nnodes)); fprintf(stream, " %d resource%s configured", nresources, s_if_plural(nresources)); if (data_set->disabled_resources || data_set->blocked_resources) { fprintf(stream, " ("); if (data_set->disabled_resources) { fprintf(stream, "%d DISABLED", data_set->disabled_resources); } if (data_set->disabled_resources && data_set->blocked_resources) { fprintf(stream, ", "); } if (data_set->blocked_resources) { fprintf(stream, "%d BLOCKED from starting due to failure", data_set->blocked_resources); } fprintf(stream, ")"); } fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n", g_list_length(data_set->nodes)); fprintf(stream, " \n", count_resources(data_set, NULL), data_set->disabled_resources, data_set->blocked_resources); break; default: break; } } /*! * \internal * \brief Print cluster-wide options * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state * * \note Currently this is only implemented for HTML and XML output, and * prints only a few options. If there is demand, more could be added. */ static void print_cluster_options(FILE *stream, pe_working_set_t *data_set) { switch (output_format) { case mon_output_plain: case mon_output_console: if (is_set(data_set->flags, pe_flag_maintenance_mode)) { print_as("\n *** Resource management is DISABLED ***"); print_as("\n The cluster will not attempt to start, stop or recover services"); print_as("\n"); } break; case mon_output_html: fprintf(stream, "

    \n

    Config Options

    \n"); fprintf(stream, " \n"); fprintf(stream, " \n", is_set(data_set->flags, pe_flag_stonith_enabled)? "enabled" : "disabled"); fprintf(stream, " \n", is_set(data_set->flags, pe_flag_symmetric_cluster)? "" : "a"); fprintf(stream, " \n"); fprintf(stream, " \n"); fprintf(stream, "
    STONITH of failed nodes%s
    Cluster is%ssymmetric
    No Quorum Policy"); switch (data_set->no_quorum_policy) { case no_quorum_freeze: fprintf(stream, "Freeze resources"); break; case no_quorum_stop: fprintf(stream, "Stop ALL resources"); break; case no_quorum_ignore: fprintf(stream, "Ignore"); break; case no_quorum_suicide: fprintf(stream, "Suicide"); break; } fprintf(stream, "
    Resource management"); if (is_set(data_set->flags, pe_flag_maintenance_mode)) { fprintf(stream, "DISABLED (the cluster will " "not attempt to start, stop or recover services)"); } else { fprintf(stream, "enabled"); } fprintf(stream, "
    \n

    \n"); break; case mon_output_xml: fprintf(stream, " flags, pe_flag_stonith_enabled)? "true" : "false"); fprintf(stream, " symmetric-cluster=\"%s\"", is_set(data_set->flags, pe_flag_symmetric_cluster)? "true" : "false"); fprintf(stream, " no-quorum-policy=\""); switch (data_set->no_quorum_policy) { case no_quorum_freeze: fprintf(stream, "freeze"); break; case no_quorum_stop: fprintf(stream, "stop"); break; case no_quorum_ignore: fprintf(stream, "ignore"); break; case no_quorum_suicide: fprintf(stream, "suicide"); break; } fprintf(stream, "\""); fprintf(stream, " maintenance-mode=\"%s\"", is_set(data_set->flags, pe_flag_maintenance_mode)? "true" : "false"); fprintf(stream, " />\n"); break; default: break; } } /*! * \internal * \brief Get the name of the stack in use (or "unknown" if not available) * * \param[in] data_set Working set of CIB state * * \return String representing stack name */ static const char * get_cluster_stack(pe_working_set_t *data_set) { xmlNode *stack = get_xpath_object("//nvpair[@name='cluster-infrastructure']", data_set->input, LOG_DEBUG); return stack? crm_element_value(stack, XML_NVPAIR_ATTR_VALUE) : "unknown"; } /*! * \internal * \brief Print a summary of cluster-wide information * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_cluster_summary(FILE *stream, pe_working_set_t *data_set) { const char *stack_s = get_cluster_stack(data_set); gboolean header_printed = FALSE; if (show & mon_show_stack) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_stack(stream, stack_s); } /* Always print DC if none, even if not requested */ if ((data_set->dc_node == NULL) || (show & mon_show_dc)) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_dc(stream, data_set); } if (show & mon_show_times) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_times(stream, data_set); } if (is_set(data_set->flags, pe_flag_maintenance_mode) || data_set->disabled_resources || data_set->blocked_resources || is_set(show, mon_show_count)) { if (header_printed == FALSE) { print_cluster_summary_header(stream); header_printed = TRUE; } print_cluster_counts(stream, data_set, stack_s); } /* There is not a separate option for showing cluster options, so show with * stack for now; a separate option could be added if there is demand */ if (show & mon_show_stack) { print_cluster_options(stream, data_set); } if (header_printed) { print_cluster_summary_footer(stream); } } /*! * \internal * \brief Print a failed action * * \param[in] stream File stream to display output to * \param[in] xml_op Root of XML tree describing failed action */ static void print_failed_action(FILE *stream, xmlNode *xml_op) { const char *op_key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); const char *op_key_attr = "op_key"; const char *last = crm_element_value(xml_op, XML_RSC_OP_LAST_CHANGE); const char *node = crm_element_value(xml_op, XML_ATTR_UNAME); const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); int rc = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), "0"); int status = crm_parse_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), "0"); char *exit_reason_cleaned; /* If no op_key was given, use id instead */ if (op_key == NULL) { op_key = ID(xml_op); op_key_attr = "id"; } /* If no exit reason was given, use "none" */ if (exit_reason == NULL) { exit_reason = "none"; } /* Print common action information */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("* %s on %s '%s' (%d): call=%s, status=%s, exitreason='%s'", op_key, node, services_ocf_exitcode_str(rc), rc, call, services_lrm_status_str(status), exit_reason); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "

  • %s on %s '%s' (%d): call=%s, status=%s, exitreason='%s'", op_key, node, services_ocf_exitcode_str(rc), rc, call, services_lrm_status_str(status), exit_reason); break; case mon_output_xml: exit_reason_cleaned = crm_xml_escape(exit_reason); fprintf(stream, " \n"); break; case mon_output_xml: fprintf(stream, " />\n"); break; default: break; } } /*! * \internal * \brief Print a section for failed actions * * \param[in] stream File stream to display output to * \param[in] data_set Working set of CIB state */ static void print_failed_actions(FILE *stream, pe_working_set_t *data_set) { xmlNode *xml_op = NULL; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nFailed Resource Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Failed Resource Actions

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } /* Print each failed action */ for (xml_op = __xml_first_child(data_set->failed); xml_op != NULL; xml_op = __xml_next(xml_op)) { print_failed_action(stream, xml_op); } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Reduce the stonith-history * for successful actions we keep the last of every action-type & target * for failed actions we record as well who had failed * for actions in progress we keep full track * * \param[in] history List of stonith actions * */ static stonith_history_t * reduce_stonith_history(stonith_history_t *history) { stonith_history_t *new = NULL, *hp, *np, *tmp; for (hp = history; hp; ) { for (np = new; np; np = np->next) { if ((hp->state == st_done) || (hp->state == st_failed)) { /* action not in progress */ if (safe_str_eq(hp->target, np->target) && safe_str_eq(hp->action, np->action) && (hp->state == np->state)) { if ((hp->state == st_done) || safe_str_eq(hp->delegate, np->delegate)) { /* replace or purge */ if (hp->completed < np->completed) { /* purge older hp */ tmp = hp->next; hp->next = NULL; stonith_history_free(hp); hp = tmp; break; } /* damn single linked list */ free(hp->target); free(hp->action); free(np->origin); np->origin = hp->origin; free(np->delegate); np->delegate = hp->delegate; free(np->client); np->client = hp->client; np->completed = hp->completed; tmp = hp; hp = hp->next; free(tmp); break; } } if (np->next) { continue; } } np = 0; /* let outer loop progress hp */ break; } /* simply move hp from history to new */ if (np == NULL) { tmp = hp->next; hp->next = new; new = hp; hp = tmp; } } return new; } /*! * \internal * \brief Sort the stonith-history * sort by competed most current on the top * pending actions lacking a completed-stamp are gathered at the top * * \param[in] history List of stonith actions * */ static stonith_history_t * sort_stonith_history(stonith_history_t *history) { stonith_history_t *new = NULL, *pending = NULL, *hp, *np, *tmp; for (hp = history; hp; ) { tmp = hp->next; if ((hp->state == st_done) || (hp->state == st_failed)) { /* sort into new */ if ((!new) || (hp->completed > new->completed)) { hp->next = new; new = hp; } else { np = new; do { if ((!np->next) || (hp->completed > np->next->completed)) { hp->next = np->next; np->next = hp; break; } np = np->next; } while (1); } } else { /* put into pending */ hp->next = pending; pending = hp; } hp = tmp; } /* pending actions don't have a completed-stamp so make them go front */ if (pending) { stonith_history_t *last_pending = pending; while (last_pending->next) { last_pending = last_pending->next; } last_pending->next = new; new = pending; } return new; } /*! * \internal * \brief Print a stonith action * * \param[in] stream File stream to display output to * \param[in] event stonith event */ static void print_stonith_action(FILE *stream, stonith_history_t *event) { const char *action_s = stonith_action_str(event->action); char *run_at_s = ctime(&event->completed); if ((run_at_s) && (run_at_s[0] != 0)) { run_at_s[strlen(run_at_s)-1] = 0; /* Overwrite the newline */ } switch(output_format) { case mon_output_xml: fprintf(stream, " target, event->action); switch(event->state) { case st_done: fprintf(stream, " state=\"success\""); break; case st_failed: fprintf(stream, " state=\"failed\""); break; default: fprintf(stream, " state=\"pending\""); } fprintf(stream, " origin=\"%s\" client=\"%s\"", event->origin, event->client); if (event->delegate) { fprintf(stream, " delegate=\"%s\"", event->delegate); } switch(event->state) { case st_done: case st_failed: fprintf(stream, " completed=\"%s\"", run_at_s?run_at_s:""); break; default: break; } fprintf(stream, " />\n"); break; case mon_output_plain: case mon_output_console: switch(event->state) { case st_done: print_as("* %s of %s successful: delegate=%s, client=%s, origin=%s,\n" " %s='%s'\n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-successful", run_at_s?run_at_s:""); break; case st_failed: print_as("* %s of %s failed: delegate=%s, client=%s, origin=%s,\n" " %s='%s'\n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-failed", run_at_s?run_at_s:""); break; default: print_as("* %s of %s pending: client=%s, origin=%s\n", action_s, event->target, event->client, event->origin); } break; case mon_output_html: case mon_output_cgi: switch(event->state) { case st_done: fprintf(stream, "
  • %s of %s successful: delegate=%s, " "client=%s, origin=%s, %s='%s'
    • \n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-successful", run_at_s?run_at_s:""); break; case st_failed: fprintf(stream, "
  • %s of %s failed: delegate=%s, " "client=%s, origin=%s, %s='%s'
    • \n", action_s, event->target, event->delegate ? event->delegate : "", event->client, event->origin, fence_full_history?"completed":"last-failed", run_at_s?run_at_s:""); break; default: fprintf(stream, "
  • %s of %s pending: client=%s, " "origin=%s
    • \n", action_s, event->target, event->client, event->origin); } break; default: /* no support for fence history for other formats so far */ break; } } /*! * \internal * \brief Print a section for failed stonith actions * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_failed_stonith_actions(FILE *stream, stonith_history_t *history) { stonith_history_t *hp; for (hp = history; hp; hp = hp->next) { if (hp->state == st_failed) { break; } } if (!hp) { return; } /* Print section heading */ switch (output_format) { /* no need to take care of xml in here as xml gets full * history anyway */ case mon_output_plain: case mon_output_console: print_as("\nFailed Fencing Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Failed Fencing Actions

    \n
      \n"); break; default: break; } /* Print each failed stonith action */ for (hp = history; hp; hp = hp->next) { if (hp->state == st_failed) { print_stonith_action(stream, hp); } } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; default: break; } } /*! * \internal * \brief Print pending stonith actions * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_stonith_pending(FILE *stream, stonith_history_t *history) { /* xml-output always shows the full history * so we'll never have to show pending-actions * separately */ if (history && (history->state != st_failed) && (history->state != st_done)) { stonith_history_t *hp; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nPending Fencing Actions:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Pending Fencing Actions

    \n
      \n"); break; default: break; } for (hp = history; hp; hp = hp->next) { if ((hp->state == st_failed) || (hp->state == st_done)) { break; } print_stonith_action(stream, hp); } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; default: break; } } } /*! * \internal * \brief Print a section for stonith-history * * \param[in] stream File stream to display output to * \param[in] history List of stonith actions * */ static void print_stonith_history(FILE *stream, stonith_history_t *history) { stonith_history_t *hp; /* Print section heading */ switch (output_format) { case mon_output_plain: case mon_output_console: print_as("\nFencing History:\n"); break; case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n

    Fencing History

    \n
      \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } for (hp = history; hp; hp = hp->next) { if ((hp->state != st_failed) || (output_format == mon_output_xml)) { print_stonith_action(stream, hp); } } /* End section */ switch (output_format) { case mon_output_html: case mon_output_cgi: fprintf(stream, "
    \n"); break; case mon_output_xml: fprintf(stream, " \n"); break; default: break; } } /*! * \internal * \brief Print cluster status to screen * * This uses the global display preferences set by command-line options * to display cluster status in a human-friendly way. * * \param[in] data_set Working set of CIB state * \param[in] stonith_history List of stonith actions */ static void print_status(pe_working_set_t * data_set, stonith_history_t *stonith_history) { GListPtr gIter = NULL; int print_opts = get_resource_display_options(); /* space-separated lists of node names */ char *online_nodes = NULL; char *online_remote_nodes = NULL; char *online_guest_nodes = NULL; char *offline_nodes = NULL; char *offline_remote_nodes = NULL; if (output_format == mon_output_console) { blank_screen(); } print_cluster_summary(stdout, data_set); print_as("\n"); /* Gather node information (and print if in bad state or grouping by node) */ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_mode = NULL; char *node_name = get_node_display_name(node); /* Get node mode */ if (node->details->unclean) { if (node->details->online) { node_mode = "UNCLEAN (online)"; } else if (node->details->pending) { node_mode = "UNCLEAN (pending)"; } else { node_mode = "UNCLEAN (offline)"; } } else if (node->details->pending) { node_mode = "pending"; } else if (node->details->standby_onfail && node->details->online) { node_mode = "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { if (node->details->running_rsc) { node_mode = "standby (with active resources)"; } else { node_mode = "standby"; } } else { node_mode = "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { node_mode = "maintenance"; } else { node_mode = "OFFLINE (maintenance)"; } } else if (node->details->online) { node_mode = "online"; if (group_by_node == FALSE) { if (pe__is_guest_node(node)) { online_guest_nodes = add_list_element(online_guest_nodes, node_name); } else if (pe__is_remote_node(node)) { online_remote_nodes = add_list_element(online_remote_nodes, node_name); } else { online_nodes = add_list_element(online_nodes, node_name); } free(node_name); continue; } } else { node_mode = "OFFLINE"; if (group_by_node == FALSE) { if (pe__is_remote_node(node)) { offline_remote_nodes = add_list_element(offline_remote_nodes, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline guest nodes */ } else { offline_nodes = add_list_element(offline_nodes, node_name); } free(node_name); continue; } } /* If we get here, node is in bad state, or we're grouping by node */ /* Print the node name and status */ if (pe__is_guest_node(node)) { print_as("Guest"); } else if (pe__is_remote_node(node)) { print_as("Remote"); } print_as("Node %s: %s\n", node_name, node_mode); /* If we're grouping by node, print its resources */ if (group_by_node) { if (print_brief) { print_rscs_brief(node->details->running_rsc, "\t", print_opts | pe_print_rsconly, stdout, FALSE); } else { GListPtr gIter2 = NULL; for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) { resource_t *rsc = (resource_t *) gIter2->data; rsc->fns->print(rsc, "\t", print_opts | pe_print_rsconly, stdout); } } } free(node_name); } /* If we're not grouping by node, summarize nodes by status */ if (online_nodes) { print_as("Online: [%s ]\n", online_nodes); free(online_nodes); } if (offline_nodes) { print_as("OFFLINE: [%s ]\n", offline_nodes); free(offline_nodes); } if (online_remote_nodes) { print_as("RemoteOnline: [%s ]\n", online_remote_nodes); free(online_remote_nodes); } if (offline_remote_nodes) { print_as("RemoteOFFLINE: [%s ]\n", offline_remote_nodes); free(offline_remote_nodes); } if (online_guest_nodes) { print_as("GuestOnline: [%s ]\n", online_guest_nodes); free(online_guest_nodes); } /* Print resources section, if needed */ print_resources(stdout, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stdout, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stdout, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stdout, data_set); } /* Print failed stonith actions */ if (fence_history) { print_failed_stonith_actions(stdout, stonith_history); } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stdout, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stdout, data_set); } /* Print stonith history */ if (fence_history) { if (show & mon_show_fence_history) { print_stonith_history(stdout, stonith_history); } else { print_stonith_pending(stdout, stonith_history); } } #if CURSES_ENABLED if (output_format == mon_output_console) { refresh(); } #endif } /*! * \internal * \brief Print cluster status in XML format * * \param[in] data_set Working set of CIB state */ static void print_xml_status(pe_working_set_t * data_set, stonith_history_t *stonith_history) { FILE *stream = stdout; GListPtr gIter = NULL; int print_opts = get_resource_display_options(); fprintf(stream, "\n"); fprintf(stream, "\n", VERSION); print_cluster_summary(stream, data_set); /*** NODES ***/ fprintf(stream, " \n"); for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; const char *node_type = "unknown"; switch (node->details->type) { case node_member: node_type = "member"; break; case node_remote: node_type = "remote"; break; case node_ping: node_type = "ping"; break; } fprintf(stream, " details->uname); fprintf(stream, "id=\"%s\" ", node->details->id); fprintf(stream, "online=\"%s\" ", node->details->online ? "true" : "false"); fprintf(stream, "standby=\"%s\" ", node->details->standby ? "true" : "false"); fprintf(stream, "standby_onfail=\"%s\" ", node->details->standby_onfail ? "true" : "false"); fprintf(stream, "maintenance=\"%s\" ", node->details->maintenance ? "true" : "false"); fprintf(stream, "pending=\"%s\" ", node->details->pending ? "true" : "false"); fprintf(stream, "unclean=\"%s\" ", node->details->unclean ? "true" : "false"); fprintf(stream, "shutdown=\"%s\" ", node->details->shutdown ? "true" : "false"); fprintf(stream, "expected_up=\"%s\" ", node->details->expected_up ? "true" : "false"); fprintf(stream, "is_dc=\"%s\" ", node->details->is_dc ? "true" : "false"); fprintf(stream, "resources_running=\"%d\" ", g_list_length(node->details->running_rsc)); fprintf(stream, "type=\"%s\" ", node_type); if (pe__is_guest_node(node)) { fprintf(stream, "id_as_resource=\"%s\" ", node->details->remote_rsc->container->id); } if (group_by_node) { GListPtr lpc2 = NULL; fprintf(stream, ">\n"); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { resource_t *rsc = (resource_t *) lpc2->data; rsc->fns->print(rsc, " ", print_opts | pe_print_rsconly, stream); } fprintf(stream, " \n"); } else { fprintf(stream, "/>\n"); } } fprintf(stream, " \n"); /* Print resources section, if needed */ print_resources(stream, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stream, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stream, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stream, data_set); } /* Print stonith history */ if (fence_history) { print_stonith_history(stdout, stonith_history); } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stream, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stream, data_set); } fprintf(stream, "\n"); fflush(stream); fclose(stream); } /*! * \internal * \brief Print cluster status in HTML format (with HTTP headers if CGI) * * \param[in] data_set Working set of CIB state * \param[in] filename Name of file to write HTML to (ignored if CGI) * * \return 0 on success, -1 on error */ static int print_html_status(pe_working_set_t * data_set, const char *filename, stonith_history_t *stonith_history) { FILE *stream; GListPtr gIter = NULL; char *filename_tmp = NULL; int print_opts = get_resource_display_options(); if (output_format == mon_output_cgi) { stream = stdout; fprintf(stream, "Content-Type: text/html\n\n"); } else { filename_tmp = crm_concat(filename, "tmp", '.'); stream = fopen(filename_tmp, "w"); if (stream == NULL) { crm_perror(LOG_ERR, "Cannot open %s for writing", filename_tmp); free(filename_tmp); return -1; } } fprintf(stream, "\n"); fprintf(stream, " \n"); fprintf(stream, " Cluster status\n"); fprintf(stream, " \n", reconnect_msec / 1000); fprintf(stream, " \n"); fprintf(stream, "\n"); print_cluster_summary(stream, data_set); /*** NODE LIST ***/ fprintf(stream, "
    \n

    Node List

    \n"); fprintf(stream, "
      \n"); for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { node_t *node = (node_t *) gIter->data; char *node_name = get_node_display_name(node); fprintf(stream, "
    • Node: %s: ", node_name); if (node->details->standby_onfail && node->details->online) { fprintf(stream, "standby (on-fail)\n"); } else if (node->details->standby && node->details->online) { fprintf(stream, "standby%s\n", node->details->running_rsc?" (with active resources)":""); } else if (node->details->standby) { fprintf(stream, "OFFLINE (standby)\n"); } else if (node->details->maintenance && node->details->online) { fprintf(stream, "maintenance\n"); } else if (node->details->maintenance) { fprintf(stream, "OFFLINE (maintenance)\n"); } else if (node->details->online) { fprintf(stream, "online\n"); } else { fprintf(stream, "OFFLINE\n"); } if (print_brief && group_by_node) { fprintf(stream, "
        \n"); print_rscs_brief(node->details->running_rsc, NULL, print_opts | pe_print_rsconly, stream, FALSE); fprintf(stream, "
      \n"); } else if (group_by_node) { GListPtr lpc2 = NULL; fprintf(stream, "
        \n"); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { resource_t *rsc = (resource_t *) lpc2->data; fprintf(stream, "
      • "); rsc->fns->print(rsc, NULL, print_opts | pe_print_rsconly, stream); fprintf(stream, "
        • \n"); } fprintf(stream, "
      \n"); } fprintf(stream, "
      • \n"); free(node_name); } fprintf(stream, "
    \n"); /* Print resources section, if needed */ print_resources(stream, data_set, print_opts); /* print Node Attributes section if requested */ if (show & mon_show_attributes) { print_node_attributes(stream, data_set); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (show & (mon_show_operations | mon_show_failcounts)) { print_node_summary(stream, data_set, ((show & mon_show_operations)? TRUE : FALSE)); } /* If there were any failed actions, print them */ if (xml_has_children(data_set->failed)) { print_failed_actions(stream, data_set); } /* Print failed stonith actions */ if (fence_history) { print_failed_stonith_actions(stream, stonith_history); } /* Print stonith history */ if (fence_history) { if (show & mon_show_fence_history) { print_stonith_history(stream, stonith_history); } else { print_stonith_pending(stdout, stonith_history); } } /* Print tickets if requested */ if (show & mon_show_tickets) { print_cluster_tickets(stream, data_set); } /* Print negative location constraints if requested */ if (show & mon_show_bans) { print_neg_locations(stream, data_set); } fprintf(stream, "\n"); fprintf(stream, "\n"); fflush(stream); fclose(stream); if (output_format != mon_output_cgi) { if (rename(filename_tmp, filename) != 0) { crm_perror(LOG_ERR, "Unable to rename %s->%s", filename_tmp, filename); } free(filename_tmp); } return 0; } static int send_custom_trap(const char *node, const char *rsc, const char *task, int target_rc, int rc, int status, const char *desc) { pid_t pid; /*setenv needs chars, these are ints */ char *rc_s = crm_itoa(rc); char *status_s = crm_itoa(status); char *target_rc_s = crm_itoa(target_rc); crm_debug("Sending external notification to '%s' via '%s'", external_recipient, external_agent); if(rsc) { setenv("CRM_notify_rsc", rsc, 1); } if (external_recipient) { setenv("CRM_notify_recipient", external_recipient, 1); } setenv("CRM_notify_node", node, 1); setenv("CRM_notify_task", task, 1); setenv("CRM_notify_desc", desc, 1); setenv("CRM_notify_rc", rc_s, 1); setenv("CRM_notify_target_rc", target_rc_s, 1); setenv("CRM_notify_status", status_s, 1); pid = fork(); if (pid == -1) { crm_perror(LOG_ERR, "notification fork() failed."); } if (pid == 0) { /* crm_debug("notification: I am the child. Executing the nofitication program."); */ execl(external_agent, external_agent, NULL); exit(CRM_EX_ERROR); } crm_trace("Finished running custom notification program '%s'.", external_agent); free(target_rc_s); free(status_s); free(rc_s); return 0; } static void handle_rsc_op(xmlNode * xml, const char *node_id) { int rc = -1; int status = -1; int target_rc = -1; gboolean notify = TRUE; char *rsc = NULL; char *task = NULL; const char *desc = NULL; const char *magic = NULL; const char *id = NULL; const char *node = NULL; xmlNode *n = xml; xmlNode * rsc_op = xml; if(strcmp((const char*)xml->name, XML_LRM_TAG_RSC_OP) != 0) { xmlNode *cIter; for(cIter = xml->children; cIter; cIter = cIter->next) { handle_rsc_op(cIter, node_id); } return; } id = crm_element_value(rsc_op, XML_LRM_ATTR_TASK_KEY); if (id == NULL) { /* Compatibility with <= 1.1.5 */ id = ID(rsc_op); } magic = crm_element_value(rsc_op, XML_ATTR_TRANSITION_MAGIC); if (magic == NULL) { /* non-change */ return; } if (!decode_transition_magic(magic, NULL, NULL, NULL, &status, &rc, &target_rc)) { crm_err("Invalid event %s detected for %s", magic, id); return; } if (parse_op_key(id, &rsc, &task, NULL) == FALSE) { crm_err("Invalid event detected for %s", id); goto bail; } node = crm_element_value(rsc_op, XML_LRM_ATTR_TARGET); while (n != NULL && safe_str_neq(XML_CIB_TAG_STATE, TYPE(n))) { n = n->parent; } if(node == NULL && n) { node = crm_element_value(n, XML_ATTR_UNAME); } if (node == NULL && n) { node = ID(n); } if (node == NULL) { node = node_id; } if (node == NULL) { crm_err("No node detected for event %s (%s)", magic, id); goto bail; } /* look up where we expected it to be? */ desc = pcmk_strerror(pcmk_ok); if (status == PCMK_LRM_OP_DONE && target_rc == rc) { crm_notice("%s of %s on %s completed: %s", task, rsc, node, desc); if (rc == PCMK_OCF_NOT_RUNNING) { notify = FALSE; } } else if (status == PCMK_LRM_OP_DONE) { desc = services_ocf_exitcode_str(rc); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } else { desc = services_lrm_status_str(status); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } if (notify && external_agent) { send_custom_trap(node, rsc, task, target_rc, rc, status, desc); } bail: free(rsc); free(task); } static gboolean mon_trigger_refresh(gpointer user_data) { mainloop_set_trigger(refresh_trigger); return FALSE; } #define NODE_PATT "/lrm[@id=" static char * get_node_from_xpath(const char *xpath) { char *nodeid = NULL; char *tmp = strstr(xpath, NODE_PATT); if(tmp) { tmp += strlen(NODE_PATT); tmp += 1; nodeid = strdup(tmp); tmp = strstr(nodeid, "\'"); CRM_ASSERT(tmp); tmp[0] = 0; } return nodeid; } static void crm_diff_update_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) { const char *name = NULL; const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); xmlNode *match = NULL; const char *node = NULL; if(op == NULL) { continue; } else if(strcmp(op, "create") == 0) { match = change->children; } else if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "delete") == 0) { continue; } else if(strcmp(op, "modify") == 0) { match = first_named_child(change, XML_DIFF_RESULT); if(match) { match = match->children; } } if(match) { name = (const char *)match->name; } crm_trace("Handling %s operation for %s %p, %s", op, xpath, match, name); if(xpath == NULL) { /* Version field, ignore */ } else if(name == NULL) { crm_debug("No result for %s operation to %s", op, xpath); CRM_ASSERT(strcmp(op, "delete") == 0 || strcmp(op, "move") == 0); } else if(strcmp(name, XML_TAG_CIB) == 0) { xmlNode *state = NULL; xmlNode *status = first_named_child(match, XML_CIB_TAG_STATUS); for (state = __xml_first_child(status); state != NULL; state = __xml_next(state)) { node = crm_element_value(state, XML_ATTR_UNAME); if (node == NULL) { node = ID(state); } handle_rsc_op(state, node); } } else if(strcmp(name, XML_CIB_TAG_STATUS) == 0) { xmlNode *state = NULL; for (state = __xml_first_child(match); state != NULL; state = __xml_next(state)) { node = crm_element_value(state, XML_ATTR_UNAME); if (node == NULL) { node = ID(state); } handle_rsc_op(state, node); } } else if(strcmp(name, XML_CIB_TAG_STATE) == 0) { node = crm_element_value(match, XML_ATTR_UNAME); if (node == NULL) { node = ID(match); } handle_rsc_op(match, node); } else if(strcmp(name, XML_CIB_TAG_LRM) == 0) { node = ID(match); handle_rsc_op(match, node); } else if(strcmp(name, XML_LRM_TAG_RESOURCES) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = get_node_from_xpath(xpath); handle_rsc_op(match, local_node); free(local_node); } else { crm_trace("Ignoring %s operation for %s %p, %s", op, xpath, match, name); } } } static void crm_diff_update_v1(const char *event, xmlNode * msg) { /* Process operation updates */ xmlXPathObject *xpathObj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_LRM_TAG_RSC_OP); int lpc = 0, max = numXpathResults(xpathObj); for (lpc = 0; lpc < max; lpc++) { xmlNode *rsc_op = getXpathResult(xpathObj, lpc); handle_rsc_op(rsc_op, NULL); } freeXpathObject(xpathObj); } static void crm_diff_update(const char *event, xmlNode * msg) { int rc = -1; static bool stale = FALSE; gboolean cib_updated = FALSE; xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT); print_dot(); if (current_cib != NULL) { rc = xml_apply_patchset(current_cib, diff, TRUE); switch (rc) { case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(current_cib); current_cib = NULL; break; case pcmk_ok: cib_updated = TRUE; break; default: crm_notice("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(current_cib); current_cib = NULL; } } if (current_cib == NULL) { crm_trace("Re-requesting the full cib"); cib->cmds->query(cib, NULL, ¤t_cib, cib_scope_local | cib_sync_call); } if (external_agent) { int format = 0; crm_element_value_int(diff, "format", &format); switch(format) { case 1: crm_diff_update_v1(event, msg); break; case 2: crm_diff_update_v2(event, msg); break; default: crm_err("Unknown patch format: %d", format); } } if (current_cib == NULL) { if(!stale) { print_as("--- Stale data ---"); } stale = TRUE; return; } stale = FALSE; kick_refresh(cib_updated); } static gboolean mon_refresh_display(gpointer user_data) { xmlNode *cib_copy = copy_xml(current_cib); stonith_history_t *stonith_history = NULL; last_refresh = time(NULL); if (cli_config_update(&cib_copy, NULL, FALSE) == FALSE) { if (cib) { cib->cmds->signoff(cib); } print_as("Upgrade failed: %s", pcmk_strerror(-pcmk_err_schema_validation)); if (output_format == mon_output_console) { sleep(2); } clean_up(CRM_EX_CONFIG); return FALSE; } /* get the stonith-history if there is evidence we need it */ while (fence_history) { if (st != NULL) { if (st->cmds->history(st, st_opt_sync_call, NULL, &stonith_history, 120)) { fprintf(stderr, "Critical: Unable to get stonith-history\n"); mon_cib_connection_destroy(NULL); } else { if ((!fence_full_history) && (output_format != mon_output_xml)) { stonith_history = reduce_stonith_history(stonith_history); } stonith_history = sort_stonith_history(stonith_history); break; /* all other cases are errors */ } } else { fprintf(stderr, "Critical: No stonith-API\n"); } free_xml(cib_copy); print_as("Reading stonith-history failed"); if (output_format == mon_output_console) { sleep(2); } return FALSE; } if (mon_data_set == NULL) { mon_data_set = pe_new_working_set(); CRM_ASSERT(mon_data_set != NULL); } mon_data_set->input = cib_copy; cluster_status(mon_data_set); /* Unpack constraints if any section will need them * (tickets may be referenced in constraints but not granted yet, * and bans need negative location constraints) */ if (show & (mon_show_bans | mon_show_tickets)) { xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, mon_data_set->input); unpack_constraints(cib_constraints, mon_data_set); } switch (output_format) { case mon_output_html: case mon_output_cgi: if (print_html_status(mon_data_set, output_filename, stonith_history) != 0) { fprintf(stderr, "Critical: Unable to output html file\n"); clean_up(CRM_EX_CANTCREAT); return FALSE; } break; case mon_output_xml: print_xml_status(mon_data_set, stonith_history); break; case mon_output_monitor: print_simple_status(mon_data_set, stonith_history); if (has_warnings) { clean_up(MON_STATUS_WARN); return FALSE; } break; case mon_output_plain: case mon_output_console: print_status(mon_data_set, stonith_history); break; case mon_output_none: break; } stonith_history_free(stonith_history); stonith_history = NULL; pe_reset_working_set(mon_data_set); return TRUE; } static void mon_st_callback_event(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else if (external_agent) { char *desc = crm_strdup_printf("Operation %s requested by %s for peer %s: %s (ref=%s)", e->operation, e->origin, e->target, pcmk_strerror(e->result), e->id); send_custom_trap(e->target, NULL, e->operation, pcmk_ok, e->result, 0, desc); free(desc); } } static void kick_refresh(gboolean data_updated) { static int updates = 0; long now = time(NULL); if (data_updated) { updates++; } if(refresh_timer == NULL) { refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL); } /* Refresh * - immediately if the last update was more than 5s ago * - every 10 cib-updates * - at most 2s after the last update */ if ((now - last_refresh) > (reconnect_msec / 1000)) { mainloop_set_trigger(refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else if(updates >= 10) { mainloop_set_trigger(refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else { mainloop_timer_start(refresh_timer); } } static void mon_st_callback_display(stonith_t * st, stonith_event_t * e) { if (st->state == stonith_disconnected) { /* disconnect cib as well and have everything reconnect */ mon_cib_connection_destroy(NULL); } else { print_dot(); kick_refresh(TRUE); } } static void clean_up_connections(void) { if (cib != NULL) { cib->cmds->signoff(cib); cib_delete(cib); cib = NULL; } if (st != NULL) { if (st->state != stonith_disconnected) { st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT); st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE); st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY); st->cmds->disconnect(st); } stonith_api_delete(st); st = NULL; } } /* * De-init ncurses, disconnect from the CIB manager, disconnect fencing, * deallocate memory and show usage-message if requested. * * We don't actually return, but nominally returning crm_exit_t allows a usage * like "return clean_up(exit_code);" which helps static analysis understand the * code flow. */ static crm_exit_t clean_up(crm_exit_t exit_code) { #if CURSES_ENABLED if (curses_console_initialized) { output_format = mon_output_plain; echo(); nocbreak(); endwin(); curses_console_initialized = FALSE; } #endif clean_up_connections(); free(output_filename); free(pid_file); pe_free_working_set(mon_data_set); mon_data_set = NULL; if (exit_code == CRM_EX_USAGE) { if (output_format == mon_output_cgi) { fprintf(stdout, "Content-Type: text/plain\n" "Status: 500\n\n"); } else { crm_help('?', CRM_EX_USAGE); } } crm_exit(exit_code); } diff --git a/tools/stonith_admin.c b/tools/stonith_admin.c index d960fb13e4..6be66c6cbb 100644 --- a/tools/stonith_admin.c +++ b/tools/stonith_admin.c @@ -1,786 +1,763 @@ /* * Copyright 2009-2019 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* *INDENT-OFF* */ static struct crm_option long_options[] = { { "help", no_argument, NULL, '?', "\tDisplay this text and exit." }, { "version", no_argument, NULL, '$', "\tDisplay version information and exit." }, { "verbose", no_argument, NULL, 'V', "\tIncrease debug output (may be specified multiple times)." }, { "quiet", no_argument, NULL, 'q', "\tBe less descriptive in output." }, { "cleanup", no_argument, NULL, 'c', "\tCleanup wherever appropriate. Requires: --history." }, { "broadcast", no_argument, NULL, 'b', "Broadcast wherever appropriate." }, PCMK__OUTPUT_OPTIONS("text, xml"), { "-spacer-", no_argument, NULL, '-', "\nDevice definition commands:" }, { "register", required_argument, NULL, 'R', "Register the named stonith device. Requires: --agent.\n" "\t\t\tOptional: --option, --env-option." }, { "deregister", required_argument, NULL, 'D', "De-register the named stonith device." }, { "register-level", required_argument, NULL, 'r', "Register a stonith level for the named target,\n" "\t\t\tspecified as one of NAME, @PATTERN, or ATTR=VALUE.\n" "\t\t\tRequires: --index and one or more --device entries." }, { "deregister-level", required_argument, NULL, 'd', "Unregister a stonith level for the named target,\n" "\t\t\tspecified as for --register-level. Requires: --index." }, { "-spacer-", no_argument, NULL, '-', "\nQueries:" }, { "list", required_argument, NULL, 'l', "List devices that can terminate the specified host.\n" "\t\t\tOptional: --timeout." }, { "list-registered", no_argument, NULL, 'L', "List all registered devices. Optional: --timeout." }, { "list-installed", no_argument, NULL, 'I', "List all installed devices. Optional: --timeout." }, { "list-targets", required_argument, NULL, 's', "List the targets that can be fenced by the\n" "\t\t\tnamed device. Optional: --timeout." }, { "metadata", no_argument, NULL, 'M', "\tShow agent metadata. Requires: --agent.\n" "\t\t\tOptional: --timeout." }, { "query", required_argument, NULL, 'Q', "Check the named device's status. Optional: --timeout." }, { "history", required_argument, NULL, 'H', "Show last successful fencing operation for named node\n" "\t\t\t(or '*' for all nodes). Optional: --timeout, --cleanup,\n" "\t\t\t--quiet (show only the operation's epoch timestamp),\n" "\t\t\t--verbose (show all recorded and pending operations),\n" "\t\t\t--broadcast (update history from all nodes available)." }, { "last", required_argument, NULL, 'h', "Indicate when the named node was last fenced.\n" "\t\t\tOptional: --as-node-id." }, { "validate", no_argument, NULL, 'K', "\tValidate a fence device configuration.\n" "\t\t\tRequires: --agent. Optional: --option, --env-option,\n" "\t\t\t--quiet (print no output, only return status).\n" }, { "-spacer-", no_argument, NULL, '-', "\nFencing Commands:" }, { "fence", required_argument, NULL, 'F', "Fence named host. Optional: --timeout, --tolerance." }, { "unfence", required_argument, NULL, 'U', "Unfence named host. Optional: --timeout, --tolerance." }, { "reboot", required_argument, NULL, 'B', "Reboot named host. Optional: --timeout, --tolerance." }, { "confirm", required_argument, NULL, 'C', "Tell cluster that named host is now safely down." }, { "-spacer-", no_argument, NULL, '-', "\nAdditional Options:" }, { "agent", required_argument, NULL, 'a', "The agent to use (for example, fence_xvm;\n" "\t\t\twith --register, --metadata, --validate)." }, { "option", required_argument, NULL, 'o', "Specify a device configuration parameter as NAME=VALUE\n" "\t\t\t(may be specified multiple times; with --register,\n" "\t\t\t--validate)." }, { "env-option", required_argument, NULL, 'e', "Specify a device configuration parameter with the\n" "\t\t\tspecified name, using the value of the\n" "\t\t\tenvironment variable of the same name prefixed with\n" "\t\t\tOCF_RESKEY_ (may be specified multiple times;\n" "\t\t\twith --register, --validate)." }, { "tag", required_argument, NULL, 'T', "Identify fencing operations in logs with the specified\n" "\t\t\ttag; useful when multiple entities might invoke\n" "\t\t\tstonith_admin (used with most commands)." }, { "device", required_argument, NULL, 'v', "Device ID (with --register-level, device to associate with\n" "\t\t\ta given host and level; may be specified multiple times)" #if SUPPORT_CIBSECRETS "\n\t\t\t(with --validate, name to use to load CIB secrets)" #endif "." }, { "index", required_argument, NULL, 'i', "The stonith level (1-9) (with --register-level,\n" "\t\t\t--deregister-level)." }, { "timeout", required_argument, NULL, 't', "Operation timeout in seconds (default 120;\n" "\t\t\tused with most commands)." }, { "as-node-id", no_argument, NULL, 'n', "(Advanced) The supplied node is the corosync node ID\n" "\t\t\t(with --last)." }, { "tolerance", required_argument, NULL, 0, "(Advanced) Do nothing if an equivalent --fence request\n" "\t\t\tsucceeded less than this many seconds earlier\n" "\t\t\t(with --fence, --unfence, --reboot)." }, { 0, 0, 0, 0 } }; /* *INDENT-ON* */ static int st_opts = st_opt_sync_call | st_opt_allow_suicide; static GMainLoop *mainloop = NULL; struct { stonith_t *st; const char *target; const char *action; char *name; int timeout; int tolerance; int rc; } async_fence_data; -static int -try_mainloop_connect(void) -{ - stonith_t *st = async_fence_data.st; - int tries = 10; - int i = 0; - int rc = 0; - - for (i = 0; i < tries; i++) { - crm_debug("Connecting as %s", async_fence_data.name); - rc = st->cmds->connect(st, async_fence_data.name, NULL); - - if (!rc) { - crm_debug("stonith client connection established"); - return 0; - } else { - crm_debug("stonith client connection failed"); - } - sleep(1); - } - - crm_err("Could not connect to the fencer"); - return -1; -} - static void notify_callback(stonith_t * st, stonith_event_t * e) { if (e->result != pcmk_ok) { return; } if (safe_str_eq(async_fence_data.target, e->target) && safe_str_eq(async_fence_data.action, e->action)) { async_fence_data.rc = e->result; g_main_loop_quit(mainloop); } } static void fence_callback(stonith_t * stonith, stonith_callback_data_t * data) { async_fence_data.rc = data->rc; g_main_loop_quit(mainloop); } static gboolean async_fence_helper(gpointer user_data) { stonith_t *st = async_fence_data.st; int call_id = 0; + int rc = stonith_api_connect_retry(st, async_fence_data.name, 10); - if (try_mainloop_connect()) { + if (rc != pcmk_ok) { + fprintf(stderr, "Could not connect to fencer: %s\n", pcmk_strerror(rc)); g_main_loop_quit(mainloop); return TRUE; } st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, notify_callback); call_id = st->cmds->fence(st, st_opt_allow_suicide, async_fence_data.target, async_fence_data.action, async_fence_data.timeout, async_fence_data.tolerance); if (call_id < 0) { g_main_loop_quit(mainloop); return TRUE; } st->cmds->register_callback(st, call_id, async_fence_data.timeout, st_opt_timeout_updates, NULL, "callback", fence_callback); return TRUE; } static int mainloop_fencing(stonith_t * st, const char *target, const char *action, int timeout, int tolerance) { crm_trigger_t *trig; async_fence_data.st = st; async_fence_data.target = target; async_fence_data.action = action; async_fence_data.timeout = timeout; async_fence_data.tolerance = tolerance; async_fence_data.rc = -1; trig = mainloop_add_trigger(G_PRIORITY_HIGH, async_fence_helper, NULL); mainloop_set_trigger(trig); mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); return async_fence_data.rc; } static int handle_level(stonith_t *st, char *target, int fence_level, stonith_key_value_t *devices, bool added) { char *node = NULL; char *pattern = NULL; char *name = NULL; char *value = strchr(target, '='); /* Determine if targeting by attribute, node name pattern or node name */ if (value != NULL) { name = target; *value++ = '\0'; } else if (*target == '@') { pattern = target + 1; } else { node = target; } /* Register or unregister level as appropriate */ if (added) { return st->cmds->register_level_full(st, st_opts, node, pattern, name, value, fence_level, devices); } return st->cmds->remove_level_full(st, st_opts, node, pattern, name, value, fence_level); } static int handle_history(stonith_t *st, const char *target, int timeout, int quiet, int verbose, int cleanup, int broadcast, pcmk__output_t *out) { stonith_history_t *history = NULL, *hp, *latest = NULL; int rc = 0; if (!quiet) { if (cleanup) { out->info(out, "cleaning up fencing-history%s%s", target ? " for node " : "", target ? target : ""); } if (broadcast) { out->info(out, "gather fencing-history from all nodes"); } } rc = st->cmds->history(st, st_opts | (cleanup?st_opt_cleanup:0) | (broadcast?st_opt_broadcast:0), (safe_str_eq(target, "*")? NULL : target), &history, timeout); out->begin_list(out, "Fencing history", "event", "events"); for (hp = history; hp; hp = hp->next) { if (hp->state == st_done) { latest = hp; } if (quiet || !verbose) { continue; } out->message(out, "stonith-event", hp); } if (latest) { if (quiet && out->supports_quiet) { out->info(out, "%lld", (long long) latest->completed); } else if (!verbose) { // already printed if verbose out->message(out, "stonith-event", latest); } } out->end_list(out); stonith_history_free(history); return rc; } static int validate(stonith_t *st, const char *agent, const char *id, stonith_key_value_t *params, int timeout, int quiet, pcmk__output_t *out) { int rc = 1; char *output = NULL; char *error_output = NULL; rc = st->cmds->validate(st, st_opt_sync_call, id, NULL, agent, params, timeout, &output, &error_output); if (quiet) { return rc; } out->message(out, "validate", agent, id, output, error_output, rc); return rc; } int main(int argc, char **argv) { int flag; int rc = 0; int quiet = 0; int cleanup = 0; int broadcast = 0; int verbose = 0; int argerr = 0; int timeout = 120; int option_index = 0; int fence_level = 0; int no_connect = 0; int tolerance = 0; int as_nodeid = FALSE; bool required_agent = false; char *name = NULL; char *value = NULL; char *target = NULL; char *lists = NULL; const char *agent = NULL; const char *device = NULL; const char *longname = NULL; char action = 0; crm_exit_t exit_code = CRM_EX_OK; stonith_t *st = NULL; stonith_key_value_t *params = NULL; stonith_key_value_t *devices = NULL; stonith_key_value_t *dIter = NULL; char *output_ty = NULL; char *output_dest = NULL; pcmk__output_t *out = NULL; crm_log_cli_init("stonith_admin"); crm_set_options(NULL, " []", long_options, "access the Pacemaker fencing API"); async_fence_data.name = strdup(crm_system_name); while (1) { flag = crm_get_option_long(argc, argv, &option_index, &longname); if (flag == -1) break; switch (flag) { case 'V': verbose = 1; crm_bump_log_level(argc, argv); break; case '$': case '?': crm_help(flag, CRM_EX_OK); break; case 'K': required_agent = true; /* fall through */ case 'I': no_connect = 1; /* fall through */ case 'L': action = flag; break; case 'q': quiet = 1; break; case 'c': cleanup = 1; break; case 'b': broadcast = 1; break; case 'R': required_agent = true; /* fall through */ case 'Q': case 'D': case 's': action = flag; device = optarg; break; case 'T': free(async_fence_data.name); async_fence_data.name = crm_strdup_printf("%s.%s", crm_system_name, optarg); break; case 'a': agent = optarg; break; case 'l': target = optarg; action = 'L'; break; case 'M': no_connect = 1; action = flag; required_agent = true; break; case 't': timeout = crm_atoi(optarg, NULL); break; case 'B': case 'F': case 'U': /* using mainloop here */ no_connect = 1; /* fall through */ case 'C': /* Always log the input arguments */ crm_log_args(argc, argv); target = optarg; action = flag; break; case 'n': as_nodeid = TRUE; break; case 'h': case 'H': case 'r': case 'd': target = optarg; action = flag; break; case 'i': fence_level = crm_atoi(optarg, NULL); break; case 'v': devices = stonith_key_value_add(devices, NULL, optarg); break; case 'o': crm_info("Scanning: -o %s", optarg); rc = pcmk_scan_nvpair(optarg, &name, &value); if (rc != 2) { crm_err("Invalid option: -o %s: %s", optarg, pcmk_strerror(rc)); ++argerr; } else { crm_info("Got: '%s'='%s'", name, value); params = stonith_key_value_add(params, name, value); } free(value); value = NULL; free(name); name = NULL; break; case 'e': { char *key = crm_concat("OCF_RESKEY", optarg, '_'); const char *env = getenv(key); if (env == NULL) { crm_err("Invalid option: -e %s", optarg); ++argerr; } else { crm_info("Got: '%s'='%s'", optarg, env); params = stonith_key_value_add(params, optarg, env); } free(key); } break; case 0: if (safe_str_eq("tolerance", longname)) { tolerance = crm_get_msec(optarg) / 1000; /* Send in seconds */ } else if (pcmk__parse_output_args(longname, optarg, &output_ty, &output_dest) == false) { fprintf(stderr, "Unknown long option used: %s\n", longname); ++argerr; } break; default: ++argerr; break; } } if (optind > argc || action == 0) { ++argerr; } if (required_agent && agent == NULL) { fprintf(stderr, "Please specify an agent to query using -a,--agent [value]\n"); ++argerr; } if (argerr) { crm_help('?', CRM_EX_USAGE); } CRM_ASSERT(pcmk__register_format("text", pcmk__mk_text_output) == 0); CRM_ASSERT(pcmk__register_format("xml", pcmk__mk_xml_output) == 0); rc = pcmk__output_new(&out, output_ty, output_dest, argv); if (rc != 0) { fprintf(stderr, "Error creating output format %s: %s\n", output_ty, pcmk_strerror(rc)); exit_code = CRM_EX_ERROR; goto done; } stonith__register_messages(out); st = stonith_api_new(); if (!no_connect) { rc = st->cmds->connect(st, async_fence_data.name, NULL); if (rc < 0) { fprintf(stderr, "Could not connect to fencer: %s\n", pcmk_strerror(rc)); exit_code = CRM_EX_DISCONNECT; goto done; } } switch (action) { case 'I': rc = st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout); if (rc < 0) { fprintf(stderr, "Failed to list installed devices: %s\n", pcmk_strerror(rc)); break; } out->begin_list(out, "Installed fence devices", "fence device", "fence devices"); for (dIter = devices; dIter; dIter = dIter->next) { out->list_item(out, "device", dIter->value); } out->end_list(out); rc = 0; stonith_key_value_freeall(devices, 1, 1); break; case 'L': rc = st->cmds->query(st, st_opts, target, &devices, timeout); if (rc < 0) { fprintf(stderr, "Failed to list registered devices: %s\n", pcmk_strerror(rc)); break; } out->begin_list(out, "Registered fence devices", "fence device", "fence devices"); for (dIter = devices; dIter; dIter = dIter->next) { out->list_item(out, "device", dIter->value); } out->end_list(out); rc = 0; stonith_key_value_freeall(devices, 1, 1); break; case 'Q': rc = st->cmds->monitor(st, st_opts, device, timeout); if (rc < 0) { rc = st->cmds->list(st, st_opts, device, NULL, timeout); } break; case 's': rc = st->cmds->list(st, st_opts, device, &lists, timeout); if (rc == 0 && lists) { char *head = lists; char *eol = NULL; out->begin_list(out, "Fence targets", "fence target", "fence targets"); do { char *line = NULL; char *elem = NULL; char *hostname = NULL; char *uuid = NULL; char *status = NULL; eol = strstr(head, "\\n"); line = strndup(head, eol-head); while ((elem = strsep(&line, " ")) != NULL) { if (strcmp(elem, "") == 0) { continue; } if (hostname == NULL) { hostname = elem; } else if (uuid == NULL) { uuid = elem; } else if (status == NULL) { char *end = NULL; status = elem; end = strchr(status, '\n'); if (end != NULL) { *end = '\0'; } } } if (hostname != NULL && uuid != NULL && status != NULL) { out->message(out, "fence-target", hostname, uuid, status); } free(line); head = eol+2; } while (eol != NULL); out->end_list(out); } else if (rc != 0) { fprintf(stderr, "List command returned error. rc : %d\n", rc); } break; case 'R': rc = st->cmds->register_device(st, st_opts, device, NULL, agent, params); break; case 'D': rc = st->cmds->remove_device(st, st_opts, device); break; case 'd': case 'r': rc = handle_level(st, target, fence_level, devices, action == 'r'); break; case 'M': { char *buffer = NULL; rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, timeout); if (rc == pcmk_ok) { out->output_xml(out, "metadata", buffer); } free(buffer); } break; case 'C': rc = st->cmds->confirm(st, st_opts, target); break; case 'B': rc = mainloop_fencing(st, target, "reboot", timeout, tolerance); break; case 'F': rc = mainloop_fencing(st, target, "off", timeout, tolerance); break; case 'U': rc = mainloop_fencing(st, target, "on", timeout, tolerance); break; case 'h': { time_t when = 0; if(as_nodeid) { uint32_t nodeid = atol(target); when = stonith_api_time(nodeid, NULL, FALSE); } else { when = stonith_api_time(0, target, FALSE); } out->message(out, "last-fenced", target, when); } break; case 'H': rc = handle_history(st, target, timeout, quiet, verbose, cleanup, broadcast, out); break; case 'K': device = (devices? devices->key : NULL); rc = validate(st, agent, device, params, timeout, quiet, out); break; } crm_info("Command returned: %s (%d)", pcmk_strerror(rc), rc); exit_code = crm_errno2exit(rc); pcmk__output_free(out, exit_code); done: free(async_fence_data.name); stonith_key_value_freeall(params, 1, 1); if (st != NULL) { st->cmds->disconnect(st); stonith_api_delete(st); } return exit_code; }