diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index a37e9dceaf..ed1d760141 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,3540 +1,3540 @@ /* * Copyright 2009-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include GHashTable *device_list = NULL; GHashTable *topology = NULL; GList *cmd_list = NULL; static GHashTable *fenced_handlers = 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) */ GList *capable; }; static gboolean stonith_device_dispatch(gpointer user_data); static void st_child_done(int pid, const pcmk__action_result_t *result, void *user_data); static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer, pcmk__client_t *client); static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence); static int get_agent_metadata(const char *agent, xmlNode **metadata); static void read_action_metadata(stonith_device_t *device); static enum fenced_target_by unpack_level_kind(xmlNode *level); typedef struct async_command_s { int id; int pid; int fd_stdout; int options; int default_timeout; /* seconds */ int timeout; /* seconds */ int start_delay; /* seconds */ 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; GList *device_list; GList *device_next; void *internal_user_data; void (*done_cb) (int pid, const pcmk__action_result_t *result, void *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 *construct_async_reply(async_command_t *cmd, const pcmk__action_result_t *result); static gboolean is_action_required(const char *action, stonith_device_t *device) { return device && device->automatic_unfencing && pcmk__str_eq(action, "on", pcmk__str_casei); } static int get_action_delay_max(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_max = 0; if (!pcmk__is_fencing_action(action)) { return 0; } value = g_hash_table_lookup(device->params, PCMK_STONITH_DELAY_MAX); if (value) { delay_max = crm_parse_interval_spec(value) / 1000; } return delay_max; } static int get_action_delay_base(stonith_device_t *device, const char *action, const char *victim) { char *hash_value = NULL; int delay_base = 0; if (!pcmk__is_fencing_action(action)) { return 0; } hash_value = g_hash_table_lookup(device->params, PCMK_STONITH_DELAY_BASE); if (hash_value) { char *value = strdup(hash_value); char *valptr = value; CRM_ASSERT(value != NULL); if (victim) { for (char *val = strtok(value, "; \t"); val != NULL; val = strtok(NULL, "; \t")) { char *mapval = strchr(val, ':'); if (mapval == NULL || mapval[1] == 0) { crm_err("pcmk_delay_base: empty value in mapping", val); continue; } if (mapval != val && strncasecmp(victim, val, (size_t)(mapval - val)) == 0) { value = mapval + 1; crm_debug("pcmk_delay_base mapped to %s for %s", value, victim); break; } } } if (strchr(value, ':') == 0) { delay_base = crm_parse_interval_spec(value) / 1000; } free(valptr); } return delay_base; } /*! * \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 (pcmk__str_eq(action, "reboot", pcmk__str_casei) && !pcmk_is_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->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; // Value -1 means disable any static/random fencing delays crm_element_value_int(msg, F_STONITH_DELAY, &(cmd->start_delay)); 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->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, PCMK_STONITH_ACTION_LIMIT); if ((value == NULL) || (pcmk__scan_min_int(value, &action_limit, INT_MIN) != pcmk_rc_ok) || (action_limit == 0)) { action_limit = 1; } return action_limit; } static int get_active_cmds(stonith_device_t * device) { int counter = 0; GList *gIter = NULL; GList *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(int pid, void *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' [%d]%s%s using %s now running with %ds timeout", cmd->action, pid, ((cmd->victim == NULL)? "" : " targeting "), ((cmd->victim == NULL)? "" : cmd->victim), device->id, cmd->timeout); cmd->active_on = device; cmd->activating_on = NULL; } static int get_agent_metadata_cb(gpointer data) { stonith_device_t *device = data; guint period_ms; switch (get_agent_metadata(device->agent, &device->agent_metadata)) { case pcmk_rc_ok: if (device->agent_metadata) { read_action_metadata(device); stonith__device_parameter_flags(&(device->flags), device->id, device->agent_metadata); } return G_SOURCE_REMOVE; case EAGAIN: period_ms = pcmk__mainloop_timer_get_period(device->timer); if (period_ms < 160 * 1000) { mainloop_timer_set_period(device->timer, 2 * period_ms); } return G_SOURCE_CONTINUE; default: return G_SOURCE_REMOVE; } } /*! * \internal * \brief Call a command's action callback for an internal (not library) result * * \param[in] cmd Command to report result for * \param[in] execution_status Execution status to use for result * \param[in] exit_status Exit status to use for result * \param[in] exit_reason Exit reason to use for result */ static void report_internal_result(async_command_t *cmd, int exit_status, int execution_status, const char *exit_reason) { pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; pcmk__set_result(&result, exit_status, execution_status, exit_reason); cmd->done_cb(0, &result, cmd); pcmk__reset_result(&result); } static gboolean stonith_device_execute(stonith_device_t * device) { int exec_rc = 0; const char *action_str = NULL; const char *host_arg = NULL; async_command_t *cmd = NULL; stonith_action_t *action = NULL; int active_cmds = 0; int action_limit = 0; GList *gIter = NULL; GList *gIterNext = NULL; 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, pcmk__plural_s(active_cmds)); return TRUE; } for (gIter = device->pending_ops; gIter != NULL; gIter = gIterNext) { async_command_t *pending_op = gIter->data; gIterNext = gIter->next; if (pending_op && pending_op->delay_id) { crm_trace("Operation '%s'%s%s using %s was asked to run too early, " "waiting for start delay of %ds", pending_op->action, ((pending_op->victim == NULL)? "" : " targeting "), ((pending_op->victim == NULL)? "" : pending_op->victim), device->id, pending_op->start_delay); continue; } device->pending_ops = g_list_remove_link(device->pending_ops, gIter); g_list_free_1(gIter); cmd = pending_op; break; } if (cmd == NULL) { crm_trace("No actions using %s are needed", device->id); return TRUE; } if (pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) { if (pcmk__is_fencing_action(cmd->action)) { if (node_does_watchdog_fencing(stonith_our_uname)) { pcmk__panic(__func__); goto done; } } else { crm_info("Faking success for %s watchdog operation", cmd->action); report_internal_result(cmd, CRM_EX_OK, PCMK_EXEC_DONE, NULL); goto done; } } #if SUPPORT_CIBSECRETS exec_rc = pcmk__substitute_secrets(device->id, device->params); if (exec_rc != pcmk_rc_ok) { if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { crm_info("Proceeding with stop operation for %s " "despite being unable to load CIB secrets (%s)", device->id, pcmk_rc_str(exec_rc)); } else { crm_err("Considering %s unconfigured " "because unable to load CIB secrets: %s", device->id, pcmk_rc_str(exec_rc)); report_internal_result(cmd, CRM_EX_ERROR, PCMK_EXEC_NO_SECRETS, "Failed to get CIB secrets"); goto done; } } #endif action_str = cmd->action; if (pcmk__str_eq(cmd->action, "reboot", pcmk__str_casei) && !pcmk_is_set(device->flags, st_device_supports_reboot)) { crm_notice("Remapping 'reboot' action%s%s using %s to 'off' " "because agent '%s' does not support reboot", ((cmd->victim == NULL)? "" : " targeting "), ((cmd->victim == NULL)? "" : cmd->victim), device->id, device->agent); action_str = "off"; } if (pcmk_is_set(device->flags, st_device_supports_parameter_port)) { host_arg = "port"; } else if (pcmk_is_set(device->flags, st_device_supports_parameter_plug)) { host_arg = "plug"; } action = stonith_action_create(device->agent, action_str, cmd->victim, cmd->victim_nodeid, cmd->timeout, device->params, device->aliases, host_arg); /* 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) { cmd->activating_on = NULL; cmd->done_cb(0, stonith__action_result(action), cmd); stonith__destroy_action(action); } done: /* Device might get triggered to work by multiple fencing commands * simultaneously. Trigger the device again to make sure any * remaining concurrent commands get executed. */ if (device->pending_ops) { mainloop_set_trigger(device->work); } 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; int requested_delay = cmd->start_delay; 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' action%s%s using %s for remote peer %s " "with op id %.8s and timeout %ds", cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "", device->id, cmd->origin, cmd->remote_op_id, cmd->timeout); } else { crm_debug("Scheduling '%s' action%s%s using %s for %s with timeout %ds", cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "", device->id, cmd->client, cmd->timeout); } device->pending_ops = g_list_append(device->pending_ops, cmd); mainloop_set_trigger(device->work); // Value -1 means disable any static/random fencing delays if (requested_delay < 0) { return; } delay_max = get_action_delay_max(device, cmd->action); delay_base = get_action_delay_base(device, cmd->action, cmd->victim); if (delay_max == 0) { delay_max = delay_base; } if (delay_max < delay_base) { crm_warn(PCMK_STONITH_DELAY_BASE " (%ds) is larger than " PCMK_STONITH_DELAY_MAX " (%ds) for %s using %s " "(limiting to maximum 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; } if (cmd->start_delay > 0) { crm_notice("Delaying '%s' action%s%s using %s for %ds " CRM_XS " timeout=%ds requested_delay=%ds base=%ds max=%ds", cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "", device->id, cmd->start_delay, cmd->timeout, requested_delay, delay_base, delay_max); cmd->delay_id = g_timeout_add_seconds(cmd->start_delay, start_delay_helper, cmd); } } static void free_device(gpointer data) { GList *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); report_internal_result(cmd, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "Device was removed before action could be executed"); } g_list_free(device->pending_ops); g_list_free_full(device->targets, free); if (device->timer) { mainloop_timer_stop(device->timer); mainloop_timer_del(device->timer); } 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(void) { if (device_list != NULL) { g_hash_table_destroy(device_list); device_list = NULL; } } void init_device_list(void) { if (device_list == NULL) { device_list = pcmk__strkey_table(NULL, free_device); } } static GHashTable * build_port_aliases(const char *hostmap, GList ** targets) { char *name = NULL; int last = 0, lpc = 0, max = 0, added = 0; GHashTable *aliases = pcmk__strikey_table(free, free); if (hostmap == NULL) { return aliases; } max = strlen(hostmap); for (; lpc <= max; lpc++) { switch (hostmap[lpc]) { /* Skip escaped chars */ case '\\': lpc++; break; /* 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; int k = 0; value = calloc(1, 1 + lpc - last); memcpy(value, hostmap + last, lpc - last); for (int i = 0; value[i] != '\0'; i++) { if (value[i] != '\\') { value[k++] = value[i]; } } value[k] = '\0'; 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; } GHashTable *metadata_cache = NULL; void free_metadata_cache(void) { if (metadata_cache != NULL) { g_hash_table_destroy(metadata_cache); metadata_cache = NULL; } } static void init_metadata_cache(void) { if (metadata_cache == NULL) { metadata_cache = pcmk__strkey_table(free, free); } } int get_agent_metadata(const char *agent, xmlNode ** metadata) { char *buffer = NULL; if (metadata == NULL) { return EINVAL; } *metadata = NULL; if (pcmk__str_eq(agent, STONITH_WATCHDOG_AGENT_INTERNAL, pcmk__str_none)) { return pcmk_rc_ok; } init_metadata_cache(); buffer = g_hash_table_lookup(metadata_cache, agent); if (buffer == NULL) { stonith_t *st = stonith_api_new(); int rc; if (st == NULL) { crm_warn("Could not get agent meta-data: " "API memory allocation failed"); return EAGAIN; } 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 EAGAIN; } g_hash_table_replace(metadata_cache, strdup(agent), buffer); } *metadata = string2xml(buffer); return pcmk_rc_ok; } 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 *action = NULL; xmlNode *match = getXpathResult(xpath, lpc); CRM_LOG_ASSERT(match != NULL); if(match == NULL) { continue; }; action = crm_element_value(match, "name"); if(pcmk__str_eq(action, "list", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_list); } else if(pcmk__str_eq(action, "status", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_status); } else if(pcmk__str_eq(action, "reboot", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_reboot); } else if (pcmk__str_eq(action, "on", pcmk__str_casei)) { /* "automatic" means the cluster will unfence node when it joins */ /* "required" is a deprecated synonym for "automatic" */ if (pcmk__xe_attr_is_true(match, "automatic") || pcmk__xe_attr_is_true(match, "required")) { device->automatic_unfencing = TRUE; } } if (action && pcmk__xe_attr_is_true(match, "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 const char * target_list_type(stonith_device_t * dev) { const char *check_type = NULL; check_type = g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_CHECK); if (check_type == NULL) { if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_LIST)) { check_type = "static-list"; } else if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_MAP)) { check_type = "static-list"; } else if (pcmk_is_set(dev->flags, st_device_supports_list)) { check_type = "dynamic-list"; } else if (pcmk_is_set(dev->flags, st_device_supports_status)) { check_type = "status"; } else { check_type = PCMK__VALUE_NONE; } } return check_type; } static stonith_device_t * build_device_from_xml(xmlNode *dev) { const char *value; 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, PCMK_STONITH_HOST_LIST); if (value) { device->targets = stonith__parse_targets(value); } value = g_hash_table_lookup(device->params, PCMK_STONITH_HOST_MAP); device->aliases = build_port_aliases(value, &(device->targets)); value = target_list_type(device); if (!pcmk__str_eq(value, "static-list", pcmk__str_casei) && device->targets) { /* Other than "static-list", dev-> targets is unnecessary. */ g_list_free_full(device->targets, free); device->targets = NULL; } switch (get_agent_metadata(device->agent, &device->agent_metadata)) { case pcmk_rc_ok: if (device->agent_metadata) { read_action_metadata(device); stonith__device_parameter_flags(&(device->flags), device->id, device->agent_metadata); } break; case EAGAIN: if (device->timer == NULL) { device->timer = mainloop_timer_add("get_agent_metadata", 10 * 1000, TRUE, get_agent_metadata_cb, device); } if (!mainloop_timer_running(device->timer)) { mainloop_timer_start(device->timer); } break; default: break; } 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 (pcmk__str_eq(value, "unfencing", pcmk__str_casei)) { + if (pcmk__str_eq(value, PCMK__VALUE_UNFENCING, pcmk__str_casei)) { device->automatic_unfencing = TRUE; } if (is_action_required("on", device)) { crm_info("Fencing device '%s' requires unfencing", device->id); } if (device->on_target_actions) { crm_info("Fencing device '%s' requires actions (%s) to be executed " "on target", 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 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) (int pid, const pcmk__action_result_t *result, void *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); pcmk__str_update(&cmd->victim, victim); 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); } // Fence agent status commands use custom exit status codes enum fence_status_code { fence_status_invalid = -1, fence_status_active = 0, fence_status_unknown = 1, fence_status_inactive = 2, }; static void status_search_cb(int pid, const pcmk__action_result_t *result, void *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 (result->execution_status != PCMK_EXEC_DONE) { crm_warn("Assuming %s cannot fence %s " "because status could not be executed: %s%s%s%s", dev->id, search->host, pcmk_exec_status_str(result->execution_status), ((result->exit_reason == NULL)? "" : " ("), ((result->exit_reason == NULL)? "" : result->exit_reason), ((result->exit_reason == NULL)? "" : ")")); search_devices_record_result(search, dev->id, FALSE); return; } switch (result->exit_status) { case fence_status_unknown: crm_trace("%s reported it cannot fence %s", dev->id, search->host); break; case fence_status_active: case fence_status_inactive: crm_trace("%s reported it can fence %s", dev->id, search->host); can = TRUE; break; default: crm_warn("Assuming %s cannot fence %s " "(status returned unknown code %d)", dev->id, search->host, result->exit_status); break; } search_devices_record_result(search, dev->id, can); } static void dynamic_list_search_cb(int pid, const pcmk__action_result_t *result, void *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 (pcmk__result_ok(result)) { crm_info("Refreshing target list for %s", dev->id); g_list_free_full(dev->targets, free); dev->targets = stonith__parse_targets(result->action_stdout); dev->targets_age = time(NULL); } else if (dev->targets != NULL) { if (result->execution_status == PCMK_EXEC_DONE) { crm_info("Reusing most recent target list for %s " "because list returned error code %d", dev->id, result->exit_status); } else { crm_info("Reusing most recent target list for %s " "because list could not be executed: %s%s%s%s", dev->id, pcmk_exec_status_str(result->execution_status), ((result->exit_reason == NULL)? "" : " ("), ((result->exit_reason == NULL)? "" : result->exit_reason), ((result->exit_reason == NULL)? "" : ")")); } } else { // We have never successfully executed list if (result->execution_status == PCMK_EXEC_DONE) { crm_warn("Assuming %s cannot fence %s " "because list returned error code %d", dev->id, search->host, result->exit_status); } else { crm_warn("Assuming %s cannot fence %s " "because list could not be executed: %s%s%s%s", dev->id, search->host, pcmk_exec_status_str(result->execution_status), ((result->exit_reason == NULL)? "" : " ("), ((result->exit_reason == NULL)? "" : result->exit_reason), ((result->exit_reason == NULL)? "" : ")")); } /* Fall back to pcmk_host_check="status" if the user didn't explicitly * specify "dynamic-list". */ if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_CHECK) == NULL) { crm_notice("Switching to pcmk_host_check='status' for %s", dev->id); g_hash_table_replace(dev->params, strdup(PCMK_STONITH_HOST_CHECK), strdup("status")); } } if (dev->targets) { const char *alias = g_hash_table_lookup(dev->aliases, search->host); if (!alias) { alias = search->host; } if (pcmk__str_in_list(alias, dev->targets, pcmk__str_casei)) { 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 || !pcmk__str_eq(other_value, value, pcmk__str_casei)) { 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 (!pcmk__str_eq(dup->agent, device->agent, pcmk__str_casei)) { 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 *dev, gboolean from_cib) { stonith_device_t *dup = NULL; stonith_device_t *device = build_device_from_xml(dev); guint ndevices = 0; int rv = pcmk_ok; CRM_CHECK(device != NULL, return -ENOMEM); /* do we have a watchdog-device? */ if (pcmk__str_eq(device->id, STONITH_WATCHDOG_ID, pcmk__str_none) || pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) do { if (stonith_watchdog_timeout_ms <= 0) { crm_err("Ignoring watchdog fence device without " "stonith-watchdog-timeout set."); rv = -ENODEV; /* fall through to cleanup & return */ } else if (!pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) { crm_err("Ignoring watchdog fence device with unknown " "agent '%s' unequal '" STONITH_WATCHDOG_AGENT "'.", device->agent?device->agent:""); rv = -ENODEV; /* fall through to cleanup & return */ } else if (!pcmk__str_eq(device->id, STONITH_WATCHDOG_ID, pcmk__str_none)) { crm_err("Ignoring watchdog fence device " "named %s !='"STONITH_WATCHDOG_ID"'.", device->id?device->id:""); rv = -ENODEV; /* fall through to cleanup & return */ } else { if (pcmk__str_eq(device->agent, STONITH_WATCHDOG_AGENT, pcmk__str_none)) { /* this either has an empty list or the targets configured for watchdog-fencing */ g_list_free_full(stonith_watchdog_targets, free); stonith_watchdog_targets = device->targets; device->targets = NULL; } if (node_does_watchdog_fencing(stonith_our_uname)) { g_list_free_full(device->targets, free); device->targets = stonith__parse_targets(stonith_our_uname); g_hash_table_replace(device->params, strdup(PCMK_STONITH_HOST_LIST), strdup(stonith_our_uname)); /* proceed as with any other stonith-device */ break; } crm_debug("Skip registration of watchdog fence device on node not in host-list."); /* cleanup and fall through to more cleanup and return */ device->targets = NULL; stonith_device_remove(device->id, from_cib); } free_device(device); return rv; } while (0); dup = device_has_duplicate(device); if (dup) { ndevices = g_hash_table_size(device_list); crm_debug("Device '%s' already in device list (%d active device%s)", device->id, ndevices, pcmk__plural_s(ndevices)); free_device(device); device = dup; dup = g_hash_table_lookup(device_list, device->id); dup->dirty = FALSE; } 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 existing entry for %s from 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); ndevices = g_hash_table_size(device_list); crm_notice("Added '%s' to device list (%d active device%s)", device->id, ndevices, pcmk__plural_s(ndevices)); } if (from_cib) { device->cib_registered = TRUE; } else { device->api_registered = TRUE; } return pcmk_ok; } void stonith_device_remove(const char *id, bool from_cib) { stonith_device_t *device = g_hash_table_lookup(device_list, id); guint ndevices = 0; if (!device) { ndevices = g_hash_table_size(device_list); crm_info("Device '%s' not found (%d active device%s)", id, ndevices, pcmk__plural_s(ndevices)); return; } 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); ndevices = g_hash_table_size(device_list); crm_info("Removed '%s' from device list (%d active device%s)", id, ndevices, pcmk__plural_s(ndevices)); } else { crm_trace("Not removing '%s' from device list (%d active) because " "still registered via:%s%s", id, g_hash_table_size(device_list), (device->cib_registered? " cib" : ""), (device->api_registered? " api" : "")); } } /*! * \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(void) { if (topology != NULL) { g_hash_table_destroy(topology); topology = NULL; } } void init_topology_list(void) { if (topology == NULL) { topology = pcmk__strkey_table(NULL, free_topology_entry); } } char * stonith_level_key(xmlNode *level, enum fenced_target_by mode) { if (mode == fenced_target_by_unknown) { mode = unpack_level_kind(level); } switch (mode) { case fenced_target_by_name: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET); case fenced_target_by_pattern: return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN); case fenced_target_by_attribute: return crm_strdup_printf("%s=%s", crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE), crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE)); default: return crm_strdup_printf("unknown-%s", ID(level)); } } /*! * \internal * \brief Parse target identification from topology level XML * * \param[in] level Topology level XML to parse * * \return How to identify target of \p level */ static int unpack_level_kind(xmlNode *level) { if (crm_element_value(level, XML_ATTR_STONITH_TARGET) != NULL) { return fenced_target_by_name; } if (crm_element_value(level, XML_ATTR_STONITH_TARGET_PATTERN) != NULL) { return fenced_target_by_pattern; } if (!stand_alone /* if standalone, there's no attribute manager */ && (crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE) != NULL) && (crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE) == NULL)) { return fenced_target_by_attribute; } return fenced_target_by_unknown; } 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 Unpack essential information from topology request XML * * \param[in] xml Request XML to search * \param[out] mode If not NULL, where to store level kind * \param[out] target If not NULL, where to store representation of target * \param[out] id If not NULL, where to store level number * \param[out] desc If not NULL, where to store log-friendly level description * * \return Topology level XML from within \p xml, or NULL if not found * \note The caller is responsible for freeing \p *target and \p *desc if set. */ static xmlNode * unpack_level_request(xmlNode *xml, enum fenced_target_by *mode, char **target, int *id, char **desc) { enum fenced_target_by local_mode = fenced_target_by_unknown; char *local_target = NULL; int local_id = 0; /* The level element can be the top element or lower. If top level, don't * search by xpath, because it might give multiple hits if the XML is the * CIB. */ if ((xml != NULL) && !pcmk__str_eq(TYPE(xml), XML_TAG_FENCING_LEVEL, pcmk__str_none)) { xml = get_xpath_object("//" XML_TAG_FENCING_LEVEL, xml, LOG_WARNING); } if (xml == NULL) { if (desc != NULL) { *desc = crm_strdup_printf("missing"); } } else { local_mode = unpack_level_kind(xml); local_target = stonith_level_key(xml, local_mode); crm_element_value_int(xml, XML_ATTR_STONITH_INDEX, &local_id); if (desc != NULL) { *desc = crm_strdup_printf("%s[%d]", local_target, local_id); } } if (mode != NULL) { *mode = local_mode; } if (id != NULL) { *id = local_id; } if (target != NULL) { *target = local_target; } else { free(local_target); } return xml; } /*! * \internal * \brief Register a fencing topology 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, set to string representation "TARGET[LEVEL]" * \param[out] result Where to set result of registration */ void fenced_register_level(xmlNode *msg, char **desc, pcmk__action_result_t *result) { int id = 0; xmlNode *level; enum fenced_target_by mode; char *target; stonith_topology_t *tp; stonith_key_value_t *dIter = NULL; stonith_key_value_t *devices = NULL; CRM_CHECK(result != NULL, return); level = unpack_level_request(msg, &mode, &target, &id, desc); if (level == NULL) { fenced_set_protocol_error(result); return; } // Ensure a valid target was specified if (mode == fenced_target_by_unknown) { crm_warn("Ignoring registration for topology level '%s' " "without valid target", pcmk__s(ID(level), "")); free(target); crm_log_xml_trace(level, "Bad level"); pcmk__format_result(result, CRM_EX_INVALID_PARAM, PCMK_EXEC_INVALID, "Invalid target for topology level '%s'", pcmk__s(ID(level), "")); return; } // Ensure level ID is in allowed range if ((id <= 0) || (id >= ST_LEVEL_MAX)) { crm_warn("Ignoring topology registration for %s with invalid level %d", target, id); free(target); crm_log_xml_trace(level, "Bad level"); pcmk__format_result(result, CRM_EX_INVALID_PARAM, PCMK_EXEC_INVALID, "Invalid level number '%s' for topology level '%s'", pcmk__s(crm_element_value(level, XML_ATTR_STONITH_INDEX), ""), pcmk__s(ID(level), "")); return; } /* Find or create topology table entry */ tp = g_hash_table_lookup(topology, target); if (tp == NULL) { tp = calloc(1, sizeof(stonith_topology_t)); if (tp == NULL) { pcmk__set_result(result, CRM_EX_ERROR, PCMK_EXEC_ERROR, strerror(ENOMEM)); free(target); return; } 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, (int) 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); { int nlevels = count_active_levels(tp); crm_info("Target %s has %d active fencing level%s", tp->target, nlevels, pcmk__plural_s(nlevels)); } pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); } /*! * \internal * \brief Unregister a fencing topology level for a target * * Given an XML request specifying the target name and level index (or 0 for all * levels), this will remove any corresponding entry for the target from the * global topology table. * * \param[in] msg XML request for STONITH level registration * \param[out] desc If not NULL, set to string representation "TARGET[LEVEL]" * \param[out] result Where to set result of unregistration */ void fenced_unregister_level(xmlNode *msg, char **desc, pcmk__action_result_t *result) { int id = -1; stonith_topology_t *tp; char *target; xmlNode *level = NULL; CRM_CHECK(result != NULL, return); level = unpack_level_request(msg, NULL, &target, &id, desc); if (level == NULL) { fenced_set_protocol_error(result); return; } // Ensure level ID is in allowed range if ((id < 0) || (id >= ST_LEVEL_MAX)) { crm_warn("Ignoring topology unregistration for %s with invalid level %d", target, id); free(target); crm_log_xml_trace(level, "Bad level"); pcmk__format_result(result, CRM_EX_INVALID_PARAM, PCMK_EXEC_INVALID, "Invalid level number '%s' for topology level '%s'", pcmk__s(crm_element_value(level, XML_ATTR_STONITH_INDEX), ""), pcmk__s(ID(level), "")); return; } tp = g_hash_table_lookup(topology, target); if (tp == NULL) { guint nentries = g_hash_table_size(topology); crm_info("No fencing topology found for %s (%d active %s)", target, nentries, pcmk__plural_alt(nentries, "entry", "entries")); } else if (id == 0 && g_hash_table_remove(topology, target)) { guint nentries = g_hash_table_size(topology); crm_info("Removed all fencing topology entries related to %s " "(%d active %s remaining)", target, nentries, pcmk__plural_alt(nentries, "entry", "entries")); } else if (tp->levels[id] != NULL) { guint nlevels; g_list_free_full(tp->levels[id], free); tp->levels[id] = NULL; nlevels = count_active_levels(tp); crm_info("Removed level %d from fencing topology for %s " "(%d active level%s remaining)", id, target, nlevels, pcmk__plural_s(nlevels)); } free(target); pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); } static char * list_to_string(GList *list, const char *delim, gboolean terminate_with_delim) { int max = g_list_length(list); size_t delim_len = delim?strlen(delim):0; size_t alloc_size = 1 + (max?((max-1+(terminate_with_delim?1:0))*delim_len):0); char *rv; GList *gIter; for (gIter = list; gIter != NULL; gIter = gIter->next) { const char *value = (const char *) gIter->data; alloc_size += strlen(value); } rv = calloc(alloc_size, sizeof(char)); if (rv) { char *pos = rv; const char *lead_delim = ""; for (gIter = list; gIter != NULL; gIter = gIter->next) { const char *value = (const char *) gIter->data; pos = &pos[sprintf(pos, "%s%s", lead_delim, value)]; lead_delim = delim; } if (max && terminate_with_delim) { sprintf(pos, "%s", delim); } } return rv; } /*! * \internal * \brief Execute a fence agent action directly (and asynchronously) * * Handle a STONITH_OP_EXEC API message by scheduling a requested agent action * directly on a specified device. Only list, monitor, and status actions are * expected to use this call, though it should work with any agent command. * * \param[in] msg Request XML specifying action * \param[out] result Where to store result of action * * \note If the action is monitor, the device must be registered via the API * (CIB registration is not sufficient), because monitor should not be * possible unless the device is "started" (API registered). */ static void execute_agent_action(xmlNode *msg, pcmk__action_result_t *result) { xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR); xmlNode *op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR); const char *id = crm_element_value(dev, F_STONITH_DEVICE); const char *action = crm_element_value(op, F_STONITH_ACTION); async_command_t *cmd = NULL; stonith_device_t *device = NULL; if ((id == NULL) || (action == NULL)) { crm_info("Malformed API action request: device %s, action %s", (id? id : "not specified"), (action? action : "not specified")); fenced_set_protocol_error(result); return; } if (pcmk__str_eq(id, STONITH_WATCHDOG_ID, pcmk__str_none)) { // Watchdog agent actions are implemented internally if (stonith_watchdog_timeout_ms <= 0) { pcmk__set_result(result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "Watchdog fence device not configured"); return; } else if (pcmk__str_eq(action, "list", pcmk__str_casei)) { pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); pcmk__set_result_output(result, list_to_string(stonith_watchdog_targets, "\n", TRUE), NULL); return; } else if (pcmk__str_eq(action, "monitor", pcmk__str_casei)) { pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); return; } } device = g_hash_table_lookup(device_list, id); if (device == NULL) { crm_info("Ignoring API '%s' action request because device %s not found", action, id); pcmk__format_result(result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "'%s' not found", id); return; } else if (!device->api_registered && !strcmp(action, "monitor")) { // Monitors may run only on "started" (API-registered) devices crm_info("Ignoring API '%s' action request because device %s not active", action, id); pcmk__format_result(result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "'%s' not active", id); return; } cmd = create_async_command(msg); if (cmd == NULL) { fenced_set_protocol_error(result); return; } schedule_stonith_command(cmd, device); pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); } 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) { guint ndevices = g_list_length(search->capable); crm_debug("Search found %d device%s that can perform '%s' targeting %s", ndevices, pcmk__plural_s(ndevices), (search->action? search->action : "unknown action"), (search->host? search->host : "any node")); 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 = pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei); if (device && action && device->on_target_actions && strstr(device->on_target_actions, action)) { if (!localhost_is_target) { crm_trace("Operation '%s' using %s can only be executed for " "local host, 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 (pcmk__str_eq(search->action, "reboot", pcmk__str_casei)) { /* 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 (pcmk__str_eq(check_type, PCMK__VALUE_NONE, pcmk__str_casei)) { can = TRUE; } else if (pcmk__str_eq(check_type, "static-list", pcmk__str_casei)) { /* 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 (pcmk__str_in_list(host, dev->targets, pcmk__str_casei)) { can = TRUE; } else if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_MAP) && g_hash_table_lookup(dev->aliases, host)) { can = TRUE; } } else if (pcmk__str_eq(check_type, "dynamic-list", pcmk__str_casei)) { time_t now = time(NULL); if (dev->targets == NULL || dev->targets_age + 60 < now) { int device_timeout = get_action_timeout(dev, "list", search->per_device_timeout); if (device_timeout > search->per_device_timeout) { crm_notice("Since the pcmk_list_timeout(%ds) parameter of %s is larger than stonith-timeout(%ds), timeout may occur", device_timeout, dev->id, search->per_device_timeout); } 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(__func__, 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 (pcmk__str_in_list(alias, dev->targets, pcmk__str_casei)) { can = TRUE; } } else if (pcmk__str_eq(check_type, "status", pcmk__str_casei)) { int device_timeout = get_action_timeout(dev, check_type, search->per_device_timeout); if (device_timeout > search->per_device_timeout) { crm_notice("Since the pcmk_status_timeout(%ds) parameter of %s is larger than stonith-timeout(%ds), timeout may occur", device_timeout, dev->id, search->per_device_timeout); } 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(__func__, 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 " PCMK_STONITH_HOST_CHECK ": %s", check_type); check_type = "Invalid " PCMK_STONITH_HOST_CHECK; } if (pcmk__str_eq(host, alias, pcmk__str_casei)) { crm_info("%s is%s eligible to fence (%s) %s: %s", dev->id, (can? "" : " not"), search->action, host, check_type); } else { crm_info("%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; guint ndevices = g_hash_table_size(device_list); if (ndevices == 0) { callback(NULL, user_data); return; } search = calloc(1, sizeof(struct device_search_s)); if (!search) { crm_crit("Cannot search for capable fence devices: %s", strerror(ENOMEM)); callback(NULL, user_data); return; } pcmk__str_update(&search->host, host); pcmk__str_update(&search->action, action); search->per_device_timeout = timeout; search->allow_suicide = suicide; search->callback = callback; search->user_data = user_data; /* We are guaranteed this many replies, even if a device is * unregistered while the search is in progress. */ search->replies_needed = ndevices; crm_debug("Searching %d device%s to see which can execute '%s' targeting %s", ndevices, pcmk__plural_s(ndevices), (search->action? search->action : "unknown action"), (search->host? search->host : "any node")); 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 * \param[in] target Fence target */ static void add_action_specific_attributes(xmlNode *xml, const char *action, stonith_device_t *device, const char *target) { 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 using %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 using %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 using %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, target); 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 using %s", action, delay_max, device->id); } else if ((delay_max == 0) && (delay_base > 0)) { crm_trace("Action '%s' has a static delay of %dms using %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 using %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' using %s is disallowed for local host", action, device->id); pcmk__xe_set_bool_attr(xml, F_STONITH_ACTION_DISALLOWED, 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, target); 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; GList *lpc = NULL; pcmk__client_t *client = NULL; if (query->client_id != NULL) { client = pcmk__find_client_by_id(query->client_id); if ((client == NULL) && (query->remote_peer == NULL)) { crm_trace("Skipping reply to %s: no longer a client", query->client_id); goto done; } } /* Pack the results into XML */ list = create_xml_node(NULL, __func__); 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 (!pcmk_is_set(device->flags, st_device_supports_reboot) && pcmk__str_eq(query->action, "reboot", pcmk__str_casei)) { 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, query->target); if (pcmk__str_eq(query->action, "reboot", pcmk__str_casei)) { /* 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, pcmk_is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "off", device, query->target, pcmk_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 device%s for target '%s'", available_devices, pcmk__plural_s(available_devices), query->target); } else { crm_debug("%d device%s installed", available_devices, pcmk__plural_s(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, client); done: 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); } /*! * \internal * \brief Log the result of an asynchronous command * * \param[in] cmd Command the result is for * \param[in] result Result of command * \param[in] pid Process ID of command, if available * \param[in] next Alternate device that will be tried if command failed * \param[in] op_merged Whether this command was merged with an earlier one */ static void log_async_result(async_command_t *cmd, const pcmk__action_result_t *result, int pid, const char *next, bool op_merged) { int log_level = LOG_ERR; int output_log_level = LOG_NEVER; guint devices_remaining = g_list_length(cmd->device_next); GString *msg = g_string_sized_new(80); // Reasonable starting size // Choose log levels appropriately if we have a result if (pcmk__result_ok(result)) { log_level = (cmd->victim == NULL)? LOG_DEBUG : LOG_NOTICE; if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_casei)) { output_log_level = LOG_DEBUG; } next = NULL; } else { log_level = (cmd->victim == NULL)? LOG_NOTICE : LOG_ERR; if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_casei)) { output_log_level = LOG_WARNING; } } // Build the log message piece by piece g_string_printf(msg, "Operation '%s' ", cmd->action); if (pid != 0) { g_string_append_printf(msg, "[%d] ", pid); } if (cmd->victim != NULL) { g_string_append_printf(msg, "targeting %s ", cmd->victim); } if (cmd->device != NULL) { g_string_append_printf(msg, "using %s ", cmd->device); } // Add exit status or execution status as appropriate if (result->execution_status == PCMK_EXEC_DONE) { g_string_append_printf(msg, "returned %d", result->exit_status); } else { g_string_append_printf(msg, "could not be executed: %s", pcmk_exec_status_str(result->execution_status)); } // Add exit reason and next device if appropriate if (result->exit_reason != NULL) { g_string_append_printf(msg, " (%s)", result->exit_reason); } if (next != NULL) { g_string_append_printf(msg, ", retrying with %s", next); } if (devices_remaining > 0) { g_string_append_printf(msg, " (%u device%s remaining)", (unsigned int) devices_remaining, pcmk__plural_s(devices_remaining)); } g_string_append_printf(msg, " " CRM_XS " %scall %d from %s", (op_merged? "merged " : ""), cmd->id, cmd->client_name); // Log the result do_crm_log(log_level, "%s", msg->str); g_string_free(msg, TRUE); // Log the output (which may have multiple lines), if appropriate if (output_log_level != LOG_NEVER) { char *prefix = crm_strdup_printf("%s[%d]", cmd->device, pid); crm_log_output(output_log_level, prefix, result->action_stdout); free(prefix); } } /*! * \internal * \brief Reply to requester after asynchronous command completion * * \param[in] cmd Command that completed * \param[in] result Result of command * \param[in] pid Process ID of command, if available * \param[in] merged If true, command was merged with another, not executed */ static void send_async_reply(async_command_t *cmd, const pcmk__action_result_t *result, int pid, bool merged) { xmlNode *reply = NULL; pcmk__client_t *client = NULL; CRM_CHECK((cmd != NULL) && (result != NULL), return); log_async_result(cmd, result, pid, NULL, merged); if (cmd->client != NULL) { client = pcmk__find_client_by_id(cmd->client); if ((client == NULL) && (cmd->origin == NULL)) { crm_trace("Skipping reply to %s: no longer a client", cmd->client); return; } } reply = construct_async_reply(cmd, result); if (merged) { pcmk__xe_set_bool_attr(reply, F_STONITH_MERGED, true); } if (!stand_alone && pcmk__is_fencing_action(cmd->action) && pcmk__str_eq(cmd->origin, cmd->victim, pcmk__str_casei)) { /* The target was also the originator, so broadcast the result on its * behalf (since it will be unable to). */ crm_trace("Broadcast '%s' result for %s (target was also originator)", cmd->action, cmd->victim); crm_xml_add(reply, F_SUBTYPE, "broadcast"); crm_xml_add(reply, F_STONITH_OPERATION, T_STONITH_NOTIFY); send_cluster_message(NULL, crm_msg_stonith_ng, reply, FALSE); } else { // Reply only to the originator stonith_send_reply(reply, cmd->options, cmd->origin, client); } crm_log_xml_trace(reply, "Reply"); free_xml(reply); if (stand_alone) { /* Do notification with a clean data object */ xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); stonith__xe_set_result(notify_data, result); 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); fenced_send_notification(T_STONITH_NOTIFY_FENCE, result, notify_data); fenced_send_notification(T_STONITH_NOTIFY_HISTORY, NULL, NULL); } } 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' action using %s", cmd->action, device->id); device->pending_ops = g_list_remove(device->pending_ops, cmd); } } static void st_child_done(int pid, const pcmk__action_result_t *result, void *user_data) { stonith_device_t *device = NULL; stonith_device_t *next_device = NULL; async_command_t *cmd = user_data; GList *gIter = NULL; GList *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 (!device->verified && pcmk__result_ok(result) && (pcmk__strcase_any_of(cmd->action, "list", "monitor", "status", NULL))) { device->verified = TRUE; } mainloop_set_trigger(device->work); } if (pcmk__result_ok(result)) { GList *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 ((cmd->device_next != NULL) && !is_action_required(cmd->action, device)) { /* 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_async_result(cmd, result, pid, next_device->id, false); schedule_stonith_command(cmd, next_device); /* Prevent cmd from being freed */ cmd = NULL; goto done; } send_async_reply(cmd, result, pid, false); if (!pcmk__result_ok(result)) { 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 (pcmk__str_eq(cmd->client, cmd_other->client, pcmk__str_casei) || !pcmk__str_eq(cmd->victim, cmd_other->victim, pcmk__str_casei) || !pcmk__str_eq(cmd->action, cmd_other->action, pcmk__str_casei) || !pcmk__str_eq(cmd->device, cmd_other->device, pcmk__str_casei)) { 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 fencing action '%s' targeting %s originating from " "client %s with identical fencing 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); send_async_reply(cmd_other, result, pid, true); 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; guint ndevices = g_list_length(devices); crm_info("Found %d matching device%s for target '%s'", ndevices, pcmk__plural_s(ndevices), cmd->victim); if (devices != NULL) { /* Order based on priority */ devices = g_list_sort(devices, sort_device_priority); device = g_hash_table_lookup(device_list, devices->data); } if (device == NULL) { // No device found pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; pcmk__format_result(&result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "No device configured for target '%s'", cmd->victim); send_async_reply(cmd, &result, 0, false); pcmk__reset_result(&result); free_async_command(cmd); g_list_free_full(devices, free); } else { // Device found, schedule it for fencing cmd->device_list = devices; cmd->device_next = devices->next; schedule_stonith_command(cmd, device); } } /*! * \internal * \brief Execute a fence action via the local node * * \param[in] msg Fencing request * \param[out] result Where to store result of fence action */ static void fence_locally(xmlNode *msg, pcmk__action_result_t *result) { 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); CRM_CHECK(result != NULL, return); if (cmd == NULL) { fenced_set_protocol_error(result); return; } device_id = crm_element_value(dev, F_STONITH_DEVICE); if (device_id != NULL) { device = g_hash_table_lookup(device_list, device_id); if (device == NULL) { crm_err("Requested device '%s' is not available", device_id); pcmk__format_result(result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, "Requested device '%s' not found", device_id); return; } schedule_stonith_command(cmd, device); } else { const char *host = crm_element_value(dev, F_STONITH_TARGET); if (pcmk_is_set(cmd->options, st_opt_cs_nodeid)) { int nodeid = 0; crm_node_t *node = NULL; pcmk__scan_min_int(host, &nodeid, 0); node = pcmk__search_known_node_cache(nodeid, NULL, CRM_GET_PEER_ANY); if (node != NULL) { 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); } pcmk__set_result(result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); } /*! * \internal * \brief Build an XML reply for a fencing operation * * \param[in] request Request that reply is for * \param[in] data If not NULL, add to reply as call data * \param[in] result Full result of fencing operation * * \return Newly created XML reply * \note The caller is responsible for freeing the result. * \note This has some overlap with construct_async_reply(), but that copies * values from an async_command_t, whereas this one copies them from the * request. */ xmlNode * fenced_construct_reply(xmlNode *request, xmlNode *data, pcmk__action_result_t *result) { xmlNode *reply = NULL; reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __func__); crm_xml_add(reply, F_TYPE, T_STONITH_NG); stonith__xe_set_result(reply, result); if (request == NULL) { /* Most likely, this is the result of a stonith operation that was * initiated before we came up. Unfortunately that means we lack enough * information to provide clients with a full result. * * @TODO Maybe synchronize this information at start-up? */ crm_warn("Missing request information for client notifications for " "operation with result '%s' (initiated before we came up?)", pcmk_exec_status_str(result->execution_status)); } else { const char *name = NULL; const char *value = NULL; // Attributes to copy from request to reply const char *names[] = { F_STONITH_OPERATION, F_STONITH_CALLID, F_STONITH_CLIENTID, F_STONITH_CLIENTNAME, F_STONITH_REMOTE_OP_ID, F_STONITH_CALLOPTS }; for (int lpc = 0; lpc < PCMK__NELEM(names); lpc++) { name = names[lpc]; value = crm_element_value(request, name); crm_xml_add(reply, name, value); } if (data != NULL) { add_message_xml(reply, F_STONITH_CALLDATA, data); } } return reply; } /*! * \internal * \brief Build an XML reply to an asynchronous fencing command * * \param[in] cmd Fencing command that reply is for * \param[in] result Command result */ static xmlNode * construct_async_reply(async_command_t *cmd, const pcmk__action_result_t *result) { xmlNode *reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __func__); 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); stonith__xe_set_result(reply, result); return reply; } bool fencing_peer_active(crm_node_t *peer) { if (peer == NULL) { return FALSE; } else if (peer->uname == NULL) { return FALSE; } else if (pcmk_is_set(peer->processes, crm_get_cluster_proc())) { return TRUE; } return FALSE; } void set_fencing_completed(remote_fencing_op_t *op) { struct timespec tv; qb_util_timespec_from_epoch_get(&tv); op->completed = tv.tv_sec; op->completed_nsec = tv.tv_nsec; } /*! * \internal * \brief Look for alternate node needed if local node shouldn't fence target * * \param[in] target Node that must be fenced * * \return Name of an alternate node that should fence \p target if any, * or NULL otherwise */ 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) && pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei)) { 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) && !pcmk__str_eq(entry->uname, target, pcmk__str_casei)) { alternate_host = entry->uname; break; } } if (alternate_host == NULL) { crm_err("No alternate host available to handle request " "for self-fencing with topology"); 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; } /*! * \internal * \brief Send a reply to a CPG peer or IPC client * * \param[in] reply XML reply to send * \param[in] call_options Send synchronously if st_opt_sync_call is set here * \param[in] remote_peer If not NULL, name of peer node to send CPG reply * \param[in] client If not NULL, client to send IPC reply */ static void stonith_send_reply(xmlNode *reply, int call_options, const char *remote_peer, pcmk__client_t *client) { CRM_CHECK((reply != NULL) && ((remote_peer != NULL) || (client != NULL)), return); if (remote_peer == NULL) { do_local_reply(reply, client, call_options); } else { send_cluster_message(crm_get_peer(0, remote_peer), crm_msg_stonith_ng, reply, FALSE); } } static void remove_relay_op(xmlNode * request) { xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, request, LOG_TRACE); const char *relay_op_id = NULL; const char *op_id = NULL; const char *client_name = NULL; const char *target = NULL; remote_fencing_op_t *relay_op = NULL; if (dev) { target = crm_element_value(dev, F_STONITH_TARGET); } relay_op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID_RELAY); op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID); client_name = crm_element_value(request, F_STONITH_CLIENTNAME); /* Delete RELAY operation. */ if (relay_op_id && target && pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei)) { relay_op = g_hash_table_lookup(stonith_remote_op_list, relay_op_id); if (relay_op) { GHashTableIter iter; remote_fencing_op_t *list_op = NULL; g_hash_table_iter_init(&iter, stonith_remote_op_list); /* If the operation to be deleted is registered as a duplicate, delete the registration. */ while (g_hash_table_iter_next(&iter, NULL, (void **)&list_op)) { GList *dup_iter = NULL; if (list_op != relay_op) { for (dup_iter = list_op->duplicates; dup_iter != NULL; dup_iter = dup_iter->next) { remote_fencing_op_t *other = dup_iter->data; if (other == relay_op) { other->duplicates = g_list_remove(other->duplicates, relay_op); break; } } } } crm_debug("Deleting relay op %s ('%s' targeting %s for %s), " "replaced by op %s ('%s' targeting %s for %s)", relay_op->id, relay_op->action, relay_op->target, relay_op->client_name, op_id, relay_op->action, target, client_name); g_hash_table_remove(stonith_remote_op_list, relay_op_id); } } } /*! * \internal * \brief Check whether an API request was sent by a privileged user * * API commands related to fencing configuration may be done only by privileged * IPC users (i.e. root or hacluster), because all other users should go through * the CIB to have ACLs applied. If no client was given, this is a peer request, * which is always allowed. * * \param[in] c IPC client that sent request (or NULL if sent by CPG peer) * \param[in] op Requested API operation (for logging only) * * \return true if sender is peer or privileged client, otherwise false */ static inline bool is_privileged(pcmk__client_t *c, const char *op) { if ((c == NULL) || pcmk_is_set(c->flags, pcmk__client_privileged)) { return true; } else { crm_warn("Rejecting IPC request '%s' from unprivileged client %s", pcmk__s(op, ""), pcmk__client_name(c)); return false; } } // CRM_OP_REGISTER static xmlNode * handle_register_request(pcmk__request_t *request) { xmlNode *reply = create_xml_node(NULL, "reply"); CRM_ASSERT(request->ipc_client != NULL); crm_xml_add(reply, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(reply, F_STONITH_CLIENTID, request->ipc_client->id); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); pcmk__set_request_flags(request, pcmk__request_reuse_options); return reply; } // STONITH_OP_EXEC static xmlNode * handle_agent_request(pcmk__request_t *request) { execute_agent_action(request->xml, &request->result); if (request->result.execution_status == PCMK_EXEC_PENDING) { return NULL; } return fenced_construct_reply(request->xml, NULL, &request->result); } // STONITH_OP_TIMEOUT_UPDATE static xmlNode * handle_update_timeout_request(pcmk__request_t *request) { const char *call_id = crm_element_value(request->xml, F_STONITH_CALLID); const char *client_id = crm_element_value(request->xml, F_STONITH_CLIENTID); int op_timeout = 0; crm_element_value_int(request->xml, F_STONITH_TIMEOUT, &op_timeout); do_stonith_async_timeout_update(client_id, call_id, op_timeout); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); return NULL; } // STONITH_OP_QUERY static xmlNode * handle_query_request(pcmk__request_t *request) { int timeout = 0; xmlNode *dev = NULL; const char *action = NULL; const char *target = NULL; const char *client_id = crm_element_value(request->xml, F_STONITH_CLIENTID); struct st_query_data *query = NULL; if (request->peer != NULL) { // Record it for the future notification create_remote_stonith_op(client_id, request->xml, TRUE); } /* Delete the DC node RELAY operation. */ remove_relay_op(request->xml); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); dev = get_xpath_object("//@" F_STONITH_ACTION, request->xml, LOG_NEVER); if (dev != NULL) { const char *device = crm_element_value(dev, F_STONITH_DEVICE); if (pcmk__str_eq(device, "manual_ack", pcmk__str_casei)) { return NULL; // No query or reply necessary } target = crm_element_value(dev, F_STONITH_TARGET); action = crm_element_value(dev, F_STONITH_ACTION); } crm_log_xml_trace(request->xml, "Query"); query = calloc(1, sizeof(struct st_query_data)); CRM_ASSERT(query != NULL); query->reply = fenced_construct_reply(request->xml, NULL, &request->result); pcmk__str_update(&query->remote_peer, request->peer); pcmk__str_update(&query->client_id, client_id); pcmk__str_update(&query->target, target); pcmk__str_update(&query->action, action); query->call_options = request->call_options; crm_element_value_int(request->xml, F_STONITH_TIMEOUT, &timeout); get_capable_devices(target, action, timeout, pcmk_is_set(query->call_options, st_opt_allow_suicide), query, stonith_query_capable_device_cb); return NULL; } // T_STONITH_NOTIFY static xmlNode * handle_notify_request(pcmk__request_t *request) { const char *flag_name = NULL; CRM_ASSERT(request->ipc_client != NULL); flag_name = crm_element_value(request->xml, F_STONITH_NOTIFY_ACTIVATE); if (flag_name != NULL) { crm_debug("Enabling %s callbacks for client %s", flag_name, pcmk__request_origin(request)); pcmk__set_client_flags(request->ipc_client, get_stonith_flag(flag_name)); } flag_name = crm_element_value(request->xml, F_STONITH_NOTIFY_DEACTIVATE); if (flag_name != NULL) { crm_debug("Disabling %s callbacks for client %s", flag_name, pcmk__request_origin(request)); pcmk__clear_client_flags(request->ipc_client, get_stonith_flag(flag_name)); } pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); pcmk__set_request_flags(request, pcmk__request_reuse_options); return pcmk__ipc_create_ack(request->ipc_flags, "ack", CRM_EX_OK); } // STONITH_OP_RELAY static xmlNode * handle_relay_request(pcmk__request_t *request) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request->xml, LOG_TRACE); crm_notice("Received forwarded fencing request from " "%s %s to fence (%s) peer %s", pcmk__request_origin_type(request), pcmk__request_origin(request), crm_element_value(dev, F_STONITH_ACTION), crm_element_value(dev, F_STONITH_TARGET)); if (initiate_remote_stonith_op(NULL, request->xml, FALSE) == NULL) { fenced_set_protocol_error(&request->result); return fenced_construct_reply(request->xml, NULL, &request->result); } pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); return NULL; } // STONITH_OP_FENCE static xmlNode * handle_fence_request(pcmk__request_t *request) { if ((request->peer != NULL) || stand_alone) { fence_locally(request->xml, &request->result); } else if (pcmk_is_set(request->call_options, st_opt_manual_ack)) { switch (fenced_handle_manual_confirmation(request->ipc_client, request->xml)) { case pcmk_rc_ok: pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); break; case EINPROGRESS: pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); break; default: fenced_set_protocol_error(&request->result); break; } } else { const char *alternate_host = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request->xml, 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 (request->ipc_client != NULL) { int tolerance = 0; crm_notice("Client %s wants to fence (%s) %s using %s", pcmk__request_origin(request), action, target, (device? device : "any device")); crm_element_value_int(dev, F_STONITH_TOLERANCE, &tolerance); if (stonith_check_fence_tolerance(tolerance, target, action)) { pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); return fenced_construct_reply(request->xml, NULL, &request->result); } } else { crm_notice("Peer %s wants to fence (%s) '%s' with device '%s'", request->peer, action, target, (device == NULL)? "(any)" : device); } alternate_host = check_alternate_host(target); if ((alternate_host != NULL) && (request->ipc_client != NULL)) { const char *client_id = NULL; remote_fencing_op_t *op = NULL; crm_notice("Forwarding self-fencing request to peer %s " "due to topology", alternate_host); if (request->ipc_client->id == 0) { client_id = crm_element_value(request->xml, F_STONITH_CLIENTID); } else { client_id = request->ipc_client->id; } /* Create a duplicate fencing operation to relay with the client ID. * When a query response is received, this operation should be * deleted to avoid keeping the duplicate around. */ op = create_remote_stonith_op(client_id, request->xml, FALSE); crm_xml_add(request->xml, F_STONITH_OPERATION, STONITH_OP_RELAY); crm_xml_add(request->xml, F_STONITH_CLIENTID, request->ipc_client->id); crm_xml_add(request->xml, F_STONITH_REMOTE_OP_ID, op->id); send_cluster_message(crm_get_peer(0, alternate_host), crm_msg_stonith_ng, request->xml, FALSE); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); } else if (initiate_remote_stonith_op(request->ipc_client, request->xml, FALSE) == NULL) { fenced_set_protocol_error(&request->result); } else { pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_PENDING, NULL); } } if (request->result.execution_status == PCMK_EXEC_PENDING) { return NULL; } return fenced_construct_reply(request->xml, NULL, &request->result); } // STONITH_OP_FENCE_HISTORY static xmlNode * handle_history_request(pcmk__request_t *request) { xmlNode *reply = NULL; xmlNode *data = NULL; stonith_fence_history(request->xml, &data, request->peer, request->call_options); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); if (!pcmk_is_set(request->call_options, st_opt_discard_reply)) { /* When the local node broadcasts its history, it sets * st_opt_discard_reply and doesn't need a reply. */ reply = fenced_construct_reply(request->xml, data, &request->result); } free_xml(data); return reply; } // STONITH_OP_DEVICE_ADD static xmlNode * handle_device_add_request(pcmk__request_t *request) { const char *op = crm_element_value(request->xml, F_STONITH_OPERATION); xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request->xml, LOG_ERR); if (is_privileged(request->ipc_client, op)) { int rc = stonith_device_register(dev, FALSE); pcmk__set_result(&request->result, ((rc == pcmk_ok)? CRM_EX_OK : CRM_EX_ERROR), stonith__legacy2status(rc), ((rc == pcmk_ok)? NULL : pcmk_strerror(rc))); } else { pcmk__set_result(&request->result, CRM_EX_INSUFFICIENT_PRIV, PCMK_EXEC_INVALID, "Unprivileged users must register device via CIB"); } fenced_send_device_notification(op, &request->result, (dev == NULL)? NULL : ID(dev)); return fenced_construct_reply(request->xml, NULL, &request->result); } // STONITH_OP_DEVICE_DEL static xmlNode * handle_device_delete_request(pcmk__request_t *request) { xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request->xml, LOG_ERR); const char *device_id = crm_element_value(dev, XML_ATTR_ID); const char *op = crm_element_value(request->xml, F_STONITH_OPERATION); if (is_privileged(request->ipc_client, op)) { stonith_device_remove(device_id, false); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); } else { pcmk__set_result(&request->result, CRM_EX_INSUFFICIENT_PRIV, PCMK_EXEC_INVALID, "Unprivileged users must delete device via CIB"); } fenced_send_device_notification(op, &request->result, device_id); return fenced_construct_reply(request->xml, NULL, &request->result); } // STONITH_OP_LEVEL_ADD static xmlNode * handle_level_add_request(pcmk__request_t *request) { char *desc = NULL; const char *op = crm_element_value(request->xml, F_STONITH_OPERATION); if (is_privileged(request->ipc_client, op)) { fenced_register_level(request->xml, &desc, &request->result); } else { unpack_level_request(request->xml, NULL, NULL, NULL, &desc); pcmk__set_result(&request->result, CRM_EX_INSUFFICIENT_PRIV, PCMK_EXEC_INVALID, "Unprivileged users must add level via CIB"); } fenced_send_level_notification(op, &request->result, desc); free(desc); return fenced_construct_reply(request->xml, NULL, &request->result); } // STONITH_OP_LEVEL_DEL static xmlNode * handle_level_delete_request(pcmk__request_t *request) { char *desc = NULL; const char *op = crm_element_value(request->xml, F_STONITH_OPERATION); if (is_privileged(request->ipc_client, op)) { fenced_unregister_level(request->xml, &desc, &request->result); } else { unpack_level_request(request->xml, NULL, NULL, NULL, &desc); pcmk__set_result(&request->result, CRM_EX_INSUFFICIENT_PRIV, PCMK_EXEC_INVALID, "Unprivileged users must delete level via CIB"); } fenced_send_level_notification(op, &request->result, desc); free(desc); return fenced_construct_reply(request->xml, NULL, &request->result); } // CRM_OP_RM_NODE_CACHE static xmlNode * handle_cache_request(pcmk__request_t *request) { int node_id = 0; const char *name = NULL; crm_element_value_int(request->xml, XML_ATTR_ID, &node_id); name = crm_element_value(request->xml, XML_ATTR_UNAME); reap_crm_member(node_id, name); pcmk__set_result(&request->result, CRM_EX_OK, PCMK_EXEC_DONE, NULL); return NULL; } static xmlNode * handle_unknown_request(pcmk__request_t *request) { const char *op = crm_element_value(request->xml, F_STONITH_OPERATION); crm_err("Unknown IPC request %s from %s %s", op, pcmk__request_origin_type(request), pcmk__request_origin(request)); pcmk__format_result(&request->result, CRM_EX_PROTOCOL, PCMK_EXEC_INVALID, "Unknown IPC request type '%s' (bug?)", pcmk__s(op, "")); return fenced_construct_reply(request->xml, NULL, &request->result); } static void fenced_register_handlers(void) { pcmk__server_command_t handlers[] = { { CRM_OP_REGISTER, handle_register_request }, { STONITH_OP_EXEC, handle_agent_request }, { STONITH_OP_TIMEOUT_UPDATE, handle_update_timeout_request }, { STONITH_OP_QUERY, handle_query_request }, { T_STONITH_NOTIFY, handle_notify_request }, { STONITH_OP_RELAY, handle_relay_request }, { STONITH_OP_FENCE, handle_fence_request }, { STONITH_OP_FENCE_HISTORY, handle_history_request }, { STONITH_OP_DEVICE_ADD, handle_device_add_request }, { STONITH_OP_DEVICE_DEL, handle_device_delete_request }, { STONITH_OP_LEVEL_ADD, handle_level_add_request }, { STONITH_OP_LEVEL_DEL, handle_level_delete_request }, { CRM_OP_RM_NODE_CACHE, handle_cache_request }, { NULL, handle_unknown_request }, }; fenced_handlers = pcmk__register_handlers(handlers); } void fenced_unregister_handlers(void) { if (fenced_handlers != NULL) { g_hash_table_destroy(fenced_handlers); fenced_handlers = NULL; } } static void handle_request(pcmk__request_t *request) { xmlNode *reply = NULL; const char *reason = NULL; if (fenced_handlers == NULL) { fenced_register_handlers(); } reply = pcmk__process_request(request, fenced_handlers); if (reply != NULL) { if (pcmk_is_set(request->flags, pcmk__request_reuse_options) && (request->ipc_client != NULL)) { /* Certain IPC-only commands must reuse the call options from the * original request rather than the ones set by stonith_send_reply() * -> do_local_reply(). */ pcmk__ipc_send_xml(request->ipc_client, request->ipc_id, reply, request->ipc_flags); request->ipc_client->request_id = 0; } else { stonith_send_reply(reply, request->call_options, request->peer, request->ipc_client); } free_xml(reply); } reason = request->result.exit_reason; crm_debug("Processed %s request from %s %s: %s%s%s%s", request->op, pcmk__request_origin_type(request), pcmk__request_origin(request), pcmk_exec_status_str(request->result.execution_status), (reason == NULL)? "" : " (", (reason == NULL)? "" : reason, (reason == NULL)? "" : ")"); } static void handle_reply(pcmk__client_t *client, xmlNode *request, const char *remote_peer) { // Copy, because request might be freed before we want to log this char *op = crm_element_value_copy(request, F_STONITH_OPERATION); if (pcmk__str_eq(op, STONITH_OP_QUERY, pcmk__str_none)) { process_remote_stonith_query(request); } else if (pcmk__str_any_of(op, T_STONITH_NOTIFY, STONITH_OP_FENCE, NULL)) { fenced_process_fencing_reply(request); } else { crm_err("Ignoring unknown %s reply from %s %s", pcmk__s(op, "untyped"), ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client))); crm_log_xml_warn(request, "UnknownOp"); free(op); return; } crm_debug("Processed %s reply from %s %s", op, ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client))); free(op); } /*! * \internal * \brief Handle a message from an IPC client or CPG peer * * \param[in] client If not NULL, IPC client that sent message * \param[in] id If from IPC client, IPC message ID * \param[in] flags Message flags * \param[in] message Message XML * \param[in] remote_peer If not NULL, CPG peer that sent message */ void stonith_command(pcmk__client_t *client, uint32_t id, uint32_t flags, xmlNode *message, const char *remote_peer) { int call_options = st_opt_none; bool is_reply = false; CRM_CHECK(message != NULL, return); if (get_xpath_object("//" T_STONITH_REPLY, message, LOG_NEVER) != NULL) { is_reply = true; } crm_element_value_int(message, F_STONITH_CALLOPTS, &call_options); crm_debug("Processing %ssynchronous %s %s %u from %s %s", pcmk_is_set(call_options, st_opt_sync_call)? "" : "a", crm_element_value(message, F_STONITH_OPERATION), (is_reply? "reply" : "request"), id, ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client))); if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (is_reply) { handle_reply(client, message, remote_peer); } else { pcmk__request_t request = { .ipc_client = client, .ipc_id = id, .ipc_flags = flags, .peer = remote_peer, .xml = message, .call_options = call_options, .result = PCMK__UNKNOWN_RESULT, }; request.op = crm_element_value(request.xml, F_STONITH_OPERATION); CRM_CHECK(request.op != NULL, return); if (pcmk_is_set(request.call_options, st_opt_sync_call)) { pcmk__set_request_flags(&request, pcmk__request_sync); } handle_request(&request); pcmk__reset_result(&request.result); } } diff --git a/include/crm/common/options_internal.h b/include/crm/common/options_internal.h index ea025417fd..494a3e7235 100644 --- a/include/crm/common/options_internal.h +++ b/include/crm/common/options_internal.h @@ -1,156 +1,160 @@ /* * Copyright 2006-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef PCMK__OPTIONS_INTERNAL__H # define PCMK__OPTIONS_INTERNAL__H # ifndef PCMK__CONFIG_H # define PCMK__CONFIG_H # include // HAVE_GETOPT, _Noreturn # endif # include // GHashTable # include // bool /* * Command-line option handling * * This will all eventually go away as everything is converted to use GOption */ # ifdef HAVE_GETOPT_H # include # else # define no_argument 0 # define required_argument 1 # endif enum pcmk__cli_option_flags { pcmk__option_default = (1 << 0), pcmk__option_hidden = (1 << 1), pcmk__option_paragraph = (1 << 2), pcmk__option_example = (1 << 3), }; typedef struct pcmk__cli_option_s { /* Fields from 'struct option' in getopt.h */ /* name of long option */ const char *name; /* * one of no_argument, required_argument, and optional_argument: * whether option takes an argument */ int has_arg; /* if not NULL, set *flag to val when option found */ int *flag; /* if flag not NULL, value to set *flag to; else return value */ int val; /* Custom fields */ const char *desc; long flags; } pcmk__cli_option_t; void pcmk__set_cli_options(const char *short_options, const char *usage, pcmk__cli_option_t *long_options, const char *app_desc); int pcmk__next_cli_option(int argc, char **argv, int *index, const char **longname); _Noreturn void pcmk__cli_help(char cmd, crm_exit_t exit_code); void pcmk__cli_option_cleanup(void); /* * Environment variable option handling */ const char *pcmk__env_option(const char *option); void pcmk__set_env_option(const char *option, const char *value); bool pcmk__env_option_enabled(const char *daemon, const char *option); /* * Cluster option handling */ typedef struct pcmk__cluster_option_s { const char *name; const char *alt_name; const char *type; const char *values; const char *default_value; bool (*is_valid)(const char *); const char *description_short; const char *description_long; } pcmk__cluster_option_t; const char *pcmk__cluster_option(GHashTable *options, pcmk__cluster_option_t *option_list, int len, const char *name); char *pcmk__format_option_metadata(const char *name, const char *desc_short, const char *desc_long, pcmk__cluster_option_t *option_list, int len); void pcmk__validate_cluster_options(GHashTable *options, pcmk__cluster_option_t *option_list, int len); bool pcmk__valid_interval_spec(const char *value); bool pcmk__valid_boolean(const char *value); bool pcmk__valid_number(const char *value); bool pcmk__valid_positive_number(const char *value); bool pcmk__valid_quorum(const char *value); bool pcmk__valid_script(const char *value); bool pcmk__valid_percentage(const char *value); // from watchdog.c long pcmk__get_sbd_timeout(void); bool pcmk__get_sbd_sync_resource_startup(void); long pcmk__auto_watchdog_timeout(void); bool pcmk__valid_sbd_timeout(const char *value); // Constants for environment variable names #define PCMK__ENV_BLACKBOX "blackbox" #define PCMK__ENV_CLUSTER_TYPE "cluster_type" #define PCMK__ENV_DEBUG "debug" #define PCMK__ENV_LOGFACILITY "logfacility" #define PCMK__ENV_LOGFILE "logfile" #define PCMK__ENV_LOGPRIORITY "logpriority" #define PCMK__ENV_MCP "mcp" #define PCMK__ENV_NODE_START_STATE "node_start_state" #define PCMK__ENV_PHYSICAL_HOST "physical_host" #define PCMK__ENV_QUORUM_TYPE "quorum_type" #define PCMK__ENV_SHUTDOWN_DELAY "shutdown_delay" #define PCMK__ENV_STDERR "stderr" // Constants for cluster option names #define PCMK__OPT_NODE_HEALTH_BASE "node-health-base" #define PCMK__OPT_NODE_HEALTH_GREEN "node-health-green" #define PCMK__OPT_NODE_HEALTH_RED "node-health-red" #define PCMK__OPT_NODE_HEALTH_STRATEGY "node-health-strategy" #define PCMK__OPT_NODE_HEALTH_YELLOW "node-health-yellow" // Constants for meta-attribute names #define PCMK__META_ALLOW_UNHEALTHY_NODES "allow-unhealthy-nodes" // Constants for enumerated values for various options #define PCMK__VALUE_CUSTOM "custom" +#define PCMK__VALUE_FENCING "fencing" #define PCMK__VALUE_GREEN "green" #define PCMK__VALUE_MIGRATE_ON_RED "migrate-on-red" #define PCMK__VALUE_NONE "none" +#define PCMK__VALUE_NOTHING "nothing" #define PCMK__VALUE_ONLY_GREEN "only-green" #define PCMK__VALUE_PROGRESSIVE "progressive" +#define PCMK__VALUE_QUORUM "quorum" #define PCMK__VALUE_RED "red" +#define PCMK__VALUE_UNFENCING "unfencing" #define PCMK__VALUE_YELLOW "yellow" #endif // PCMK__OPTIONS_INTERNAL__H diff --git a/lib/pengine/complex.c b/lib/pengine/complex.c index fc9028e81e..305ff46978 100644 --- a/lib/pengine/complex.c +++ b/lib/pengine/complex.c @@ -1,1134 +1,1134 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include void populate_hash(xmlNode * nvpair_list, GHashTable * hash, const char **attrs, int attrs_length); resource_object_functions_t resource_class_functions[] = { { native_unpack, native_find_rsc, native_parameter, native_print, native_active, native_resource_state, native_location, native_free, pe__count_common, pe__native_is_filtered, }, { group_unpack, native_find_rsc, native_parameter, group_print, group_active, group_resource_state, native_location, group_free, pe__count_common, pe__group_is_filtered, }, { clone_unpack, native_find_rsc, native_parameter, clone_print, clone_active, clone_resource_state, native_location, clone_free, pe__count_common, pe__clone_is_filtered, }, { pe__unpack_bundle, native_find_rsc, native_parameter, pe__print_bundle, pe__bundle_active, pe__bundle_resource_state, native_location, pe__free_bundle, pe__count_bundle, pe__bundle_is_filtered, } }; static enum pe_obj_types get_resource_type(const char *name) { if (pcmk__str_eq(name, XML_CIB_TAG_RESOURCE, pcmk__str_casei)) { return pe_native; } else if (pcmk__str_eq(name, XML_CIB_TAG_GROUP, pcmk__str_casei)) { return pe_group; } else if (pcmk__str_eq(name, XML_CIB_TAG_INCARNATION, pcmk__str_casei)) { return pe_clone; } else if (pcmk__str_eq(name, PCMK_XE_PROMOTABLE_LEGACY, pcmk__str_casei)) { // @COMPAT deprecated since 2.0.0 return pe_clone; } else if (pcmk__str_eq(name, XML_CIB_TAG_CONTAINER, pcmk__str_casei)) { return pe_container; } return pe_unknown; } static void dup_attr(gpointer key, gpointer value, gpointer user_data) { add_hash_param(user_data, key, value); } static void expand_parents_fixed_nvpairs(pe_resource_t * rsc, pe_rule_eval_data_t * rule_data, GHashTable * meta_hash, pe_working_set_t * data_set) { GHashTable *parent_orig_meta = pcmk__strkey_table(free, free); pe_resource_t *p = rsc->parent; if (p == NULL) { return ; } /* Search all parent resources, get the fixed value of "meta_attributes" set only in the original xml, and stack it in the hash table. */ /* The fixed value of the lower parent resource takes precedence and is not overwritten. */ while(p != NULL) { /* A hash table for comparison is generated, including the id-ref. */ pe__unpack_dataset_nvpairs(p->xml, XML_TAG_META_SETS, rule_data, parent_orig_meta, NULL, FALSE, data_set); p = p->parent; } /* If there is a fixed value of "meta_attributes" of the parent resource, it will be processed. */ if (parent_orig_meta != NULL) { GHashTableIter iter; char *key = NULL; char *value = NULL; g_hash_table_iter_init(&iter, parent_orig_meta); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { /* Parameters set in the original xml of the parent resource will also try to overwrite the child resource. */ /* Attributes that already exist in the child lease are not updated. */ dup_attr(key, value, meta_hash); } } if (parent_orig_meta != NULL) { g_hash_table_destroy(parent_orig_meta); } return ; } void get_meta_attributes(GHashTable * meta_hash, pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set) { pe_rsc_eval_data_t rsc_rule_data = { .standard = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS), .provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER), .agent = crm_element_value(rsc->xml, XML_EXPR_ATTR_TYPE) }; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = &rsc_rule_data, .op_data = NULL }; if (node) { rule_data.node_hash = node->details->attrs; } for (xmlAttrPtr a = pcmk__xe_first_attr(rsc->xml); a != NULL; a = a->next) { const char *prop_name = (const char *) a->name; const char *prop_value = crm_element_value(rsc->xml, prop_name); add_hash_param(meta_hash, prop_name, prop_value); } pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_META_SETS, &rule_data, meta_hash, NULL, FALSE, data_set); /* Set the "meta_attributes" explicitly set in the parent resource to the hash table of the child resource. */ /* If it is already explicitly set as a child, it will not be overwritten. */ if (rsc->parent != NULL) { expand_parents_fixed_nvpairs(rsc, &rule_data, meta_hash, data_set); } /* check the defaults */ pe__unpack_dataset_nvpairs(data_set->rsc_defaults, XML_TAG_META_SETS, &rule_data, meta_hash, NULL, FALSE, data_set); /* If there is "meta_attributes" that the parent resource has not explicitly set, set a value that is not set from rsc_default either. */ /* The values already set up to this point will not be overwritten. */ if (rsc->parent) { g_hash_table_foreach(rsc->parent->meta, dup_attr, meta_hash); } } void get_rsc_attributes(GHashTable * meta_hash, pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set) { pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; if (node) { rule_data.node_hash = node->details->attrs; } pe__unpack_dataset_nvpairs(rsc->xml, XML_TAG_ATTR_SETS, &rule_data, meta_hash, NULL, FALSE, data_set); /* set anything else based on the parent */ if (rsc->parent != NULL) { get_rsc_attributes(meta_hash, rsc->parent, node, data_set); } else { /* and finally check the defaults */ pe__unpack_dataset_nvpairs(data_set->rsc_defaults, XML_TAG_ATTR_SETS, &rule_data, meta_hash, NULL, FALSE, data_set); } } #if ENABLE_VERSIONED_ATTRS void pe_get_versioned_attributes(xmlNode * meta_hash, pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set) { pe_rule_eval_data_t rule_data = { .node_hash = (node == NULL)? NULL : node->details->attrs, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; pe_eval_versioned_attributes(data_set->input, rsc->xml, XML_TAG_ATTR_SETS, &rule_data, meta_hash, NULL); /* set anything else based on the parent */ if (rsc->parent != NULL) { pe_get_versioned_attributes(meta_hash, rsc->parent, node, data_set); } else { /* and finally check the defaults */ pe_eval_versioned_attributes(data_set->input, data_set->rsc_defaults, XML_TAG_ATTR_SETS, &rule_data, meta_hash, NULL); } } #endif static char * template_op_key(xmlNode * op) { const char *name = crm_element_value(op, "name"); const char *role = crm_element_value(op, "role"); char *key = NULL; if ((role == NULL) || pcmk__strcase_any_of(role, RSC_ROLE_STARTED_S, RSC_ROLE_UNPROMOTED_S, RSC_ROLE_UNPROMOTED_LEGACY_S, NULL)) { role = RSC_ROLE_UNKNOWN_S; } key = crm_strdup_printf("%s-%s", name, role); return key; } static gboolean unpack_template(xmlNode * xml_obj, xmlNode ** expanded_xml, pe_working_set_t * data_set) { xmlNode *cib_resources = NULL; xmlNode *template = NULL; xmlNode *new_xml = NULL; xmlNode *child_xml = NULL; xmlNode *rsc_ops = NULL; xmlNode *template_ops = NULL; const char *template_ref = NULL; const char *clone = NULL; const char *id = NULL; if (xml_obj == NULL) { pe_err("No resource object for template unpacking"); return FALSE; } template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE); if (template_ref == NULL) { return TRUE; } id = ID(xml_obj); if (id == NULL) { pe_err("'%s' object must have a id", crm_element_name(xml_obj)); return FALSE; } if (pcmk__str_eq(template_ref, id, pcmk__str_none)) { pe_err("The resource object '%s' should not reference itself", id); return FALSE; } cib_resources = get_xpath_object("//"XML_CIB_TAG_RESOURCES, data_set->input, LOG_TRACE); if (cib_resources == NULL) { pe_err("No resources configured"); return FALSE; } template = pcmk__xe_match(cib_resources, XML_CIB_TAG_RSC_TEMPLATE, XML_ATTR_ID, template_ref); if (template == NULL) { pe_err("No template named '%s'", template_ref); return FALSE; } new_xml = copy_xml(template); xmlNodeSetName(new_xml, xml_obj->name); crm_xml_replace(new_xml, XML_ATTR_ID, id); clone = crm_element_value(xml_obj, XML_RSC_ATTR_INCARNATION); if(clone) { crm_xml_add(new_xml, XML_RSC_ATTR_INCARNATION, clone); } template_ops = find_xml_node(new_xml, "operations", FALSE); for (child_xml = pcmk__xe_first_child(xml_obj); child_xml != NULL; child_xml = pcmk__xe_next(child_xml)) { xmlNode *new_child = NULL; new_child = add_node_copy(new_xml, child_xml); if (pcmk__str_eq((const char *)new_child->name, "operations", pcmk__str_none)) { rsc_ops = new_child; } } if (template_ops && rsc_ops) { xmlNode *op = NULL; GHashTable *rsc_ops_hash = pcmk__strkey_table(free, NULL); for (op = pcmk__xe_first_child(rsc_ops); op != NULL; op = pcmk__xe_next(op)) { char *key = template_op_key(op); g_hash_table_insert(rsc_ops_hash, key, op); } for (op = pcmk__xe_first_child(template_ops); op != NULL; op = pcmk__xe_next(op)) { char *key = template_op_key(op); if (g_hash_table_lookup(rsc_ops_hash, key) == NULL) { add_node_copy(rsc_ops, op); } free(key); } if (rsc_ops_hash) { g_hash_table_destroy(rsc_ops_hash); } free_xml(template_ops); } /*free_xml(*expanded_xml); */ *expanded_xml = new_xml; /* Disable multi-level templates for now */ /*if(unpack_template(new_xml, expanded_xml, data_set) == FALSE) { free_xml(*expanded_xml); *expanded_xml = NULL; return FALSE; } */ return TRUE; } static gboolean add_template_rsc(xmlNode * xml_obj, pe_working_set_t * data_set) { const char *template_ref = NULL; const char *id = NULL; if (xml_obj == NULL) { pe_err("No resource object for processing resource list of template"); return FALSE; } template_ref = crm_element_value(xml_obj, XML_CIB_TAG_RSC_TEMPLATE); if (template_ref == NULL) { return TRUE; } id = ID(xml_obj); if (id == NULL) { pe_err("'%s' object must have a id", crm_element_name(xml_obj)); return FALSE; } if (pcmk__str_eq(template_ref, id, pcmk__str_none)) { pe_err("The resource object '%s' should not reference itself", id); return FALSE; } if (add_tag_ref(data_set->template_rsc_sets, template_ref, id) == FALSE) { return FALSE; } return TRUE; } static bool detect_promotable(pe_resource_t *rsc) { const char *promotable = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_PROMOTABLE); if (crm_is_true(promotable)) { return TRUE; } // @COMPAT deprecated since 2.0.0 if (pcmk__str_eq(crm_element_name(rsc->xml), PCMK_XE_PROMOTABLE_LEGACY, pcmk__str_casei)) { /* @TODO in some future version, pe_warn_once() here, * then drop support in even later version */ g_hash_table_insert(rsc->meta, strdup(XML_RSC_ATTR_PROMOTABLE), strdup(XML_BOOLEAN_TRUE)); return TRUE; } return FALSE; } static void free_params_table(gpointer data) { g_hash_table_destroy((GHashTable *) data); } /*! * \brief Get a table of resource parameters * * \param[in] rsc Resource to query * \param[in] node Node for evaluating rules (NULL for defaults) * \param[in] data_set Cluster working set * * \return Hash table containing resource parameter names and values * (or NULL if \p rsc or \p data_set is NULL) * \note The returned table will be destroyed when the resource is freed, so * callers should not destroy it. */ GHashTable * pe_rsc_params(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { GHashTable *params_on_node = NULL; /* A NULL node is used to request the resource's default parameters * (not evaluated for node), but we always want something non-NULL * as a hash table key. */ const char *node_name = ""; // Sanity check if ((rsc == NULL) || (data_set == NULL)) { return NULL; } if ((node != NULL) && (node->details->uname != NULL)) { node_name = node->details->uname; } // Find the parameter table for given node if (rsc->parameter_cache == NULL) { rsc->parameter_cache = pcmk__strikey_table(free, free_params_table); } else { params_on_node = g_hash_table_lookup(rsc->parameter_cache, node_name); } // If none exists yet, create one with parameters evaluated for node if (params_on_node == NULL) { params_on_node = pcmk__strkey_table(free, free); get_rsc_attributes(params_on_node, rsc, node, data_set); g_hash_table_insert(rsc->parameter_cache, strdup(node_name), params_on_node); } return params_on_node; } gboolean common_unpack(xmlNode * xml_obj, pe_resource_t ** rsc, pe_resource_t * parent, pe_working_set_t * data_set) { bool isdefault = FALSE; xmlNode *expanded_xml = NULL; xmlNode *ops = NULL; const char *value = NULL; const char *rclass = NULL; /* Look for this after any templates have been expanded */ const char *id = crm_element_value(xml_obj, XML_ATTR_ID); bool guest_node = FALSE; bool remote_node = FALSE; bool has_versioned_params = FALSE; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; crm_log_xml_trace(xml_obj, "Processing resource input..."); if (id == NULL) { pe_err("Must specify id tag in "); return FALSE; } else if (rsc == NULL) { pe_err("Nowhere to unpack resource into"); return FALSE; } if (unpack_template(xml_obj, &expanded_xml, data_set) == FALSE) { return FALSE; } *rsc = calloc(1, sizeof(pe_resource_t)); (*rsc)->cluster = data_set; if (expanded_xml) { crm_log_xml_trace(expanded_xml, "Expanded resource..."); (*rsc)->xml = expanded_xml; (*rsc)->orig_xml = xml_obj; } else { (*rsc)->xml = xml_obj; (*rsc)->orig_xml = NULL; } /* Do not use xml_obj from here on, use (*rsc)->xml in case templates are involved */ rclass = crm_element_value((*rsc)->xml, XML_AGENT_ATTR_CLASS); (*rsc)->parent = parent; ops = find_xml_node((*rsc)->xml, "operations", FALSE); (*rsc)->ops_xml = expand_idref(ops, data_set->input); (*rsc)->variant = get_resource_type(crm_element_name((*rsc)->xml)); if ((*rsc)->variant == pe_unknown) { pe_err("Unknown resource type: %s", crm_element_name((*rsc)->xml)); free(*rsc); return FALSE; } #if ENABLE_VERSIONED_ATTRS (*rsc)->versioned_parameters = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS); #endif (*rsc)->meta = pcmk__strkey_table(free, free); (*rsc)->allowed_nodes = pcmk__strkey_table(NULL, free); (*rsc)->known_on = pcmk__strkey_table(NULL, free); value = crm_element_value((*rsc)->xml, XML_RSC_ATTR_INCARNATION); if (value) { (*rsc)->id = crm_strdup_printf("%s:%s", id, value); add_hash_param((*rsc)->meta, XML_RSC_ATTR_INCARNATION, value); } else { (*rsc)->id = strdup(id); } (*rsc)->fns = &resource_class_functions[(*rsc)->variant]; pe_rsc_trace((*rsc), "Unpacking resource..."); get_meta_attributes((*rsc)->meta, *rsc, NULL, data_set); (*rsc)->parameters = pe_rsc_params(*rsc, NULL, data_set); // \deprecated #if ENABLE_VERSIONED_ATTRS pe_get_versioned_attributes((*rsc)->versioned_parameters, *rsc, NULL, data_set); #endif (*rsc)->flags = 0; pe__set_resource_flags(*rsc, pe_rsc_runnable|pe_rsc_provisional); if (!pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { pe__set_resource_flags(*rsc, pe_rsc_managed); } (*rsc)->rsc_cons = NULL; (*rsc)->rsc_tickets = NULL; (*rsc)->actions = NULL; (*rsc)->role = RSC_ROLE_STOPPED; (*rsc)->next_role = RSC_ROLE_UNKNOWN; (*rsc)->recovery_type = recovery_stop_start; (*rsc)->stickiness = 0; (*rsc)->migration_threshold = INFINITY; (*rsc)->failure_timeout = 0; value = g_hash_table_lookup((*rsc)->meta, XML_CIB_ATTR_PRIORITY); (*rsc)->priority = char2score(value); value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_CRITICAL); if ((value == NULL) || crm_is_true(value)) { pe__set_resource_flags(*rsc, pe_rsc_critical); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_NOTIFY); if (crm_is_true(value)) { pe__set_resource_flags(*rsc, pe_rsc_notify); } if (xml_contains_remote_node((*rsc)->xml)) { (*rsc)->is_remote_node = TRUE; if (g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_CONTAINER)) { guest_node = TRUE; } else { remote_node = TRUE; } } value = g_hash_table_lookup((*rsc)->meta, XML_OP_ATTR_ALLOW_MIGRATE); #if ENABLE_VERSIONED_ATTRS has_versioned_params = xml_has_children((*rsc)->versioned_parameters); #endif if (crm_is_true(value) && has_versioned_params) { pe_rsc_trace((*rsc), "Migration is disabled for resources with versioned parameters"); } else if (crm_is_true(value)) { pe__set_resource_flags(*rsc, pe_rsc_allow_migrate); } else if ((value == NULL) && remote_node && !has_versioned_params) { /* By default, we want remote nodes to be able * to float around the cluster without having to stop all the * resources within the remote-node before moving. Allowing * migration support enables this feature. If this ever causes * problems, migration support can be explicitly turned off with * allow-migrate=false. * We don't support migration for versioned resources, though. */ pe__set_resource_flags(*rsc, pe_rsc_allow_migrate); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MANAGED); if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) { if (crm_is_true(value)) { pe__set_resource_flags(*rsc, pe_rsc_managed); } else { pe__clear_resource_flags(*rsc, pe_rsc_managed); } } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MAINTENANCE); if (crm_is_true(value)) { pe__clear_resource_flags(*rsc, pe_rsc_managed); pe__set_resource_flags(*rsc, pe_rsc_maintenance); } if (pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { pe__clear_resource_flags(*rsc, pe_rsc_managed); pe__set_resource_flags(*rsc, pe_rsc_maintenance); } if (pe_rsc_is_clone(uber_parent(*rsc))) { value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_UNIQUE); if (crm_is_true(value)) { pe__set_resource_flags(*rsc, pe_rsc_unique); } if (detect_promotable(*rsc)) { pe__set_resource_flags(*rsc, pe_rsc_promotable); } } else { pe__set_resource_flags(*rsc, pe_rsc_unique); } pe_rsc_trace((*rsc), "Options for %s", (*rsc)->id); value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_RESTART); if (pcmk__str_eq(value, "restart", pcmk__str_casei)) { (*rsc)->restart_type = pe_restart_restart; pe_rsc_trace((*rsc), "\tDependency restart handling: restart"); pe_warn_once(pe_wo_restart_type, "Support for restart-type is deprecated and will be removed in a future release"); } else { (*rsc)->restart_type = pe_restart_ignore; pe_rsc_trace((*rsc), "\tDependency restart handling: ignore"); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_MULTIPLE); if (pcmk__str_eq(value, "stop_only", pcmk__str_casei)) { (*rsc)->recovery_type = recovery_stop_only; pe_rsc_trace((*rsc), "\tMultiple running resource recovery: stop only"); } else if (pcmk__str_eq(value, "block", pcmk__str_casei)) { (*rsc)->recovery_type = recovery_block; pe_rsc_trace((*rsc), "\tMultiple running resource recovery: block"); } else if (pcmk__str_eq(value, "stop_unexpected", pcmk__str_casei)) { (*rsc)->recovery_type = recovery_stop_unexpected; pe_rsc_trace((*rsc), "\tMultiple running resource recovery: " "stop unexpected instances"); } else { // "stop_start" if (!pcmk__str_eq(value, "stop_start", pcmk__str_casei|pcmk__str_null_matches)) { pe_warn("%s is not a valid value for " XML_RSC_ATTR_MULTIPLE ", using default of \"stop_start\"", value); } (*rsc)->recovery_type = recovery_stop_start; pe_rsc_trace((*rsc), "\tMultiple running resource recovery: stop/start"); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_STICKINESS); if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) { (*rsc)->stickiness = char2score(value); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_FAIL_STICKINESS); if (value != NULL && !pcmk__str_eq("default", value, pcmk__str_casei)) { (*rsc)->migration_threshold = char2score(value); if ((*rsc)->migration_threshold < 0) { /* @TODO We use 1 here to preserve previous behavior, but this * should probably use the default (INFINITY) or 0 (to disable) * instead. */ pe_warn_once(pe_wo_neg_threshold, XML_RSC_ATTR_FAIL_STICKINESS " must be non-negative, using 1 instead"); (*rsc)->migration_threshold = 1; } } if (pcmk__str_eq(rclass, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { pe__set_working_set_flags(data_set, pe_flag_have_stonith_resource); pe__set_resource_flags(*rsc, pe_rsc_fence_device); } value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_REQUIRES); handle_requires_pref: - if (pcmk__str_eq(value, "nothing", pcmk__str_casei)) { + if (pcmk__str_eq(value, PCMK__VALUE_NOTHING, pcmk__str_casei)) { - } else if (pcmk__str_eq(value, "quorum", pcmk__str_casei)) { + } else if (pcmk__str_eq(value, PCMK__VALUE_QUORUM, pcmk__str_casei)) { pe__set_resource_flags(*rsc, pe_rsc_needs_quorum); - } else if (pcmk__str_eq(value, "unfencing", pcmk__str_casei)) { + } else if (pcmk__str_eq(value, PCMK__VALUE_UNFENCING, pcmk__str_casei)) { if (pcmk_is_set((*rsc)->flags, pe_rsc_fence_device)) { - pcmk__config_warn("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s " - "to 'quorum' because fencing devices cannot " - "require unfencing", (*rsc)->id); - value = "quorum"; + pcmk__config_warn("Resetting \"" XML_RSC_ATTR_REQUIRES "\" for %s " + "to \"" PCMK__VALUE_QUORUM "\" because fencing " + "devices cannot require unfencing", (*rsc)->id); + value = PCMK__VALUE_QUORUM; isdefault = TRUE; goto handle_requires_pref; } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { - pcmk__config_warn("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s " - "to 'quorum' because fencing is disabled", - (*rsc)->id); - value = "quorum"; + pcmk__config_warn("Resetting \"" XML_RSC_ATTR_REQUIRES "\" for %s " + "to \"" PCMK__VALUE_QUORUM "\" because fencing " + "is disabled", (*rsc)->id); + value = PCMK__VALUE_QUORUM; isdefault = TRUE; goto handle_requires_pref; } else { pe__set_resource_flags(*rsc, pe_rsc_needs_fencing |pe_rsc_needs_unfencing); } - } else if (pcmk__str_eq(value, "fencing", pcmk__str_casei)) { + } else if (pcmk__str_eq(value, PCMK__VALUE_FENCING, pcmk__str_casei)) { pe__set_resource_flags(*rsc, pe_rsc_needs_fencing); if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { pcmk__config_warn("%s requires fencing but fencing is disabled", (*rsc)->id); } } else { const char *orig_value = value; isdefault = TRUE; if (pcmk_is_set((*rsc)->flags, pe_rsc_fence_device)) { - value = "quorum"; + value = PCMK__VALUE_QUORUM; } else if (((*rsc)->variant == pe_native) && pcmk__str_eq(crm_element_value((*rsc)->xml, XML_AGENT_ATTR_CLASS), PCMK_RESOURCE_CLASS_OCF, pcmk__str_casei) && pcmk__str_eq(crm_element_value((*rsc)->xml, XML_AGENT_ATTR_PROVIDER), "pacemaker", pcmk__str_casei) && pcmk__str_eq(crm_element_value((*rsc)->xml, XML_ATTR_TYPE), "remote", pcmk__str_casei) ) { - value = "quorum"; + value = PCMK__VALUE_QUORUM; } else if (pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { - value = "unfencing"; + value = PCMK__VALUE_UNFENCING; } else if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { - value = "fencing"; + value = PCMK__VALUE_FENCING; } else if (data_set->no_quorum_policy == no_quorum_ignore) { - value = "nothing"; + value = PCMK__VALUE_NOTHING; } else { - value = "quorum"; + value = PCMK__VALUE_QUORUM; } if (orig_value != NULL) { pcmk__config_err("Resetting '" XML_RSC_ATTR_REQUIRES "' for %s " "to '%s' because '%s' is not valid", (*rsc)->id, value, orig_value); } goto handle_requires_pref; } pe_rsc_trace((*rsc), "\tRequired to start: %s%s", value, isdefault?" (default)":""); value = g_hash_table_lookup((*rsc)->meta, XML_RSC_ATTR_FAIL_TIMEOUT); if (value != NULL) { // Stored as seconds (*rsc)->failure_timeout = (int) (crm_parse_interval_spec(value) / 1000); } if (remote_node) { GHashTable *params = pe_rsc_params(*rsc, NULL, data_set); /* Grabbing the value now means that any rules based on node attributes * will evaluate to false, so such rules should not be used with * reconnect_interval. * * @TODO Evaluate per node before using */ value = g_hash_table_lookup(params, XML_REMOTE_ATTR_RECONNECT_INTERVAL); if (value) { /* reconnect delay works by setting failure_timeout and preventing the * connection from starting until the failure is cleared. */ (*rsc)->remote_reconnect_ms = crm_parse_interval_spec(value); /* we want to override any default failure_timeout in use when remote * reconnect_interval is in use. */ (*rsc)->failure_timeout = (*rsc)->remote_reconnect_ms / 1000; } } get_target_role(*rsc, &((*rsc)->next_role)); pe_rsc_trace((*rsc), "\tDesired next state: %s", (*rsc)->next_role != RSC_ROLE_UNKNOWN ? role2text((*rsc)->next_role) : "default"); if ((*rsc)->fns->unpack(*rsc, data_set) == FALSE) { return FALSE; } if (pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster)) { // This tag must stay exactly the same because it is tested elsewhere resource_location(*rsc, NULL, 0, "symmetric_default", data_set); } else if (guest_node) { /* remote resources tied to a container resource must always be allowed * to opt-in to the cluster. Whether the connection resource is actually * allowed to be placed on a node is dependent on the container resource */ resource_location(*rsc, NULL, 0, "remote_connection_default", data_set); } pe_rsc_trace((*rsc), "\tAction notification: %s", pcmk_is_set((*rsc)->flags, pe_rsc_notify)? "required" : "not required"); (*rsc)->utilization = pcmk__strkey_table(free, free); pe__unpack_dataset_nvpairs((*rsc)->xml, XML_TAG_UTILIZATION, &rule_data, (*rsc)->utilization, NULL, FALSE, data_set); /* data_set->resources = g_list_append(data_set->resources, (*rsc)); */ if (expanded_xml) { if (add_template_rsc(xml_obj, data_set) == FALSE) { return FALSE; } } return TRUE; } void common_update_score(pe_resource_t * rsc, const char *id, int score) { pe_node_t *node = NULL; node = pe_hash_table_lookup(rsc->allowed_nodes, id); if (node != NULL) { pe_rsc_trace(rsc, "Updating score for %s on %s: %d + %d", rsc->id, id, node->weight, score); node->weight = pcmk__add_scores(node->weight, score); } if (rsc->children) { GList *gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; common_update_score(child_rsc, id, score); } } } gboolean is_parent(pe_resource_t *child, pe_resource_t *rsc) { pe_resource_t *parent = child; if (parent == NULL || rsc == NULL) { return FALSE; } while (parent->parent != NULL) { if (parent->parent == rsc) { return TRUE; } parent = parent->parent; } return FALSE; } pe_resource_t * uber_parent(pe_resource_t * rsc) { pe_resource_t *parent = rsc; if (parent == NULL) { return NULL; } while (parent->parent != NULL && parent->parent->variant != pe_container) { parent = parent->parent; } return parent; } void common_free(pe_resource_t * rsc) { if (rsc == NULL) { return; } pe_rsc_trace(rsc, "Freeing %s %d", rsc->id, rsc->variant); g_list_free(rsc->rsc_cons); g_list_free(rsc->rsc_cons_lhs); g_list_free(rsc->rsc_tickets); g_list_free(rsc->dangling_migrations); if (rsc->parameter_cache != NULL) { g_hash_table_destroy(rsc->parameter_cache); } #if ENABLE_VERSIONED_ATTRS if (rsc->versioned_parameters != NULL) { free_xml(rsc->versioned_parameters); } #endif if (rsc->meta != NULL) { g_hash_table_destroy(rsc->meta); } if (rsc->utilization != NULL) { g_hash_table_destroy(rsc->utilization); } if ((rsc->parent == NULL) && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { free_xml(rsc->xml); rsc->xml = NULL; free_xml(rsc->orig_xml); rsc->orig_xml = NULL; /* if rsc->orig_xml, then rsc->xml is an expanded xml from a template */ } else if (rsc->orig_xml) { free_xml(rsc->xml); rsc->xml = NULL; } if (rsc->running_on) { g_list_free(rsc->running_on); rsc->running_on = NULL; } if (rsc->known_on) { g_hash_table_destroy(rsc->known_on); rsc->known_on = NULL; } if (rsc->actions) { g_list_free(rsc->actions); rsc->actions = NULL; } if (rsc->allowed_nodes) { g_hash_table_destroy(rsc->allowed_nodes); rsc->allowed_nodes = NULL; } g_list_free(rsc->fillers); g_list_free(rsc->rsc_location); pe_rsc_trace(rsc, "Resource freed"); free(rsc->id); free(rsc->clone_name); free(rsc->allocated_to); free(rsc->variant_opaque); free(rsc->pending_task); free(rsc); } /*! * \brief * \internal Find a node (and optionally count all) where resource is active * * \param[in] rsc Resource to check * \param[out] count_all If not NULL, will be set to count of active nodes * \param[out] count_clean If not NULL, will be set to count of clean nodes * * \return An active node (or NULL if resource is not active anywhere) * * \note The order of preference is: an active node that is the resource's * partial migration source; if the resource's "requires" is "quorum" or * "nothing", the first active node in the list that is clean and online; * the first active node in the list. */ pe_node_t * pe__find_active_on(const pe_resource_t *rsc, unsigned int *count_all, unsigned int *count_clean) { pe_node_t *active = NULL; pe_node_t *node = NULL; bool keep_looking = FALSE; bool is_happy = FALSE; if (count_all) { *count_all = 0; } if (count_clean) { *count_clean = 0; } if (rsc == NULL) { return NULL; } for (GList *node_iter = rsc->running_on; node_iter != NULL; node_iter = node_iter->next) { node = node_iter->data; keep_looking = FALSE; is_happy = node->details->online && !node->details->unclean; if (count_all) { ++*count_all; } if (count_clean && is_happy) { ++*count_clean; } if (count_all || count_clean) { // If we're counting, we need to go through entire list keep_looking = TRUE; } if (rsc->partial_migration_source != NULL) { if (node->details == rsc->partial_migration_source->details) { // This is the migration source active = node; } else { keep_looking = TRUE; } } else if (!pcmk_is_set(rsc->flags, pe_rsc_needs_fencing)) { if (is_happy && (!active || !active->details->online || active->details->unclean)) { // This is the first clean node active = node; } else { keep_looking = TRUE; } } if (active == NULL) { // This is first node in list active = node; } if (keep_looking == FALSE) { // Don't waste time iterating if we don't have to break; } } return active; } /*! * \brief * \internal Find and count active nodes according to "requires" * * \param[in] rsc Resource to check * \param[out] count If not NULL, will be set to count of active nodes * * \return An active node (or NULL if resource is not active anywhere) * * \note This is a convenience wrapper for pe__find_active_on() where the count * of all active nodes or only clean active nodes is desired according to * the "requires" meta-attribute. */ pe_node_t * pe__find_active_requires(const pe_resource_t *rsc, unsigned int *count) { if (rsc && !pcmk_is_set(rsc->flags, pe_rsc_needs_fencing)) { return pe__find_active_on(rsc, NULL, count); } return pe__find_active_on(rsc, count, NULL); } void pe__count_common(pe_resource_t *rsc) { if (rsc->children != NULL) { for (GList *item = rsc->children; item != NULL; item = item->next) { ((pe_resource_t *) item->data)->fns->count(item->data); } } else if (!pcmk_is_set(rsc->flags, pe_rsc_orphan) || (rsc->role > RSC_ROLE_STOPPED)) { rsc->cluster->ninstances++; if (pe__resource_is_disabled(rsc)) { rsc->cluster->disabled_resources++; } if (pcmk_is_set(rsc->flags, pe_rsc_block)) { rsc->cluster->blocked_resources++; } } } /*! * \internal * \brief Update a resource's next role * * \param[in,out] rsc Resource to be updated * \param[in] role Resource's new next role * \param[in] why Human-friendly reason why role is changing (for logs) */ void pe__set_next_role(pe_resource_t *rsc, enum rsc_role_e role, const char *why) { CRM_ASSERT((rsc != NULL) && (why != NULL)); if (rsc->next_role != role) { pe_rsc_trace(rsc, "Resetting next role for %s from %s to %s (%s)", rsc->id, role2text(rsc->next_role), role2text(role), why); rsc->next_role = role; } } diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c index 2204493f27..094469e521 100644 --- a/lib/pengine/unpack.c +++ b/lib/pengine/unpack.c @@ -1,4219 +1,4219 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include CRM_TRACE_INIT_DATA(pe_status); /* This uses pcmk__set_flags_as()/pcmk__clear_flags_as() directly rather than * use pe__set_working_set_flags()/pe__clear_working_set_flags() so that the * flag is stringified more readably in log messages. */ #define set_config_flag(data_set, option, flag) do { \ const char *scf_value = pe_pref((data_set)->config_hash, (option)); \ if (scf_value != NULL) { \ if (crm_is_true(scf_value)) { \ (data_set)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Working set", \ crm_system_name, (data_set)->flags, \ (flag), #flag); \ } else { \ (data_set)->flags = pcmk__clear_flags_as(__func__, __LINE__,\ LOG_TRACE, "Working set", \ crm_system_name, (data_set)->flags, \ (flag), #flag); \ } \ } \ } while(0) static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *failed, pe_working_set_t *data_set); static void determine_remote_online_status(pe_working_set_t *data_set, pe_node_t *this_node); static void add_node_attrs(xmlNode *attrs, pe_node_t *node, bool overwrite, pe_working_set_t *data_set); static void determine_online_status(xmlNode *node_state, pe_node_t *this_node, pe_working_set_t *data_set); static void unpack_node_lrm(pe_node_t *node, xmlNode *xml, pe_working_set_t *data_set); // Bitmask for warnings we only want to print once uint32_t pe_wo = 0; static gboolean is_dangling_guest_node(pe_node_t *node) { /* we are looking for a remote-node that was supposed to be mapped to a * container resource, but all traces of that container have disappeared * from both the config and the status section. */ if (pe__is_guest_or_remote_node(node) && node->details->remote_rsc && node->details->remote_rsc->container == NULL && pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_orphan_container_filler)) { return TRUE; } return FALSE; } /*! * \brief Schedule a fence action for a node * * \param[in,out] data_set Current working set of cluster * \param[in,out] node Node to fence * \param[in] reason Text description of why fencing is needed * \param[in] priority_delay Whether to consider `priority-fencing-delay` */ void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay) { CRM_CHECK(node, return); /* A guest node is fenced by marking its container as failed */ if (pe__is_guest_node(node)) { pe_resource_t *rsc = node->details->remote_rsc->container; if (!pcmk_is_set(rsc->flags, pe_rsc_failed)) { if (!pcmk_is_set(rsc->flags, pe_rsc_managed)) { crm_notice("Not fencing guest node %s " "(otherwise would because %s): " "its guest resource %s is unmanaged", node->details->uname, reason, rsc->id); } else { crm_warn("Guest node %s will be fenced " "(by recovering its guest resource %s): %s", node->details->uname, rsc->id, reason); /* We don't mark the node as unclean because that would prevent the * node from running resources. We want to allow it to run resources * in this transition if the recovery succeeds. */ node->details->remote_requires_reset = TRUE; pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); } } } else if (is_dangling_guest_node(node)) { crm_info("Cleaning up dangling connection for guest node %s: " "fencing was already done because %s, " "and guest resource no longer exists", node->details->uname, reason); pe__set_resource_flags(node->details->remote_rsc, pe_rsc_failed|pe_rsc_stop); } else if (pe__is_remote_node(node)) { pe_resource_t *rsc = node->details->remote_rsc; if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_managed)) { crm_notice("Not fencing remote node %s " "(otherwise would because %s): connection is unmanaged", node->details->uname, reason); } else if(node->details->remote_requires_reset == FALSE) { node->details->remote_requires_reset = TRUE; crm_warn("Remote node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); } node->details->unclean = TRUE; // No need to apply `priority-fencing-delay` for remote nodes pe_fence_op(node, NULL, TRUE, reason, FALSE, data_set); } else if (node->details->unclean) { crm_trace("Cluster node %s %s because %s", node->details->uname, pe_can_fence(data_set, node)? "would also be fenced" : "also is unclean", reason); } else { crm_warn("Cluster node %s %s: %s", node->details->uname, pe_can_fence(data_set, node)? "will be fenced" : "is unclean", reason); node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, priority_delay, data_set); } } // @TODO xpaths can't handle templates, rules, or id-refs // nvpair with provides or requires set to unfencing #define XPATH_UNFENCING_NVPAIR XML_CIB_TAG_NVPAIR \ "[(@" XML_NVPAIR_ATTR_NAME "='" PCMK_STONITH_PROVIDES "'" \ "or @" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_REQUIRES "') " \ - "and @" XML_NVPAIR_ATTR_VALUE "='unfencing']" + "and @" XML_NVPAIR_ATTR_VALUE "='" PCMK__VALUE_UNFENCING "']" // unfencing in rsc_defaults or any resource #define XPATH_ENABLE_UNFENCING \ "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \ "/" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR static void set_if_xpath(uint64_t flag, const char *xpath, pe_working_set_t *data_set) { xmlXPathObjectPtr result = NULL; if (!pcmk_is_set(data_set->flags, flag)) { result = xpath_search(data_set->input, xpath); if (result && (numXpathResults(result) > 0)) { pe__set_working_set_flags(data_set, flag); } freeXpathObject(result); } } gboolean unpack_config(xmlNode * config, pe_working_set_t * data_set) { const char *value = NULL; GHashTable *config_hash = pcmk__strkey_table(free, free); pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; data_set->config_hash = config_hash; pe__unpack_dataset_nvpairs(config, XML_CIB_TAG_PROPSET, &rule_data, config_hash, CIB_OPTIONS_FIRST, FALSE, data_set); verify_pe_options(data_set->config_hash); set_config_flag(data_set, "enable-startup-probes", pe_flag_startup_probes); if (!pcmk_is_set(data_set->flags, pe_flag_startup_probes)) { crm_info("Startup probes: disabled (dangerous)"); } value = pe_pref(data_set->config_hash, XML_ATTR_HAVE_WATCHDOG); if (value && crm_is_true(value)) { crm_info("Watchdog-based self-fencing will be performed via SBD if " "fencing is required and stonith-watchdog-timeout is nonzero"); pe__set_working_set_flags(data_set, pe_flag_have_stonith_resource); } /* Set certain flags via xpath here, so they can be used before the relevant * configuration sections are unpacked. */ set_if_xpath(pe_flag_enable_unfencing, XPATH_ENABLE_UNFENCING, data_set); value = pe_pref(data_set->config_hash, "stonith-timeout"); data_set->stonith_timeout = (int) crm_parse_interval_spec(value); crm_debug("STONITH timeout: %d", data_set->stonith_timeout); set_config_flag(data_set, "stonith-enabled", pe_flag_stonith_enabled); crm_debug("STONITH of failed nodes is %s", pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)? "enabled" : "disabled"); data_set->stonith_action = pe_pref(data_set->config_hash, "stonith-action"); if (!strcmp(data_set->stonith_action, "poweroff")) { pe_warn_once(pe_wo_poweroff, "Support for stonith-action of 'poweroff' is deprecated " "and will be removed in a future release (use 'off' instead)"); data_set->stonith_action = "off"; } crm_trace("STONITH will %s nodes", data_set->stonith_action); set_config_flag(data_set, "concurrent-fencing", pe_flag_concurrent_fencing); crm_debug("Concurrent fencing is %s", pcmk_is_set(data_set->flags, pe_flag_concurrent_fencing)? "enabled" : "disabled"); value = pe_pref(data_set->config_hash, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY); if (value) { data_set->priority_fencing_delay = crm_parse_interval_spec(value) / 1000; crm_trace("Priority fencing delay is %ds", data_set->priority_fencing_delay); } set_config_flag(data_set, "stop-all-resources", pe_flag_stop_everything); crm_debug("Stop all active resources: %s", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_stop_everything))); set_config_flag(data_set, "symmetric-cluster", pe_flag_symmetric_cluster); if (pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster)) { crm_debug("Cluster is symmetric" " - resources can run anywhere by default"); } value = pe_pref(data_set->config_hash, "no-quorum-policy"); if (pcmk__str_eq(value, "ignore", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_ignore; } else if (pcmk__str_eq(value, "freeze", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_freeze; } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { data_set->no_quorum_policy = no_quorum_demote; } else if (pcmk__str_eq(value, "suicide", pcmk__str_casei)) { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { int do_panic = 0; crm_element_value_int(data_set->input, XML_ATTR_QUORUM_PANIC, &do_panic); if (do_panic || pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { data_set->no_quorum_policy = no_quorum_suicide; } else { crm_notice("Resetting no-quorum-policy to 'stop': cluster has never had quorum"); data_set->no_quorum_policy = no_quorum_stop; } } else { pcmk__config_err("Resetting no-quorum-policy to 'stop' because " "fencing is disabled"); data_set->no_quorum_policy = no_quorum_stop; } } else { data_set->no_quorum_policy = no_quorum_stop; } switch (data_set->no_quorum_policy) { case no_quorum_freeze: crm_debug("On loss of quorum: Freeze resources"); break; case no_quorum_stop: crm_debug("On loss of quorum: Stop ALL resources"); break; case no_quorum_demote: crm_debug("On loss of quorum: " "Demote promotable resources and stop other resources"); break; case no_quorum_suicide: crm_notice("On loss of quorum: Fence all remaining nodes"); break; case no_quorum_ignore: crm_notice("On loss of quorum: Ignore"); break; } set_config_flag(data_set, "stop-orphan-resources", pe_flag_stop_rsc_orphans); crm_trace("Orphan resources are %s", pcmk_is_set(data_set->flags, pe_flag_stop_rsc_orphans)? "stopped" : "ignored"); set_config_flag(data_set, "stop-orphan-actions", pe_flag_stop_action_orphans); crm_trace("Orphan resource actions are %s", pcmk_is_set(data_set->flags, pe_flag_stop_action_orphans)? "stopped" : "ignored"); value = pe_pref(data_set->config_hash, "remove-after-stop"); if (value != NULL) { if (crm_is_true(value)) { pe__set_working_set_flags(data_set, pe_flag_remove_after_stop); #ifndef PCMK__COMPAT_2_0 pe_warn_once(pe_wo_remove_after, "Support for the remove-after-stop cluster property is" " deprecated and will be removed in a future release"); #endif } else { pe__clear_working_set_flags(data_set, pe_flag_remove_after_stop); } } set_config_flag(data_set, "maintenance-mode", pe_flag_maintenance_mode); crm_trace("Maintenance mode: %s", pcmk__btoa(pcmk_is_set(data_set->flags, pe_flag_maintenance_mode))); set_config_flag(data_set, "start-failure-is-fatal", pe_flag_start_failure_fatal); crm_trace("Start failures are %s", pcmk_is_set(data_set->flags, pe_flag_start_failure_fatal)? "always fatal" : "handled by failcount"); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { set_config_flag(data_set, "startup-fencing", pe_flag_startup_fencing); } if (pcmk_is_set(data_set->flags, pe_flag_startup_fencing)) { crm_trace("Unseen nodes will be fenced"); } else { pe_warn_once(pe_wo_blind, "Blind faith: not fencing unseen nodes"); } pe__unpack_node_health_scores(data_set); data_set->placement_strategy = pe_pref(data_set->config_hash, "placement-strategy"); crm_trace("Placement strategy: %s", data_set->placement_strategy); set_config_flag(data_set, "shutdown-lock", pe_flag_shutdown_lock); crm_trace("Resources will%s be locked to cleanly shut down nodes", (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)? "" : " not")); if (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { value = pe_pref(data_set->config_hash, XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT); data_set->shutdown_lock = crm_parse_interval_spec(value) / 1000; crm_trace("Shutdown locks expire after %us", data_set->shutdown_lock); } return TRUE; } pe_node_t * pe_create_node(const char *id, const char *uname, const char *type, const char *score, pe_working_set_t * data_set) { pe_node_t *new_node = NULL; if (pe_find_node(data_set->nodes, uname) != NULL) { pcmk__config_warn("More than one node entry has name '%s'", uname); } new_node = calloc(1, sizeof(pe_node_t)); if (new_node == NULL) { return NULL; } new_node->weight = char2score(score); new_node->fixed = FALSE; new_node->details = calloc(1, sizeof(struct pe_node_shared_s)); if (new_node->details == NULL) { free(new_node); return NULL; } crm_trace("Creating node for entry %s/%s", uname, id); new_node->details->id = id; new_node->details->uname = uname; new_node->details->online = FALSE; new_node->details->shutdown = FALSE; new_node->details->rsc_discovery_enabled = TRUE; new_node->details->running_rsc = NULL; new_node->details->data_set = data_set; if (pcmk__str_eq(type, "member", pcmk__str_null_matches | pcmk__str_casei)) { new_node->details->type = node_member; } else if (pcmk__str_eq(type, "remote", pcmk__str_casei)) { new_node->details->type = node_remote; pe__set_working_set_flags(data_set, pe_flag_have_remote_nodes); } else { /* @COMPAT 'ping' is the default for backward compatibility, but it * should be changed to 'member' at a compatibility break */ if (!pcmk__str_eq(type, "ping", pcmk__str_casei)) { pcmk__config_warn("Node %s has unrecognized type '%s', " "assuming 'ping'", pcmk__s(uname, "without name"), type); } pe_warn_once(pe_wo_ping_node, "Support for nodes of type 'ping' (such as %s) is " "deprecated and will be removed in a future release", pcmk__s(uname, "unnamed node")); new_node->details->type = node_ping; } new_node->details->attrs = pcmk__strkey_table(free, free); if (pe__is_guest_or_remote_node(new_node)) { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("remote")); } else { g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("cluster")); } new_node->details->utilization = pcmk__strkey_table(free, free); new_node->details->digest_cache = pcmk__strkey_table(free, pe__free_digests); data_set->nodes = g_list_insert_sorted(data_set->nodes, new_node, sort_node_uname); return new_node; } static const char * expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pe_working_set_t *data) { xmlNode *attr_set = NULL; xmlNode *attr = NULL; const char *container_id = ID(xml_obj); const char *remote_name = NULL; const char *remote_server = NULL; const char *remote_port = NULL; const char *connect_timeout = "60s"; const char *remote_allow_migrate=NULL; const char *is_managed = NULL; for (attr_set = pcmk__xe_first_child(xml_obj); attr_set != NULL; attr_set = pcmk__xe_next(attr_set)) { if (!pcmk__str_eq((const char *)attr_set->name, XML_TAG_META_SETS, pcmk__str_casei)) { continue; } for (attr = pcmk__xe_first_child(attr_set); attr != NULL; attr = pcmk__xe_next(attr)) { const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); if (pcmk__str_eq(name, XML_RSC_ATTR_REMOTE_NODE, pcmk__str_casei)) { remote_name = value; } else if (pcmk__str_eq(name, "remote-addr", pcmk__str_casei)) { remote_server = value; } else if (pcmk__str_eq(name, "remote-port", pcmk__str_casei)) { remote_port = value; } else if (pcmk__str_eq(name, "remote-connect-timeout", pcmk__str_casei)) { connect_timeout = value; } else if (pcmk__str_eq(name, "remote-allow-migrate", pcmk__str_casei)) { remote_allow_migrate=value; } else if (pcmk__str_eq(name, XML_RSC_ATTR_MANAGED, pcmk__str_casei)) { is_managed = value; } } } if (remote_name == NULL) { return NULL; } if (pe_find_resource(data->resources, remote_name) != NULL) { return NULL; } pe_create_remote_xml(parent, remote_name, container_id, remote_allow_migrate, is_managed, connect_timeout, remote_server, remote_port); return remote_name; } static void handle_startup_fencing(pe_working_set_t *data_set, pe_node_t *new_node) { if ((new_node->details->type == node_remote) && (new_node->details->remote_rsc == NULL)) { /* Ignore fencing for remote nodes that don't have a connection resource * associated with them. This happens when remote node entries get left * in the nodes section after the connection resource is removed. */ return; } if (pcmk_is_set(data_set->flags, pe_flag_startup_fencing)) { // All nodes are unclean until we've seen their status entry new_node->details->unclean = TRUE; } else { // Blind faith ... new_node->details->unclean = FALSE; } /* We need to be able to determine if a node's status section * exists or not separate from whether the node is unclean. */ new_node->details->unseen = TRUE; } gboolean unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; pe_node_t *new_node = NULL; const char *id = NULL; const char *uname = NULL; const char *type = NULL; const char *score = NULL; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; for (xml_obj = pcmk__xe_first_child(xml_nodes); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_NODE, pcmk__str_none)) { new_node = NULL; id = crm_element_value(xml_obj, XML_ATTR_ID); uname = crm_element_value(xml_obj, XML_ATTR_UNAME); type = crm_element_value(xml_obj, XML_ATTR_TYPE); score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE); crm_trace("Processing node %s/%s", uname, id); if (id == NULL) { pcmk__config_err("Ignoring <" XML_CIB_TAG_NODE "> entry in configuration without id"); continue; } new_node = pe_create_node(id, uname, type, score, data_set); if (new_node == NULL) { return FALSE; } /* if(data_set->have_quorum == FALSE */ /* && data_set->no_quorum_policy == no_quorum_stop) { */ /* /\* start shutting resources down *\/ */ /* new_node->weight = -INFINITY; */ /* } */ handle_startup_fencing(data_set, new_node); add_node_attrs(xml_obj, new_node, FALSE, data_set); pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_UTILIZATION, &rule_data, new_node->details->utilization, NULL, FALSE, data_set); crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME)); } } if (data_set->localhost && pe_find_node(data_set->nodes, data_set->localhost) == NULL) { crm_info("Creating a fake local node"); pe_create_node(data_set->localhost, data_set->localhost, NULL, 0, data_set); } return TRUE; } static void setup_container(pe_resource_t * rsc, pe_working_set_t * data_set) { const char *container_id = NULL; if (rsc->children) { g_list_foreach(rsc->children, (GFunc) setup_container, data_set); return; } container_id = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_CONTAINER); if (container_id && !pcmk__str_eq(container_id, rsc->id, pcmk__str_casei)) { pe_resource_t *container = pe_find_resource(data_set->resources, container_id); if (container) { rsc->container = container; pe__set_resource_flags(container, pe_rsc_is_container); container->fillers = g_list_append(container->fillers, rsc); pe_rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id); } else { pe_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id); } } } gboolean unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; /* Create remote nodes and guest nodes from the resource configuration * before unpacking resources. */ for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { const char *new_node_id = NULL; /* Check for remote nodes, which are defined by ocf:pacemaker:remote * primitives. */ if (xml_contains_remote_node(xml_obj)) { new_node_id = ID(xml_obj); /* The "pe_find_node" check is here to make sure we don't iterate over * an expanded node that has already been added to the node list. */ if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found remote node %s defined by resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes, which are defined by special meta-attributes * of a primitive of any type (for example, VirtualDomain or Xen). */ if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_RESOURCE, pcmk__str_none)) { /* This will add an ocf:pacemaker:remote primitive to the * configuration for the guest node's connection, to be unpacked * later. */ new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s", new_node_id, ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } continue; } /* Check for guest nodes inside a group. Clones are currently not * supported as guest nodes. */ if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_GROUP, pcmk__str_none)) { xmlNode *xml_obj2 = NULL; for (xml_obj2 = pcmk__xe_first_child(xml_obj); xml_obj2 != NULL; xml_obj2 = pcmk__xe_next(xml_obj2)) { new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, data_set); if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) { crm_trace("Found guest node %s in resource %s inside group %s", new_node_id, ID(xml_obj2), ID(xml_obj)); pe_create_node(new_node_id, new_node_id, "remote", NULL, data_set); } } } } return TRUE; } /* Call this after all the nodes and resources have been * unpacked, but before the status section is read. * * A remote node's online status is reflected by the state * of the remote node's connection resource. We need to link * the remote node to this connection resource so we can have * easy access to the connection resource during the scheduler calculations. */ static void link_rsc2remotenode(pe_working_set_t *data_set, pe_resource_t *new_rsc) { pe_node_t *remote_node = NULL; if (new_rsc->is_remote_node == FALSE) { return; } if (pcmk_is_set(data_set->flags, pe_flag_quick_location)) { /* remote_nodes and remote_resources are not linked in quick location calculations */ return; } remote_node = pe_find_node(data_set->nodes, new_rsc->id); CRM_CHECK(remote_node != NULL, return); pe_rsc_trace(new_rsc, "Linking remote connection resource %s to node %s", new_rsc->id, remote_node->details->uname); remote_node->details->remote_rsc = new_rsc; if (new_rsc->container == NULL) { /* Handle start-up fencing for remote nodes (as opposed to guest nodes) * the same as is done for cluster nodes. */ handle_startup_fencing(data_set, remote_node); } else { /* pe_create_node() marks the new node as "remote" or "cluster"; now * that we know the node is a guest node, update it correctly. */ g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND), strdup("container")); } } static void destroy_tag(gpointer data) { pe_tag_t *tag = data; if (tag) { free(tag->id); g_list_free_full(tag->refs, free); free(tag); } } /*! * \internal * \brief Parse configuration XML for resource information * * \param[in] xml_resources Top of resource configuration XML * \param[in,out] data_set Where to put resource information * * \return TRUE * * \note unpack_remote_nodes() MUST be called before this, so that the nodes can * be used when common_unpack() calls resource_location() */ gboolean unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; GList *gIter = NULL; data_set->template_rsc_sets = pcmk__strkey_table(free, destroy_tag); for (xml_obj = pcmk__xe_first_child(xml_resources); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { pe_resource_t *new_rsc = NULL; if (pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, pcmk__str_none)) { const char *template_id = ID(xml_obj); if (template_id && g_hash_table_lookup_extended(data_set->template_rsc_sets, template_id, NULL, NULL) == FALSE) { /* Record the template's ID for the knowledge of its existence anyway. */ g_hash_table_insert(data_set->template_rsc_sets, strdup(template_id), NULL); } continue; } crm_trace("Beginning unpack... <%s id=%s... >", crm_element_name(xml_obj), ID(xml_obj)); if (common_unpack(xml_obj, &new_rsc, NULL, data_set) && (new_rsc != NULL)) { data_set->resources = g_list_append(data_set->resources, new_rsc); pe_rsc_trace(new_rsc, "Added resource %s", new_rsc->id); } else { pcmk__config_err("Ignoring <%s> resource '%s' " "because configuration is invalid", crm_element_name(xml_obj), pcmk__s(ID(xml_obj), "(unnamed)")); if (new_rsc != NULL && new_rsc->fns != NULL) { new_rsc->fns->free(new_rsc); } } } for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; setup_container(rsc, data_set); link_rsc2remotenode(data_set, rsc); } data_set->resources = g_list_sort(data_set->resources, sort_rsc_priority); if (pcmk_is_set(data_set->flags, pe_flag_quick_location)) { /* Ignore */ } else if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && !pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined"); pcmk__config_err("Either configure some or disable STONITH with the stonith-enabled option"); pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity"); } return TRUE; } gboolean unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set) { xmlNode *xml_tag = NULL; data_set->tags = pcmk__strkey_table(free, destroy_tag); for (xml_tag = pcmk__xe_first_child(xml_tags); xml_tag != NULL; xml_tag = pcmk__xe_next(xml_tag)) { xmlNode *xml_obj_ref = NULL; const char *tag_id = ID(xml_tag); if (!pcmk__str_eq((const char *)xml_tag->name, XML_CIB_TAG_TAG, pcmk__str_none)) { continue; } if (tag_id == NULL) { pcmk__config_err("Ignoring <%s> without " XML_ATTR_ID, crm_element_name(xml_tag)); continue; } for (xml_obj_ref = pcmk__xe_first_child(xml_tag); xml_obj_ref != NULL; xml_obj_ref = pcmk__xe_next(xml_obj_ref)) { const char *obj_ref = ID(xml_obj_ref); if (!pcmk__str_eq((const char *)xml_obj_ref->name, XML_CIB_TAG_OBJ_REF, pcmk__str_none)) { continue; } if (obj_ref == NULL) { pcmk__config_err("Ignoring <%s> for tag '%s' without " XML_ATTR_ID, crm_element_name(xml_obj_ref), tag_id); continue; } if (add_tag_ref(data_set->tags, tag_id, obj_ref) == FALSE) { return FALSE; } } } return TRUE; } /* The ticket state section: * "/cib/status/tickets/ticket_state" */ static gboolean unpack_ticket_state(xmlNode * xml_ticket, pe_working_set_t * data_set) { const char *ticket_id = NULL; const char *granted = NULL; const char *last_granted = NULL; const char *standby = NULL; xmlAttrPtr xIter = NULL; pe_ticket_t *ticket = NULL; ticket_id = ID(xml_ticket); if (pcmk__str_empty(ticket_id)) { return FALSE; } crm_trace("Processing ticket state for %s", ticket_id); ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = ticket_new(ticket_id, data_set); if (ticket == NULL) { return FALSE; } } for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_ticket, prop_name); if (pcmk__str_eq(prop_name, XML_ATTR_ID, pcmk__str_none)) { continue; } g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value)); } granted = g_hash_table_lookup(ticket->state, "granted"); if (granted && crm_is_true(granted)) { ticket->granted = TRUE; crm_info("We have ticket '%s'", ticket->id); } else { ticket->granted = FALSE; crm_info("We do not have ticket '%s'", ticket->id); } last_granted = g_hash_table_lookup(ticket->state, "last-granted"); if (last_granted) { long long last_granted_ll; pcmk__scan_ll(last_granted, &last_granted_ll, 0LL); ticket->last_granted = (time_t) last_granted_ll; } standby = g_hash_table_lookup(ticket->state, "standby"); if (standby && crm_is_true(standby)) { ticket->standby = TRUE; if (ticket->granted) { crm_info("Granted ticket '%s' is in standby-mode", ticket->id); } } else { ticket->standby = FALSE; } crm_trace("Done with ticket state for %s", ticket_id); return TRUE; } static gboolean unpack_tickets_state(xmlNode * xml_tickets, pe_working_set_t * data_set) { xmlNode *xml_obj = NULL; for (xml_obj = pcmk__xe_first_child(xml_tickets); xml_obj != NULL; xml_obj = pcmk__xe_next(xml_obj)) { if (!pcmk__str_eq((const char *)xml_obj->name, XML_CIB_TAG_TICKET_STATE, pcmk__str_none)) { continue; } unpack_ticket_state(xml_obj, data_set); } return TRUE; } static void unpack_handle_remote_attrs(pe_node_t *this_node, xmlNode *state, pe_working_set_t * data_set) { const char *resource_discovery_enabled = NULL; xmlNode *attrs = NULL; pe_resource_t *rsc = NULL; if (!pcmk__str_eq((const char *)state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { return; } if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) { return; } crm_trace("Processing remote node id=%s, uname=%s", this_node->details->id, this_node->details->uname); pcmk__scan_min_int(crm_element_value(state, XML_NODE_IS_MAINTENANCE), &(this_node->details->remote_maintenance), 0); rsc = this_node->details->remote_rsc; if (this_node->details->remote_requires_reset == FALSE) { this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; } attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, this_node, TRUE, data_set); if (pe__shutdown_requested(this_node)) { crm_info("Node %s is shutting down", this_node->details->uname); this_node->details->shutdown = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) { crm_info("Node %s is in standby-mode", this_node->details->uname); this_node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) || ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_managed))) { crm_info("Node %s is in maintenance-mode", this_node->details->uname); this_node->details->maintenance = TRUE; } resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY); if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) { if (pe__is_remote_node(this_node) && !pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { crm_warn("Ignoring %s attribute on remote node %s because stonith is disabled", XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname); } else { /* This is either a remote node with fencing enabled, or a guest * node. We don't care whether fencing is enabled when fencing guest * nodes, because they are "fenced" by recovering their containing * resource. */ crm_info("Node %s has resource discovery disabled", this_node->details->uname); this_node->details->rsc_discovery_enabled = FALSE; } } } /*! * \internal * \brief Unpack a cluster node's transient attributes * * \param[in] state CIB node state XML * \param[in] node Cluster node whose attributes are being unpacked * \param[in] data_set Cluster working set */ static void unpack_transient_attributes(xmlNode *state, pe_node_t *node, pe_working_set_t *data_set) { const char *discovery = NULL; xmlNode *attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, node, TRUE, data_set); if (crm_is_true(pe_node_attribute_raw(node, "standby"))) { crm_info("Node %s is in standby-mode", node->details->uname); node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(node, "maintenance"))) { crm_info("Node %s is in maintenance-mode", node->details->uname); node->details->maintenance = TRUE; } discovery = pe_node_attribute_raw(node, XML_NODE_ATTR_RSC_DISCOVERY); if ((discovery != NULL) && !crm_is_true(discovery)) { crm_warn("Ignoring %s attribute for node %s because disabling " "resource discovery is not allowed for cluster nodes", XML_NODE_ATTR_RSC_DISCOVERY, node->details->uname); } } /*! * \internal * \brief Unpack a node state entry (first pass) * * Unpack one node state entry from status. This unpacks information from the * node_state element itself and node attributes inside it, but not the * resource history inside it. Multiple passes through the status are needed to * fully unpack everything. * * \param[in] state CIB node state XML * \param[in] data_set Cluster working set */ static void unpack_node_state(xmlNode *state, pe_working_set_t *data_set) { const char *id = NULL; const char *uname = NULL; pe_node_t *this_node = NULL; id = crm_element_value(state, XML_ATTR_ID); if (id == NULL) { crm_warn("Ignoring malformed " XML_CIB_TAG_STATE " entry without " XML_ATTR_ID); return; } uname = crm_element_value(state, XML_ATTR_UNAME); if (uname == NULL) { crm_warn("Ignoring malformed " XML_CIB_TAG_STATE " entry without " XML_ATTR_UNAME); return; } this_node = pe_find_node_any(data_set->nodes, id, uname); if (this_node == NULL) { pcmk__config_warn("Ignoring recorded node state for '%s' because " "it is no longer in the configuration", uname); return; } if (pe__is_guest_or_remote_node(this_node)) { /* We can't determine the online status of Pacemaker Remote nodes until * after all resource history has been unpacked. In this first pass, we * do need to mark whether the node has been fenced, as this plays a * role during unpacking cluster node resource state. */ pcmk__scan_min_int(crm_element_value(state, XML_NODE_IS_FENCED), &(this_node->details->remote_was_fenced), 0); return; } unpack_transient_attributes(state, this_node, data_set); /* Provisionally mark this cluster node as clean. We have at least seen it * in the current cluster's lifetime. */ this_node->details->unclean = FALSE; this_node->details->unseen = FALSE; crm_trace("Determining online status of cluster node %s (id %s)", this_node->details->uname, id); determine_online_status(state, this_node, data_set); if (!pcmk_is_set(data_set->flags, pe_flag_have_quorum) && this_node->details->online && (data_set->no_quorum_policy == no_quorum_suicide)) { /* Everything else should flow from this automatically * (at least until the scheduler becomes able to migrate off * healthy resources) */ pe_fence_node(data_set, this_node, "cluster does not have quorum", FALSE); } } /*! * \internal * \brief Unpack nodes' resource history as much as possible * * Unpack as many nodes' resource history as possible in one pass through the * status. We need to process Pacemaker Remote nodes' connections/containers * before unpacking their history; the connection/container history will be * in another node's history, so it might take multiple passes to unpack * everything. * * \param[in] status CIB XML status section * \param[in] fence If true, treat any not-yet-unpacked nodes as unseen * \param[in] data_set Cluster working set * * \return Standard Pacemaker return code (specifically pcmk_rc_ok if done, * or EAGAIN if more unpacking remains to be done) */ static int unpack_node_history(xmlNode *status, bool fence, pe_working_set_t *data_set) { int rc = pcmk_rc_ok; // Loop through all node_state entries in CIB status for (xmlNode *state = first_named_child(status, XML_CIB_TAG_STATE); state != NULL; state = crm_next_same_xml(state)) { const char *id = ID(state); const char *uname = crm_element_value(state, XML_ATTR_UNAME); pe_node_t *this_node = NULL; if ((id == NULL) || (uname == NULL)) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history from malformed " XML_CIB_TAG_STATE " without id and/or uname"); continue; } this_node = pe_find_node_any(data_set->nodes, id, uname); if (this_node == NULL) { // Warning already logged in first pass through status section crm_trace("Not unpacking resource history for node %s because " "no longer in configuration", id); continue; } if (this_node->details->unpacked) { crm_trace("Not unpacking resource history for node %s because " "already unpacked", id); continue; } if (fence) { // We're processing all remaining nodes } else if (pe__is_guest_node(this_node)) { /* We can unpack a guest node's history only after we've unpacked * other resource history to the point that we know that the node's * connection and containing resource are both up. */ pe_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (rsc->role != RSC_ROLE_STARTED) || (rsc->container->role != RSC_ROLE_STARTED)) { crm_trace("Not unpacking resource history for guest node %s " "because container and connection are not known to " "be up", id); continue; } } else if (pe__is_remote_node(this_node)) { /* We can unpack a remote node's history only after we've unpacked * other resource history to the point that we know that the node's * connection is up, with the exception of when shutdown locks are * in use. */ pe_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (!pcmk_is_set(data_set->flags, pe_flag_shutdown_lock) && (rsc->role != RSC_ROLE_STARTED))) { crm_trace("Not unpacking resource history for remote node %s " "because connection is not known to be up", id); continue; } /* If fencing and shutdown locks are disabled and we're not processing * unseen nodes, then we don't want to unpack offline nodes until online * nodes have been unpacked. This allows us to number active clone * instances first. */ } else if (!pcmk_any_flags_set(data_set->flags, pe_flag_stonith_enabled |pe_flag_shutdown_lock) && !this_node->details->online) { crm_trace("Not unpacking resource history for offline " "cluster node %s", id); continue; } if (pe__is_guest_or_remote_node(this_node)) { determine_remote_online_status(data_set, this_node); unpack_handle_remote_attrs(this_node, state, data_set); } crm_trace("Unpacking resource history for %snode %s", (fence? "unseen " : ""), id); this_node->details->unpacked = TRUE; unpack_node_lrm(this_node, state, data_set); rc = EAGAIN; // Other node histories might depend on this one } return rc; } /* remove nodes that are down, stopping */ /* create positive rsc_to_node constraints between resources and the nodes they are running on */ /* anything else? */ gboolean unpack_status(xmlNode * status, pe_working_set_t * data_set) { xmlNode *state = NULL; crm_trace("Beginning unpack"); if (data_set->tickets == NULL) { data_set->tickets = pcmk__strkey_table(free, destroy_ticket); } for (state = pcmk__xe_first_child(status); state != NULL; state = pcmk__xe_next(state)) { if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_TICKETS, pcmk__str_none)) { unpack_tickets_state((xmlNode *) state, data_set); } else if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { unpack_node_state(state, data_set); } } while (unpack_node_history(status, FALSE, data_set) == EAGAIN) { crm_trace("Another pass through node resource histories is needed"); } // Now catch any nodes we didn't see unpack_node_history(status, pcmk_is_set(data_set->flags, pe_flag_stonith_enabled), data_set); /* Now that we know where resources are, we can schedule stops of containers * with failed bundle connections */ if (data_set->stop_needed != NULL) { for (GList *item = data_set->stop_needed; item; item = item->next) { pe_resource_t *container = item->data; pe_node_t *node = pe__current_node(container); if (node) { stop_action(container, node, FALSE); } } g_list_free(data_set->stop_needed); data_set->stop_needed = NULL; } /* Now that we know status of all Pacemaker Remote connections and nodes, * we can stop connections for node shutdowns, and check the online status * of remote/guest nodes that didn't have any node history to unpack. */ for (GList *gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *this_node = gIter->data; if (!pe__is_guest_or_remote_node(this_node)) { continue; } if (this_node->details->shutdown && (this_node->details->remote_rsc != NULL)) { pe__set_next_role(this_node->details->remote_rsc, RSC_ROLE_STOPPED, "remote shutdown"); } if (!this_node->details->unpacked) { determine_remote_online_status(data_set, this_node); } } return TRUE; } static gboolean determine_online_status_no_fencing(pe_working_set_t * data_set, xmlNode * node_state, pe_node_t * this_node) { gboolean online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); if (!crm_is_true(in_cluster)) { crm_trace("Node is down: in_cluster=%s", pcmk__s(in_cluster, "")); } else if (pcmk__str_eq(is_peer, ONLINESTATUS, pcmk__str_casei)) { if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { online = TRUE; } else { crm_debug("Node is not ready to run resources: %s", join); } } else if (this_node->details->expected_up == FALSE) { crm_trace("Controller is down: " "in_cluster=%s is_peer=%s join=%s expected=%s", pcmk__s(in_cluster, ""), pcmk__s(is_peer, ""), pcmk__s(join, ""), pcmk__s(exp_state, "")); } else { /* mark it unclean */ pe_fence_node(data_set, this_node, "peer is unexpectedly down", FALSE); crm_info("in_cluster=%s is_peer=%s join=%s expected=%s", pcmk__s(in_cluster, ""), pcmk__s(is_peer, ""), pcmk__s(join, ""), pcmk__s(exp_state, "")); } return online; } static gboolean determine_online_status_fencing(pe_working_set_t * data_set, xmlNode * node_state, pe_node_t * this_node) { gboolean online = FALSE; gboolean do_terminate = FALSE; bool crmd_online = FALSE; const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE); const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER); const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER); const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); const char *terminate = pe_node_attribute_raw(this_node, "terminate"); /* - XML_NODE_IN_CLUSTER ::= true|false - XML_NODE_IS_PEER ::= online|offline - XML_NODE_JOIN_STATE ::= member|down|pending|banned - XML_NODE_EXPECTED ::= member|down */ if (crm_is_true(terminate)) { do_terminate = TRUE; } else if (terminate != NULL && strlen(terminate) > 0) { /* could be a time() value */ char t = terminate[0]; if (t != '0' && isdigit(t)) { do_terminate = TRUE; } } crm_trace("%s: in_cluster=%s is_peer=%s join=%s expected=%s term=%d", this_node->details->uname, pcmk__s(in_cluster, ""), pcmk__s(is_peer, ""), pcmk__s(join, ""), pcmk__s(exp_state, ""), do_terminate); online = crm_is_true(in_cluster); crmd_online = pcmk__str_eq(is_peer, ONLINESTATUS, pcmk__str_casei); if (exp_state == NULL) { exp_state = CRMD_JOINSTATE_DOWN; } if (this_node->details->shutdown) { crm_debug("%s is shutting down", this_node->details->uname); /* Slightly different criteria since we can't shut down a dead peer */ online = crmd_online; } else if (in_cluster == NULL) { pe_fence_node(data_set, this_node, "peer has not been seen by the cluster", FALSE); } else if (pcmk__str_eq(join, CRMD_JOINSTATE_NACK, pcmk__str_casei)) { pe_fence_node(data_set, this_node, "peer failed Pacemaker membership criteria", FALSE); } else if (do_terminate == FALSE && pcmk__str_eq(exp_state, CRMD_JOINSTATE_DOWN, pcmk__str_casei)) { if (crm_is_true(in_cluster) || crmd_online) { crm_info("- Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { crm_trace("%s is down or still coming up", this_node->details->uname); } } else if (do_terminate && pcmk__str_eq(join, CRMD_JOINSTATE_DOWN, pcmk__str_casei) && crm_is_true(in_cluster) == FALSE && !crmd_online) { crm_info("Node %s was just shot", this_node->details->uname); online = FALSE; } else if (crm_is_true(in_cluster) == FALSE) { // Consider `priority-fencing-delay` for lost nodes pe_fence_node(data_set, this_node, "peer is no longer part of the cluster", TRUE); } else if (!crmd_online) { pe_fence_node(data_set, this_node, "peer process is no longer available", FALSE); /* Everything is running at this point, now check join state */ } else if (do_terminate) { pe_fence_node(data_set, this_node, "termination was requested", FALSE); } else if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { crm_info("Node %s is active", this_node->details->uname); } else if (pcmk__strcase_any_of(join, CRMD_JOINSTATE_PENDING, CRMD_JOINSTATE_DOWN, NULL)) { crm_info("Node %s is not ready to run resources", this_node->details->uname); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { pe_fence_node(data_set, this_node, "peer was in an unknown state", FALSE); crm_warn("%s: in-cluster=%s is-peer=%s join=%s expected=%s term=%d shutdown=%d", this_node->details->uname, pcmk__s(in_cluster, ""), pcmk__s(is_peer, ""), pcmk__s(join, ""), pcmk__s(exp_state, ""), do_terminate, this_node->details->shutdown); } return online; } static void determine_remote_online_status(pe_working_set_t * data_set, pe_node_t * this_node) { pe_resource_t *rsc = this_node->details->remote_rsc; pe_resource_t *container = NULL; pe_node_t *host = NULL; /* If there is a node state entry for a (former) Pacemaker Remote node * but no resource creating that node, the node's connection resource will * be NULL. Consider it an offline remote node in that case. */ if (rsc == NULL) { this_node->details->online = FALSE; goto remote_online_done; } container = rsc->container; if (container && pcmk__list_of_1(rsc->running_on)) { host = rsc->running_on->data; } /* If the resource is currently started, mark it online. */ if (rsc->role == RSC_ROLE_STARTED) { crm_trace("%s node %s presumed ONLINE because connection resource is started", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = TRUE; } /* consider this node shutting down if transitioning start->stop */ if (rsc->role == RSC_ROLE_STARTED && rsc->next_role == RSC_ROLE_STOPPED) { crm_trace("%s node %s shutting down because connection resource is stopping", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->shutdown = TRUE; } /* Now check all the failure conditions. */ if(container && pcmk_is_set(container->flags, pe_rsc_failed)) { crm_trace("Guest node %s UNCLEAN because guest resource failed", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } else if (pcmk_is_set(rsc->flags, pe_rsc_failed)) { crm_trace("%s node %s OFFLINE because connection resource failed", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; } else if (rsc->role == RSC_ROLE_STOPPED || (container && container->role == RSC_ROLE_STOPPED)) { crm_trace("%s node %s OFFLINE because its resource is stopped", (container? "Guest" : "Remote"), this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = FALSE; } else if (host && (host->details->online == FALSE) && host->details->unclean) { crm_trace("Guest node %s UNCLEAN because host is unclean", this_node->details->id); this_node->details->online = FALSE; this_node->details->remote_requires_reset = TRUE; } remote_online_done: crm_trace("Remote node %s online=%s", this_node->details->id, this_node->details->online ? "TRUE" : "FALSE"); } static void determine_online_status(xmlNode * node_state, pe_node_t * this_node, pe_working_set_t * data_set) { gboolean online = FALSE; const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED); CRM_CHECK(this_node != NULL, return); this_node->details->shutdown = FALSE; this_node->details->expected_up = FALSE; if (pe__shutdown_requested(this_node)) { this_node->details->shutdown = TRUE; } else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { this_node->details->expected_up = TRUE; } if (this_node->details->type == node_ping) { this_node->details->unclean = FALSE; online = FALSE; /* As far as resource management is concerned, * the node is safely offline. * Anyone caught abusing this logic will be shot */ } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { online = determine_online_status_no_fencing(data_set, node_state, this_node); } else { online = determine_online_status_fencing(data_set, node_state, this_node); } if (online) { this_node->details->online = TRUE; } else { /* remove node from contention */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (online && this_node->details->shutdown) { /* don't run resources here */ this_node->fixed = TRUE; this_node->weight = -INFINITY; } if (this_node->details->type == node_ping) { crm_info("Node %s is not a Pacemaker node", this_node->details->uname); } else if (this_node->details->unclean) { pe_proc_warn("Node %s is unclean", this_node->details->uname); } else if (this_node->details->online) { crm_info("Node %s is %s", this_node->details->uname, this_node->details->shutdown ? "shutting down" : this_node->details->pending ? "pending" : this_node->details->standby ? "standby" : this_node->details->maintenance ? "maintenance" : "online"); } else { crm_trace("Node %s is offline", this_node->details->uname); } } /*! * \internal * \brief Find the end of a resource's name, excluding any clone suffix * * \param[in] id Resource ID to check * * \return Pointer to last character of resource's base name */ const char * pe_base_name_end(const char *id) { if (!pcmk__str_empty(id)) { const char *end = id + strlen(id) - 1; for (const char *s = end; s > id; --s) { switch (*s) { case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': break; case ':': return (s == end)? s : (s - 1); default: return end; } } return end; } return NULL; } /*! * \internal * \brief Get a resource name excluding any clone suffix * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_strip(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); char *basename = NULL; CRM_ASSERT(end); basename = strndup(last_rsc_id, end - last_rsc_id + 1); CRM_ASSERT(basename); return basename; } /*! * \internal * \brief Get the name of the first instance of a cloned resource * * \param[in] last_rsc_id Resource ID to check * * \return Pointer to newly allocated string with resource's base name plus :0 * \note It is the caller's responsibility to free() the result. * This asserts on error, so callers can assume result is not NULL. */ char * clone_zero(const char *last_rsc_id) { const char *end = pe_base_name_end(last_rsc_id); size_t base_name_len = end - last_rsc_id + 1; char *zero = NULL; CRM_ASSERT(end); zero = calloc(base_name_len + 3, sizeof(char)); CRM_ASSERT(zero); memcpy(zero, last_rsc_id, base_name_len); zero[base_name_len] = ':'; zero[base_name_len + 1] = '0'; return zero; } static pe_resource_t * create_fake_resource(const char *rsc_id, xmlNode * rsc_entry, pe_working_set_t * data_set) { pe_resource_t *rsc = NULL; xmlNode *xml_rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE); copy_in_properties(xml_rsc, rsc_entry); crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id); crm_log_xml_debug(xml_rsc, "Orphan resource"); if (!common_unpack(xml_rsc, &rsc, NULL, data_set)) { return NULL; } if (xml_contains_remote_node(xml_rsc)) { pe_node_t *node; crm_debug("Detected orphaned remote node %s", rsc_id); node = pe_find_node(data_set->nodes, rsc_id); if (node == NULL) { node = pe_create_node(rsc_id, rsc_id, "remote", NULL, data_set); } link_rsc2remotenode(data_set, rsc); if (node) { crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id); node->details->shutdown = TRUE; } } if (crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER)) { /* This orphaned rsc needs to be mapped to a container. */ crm_trace("Detected orphaned container filler %s", rsc_id); pe__set_resource_flags(rsc, pe_rsc_orphan_container_filler); } pe__set_resource_flags(rsc, pe_rsc_orphan); data_set->resources = g_list_append(data_set->resources, rsc); return rsc; } /*! * \internal * \brief Create orphan instance for anonymous clone resource history */ static pe_resource_t * create_anonymous_orphan(pe_resource_t *parent, const char *rsc_id, pe_node_t *node, pe_working_set_t *data_set) { pe_resource_t *top = pe__create_clone_child(parent, data_set); // find_rsc() because we might be a cloned group pe_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pe_find_clone); pe_rsc_debug(parent, "Created orphan %s for %s: %s on %s", top->id, parent->id, rsc_id, node->details->uname); return orphan; } /*! * \internal * \brief Check a node for an instance of an anonymous clone * * Return a child instance of the specified anonymous clone, in order of * preference: (1) the instance running on the specified node, if any; * (2) an inactive instance (i.e. within the total of clone-max instances); * (3) a newly created orphan (i.e. clone-max instances are already active). * * \param[in] data_set Cluster information * \param[in] node Node on which to check for instance * \param[in] parent Clone to check * \param[in] rsc_id Name of cloned resource in history (without instance) */ static pe_resource_t * find_anonymous_clone(pe_working_set_t * data_set, pe_node_t * node, pe_resource_t * parent, const char *rsc_id) { GList *rIter = NULL; pe_resource_t *rsc = NULL; pe_resource_t *inactive_instance = NULL; gboolean skip_inactive = FALSE; CRM_ASSERT(parent != NULL); CRM_ASSERT(pe_rsc_is_clone(parent)); CRM_ASSERT(!pcmk_is_set(parent->flags, pe_rsc_unique)); // Check for active (or partially active, for cloned groups) instance pe_rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, node->details->uname, parent->id); for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) { GList *locations = NULL; pe_resource_t *child = rIter->data; /* Check whether this instance is already known to be active or pending * anywhere, at this stage of unpacking. Because this function is called * for a resource before the resource's individual operation history * entries are unpacked, locations will generally not contain the * desired node. * * However, there are three exceptions: * (1) when child is a cloned group and we have already unpacked the * history of another member of the group on the same node; * (2) when we've already unpacked the history of another numbered * instance on the same node (which can happen if globally-unique * was flipped from true to false); and * (3) when we re-run calculations on the same data set as part of a * simulation. */ child->fns->location(child, &locations, 2); if (locations) { /* We should never associate the same numbered anonymous clone * instance with multiple nodes, and clone instances can't migrate, * so there must be only one location, regardless of history. */ CRM_LOG_ASSERT(locations->next == NULL); if (((pe_node_t *)locations->data)->details == node->details) { /* This child instance is active on the requested node, so check * for a corresponding configured resource. We use find_rsc() * instead of child because child may be a cloned group, and we * need the particular member corresponding to rsc_id. * * If the history entry is orphaned, rsc will be NULL. */ rsc = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); if (rsc) { /* If there are multiple instance history entries for an * anonymous clone in a single node's history (which can * happen if globally-unique is switched from true to * false), we want to consider the instances beyond the * first as orphans, even if there are inactive instance * numbers available. */ if (rsc->running_on) { crm_notice("Active (now-)anonymous clone %s has " "multiple (orphan) instance histories on %s", parent->id, node->details->uname); skip_inactive = TRUE; rsc = NULL; } else { pe_rsc_trace(parent, "Resource %s, active", rsc->id); } } } g_list_free(locations); } else { pe_rsc_trace(parent, "Resource %s, skip inactive", child->id); if (!skip_inactive && !inactive_instance && !pcmk_is_set(child->flags, pe_rsc_block)) { // Remember one inactive instance in case we don't find active inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone); /* ... but don't use it if it was already associated with a * pending action on another node */ if (inactive_instance && inactive_instance->pending_node && (inactive_instance->pending_node->details != node->details)) { inactive_instance = NULL; } } } } if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) { pe_rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id); rsc = inactive_instance; } /* If the resource has "requires" set to "quorum" or "nothing", and we don't * have a clone instance for every node, we don't want to consume a valid * instance number for unclean nodes. Such instances may appear to be active * according to the history, but should be considered inactive, so we can * start an instance elsewhere. Treat such instances as orphans. * * An exception is instances running on guest nodes -- since guest node * "fencing" is actually just a resource stop, requires shouldn't apply. * * @TODO Ideally, we'd use an inactive instance number if it is not needed * for any clean instances. However, we don't know that at this point. */ if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_needs_fencing) && (!node->details->online || node->details->unclean) && !pe__is_guest_node(node) && !pe__is_universal_clone(parent, data_set)) { rsc = NULL; } if (rsc == NULL) { rsc = create_anonymous_orphan(parent, rsc_id, node, data_set); pe_rsc_trace(parent, "Resource %s, orphan", rsc->id); } return rsc; } static pe_resource_t * unpack_find_resource(pe_working_set_t * data_set, pe_node_t * node, const char *rsc_id, xmlNode * rsc_entry) { pe_resource_t *rsc = NULL; pe_resource_t *parent = NULL; crm_trace("looking for %s", rsc_id); rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL) { /* If we didn't find the resource by its name in the operation history, * check it again as a clone instance. Even when clone-max=0, we create * a single :0 orphan to match against here. */ char *clone0_id = clone_zero(rsc_id); pe_resource_t *clone0 = pe_find_resource(data_set->resources, clone0_id); if (clone0 && !pcmk_is_set(clone0->flags, pe_rsc_unique)) { rsc = clone0; parent = uber_parent(clone0); crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id); } else { crm_trace("%s is not known as %s either (orphan)", rsc_id, clone0_id); } free(clone0_id); } else if (rsc->variant > pe_native) { crm_trace("Resource history for %s is orphaned because it is no longer primitive", rsc_id); return NULL; } else { parent = uber_parent(rsc); } if (pe_rsc_is_anon_clone(parent)) { if (pe_rsc_is_bundled(parent)) { rsc = pe__find_bundle_replica(parent->parent, node); } else { char *base = clone_strip(rsc_id); rsc = find_anonymous_clone(data_set, node, parent, base); free(base); CRM_ASSERT(rsc != NULL); } } if (rsc && !pcmk__str_eq(rsc_id, rsc->id, pcmk__str_casei) && !pcmk__str_eq(rsc_id, rsc->clone_name, pcmk__str_casei)) { pcmk__str_update(&rsc->clone_name, rsc_id); pe_rsc_debug(rsc, "Internally renamed %s on %s to %s%s", rsc_id, node->details->uname, rsc->id, (pcmk_is_set(rsc->flags, pe_rsc_orphan)? " (ORPHAN)" : "")); } return rsc; } static pe_resource_t * process_orphan_resource(xmlNode * rsc_entry, pe_node_t * node, pe_working_set_t * data_set) { pe_resource_t *rsc = NULL; const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); crm_debug("Detected orphan resource %s on %s", rsc_id, node->details->uname); rsc = create_fake_resource(rsc_id, rsc_entry, data_set); if (rsc == NULL) { return NULL; } if (!pcmk_is_set(data_set->flags, pe_flag_stop_rsc_orphans)) { pe__clear_resource_flags(rsc, pe_rsc_managed); } else { CRM_CHECK(rsc != NULL, return NULL); pe_rsc_trace(rsc, "Added orphan %s", rsc->id); resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__", data_set); } return rsc; } static void process_rsc_state(pe_resource_t * rsc, pe_node_t * node, enum action_fail_response on_fail, xmlNode * migrate_op, pe_working_set_t * data_set) { pe_node_t *tmpnode = NULL; char *reason = NULL; enum action_fail_response save_on_fail = action_fail_ignore; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s", rsc->id, role2text(rsc->role), node->details->uname, fail2text(on_fail)); /* process current state */ if (rsc->role != RSC_ROLE_UNKNOWN) { pe_resource_t *iter = rsc; while (iter) { if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) { pe_node_t *n = pe__copy_node(node); pe_rsc_trace(rsc, "%s%s%s known on %s", rsc->id, ((rsc->clone_name == NULL)? "" : " also known as "), ((rsc->clone_name == NULL)? "" : rsc->clone_name), n->details->uname); g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n); } if (pcmk_is_set(iter->flags, pe_rsc_unique)) { break; } iter = iter->parent; } } /* If a managed resource is believed to be running, but node is down ... */ if (rsc->role > RSC_ROLE_STOPPED && node->details->online == FALSE && node->details->maintenance == FALSE && pcmk_is_set(rsc->flags, pe_rsc_managed)) { gboolean should_fence = FALSE; /* If this is a guest node, fence it (regardless of whether fencing is * enabled, because guest node fencing is done by recovery of the * container resource rather than by the fencer). Mark the resource * we're processing as failed. When the guest comes back up, its * operation history in the CIB will be cleared, freeing the affected * resource to run again once we are sure we know its state. */ if (pe__is_guest_node(node)) { pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); should_fence = TRUE; } else if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { if (pe__is_remote_node(node) && node->details->remote_rsc && !pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_failed)) { /* Setting unseen means that fencing of the remote node will * occur only if the connection resource is not going to start * somewhere. This allows connection resources on a failed * cluster node to move to another node without requiring the * remote nodes to be fenced as well. */ node->details->unseen = TRUE; reason = crm_strdup_printf("%s is active there (fencing will be" " revoked if remote connection can " "be re-established elsewhere)", rsc->id); } should_fence = TRUE; } if (should_fence) { if (reason == NULL) { reason = crm_strdup_printf("%s is thought to be active there", rsc->id); } pe_fence_node(data_set, node, reason, FALSE); } free(reason); } /* In order to calculate priority_fencing_delay correctly, save the failure information and pass it to native_add_running(). */ save_on_fail = on_fail; if (node->details->unclean) { /* No extra processing needed * Also allows resources to be started again after a node is shot */ on_fail = action_fail_ignore; } switch (on_fail) { case action_fail_ignore: /* nothing to do */ break; case action_fail_demote: pe__set_resource_flags(rsc, pe_rsc_failed); demote_action(rsc, node, FALSE); break; case action_fail_fence: /* treat it as if it is still running * but also mark the node as unclean */ reason = crm_strdup_printf("%s failed there", rsc->id); pe_fence_node(data_set, node, reason, FALSE); free(reason); break; case action_fail_standby: node->details->standby = TRUE; node->details->standby_onfail = TRUE; break; case action_fail_block: /* is_managed == FALSE will prevent any * actions being sent for the resource */ pe__clear_resource_flags(rsc, pe_rsc_managed); pe__set_resource_flags(rsc, pe_rsc_block); break; case action_fail_migrate: /* make sure it comes up somewhere else * or not at all */ resource_location(rsc, node, -INFINITY, "__action_migration_auto__", data_set); break; case action_fail_stop: pe__set_next_role(rsc, RSC_ROLE_STOPPED, "on-fail=stop"); break; case action_fail_recover: if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); stop_action(rsc, node, FALSE); } break; case action_fail_restart_container: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); if (rsc->container && pe_rsc_is_bundled(rsc)) { /* A bundle's remote connection can run on a different node than * the bundle's container. We don't necessarily know where the * container is running yet, so remember it and add a stop * action for it later. */ data_set->stop_needed = g_list_prepend(data_set->stop_needed, rsc->container); } else if (rsc->container) { stop_action(rsc->container, node, FALSE); } else if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { stop_action(rsc, node, FALSE); } break; case action_fail_reset_remote: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { tmpnode = NULL; if (rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); } if (tmpnode && pe__is_remote_node(tmpnode) && tmpnode->details->remote_was_fenced == 0) { /* The remote connection resource failed in a way that * should result in fencing the remote node. */ pe_fence_node(data_set, tmpnode, "remote connection is unrecoverable", FALSE); } } /* require the stop action regardless if fencing is occurring or not. */ if (rsc->role > RSC_ROLE_STOPPED) { stop_action(rsc, node, FALSE); } /* if reconnect delay is in use, prevent the connection from exiting the * "STOPPED" role until the failure is cleared by the delay timeout. */ if (rsc->remote_reconnect_ms) { pe__set_next_role(rsc, RSC_ROLE_STOPPED, "remote reset"); } break; } /* ensure a remote-node connection failure forces an unclean remote-node * to be fenced. By setting unseen = FALSE, the remote-node failure will * result in a fencing operation regardless if we're going to attempt to * reconnect to the remote-node in this transition or not. */ if (pcmk_is_set(rsc->flags, pe_rsc_failed) && rsc->is_remote_node) { tmpnode = pe_find_node(data_set->nodes, rsc->id); if (tmpnode && tmpnode->details->unclean) { tmpnode->details->unseen = FALSE; } } if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) { if (pcmk_is_set(rsc->flags, pe_rsc_orphan)) { if (pcmk_is_set(rsc->flags, pe_rsc_managed)) { pcmk__config_warn("Detected active orphan %s running on %s", rsc->id, node->details->uname); } else { pcmk__config_warn("Resource '%s' must be stopped manually on " "%s because cluster is configured not to " "stop active orphans", rsc->id, node->details->uname); } } native_add_running(rsc, node, data_set, (save_on_fail != action_fail_ignore)); switch (on_fail) { case action_fail_ignore: break; case action_fail_demote: case action_fail_block: pe__set_resource_flags(rsc, pe_rsc_failed); break; default: pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); break; } } else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) { /* Only do this for older status sections that included instance numbers * Otherwise stopped instances will appear as orphans */ pe_rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)", rsc->clone_name, rsc->id); free(rsc->clone_name); rsc->clone_name = NULL; } else { GList *possible_matches = pe__resource_actions(rsc, node, RSC_STOP, FALSE); GList *gIter = possible_matches; for (; gIter != NULL; gIter = gIter->next) { pe_action_t *stop = (pe_action_t *) gIter->data; pe__set_action_flags(stop, pe_action_optional); } g_list_free(possible_matches); } } /* create active recurring operations as optional */ static void process_recurring(pe_node_t * node, pe_resource_t * rsc, int start_index, int stop_index, GList *sorted_op_list, pe_working_set_t * data_set) { int counter = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; CRM_ASSERT(rsc); pe_rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index); for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; guint interval_ms = 0; char *key = NULL; const char *id = ID(rsc_op); counter++; if (node->details->online == FALSE) { pe_rsc_trace(rsc, "Skipping %s/%s: node is offline", rsc->id, node->details->uname); break; /* Need to check if there's a monitor for role="Stopped" */ } else if (start_index < stop_index && counter <= stop_index) { pe_rsc_trace(rsc, "Skipping %s/%s: resource is not active", id, node->details->uname); continue; } else if (counter < start_index) { pe_rsc_trace(rsc, "Skipping %s/%s: old %d", id, node->details->uname, counter); continue; } crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms == 0) { pe_rsc_trace(rsc, "Skipping %s/%s: non-recurring", id, node->details->uname); continue; } status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (pcmk__str_eq(status, "-1", pcmk__str_casei)) { pe_rsc_trace(rsc, "Skipping %s/%s: status", id, node->details->uname); continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); /* create the action */ key = pcmk__op_key(rsc->id, task, interval_ms); pe_rsc_trace(rsc, "Creating %s/%s", key, node->details->uname); custom_action(rsc, key, task, node, TRUE, TRUE, data_set); } } void calculate_active_ops(GList *sorted_op_list, int *start_index, int *stop_index) { int counter = -1; int implied_monitor_start = -1; int implied_clone_start = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; *stop_index = -1; *start_index = -1; for (; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (pcmk__str_eq(task, CRMD_ACTION_STOP, pcmk__str_casei) && pcmk__str_eq(status, "0", pcmk__str_casei)) { *stop_index = counter; } else if (pcmk__strcase_any_of(task, CRMD_ACTION_START, CRMD_ACTION_MIGRATED, NULL)) { *start_index = counter; } else if ((implied_monitor_start <= *stop_index) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { const char *rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC); if (pcmk__strcase_any_of(rc, "0", "8", NULL)) { implied_monitor_start = counter; } } else if (pcmk__strcase_any_of(task, CRMD_ACTION_PROMOTE, CRMD_ACTION_DEMOTE, NULL)) { implied_clone_start = counter; } } if (*start_index == -1) { if (implied_clone_start != -1) { *start_index = implied_clone_start; } else if (implied_monitor_start != -1) { *start_index = implied_monitor_start; } } } // If resource history entry has shutdown lock, remember lock node and time static void unpack_shutdown_lock(xmlNode *rsc_entry, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { time_t lock_time = 0; // When lock started (i.e. node shutdown time) if ((crm_element_value_epoch(rsc_entry, XML_CONFIG_ATTR_SHUTDOWN_LOCK, &lock_time) == pcmk_ok) && (lock_time != 0)) { if ((data_set->shutdown_lock > 0) && (get_effective_time(data_set) > (lock_time + data_set->shutdown_lock))) { pe_rsc_info(rsc, "Shutdown lock for %s on %s expired", rsc->id, node->details->uname); pe__clear_resource_history(rsc, node, data_set); } else { rsc->lock_node = node; rsc->lock_time = lock_time; } } } /*! * \internal * \brief Unpack one lrm_resource entry from a node's CIB status * * \param[in] node Node whose status is being unpacked * \param[in] rsc_entry lrm_resource XML being unpacked * \param[in] data_set Cluster working set * * \return Resource corresponding to the entry, or NULL if no operation history */ static pe_resource_t * unpack_lrm_resource(pe_node_t *node, xmlNode *lrm_resource, pe_working_set_t *data_set) { GList *gIter = NULL; int stop_index = -1; int start_index = -1; enum rsc_role_e req_role = RSC_ROLE_UNKNOWN; const char *task = NULL; const char *rsc_id = ID(lrm_resource); pe_resource_t *rsc = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; xmlNode *migrate_op = NULL; xmlNode *rsc_op = NULL; xmlNode *last_failure = NULL; enum action_fail_response on_fail = action_fail_ignore; enum rsc_role_e saved_role = RSC_ROLE_UNKNOWN; if (rsc_id == NULL) { crm_warn("Ignoring malformed " XML_LRM_TAG_RESOURCE " entry without id"); return NULL; } crm_trace("Unpacking " XML_LRM_TAG_RESOURCE " for %s on %s", rsc_id, node->details->uname); // Build a list of individual lrm_rsc_op entries, so we can sort them for (rsc_op = first_named_child(lrm_resource, XML_LRM_TAG_RSC_OP); rsc_op != NULL; rsc_op = crm_next_same_xml(rsc_op)) { op_list = g_list_prepend(op_list, rsc_op); } if (!pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } } /* find the resource */ rsc = unpack_find_resource(data_set, node, rsc_id, lrm_resource); if (rsc == NULL) { if (op_list == NULL) { // If there are no operations, there is nothing to do return NULL; } else { rsc = process_orphan_resource(lrm_resource, node, data_set); } } CRM_ASSERT(rsc != NULL); // Check whether the resource is "shutdown-locked" to this node if (pcmk_is_set(data_set->flags, pe_flag_shutdown_lock)) { unpack_shutdown_lock(lrm_resource, rsc, node, data_set); } /* process operations */ saved_role = rsc->role; rsc->role = RSC_ROLE_UNKNOWN; sorted_op_list = g_list_sort(op_list, sort_op_by_callid); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); if (pcmk__str_eq(task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { migrate_op = rsc_op; } unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail, data_set); } /* create active recurring operations as optional */ calculate_active_ops(sorted_op_list, &start_index, &stop_index); process_recurring(node, rsc, start_index, stop_index, sorted_op_list, data_set); /* no need to free the contents */ g_list_free(sorted_op_list); process_rsc_state(rsc, node, on_fail, migrate_op, data_set); if (get_target_role(rsc, &req_role)) { if (rsc->next_role == RSC_ROLE_UNKNOWN || req_role < rsc->next_role) { pe__set_next_role(rsc, req_role, XML_RSC_ATTR_TARGET_ROLE); } else if (req_role > rsc->next_role) { pe_rsc_info(rsc, "%s: Not overwriting calculated next role %s" " with requested next role %s", rsc->id, role2text(rsc->next_role), role2text(req_role)); } } if (saved_role > rsc->role) { rsc->role = saved_role; } return rsc; } static void handle_orphaned_container_fillers(xmlNode * lrm_rsc_list, pe_working_set_t * data_set) { xmlNode *rsc_entry = NULL; for (rsc_entry = pcmk__xe_first_child(lrm_rsc_list); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry)) { pe_resource_t *rsc; pe_resource_t *container; const char *rsc_id; const char *container_id; if (!pcmk__str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, pcmk__str_casei)) { continue; } container_id = crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER); rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID); if (container_id == NULL || rsc_id == NULL) { continue; } container = pe_find_resource(data_set->resources, container_id); if (container == NULL) { continue; } rsc = pe_find_resource(data_set->resources, rsc_id); if (rsc == NULL || !pcmk_is_set(rsc->flags, pe_rsc_orphan_container_filler) || rsc->container != NULL) { continue; } pe_rsc_trace(rsc, "Mapped container of orphaned resource %s to %s", rsc->id, container_id); rsc->container = container; container->fillers = g_list_append(container->fillers, rsc); } } /*! * \internal * \brief Unpack one node's lrm status section * * \param[in] node Node whose status is being unpacked * \param[in] xml CIB node state XML * \param[in] data_set Cluster working set */ static void unpack_node_lrm(pe_node_t *node, xmlNode *xml, pe_working_set_t *data_set) { bool found_orphaned_container_filler = false; // Drill down to lrm_resources section xml = find_xml_node(xml, XML_CIB_TAG_LRM, FALSE); if (xml == NULL) { return; } xml = find_xml_node(xml, XML_LRM_TAG_RESOURCES, FALSE); if (xml == NULL) { return; } // Unpack each lrm_resource entry for (xmlNode *rsc_entry = first_named_child(xml, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { pe_resource_t *rsc = unpack_lrm_resource(node, rsc_entry, data_set); if ((rsc != NULL) && pcmk_is_set(rsc->flags, pe_rsc_orphan_container_filler)) { found_orphaned_container_filler = true; } } /* Now that all resource state has been unpacked for this node, map any * orphaned container fillers to their container resource. */ if (found_orphaned_container_filler) { handle_orphaned_container_fillers(xml, data_set); } } static void set_active(pe_resource_t * rsc) { pe_resource_t *top = uber_parent(rsc); if (top && pcmk_is_set(top->flags, pe_rsc_promotable)) { rsc->role = RSC_ROLE_UNPROMOTED; } else { rsc->role = RSC_ROLE_STARTED; } } static void set_node_score(gpointer key, gpointer value, gpointer user_data) { pe_node_t *node = value; int *score = user_data; node->weight = *score; } #define STATUS_PATH_MAX 1024 static xmlNode * find_lrm_op(const char *resource, const char *op, const char *node, const char *source, int target_rc, pe_working_set_t *data_set) { int offset = 0; char xpath[STATUS_PATH_MAX]; xmlNode *xml = NULL; offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node); offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']", resource); /* Need to check against transition_magic too? */ if (source && pcmk__str_eq(op, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_target='%s']", op, source); } else if (source && pcmk__str_eq(op, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_source='%s']", op, source); } else { offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "/" XML_LRM_TAG_RSC_OP "[@operation='%s']", op); } CRM_LOG_ASSERT(offset > 0); xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG); if (xml && target_rc >= 0) { int rc = PCMK_OCF_UNKNOWN_ERROR; int status = PCMK_EXEC_ERROR; crm_element_value_int(xml, XML_LRM_ATTR_RC, &rc); crm_element_value_int(xml, XML_LRM_ATTR_OPSTATUS, &status); if ((rc != target_rc) || (status != PCMK_EXEC_DONE)) { return NULL; } } return xml; } static xmlNode * find_lrm_resource(const char *rsc_id, const char *node_name, pe_working_set_t *data_set) { int offset = 0; char xpath[STATUS_PATH_MAX]; xmlNode *xml = NULL; offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node_name); offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']", rsc_id); CRM_LOG_ASSERT(offset > 0); xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG); return xml; } static bool unknown_on_node(const char *rsc_id, const char *node_name, pe_working_set_t *data_set) { xmlNode *lrm_resource = NULL; lrm_resource = find_lrm_resource(rsc_id, node_name, data_set); /* If the resource has no lrm_rsc_op history on the node, that means its * state is unknown there. */ return (lrm_resource == NULL || first_named_child(lrm_resource, XML_LRM_TAG_RSC_OP) == NULL); } /*! * \brief Check whether a probe/monitor indicating the resource was not running * on a node happened after some event * * \param[in] rsc_id Resource being checked * \param[in] node_name Node being checked * \param[in] xml_op Event that monitor is being compared to * \param[in] data_set Cluster working set * * \return true if such a monitor happened after event, false otherwise */ static bool monitor_not_running_after(const char *rsc_id, const char *node_name, xmlNode *xml_op, pe_working_set_t *data_set) { /* Any probe/monitor operation on the node indicating it was not running * there */ xmlNode *monitor = find_lrm_op(rsc_id, CRMD_ACTION_STATUS, node_name, NULL, PCMK_OCF_NOT_RUNNING, data_set); return (monitor && pe__is_newer_op(monitor, xml_op) > 0); } static int pe__call_id(xmlNode *op_xml) { int id = 0; if (op_xml) { crm_element_value_int(op_xml, XML_LRM_ATTR_CALLID, &id); } return id; } /*! * \brief Check whether a stop happened on the same node after some event * * \param[in] rsc Resource being checked * \param[in] node Node being checked * \param[in] xml_op Event that stop is being compared to * \param[in] data_set Cluster working set * * \return TRUE if stop happened after event, FALSE otherwise * * \note This is really unnecessary, but kept as a safety mechanism. We * currently don't save more than one successful event in history, so this * only matters when processing really old CIB files that we don't * technically support anymore, or as preparation for logging an extended * history in the future. */ static bool stop_happened_after(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP, node->details->uname, NULL, PCMK_OCF_OK, data_set); return (stop_op && (pe__call_id(stop_op) > pe__call_id(xml_op))); } static void unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { /* A successful migration sequence is: * migrate_to on source node * migrate_from on target node * stop on source node * * If a migrate_to is followed by a stop, the entire migration (successful * or failed) is complete, and we don't care what happened on the target. * * If no migrate_from has happened, the migration is considered to be * "partial". If the migrate_from failed, make sure the resource gets * stopped on both source and target (if up). * * If the migrate_to and migrate_from both succeeded (which also implies the * resource is no longer running on the source), but there is no stop, the * migration is considered to be "dangling". Schedule a stop on the source * in this case. */ int from_rc = 0; int from_status = 0; pe_node_t *target_node = NULL; pe_node_t *source_node = NULL; xmlNode *migrate_from = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(source, node->details->uname), return); if (stop_happened_after(rsc, node, xml_op, data_set)) { return; } // Clones are not allowed to migrate, so role can't be promoted rsc->role = RSC_ROLE_STARTED; target_node = pe_find_node(data_set->nodes, target); source_node = pe_find_node(data_set->nodes, source); // Check whether there was a migrate_from action on the target migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target, source, -1, data_set); if (migrate_from) { crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc); crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status); pe_rsc_trace(rsc, "%s op on %s exited with status=%d, rc=%d", ID(migrate_from), target, from_status, from_rc); } if (migrate_from && from_rc == PCMK_OCF_OK && (from_status == PCMK_EXEC_DONE)) { /* The migrate_to and migrate_from both succeeded, so mark the migration * as "dangling". This will be used to schedule a stop action on the * source without affecting the target. */ pe_rsc_trace(rsc, "Detected dangling migration op: %s on %s", ID(xml_op), source); rsc->role = RSC_ROLE_STOPPED; rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } else if (migrate_from && (from_status != PCMK_EXEC_PENDING)) { // Failed if (target_node && target_node->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node, target_node->details->online); native_add_running(rsc, target_node, data_set, TRUE); } } else { // Pending, or complete but erased if (target_node && target_node->details->online) { pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node, target_node->details->online); native_add_running(rsc, target_node, data_set, FALSE); if (source_node && source_node->details->online) { /* This is a partial migration: the migrate_to completed * successfully on the source, but the migrate_from has not * completed. Remember the source and target; if the newly * chosen target remains the same when we schedule actions * later, we may continue with the migration. */ rsc->partial_migration_target = target_node; rsc->partial_migration_source = source_node; } } else { /* Consider it failed here - forces a restart, prevents migration */ pe__set_resource_flags(rsc, pe_rsc_failed|pe_rsc_stop); pe__clear_resource_flags(rsc, pe_rsc_allow_migrate); } } } // Is there an action_name in node_name's rsc history newer than call_id? static bool newer_op(pe_resource_t *rsc, const char *action_name, const char *node_name, int call_id, pe_working_set_t *data_set) { xmlNode *action = find_lrm_op(rsc->id, action_name, node_name, NULL, PCMK_OCF_OK, data_set); return pe__call_id(action) > call_id; } static void unpack_migrate_to_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { int target_stop_id = 0; int target_migrate_from_id = 0; xmlNode *target_stop = NULL; xmlNode *target_migrate_from = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(source, node->details->uname), return); /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ rsc->role = RSC_ROLE_STARTED; // Check for stop on the target target_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, target, NULL, PCMK_OCF_OK, data_set); target_stop_id = pe__call_id(target_stop); // Check for migrate_from on the target target_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target, source, PCMK_OCF_OK, data_set); target_migrate_from_id = pe__call_id(target_migrate_from); if (/* If the resource state is unknown on the target, it will likely be * probed there. * Don't just consider it running there. We will get back here anyway in * case the probe detects it's running there. */ !unknown_on_node(rsc->id, target, data_set) /* If there's any probe/monitor operation on the target newer than this * failed migrate_to indicating it was not running there, this migrate_to * failure no longer matters for the target. */ && !monitor_not_running_after(rsc->id, target, xml_op, data_set) && ((target_stop == NULL) || (target_stop_id < target_migrate_from_id))) { /* There was no stop on the target, or a stop that happened before a * migrate_from, so assume the resource is still active on the target * (if it is up). */ pe_node_t *target_node = pe_find_node(data_set->nodes, target); pe_rsc_trace(rsc, "stop (%d) + migrate_from (%d)", target_stop_id, target_migrate_from_id); if (target_node && target_node->details->online) { native_add_running(rsc, target_node, data_set, FALSE); } } else if (target_migrate_from == NULL) { /* We know there was a stop on the target, but there may not have been a * migrate_from (the stop could have happened before migrate_from was * scheduled or attempted). * * That means this could be a "dangling" migration. But first, check * whether there is a newer successful stop, start, or migrate_from on * the source node -- it's possible the failed migration was followed by * a successful stop, full restart, or migration in the reverse * direction, in which case we don't want to force a stop. */ int source_migrate_to_id = pe__call_id(xml_op); if (newer_op(rsc, CRMD_ACTION_MIGRATED, source, source_migrate_to_id, data_set) || newer_op(rsc, CRMD_ACTION_START, source, source_migrate_to_id, data_set) || newer_op(rsc, CRMD_ACTION_STOP, source, source_migrate_to_id, data_set)) { return; } // Mark node as having dangling migration so we can force a stop later rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node); } } static void unpack_migrate_from_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, pe_working_set_t *data_set) { xmlNode *source_stop = NULL; xmlNode *source_migrate_to = NULL; const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); // Sanity check CRM_CHECK(source && target && !strcmp(target, node->details->uname), return); /* If a migration failed, we have to assume the resource is active. Clones * are not allowed to migrate, so role can't be promoted. */ rsc->role = RSC_ROLE_STARTED; // Check for a stop on the source source_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, source, NULL, PCMK_OCF_OK, data_set); // Check for a migrate_to on the source source_migrate_to = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATE, source, target, PCMK_OCF_OK, data_set); if ((source_stop == NULL) || (pe__call_id(source_stop) < pe__call_id(source_migrate_to))) { /* There was no stop on the source, or a stop that happened before * migrate_to, so assume the resource is still active on the source (if * it is up). */ pe_node_t *source_node = pe_find_node(data_set->nodes, source); if (source_node && source_node->details->online) { native_add_running(rsc, source_node, data_set, TRUE); } } } static void record_failed_op(xmlNode *op, const pe_node_t *node, const pe_resource_t *rsc, pe_working_set_t *data_set) { xmlNode *xIter = NULL; const char *op_key = crm_element_value(op, XML_LRM_ATTR_TASK_KEY); if (node->details->online == FALSE) { return; } for (xIter = data_set->failed->children; xIter; xIter = xIter->next) { const char *key = crm_element_value(xIter, XML_LRM_ATTR_TASK_KEY); const char *uname = crm_element_value(xIter, XML_ATTR_UNAME); if(pcmk__str_eq(op_key, key, pcmk__str_casei) && pcmk__str_eq(uname, node->details->uname, pcmk__str_casei)) { crm_trace("Skipping duplicate entry %s on %s", op_key, node->details->uname); return; } } crm_trace("Adding entry %s on %s", op_key, node->details->uname); crm_xml_add(op, XML_ATTR_UNAME, node->details->uname); crm_xml_add(op, XML_LRM_ATTR_RSCID, rsc->id); add_node_copy(data_set->failed, op); } static const char *get_op_key(xmlNode *xml_op) { const char *key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if(key == NULL) { key = ID(xml_op); } return key; } static const char * last_change_str(xmlNode *xml_op) { time_t when; const char *when_s = NULL; if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &when) == pcmk_ok) { when_s = pcmk__epoch2str(&when); if (when_s) { // Skip day of week to make message shorter when_s = strchr(when_s, ' '); if (when_s) { ++when_s; } } } return ((when_s && *when_s)? when_s : "unknown time"); } /*! * \internal * \brief Compare two on-fail values * * \param[in] first One on-fail value to compare * \param[in] second The other on-fail value to compare * * \return A negative number if second is more severe than first, zero if they * are equal, or a positive number if first is more severe than second. * \note This is only needed until the action_fail_response values can be * renumbered at the next API compatibility break. */ static int cmp_on_fail(enum action_fail_response first, enum action_fail_response second) { switch (first) { case action_fail_demote: switch (second) { case action_fail_ignore: return 1; case action_fail_demote: return 0; default: return -1; } break; case action_fail_reset_remote: switch (second) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: return 1; case action_fail_reset_remote: return 0; default: return -1; } break; case action_fail_restart_container: switch (second) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_reset_remote: return 1; case action_fail_restart_container: return 0; default: return -1; } break; default: break; } switch (second) { case action_fail_demote: return (first == action_fail_ignore)? -1 : 1; case action_fail_reset_remote: switch (first) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: return -1; default: return 1; } break; case action_fail_restart_container: switch (first) { case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_reset_remote: return -1; default: return 1; } break; default: break; } return first - second; } static void unpack_rsc_op_failure(pe_resource_t * rsc, pe_node_t * node, int rc, xmlNode * xml_op, xmlNode ** last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { bool is_probe = false; pe_action_t *action = NULL; const char *key = get_op_key(xml_op); const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); CRM_ASSERT(rsc); CRM_CHECK(task != NULL, return); *last_failure = xml_op; is_probe = pcmk_xe_is_probe(xml_op); if (exit_reason == NULL) { exit_reason = ""; } if (!pcmk_is_set(data_set->flags, pe_flag_symmetric_cluster) && (rc == PCMK_OCF_NOT_INSTALLED)) { crm_trace("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " CRM_XS " rc=%d id=%s", services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, (is_probe? "probe" : task), rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); } else { crm_warn("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " CRM_XS " rc=%d id=%s", services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, (is_probe? "probe" : task), rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); if (is_probe && (rc != PCMK_OCF_OK) && (rc != PCMK_OCF_NOT_RUNNING) && (rc != PCMK_OCF_RUNNING_PROMOTED)) { /* A failed (not just unexpected) probe result could mean the user * didn't know resources will be probed even where they can't run. */ crm_notice("If it is not possible for %s to run on %s, see " "the resource-discovery option for location constraints", rsc->id, node->details->uname); } record_failed_op(xml_op, node, rsc, data_set); } action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); if (cmp_on_fail(*on_fail, action->on_fail) < 0) { pe_rsc_trace(rsc, "on-fail %s -> %s for %s (%s)", fail2text(*on_fail), fail2text(action->on_fail), action->uuid, key); *on_fail = action->on_fail; } if (!strcmp(task, CRMD_ACTION_STOP)) { resource_location(rsc, node, -INFINITY, "__stop_fail__", data_set); } else if (!strcmp(task, CRMD_ACTION_MIGRATE)) { unpack_migrate_to_failure(rsc, node, xml_op, data_set); } else if (!strcmp(task, CRMD_ACTION_MIGRATED)) { unpack_migrate_from_failure(rsc, node, xml_op, data_set); } else if (!strcmp(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_PROMOTED; } else if (!strcmp(task, CRMD_ACTION_DEMOTE)) { if (action->on_fail == action_fail_block) { rsc->role = RSC_ROLE_PROMOTED; pe__set_next_role(rsc, RSC_ROLE_STOPPED, "demote with on-fail=block"); } else if(rc == PCMK_OCF_NOT_RUNNING) { rsc->role = RSC_ROLE_STOPPED; } else { /* Staying in the promoted role would put the scheduler and * controller into a loop. Setting the role to unpromoted is not * dangerous because the resource will be stopped as part of * recovery, and any promotion will be ordered after that stop. */ rsc->role = RSC_ROLE_UNPROMOTED; } } if(is_probe && rc == PCMK_OCF_NOT_INSTALLED) { /* leave stopped */ pe_rsc_trace(rsc, "Leaving %s stopped", rsc->id); rsc->role = RSC_ROLE_STOPPED; } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "Setting %s active", rsc->id); set_active(rsc); } pe_rsc_trace(rsc, "Resource %s: role=%s, unclean=%s, on_fail=%s, fail_role=%s", rsc->id, role2text(rsc->role), pcmk__btoa(node->details->unclean), fail2text(action->on_fail), role2text(action->fail_role)); if (action->fail_role != RSC_ROLE_STARTED && rsc->next_role < action->fail_role) { pe__set_next_role(rsc, action->fail_role, "failure"); } if (action->fail_role == RSC_ROLE_STOPPED) { int score = -INFINITY; pe_resource_t *fail_rsc = rsc; if (fail_rsc->parent) { pe_resource_t *parent = uber_parent(fail_rsc); if (pe_rsc_is_clone(parent) && !pcmk_is_set(parent->flags, pe_rsc_unique)) { /* For clone resources, if a child fails on an operation * with on-fail = stop, all the resources fail. Do this by preventing * the parent from coming up again. */ fail_rsc = parent; } } crm_notice("%s will not be started under current conditions", fail_rsc->id); /* make sure it doesn't come up again */ if (fail_rsc->allowed_nodes != NULL) { g_hash_table_destroy(fail_rsc->allowed_nodes); } fail_rsc->allowed_nodes = pe__node_list2table(data_set->nodes); g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score); } pe_free_action(action); } /*! * \internal * \brief Remap informational monitor results and operation status * * For the monitor results, certain OCF codes are for providing extended information * to the user about services that aren't yet failed but not entirely healthy either. * These must be treated as the "normal" result by Pacemaker. * * For operation status, the action result can be used to determine an appropriate * status for the purposes of responding to the action. The status provided by the * executor is not directly usable since the executor does not know what was expected. * * \param[in] xml_op Operation history entry XML from CIB status * \param[in,out] rsc Resource that operation history entry is for * \param[in] node Node where operation was executed * \param[in] data_set Current cluster working set * \param[in,out] on_fail What should be done about the result * \param[in] target_rc Expected return code of operation * \param[in,out] rc Actual return code of operation * \param[in,out] status Operation execution status * * \note If the result is remapped and the node is not shutting down or failed, * the operation will be recorded in the data set's list of failed operations * to highlight it for the user. * * \note This may update the resource's current and next role. */ static void remap_operation(xmlNode *xml_op, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set, enum action_fail_response *on_fail, int target_rc, int *rc, int *status) { bool is_probe = false; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *key = get_op_key(xml_op); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_none)) { int remapped_rc = pcmk__effective_rc(*rc); if (*rc != remapped_rc) { crm_trace("Remapping monitor result %d to %d", *rc, remapped_rc); if (!node->details->shutdown || node->details->online) { record_failed_op(xml_op, node, rsc, data_set); } *rc = remapped_rc; } } if (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op)) { *status = PCMK_EXEC_DONE; *rc = PCMK_OCF_NOT_RUNNING; } /* If the executor reported an operation status of anything but done or * error, consider that final. But for done or error, we know better whether * it should be treated as a failure or not, because we know the expected * result. */ if (*status != PCMK_EXEC_DONE && *status != PCMK_EXEC_ERROR) { return; } CRM_ASSERT(rsc); CRM_CHECK(task != NULL, *status = PCMK_EXEC_ERROR; return); *status = PCMK_EXEC_DONE; if (exit_reason == NULL) { exit_reason = ""; } is_probe = pcmk_xe_is_probe(xml_op); if (is_probe) { task = "probe"; } if (target_rc < 0) { /* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with * Heartbeat 2.0.7 or earlier as the cluster layer, did not include the * target_rc in the transition key, which (along with the similar case * of a corrupted transition key in the CIB) will be reported to this * function as -1. Pacemaker 2.0+ does not support rolling upgrades from * those versions or processing of saved CIB files from those versions, * so we do not need to care much about this case. */ *status = PCMK_EXEC_ERROR; crm_warn("Expected result not found for %s on %s (corrupt or obsolete CIB?)", key, node->details->uname); } else if (target_rc != *rc) { *status = PCMK_EXEC_ERROR; pe_rsc_debug(rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)", key, node->details->uname, target_rc, services_ocf_exitcode_str(target_rc), *rc, services_ocf_exitcode_str(*rc), (*exit_reason? ": " : ""), exit_reason); } switch (*rc) { case PCMK_OCF_OK: if (is_probe && (target_rc == PCMK_OCF_NOT_RUNNING)) { *status = PCMK_EXEC_DONE; pe_rsc_info(rsc, "Probe found %s active on %s at %s", rsc->id, node->details->uname, last_change_str(xml_op)); } break; case PCMK_OCF_NOT_RUNNING: if (is_probe || (target_rc == *rc) || !pcmk_is_set(rsc->flags, pe_rsc_managed)) { *status = PCMK_EXEC_DONE; rsc->role = RSC_ROLE_STOPPED; /* clear any previous failure actions */ *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "not running"); } break; case PCMK_OCF_RUNNING_PROMOTED: if (is_probe && (*rc != target_rc)) { *status = PCMK_EXEC_DONE; pe_rsc_info(rsc, "Probe found %s active and promoted on %s at %s", rsc->id, node->details->uname, last_change_str(xml_op)); } rsc->role = RSC_ROLE_PROMOTED; break; case PCMK_OCF_DEGRADED_PROMOTED: case PCMK_OCF_FAILED_PROMOTED: rsc->role = RSC_ROLE_PROMOTED; *status = PCMK_EXEC_ERROR; break; case PCMK_OCF_NOT_CONFIGURED: *status = PCMK_EXEC_ERROR_FATAL; break; case PCMK_OCF_UNIMPLEMENT_FEATURE: { guint interval_ms = 0; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms > 0) { *status = PCMK_EXEC_NOT_SUPPORTED; break; } // fall through } case PCMK_OCF_NOT_INSTALLED: case PCMK_OCF_INVALID_PARAM: case PCMK_OCF_INSUFFICIENT_PRIV: if (!pe_can_fence(data_set, node) && !strcmp(task, CRMD_ACTION_STOP)) { /* If a stop fails and we can't fence, there's nothing else we can do */ pe_proc_err("No further recovery can be attempted for %s " "because %s on %s failed (%s%s%s) at %s " CRM_XS " rc=%d id=%s", rsc->id, task, node->details->uname, services_ocf_exitcode_str(*rc), (*exit_reason? ": " : ""), exit_reason, last_change_str(xml_op), *rc, ID(xml_op)); pe__clear_resource_flags(rsc, pe_rsc_managed); pe__set_resource_flags(rsc, pe_rsc_block); } *status = PCMK_EXEC_ERROR_HARD; break; default: if (*status == PCMK_EXEC_DONE) { crm_info("Treating unknown exit status %d from %s of %s " "on %s at %s as failure", *rc, task, rsc->id, node->details->uname, last_change_str(xml_op)); *status = PCMK_EXEC_ERROR; } break; } pe_rsc_trace(rsc, "Remapped %s status to '%s'", key, pcmk_exec_status_str(*status)); } // return TRUE if start or monitor last failure but parameters changed static bool should_clear_for_param_change(xmlNode *xml_op, const char *task, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { if (!strcmp(task, "start") || !strcmp(task, "monitor")) { if (pe__bundle_needs_remote_name(rsc, data_set)) { /* We haven't allocated resources yet, so we can't reliably * substitute addr parameters for the REMOTE_CONTAINER_HACK. * When that's needed, defer the check until later. */ pe__add_param_check(xml_op, rsc, node, pe_check_last_failure, data_set); } else { op_digest_cache_t *digest_data = NULL; digest_data = rsc_action_digest_cmp(rsc, xml_op, node, data_set); switch (digest_data->rc) { case RSC_DIGEST_UNKNOWN: crm_trace("Resource %s history entry %s on %s" " has no digest to compare", rsc->id, get_op_key(xml_op), node->details->id); break; case RSC_DIGEST_MATCH: break; default: return TRUE; } } } return FALSE; } // Order action after fencing of remote node, given connection rsc static void order_after_remote_fencing(pe_action_t *action, pe_resource_t *remote_conn, pe_working_set_t *data_set) { pe_node_t *remote_node = pe_find_node(data_set->nodes, remote_conn->id); if (remote_node) { pe_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL, FALSE, data_set); order_actions(fence, action, pe_order_implies_then); } } static bool should_ignore_failure_timeout(pe_resource_t *rsc, xmlNode *xml_op, const char *task, guint interval_ms, bool is_last_failure, pe_working_set_t *data_set) { /* Clearing failures of recurring monitors has special concerns. The * executor reports only changes in the monitor result, so if the * monitor is still active and still getting the same failure result, * that will go undetected after the failure is cleared. * * Also, the operation history will have the time when the recurring * monitor result changed to the given code, not the time when the * result last happened. * * @TODO We probably should clear such failures only when the failure * timeout has passed since the last occurrence of the failed result. * However we don't record that information. We could maybe approximate * that by clearing only if there is a more recent successful monitor or * stop result, but we don't even have that information at this point * since we are still unpacking the resource's operation history. * * This is especially important for remote connection resources with a * reconnect interval, so in that case, we skip clearing failures * if the remote node hasn't been fenced. */ if (rsc->remote_reconnect_ms && pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && (interval_ms != 0) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id); if (remote_node && !remote_node->details->remote_was_fenced) { if (is_last_failure) { crm_info("Waiting to clear monitor failure for remote node %s" " until fencing has occurred", rsc->id); } return TRUE; } } return FALSE; } /*! * \internal * \brief Check operation age and schedule failure clearing when appropriate * * This function has two distinct purposes. The first is to check whether an * operation history entry is expired (i.e. the resource has a failure timeout, * the entry is older than the timeout, and the resource either has no fail * count or its fail count is entirely older than the timeout). The second is to * schedule fail count clearing when appropriate (i.e. the operation is expired * and either the resource has an expired fail count or the operation is a * last_failure for a remote connection resource with a reconnect interval, * or the operation is a last_failure for a start or monitor operation and the * resource's parameters have changed since the operation). * * \param[in] rsc Resource that operation happened to * \param[in] node Node that operation happened on * \param[in] rc Actual result of operation * \param[in] xml_op Operation history entry XML * \param[in] data_set Current working set * * \return TRUE if operation history entry is expired, FALSE otherwise */ static bool check_operation_expiry(pe_resource_t *rsc, pe_node_t *node, int rc, xmlNode *xml_op, pe_working_set_t *data_set) { bool expired = FALSE; bool is_last_failure = pcmk__ends_with(ID(xml_op), "_last_failure_0"); time_t last_run = 0; guint interval_ms = 0; int unexpired_fail_count = 0; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *clear_reason = NULL; crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((rsc->failure_timeout > 0) && (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_run) == 0)) { // Resource has a failure-timeout, and history entry has a timestamp time_t now = get_effective_time(data_set); time_t last_failure = 0; // Is this particular operation history older than the failure timeout? if ((now >= (last_run + rsc->failure_timeout)) && !should_ignore_failure_timeout(rsc, xml_op, task, interval_ms, is_last_failure, data_set)) { expired = TRUE; } // Does the resource as a whole have an unexpired fail count? unexpired_fail_count = pe_get_failcount(node, rsc, &last_failure, pe_fc_effective, xml_op, data_set); // Update scheduler recheck time according to *last* failure crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds" " last-failure@%lld", ID(xml_op), (long long) last_run, (expired? "" : "not "), (long long) now, unexpired_fail_count, rsc->failure_timeout, (long long) last_failure); last_failure += rsc->failure_timeout + 1; if (unexpired_fail_count && (now < last_failure)) { pe__update_recheck_time(last_failure, data_set); } } if (expired) { if (pe_get_failcount(node, rsc, NULL, pe_fc_default, xml_op, data_set)) { // There is a fail count ignoring timeout if (unexpired_fail_count == 0) { // There is no fail count considering timeout clear_reason = "it expired"; } else { /* This operation is old, but there is an unexpired fail count. * In a properly functioning cluster, this should only be * possible if this operation is not a failure (otherwise the * fail count should be expired too), so this is really just a * failsafe. */ expired = FALSE; } } else if (is_last_failure && rsc->remote_reconnect_ms) { /* Clear any expired last failure when reconnect interval is set, * even if there is no fail count. */ clear_reason = "reconnect interval is set"; } } if (!expired && is_last_failure && should_clear_for_param_change(xml_op, task, rsc, node, data_set)) { clear_reason = "resource parameters have changed"; } if (clear_reason != NULL) { // Schedule clearing of the fail count pe_action_t *clear_op = pe__clear_failcount(rsc, node, clear_reason, data_set); if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled) && rsc->remote_reconnect_ms) { /* If we're clearing a remote connection due to a reconnect * interval, we want to wait until any scheduled fencing * completes. * * We could limit this to remote_node->details->unclean, but at * this point, that's always true (it won't be reliable until * after unpack_node_history() is done). */ crm_info("Clearing %s failure will wait until any scheduled " "fencing of %s completes", task, rsc->id); order_after_remote_fencing(clear_op, rsc, data_set); } } if (expired && (interval_ms == 0) && pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { switch(rc) { case PCMK_OCF_OK: case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Don't expire probes that return these values expired = FALSE; break; } } return expired; } int pe__target_rc_from_xml(xmlNode *xml_op) { int target_rc = 0; const char *key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY); if (key == NULL) { return -1; } decode_transition_key(key, NULL, NULL, NULL, &target_rc); return target_rc; } static enum action_fail_response get_action_on_fail(pe_resource_t *rsc, const char *key, const char *task, pe_working_set_t * data_set) { enum action_fail_response result = action_fail_recover; pe_action_t *action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set); result = action->on_fail; pe_free_action(action); return result; } static void update_resource_state(pe_resource_t * rsc, pe_node_t * node, xmlNode * xml_op, const char * task, int rc, xmlNode * last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set) { gboolean clear_past_failure = FALSE; CRM_ASSERT(rsc); CRM_ASSERT(xml_op); if (rc == PCMK_OCF_NOT_INSTALLED || (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op))) { rsc->role = RSC_ROLE_STOPPED; } else if (rc == PCMK_OCF_NOT_RUNNING) { clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_STATUS, pcmk__str_casei)) { if (last_failure) { const char *op_key = get_op_key(xml_op); const char *last_failure_key = get_op_key(last_failure); if (pcmk__str_eq(op_key, last_failure_key, pcmk__str_casei)) { clear_past_failure = TRUE; } } if (rsc->role < RSC_ROLE_STARTED) { set_active(rsc); } } else if (pcmk__str_eq(task, CRMD_ACTION_START, pcmk__str_casei)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_STOP, pcmk__str_casei)) { rsc->role = RSC_ROLE_STOPPED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_PROMOTE, pcmk__str_casei)) { rsc->role = RSC_ROLE_PROMOTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_DEMOTE, pcmk__str_casei)) { if (*on_fail == action_fail_demote) { // Demote clears an error only if on-fail=demote clear_past_failure = TRUE; } rsc->role = RSC_ROLE_UNPROMOTED; } else if (pcmk__str_eq(task, CRMD_ACTION_MIGRATED, pcmk__str_casei)) { rsc->role = RSC_ROLE_STARTED; clear_past_failure = TRUE; } else if (pcmk__str_eq(task, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { unpack_migrate_to_success(rsc, node, xml_op, data_set); } else if (rsc->role < RSC_ROLE_STARTED) { pe_rsc_trace(rsc, "%s active on %s", rsc->id, node->details->uname); set_active(rsc); } /* clear any previous failure actions */ if (clear_past_failure) { switch (*on_fail) { case action_fail_stop: case action_fail_fence: case action_fail_migrate: case action_fail_standby: pe_rsc_trace(rsc, "%s.%s is not cleared by a completed stop", rsc->id, fail2text(*on_fail)); break; case action_fail_block: case action_fail_ignore: case action_fail_demote: case action_fail_recover: case action_fail_restart_container: *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "clear past failures"); break; case action_fail_reset_remote: if (rsc->remote_reconnect_ms == 0) { /* With no reconnect interval, the connection is allowed to * start again after the remote node is fenced and * completely stopped. (With a reconnect interval, we wait * for the failure to be cleared entirely before attempting * to reconnect.) */ *on_fail = action_fail_ignore; pe__set_next_role(rsc, RSC_ROLE_UNKNOWN, "clear past failures and reset remote"); } break; } } } static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *on_fail, pe_working_set_t *data_set) { int rc = 0; int old_rc = 0; int task_id = 0; int target_rc = 0; int old_target_rc = 0; int status = PCMK_EXEC_UNKNOWN; guint interval_ms = 0; const char *task = NULL; const char *task_key = NULL; const char *exit_reason = NULL; bool expired = false; pe_resource_t *parent = rsc; enum action_fail_response failure_strategy = action_fail_recover; bool maskable_probe_failure = false; CRM_CHECK(rsc && node && xml_op, return); target_rc = pe__target_rc_from_xml(xml_op); task_key = get_op_key(xml_op); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); if (exit_reason == NULL) { exit_reason = ""; } crm_element_value_int(xml_op, XML_LRM_ATTR_RC, &rc); crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id); crm_element_value_int(xml_op, XML_LRM_ATTR_OPSTATUS, &status); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); CRM_CHECK(task != NULL, return); CRM_CHECK((status >= PCMK_EXEC_PENDING) && (status <= PCMK_EXEC_MAX), return); if (!strcmp(task, CRMD_ACTION_NOTIFY) || !strcmp(task, CRMD_ACTION_METADATA)) { /* safe to ignore these */ return; } if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { parent = uber_parent(rsc); } pe_rsc_trace(rsc, "Unpacking task %s/%s (call_id=%d, status=%d, rc=%d) on %s (role=%s)", task_key, task, task_id, status, rc, node->details->uname, role2text(rsc->role)); if (node->details->unclean) { pe_rsc_trace(rsc, "Node %s (where %s is running) is unclean." " Further action depends on the value of the stop's on-fail attribute", node->details->uname, rsc->id); } /* It should be possible to call remap_operation() first then call * check_operation_expiry() only if rc != target_rc, because there should * never be a fail count without at least one unexpected result in the * resource history. That would be more efficient by avoiding having to call * check_operation_expiry() for expected results. * * However, we do have such configurations in the scheduler regression * tests, even if it shouldn't be possible with the current code. It's * probably a good idea anyway, but that would require updating the test * inputs to something currently possible. */ if ((status != PCMK_EXEC_NOT_INSTALLED) && check_operation_expiry(rsc, node, rc, xml_op, data_set)) { expired = true; } old_rc = rc; old_target_rc = target_rc; remap_operation(xml_op, rsc, node, data_set, on_fail, target_rc, &rc, &status); maskable_probe_failure = !pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op); if (expired && maskable_probe_failure && old_rc != old_target_rc) { if (rsc->role <= RSC_ROLE_STOPPED) { rsc->role = RSC_ROLE_UNKNOWN; } goto done; } else if (expired && (rc != target_rc)) { const char *magic = crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC); if (interval_ms == 0) { crm_notice("Ignoring expired %s failure on %s " CRM_XS " actual=%d expected=%d magic=%s", task_key, node->details->uname, rc, target_rc, magic); goto done; } else if(node->details->online && node->details->unclean == FALSE) { /* Reschedule the recurring monitor. schedule_cancel() won't work at * this stage, so as a hacky workaround, forcibly change the restart * digest so pcmk__check_action_config() does what we want later. * * @TODO We should skip this if there is a newer successful monitor. * Also, this causes rescheduling only if the history entry * has an op-digest (which the expire-non-blocked-failure * scheduler regression test doesn't, but that may not be a * realistic scenario in production). */ crm_notice("Rescheduling %s after failure expired on %s " CRM_XS " actual=%d expected=%d magic=%s", task_key, node->details->uname, rc, target_rc, magic); crm_xml_add(xml_op, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout"); goto done; } } if (maskable_probe_failure) { crm_notice("Treating probe result '%s' for %s on %s as 'not running'", services_ocf_exitcode_str(old_rc), rsc->id, node->details->uname); update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set); crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); record_failed_op(xml_op, node, rsc, data_set); resource_location(parent, node, -INFINITY, "masked-probe-failure", data_set); goto done; } switch (status) { case PCMK_EXEC_CANCELLED: // Should never happen pe_err("Resource history contains cancellation '%s' " "(%s of %s on %s at %s)", ID(xml_op), task, rsc->id, node->details->uname, last_change_str(xml_op)); goto done; case PCMK_EXEC_PENDING: if (!strcmp(task, CRMD_ACTION_START)) { pe__set_resource_flags(rsc, pe_rsc_start_pending); set_active(rsc); } else if (!strcmp(task, CRMD_ACTION_PROMOTE)) { rsc->role = RSC_ROLE_PROMOTED; } else if (!strcmp(task, CRMD_ACTION_MIGRATE) && node->details->unclean) { /* If a pending migrate_to action is out on a unclean node, * we have to force the stop action on the target. */ const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); pe_node_t *target = pe_find_node(data_set->nodes, migrate_target); if (target) { stop_action(rsc, target, FALSE); } } if (rsc->pending_task == NULL) { if ((interval_ms != 0) || strcmp(task, CRMD_ACTION_STATUS)) { rsc->pending_task = strdup(task); rsc->pending_node = node; } else { /* Pending probes are not printed, even if pending * operations are requested. If someone ever requests that * behavior, enable the below and the corresponding part of * native.c:native_pending_task(). */ #if 0 rsc->pending_task = strdup("probe"); rsc->pending_node = node; #endif } } goto done; case PCMK_EXEC_DONE: pe_rsc_trace(rsc, "%s of %s on %s completed at %s " CRM_XS " id=%s", task, rsc->id, node->details->uname, last_change_str(xml_op), ID(xml_op)); update_resource_state(rsc, node, xml_op, task, rc, *last_failure, on_fail, data_set); goto done; case PCMK_EXEC_NOT_INSTALLED: failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if (failure_strategy == action_fail_ignore) { crm_warn("Cannot ignore failed %s of %s on %s: " "Resource agent doesn't exist " CRM_XS " status=%d rc=%d id=%s", task, rsc->id, node->details->uname, status, rc, ID(xml_op)); /* Also for printing it as "FAILED" by marking it as pe_rsc_failed later */ *on_fail = action_fail_migrate; } resource_location(parent, node, -INFINITY, "hard-error", data_set); unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); goto done; case PCMK_EXEC_NOT_CONNECTED: if (pe__is_guest_or_remote_node(node) && pcmk_is_set(node->details->remote_rsc->flags, pe_rsc_managed)) { /* We should never get into a situation where a managed remote * connection resource is considered OK but a resource action * behind the connection gets a "not connected" status. But as a * fail-safe in case a bug or unusual circumstances do lead to * that, ensure the remote connection is considered failed. */ pe__set_resource_flags(node->details->remote_rsc, pe_rsc_failed|pe_rsc_stop); } break; // Not done, do error handling case PCMK_EXEC_ERROR: case PCMK_EXEC_ERROR_HARD: case PCMK_EXEC_ERROR_FATAL: case PCMK_EXEC_TIMEOUT: case PCMK_EXEC_NOT_SUPPORTED: case PCMK_EXEC_INVALID: break; // Not done, do error handling case PCMK_EXEC_NO_FENCE_DEVICE: case PCMK_EXEC_NO_SECRETS: status = PCMK_EXEC_ERROR_HARD; break; // Not done, do error handling } failure_strategy = get_action_on_fail(rsc, task_key, task, data_set); if ((failure_strategy == action_fail_ignore) || (failure_strategy == action_fail_restart_container && !strcmp(task, CRMD_ACTION_STOP))) { crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s " "succeeded " CRM_XS " rc=%d id=%s", task, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rsc->id, node->details->uname, last_change_str(xml_op), rc, ID(xml_op)); update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set); crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); pe__set_resource_flags(rsc, pe_rsc_failure_ignored); record_failed_op(xml_op, node, rsc, data_set); if ((failure_strategy == action_fail_restart_container) && cmp_on_fail(*on_fail, action_fail_recover) <= 0) { *on_fail = failure_strategy; } } else { unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set); if (status == PCMK_EXEC_ERROR_HARD) { do_crm_log(rc != PCMK_OCF_NOT_INSTALLED?LOG_ERR:LOG_NOTICE, "Preventing %s from restarting on %s because " "of hard failure (%s%s%s)" CRM_XS " rc=%d id=%s", parent->id, node->details->uname, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rc, ID(xml_op)); resource_location(parent, node, -INFINITY, "hard-error", data_set); } else if (status == PCMK_EXEC_ERROR_FATAL) { crm_err("Preventing %s from restarting anywhere because " "of fatal failure (%s%s%s) " CRM_XS " rc=%d id=%s", parent->id, services_ocf_exitcode_str(rc), (*exit_reason? ": " : ""), exit_reason, rc, ID(xml_op)); resource_location(parent, NULL, -INFINITY, "fatal-error", data_set); } } done: pe_rsc_trace(rsc, "Resource %s after %s: role=%s, next=%s", rsc->id, task, role2text(rsc->role), role2text(rsc->next_role)); } static void add_node_attrs(xmlNode *xml_obj, pe_node_t *node, bool overwrite, pe_working_set_t *data_set) { const char *cluster_name = NULL; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_UNAME), strdup(node->details->uname)); g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID), strdup(node->details->id)); if (pcmk__str_eq(node->details->id, data_set->dc_uuid, pcmk__str_casei)) { data_set->dc_node = node; node->details->is_dc = TRUE; g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_TRUE)); } else { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_FALSE)); } cluster_name = g_hash_table_lookup(data_set->config_hash, "cluster-name"); if (cluster_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME), strdup(cluster_name)); } pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_ATTR_SETS, &rule_data, node->details->attrs, NULL, overwrite, data_set); if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) { const char *site_name = pe_node_attribute_raw(node, "site-name"); if (site_name) { g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(site_name)); } else if (cluster_name) { /* Default to cluster-name if unset */ g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_SITE_NAME), strdup(cluster_name)); } } } static GList * extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter) { int counter = -1; int stop_index = -1; int start_index = -1; xmlNode *rsc_op = NULL; GList *gIter = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; /* extract operations */ op_list = NULL; sorted_op_list = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op)) { if (pcmk__str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, pcmk__str_none)) { crm_xml_add(rsc_op, "resource", rsc); crm_xml_add(rsc_op, XML_ATTR_UNAME, node); op_list = g_list_prepend(op_list, rsc_op); } } if (op_list == NULL) { /* if there are no operations, there is nothing to do */ return NULL; } sorted_op_list = g_list_sort(op_list, sort_op_by_callid); /* create active recurring operations as optional */ if (active_filter == FALSE) { return sorted_op_list; } op_list = NULL; calculate_active_ops(sorted_op_list, &start_index, &stop_index); for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) { xmlNode *rsc_op = (xmlNode *) gIter->data; counter++; if (start_index < stop_index) { crm_trace("Skipping %s: not active", ID(rsc_entry)); break; } else if (counter < start_index) { crm_trace("Skipping %s: old", ID(rsc_op)); continue; } op_list = g_list_append(op_list, rsc_op); } g_list_free(sorted_op_list); return op_list; } GList * find_operations(const char *rsc, const char *node, gboolean active_filter, pe_working_set_t * data_set) { GList *output = NULL; GList *intermediate = NULL; xmlNode *tmp = NULL; xmlNode *status = find_xml_node(data_set->input, XML_CIB_TAG_STATUS, TRUE); pe_node_t *this_node = NULL; xmlNode *node_state = NULL; for (node_state = pcmk__xe_first_child(status); node_state != NULL; node_state = pcmk__xe_next(node_state)) { if (pcmk__str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { const char *uname = crm_element_value(node_state, XML_ATTR_UNAME); if (node != NULL && !pcmk__str_eq(uname, node, pcmk__str_casei)) { continue; } this_node = pe_find_node(data_set->nodes, uname); if(this_node == NULL) { CRM_LOG_ASSERT(this_node != NULL); continue; } else if (pe__is_guest_or_remote_node(this_node)) { determine_remote_online_status(data_set, this_node); } else { determine_online_status(node_state, this_node, data_set); } if (this_node->details->online || pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { /* offline nodes run no resources... * unless stonith is enabled in which case we need to * make sure rsc start events happen after the stonith */ xmlNode *lrm_rsc = NULL; tmp = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); tmp = find_xml_node(tmp, XML_LRM_TAG_RESOURCES, FALSE); for (lrm_rsc = pcmk__xe_first_child(tmp); lrm_rsc != NULL; lrm_rsc = pcmk__xe_next(lrm_rsc)) { if (pcmk__str_eq((const char *)lrm_rsc->name, XML_LRM_TAG_RESOURCE, pcmk__str_none)) { const char *rsc_id = crm_element_value(lrm_rsc, XML_ATTR_ID); if (rsc != NULL && !pcmk__str_eq(rsc_id, rsc, pcmk__str_casei)) { continue; } intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter); output = g_list_concat(output, intermediate); } } } } } return output; } diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c index c7be84a775..39d103658a 100644 --- a/lib/pengine/utils.c +++ b/lib/pengine/utils.c @@ -1,2626 +1,2630 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include "pe_status_private.h" extern bool pcmk__is_daemon; void print_str_str(gpointer key, gpointer value, gpointer user_data); gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data); static void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container, pe_working_set_t * data_set, guint interval_ms); static xmlNode *find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled); #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t * pe_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; CRM_CHECK(action != NULL, return NULL); if (action->action_details == NULL) { action->action_details = calloc(1, sizeof(pe_rsc_action_details_t)); CRM_CHECK(action->action_details != NULL, return NULL); } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters == NULL) { details->versioned_parameters = create_xml_node(NULL, XML_TAG_OP_VER_ATTRS); } if (details->versioned_meta == NULL) { details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META); } return details; } static void pe_free_rsc_action_details(pe_action_t *action) { pe_rsc_action_details_t *details; if ((action == NULL) || (action->action_details == NULL)) { return; } details = (pe_rsc_action_details_t *) action->action_details; if (details->versioned_parameters) { free_xml(details->versioned_parameters); } if (details->versioned_meta) { free_xml(details->versioned_meta); } action->action_details = NULL; } #endif /*! * \internal * \brief Check whether we can fence a particular node * * \param[in] data_set Working set for cluster * \param[in] node Name of node to check * * \return true if node can be fenced, false otherwise */ bool pe_can_fence(pe_working_set_t *data_set, pe_node_t *node) { if (pe__is_guest_node(node)) { /* Guest nodes are fenced by stopping their container resource. We can * do that if the container's host is either online or fenceable. */ pe_resource_t *rsc = node->details->remote_rsc->container; for (GList *n = rsc->running_on; n != NULL; n = n->next) { pe_node_t *container_node = n->data; if (!container_node->details->online && !pe_can_fence(data_set, container_node)) { return false; } } return true; } else if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { return false; /* Turned off */ } else if (!pcmk_is_set(data_set->flags, pe_flag_have_stonith_resource)) { return false; /* No devices */ } else if (pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { return true; } else if (data_set->no_quorum_policy == no_quorum_ignore) { return true; } else if(node == NULL) { return false; } else if(node->details->online) { crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname); return true; } crm_trace("Cannot fence %s", node->details->uname); return false; } /*! * \internal * \brief Copy a node object * * \param[in] this_node Node object to copy * * \return Newly allocated shallow copy of this_node * \note This function asserts on errors and is guaranteed to return non-NULL. */ pe_node_t * pe__copy_node(const pe_node_t *this_node) { pe_node_t *new_node = NULL; CRM_ASSERT(this_node != NULL); new_node = calloc(1, sizeof(pe_node_t)); CRM_ASSERT(new_node != NULL); new_node->rsc_discover_mode = this_node->rsc_discover_mode; new_node->weight = this_node->weight; new_node->fixed = this_node->fixed; new_node->details = this_node->details; return new_node; } /* any node in list1 or list2 and not in the other gets a score of -INFINITY */ void node_list_exclude(GHashTable * hash, GList *list, gboolean merge_scores) { GHashTable *result = hash; pe_node_t *other_node = NULL; GList *gIter = list; GHashTableIter iter; pe_node_t *node = NULL; g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { other_node = pe_find_node_id(list, node->details->id); if (other_node == NULL) { node->weight = -INFINITY; } else if (merge_scores) { node->weight = pcmk__add_scores(node->weight, other_node->weight); } } for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; other_node = pe_hash_table_lookup(result, node->details->id); if (other_node == NULL) { pe_node_t *new_node = pe__copy_node(node); new_node->weight = -INFINITY; g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } } } /*! * \internal * \brief Create a node hash table from a node list * * \param[in] list Node list * * \return Hash table equivalent of node list */ GHashTable * pe__node_list2table(GList *list) { GHashTable *result = NULL; result = pcmk__strkey_table(NULL, free); for (GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = pe__copy_node((pe_node_t *) gIter->data); g_hash_table_insert(result, (gpointer) new_node->details->id, new_node); } return result; } gint sort_node_uname(gconstpointer a, gconstpointer b) { return pcmk__numeric_strcasecmp(((const pe_node_t *) a)->details->uname, ((const pe_node_t *) b)->details->uname); } /*! * \internal * \brief Output node weights to stdout * * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes If rsc is not specified, use these nodes */ static void pe__output_node_weights(pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set) { pcmk__output_t *out = data_set->priv; char score[128]; // Stack-allocated since this is called frequently // Sort the nodes so the output is consistent for regression tests GList *list = g_list_sort(g_hash_table_get_values(nodes), sort_node_uname); for (GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *node = (pe_node_t *) gIter->data; score2char_stack(node->weight, score, sizeof(score)); out->message(out, "node-weight", rsc, comment, node->details->uname, score); } g_list_free(list); } /*! * \internal * \brief Log node weights at trace level * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes If rsc is not specified, use these nodes */ static void pe__log_node_weights(const char *file, const char *function, int line, pe_resource_t *rsc, const char *comment, GHashTable *nodes) { GHashTableIter iter; pe_node_t *node = NULL; char score[128]; // Stack-allocated since this is called frequently // Don't waste time if we're not tracing at this point pcmk__log_else(LOG_TRACE, return); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { score2char_stack(node->weight, score, sizeof(score)); if (rsc) { qb_log_from_external_source(function, file, "%s: %s allocation score on %s: %s", LOG_TRACE, line, 0, comment, rsc->id, node->details->uname, score); } else { qb_log_from_external_source(function, file, "%s: %s = %s", LOG_TRACE, line, 0, comment, node->details->uname, score); } } } /*! * \internal * \brief Log or output node weights * * \param[in] file Caller's filename * \param[in] function Caller's function name * \param[in] line Caller's line number * \param[in] to_log Log if true, otherwise output * \param[in] rsc Use allowed nodes for this resource * \param[in] comment Text description to prefix lines with * \param[in] nodes Use these nodes */ void pe__show_node_weights_as(const char *file, const char *function, int line, bool to_log, pe_resource_t *rsc, const char *comment, GHashTable *nodes, pe_working_set_t *data_set) { if (rsc != NULL && pcmk_is_set(rsc->flags, pe_rsc_orphan)) { // Don't show allocation scores for orphans return; } if (nodes == NULL) { // Nothing to show return; } if (to_log) { pe__log_node_weights(file, function, line, rsc, comment, nodes); } else { pe__output_node_weights(rsc, comment, nodes, data_set); } // If this resource has children, repeat recursively for each if (rsc && rsc->children) { for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe_resource_t *child = (pe_resource_t *) gIter->data; pe__show_node_weights_as(file, function, line, to_log, child, comment, child->allowed_nodes, data_set); } } } gint sort_rsc_index(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->sort_index > resource2->sort_index) { return -1; } if (resource1->sort_index < resource2->sort_index) { return 1; } return 0; } gint sort_rsc_priority(gconstpointer a, gconstpointer b) { const pe_resource_t *resource1 = (const pe_resource_t *)a; const pe_resource_t *resource2 = (const pe_resource_t *)b; if (a == NULL && b == NULL) { return 0; } if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (resource1->priority > resource2->priority) { return -1; } if (resource1->priority < resource2->priority) { return 1; } return 0; } static enum pe_quorum_policy effective_quorum_policy(pe_resource_t *rsc, pe_working_set_t *data_set) { enum pe_quorum_policy policy = data_set->no_quorum_policy; if (pcmk_is_set(data_set->flags, pe_flag_have_quorum)) { policy = no_quorum_ignore; } else if (data_set->no_quorum_policy == no_quorum_demote) { switch (rsc->role) { case RSC_ROLE_PROMOTED: case RSC_ROLE_UNPROMOTED: if (rsc->next_role > RSC_ROLE_UNPROMOTED) { pe__set_next_role(rsc, RSC_ROLE_UNPROMOTED, "no-quorum-policy=demote"); } policy = no_quorum_ignore; break; default: policy = no_quorum_stop; break; } } return policy; } static void add_singleton(pe_working_set_t *data_set, pe_action_t *action) { if (data_set->singletons == NULL) { data_set->singletons = pcmk__strkey_table(NULL, NULL); } g_hash_table_insert(data_set->singletons, action->uuid, action); } static pe_action_t * lookup_singleton(pe_working_set_t *data_set, const char *action_uuid) { if (data_set->singletons == NULL) { return NULL; } return g_hash_table_lookup(data_set->singletons, action_uuid); } /*! * \internal * \brief Find an existing action that matches arguments * * \param[in] key Action key to match * \param[in] rsc Resource to match (if any) * \param[in] node Node to match (if any) * \param[in] data_set Cluster working set * * \return Existing action that matches arguments (or NULL if none) */ static pe_action_t * find_existing_action(const char *key, pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { GList *matches = NULL; pe_action_t *action = NULL; /* When rsc is NULL, it would be quicker to check data_set->singletons, * but checking all data_set->actions takes the node into account. */ matches = find_actions(((rsc == NULL)? data_set->actions : rsc->actions), key, node); if (matches == NULL) { return NULL; } CRM_LOG_ASSERT(!pcmk__list_of_multiple(matches)); action = matches->data; g_list_free(matches); return action; } /*! * \internal * \brief Create a new action object * * \param[in] key Action key * \param[in] task Action name * \param[in] rsc Resource that action is for (if any) * \param[in] node Node that action is on (if any) * \param[in] optional Whether action should be considered optional * \param[in] for_graph Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \return Newly allocated action * \note This function takes ownership of \p key. It is the caller's * responsibility to free the return value with pe_free_action(). */ static pe_action_t * new_action(char *key, const char *task, pe_resource_t *rsc, pe_node_t *node, bool optional, bool for_graph, pe_working_set_t *data_set) { pe_action_t *action = calloc(1, sizeof(pe_action_t)); CRM_ASSERT(action != NULL); action->rsc = rsc; action->task = strdup(task); CRM_ASSERT(action->task != NULL); action->uuid = key; action->extra = pcmk__strkey_table(free, free); action->meta = pcmk__strkey_table(free, free); if (node) { action->node = pe__copy_node(node); } if (pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { // Resource history deletion for a node can be done on the DC pe__set_action_flags(action, pe_action_dc); } pe__set_action_flags(action, pe_action_runnable); if (optional) { pe__set_action_flags(action, pe_action_optional); } else { pe__clear_action_flags(action, pe_action_optional); } if (rsc != NULL) { guint interval_ms = 0; action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE); parse_op_key(key, NULL, NULL, &interval_ms); unpack_operation(action, action->op_entry, rsc->container, data_set, interval_ms); } if (for_graph) { pe_rsc_trace(rsc, "Created %s action %d (%s): %s for %s on %s", (optional? "optional" : "required"), data_set->action_id, key, task, ((rsc == NULL)? "no resource" : rsc->id), ((node == NULL)? "no node" : node->details->uname)); action->id = data_set->action_id++; data_set->actions = g_list_prepend(data_set->actions, action); if (rsc == NULL) { add_singleton(data_set, action); } else { rsc->actions = g_list_prepend(rsc->actions, action); } } return action; } /*! * \internal * \brief Evaluate node attribute values for an action * * \param[in] action Action to unpack attributes for * \param[in] data_set Cluster working set */ static void unpack_action_node_attributes(pe_action_t *action, pe_working_set_t *data_set) { if (!pcmk_is_set(action->flags, pe_action_have_node_attrs) && (action->op_entry != NULL)) { pe_rule_eval_data_t rule_data = { .node_hash = action->node->details->attrs, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; pe__set_action_flags(action, pe_action_have_node_attrs); pe__unpack_dataset_nvpairs(action->op_entry, XML_TAG_ATTR_SETS, &rule_data, action->extra, NULL, FALSE, data_set); } } /*! * \internal * \brief Update an action's optional flag * * \param[in] action Action to update * \param[in] optional Requested optional status */ static void update_action_optional(pe_action_t *action, gboolean optional) { // Force a non-recurring action to be optional if its resource is unmanaged if ((action->rsc != NULL) && (action->node != NULL) && !pcmk_is_set(action->flags, pe_action_pseudo) && !pcmk_is_set(action->rsc->flags, pe_rsc_managed) && (g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS) == NULL)) { pe_rsc_debug(action->rsc, "%s on %s is optional (%s is unmanaged)", action->uuid, action->node->details->uname, action->rsc->id); pe__set_action_flags(action, pe_action_optional); // We shouldn't clear runnable here because ... something // Otherwise require the action if requested } else if (!optional) { pe__clear_action_flags(action, pe_action_optional); } } /*! * \internal * \brief Update a resource action's runnable flag * * \param[in] action Action to update * \param[in] for_graph Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \note This may also schedule fencing if a stop is unrunnable. */ static void update_resource_action_runnable(pe_action_t *action, bool for_graph, pe_working_set_t *data_set) { if (pcmk_is_set(action->flags, pe_action_pseudo)) { return; } if (action->node == NULL) { pe_rsc_trace(action->rsc, "%s is unrunnable (unallocated)", action->uuid); pe__clear_action_flags(action, pe_action_runnable); } else if (!pcmk_is_set(action->flags, pe_action_dc) && !(action->node->details->online) && (!pe__is_guest_node(action->node) || action->node->details->remote_requires_reset)) { pe__clear_action_flags(action, pe_action_runnable); do_crm_log((for_graph? LOG_WARNING: LOG_TRACE), "%s on %s is unrunnable (node is offline)", action->uuid, action->node->details->uname); if (pcmk_is_set(action->rsc->flags, pe_rsc_managed) && for_graph && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei) && !(action->node->details->unclean)) { pe_fence_node(data_set, action->node, "stop is unrunnable", false); } } else if (!pcmk_is_set(action->flags, pe_action_dc) && action->node->details->pending) { pe__clear_action_flags(action, pe_action_runnable); do_crm_log((for_graph? LOG_WARNING: LOG_TRACE), "Action %s on %s is unrunnable (node is pending)", action->uuid, action->node->details->uname); } else if (action->needs == rsc_req_nothing) { pe_action_set_reason(action, NULL, TRUE); if (pe__is_guest_node(action->node) && !pe_can_fence(data_set, action->node)) { /* An action that requires nothing usually does not require any * fencing in order to be runnable. However, there is an exception: * such an action cannot be completed if it is on a guest node whose * host is unclean and cannot be fenced. */ pe_rsc_debug(action->rsc, "%s on %s is unrunnable " "(node's host cannot be fenced)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); } else { pe_rsc_trace(action->rsc, "%s on %s does not require fencing or quorum", action->uuid, action->node->details->uname); pe__set_action_flags(action, pe_action_runnable); } } else { switch (effective_quorum_policy(action->rsc, data_set)) { case no_quorum_stop: pe_rsc_debug(action->rsc, "%s on %s is unrunnable (no quorum)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); pe_action_set_reason(action, "no quorum", true); break; case no_quorum_freeze: if (!action->rsc->fns->active(action->rsc, TRUE) || (action->rsc->next_role > action->rsc->role)) { pe_rsc_debug(action->rsc, "%s on %s is unrunnable (no quorum)", action->uuid, action->node->details->uname); pe__clear_action_flags(action, pe_action_runnable); pe_action_set_reason(action, "quorum freeze", true); } break; default: //pe_action_set_reason(action, NULL, TRUE); pe__set_action_flags(action, pe_action_runnable); break; } } } /*! * \internal * \brief Update a resource object's flags for a new action on it * * \param[in] rsc Resource that action is for (if any) * \param[in] action New action */ static void update_resource_flags_for_action(pe_resource_t *rsc, pe_action_t *action) { /* @COMPAT pe_rsc_starting and pe_rsc_stopping are not actually used * within Pacemaker, and should be deprecated and eventually removed */ if (pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { pe__set_resource_flags(rsc, pe_rsc_stopping); } else if (pcmk__str_eq(action->task, CRMD_ACTION_START, pcmk__str_casei)) { if (pcmk_is_set(action->flags, pe_action_runnable)) { pe__set_resource_flags(rsc, pe_rsc_starting); } else { pe__clear_resource_flags(rsc, pe_rsc_starting); } } } /*! * \brief Create or update an action object * * \param[in] rsc Resource that action is for (if any) * \param[in] key Action key (must be non-NULL) * \param[in] task Action name (must be non-NULL) * \param[in] on_node Node that action is on (if any) * \param[in] optional Whether action should be considered optional * \param[in] save_action Whether action should be recorded in transition graph * \param[in] data_set Cluster working set * * \return Action object corresponding to arguments * \note This function takes ownership of (and might free) \p key. If * \p save_action is true, \p data_set will own the returned action, * otherwise it is the caller's responsibility to free the return value * with pe_free_action(). */ pe_action_t * custom_action(pe_resource_t *rsc, char *key, const char *task, pe_node_t *on_node, gboolean optional, gboolean save_action, pe_working_set_t *data_set) { pe_action_t *action = NULL; CRM_ASSERT((key != NULL) && (task != NULL) && (data_set != NULL)); if (save_action) { action = find_existing_action(key, rsc, on_node, data_set); } if (action == NULL) { action = new_action(key, task, rsc, on_node, optional, save_action, data_set); } else { free(key); } update_action_optional(action, optional); if (rsc != NULL) { if (action->node != NULL) { unpack_action_node_attributes(action, data_set); } update_resource_action_runnable(action, save_action, data_set); if (save_action) { update_resource_flags_for_action(rsc, action); } } return action; } static bool valid_stop_on_fail(const char *value) { return !pcmk__strcase_any_of(value, "standby", "demote", "stop", NULL); } static const char * unpack_operation_on_fail(pe_action_t * action) { const char *name = NULL; const char *role = NULL; const char *on_fail = NULL; const char *interval_spec = NULL; const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL); if (pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei) && !valid_stop_on_fail(value)) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s stop " "action to default value because '%s' is not " "allowed for stop", action->rsc->id, value); return NULL; } else if (pcmk__str_eq(action->task, CRMD_ACTION_DEMOTE, pcmk__str_casei) && !value) { // demote on_fail defaults to monitor value for promoted role if present xmlNode *operation = NULL; CRM_CHECK(action->rsc != NULL, return NULL); for (operation = pcmk__xe_first_child(action->rsc->ops_xml); (operation != NULL) && (value == NULL); operation = pcmk__xe_next(operation)) { bool enabled = false; if (!pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { continue; } name = crm_element_value(operation, "name"); role = crm_element_value(operation, "role"); on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!on_fail) { continue; } else if (pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } else if (!pcmk__str_eq(name, "monitor", pcmk__str_casei) || !pcmk__strcase_any_of(role, RSC_ROLE_PROMOTED_S, RSC_ROLE_PROMOTED_LEGACY_S, NULL)) { continue; } else if (crm_parse_interval_spec(interval_spec) == 0) { continue; } else if (pcmk__str_eq(on_fail, "demote", pcmk__str_casei)) { continue; } value = on_fail; } } else if (pcmk__str_eq(action->task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { value = "ignore"; } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { name = crm_element_value(action->op_entry, "name"); role = crm_element_value(action->op_entry, "role"); interval_spec = crm_element_value(action->op_entry, XML_LRM_ATTR_INTERVAL); if (!pcmk__str_eq(name, CRMD_ACTION_PROMOTE, pcmk__str_casei) && (!pcmk__str_eq(name, CRMD_ACTION_STATUS, pcmk__str_casei) || !pcmk__strcase_any_of(role, RSC_ROLE_PROMOTED_S, RSC_ROLE_PROMOTED_LEGACY_S, NULL) || (crm_parse_interval_spec(interval_spec) == 0))) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s %s " "action to default value because 'demote' is not " "allowed for it", action->rsc->id, name); return NULL; } } return value; } static xmlNode * find_min_interval_mon(pe_resource_t * rsc, gboolean include_disabled) { guint interval_ms = 0; guint min_interval_ms = G_MAXUINT; const char *name = NULL; const char *interval_spec = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { bool enabled = false; name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!include_disabled && pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } if (!pcmk__str_eq(name, RSC_STATUS, pcmk__str_casei)) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); if (interval_ms && (interval_ms < min_interval_ms)) { min_interval_ms = interval_ms; op = operation; } } } return op; } static int unpack_start_delay(const char *value, GHashTable *meta) { int start_delay = 0; if (value != NULL) { start_delay = crm_get_msec(value); if (start_delay < 0) { start_delay = 0; } if (meta) { g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), pcmk__itoa(start_delay)); } } return start_delay; } // true if value contains valid, non-NULL interval origin for recurring op static bool unpack_interval_origin(const char *value, xmlNode *xml_obj, guint interval_ms, crm_time_t *now, long long *start_delay) { long long result = 0; guint interval_sec = interval_ms / 1000; crm_time_t *origin = NULL; // Ignore unspecified values and non-recurring operations if ((value == NULL) || (interval_ms == 0) || (now == NULL)) { return false; } // Parse interval origin from text origin = crm_time_new(value); if (origin == NULL) { pcmk__config_err("Ignoring '" XML_OP_ATTR_ORIGIN "' for operation " "'%s' because '%s' is not valid", (ID(xml_obj)? ID(xml_obj) : "(missing ID)"), value); return false; } // Get seconds since origin (negative if origin is in the future) result = crm_time_get_seconds(now) - crm_time_get_seconds(origin); crm_time_free(origin); // Calculate seconds from closest interval to now result = result % interval_sec; // Calculate seconds remaining until next interval result = ((result <= 0)? 0 : interval_sec) - result; crm_info("Calculated a start delay of %llds for operation '%s'", result, (ID(xml_obj)? ID(xml_obj) : "(unspecified)")); if (start_delay != NULL) { *start_delay = result * 1000; // milliseconds } return true; } static int unpack_timeout(const char *value) { int timeout_ms = crm_get_msec(value); if (timeout_ms < 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } return timeout_ms; } int pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set) { xmlNode *child = NULL; GHashTable *action_meta = NULL; const char *timeout_spec = NULL; int timeout_ms = 0; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP); child != NULL; child = crm_next_same_xml(child)) { if (pcmk__str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME), pcmk__str_casei)) { timeout_spec = crm_element_value(child, XML_ATTR_TIMEOUT); break; } } if (timeout_spec == NULL && data_set->op_defaults) { action_meta = pcmk__strkey_table(free, free); pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, &rule_data, action_meta, NULL, FALSE, data_set); timeout_spec = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT); } // @TODO check meta-attributes (including versioned meta-attributes) // @TODO maybe use min-interval monitor timeout as default for monitors timeout_ms = crm_get_msec(timeout_spec); if (timeout_ms < 0) { timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S); } if (action_meta != NULL) { g_hash_table_destroy(action_meta); } return timeout_ms; } #if ENABLE_VERSIONED_ATTRS static void unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj, guint interval_ms, crm_time_t *now) { xmlNode *attrs = NULL; xmlNode *attr = NULL; for (attrs = pcmk__xe_first_child(versioned_meta); attrs != NULL; attrs = pcmk__xe_next(attrs)) { for (attr = pcmk__xe_first_child(attrs); attr != NULL; attr = pcmk__xe_next(attr)) { const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME); const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE); if (pcmk__str_eq(name, XML_OP_ATTR_START_DELAY, pcmk__str_casei)) { int start_delay = unpack_start_delay(value, NULL); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } else if (pcmk__str_eq(name, XML_OP_ATTR_ORIGIN, pcmk__str_casei)) { long long start_delay = 0; if (unpack_interval_origin(value, xml_obj, interval_ms, now, &start_delay)) { crm_xml_add(attr, XML_NVPAIR_ATTR_NAME, XML_OP_ATTR_START_DELAY); crm_xml_add_ll(attr, XML_NVPAIR_ATTR_VALUE, start_delay); } } else if (pcmk__str_eq(name, XML_ATTR_TIMEOUT, pcmk__str_casei)) { int timeout_ms = unpack_timeout(value); crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout_ms); } } } } #endif /*! * \brief Unpack operation XML into an action structure * * Unpack an operation's meta-attributes (normalizing the interval, timeout, * and start delay values as integer milliseconds), requirements, and * failure policy. * * \param[in,out] action Action to unpack into * \param[in] xml_obj Operation XML (or NULL if all defaults) * \param[in] container Resource that contains affected resource, if any * \param[in] data_set Cluster state * \param[in] interval_ms How frequently to perform the operation */ static void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container, pe_working_set_t * data_set, guint interval_ms) { int timeout_ms = 0; const char *value = NULL; bool is_probe = false; #if ENABLE_VERSIONED_ATTRS pe_rsc_action_details_t *rsc_details = NULL; #endif pe_rsc_eval_data_t rsc_rule_data = { .standard = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_CLASS), .provider = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_PROVIDER), .agent = crm_element_value(action->rsc->xml, XML_EXPR_ATTR_TYPE) }; pe_op_eval_data_t op_rule_data = { .op_name = action->task, .interval = interval_ms }; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = RSC_ROLE_UNKNOWN, .now = data_set->now, .match_data = NULL, .rsc_data = &rsc_rule_data, .op_data = &op_rule_data }; CRM_CHECK(action && action->rsc, return); is_probe = pcmk_is_probe(action->task, interval_ms); // Cluster-wide pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, &rule_data, action->meta, NULL, FALSE, data_set); // Determine probe default timeout differently if (is_probe) { xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE); if (min_interval_mon) { value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT); if (value) { crm_trace("\t%s: Setting default timeout to minimum-interval " "monitor's timeout '%s'", action->uuid, value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), strdup(value)); } } } if (xml_obj) { xmlAttrPtr xIter = NULL; // take precedence over defaults pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_META_SETS, &rule_data, action->meta, NULL, TRUE, data_set); #if ENABLE_VERSIONED_ATTRS rsc_details = pe_rsc_action_details(action); /* Non-versioned attributes also unpack XML_TAG_ATTR_SETS, but that * capability is deprecated, so we don't need to extend that support to * versioned attributes. */ pe_eval_versioned_attributes(data_set->input, xml_obj, XML_TAG_META_SETS, &rule_data, rsc_details->versioned_meta, NULL); #endif /* Anything set as an XML property has highest precedence. * This ensures we use the name and interval from the tag. */ for (xIter = xml_obj->properties; xIter; xIter = xIter->next) { const char *prop_name = (const char *)xIter->name; const char *prop_value = crm_element_value(xml_obj, prop_name); g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value)); } } g_hash_table_remove(action->meta, "id"); // Normalize interval to milliseconds if (interval_ms > 0) { g_hash_table_replace(action->meta, strdup(XML_LRM_ATTR_INTERVAL), crm_strdup_printf("%u", interval_ms)); } else { g_hash_table_remove(action->meta, XML_LRM_ATTR_INTERVAL); } /* * Timeout order of precedence: * 1. pcmk_monitor_timeout (if rsc has pcmk_ra_cap_fence_params * and task is start or a probe; pcmk_monitor_timeout works * by default for a recurring monitor) * 2. explicit op timeout on the primitive * 3. default op timeout * a. if probe, then min-interval monitor's timeout * b. else, in XML_CIB_TAG_OPCONFIG * 4. CRM_DEFAULT_OP_TIMEOUT_S * * #1 overrides general rule of XML property having highest * precedence. */ if (pcmk_is_set(pcmk_get_ra_caps(rsc_rule_data.standard), pcmk_ra_cap_fence_params) && (pcmk__str_eq(action->task, RSC_START, pcmk__str_casei) || is_probe)) { GHashTable *params = pe_rsc_params(action->rsc, action->node, data_set); value = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (value) { crm_trace("\t%s: Setting timeout to pcmk_monitor_timeout '%s', " "overriding default", action->uuid, value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), strdup(value)); } } // Normalize timeout to positive milliseconds value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT); timeout_ms = unpack_timeout(value); g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), pcmk__itoa(timeout_ms)); if (!pcmk__strcase_any_of(action->task, RSC_START, RSC_PROMOTE, NULL)) { action->needs = rsc_req_nothing; value = "nothing (not start or promote)"; } else if (pcmk_is_set(action->rsc->flags, pe_rsc_needs_fencing)) { action->needs = rsc_req_stonith; value = "fencing"; } else if (pcmk_is_set(action->rsc->flags, pe_rsc_needs_quorum)) { action->needs = rsc_req_quorum; value = "quorum"; } else { action->needs = rsc_req_nothing; value = "nothing"; } pe_rsc_trace(action->rsc, "%s requires %s", action->uuid, value); value = unpack_operation_on_fail(action); if (value == NULL) { } else if (pcmk__str_eq(value, "block", pcmk__str_casei)) { action->on_fail = action_fail_block; g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block")); value = "block"; // The above could destroy the original string } else if (pcmk__str_eq(value, "fence", pcmk__str_casei)) { action->on_fail = action_fail_fence; value = "node fencing"; if (!pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for " "operation '%s' to 'stop' because 'fence' is not " "valid when fencing is disabled", action->uuid); action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } } else if (pcmk__str_eq(value, "standby", pcmk__str_casei)) { action->on_fail = action_fail_standby; value = "node standby"; - } else if (pcmk__strcase_any_of(value, "ignore", "nothing", NULL)) { + } else if (pcmk__strcase_any_of(value, "ignore", PCMK__VALUE_NOTHING, + NULL)) { action->on_fail = action_fail_ignore; value = "ignore"; } else if (pcmk__str_eq(value, "migrate", pcmk__str_casei)) { action->on_fail = action_fail_migrate; value = "force migration"; } else if (pcmk__str_eq(value, "stop", pcmk__str_casei)) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop resource"; } else if (pcmk__str_eq(value, "restart", pcmk__str_casei)) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate)"; } else if (pcmk__str_eq(value, "restart-container", pcmk__str_casei)) { if (container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate)"; } else { value = NULL; } } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { action->on_fail = action_fail_demote; value = "demote instance"; } else { pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value); value = NULL; } /* defaults */ if (value == NULL && container) { action->on_fail = action_fail_restart_container; value = "restart container (and possibly migrate) (default)"; /* For remote nodes, ensure that any failure that results in dropping an * active connection to the node results in fencing of the node. * * There are only two action failures that don't result in fencing. * 1. probes - probe failures are expected. * 2. start - a start failure indicates that an active connection does not already * exist. The user can set op on-fail=fence if they really want to fence start * failures. */ } else if (((value == NULL) || !pcmk_is_set(action->rsc->flags, pe_rsc_managed)) && pe__resource_is_remote_conn(action->rsc, data_set) && !(pcmk__str_eq(action->task, CRMD_ACTION_STATUS, pcmk__str_casei) && (interval_ms == 0)) && !pcmk__str_eq(action->task, CRMD_ACTION_START, pcmk__str_casei)) { if (!pcmk_is_set(action->rsc->flags, pe_rsc_managed)) { action->on_fail = action_fail_stop; action->fail_role = RSC_ROLE_STOPPED; value = "stop unmanaged remote node (enforcing default)"; } else { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { value = "fence remote node (default)"; } else { value = "recover remote node connection (default)"; } if (action->rsc->remote_reconnect_ms) { action->fail_role = RSC_ROLE_STOPPED; } action->on_fail = action_fail_reset_remote; } } else if (value == NULL && pcmk__str_eq(action->task, CRMD_ACTION_STOP, pcmk__str_casei)) { if (pcmk_is_set(data_set->flags, pe_flag_stonith_enabled)) { action->on_fail = action_fail_fence; value = "resource fence (default)"; } else { action->on_fail = action_fail_block; value = "resource block (default)"; } } else if (value == NULL) { action->on_fail = action_fail_recover; value = "restart (and possibly migrate) (default)"; } pe_rsc_trace(action->rsc, "%s failure handling: %s", action->uuid, value); value = NULL; if (xml_obj != NULL) { value = g_hash_table_lookup(action->meta, "role_after_failure"); if (value) { pe_warn_once(pe_wo_role_after, "Support for role_after_failure is deprecated and will be removed in a future release"); } } if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) { action->fail_role = text2role(value); } /* defaults */ if (action->fail_role == RSC_ROLE_UNKNOWN) { if (pcmk__str_eq(action->task, CRMD_ACTION_PROMOTE, pcmk__str_casei)) { action->fail_role = RSC_ROLE_UNPROMOTED; } else { action->fail_role = RSC_ROLE_STARTED; } } pe_rsc_trace(action->rsc, "%s failure results in: %s", action->uuid, role2text(action->fail_role)); value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY); if (value) { unpack_start_delay(value, action->meta); } else { long long start_delay = 0; value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN); if (unpack_interval_origin(value, xml_obj, interval_ms, data_set->now, &start_delay)) { g_hash_table_replace(action->meta, strdup(XML_OP_ATTR_START_DELAY), crm_strdup_printf("%lld", start_delay)); } } #if ENABLE_VERSIONED_ATTRS unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms, data_set->now); #endif } static xmlNode * find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled) { guint interval_ms = 0; gboolean do_retry = TRUE; char *local_key = NULL; const char *name = NULL; const char *interval_spec = NULL; char *match_key = NULL; xmlNode *op = NULL; xmlNode *operation = NULL; retry: for (operation = pcmk__xe_first_child(rsc->ops_xml); operation != NULL; operation = pcmk__xe_next(operation)) { if (pcmk__str_eq((const char *)operation->name, "op", pcmk__str_none)) { bool enabled = false; name = crm_element_value(operation, "name"); interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL); if (!include_disabled && pcmk__xe_get_bool_attr(operation, "enabled", &enabled) == pcmk_rc_ok && !enabled) { continue; } interval_ms = crm_parse_interval_spec(interval_spec); match_key = pcmk__op_key(rsc->id, name, interval_ms); if (pcmk__str_eq(key, match_key, pcmk__str_casei)) { op = operation; } free(match_key); if (rsc->clone_name) { match_key = pcmk__op_key(rsc->clone_name, name, interval_ms); if (pcmk__str_eq(key, match_key, pcmk__str_casei)) { op = operation; } free(match_key); } if (op != NULL) { free(local_key); return op; } } } free(local_key); if (do_retry == FALSE) { return NULL; } do_retry = FALSE; if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) { local_key = pcmk__op_key(rsc->id, "migrate", 0); key = local_key; goto retry; } else if (strstr(key, "_notify_")) { local_key = pcmk__op_key(rsc->id, "notify", 0); key = local_key; goto retry; } return NULL; } xmlNode * find_rsc_op_entry(pe_resource_t * rsc, const char *key) { return find_rsc_op_entry_helper(rsc, key, FALSE); } /* * Used by the HashTable for-loop */ void print_str_str(gpointer key, gpointer value, gpointer user_data) { crm_trace("%s%s %s ==> %s", user_data == NULL ? "" : (char *)user_data, user_data == NULL ? "" : ": ", (char *)key, (char *)value); } void pe_free_action(pe_action_t * action) { if (action == NULL) { return; } g_list_free_full(action->actions_before, free); /* pe_action_wrapper_t* */ g_list_free_full(action->actions_after, free); /* pe_action_wrapper_t* */ if (action->extra) { g_hash_table_destroy(action->extra); } if (action->meta) { g_hash_table_destroy(action->meta); } #if ENABLE_VERSIONED_ATTRS if (action->rsc) { pe_free_rsc_action_details(action); } #endif free(action->cancel_task); free(action->reason); free(action->task); free(action->uuid); free(action->node); free(action); } GList * find_recurring_actions(GList *input, pe_node_t * not_on_node) { const char *value = NULL; GList *result = NULL; GList *gIter = input; CRM_CHECK(input != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS); if (value == NULL) { /* skip */ } else if (pcmk__str_eq(value, "0", pcmk__str_casei)) { /* skip */ } else if (pcmk__str_eq(CRMD_ACTION_CANCEL, action->task, pcmk__str_casei)) { /* skip */ } else if (not_on_node == NULL) { crm_trace("(null) Found: %s", action->uuid); result = g_list_prepend(result, action); } else if (action->node == NULL) { /* skip */ } else if (action->node->details != not_on_node->details) { crm_trace("Found: %s", action->uuid); result = g_list_prepend(result, action); } } return result; } enum action_tasks get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic) { enum action_tasks task = text2task(name); if (rsc == NULL) { return task; } else if (allow_non_atomic == FALSE || rsc->variant == pe_native) { switch (task) { case stopped_rsc: case started_rsc: case action_demoted: case action_promoted: crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id); return task - 1; default: break; } } return task; } pe_action_t * find_first_action(GList *input, const char *uuid, const char *task, pe_node_t * on_node) { GList *gIter = NULL; CRM_CHECK(uuid || task, return NULL); for (gIter = input; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (uuid != NULL && !pcmk__str_eq(uuid, action->uuid, pcmk__str_casei)) { continue; } else if (task != NULL && !pcmk__str_eq(task, action->task, pcmk__str_casei)) { continue; } else if (on_node == NULL) { return action; } else if (action->node == NULL) { continue; } else if (on_node->details == action->node->details) { return action; } } return NULL; } GList * find_actions(GList *input, const char *key, const pe_node_t *on_node) { GList *gIter = input; GList *result = NULL; CRM_CHECK(key != NULL, return NULL); for (; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if (!pcmk__str_eq(key, action->uuid, pcmk__str_casei)) { continue; } else if (on_node == NULL) { crm_trace("Action %s matches (ignoring node)", key); result = g_list_prepend(result, action); } else if (action->node == NULL) { crm_trace("Action %s matches (unallocated, assigning to %s)", key, on_node->details->uname); action->node = pe__copy_node(on_node); result = g_list_prepend(result, action); } else if (on_node->details == action->node->details) { crm_trace("Action %s on %s matches", key, on_node->details->uname); result = g_list_prepend(result, action); } } return result; } GList * find_actions_exact(GList *input, const char *key, const pe_node_t *on_node) { GList *result = NULL; CRM_CHECK(key != NULL, return NULL); if (on_node == NULL) { return NULL; } for (GList *gIter = input; gIter != NULL; gIter = gIter->next) { pe_action_t *action = (pe_action_t *) gIter->data; if ((action->node != NULL) && pcmk__str_eq(key, action->uuid, pcmk__str_casei) && pcmk__str_eq(on_node->details->id, action->node->details->id, pcmk__str_casei)) { crm_trace("Action %s on %s matches", key, on_node->details->uname); result = g_list_prepend(result, action); } } return result; } /*! * \brief Find all actions of given type for a resource * * \param[in] rsc Resource to search * \param[in] node Find only actions scheduled on this node * \param[in] task Action name to search for * \param[in] require_node If TRUE, NULL node or action node will not match * * \return List of actions found (or NULL if none) * \note If node is not NULL and require_node is FALSE, matching actions * without a node will be assigned to node. */ GList * pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node, const char *task, bool require_node) { GList *result = NULL; char *key = pcmk__op_key(rsc->id, task, 0); if (require_node) { result = find_actions_exact(rsc->actions, key, node); } else { result = find_actions(rsc->actions, key, node); } free(key); return result; } static void resource_node_score(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag) { pe_node_t *match = NULL; if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never)) && pcmk__str_eq(tag, "symmetric_default", pcmk__str_casei)) { /* This string comparision may be fragile, but exclusive resources and * exclusive nodes should not have the symmetric_default constraint * applied to them. */ return; } else if (rsc->children) { GList *gIter = rsc->children; for (; gIter != NULL; gIter = gIter->next) { pe_resource_t *child_rsc = (pe_resource_t *) gIter->data; resource_node_score(child_rsc, node, score, tag); } } pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score); match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id); if (match == NULL) { match = pe__copy_node(node); g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match); } match->weight = pcmk__add_scores(match->weight, score); } void resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag, pe_working_set_t * data_set) { if (node != NULL) { resource_node_score(rsc, node, score, tag); } else if (data_set != NULL) { GList *gIter = data_set->nodes; for (; gIter != NULL; gIter = gIter->next) { pe_node_t *node_iter = (pe_node_t *) gIter->data; resource_node_score(rsc, node_iter, score, tag); } } else { GHashTableIter iter; pe_node_t *node_iter = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) { resource_node_score(rsc, node_iter, score, tag); } } if (node == NULL && score == -INFINITY) { if (rsc->allocated_to) { crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname); free(rsc->allocated_to); rsc->allocated_to = NULL; } } } #define sort_return(an_int, why) do { \ free(a_uuid); \ free(b_uuid); \ crm_trace("%s (%d) %c %s (%d) : %s", \ a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \ b_xml_id, b_call_id, why); \ return an_int; \ } while(0) int pe__is_newer_op(const xmlNode *xml_a, const xmlNode *xml_b) { int a_call_id = -1; int b_call_id = -1; char *a_uuid = NULL; char *b_uuid = NULL; const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID); const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID); const char *a_node = crm_element_value(xml_a, XML_LRM_ATTR_TARGET); const char *b_node = crm_element_value(xml_b, XML_LRM_ATTR_TARGET); bool same_node = pcmk__str_eq(a_node, b_node, pcmk__str_casei); if (same_node && pcmk__str_eq(a_xml_id, b_xml_id, pcmk__str_none)) { /* We have duplicate lrm_rsc_op entries in the status * section which is unlikely to be a good thing * - we can handle it easily enough, but we need to get * to the bottom of why it's happening. */ pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id); sort_return(0, "duplicate"); } crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id); crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id); if (a_call_id == -1 && b_call_id == -1) { /* both are pending ops so it doesn't matter since * stops are never pending */ sort_return(0, "pending"); } else if (same_node && a_call_id >= 0 && a_call_id < b_call_id) { sort_return(-1, "call id"); } else if (same_node && b_call_id >= 0 && a_call_id > b_call_id) { sort_return(1, "call id"); } else if (a_call_id >= 0 && b_call_id >= 0 && (!same_node || a_call_id == b_call_id)) { /* * The op and last_failed_op are the same * Order on last-rc-change */ time_t last_a = -1; time_t last_b = -1; crm_element_value_epoch(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a); crm_element_value_epoch(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b); crm_trace("rc-change: %lld vs %lld", (long long) last_a, (long long) last_b); if (last_a >= 0 && last_a < last_b) { sort_return(-1, "rc-change"); } else if (last_b >= 0 && last_a > last_b) { sort_return(1, "rc-change"); } sort_return(0, "rc-change"); } else { /* One of the inputs is a pending operation * Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other */ int a_id = -1; int b_id = -1; const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC); const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC); CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic")); if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL, NULL)) { sort_return(0, "bad magic a"); } if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL, NULL)) { sort_return(0, "bad magic b"); } /* try to determine the relative age of the operation... * some pending operations (e.g. a start) may have been superseded * by a subsequent stop * * [a|b]_id == -1 means it's a shutdown operation and _always_ comes last */ if (!pcmk__str_eq(a_uuid, b_uuid, pcmk__str_casei) || a_id == b_id) { /* * some of the logic in here may be redundant... * * if the UUID from the TE doesn't match then one better * be a pending operation. * pending operations don't survive between elections and joins * because we query the LRM directly */ if (b_call_id == -1) { sort_return(-1, "transition + call"); } else if (a_call_id == -1) { sort_return(1, "transition + call"); } } else if ((a_id >= 0 && a_id < b_id) || b_id == -1) { sort_return(-1, "transition"); } else if ((b_id >= 0 && a_id > b_id) || a_id == -1) { sort_return(1, "transition"); } } /* we should never end up here */ CRM_CHECK(FALSE, sort_return(0, "default")); } gint sort_op_by_callid(gconstpointer a, gconstpointer b) { const xmlNode *xml_a = a; const xmlNode *xml_b = b; return pe__is_newer_op(xml_a, xml_b); } time_t get_effective_time(pe_working_set_t * data_set) { if(data_set) { if (data_set->now == NULL) { crm_trace("Recording a new 'now'"); data_set->now = crm_time_new(NULL); } return crm_time_get_seconds_since_epoch(data_set->now); } crm_trace("Defaulting to 'now'"); return time(NULL); } gboolean get_target_role(pe_resource_t * rsc, enum rsc_role_e * role) { enum rsc_role_e local_role = RSC_ROLE_UNKNOWN; const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); CRM_CHECK(role != NULL, return FALSE); if (pcmk__str_eq(value, "started", pcmk__str_null_matches | pcmk__str_casei) || pcmk__str_eq("default", value, pcmk__str_casei)) { return FALSE; } local_role = text2role(value); if (local_role == RSC_ROLE_UNKNOWN) { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' is not valid", rsc->id, value); return FALSE; } else if (local_role > RSC_ROLE_STARTED) { if (pcmk_is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) { if (local_role > RSC_ROLE_UNPROMOTED) { /* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */ return FALSE; } } else { pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s " "because '%s' only makes sense for promotable " "clones", rsc->id, value); return FALSE; } } *role = local_role; return TRUE; } gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order) { GList *gIter = NULL; pe_action_wrapper_t *wrapper = NULL; GList *list = NULL; if (order == pe_order_none) { return FALSE; } if (lh_action == NULL || rh_action == NULL) { return FALSE; } crm_trace("Creating action wrappers for ordering: %s then %s", lh_action->uuid, rh_action->uuid); /* Ensure we never create a dependency on ourselves... it's happened */ CRM_ASSERT(lh_action != rh_action); /* Filter dups, otherwise update_action_states() has too much work to do */ gIter = lh_action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data; if (after->action == rh_action && (after->type & order)) { return FALSE; } } wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = rh_action; wrapper->type = order; list = lh_action->actions_after; list = g_list_prepend(list, wrapper); lh_action->actions_after = list; wrapper = calloc(1, sizeof(pe_action_wrapper_t)); wrapper->action = lh_action; wrapper->type = order; list = rh_action->actions_before; list = g_list_prepend(list, wrapper); rh_action->actions_before = list; return TRUE; } pe_action_t * get_pseudo_op(const char *name, pe_working_set_t * data_set) { pe_action_t *op = lookup_singleton(data_set, name); if (op == NULL) { op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set); pe__set_action_flags(op, pe_action_pseudo|pe_action_runnable); } return op; } void destroy_ticket(gpointer data) { pe_ticket_t *ticket = data; if (ticket->state) { g_hash_table_destroy(ticket->state); } free(ticket->id); free(ticket); } pe_ticket_t * ticket_new(const char *ticket_id, pe_working_set_t * data_set) { pe_ticket_t *ticket = NULL; if (pcmk__str_empty(ticket_id)) { return NULL; } if (data_set->tickets == NULL) { data_set->tickets = pcmk__strkey_table(free, destroy_ticket); } ticket = g_hash_table_lookup(data_set->tickets, ticket_id); if (ticket == NULL) { ticket = calloc(1, sizeof(pe_ticket_t)); if (ticket == NULL) { crm_err("Cannot allocate ticket '%s'", ticket_id); return NULL; } crm_trace("Creaing ticket entry for %s", ticket_id); ticket->id = strdup(ticket_id); ticket->granted = FALSE; ticket->last_granted = -1; ticket->standby = FALSE; ticket->state = pcmk__strkey_table(free, free); g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket); } return ticket; } const char *rsc_printable_id(pe_resource_t *rsc) { if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { return ID(rsc->xml); } return rsc->id; } void pe__clear_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__clear_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__clear_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } void pe__clear_resource_flags_on_all(pe_working_set_t *data_set, uint64_t flag) { for (GList *lpc = data_set->resources; lpc != NULL; lpc = lpc->next) { pe_resource_t *r = (pe_resource_t *) lpc->data; pe__clear_resource_flags_recursive(r, flag); } } void pe__set_resource_flags_recursive(pe_resource_t *rsc, uint64_t flags) { pe__set_resource_flags(rsc, flags); for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) { pe__set_resource_flags_recursive((pe_resource_t *) gIter->data, flags); } } static GList * find_unfencing_devices(GList *candidates, GList *matches) { for (GList *gIter = candidates; gIter != NULL; gIter = gIter->next) { pe_resource_t *candidate = gIter->data; const char *provides = g_hash_table_lookup(candidate->meta, PCMK_STONITH_PROVIDES); const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES); if(candidate->children) { matches = find_unfencing_devices(candidate->children, matches); } else if (!pcmk_is_set(candidate->flags, pe_rsc_fence_device)) { continue; - } else if (pcmk__str_eq(provides, "unfencing", pcmk__str_casei) || pcmk__str_eq(requires, "unfencing", pcmk__str_casei)) { + } else if (pcmk__str_eq(provides, PCMK__VALUE_UNFENCING, + pcmk__str_casei) + || pcmk__str_eq(requires, PCMK__VALUE_UNFENCING, + pcmk__str_casei)) { matches = g_list_prepend(matches, candidate); } } return matches; } static int node_priority_fencing_delay(pe_node_t * node, pe_working_set_t * data_set) { int member_count = 0; int online_count = 0; int top_priority = 0; int lowest_priority = 0; GList *gIter = NULL; // `priority-fencing-delay` is disabled if (data_set->priority_fencing_delay <= 0) { return 0; } /* No need to request a delay if the fencing target is not a normal cluster * member, for example if it's a remote node or a guest node. */ if (node->details->type != node_member) { return 0; } // No need to request a delay if the fencing target is in our partition if (node->details->online) { return 0; } for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) { pe_node_t *n = gIter->data; if (n->details->type != node_member) { continue; } member_count ++; if (n->details->online) { online_count++; } if (member_count == 1 || n->details->priority > top_priority) { top_priority = n->details->priority; } if (member_count == 1 || n->details->priority < lowest_priority) { lowest_priority = n->details->priority; } } // No need to delay if we have more than half of the cluster members if (online_count > member_count / 2) { return 0; } /* All the nodes have equal priority. * Any configured corresponding `pcmk_delay_base/max` will be applied. */ if (lowest_priority == top_priority) { return 0; } if (node->details->priority < top_priority) { return 0; } return data_set->priority_fencing_delay; } pe_action_t * pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t * data_set) { char *op_key = NULL; pe_action_t *stonith_op = NULL; if(op == NULL) { op = data_set->stonith_action; } op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op); stonith_op = lookup_singleton(data_set, op_key); if(stonith_op == NULL) { stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname); add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id); add_hash_param(stonith_op->meta, "stonith_action", op); if (pe__is_guest_or_remote_node(node) && pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { /* Extra work to detect device changes on remotes * * We may do this for all nodes in the future, but for now * the pcmk__check_action_config() based stuff works fine. */ long max = 1024; long digests_all_offset = 0; long digests_secure_offset = 0; char *digests_all = calloc(max, sizeof(char)); char *digests_secure = calloc(max, sizeof(char)); GList *matches = find_unfencing_devices(data_set->resources, NULL); for (GList *gIter = matches; gIter != NULL; gIter = gIter->next) { pe_resource_t *match = gIter->data; const char *agent = g_hash_table_lookup(match->meta, XML_ATTR_TYPE); op_digest_cache_t *data = NULL; data = pe__compare_fencing_digest(match, agent, node, data_set); if(data->rc == RSC_DIGEST_ALL) { optional = FALSE; crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id); if (!pcmk__is_daemon && data_set->priv != NULL) { pcmk__output_t *out = data_set->priv; out->info(out, "notice: Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id); } } digests_all_offset += snprintf( digests_all+digests_all_offset, max-digests_all_offset, "%s:%s:%s,", match->id, agent, data->digest_all_calc); digests_secure_offset += snprintf( digests_secure+digests_secure_offset, max-digests_secure_offset, "%s:%s:%s,", match->id, agent, data->digest_secure_calc); } g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_ALL), digests_all); g_hash_table_insert(stonith_op->meta, strdup(XML_OP_ATTR_DIGESTS_SECURE), digests_secure); } } else { free(op_key); } if (data_set->priority_fencing_delay > 0 /* It's a suitable case where `priority-fencing-delay` applies. * At least add `priority-fencing-delay` field as an indicator. */ && (priority_delay /* The priority delay needs to be recalculated if this function has * been called by schedule_fencing_and_shutdowns() after node * priority has already been calculated by native_add_running(). */ || g_hash_table_lookup(stonith_op->meta, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY) != NULL)) { /* Add `priority-fencing-delay` to the fencing op even if it's 0 for * the targeting node. So that it takes precedence over any possible * `pcmk_delay_base/max`. */ char *delay_s = pcmk__itoa(node_priority_fencing_delay(node, data_set)); g_hash_table_insert(stonith_op->meta, strdup(XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY), delay_s); } if(optional == FALSE && pe_can_fence(data_set, node)) { pe__clear_action_flags(stonith_op, pe_action_optional); pe_action_set_reason(stonith_op, reason, false); } else if(reason && stonith_op->reason == NULL) { stonith_op->reason = strdup(reason); } return stonith_op; } void trigger_unfencing( pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set) { if (!pcmk_is_set(data_set->flags, pe_flag_enable_unfencing)) { /* No resources require it */ return; } else if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pe_rsc_fence_device)) { /* Wasn't a stonith device */ return; } else if(node && node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set); if(dependency) { order_actions(unfence, dependency, pe_order_optional); } } else if(rsc) { GHashTableIter iter; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) { trigger_unfencing(rsc, node, reason, dependency, data_set); } } } } gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref) { pe_tag_t *tag = NULL; GList *gIter = NULL; gboolean is_existing = FALSE; CRM_CHECK(tags && tag_name && obj_ref, return FALSE); tag = g_hash_table_lookup(tags, tag_name); if (tag == NULL) { tag = calloc(1, sizeof(pe_tag_t)); if (tag == NULL) { return FALSE; } tag->id = strdup(tag_name); tag->refs = NULL; g_hash_table_insert(tags, strdup(tag_name), tag); } for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) { const char *existing_ref = (const char *) gIter->data; if (pcmk__str_eq(existing_ref, obj_ref, pcmk__str_none)){ is_existing = TRUE; break; } } if (is_existing == FALSE) { tag->refs = g_list_append(tag->refs, strdup(obj_ref)); crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref); } return TRUE; } /*! * \internal * \brief Create an action reason string based on the action itself * * \param[in] action Action to create reason string for * \param[in] flag Action flag that was cleared * * \return Newly allocated string suitable for use as action reason * \note It is the caller's responsibility to free() the result. */ char * pe__action2reason(pe_action_t *action, enum pe_action_flags flag) { const char *change = NULL; switch (flag) { case pe_action_runnable: case pe_action_migrate_runnable: change = "unrunnable"; break; case pe_action_optional: change = "required"; break; default: // Bug: caller passed unsupported flag CRM_CHECK(change != NULL, change = ""); break; } return crm_strdup_printf("%s%s%s %s", change, (action->rsc == NULL)? "" : " ", (action->rsc == NULL)? "" : action->rsc->id, action->task); } void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite) { if (action->reason != NULL && overwrite) { pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'", action->uuid, action->reason, pcmk__s(reason, "(none)")); } else if (action->reason == NULL) { pe_rsc_trace(action->rsc, "Set %s reason to '%s'", action->uuid, pcmk__s(reason, "(none)")); } else { // crm_assert(action->reason != NULL && !overwrite); return; } pcmk__str_update(&action->reason, reason); } /*! * \internal * \brief Check whether shutdown has been requested for a node * * \param[in] node Node to check * * \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise * \note This differs from simply using node->details->shutdown in that it can * be used before that has been determined (and in fact to determine it), * and it can also be used to distinguish requested shutdown from implicit * shutdown of remote nodes by virtue of their connection stopping. */ bool pe__shutdown_requested(pe_node_t *node) { const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN); return !pcmk__str_eq(shutdown, "0", pcmk__str_null_matches); } /*! * \internal * \brief Update a data set's "recheck by" time * * \param[in] recheck Epoch time when recheck should happen * \param[in,out] data_set Current working set */ void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set) { if ((recheck > get_effective_time(data_set)) && ((data_set->recheck_by == 0) || (data_set->recheck_by > recheck))) { data_set->recheck_by = recheck; } } /*! * \internal * \brief Wrapper for pe_unpack_nvpairs() using a cluster working set */ void pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name, pe_rule_eval_data_t *rule_data, GHashTable *hash, const char *always_first, gboolean overwrite, pe_working_set_t *data_set) { crm_time_t *next_change = crm_time_new_undefined(); pe_eval_nvpairs(data_set->input, xml_obj, set_name, rule_data, hash, always_first, overwrite, next_change); if (crm_time_is_defined(next_change)) { time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change); pe__update_recheck_time(recheck, data_set); } crm_time_free(next_change); } bool pe__resource_is_disabled(pe_resource_t *rsc) { const char *target_role = NULL; CRM_CHECK(rsc != NULL, return false); target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE); if (target_role) { enum rsc_role_e target_role_e = text2role(target_role); if ((target_role_e == RSC_ROLE_STOPPED) || ((target_role_e == RSC_ROLE_UNPROMOTED) && pcmk_is_set(uber_parent(rsc)->flags, pe_rsc_promotable))) { return true; } } return false; } /*! * \internal * \brief Create an action to clear a resource's history from CIB * * \param[in] rsc Resource to clear * \param[in] node Node to clear history on * * \return New action to clear resource history */ pe_action_t * pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set) { char *key = NULL; CRM_ASSERT(rsc && node); key = pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0); return custom_action(rsc, key, CRM_OP_LRM_DELETE, node, FALSE, TRUE, data_set); } bool pe__rsc_running_on_any(pe_resource_t *rsc, GList *node_list) { for (GList *ele = rsc->running_on; ele; ele = ele->next) { pe_node_t *node = (pe_node_t *) ele->data; if (pcmk__str_in_list(node->details->uname, node_list, pcmk__str_star_matches|pcmk__str_casei)) { return true; } } return false; } bool pcmk__rsc_filtered_by_node(pe_resource_t *rsc, GList *only_node) { return (rsc->fns->active(rsc, FALSE) && !pe__rsc_running_on_any(rsc, only_node)); } GList * pe__filter_rsc_list(GList *rscs, GList *filter) { GList *retval = NULL; for (GList *gIter = rscs; gIter; gIter = gIter->next) { pe_resource_t *rsc = (pe_resource_t *) gIter->data; /* I think the second condition is safe here for all callers of this * function. If not, it needs to move into pe__node_text. */ if (pcmk__str_in_list(rsc_printable_id(rsc), filter, pcmk__str_star_matches) || (rsc->parent && pcmk__str_in_list(rsc_printable_id(rsc->parent), filter, pcmk__str_star_matches))) { retval = g_list_prepend(retval, rsc); } } return retval; } GList * pe__build_node_name_list(pe_working_set_t *data_set, const char *s) { GList *nodes = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { /* Nothing was given so return a list of all node names. Or, '*' was * given. This would normally fall into the pe__unames_with_tag branch * where it will return an empty list. Catch it here instead. */ nodes = g_list_prepend(nodes, strdup("*")); } else { pe_node_t *node = pe_find_node(data_set->nodes, s); if (node) { /* The given string was a valid uname for a node. Return a * singleton list containing just that uname. */ nodes = g_list_prepend(nodes, strdup(s)); } else { /* The given string was not a valid uname. It's either a tag or * it's a typo or something. In the first case, we'll return a * list of all the unames of the nodes with the given tag. In the * second case, we'll return a NULL pointer and nothing will * get displayed. */ nodes = pe__unames_with_tag(data_set, s); } } return nodes; } GList * pe__build_rsc_list(pe_working_set_t *data_set, const char *s) { GList *resources = NULL; if (pcmk__str_eq(s, "*", pcmk__str_null_matches)) { resources = g_list_prepend(resources, strdup("*")); } else { pe_resource_t *rsc = pe_find_resource_with_flags(data_set->resources, s, pe_find_renamed|pe_find_any); if (rsc) { /* A colon in the name we were given means we're being asked to filter * on a specific instance of a cloned resource. Put that exact string * into the filter list. Otherwise, use the printable ID of whatever * resource was found that matches what was asked for. */ if (strstr(s, ":") != NULL) { resources = g_list_prepend(resources, strdup(rsc->id)); } else { resources = g_list_prepend(resources, strdup(rsc_printable_id(rsc))); } } else { /* The given string was not a valid resource name. It's either * a tag or it's a typo or something. See build_uname_list for * more detail. */ resources = pe__rscs_with_tag(data_set, s); } } return resources; } xmlNode * pe__failed_probe_for_rsc(pe_resource_t *rsc, const char *name) { pe_resource_t *parent = uber_parent(rsc); const char *rsc_id = rsc->id; if (rsc->variant == pe_clone) { rsc_id = pe__clone_child_id(rsc); } else if (parent->variant == pe_clone) { rsc_id = pe__clone_child_id(parent); } for (xmlNode *xml_op = pcmk__xml_first_child(rsc->cluster->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { const char *value = NULL; char *op_id = NULL; /* This resource operation is not a failed probe. */ if (!pcmk_xe_mask_probe_failure(xml_op)) { continue; } /* This resource operation was not run on the given node. Note that if name is * NULL, this will always succeed. */ value = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if (value == NULL || !pcmk__str_eq(value, name, pcmk__str_casei|pcmk__str_null_matches)) { continue; } /* This resource operation has no operation_key. */ value = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY); if (!parse_op_key(value ? value : ID(xml_op), &op_id, NULL, NULL)) { continue; } /* This resource operation's ID does not match the rsc_id we are looking for. */ if (!pcmk__str_eq(op_id, rsc_id, pcmk__str_none)) { free(op_id); continue; } free(op_id); return xml_op; } return NULL; } diff --git a/tools/crm_mon.c b/tools/crm_mon.c index eaf79bd99f..134e76b260 100644 --- a/tools/crm_mon.c +++ b/tools/crm_mon.c @@ -1,2202 +1,2205 @@ /* * Copyright 2004-2022 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include // pcmk__ends_with_ext() #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "crm_mon.h" #define SUMMARY "Provides a summary of cluster's current state.\n\n" \ "Outputs varying levels of detail in a number of different formats." /* * Definitions indicating which items to print */ static uint32_t show; static uint32_t show_opts = pcmk_show_pending; /* * Definitions indicating how to output */ static mon_output_format_t output_format = mon_output_unset; /* other globals */ static GIOChannel *io_channel = NULL; static GMainLoop *mainloop = NULL; static guint reconnect_timer = 0; static mainloop_timer_t *refresh_timer = NULL; static cib_t *cib = NULL; static stonith_t *st = NULL; static xmlNode *current_cib = NULL; static GError *error = NULL; static pcmk__common_args_t *args = NULL; static pcmk__output_t *out = NULL; static GOptionContext *context = NULL; static gchar **processed_args = NULL; static time_t last_refresh = 0; volatile crm_trigger_t *refresh_trigger = NULL; static enum pcmk__fence_history fence_history = pcmk__fence_history_none; static gboolean on_remote_node = FALSE; static gboolean use_cib_native = FALSE; int interactive_fence_level = 0; static pcmk__supported_format_t formats[] = { #if CURSES_ENABLED CRM_MON_SUPPORTED_FORMAT_CURSES, #endif PCMK__SUPPORTED_FORMAT_HTML, PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, 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 #define RECONNECT_MSECS 5000 struct { guint reconnect_ms; gboolean daemonize; gboolean fence_connect; gboolean one_shot; gboolean print_pending; gboolean show_bans; gboolean watch_fencing; char *pid_file; char *external_agent; char *external_recipient; char *neg_location_prefix; char *only_node; char *only_rsc; GSList *user_includes_excludes; GSList *includes_excludes; } options = { .fence_connect = TRUE, .reconnect_ms = RECONNECT_MSECS }; static void clean_up_fencing_connection(void); static crm_exit_t clean_up(crm_exit_t exit_code); static void crm_diff_update(const char *event, xmlNode * msg); static void clean_up_on_connection_failure(int rc); static int mon_refresh_display(gpointer user_data); static int cib_connect(void); static int fencing_connect(void); static int pacemakerd_status(void); 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 refresh_after_event(gboolean data_updated, gboolean enforce); static uint32_t all_includes(mon_output_format_t fmt) { if (fmt == mon_output_monitor || fmt == mon_output_plain || fmt == mon_output_console) { return ~pcmk_section_options; } else { return pcmk_section_all; } } static uint32_t default_includes(mon_output_format_t fmt) { switch (fmt) { case mon_output_monitor: case mon_output_plain: case mon_output_console: return pcmk_section_summary | pcmk_section_nodes | pcmk_section_resources | pcmk_section_failures; case mon_output_xml: case mon_output_legacy_xml: return all_includes(fmt); case mon_output_html: case mon_output_cgi: return pcmk_section_summary | pcmk_section_nodes | pcmk_section_resources | pcmk_section_failures; default: return 0; } } struct { const char *name; uint32_t bit; } sections[] = { { "attributes", pcmk_section_attributes }, { "bans", pcmk_section_bans }, { "counts", pcmk_section_counts }, { "dc", pcmk_section_dc }, { "failcounts", pcmk_section_failcounts }, { "failures", pcmk_section_failures }, - { "fencing", pcmk_section_fencing_all }, + { PCMK__VALUE_FENCING, pcmk_section_fencing_all }, { "fencing-failed", pcmk_section_fence_failed }, { "fencing-pending", pcmk_section_fence_pending }, { "fencing-succeeded", pcmk_section_fence_worked }, { "maint-mode", pcmk_section_maint_mode }, { "nodes", pcmk_section_nodes }, { "operations", pcmk_section_operations }, { "options", pcmk_section_options }, { "resources", pcmk_section_resources }, { "stack", pcmk_section_stack }, { "summary", pcmk_section_summary }, { "tickets", pcmk_section_tickets }, { "times", pcmk_section_times }, { NULL } }; static uint32_t find_section_bit(const char *name) { for (int i = 0; sections[i].name != NULL; i++) { if (pcmk__str_eq(sections[i].name, name, pcmk__str_casei)) { return sections[i].bit; } } return 0; } static gboolean apply_exclude(const gchar *excludes, GError **error) { char **parts = NULL; gboolean result = TRUE; parts = g_strsplit(excludes, ",", 0); for (char **s = parts; *s != NULL; s++) { uint32_t bit = find_section_bit(*s); if (pcmk__str_eq(*s, "all", pcmk__str_none)) { show = 0; } else if (pcmk__str_eq(*s, PCMK__VALUE_NONE, pcmk__str_none)) { show = all_includes(output_format); } else if (bit != 0) { show &= ~bit; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--exclude options: all, attributes, bans, counts, dc, " "failcounts, failures, fencing, fencing-failed, " "fencing-pending, fencing-succeeded, maint-mode, nodes, " PCMK__VALUE_NONE ", operations, options, resources, " "stack, summary, tickets, times"); result = FALSE; break; } } g_strfreev(parts); return result; } static gboolean apply_include(const gchar *includes, GError **error) { char **parts = NULL; gboolean result = TRUE; parts = g_strsplit(includes, ",", 0); for (char **s = parts; *s != NULL; s++) { uint32_t bit = find_section_bit(*s); if (pcmk__str_eq(*s, "all", pcmk__str_none)) { show = all_includes(output_format); } else if (pcmk__starts_with(*s, "bans")) { show |= pcmk_section_bans; if (options.neg_location_prefix != NULL) { free(options.neg_location_prefix); options.neg_location_prefix = NULL; } if (strlen(*s) > 4 && (*s)[4] == ':') { options.neg_location_prefix = strdup(*s+5); } } else if (pcmk__str_any_of(*s, "default", "defaults", NULL)) { show |= default_includes(output_format); } else if (pcmk__str_eq(*s, PCMK__VALUE_NONE, pcmk__str_none)) { show = 0; } else if (bit != 0) { show |= bit; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--include options: all, attributes, bans[:PREFIX], counts, dc, " "default, failcounts, failures, fencing, fencing-failed, " "fencing-pending, fencing-succeeded, maint-mode, nodes, " PCMK__VALUE_NONE ", operations, options, resources, " "stack, summary, tickets, times"); result = FALSE; break; } } g_strfreev(parts); return result; } static gboolean apply_include_exclude(GSList *lst, GError **error) { gboolean rc = TRUE; GSList *node = lst; while (node != NULL) { char *s = node->data; if (pcmk__starts_with(s, "--include=")) { rc = apply_include(s+10, error); } else if (pcmk__starts_with(s, "-I=")) { rc = apply_include(s+3, error); } else if (pcmk__starts_with(s, "--exclude=")) { rc = apply_exclude(s+10, error); } else if (pcmk__starts_with(s, "-U=")) { rc = apply_exclude(s+3, error); } if (rc != TRUE) { break; } node = node->next; } return rc; } static gboolean user_include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { char *s = crm_strdup_printf("%s=%s", option_name, optarg); options.user_includes_excludes = g_slist_append(options.user_includes_excludes, s); return TRUE; } static gboolean include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { char *s = crm_strdup_printf("%s=%s", option_name, optarg); options.includes_excludes = g_slist_append(options.includes_excludes, s); return TRUE; } static gboolean as_cgi_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&args->output_ty, "html"); output_format = mon_output_cgi; options.one_shot = TRUE; return TRUE; } static gboolean as_html_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&args->output_dest, optarg); pcmk__str_update(&args->output_ty, "html"); output_format = mon_output_html; umask(S_IWGRP | S_IWOTH); return TRUE; } static gboolean as_simple_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_monitor; options.one_shot = TRUE; return TRUE; } static gboolean as_xml_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { pcmk__str_update(&args->output_ty, "xml"); output_format = mon_output_legacy_xml; return TRUE; } static gboolean fence_history_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (optarg == NULL) { interactive_fence_level = 2; } else { pcmk__scan_min_int(optarg, &interactive_fence_level, 0); } switch (interactive_fence_level) { case 3: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; - return include_exclude_cb("--include", "fencing", data, err); + return include_exclude_cb("--include", PCMK__VALUE_FENCING, data, + err); case 2: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; - return include_exclude_cb("--include", "fencing", data, err); + return include_exclude_cb("--include", PCMK__VALUE_FENCING, data, + err); case 1: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; return include_exclude_cb("--include", "fencing-failed,fencing-pending", data, err); case 0: options.fence_connect = FALSE; fence_history = pcmk__fence_history_none; - return include_exclude_cb("--exclude", "fencing", data, err); + return include_exclude_cb("--exclude", PCMK__VALUE_FENCING, data, + err); default: g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Fence history must be 0-3"); return FALSE; } } static gboolean group_by_node_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_rscs_by_node; return TRUE; } static gboolean hide_headers_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--exclude", "summary", data, err); } static gboolean inactive_resources_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_inactive_rscs; return TRUE; } static gboolean no_curses_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { output_format = mon_output_plain; return TRUE; } static gboolean print_brief_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_brief; return TRUE; } static gboolean print_detail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_details; return TRUE; } static gboolean print_timing_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_timing; return user_include_exclude_cb("--include", "operations", data, err); } static gboolean reconnect_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { int rc = crm_get_msec(optarg); if (rc == -1) { g_set_error(err, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Invalid value for -i: %s", optarg); return FALSE; } else { options.reconnect_ms = crm_parse_interval_spec(optarg); } return TRUE; } static gboolean show_attributes_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "attributes", data, err); } static gboolean show_bans_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { if (optarg != NULL) { char *s = crm_strdup_printf("bans:%s", optarg); gboolean rc = user_include_exclude_cb("--include", s, data, err); free(s); return rc; } else { return user_include_exclude_cb("--include", "bans", data, err); } } static gboolean show_failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "failcounts", data, err); } static gboolean show_operations_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "failcounts,operations", data, err); } static gboolean show_tickets_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { return user_include_exclude_cb("--include", "tickets", data, err); } static gboolean use_cib_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { setenv("CIB_file", optarg, 1); options.one_shot = TRUE; return TRUE; } #define INDENT " " /* *INDENT-OFF* */ static GOptionEntry addl_entries[] = { { "interval", 'i', 0, G_OPTION_ARG_CALLBACK, reconnect_cb, "Update frequency (default is 5 seconds)", "TIMESPEC" }, { "one-shot", '1', 0, G_OPTION_ARG_NONE, &options.one_shot, "Display the cluster status once on the console and exit", NULL }, { "daemonize", 'd', 0, G_OPTION_ARG_NONE, &options.daemonize, "Run in the background as a daemon.\n" INDENT "Requires at least one of --output-to and --external-agent.", NULL }, { "pid-file", 'p', 0, G_OPTION_ARG_FILENAME, &options.pid_file, "(Advanced) Daemon pid file location", "FILE" }, { "external-agent", 'E', 0, G_OPTION_ARG_FILENAME, &options.external_agent, "A program to run when resource operations take place", "FILE" }, { "external-recipient", 'e', 0, G_OPTION_ARG_STRING, &options.external_recipient, "A recipient for your program (assuming you want the program to send something to someone).", "RCPT" }, { "watch-fencing", 'W', 0, G_OPTION_ARG_NONE, &options.watch_fencing, "Listen for fencing events. For use with --external-agent.", NULL }, { "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, use_cib_file_cb, NULL, NULL }, { NULL } }; static GOptionEntry display_entries[] = { { "include", 'I', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb, "A list of sections to include in the output.\n" INDENT "See `Output Control` help for more information.", "SECTION(s)" }, { "exclude", 'U', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb, "A list of sections to exclude from the output.\n" INDENT "See `Output Control` help for more information.", "SECTION(s)" }, { "node", 0, 0, G_OPTION_ARG_STRING, &options.only_node, "When displaying information about nodes, show only what's related to the given\n" INDENT "node, or to all nodes tagged with the given tag", "NODE" }, { "resource", 0, 0, G_OPTION_ARG_STRING, &options.only_rsc, "When displaying information about resources, show only what's related to the given\n" INDENT "resource, or to all resources tagged with the given tag", "RSC" }, { "group-by-node", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, group_by_node_cb, "Group resources by node", NULL }, { "inactive", 'r', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, inactive_resources_cb, "Display inactive resources", NULL }, { "failcounts", 'f', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_failcounts_cb, "Display resource fail counts", NULL }, { "operations", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_operations_cb, "Display resource operation history", NULL }, { "timing-details", 't', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_timing_cb, "Display resource operation history with timing details", NULL }, { "tickets", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_tickets_cb, "Display cluster tickets", NULL }, { "fence-history", 'm', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, fence_history_cb, "Show fence history:\n" INDENT "0=off, 1=failures and pending (default without option),\n" INDENT "2=add successes (default without value for option),\n" INDENT "3=show full history without reduction to most recent of each flavor", "LEVEL" }, { "neg-locations", 'L', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, show_bans_cb, "Display negative location constraints [optionally filtered by id prefix]", NULL }, { "show-node-attributes", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_attributes_cb, "Display node attributes", NULL }, { "hide-headers", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, hide_headers_cb, "Hide all headers", NULL }, { "show-detail", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_detail_cb, "Show more details (node IDs, individual clone instances)", NULL }, { "brief", 'b', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_brief_cb, "Brief output", NULL }, { "pending", 'j', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.print_pending, "Display pending state if 'record-pending' is enabled", NULL }, { "simple-status", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_simple_cb, "Display the cluster status once as a simple one line output (suitable for nagios)", NULL }, { NULL } }; static GOptionEntry deprecated_entries[] = { { "as-html", 'h', G_OPTION_FLAG_FILENAME, G_OPTION_ARG_CALLBACK, as_html_cb, "Write cluster status to the named HTML file.\n" INDENT "Use --output-as=html --output-to=FILE instead.", "FILE" }, { "as-xml", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_xml_cb, "Write cluster status as XML to stdout. This will enable one-shot mode.\n" INDENT "Use --output-as=xml instead.", NULL }, { "disable-ncurses", 'N', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, no_curses_cb, "Disable the use of ncurses.\n" INDENT "Use --output-as=text instead.", NULL }, { "web-cgi", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_cgi_cb, "Web mode with output suitable for CGI (preselected when run as *.cgi).\n" INDENT "Use --output-as=html --html-cgi instead.", NULL }, { NULL } }; /* *INDENT-ON* */ /* Reconnect to the CIB and fencing agent after reconnect_ms has passed. This sounds * like it would be more broadly useful, but only ever happens after a disconnect via * mon_cib_connection_destroy. */ static gboolean reconnect_after_timeout(gpointer data) { #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif out->info(out, "Reconnecting..."); if (pacemakerd_status() == pcmk_rc_ok) { fencing_connect(); if (cib_connect() == pcmk_rc_ok) { /* trigger redrawing the screen (needs reconnect_timer == 0) */ reconnect_timer = 0; refresh_after_event(FALSE, TRUE); return G_SOURCE_REMOVE; } } reconnect_timer = g_timeout_add(options.reconnect_ms, reconnect_after_timeout, NULL); return G_SOURCE_REMOVE; } /* Called from various places when we are disconnected from the CIB or from the * fencing agent. If the CIB connection is still valid, this function will also * attempt to sign off and reconnect. */ static void mon_cib_connection_destroy(gpointer user_data) { out->info(out, "Connection to the cluster-daemons terminated"); if (refresh_timer != NULL) { /* we'll trigger a refresh after reconnect */ mainloop_timer_stop(refresh_timer); } if (reconnect_timer) { /* we'll trigger a new reconnect-timeout at the end */ g_source_remove(reconnect_timer); reconnect_timer = 0; } if (st) { /* the client API won't properly reconnect notifications * if they are still in the table - so remove them */ clean_up_fencing_connection(); } if (cib) { cib->cmds->signoff(cib); reconnect_timer = g_timeout_add(options.reconnect_ms, reconnect_after_timeout, NULL); } return; } /* Signal handler installed into the mainloop for normal program shutdown */ static void mon_shutdown(int nsig) { clean_up(CRM_EX_OK); } #if CURSES_ENABLED static volatile sighandler_t ncurses_winch_handler; /* Signal handler installed the regular way (not into the main loop) for when * the screen is resized. Commonly, this happens when running in an xterm and * the user changes its size. */ 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); /* Alert the mainloop code we'd like the refresh_trigger to run next * time the mainloop gets around to checking. */ mainloop_set_trigger((crm_trigger_t *) refresh_trigger); } not_done--; } #endif static int fencing_connect(void) { int rc = pcmk_ok; if (options.fence_connect && st == NULL) { st = stonith_api_new(); } if (!options.fence_connect || st == NULL || st->state != stonith_disconnected) { return rc; } rc = st->cmds->connect(st, crm_system_name, NULL); if (rc == pcmk_ok) { crm_trace("Setting up stonith callbacks"); if (options.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); } } else { clean_up_fencing_connection(); } return rc; } static int cib_connect(void) { int rc = pcmk_rc_ok; CRM_CHECK(cib != NULL, return EINVAL); if (cib->state == cib_connected_query || cib->state == cib_connected_command) { return rc; } crm_trace("Connecting to the CIB"); rc = pcmk_legacy2rc(cib->cmds->signon(cib, crm_system_name, cib_query)); if (rc != pcmk_rc_ok) { out->err(out, "Could not connect to the CIB: %s", pcmk_rc_str(rc)); return rc; } /* just show this if refresh is gonna remove all traces */ if (output_format == mon_output_console) { out->info(out,"Waiting for CIB ..."); } rc = pcmk_legacy2rc(cib->cmds->query(cib, NULL, ¤t_cib, cib_scope_local | cib_sync_call)); if (rc == pcmk_rc_ok) { rc = pcmk_legacy2rc(cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy)); if (rc == EPROTONOSUPPORT) { out->err(out, "Notification setup not supported, won't be " "able to reconnect after failure"); if (output_format == mon_output_console) { sleep(2); } rc = pcmk_rc_ok; } if (rc == pcmk_rc_ok) { cib->cmds->del_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update); rc = pcmk_legacy2rc(cib->cmds->add_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update)); } if (rc != pcmk_rc_ok) { out->err(out, "Notification setup failed, could not monitor CIB actions"); cib__clean_up_connection(&cib); clean_up_fencing_connection(); } } return rc; } /* This is used to set up the fencing options after the interactive UI has been stared. * fence_history_cb can't be used because it builds up a list of includes/excludes that * then have to be processed with apply_include_exclude and that could affect other * things. */ static void set_fencing_options(int level) { switch (level) { case 3: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fencing_all; break; case 2: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fencing_all; break; case 1: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; show |= pcmk_section_fence_failed | pcmk_section_fence_pending; break; default: interactive_fence_level = 0; options.fence_connect = FALSE; fence_history = pcmk__fence_history_none; show &= ~pcmk_section_fencing_all; break; } } /* Before trying to connect to fencer or cib check for state of pacemakerd - just no sense in trying till pacemakerd has taken care of starting all the sub-processes Only noteworthy thing to show here is when pacemakerd is waiting for startup-trigger from SBD. */ static void pacemakerd_event_cb(pcmk_ipc_api_t *pacemakerd_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk_pacemakerd_api_reply_t *reply = event_data; enum pcmk_pacemakerd_state *state = (enum pcmk_pacemakerd_state *) user_data; /* we are just interested in the latest reply */ *state = pcmk_pacemakerd_state_invalid; switch (event_type) { case pcmk_ipc_event_reply: break; default: return; } if (status != CRM_EX_OK) { out->err(out, "Bad reply from pacemakerd: %s", crm_exit_str(status)); return; } if (reply->reply_type != pcmk_pacemakerd_reply_ping) { out->err(out, "Unknown reply type %d from pacemakerd", reply->reply_type); } else { if ((reply->data.ping.last_good != (time_t) 0) && (reply->data.ping.status == pcmk_rc_ok)) { *state = reply->data.ping.state; } } } static int pacemakerd_status(void) { int rc = pcmk_rc_ok; pcmk_ipc_api_t *pacemakerd_api = NULL; enum pcmk_pacemakerd_state state = pcmk_pacemakerd_state_invalid; if (!use_cib_native) { /* we don't need fully functional pacemakerd otherwise */ return rc; } if (cib != NULL && (cib->state == cib_connected_query || cib->state == cib_connected_command)) { /* As long as we have a cib-connection let's go with * that to fetch further cluster-status and avoid * unnecessary pings to pacemakerd. * If cluster is going down and fencer is down already * this will lead to a silently failing fencer reconnect. * On cluster startup we shouldn't see this situation * as first we do is wait for pacemakerd to report all * daemons running. */ return rc; } rc = pcmk_new_ipc_api(&pacemakerd_api, pcmk_ipc_pacemakerd); if (pacemakerd_api == NULL) { out->err(out, "Could not connect to pacemakerd: %s", pcmk_rc_str(rc)); /* this is unrecoverable so return with rc we have */ return rc; } pcmk_register_ipc_callback(pacemakerd_api, pacemakerd_event_cb, (void *) &state); rc = pcmk_connect_ipc(pacemakerd_api, pcmk_ipc_dispatch_poll); switch (rc) { case pcmk_rc_ok: rc = pcmk_pacemakerd_api_ping(pacemakerd_api, crm_system_name); if (rc == pcmk_rc_ok) { rc = pcmk_poll_ipc(pacemakerd_api, options.reconnect_ms/2); if (rc == pcmk_rc_ok) { pcmk_dispatch_ipc(pacemakerd_api); rc = ENOTCONN; if ((output_format == mon_output_console) || (output_format == mon_output_plain)) { switch (state) { case pcmk_pacemakerd_state_running: rc = pcmk_rc_ok; break; case pcmk_pacemakerd_state_starting_daemons: out->info(out,"Pacemaker daemons starting ..."); break; case pcmk_pacemakerd_state_wait_for_ping: out->info(out,"Waiting for startup-trigger from SBD ..."); break; case pcmk_pacemakerd_state_shutting_down: out->info(out,"Pacemaker daemons shutting down ..."); /* try our luck maybe CIB is still accessible */ rc = pcmk_rc_ok; break; case pcmk_pacemakerd_state_shutdown_complete: /* assuming pacemakerd doesn't dispatch any pings after entering * that state unless it is waiting for SBD */ out->info(out,"Pacemaker daemons shut down - reporting to SBD ..."); break; default: break; } } else { switch (state) { case pcmk_pacemakerd_state_running: rc = pcmk_rc_ok; break; case pcmk_pacemakerd_state_shutting_down: /* try our luck maybe CIB is still accessible */ rc = pcmk_rc_ok; break; default: break; } } } } break; case EREMOTEIO: rc = pcmk_rc_ok; on_remote_node = TRUE; /* just show this if refresh is gonna remove all traces */ if (output_format == mon_output_console) { out->info(out, "Running on remote-node waiting to be connected by cluster ..."); } break; default: break; } pcmk_free_ipc_api(pacemakerd_api); /* returning with ENOTCONN triggers a retry */ return (rc == pcmk_rc_ok)?rc:ENOTCONN; } #if CURSES_ENABLED static const char * get_option_desc(char c) { const char *desc = "No help available"; for (GOptionEntry *entry = display_entries; entry != NULL; entry++) { if (entry->short_name == c) { desc = entry->description; break; } } return desc; } #define print_option_help(out, option, condition) \ curses_formatted_printf(out, "%c %c: \t%s\n", ((condition)? '*': ' '), option, get_option_desc(option)); /* This function is called from the main loop when there is something to be read * on stdin, like an interactive user's keystroke. All it does is read the keystroke, * set flags (or show the page showing which keystrokes are valid), and redraw the * screen. It does not do anything with connections to the CIB or fencing agent * agent what would happen in mon_refresh_display. */ static gboolean detect_user_input(GIOChannel *channel, GIOCondition condition, gpointer user_data) { int c; gboolean config_mode = FALSE; while (1) { /* Get user input */ c = getchar(); switch (c) { case 'm': interactive_fence_level++; if (interactive_fence_level > 3) { interactive_fence_level = 0; } set_fencing_options(interactive_fence_level); break; case 'c': show ^= pcmk_section_tickets; break; case 'f': show ^= pcmk_section_failcounts; break; case 'n': show_opts ^= pcmk_show_rscs_by_node; break; case 'o': show ^= pcmk_section_operations; if (!pcmk_is_set(show, pcmk_section_operations)) { show_opts &= ~pcmk_show_timing; } break; case 'r': show_opts ^= pcmk_show_inactive_rscs; break; case 'R': show_opts ^= pcmk_show_details; #ifdef PCMK__COMPAT_2_0 // Keep failed action output the same as 2.0.x show_opts |= pcmk_show_failed_detail; #endif break; case 't': show_opts ^= pcmk_show_timing; if (pcmk_is_set(show_opts, pcmk_show_timing)) { show |= pcmk_section_operations; } break; case 'A': show ^= pcmk_section_attributes; break; case 'L': show ^= pcmk_section_bans; break; case 'D': /* If any header is shown, clear them all, otherwise set them all */ if (pcmk_any_flags_set(show, pcmk_section_summary)) { show &= ~pcmk_section_summary; } else { show |= pcmk_section_summary; } /* Regardless, we don't show options in console mode. */ show &= ~pcmk_section_options; break; case 'b': show_opts ^= pcmk_show_brief; break; case 'j': show_opts ^= pcmk_show_pending; break; case '?': config_mode = TRUE; break; default: /* All other keys just redraw the screen. */ goto refresh; } if (!config_mode) goto refresh; blank_screen(); curses_formatted_printf(out, "%s", "Display option change mode\n"); print_option_help(out, 'c', pcmk_is_set(show, pcmk_section_tickets)); print_option_help(out, 'f', pcmk_is_set(show, pcmk_section_failcounts)); print_option_help(out, 'n', pcmk_is_set(show_opts, pcmk_show_rscs_by_node)); print_option_help(out, 'o', pcmk_is_set(show, pcmk_section_operations)); print_option_help(out, 'r', pcmk_is_set(show_opts, pcmk_show_inactive_rscs)); print_option_help(out, 't', pcmk_is_set(show_opts, pcmk_show_timing)); print_option_help(out, 'A', pcmk_is_set(show, pcmk_section_attributes)); print_option_help(out, 'L', pcmk_is_set(show, pcmk_section_bans)); print_option_help(out, 'D', !pcmk_is_set(show, pcmk_section_summary)); #ifdef PCMK__COMPAT_2_0 print_option_help(out, 'R', pcmk_any_flags_set(show_opts, pcmk_show_details & ~pcmk_show_failed_detail)); #else print_option_help(out, 'R', pcmk_any_flags_set(show_opts, pcmk_show_details)); #endif print_option_help(out, 'b', pcmk_is_set(show_opts, pcmk_show_brief)); print_option_help(out, 'j', pcmk_is_set(show_opts, pcmk_show_pending)); curses_formatted_printf(out, "%d m: \t%s\n", interactive_fence_level, get_option_desc('m')); curses_formatted_printf(out, "%s", "\nToggle fields via field letter, type any other key to return\n"); } refresh: refresh_after_event(FALSE, TRUE); return TRUE; } #endif // Basically crm_signal_handler(SIGCHLD, SIG_IGN) plus the SA_NOCLDWAIT flag static void avoid_zombies(void) { struct sigaction sa; memset(&sa, 0, sizeof(struct sigaction)); if (sigemptyset(&sa.sa_mask) < 0) { crm_warn("Cannot avoid zombies: %s", pcmk_rc_str(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_rc_str(errno)); } } static GOptionContext * build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) { GOptionContext *context = NULL; GOptionEntry extra_prog_entries[] = { { "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet), "Be less descriptive in output.", NULL }, { NULL } }; const char *description = "Notes:\n\n" "If this program is called as crm_mon.cgi, --output-as=html --html-cgi will\n" "automatically be added to the command line arguments.\n\n" "Time Specification:\n\n" "The TIMESPEC in any command line option can be specified in many different\n" "formats. It can be just an integer number of seconds, a number plus units\n" "(ms/msec/us/usec/s/sec/m/min/h/hr), or an ISO 8601 period specification.\n\n" "Output Control:\n\n" "By default, a certain list of sections are written to the output destination.\n" "The default varies based on the output format - XML includes everything, while\n" "other output formats will display less. This list can be modified with the\n" "--include and --exclude command line options. Each option may be given multiple\n" "times on the command line, and each can give a comma-separated list of sections.\n" "The options are applied to the default set, from left to right as seen on the\n" "command line. For a list of valid sections, pass --include=list or --exclude=list.\n\n" "Interactive Use:\n\n" "When run interactively, crm_mon can be told to hide and display various sections\n" "of output. To see a help screen explaining the options, hit '?'. Any key stroke\n" "aside from those listed will cause the screen to refresh.\n\n" "Examples:\n\n" "Display the cluster status on the console with updates as they occur:\n\n" "\tcrm_mon\n\n" "Display the cluster status on the console just once then exit:\n\n" "\tcrm_mon -1\n\n" "Display your cluster status, group resources by node, and include inactive resources in the list:\n\n" "\tcrm_mon --group-by-node --inactive\n\n" "Start crm_mon as a background daemon and have it write the cluster status to an HTML file:\n\n" "\tcrm_mon --daemonize --output-as html --output-to /path/to/docroot/filename.html\n\n" "Start crm_mon and export the current cluster status as XML to stdout, then exit:\n\n" "\tcrm_mon --output-as xml\n\n"; #if CURSES_ENABLED context = pcmk__build_arg_context(args, "console (default), html, text, xml", group, NULL); #else context = pcmk__build_arg_context(args, "text (default), html, xml", group, NULL); #endif pcmk__add_main_args(context, extra_prog_entries); g_option_context_set_description(context, description); pcmk__add_arg_group(context, "display", "Display Options:", "Show display options", display_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); pcmk__add_arg_group(context, "deprecated", "Deprecated Options:", "Show deprecated options", deprecated_entries); return context; } /* If certain format options were specified, we want to set some extra * options. We can just process these like they were given on the * command line. */ static void add_output_args(void) { GError *err = NULL; if (output_format == mon_output_plain) { if (!pcmk__force_args(context, &err, "%s --text-fancy", g_get_prgname())) { g_propagate_error(&error, err); clean_up(CRM_EX_USAGE); } } else if (output_format == mon_output_cgi) { if (!pcmk__force_args(context, &err, "%s --html-cgi", g_get_prgname())) { g_propagate_error(&error, err); clean_up(CRM_EX_USAGE); } } else if (output_format == mon_output_xml) { if (!pcmk__force_args(context, &err, "%s --xml-simple-list --xml-substitute", g_get_prgname())) { g_propagate_error(&error, err); clean_up(CRM_EX_USAGE); } } else if (output_format == mon_output_legacy_xml) { output_format = mon_output_xml; if (!pcmk__force_args(context, &err, "%s --xml-legacy --xml-substitute", g_get_prgname())) { g_propagate_error(&error, err); clean_up(CRM_EX_USAGE); } } } /* Which output format to use could come from two places: The --as-xml * style arguments we gave in deprecated_entries above, or the formatted output * arguments added by pcmk__register_formats. If the latter were used, * output_format will be mon_output_unset. * * Call the callbacks as if those older style arguments were provided so * the various things they do get done. */ static void reconcile_output_format(pcmk__common_args_t *args) { gboolean retval = TRUE; GError *err = NULL; if (output_format != mon_output_unset) { return; } if (pcmk__str_eq(args->output_ty, "html", pcmk__str_casei)) { char *dest = NULL; pcmk__str_update(&dest, args->output_dest); retval = as_html_cb("h", dest, NULL, &err); free(dest); } else if (pcmk__str_eq(args->output_ty, "text", pcmk__str_casei)) { retval = no_curses_cb("N", NULL, NULL, &err); } else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_casei)) { pcmk__str_update(&args->output_ty, "xml"); output_format = mon_output_xml; } else if (options.one_shot) { pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } else if (!options.daemonize && args->output_dest != NULL) { options.one_shot = TRUE; pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } else { /* Neither old nor new arguments were given, so set the default. */ pcmk__str_update(&args->output_ty, "console"); output_format = mon_output_console; } if (!retval) { g_propagate_error(&error, err); clean_up(CRM_EX_USAGE); } } static void clean_up_on_connection_failure(int rc) { if (output_format == mon_output_monitor) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "CLUSTER CRIT: Connection to cluster failed: %s", pcmk_rc_str(rc)); clean_up(MON_STATUS_CRIT); } else if (rc == ENOTCONN) { if (on_remote_node) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error: remote-node not connected to cluster"); } else { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error: cluster is not available on this node"); } } else { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Connection to cluster failed: %s", pcmk_rc_str(rc)); } clean_up(pcmk_rc2exitc(rc)); } static void one_shot(void) { int rc = pcmk__status(out, cib, fence_history, show, show_opts, options.only_node, options.only_rsc, options.neg_location_prefix, output_format == mon_output_monitor); if (rc == pcmk_rc_ok) { clean_up(pcmk_rc2exitc(rc)); } else { clean_up_on_connection_failure(rc); } } static void exit_on_invalid_cib(void) { if (cib != NULL) { return; } // Shouldn't really be possible g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Invalid CIB source"); clean_up(CRM_EX_ERROR); } int main(int argc, char **argv) { int rc = pcmk_rc_ok; GOptionGroup *output_group = NULL; args = pcmk__new_common_args(SUMMARY); context = build_arg_context(args, &output_group); pcmk__register_formats(output_group, formats); options.pid_file = strdup("/tmp/ClusterMon.pid"); pcmk__cli_init_logging("crm_mon", 0); // Avoid needing to wait for subprocesses forked for -E/--external-agent avoid_zombies(); if (pcmk__ends_with_ext(argv[0], ".cgi")) { output_format = mon_output_cgi; options.one_shot = TRUE; } processed_args = pcmk__cmdline_preproc(argv, "ehimpxEILU"); fence_history_cb("--fence-history", "1", NULL, NULL); /* Set an HTML title regardless of what format we will eventually use. This can't * be done in add_output_args. That function is called after command line * arguments are processed in the next block, which means it'll override whatever * title the user provides. Doing this here means the user can give their own * title on the command line. */ if (!pcmk__force_args(context, &error, "%s --html-title \"Cluster Status\"", g_get_prgname())) { return clean_up(CRM_EX_USAGE); } if (!g_option_context_parse_strv(context, &processed_args, &error)) { return clean_up(CRM_EX_USAGE); } for (int i = 0; i < args->verbosity; i++) { crm_bump_log_level(argc, argv); } if (!args->version) { if (args->quiet) { include_exclude_cb("--exclude", "times", NULL, NULL); } if (options.watch_fencing) { fence_history_cb("--fence-history", "0", NULL, NULL); options.fence_connect = TRUE; } /* create the cib-object early to be able to do further * decisions based on the cib-source */ cib = cib_new(); exit_on_invalid_cib(); switch (cib->variant) { case cib_native: /* cib & fencing - everything available */ use_cib_native = TRUE; 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_cb("--fence-history", "0", NULL, NULL); options.one_shot = TRUE; break; case cib_remote: /* updates coming in but no fencing */ fence_history_cb("--fence-history", "0", NULL, NULL); break; case cib_undefined: case cib_database: default: /* something is odd */ exit_on_invalid_cib(); break; } if (options.one_shot) { if (output_format == mon_output_console) { output_format = mon_output_plain; } } else if (options.daemonize) { if (pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches|pcmk__str_casei) && !options.external_agent) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires at least one of --output-to and --external-agent"); return clean_up(CRM_EX_USAGE); } } else if (output_format == mon_output_console) { #if CURSES_ENABLED crm_enable_stderr(FALSE); #else options.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 } } reconcile_output_format(args); add_output_args(); /* output_format MUST NOT BE CHANGED AFTER THIS POINT. */ if (args->version && output_format == mon_output_console) { /* Use the text output format here if we are in curses mode but were given * --version. Displaying version information uses printf, and then we * immediately exit. We don't want to initialize curses for that. */ rc = pcmk__output_new(&out, "text", args->output_dest, argv); } else { rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); } if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_ERROR, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); return clean_up(CRM_EX_ERROR); } if (options.daemonize) { if (!options.external_agent && (output_format == mon_output_console || output_format == mon_output_unset || output_format == mon_output_none)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires --output-as=[html|text|xml]"); return clean_up(CRM_EX_USAGE); } crm_enable_stderr(FALSE); cib_delete(cib); cib = NULL; pcmk__daemonize(crm_system_name, options.pid_file); cib = cib_new(); exit_on_invalid_cib(); } show = default_includes(output_format); /* Apply --include/--exclude flags we used internally. There's no error reporting * here because this would be a programming error. */ apply_include_exclude(options.includes_excludes, &error); /* And now apply any --include/--exclude flags the user gave on the command line. * These are done in a separate pass from the internal ones because we want to * make sure whatever the user specifies overrides whatever we do. */ if (!apply_include_exclude(options.user_includes_excludes, &error)) { return clean_up(CRM_EX_USAGE); } /* Sync up the initial value of interactive_fence_level with whatever was set with * --include/--exclude= options. */ if (pcmk_all_flags_set(show, pcmk_section_fencing_all)) { interactive_fence_level = 3; } else if (pcmk_is_set(show, pcmk_section_fence_worked)) { interactive_fence_level = 2; } else if (pcmk_any_flags_set(show, pcmk_section_fence_failed | pcmk_section_fence_pending)) { interactive_fence_level = 1; } else { interactive_fence_level = 0; } pcmk__register_lib_messages(out); crm_mon_register_messages(out); pe__register_messages(out); stonith__register_messages(out); if (args->version) { out->version(out, false); return clean_up(CRM_EX_OK); } /* Extra sanity checks when in CGI mode */ if (output_format == mon_output_cgi) { if (cib->variant == cib_file) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "CGI mode used with CIB file"); return clean_up(CRM_EX_USAGE); } else if (options.external_agent != NULL) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "CGI mode cannot be used with --external-agent"); return clean_up(CRM_EX_USAGE); } else if (options.daemonize == TRUE) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "CGI mode cannot be used with -d"); return clean_up(CRM_EX_USAGE); } } if (output_format == mon_output_xml || output_format == mon_output_legacy_xml) { show_opts |= pcmk_show_inactive_rscs | pcmk_show_timing; if (!options.daemonize) { options.one_shot = TRUE; } } if ((output_format == mon_output_html || output_format == mon_output_cgi) && out->dest != stdout) { pcmk__html_add_header("meta", "http-equiv", "refresh", "content", pcmk__itoa(options.reconnect_ms / 1000), NULL); } #ifdef PCMK__COMPAT_2_0 // Keep failed action output the same as 2.0.x show_opts |= pcmk_show_failed_detail; #endif crm_info("Starting %s", crm_system_name); cib__set_output(cib, out); if (options.one_shot) { one_shot(); } do { out->info(out,"Waiting until cluster is available on this node ..."); rc = pacemakerd_status(); if (rc == pcmk_rc_ok) { fencing_connect(); rc = cib_connect(); } if (rc != pcmk_rc_ok) { pcmk__sleep_ms(options.reconnect_ms); #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif } else if (output_format == mon_output_html && out->dest != stdout) { printf("Writing html to %s ...\n", args->output_dest); } } while (rc == ENOTCONN); if (rc != pcmk_rc_ok) { clean_up_on_connection_failure(rc); } set_fencing_options(interactive_fence_level); mon_refresh_display(NULL); 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; io_channel = g_io_channel_unix_new(STDIN_FILENO); g_io_add_watch(io_channel, G_IO_IN, detect_user_input, NULL); } #endif /* When refresh_trigger->trigger is set to TRUE, call mon_refresh_display. In * this file, that is anywhere mainloop_set_trigger is called. */ refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_display, NULL); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); if (io_channel != NULL) { g_io_channel_shutdown(io_channel, TRUE, NULL); } crm_info("Exiting %s", crm_system_name); return clean_up(CRM_EX_OK); } 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 = pcmk__itoa(rc); char *status_s = pcmk__itoa(status); char *target_rc_s = pcmk__itoa(target_rc); crm_debug("Sending external notification to '%s' via '%s'", options.external_recipient, options.external_agent); if(rsc) { setenv("CRM_notify_rsc", rsc, 1); } if (options.external_recipient) { setenv("CRM_notify_recipient", options.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(options.external_agent, options.external_agent, NULL); exit(CRM_EX_ERROR); } crm_trace("Finished running custom notification program '%s'.", options.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 && !pcmk__str_eq(XML_CIB_TAG_STATE, TYPE(n), pcmk__str_casei)) { 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_rc_str(pcmk_rc_ok); if ((status == PCMK_EXEC_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_EXEC_DONE) { desc = services_ocf_exitcode_str(rc); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } else { desc = pcmk_exec_status_str(status); crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc); } if (notify && options.external_agent) { send_custom_trap(node, rsc, task, target_rc, rc, status, desc); } bail: free(rsc); free(task); } /* This function is just a wrapper around mainloop_set_trigger so that it can be * called from a mainloop directly. It's simply another way of ensuring the screen * gets redrawn. */ static gboolean mon_trigger_refresh(gpointer user_data) { mainloop_set_trigger((crm_trigger_t *) refresh_trigger); return FALSE; } 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 = pcmk__xml_first_child(diff); change != NULL; change = pcmk__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 = pcmk__xe_first_child(status); state != NULL; state = pcmk__xe_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 = pcmk__xe_first_child(match); state != NULL; state = pcmk__xe_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 = pcmk__xpath_node_id(xpath, "lrm"); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RESOURCE) == 0) { char *local_node = pcmk__xpath_node_id(xpath, "lrm"); handle_rsc_op(match, local_node); free(local_node); } else if(strcmp(name, XML_LRM_TAG_RSC_OP) == 0) { char *local_node = pcmk__xpath_node_id(xpath, "lrm"); 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); out->progress(out, false); 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 (options.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) { out->info(out, "--- Stale data ---"); } stale = TRUE; return; } stale = FALSE; refresh_after_event(cib_updated, FALSE); } static int mon_refresh_display(gpointer user_data) { int rc = pcmk_rc_ok; last_refresh = time(NULL); if (output_format == mon_output_none || output_format == mon_output_unset) { return G_SOURCE_REMOVE; } if (fence_history == pcmk__fence_history_full && !pcmk_all_flags_set(show, pcmk_section_fencing_all) && output_format != mon_output_xml) { fence_history = pcmk__fence_history_reduced; } if (out->dest != stdout) { out->reset(out); } rc = pcmk__output_cluster_status(out, st, cib, current_cib, fence_history, show, show_opts, options.only_node, options.only_rsc, options.neg_location_prefix, output_format == mon_output_monitor); if (output_format == mon_output_monitor && rc != pcmk_rc_ok) { clean_up(MON_STATUS_WARN); return G_SOURCE_REMOVE; } else if (rc == pcmk_rc_schema_validation) { clean_up(CRM_EX_CONFIG); return G_SOURCE_REMOVE; } if (out->dest != stdout) { out->finish(out, CRM_EX_OK, true, NULL); } return G_SOURCE_CONTINUE; } /* This function is called for fencing events (see fencing_connect for which ones) when * --watch-fencing is used on the command line. */ 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 (options.external_agent) { char *desc = stonith__event_description(e); send_custom_trap(e->target, NULL, e->operation, pcmk_ok, e->result, 0, desc); free(desc); } } /* Cause the screen to be redrawn (via mainloop_set_trigger) when various conditions are met: * * - If the last update occurred more than reconnect_ms ago (defaults to 5s, but * can be changed via the -i command line option), or * - After every 10 CIB updates, or * - If it's been 2s since the last update * * This function sounds like it would be more broadly useful, but it is only called when a * fencing event is received or a CIB diff occurrs. */ static void refresh_after_event(gboolean data_updated, gboolean enforce) { static int updates = 0; time_t now = time(NULL); if (data_updated) { updates++; } if(refresh_timer == NULL) { refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL); } if (reconnect_timer > 0) { /* we will receive a refresh request after successful reconnect */ mainloop_timer_stop(refresh_timer); return; } /* as we're not handling initial failure of fencer-connection as * fatal give it a retry here * not getting here if cib-reconnection is already on the way */ fencing_connect(); if (enforce || ((now - last_refresh) > (options.reconnect_ms / 1000)) || updates >= 10) { mainloop_set_trigger((crm_trigger_t *) refresh_trigger); mainloop_timer_stop(refresh_timer); updates = 0; } else { mainloop_timer_start(refresh_timer); } } /* This function is called for fencing events (see fencing_connect for which ones) when * --watch-fencing is NOT used on the command line. */ 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 { out->progress(out, false); refresh_after_event(TRUE, FALSE); } } static void clean_up_fencing_connection(void) { if (st == NULL) { return; } if (st->state != stonith_disconnected) { st->cmds->remove_notification(st, NULL); 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) { /* Quitting crm_mon is much more complicated than it ought to be. */ /* (1) Close connections, free things, etc. */ cib__clean_up_connection(&cib); clean_up_fencing_connection(); free(options.neg_location_prefix); free(options.only_node); free(options.only_rsc); free(options.pid_file); g_slist_free_full(options.includes_excludes, free); g_strfreev(processed_args); /* (2) If this is abnormal termination and we're in curses mode, shut down * curses first. Any messages displayed to the screen before curses is shut * down will be lost because doing the shut down will also restore the * screen to whatever it looked like before crm_mon was started. */ if ((error != NULL || exit_code == CRM_EX_USAGE) && output_format == mon_output_console) { out->finish(out, exit_code, false, NULL); pcmk__output_free(out); out = NULL; } /* (3) If this is a command line usage related failure, print the usage * message. */ if (exit_code == CRM_EX_USAGE && (output_format == mon_output_console || output_format == mon_output_plain)) { char *help = g_option_context_get_help(context, TRUE, NULL); fprintf(stderr, "%s", help); g_free(help); } pcmk__free_arg_context(context); /* (4) If this is any kind of error, print the error out and exit. Make * sure to handle situations both before and after formatted output is * set up. We want errors to appear formatted if at all possible. */ if (error != NULL) { if (out != NULL) { out->err(out, "%s: %s", g_get_prgname(), error->message); out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } else { fprintf(stderr, "%s: %s\n", g_get_prgname(), error->message); } g_clear_error(&error); crm_exit(exit_code); } /* (5) Print formatted output to the screen if we made it far enough in * crm_mon to be able to do so. */ if (out != NULL) { if (!options.daemonize) { out->finish(out, exit_code, true, NULL); } pcmk__output_free(out); pcmk__unregister_formats(); } crm_exit(exit_code); }