diff --git a/daemons/controld/controld_cib.c b/daemons/controld/controld_cib.c index 5aab5f0d47..93477aa199 100644 --- a/daemons/controld/controld_cib.c +++ b/daemons/controld/controld_cib.c @@ -1,1067 +1,1068 @@ /* * Copyright 2004-2024 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 /* sleep */ #include #include #include #include #include #include // Call ID of the most recent in-progress CIB resource update (or 0 if none) static int pending_rsc_update = 0; /*! * \internal * \brief Respond to a dropped CIB connection * * \param[in] user_data CIB connection that dropped */ static void handle_cib_disconnect(gpointer user_data) { CRM_LOG_ASSERT(user_data == controld_globals.cib_conn); controld_trigger_fsa(); controld_globals.cib_conn->state = cib_disconnected; if (pcmk_is_set(controld_globals.fsa_input_register, R_CIB_CONNECTED)) { // @TODO This should trigger a reconnect, not a shutdown crm_crit("Lost connection to the CIB manager, shutting down"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); controld_clear_fsa_input_flags(R_CIB_CONNECTED); } else { // Expected crm_info("Disconnected from the CIB manager"); } } static void do_cib_updated(const char *event, xmlNode * msg) { const xmlNode *patchset = NULL; const char *client_name = NULL; crm_debug("Received CIB diff notification: DC=%s", pcmk__btoa(AM_I_DC)); if (cib__get_notify_patchset(msg, &patchset) != pcmk_rc_ok) { return; } if (cib__element_in_patchset(patchset, XML_CIB_TAG_ALERTS) || cib__element_in_patchset(patchset, XML_CIB_TAG_CRMCONFIG)) { controld_trigger_config(); } if (!AM_I_DC) { // We're not in control of the join sequence return; } client_name = crm_element_value(msg, F_CIB_CLIENTNAME); if (!cib__client_triggers_refresh(client_name)) { // The CIB is still accurate return; } if (cib__element_in_patchset(patchset, XML_CIB_TAG_NODES) || cib__element_in_patchset(patchset, XML_CIB_TAG_STATUS)) { /* An unsafe client modified the nodes or status section. Ensure the * node list is up-to-date, and start the join process again so we get * everyone's current resource history. */ if (client_name == NULL) { client_name = crm_element_value(msg, F_CIB_CLIENTID); } crm_notice("Populating nodes and starting an election after %s event " "triggered by %s", event, pcmk__s(client_name, "(unidentified client)")); populate_cib_nodes(node_update_quick|node_update_all, __func__); register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL); } } void controld_disconnect_cib_manager(void) { cib_t *cib_conn = controld_globals.cib_conn; CRM_ASSERT(cib_conn != NULL); crm_debug("Disconnecting from the CIB manager"); controld_clear_fsa_input_flags(R_CIB_CONNECTED); cib_conn->cmds->del_notify_callback(cib_conn, T_CIB_DIFF_NOTIFY, do_cib_updated); cib_free_callbacks(cib_conn); if (cib_conn->state != cib_disconnected) { cib_conn->cmds->set_secondary(cib_conn, cib_scope_local|cib_discard_reply); cib_conn->cmds->signoff(cib_conn); } } /* A_CIB_STOP, A_CIB_START, O_CIB_RESTART */ void do_cib_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { static int cib_retries = 0; cib_t *cib_conn = controld_globals.cib_conn; void (*dnotify_fn) (gpointer user_data) = handle_cib_disconnect; void (*update_cb) (const char *event, xmlNodePtr msg) = do_cib_updated; int rc = pcmk_ok; CRM_ASSERT(cib_conn != NULL); if (pcmk_is_set(action, A_CIB_STOP)) { if ((cib_conn->state != cib_disconnected) && (pending_rsc_update != 0)) { crm_info("Waiting for resource update %d to complete", pending_rsc_update); crmd_fsa_stall(FALSE); return; } controld_disconnect_cib_manager(); } if (!pcmk_is_set(action, A_CIB_START)) { return; } if (cur_state == S_STOPPING) { crm_err("Ignoring request to connect to the CIB manager after " "shutdown"); return; } rc = cib_conn->cmds->signon(cib_conn, CRM_SYSTEM_CRMD, cib_command_nonblocking); if (rc != pcmk_ok) { // A short wait that usually avoids stalling the FSA sleep(1); rc = cib_conn->cmds->signon(cib_conn, CRM_SYSTEM_CRMD, cib_command_nonblocking); } if (rc != pcmk_ok) { crm_info("Could not connect to the CIB manager: %s", pcmk_strerror(rc)); } else if (cib_conn->cmds->set_connection_dnotify(cib_conn, dnotify_fn) != pcmk_ok) { crm_err("Could not set dnotify callback"); } else if (cib_conn->cmds->add_notify_callback(cib_conn, T_CIB_DIFF_NOTIFY, update_cb) != pcmk_ok) { crm_err("Could not set CIB notification callback (update)"); } else { controld_set_fsa_input_flags(R_CIB_CONNECTED); cib_retries = 0; } if (!pcmk_is_set(controld_globals.fsa_input_register, R_CIB_CONNECTED)) { cib_retries++; if (cib_retries < 30) { crm_warn("Couldn't complete CIB registration %d times... " "pause and retry", cib_retries); controld_start_wait_timer(); crmd_fsa_stall(FALSE); } else { crm_err("Could not complete CIB registration %d times... " "hard error", cib_retries); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } } #define MIN_CIB_OP_TIMEOUT (30) /*! * \internal * \brief Get the timeout (in seconds) that should be used with CIB operations * * \return The maximum of 30 seconds, the value of the PCMK_cib_timeout * environment variable, or 10 seconds times one more than the number of * nodes in the cluster. */ unsigned int cib_op_timeout(void) { // @COMPAT: Drop env_timeout at 3.0.0 static int env_timeout = -1; unsigned int calculated_timeout = 0; if (env_timeout == -1) { const char *env = pcmk__env_option(PCMK__ENV_CIB_TIMEOUT); pcmk__scan_min_int(env, &env_timeout, MIN_CIB_OP_TIMEOUT); crm_trace("Minimum CIB op timeout: %ds (environment: %s)", env_timeout, (env? env : "none")); } calculated_timeout = 1 + crm_active_peers(); if (crm_remote_peer_cache) { calculated_timeout += g_hash_table_size(crm_remote_peer_cache); } calculated_timeout *= 10; calculated_timeout = QB_MAX(calculated_timeout, env_timeout); crm_trace("Calculated timeout: %us", calculated_timeout); if (controld_globals.cib_conn) { controld_globals.cib_conn->call_timeout = calculated_timeout; } return calculated_timeout; } /*! * \internal * \brief Get CIB call options to use local scope if primary is unavailable * * \return CIB call options */ int crmd_cib_smart_opt(void) { int call_opt = cib_none; if ((controld_globals.fsa_state == S_ELECTION) || (controld_globals.fsa_state == S_PENDING)) { crm_info("Sending update to local CIB in state: %s", fsa_state2string(controld_globals.fsa_state)); cib__set_call_options(call_opt, "update", cib_scope_local); } return call_opt; } static void cib_delete_callback(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { char *desc = user_data; if (rc == 0) { crm_debug("Deletion of %s (via CIB call %d) succeeded", desc, call_id); } else { crm_warn("Deletion of %s (via CIB call %d) failed: %s " CRM_XS " rc=%d", desc, call_id, pcmk_strerror(rc), rc); } } // Searches for various portions of node_state to delete // Match a particular node's node_state (takes node name 1x) -#define XPATH_NODE_STATE "//" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" +#define XPATH_NODE_STATE "//" XML_CIB_TAG_STATE \ + "[@" PCMK_XA_UNAME "='%s']" // Node's lrm section (name 1x) #define XPATH_NODE_LRM XPATH_NODE_STATE "/" XML_CIB_TAG_LRM /* Node's lrm_rsc_op entries and lrm_resource entries without unexpired lock * (name 2x, (seconds_since_epoch - PCMK_OPT_SHUTDOWN_LOCK_LIMIT) 1x) */ #define XPATH_NODE_LRM_UNLOCKED XPATH_NODE_STATE "//" XML_LRM_TAG_RSC_OP \ "|" XPATH_NODE_STATE \ "//" XML_LRM_TAG_RESOURCE \ "[not(@" PCMK_OPT_SHUTDOWN_LOCK ") " \ "or " PCMK_OPT_SHUTDOWN_LOCK "<%lld]" // Node's transient_attributes section (name 1x) #define XPATH_NODE_ATTRS XPATH_NODE_STATE "/" XML_TAG_TRANSIENT_NODEATTRS // Everything under node_state (name 1x) #define XPATH_NODE_ALL XPATH_NODE_STATE "/*" /* Unlocked history + transient attributes * (name 2x, (seconds_since_epoch - PCMK_OPT_SHUTDOWN_LOCK_LIMIT) 1x, name 1x) */ #define XPATH_NODE_ALL_UNLOCKED XPATH_NODE_LRM_UNLOCKED "|" XPATH_NODE_ATTRS /*! * \internal * \brief Get the XPath and description of a node state section to be deleted * * \param[in] uname Desired node * \param[in] section Subsection of node_state to be deleted * \param[out] xpath Where to store XPath of \p section * \param[out] desc If not \c NULL, where to store description of \p section */ void controld_node_state_deletion_strings(const char *uname, enum controld_section_e section, char **xpath, char **desc) { const char *desc_pre = NULL; // Shutdown locks that started before this time are expired long long expire = (long long) time(NULL) - controld_globals.shutdown_lock_limit; switch (section) { case controld_section_lrm: *xpath = crm_strdup_printf(XPATH_NODE_LRM, uname); desc_pre = "resource history"; break; case controld_section_lrm_unlocked: *xpath = crm_strdup_printf(XPATH_NODE_LRM_UNLOCKED, uname, uname, expire); desc_pre = "resource history (other than shutdown locks)"; break; case controld_section_attrs: *xpath = crm_strdup_printf(XPATH_NODE_ATTRS, uname); desc_pre = "transient attributes"; break; case controld_section_all: *xpath = crm_strdup_printf(XPATH_NODE_ALL, uname); desc_pre = "all state"; break; case controld_section_all_unlocked: *xpath = crm_strdup_printf(XPATH_NODE_ALL_UNLOCKED, uname, uname, expire, uname); desc_pre = "all state (other than shutdown locks)"; break; default: // We called this function incorrectly CRM_ASSERT(false); break; } if (desc != NULL) { *desc = crm_strdup_printf("%s for node %s", desc_pre, uname); } } /*! * \internal * \brief Delete subsection of a node's CIB node_state * * \param[in] uname Desired node * \param[in] section Subsection of node_state to delete * \param[in] options CIB call options to use */ void controld_delete_node_state(const char *uname, enum controld_section_e section, int options) { cib_t *cib = controld_globals.cib_conn; char *xpath = NULL; char *desc = NULL; int cib_rc = pcmk_ok; CRM_ASSERT((uname != NULL) && (cib != NULL)); controld_node_state_deletion_strings(uname, section, &xpath, &desc); cib__set_call_options(options, "node state deletion", cib_xpath|cib_multiple); cib_rc = cib->cmds->remove(cib, xpath, NULL, options); fsa_register_cib_callback(cib_rc, desc, cib_delete_callback); crm_info("Deleting %s (via CIB call %d) " CRM_XS " xpath=%s", desc, cib_rc, xpath); // CIB library handles freeing desc free(xpath); } // Takes node name and resource ID #define XPATH_RESOURCE_HISTORY "//" XML_CIB_TAG_STATE \ - "[@" XML_ATTR_UNAME "='%s']/" \ + "[@" PCMK_XA_UNAME "='%s']/" \ XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE \ "[@" PCMK_XA_ID "='%s']" // @TODO could add "and @PCMK_OPT_SHUTDOWN_LOCK" to limit to locks /*! * \internal * \brief Clear resource history from CIB for a given resource and node * * \param[in] rsc_id ID of resource to be cleared * \param[in] node Node whose resource history should be cleared * \param[in] user_name ACL user name to use * \param[in] call_options CIB call options * * \return Standard Pacemaker return code */ int controld_delete_resource_history(const char *rsc_id, const char *node, const char *user_name, int call_options) { char *desc = NULL; char *xpath = NULL; int rc = pcmk_rc_ok; cib_t *cib = controld_globals.cib_conn; CRM_CHECK((rsc_id != NULL) && (node != NULL), return EINVAL); desc = crm_strdup_printf("resource history for %s on %s", rsc_id, node); if (cib == NULL) { crm_err("Unable to clear %s: no CIB connection", desc); free(desc); return ENOTCONN; } // Ask CIB to delete the entry xpath = crm_strdup_printf(XPATH_RESOURCE_HISTORY, node, rsc_id); cib->cmds->set_user(cib, user_name); rc = cib->cmds->remove(cib, xpath, NULL, call_options|cib_xpath); cib->cmds->set_user(cib, NULL); if (rc < 0) { rc = pcmk_legacy2rc(rc); crm_err("Could not delete resource status of %s on %s%s%s: %s " CRM_XS " rc=%d", rsc_id, node, (user_name? " for user " : ""), (user_name? user_name : ""), pcmk_rc_str(rc), rc); free(desc); free(xpath); return rc; } if (pcmk_is_set(call_options, cib_sync_call)) { if (pcmk_is_set(call_options, cib_dryrun)) { crm_debug("Deletion of %s would succeed", desc); } else { crm_debug("Deletion of %s succeeded", desc); } free(desc); } else { crm_info("Clearing %s (via CIB call %d) " CRM_XS " xpath=%s", desc, rc, xpath); fsa_register_cib_callback(rc, desc, cib_delete_callback); // CIB library handles freeing desc } free(xpath); return pcmk_rc_ok; } /*! * \internal * \brief Build XML and string of parameters meeting some criteria, for digest * * \param[in] op Executor event with parameter table to use * \param[in] metadata Parsed meta-data for executed resource agent * \param[in] param_type Flag used for selection criteria * \param[out] result Will be set to newly created XML with selected * parameters as attributes * * \return Newly allocated space-separated string of parameter names * \note Selection criteria varies by param_type: for the restart digest, we * want parameters that are *not* marked reloadable (OCF 1.1) or that * *are* marked unique (pre-1.1), for both string and XML results; for the * secure digest, we want parameters that *are* marked private for the * string, but parameters that are *not* marked private for the XML. * \note It is the caller's responsibility to free the string return value with * \p g_string_free() and the XML result with \p free_xml(). */ static GString * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, enum ra_param_flags_e param_type, xmlNode **result) { GString *list = NULL; *result = create_xml_node(NULL, XML_TAG_PARAMS); /* Consider all parameters only except private ones to be consistent with * what scheduler does with calculate_secure_digest(). */ if (param_type == ra_param_private && compare_version(controld_globals.dc_version, "3.16.0") >= 0) { g_hash_table_foreach(op->params, hash2field, *result); pcmk__filter_op_for_digest(*result); } for (GList *iter = metadata->ra_params; iter != NULL; iter = iter->next) { struct ra_param_s *param = (struct ra_param_s *) iter->data; bool accept_for_list = false; bool accept_for_xml = false; switch (param_type) { case ra_param_reloadable: accept_for_list = !pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_unique: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_private: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = !accept_for_list; break; } if (accept_for_list) { crm_trace("Attr %s is %s", param->rap_name, ra_param_flag2text(param_type)); if (list == NULL) { // We will later search for " WORD ", so start list with a space pcmk__add_word(&list, 256, " "); } pcmk__add_word(&list, 0, param->rap_name); } else { crm_trace("Rejecting %s for %s", param->rap_name, ra_param_flag2text(param_type)); } if (accept_for_xml) { const char *v = g_hash_table_lookup(op->params, param->rap_name); if (v != NULL) { crm_trace("Adding attr %s=%s to the xml result", param->rap_name, v); crm_xml_add(*result, param->rap_name, v); } } else { crm_trace("Removing attr %s from the xml result", param->rap_name); xml_remove_prop(*result, param->rap_name); } } if (list != NULL) { // We will later search for " WORD ", so end list with a space pcmk__add_word(&list, 0, " "); } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { GString *list = NULL; char *digest = NULL; xmlNode *restart = NULL; CRM_LOG_ASSERT(op->params != NULL); if (op->interval_ms > 0) { /* monitors are not reloadable */ return; } if (pcmk_is_set(metadata->ra_flags, ra_supports_reload_agent)) { // Add parameters not marked reloadable to the "op-force-restart" list list = build_parameter_list(op, metadata, ra_param_reloadable, &restart); } else if (pcmk_is_set(metadata->ra_flags, ra_supports_legacy_reload)) { /* @COMPAT pre-OCF-1.1 resource agents * * Before OCF 1.1, Pacemaker abused "unique=0" to indicate * reloadability. Add any parameters with unique="1" to the * "op-force-restart" list. */ list = build_parameter_list(op, metadata, ra_param_unique, &restart); } else { // Resource does not support agent reloads return; } digest = calculate_operation_digest(restart, version); /* Add "op-force-restart" and "op-restart-digest" to indicate the resource supports reload, * no matter if it actually supports any parameters with unique="1"). */ crm_xml_add(update, XML_LRM_ATTR_OP_RESTART, (list == NULL)? "" : (const char *) list->str); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); if ((list != NULL) && (list->len > 0)) { crm_trace("%s: %s, %s", op->rsc_id, digest, (const char *) list->str); } else { crm_trace("%s: %s", op->rsc_id, digest); } if (list != NULL) { g_string_free(list, TRUE); } free_xml(restart); free(digest); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { GString *list = NULL; char *digest = NULL; xmlNode *secure = NULL; CRM_LOG_ASSERT(op->params != NULL); /* * To keep XML_LRM_ATTR_OP_SECURE short, we want it to contain the * secure parameters but XML_LRM_ATTR_SECURE_DIGEST to be based on * the insecure ones */ list = build_parameter_list(op, metadata, ra_param_private, &secure); if (list != NULL) { digest = calculate_operation_digest(secure, version); crm_xml_add(update, XML_LRM_ATTR_OP_SECURE, (const char *) list->str); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, (const char *) list->str); g_string_free(list, TRUE); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); } /*! * \internal * \brief Create XML for a resource history entry * * \param[in] func Function name of caller * \param[in,out] parent XML to add entry to * \param[in] rsc Affected resource * \param[in,out] op Action to add an entry for (or NULL to do nothing) * \param[in] node_name Node where action occurred */ void controld_add_resource_history_xml_as(const char *func, xmlNode *parent, const lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, const char *node_name) { int target_rc = 0; xmlNode *xml_op = NULL; struct ra_metadata_s *metadata = NULL; const char *caller_version = NULL; lrm_state_t *lrm_state = NULL; if (op == NULL) { return; } target_rc = rsc_op_expected_rc(op); caller_version = g_hash_table_lookup(op->params, PCMK_XA_CRM_FEATURE_SET); CRM_CHECK(caller_version != NULL, caller_version = CRM_FEATURE_SET); xml_op = pcmk__create_history_xml(parent, op, caller_version, target_rc, controld_globals.our_nodename, func); if (xml_op == NULL) { return; } if ((rsc == NULL) || (op->params == NULL) || !crm_op_needs_metadata(rsc->standard, op->op_type)) { crm_trace("No digests needed for %s action on %s (params=%p rsc=%p)", op->op_type, op->rsc_id, op->params, rsc); return; } lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { crm_warn("Cannot calculate digests for operation " PCMK__OP_FMT " because we have no connection to executor for %s", op->rsc_id, op->op_type, op->interval_ms, node_name); return; } /* Ideally the metadata is cached, and the agent is just a fallback. * * @TODO Go through all callers and ensure they get metadata asynchronously * first. */ metadata = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_agent |controld_metadata_from_cache); if (metadata == NULL) { return; } crm_trace("Including additional digests for %s:%s:%s", rsc->standard, rsc->provider, rsc->type); append_restart_list(op, metadata, xml_op, caller_version); append_secure_list(op, metadata, xml_op, caller_version); return; } /*! * \internal * \brief Record an action as pending in the CIB, if appropriate * * \param[in] node_name Node where the action is pending * \param[in] rsc Resource that action is for * \param[in,out] op Pending action * * \return true if action was recorded in CIB, otherwise false */ bool controld_record_pending_op(const char *node_name, const lrmd_rsc_info_t *rsc, lrmd_event_data_t *op) { const char *record_pending = NULL; CRM_CHECK((node_name != NULL) && (rsc != NULL) && (op != NULL), return false); // Never record certain operation types as pending if ((op->op_type == NULL) || (op->params == NULL) || !controld_action_is_recordable(op->op_type)) { return false; } // Check action's record-pending meta-attribute (defaults to true) record_pending = crm_meta_value(op->params, XML_OP_ATTR_PENDING); if ((record_pending != NULL) && !crm_is_true(record_pending)) { return false; } op->call_id = -1; op->t_run = time(NULL); op->t_rcchange = op->t_run; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); crm_debug("Recording pending %s-interval %s for %s on %s in the CIB", pcmk__readable_interval(op->interval_ms), op->op_type, op->rsc_id, node_name); controld_update_resource_history(node_name, rsc, op, 0); return true; } static void cib_rsc_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { switch (rc) { case pcmk_ok: case -pcmk_err_diff_failed: case -pcmk_err_diff_resync: crm_trace("Resource history update completed (call=%d rc=%d)", call_id, rc); break; default: if (call_id > 0) { crm_warn("Resource history update %d failed: %s " CRM_XS " rc=%d", call_id, pcmk_strerror(rc), rc); } else { crm_warn("Resource history update failed: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); } } if (call_id == pending_rsc_update) { pending_rsc_update = 0; controld_trigger_fsa(); } } /* Only successful stops, and probes that found the resource inactive, get locks * recorded in the history. This ensures the resource stays locked to the node * until it is active there again after the node comes back up. */ static bool should_preserve_lock(lrmd_event_data_t *op) { if (!pcmk_is_set(controld_globals.flags, controld_shutdown_lock_enabled)) { return false; } if (!strcmp(op->op_type, PCMK_ACTION_STOP) && (op->rc == PCMK_OCF_OK)) { return true; } if (!strcmp(op->op_type, PCMK_ACTION_MONITOR) && (op->rc == PCMK_OCF_NOT_RUNNING)) { return true; } return false; } /*! * \internal * \brief Request a CIB update * * \param[in] section Section of CIB to update * \param[in] data New XML of CIB section to update * \param[in] options CIB call options * \param[in] callback If not \c NULL, set this as the operation callback * * \return Standard Pacemaker return code * * \note If \p callback is \p cib_rsc_callback(), the CIB update's call ID is * stored in \p pending_rsc_update on success. */ int controld_update_cib(const char *section, xmlNode *data, int options, void (*callback)(xmlNode *, int, int, xmlNode *, void *)) { cib_t *cib = controld_globals.cib_conn; int cib_rc = -ENOTCONN; CRM_ASSERT(data != NULL); if (cib != NULL) { cib_rc = cib->cmds->modify(cib, section, data, options); if (cib_rc >= 0) { crm_debug("Submitted CIB update %d for %s section", cib_rc, section); } } if (callback == NULL) { if (cib_rc < 0) { crm_err("Failed to update CIB %s section: %s", section, pcmk_rc_str(pcmk_legacy2rc(cib_rc))); } } else { if ((cib_rc >= 0) && (callback == cib_rsc_callback)) { /* Checking for a particular callback is a little hacky, but it * didn't seem worth adding an output argument for cib_rc for just * one use case. */ pending_rsc_update = cib_rc; } fsa_register_cib_callback(cib_rc, NULL, callback); } return (cib_rc >= 0)? pcmk_rc_ok : pcmk_legacy2rc(cib_rc); } /*! * \internal * \brief Update resource history entry in CIB * * \param[in] node_name Node where action occurred * \param[in] rsc Resource that action is for * \param[in,out] op Action to record * \param[in] lock_time If nonzero, when resource was locked to node * * \note On success, the CIB update's call ID will be stored in * pending_rsc_update. */ void controld_update_resource_history(const char *node_name, const lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time) { xmlNode *update = NULL; xmlNode *xml = NULL; int call_opt = crmd_cib_smart_opt(); const char *node_id = NULL; const char *container = NULL; CRM_CHECK((node_name != NULL) && (op != NULL), return); if (rsc == NULL) { crm_warn("Resource %s no longer exists in the executor", op->rsc_id); controld_ack_event_directly(NULL, NULL, rsc, op, op->rsc_id); return; } // update = create_xml_node(NULL, XML_CIB_TAG_STATUS); // xml = create_xml_node(update, XML_CIB_TAG_STATE); if (pcmk__str_eq(node_name, controld_globals.our_nodename, pcmk__str_casei)) { node_id = controld_globals.our_uuid; } else { node_id = node_name; pcmk__xe_set_bool_attr(xml, XML_NODE_IS_REMOTE, true); } crm_xml_add(xml, PCMK_XA_ID, node_id); - crm_xml_add(xml, XML_ATTR_UNAME, node_name); + crm_xml_add(xml, PCMK_XA_UNAME, node_name); crm_xml_add(xml, PCMK_XA_CRM_DEBUG_ORIGIN, __func__); // xml = create_xml_node(xml, XML_CIB_TAG_LRM); crm_xml_add(xml, PCMK_XA_ID, node_id); // xml = create_xml_node(xml, XML_LRM_TAG_RESOURCES); // xml = create_xml_node(xml, XML_LRM_TAG_RESOURCE); crm_xml_add(xml, PCMK_XA_ID, op->rsc_id); crm_xml_add(xml, PCMK_XA_CLASS, rsc->standard); crm_xml_add(xml, PCMK_XA_PROVIDER, rsc->provider); crm_xml_add(xml, PCMK_XA_TYPE, rsc->type); if (lock_time != 0) { /* Actions on a locked resource should either preserve the lock by * recording it with the action result, or clear it. */ if (!should_preserve_lock(op)) { lock_time = 0; } crm_xml_add_ll(xml, PCMK_OPT_SHUTDOWN_LOCK, (long long) lock_time); } if (op->params != NULL) { container = g_hash_table_lookup(op->params, CRM_META "_" PCMK__META_CONTAINER); if (container != NULL) { crm_trace("Resource %s is a part of container resource %s", op->rsc_id, container); crm_xml_add(xml, PCMK__META_CONTAINER, container); } } // (possibly more than one) controld_add_resource_history_xml(xml, rsc, op, node_name); /* Update CIB asynchronously. Even if it fails, the resource state should be * discovered during the next election. Worst case, the node is wrongly * fenced for running a resource it isn't. */ crm_log_xml_trace(update, __func__); controld_update_cib(XML_CIB_TAG_STATUS, update, call_opt, cib_rsc_callback); free_xml(update); } /*! * \internal * \brief Erase an LRM history entry from the CIB, given the operation data * * \param[in] op Operation whose history should be deleted */ void controld_delete_action_history(const lrmd_event_data_t *op) { xmlNode *xml_top = NULL; CRM_CHECK(op != NULL, return); xml_top = create_xml_node(NULL, XML_LRM_TAG_RSC_OP); crm_xml_add_int(xml_top, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add(xml_top, XML_ATTR_TRANSITION_KEY, op->user_data); if (op->interval_ms > 0) { char *op_id = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); /* Avoid deleting last_failure too (if it was a result of this recurring op failing) */ crm_xml_add(xml_top, PCMK_XA_ID, op_id); free(op_id); } crm_debug("Erasing resource operation history for " PCMK__OP_FMT " (call=%d)", op->rsc_id, op->op_type, op->interval_ms, op->call_id); controld_globals.cib_conn->cmds->remove(controld_globals.cib_conn, XML_CIB_TAG_STATUS, xml_top, cib_none); crm_log_xml_trace(xml_top, "op:cancel"); free_xml(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ - "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ + "/" XML_CIB_TAG_STATE "[@" PCMK_XA_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" PCMK_XA_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP /* ... and also by operation key */ #define XPATH_HISTORY_ID XPATH_HISTORY "[@" PCMK_XA_ID "='%s']" /* ... and also by operation key and operation call ID */ #define XPATH_HISTORY_CALL XPATH_HISTORY \ "[@" PCMK_XA_ID "='%s' and @" XML_LRM_ATTR_CALLID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" PCMK_XA_ID "='%s' and @" XML_LRM_ATTR_TASK_KEY "='%s']" /*! * \internal * \brief Delete a last_failure resource history entry from the CIB * * \param[in] rsc_id Name of resource to clear history for * \param[in] node Name of node to clear history for * \param[in] action If specified, delete only if this was failed action * \param[in] interval_ms If \p action is specified, it has this interval */ void controld_cib_delete_last_failure(const char *rsc_id, const char *node, const char *action, guint interval_ms) { char *xpath = NULL; char *last_failure_key = NULL; CRM_CHECK((rsc_id != NULL) && (node != NULL), return); // Generate XPath to match desired entry last_failure_key = pcmk__op_key(rsc_id, "last_failure", 0); if (action == NULL) { xpath = crm_strdup_printf(XPATH_HISTORY_ID, node, rsc_id, last_failure_key); } else { char *action_key = pcmk__op_key(rsc_id, action, interval_ms); xpath = crm_strdup_printf(XPATH_HISTORY_ORIG, node, rsc_id, last_failure_key, action_key); free(action_key); } free(last_failure_key); controld_globals.cib_conn->cmds->remove(controld_globals.cib_conn, xpath, NULL, cib_xpath); free(xpath); } /*! * \internal * \brief Delete resource history entry from the CIB, given operation key * * \param[in] rsc_id Name of resource to clear history for * \param[in] node Name of node to clear history for * \param[in] key Operation key of operation to clear history for * \param[in] call_id If specified, delete entry only if it has this call ID */ void controld_delete_action_history_by_key(const char *rsc_id, const char *node, const char *key, int call_id) { char *xpath = NULL; CRM_CHECK((rsc_id != NULL) && (node != NULL) && (key != NULL), return); if (call_id > 0) { xpath = crm_strdup_printf(XPATH_HISTORY_CALL, node, rsc_id, key, call_id); } else { xpath = crm_strdup_printf(XPATH_HISTORY_ID, node, rsc_id, key); } controld_globals.cib_conn->cmds->remove(controld_globals.cib_conn, xpath, NULL, cib_xpath); free(xpath); } diff --git a/daemons/controld/controld_execd_state.c b/daemons/controld/controld_execd_state.c index 7e9b19edeb..af2f071466 100644 --- a/daemons/controld/controld_execd_state.c +++ b/daemons/controld/controld_execd_state.c @@ -1,815 +1,815 @@ /* * Copyright 2012-2024 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 static GHashTable *lrm_state_table = NULL; extern GHashTable *proxy_table; int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg); void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg)); static void free_rsc_info(gpointer value) { lrmd_rsc_info_t *rsc_info = value; lrmd_free_rsc_info(rsc_info); } static void free_deletion_op(gpointer value) { struct pending_deletion_op_s *op = value; free(op->rsc); delete_ha_msg_input(op->input); free(op); } static void free_recurring_op(gpointer value) { active_op_t *op = value; free(op->user_data); free(op->rsc_id); free(op->op_type); free(op->op_key); if (op->params) { g_hash_table_destroy(op->params); } free(op); } static gboolean fail_pending_op(gpointer key, gpointer value, gpointer user_data) { lrmd_event_data_t event = { 0, }; lrm_state_t *lrm_state = user_data; active_op_t *op = value; crm_trace("Pre-emptively failing " PCMK__OP_FMT " on %s (call=%s, %s)", op->rsc_id, op->op_type, op->interval_ms, lrm_state->node_name, (char*)key, op->user_data); event.type = lrmd_event_exec_complete; event.rsc_id = op->rsc_id; event.op_type = op->op_type; event.user_data = op->user_data; event.timeout = 0; event.interval_ms = op->interval_ms; lrmd__set_result(&event, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_NOT_CONNECTED, "Action was pending when executor connection was dropped"); event.t_run = (unsigned int) op->start_time; event.t_rcchange = (unsigned int) op->start_time; event.call_id = op->call_id; event.remote_nodename = lrm_state->node_name; event.params = op->params; process_lrm_event(lrm_state, &event, op, NULL); lrmd__reset_result(&event); return TRUE; } gboolean lrm_state_is_local(lrm_state_t *lrm_state) { return (lrm_state != NULL) && pcmk__str_eq(lrm_state->node_name, controld_globals.our_nodename, pcmk__str_casei); } /*! * \internal * \brief Create executor state entry for a node and add it to the state table * * \param[in] node_name Node to create entry for * * \return Newly allocated executor state object initialized for \p node_name */ static lrm_state_t * lrm_state_create(const char *node_name) { lrm_state_t *state = NULL; if (!node_name) { crm_err("No node name given for lrm state object"); return NULL; } state = calloc(1, sizeof(lrm_state_t)); if (!state) { return NULL; } state->node_name = strdup(node_name); state->rsc_info_cache = pcmk__strkey_table(NULL, free_rsc_info); state->deletion_ops = pcmk__strkey_table(free, free_deletion_op); state->active_ops = pcmk__strkey_table(free, free_recurring_op); state->resource_history = pcmk__strkey_table(NULL, history_free); state->metadata_cache = metadata_cache_new(); g_hash_table_insert(lrm_state_table, (char *)state->node_name, state); return state; } static gboolean remote_proxy_remove_by_node(gpointer key, gpointer value, gpointer user_data) { remote_proxy_t *proxy = value; const char *node_name = user_data; if (pcmk__str_eq(node_name, proxy->node_name, pcmk__str_casei)) { return TRUE; } return FALSE; } static remote_proxy_t * find_connected_proxy_by_node(const char * node_name) { GHashTableIter gIter; remote_proxy_t *proxy = NULL; CRM_CHECK(proxy_table != NULL, return NULL); g_hash_table_iter_init(&gIter, proxy_table); while (g_hash_table_iter_next(&gIter, NULL, (gpointer *) &proxy)) { if (proxy->source && pcmk__str_eq(node_name, proxy->node_name, pcmk__str_casei)) { return proxy; } } return NULL; } static void remote_proxy_disconnect_by_node(const char * node_name) { remote_proxy_t *proxy = NULL; CRM_CHECK(proxy_table != NULL, return); while ((proxy = find_connected_proxy_by_node(node_name)) != NULL) { /* mainloop_del_ipc_client() eventually calls remote_proxy_disconnected() * , which removes the entry from proxy_table. * Do not do this in a g_hash_table_iter_next() loop. */ if (proxy->source) { mainloop_del_ipc_client(proxy->source); } } return; } static void internal_lrm_state_destroy(gpointer data) { lrm_state_t *lrm_state = data; if (!lrm_state) { return; } /* Rather than directly remove the recorded proxy entries from proxy_table, * make sure any connected proxies get disconnected. So that * remote_proxy_disconnected() will be called and as well remove the * entries from proxy_table. */ remote_proxy_disconnect_by_node(lrm_state->node_name); crm_trace("Destroying proxy table %s with %u members", lrm_state->node_name, g_hash_table_size(proxy_table)); // Just in case there's still any leftovers in proxy_table g_hash_table_foreach_remove(proxy_table, remote_proxy_remove_by_node, (char *) lrm_state->node_name); remote_ra_cleanup(lrm_state); lrmd_api_delete(lrm_state->conn); if (lrm_state->rsc_info_cache) { crm_trace("Destroying rsc info cache with %u members", g_hash_table_size(lrm_state->rsc_info_cache)); g_hash_table_destroy(lrm_state->rsc_info_cache); } if (lrm_state->resource_history) { crm_trace("Destroying history op cache with %u members", g_hash_table_size(lrm_state->resource_history)); g_hash_table_destroy(lrm_state->resource_history); } if (lrm_state->deletion_ops) { crm_trace("Destroying deletion op cache with %u members", g_hash_table_size(lrm_state->deletion_ops)); g_hash_table_destroy(lrm_state->deletion_ops); } if (lrm_state->active_ops != NULL) { crm_trace("Destroying pending op cache with %u members", g_hash_table_size(lrm_state->active_ops)); g_hash_table_destroy(lrm_state->active_ops); } metadata_cache_free(lrm_state->metadata_cache); free((char *)lrm_state->node_name); free(lrm_state); } void lrm_state_reset_tables(lrm_state_t * lrm_state, gboolean reset_metadata) { if (lrm_state->resource_history) { crm_trace("Resetting resource history cache with %u members", g_hash_table_size(lrm_state->resource_history)); g_hash_table_remove_all(lrm_state->resource_history); } if (lrm_state->deletion_ops) { crm_trace("Resetting deletion operations cache with %u members", g_hash_table_size(lrm_state->deletion_ops)); g_hash_table_remove_all(lrm_state->deletion_ops); } if (lrm_state->active_ops != NULL) { crm_trace("Resetting active operations cache with %u members", g_hash_table_size(lrm_state->active_ops)); g_hash_table_remove_all(lrm_state->active_ops); } if (lrm_state->rsc_info_cache) { crm_trace("Resetting resource information cache with %u members", g_hash_table_size(lrm_state->rsc_info_cache)); g_hash_table_remove_all(lrm_state->rsc_info_cache); } if (reset_metadata) { metadata_cache_reset(lrm_state->metadata_cache); } } gboolean lrm_state_init_local(void) { if (lrm_state_table) { return TRUE; } lrm_state_table = pcmk__strikey_table(NULL, internal_lrm_state_destroy); if (!lrm_state_table) { return FALSE; } proxy_table = pcmk__strikey_table(NULL, remote_proxy_free); if (!proxy_table) { g_hash_table_destroy(lrm_state_table); lrm_state_table = NULL; return FALSE; } return TRUE; } void lrm_state_destroy_all(void) { if (lrm_state_table) { crm_trace("Destroying state table with %u members", g_hash_table_size(lrm_state_table)); g_hash_table_destroy(lrm_state_table); lrm_state_table = NULL; } if(proxy_table) { crm_trace("Destroying proxy table with %u members", g_hash_table_size(proxy_table)); g_hash_table_destroy(proxy_table); proxy_table = NULL; } } lrm_state_t * lrm_state_find(const char *node_name) { if ((node_name == NULL) || (lrm_state_table == NULL)) { return NULL; } return g_hash_table_lookup(lrm_state_table, node_name); } lrm_state_t * lrm_state_find_or_create(const char *node_name) { lrm_state_t *lrm_state; CRM_CHECK(lrm_state_table != NULL, return NULL); lrm_state = g_hash_table_lookup(lrm_state_table, node_name); if (!lrm_state) { lrm_state = lrm_state_create(node_name); } return lrm_state; } GList * lrm_state_get_list(void) { if (lrm_state_table == NULL) { return NULL; } return g_hash_table_get_values(lrm_state_table); } void lrm_state_disconnect_only(lrm_state_t * lrm_state) { int removed = 0; if (!lrm_state->conn) { return; } crm_trace("Disconnecting %s", lrm_state->node_name); remote_proxy_disconnect_by_node(lrm_state->node_name); ((lrmd_t *) lrm_state->conn)->cmds->disconnect(lrm_state->conn); if (!pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { removed = g_hash_table_foreach_remove(lrm_state->active_ops, fail_pending_op, lrm_state); crm_trace("Synthesized %d operation failures for %s", removed, lrm_state->node_name); } } void lrm_state_disconnect(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return; } lrm_state_disconnect_only(lrm_state); lrmd_api_delete(lrm_state->conn); lrm_state->conn = NULL; } int lrm_state_is_connected(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return FALSE; } return ((lrmd_t *) lrm_state->conn)->cmds->is_connected(lrm_state->conn); } int lrm_state_poke_connection(lrm_state_t * lrm_state) { if (!lrm_state->conn) { return -ENOTCONN; } return ((lrmd_t *) lrm_state->conn)->cmds->poke_connection(lrm_state->conn); } // \return Standard Pacemaker return code int controld_connect_local_executor(lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; if (lrm_state->conn == NULL) { lrmd_t *api = NULL; rc = lrmd__new(&api, NULL, NULL, 0); if (rc != pcmk_rc_ok) { return rc; } api->cmds->set_callback(api, lrm_op_callback); lrm_state->conn = api; } rc = ((lrmd_t *) lrm_state->conn)->cmds->connect(lrm_state->conn, CRM_SYSTEM_CRMD, NULL); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { lrm_state->num_lrm_register_fails = 0; } else { lrm_state->num_lrm_register_fails++; } return rc; } static remote_proxy_t * crmd_remote_proxy_new(lrmd_t *lrmd, const char *node_name, const char *session_id, const char *channel) { struct ipc_client_callbacks proxy_callbacks = { .dispatch = remote_proxy_dispatch, .destroy = remote_proxy_disconnected }; remote_proxy_t *proxy = remote_proxy_new(lrmd, &proxy_callbacks, node_name, session_id, channel); return proxy; } gboolean crmd_is_proxy_session(const char *session) { return g_hash_table_lookup(proxy_table, session) ? TRUE : FALSE; } void crmd_proxy_send(const char *session, xmlNode *msg) { remote_proxy_t *proxy = g_hash_table_lookup(proxy_table, session); lrm_state_t *lrm_state = NULL; if (!proxy) { return; } crm_log_xml_trace(msg, "to-proxy"); lrm_state = lrm_state_find(proxy->node_name); if (lrm_state) { crm_trace("Sending event to %.8s on %s", proxy->session_id, proxy->node_name); remote_proxy_relay_event(proxy, msg); } } static void crmd_proxy_dispatch(const char *session, xmlNode *msg) { crm_trace("Processing proxied IPC message from session %s", session); crm_log_xml_trace(msg, "controller[inbound]"); crm_xml_add(msg, F_CRM_SYS_FROM, session); if (controld_authorize_ipc_message(msg, NULL, session)) { route_message(C_IPC_MESSAGE, msg); } controld_trigger_fsa(); } static void remote_config_check(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { if (rc != pcmk_ok) { crm_err("Query resulted in an error: %s", pcmk_strerror(rc)); if (rc == -EACCES || rc == -pcmk_err_schema_validation) { crm_err("The cluster is mis-configured - shutting down and staying down"); } } else { lrmd_t * lrmd = (lrmd_t *)user_data; crm_time_t *now = crm_time_new(NULL); GHashTable *config_hash = pcmk__strkey_table(free, free); crm_debug("Call %d : Parsing CIB options", call_id); pe_unpack_nvpairs(output, output, XML_CIB_TAG_PROPSET, NULL, config_hash, CIB_OPTIONS_FIRST, FALSE, now, NULL); /* Now send it to the remote peer */ lrmd__validate_remote_settings(lrmd, config_hash); g_hash_table_destroy(config_hash); crm_time_free(now); } } static void crmd_remote_proxy_cb(lrmd_t *lrmd, void *userdata, xmlNode *msg) { lrm_state_t *lrm_state = userdata; const char *session = crm_element_value(msg, F_LRMD_IPC_SESSION); remote_proxy_t *proxy = g_hash_table_lookup(proxy_table, session); const char *op = crm_element_value(msg, F_LRMD_IPC_OP); if (pcmk__str_eq(op, LRMD_IPC_OP_NEW, pcmk__str_casei)) { const char *channel = crm_element_value(msg, F_LRMD_IPC_IPC_SERVER); proxy = crmd_remote_proxy_new(lrmd, lrm_state->node_name, session, channel); if (!remote_ra_controlling_guest(lrm_state)) { if (proxy != NULL) { cib_t *cib_conn = controld_globals.cib_conn; /* Look up PCMK_OPT_STONITH_WATCHDOG_TIMEOUT and send to the * remote peer for validation */ int rc = cib_conn->cmds->query(cib_conn, XML_CIB_TAG_CRMCONFIG, NULL, cib_scope_local); cib_conn->cmds->register_callback_full(cib_conn, rc, 10, FALSE, lrmd, "remote_config_check", remote_config_check, NULL); } } else { crm_debug("Skipping remote_config_check for guest-nodes"); } } else if (pcmk__str_eq(op, LRMD_IPC_OP_SHUTDOWN_REQ, pcmk__str_casei)) { char *now_s = NULL; crm_notice("%s requested shutdown of its remote connection", lrm_state->node_name); if (!remote_ra_is_in_maintenance(lrm_state)) { now_s = pcmk__ttoa(time(NULL)); update_attrd(lrm_state->node_name, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, TRUE); free(now_s); remote_proxy_ack_shutdown(lrmd); crm_warn("Reconnection attempts to %s may result in failures that must be cleared", lrm_state->node_name); } else { remote_proxy_nack_shutdown(lrmd); crm_notice("Remote resource for %s is not managed so no ordered shutdown happening", lrm_state->node_name); } return; } else if (pcmk__str_eq(op, LRMD_IPC_OP_REQUEST, pcmk__str_casei) && proxy && proxy->is_local) { /* This is for the controller, which we are, so don't try * to send to ourselves over IPC -- do it directly. */ int flags = 0; xmlNode *request = get_message_xml(msg, F_LRMD_IPC_MSG); CRM_CHECK(request != NULL, return); CRM_CHECK(lrm_state->node_name, return); crm_xml_add(request, XML_ACL_TAG_ROLE, "pacemaker-remote"); pcmk__update_acl_user(request, F_LRMD_IPC_USER, lrm_state->node_name); /* Pacemaker Remote nodes don't know their own names (as known to the * cluster). When getting a node info request with no name or ID, add * the name, so we don't return info for ourselves instead of the * Pacemaker Remote node. */ if (pcmk__str_eq(crm_element_value(request, F_CRM_TASK), CRM_OP_NODE_INFO, pcmk__str_casei)) { int node_id = 0; crm_element_value_int(request, PCMK_XA_ID, &node_id); if ((node_id <= 0) - && (crm_element_value(request, XML_ATTR_UNAME) == NULL)) { - crm_xml_add(request, XML_ATTR_UNAME, lrm_state->node_name); + && (crm_element_value(request, PCMK_XA_UNAME) == NULL)) { + crm_xml_add(request, PCMK_XA_UNAME, lrm_state->node_name); } } crmd_proxy_dispatch(session, request); crm_element_value_int(msg, F_LRMD_IPC_MSG_FLAGS, &flags); if (flags & crm_ipc_client_response) { int msg_id = 0; xmlNode *op_reply = create_xml_node(NULL, "ack"); crm_xml_add(op_reply, "function", __func__); crm_xml_add_int(op_reply, "line", __LINE__); crm_element_value_int(msg, F_LRMD_IPC_MSG_ID, &msg_id); remote_proxy_relay_response(proxy, op_reply, msg_id); free_xml(op_reply); } } else { remote_proxy_cb(lrmd, lrm_state->node_name, msg); } } // \return Standard Pacemaker return code int controld_connect_remote_executor(lrm_state_t *lrm_state, const char *server, int port, int timeout_ms) { int rc = pcmk_rc_ok; if (lrm_state->conn == NULL) { lrmd_t *api = NULL; rc = lrmd__new(&api, lrm_state->node_name, server, port); if (rc != pcmk_rc_ok) { crm_warn("Pacemaker Remote connection to %s:%s failed: %s " CRM_XS " rc=%d", server, port, pcmk_rc_str(rc), rc); return rc; } lrm_state->conn = api; api->cmds->set_callback(api, remote_lrm_op_callback); lrmd_internal_set_proxy_callback(api, lrm_state, crmd_remote_proxy_cb); } crm_trace("Initiating remote connection to %s:%d with timeout %dms", server, port, timeout_ms); rc = ((lrmd_t *) lrm_state->conn)->cmds->connect_async(lrm_state->conn, lrm_state->node_name, timeout_ms); if (rc == pcmk_ok) { lrm_state->num_lrm_register_fails = 0; } else { lrm_state->num_lrm_register_fails++; // Ignored for remote connections } return pcmk_legacy2rc(rc); } int lrm_state_get_metadata(lrm_state_t * lrm_state, const char *class, const char *provider, const char *agent, char **output, enum lrmd_call_options options) { lrmd_key_value_t *params = NULL; if (!lrm_state->conn) { return -ENOTCONN; } /* Add the node name to the environment, as is done with normal resource * action calls. Meta-data calls shouldn't need it, but some agents are * written with an ocf_local_nodename call at the beginning regardless of * action. Without the environment variable, the agent would try to contact * the controller to get the node name -- but the controller would be * blocking on the synchronous meta-data call. * * At this point, we have to assume that agents are unlikely to make other * calls that require the controller, such as crm_node --quorum or * --cluster-id. * * @TODO Make meta-data calls asynchronous. (This will be part of a larger * project to make meta-data calls via the executor rather than directly.) */ params = lrmd_key_value_add(params, CRM_META "_" XML_LRM_ATTR_TARGET, lrm_state->node_name); return ((lrmd_t *) lrm_state->conn)->cmds->get_metadata_params(lrm_state->conn, class, provider, agent, output, options, params); } int lrm_state_cancel(lrm_state_t *lrm_state, const char *rsc_id, const char *action, guint interval_ms) { if (!lrm_state->conn) { return -ENOTCONN; } /* Figure out a way to make this async? * NOTICE: Currently it's synced and directly acknowledged in do_lrm_invoke(). */ if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { return remote_ra_cancel(lrm_state, rsc_id, action, interval_ms); } return ((lrmd_t *) lrm_state->conn)->cmds->cancel(lrm_state->conn, rsc_id, action, interval_ms); } lrmd_rsc_info_t * lrm_state_get_rsc_info(lrm_state_t * lrm_state, const char *rsc_id, enum lrmd_call_options options) { lrmd_rsc_info_t *rsc = NULL; if (!lrm_state->conn) { return NULL; } if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { return remote_ra_get_rsc_info(lrm_state, rsc_id); } rsc = g_hash_table_lookup(lrm_state->rsc_info_cache, rsc_id); if (rsc == NULL) { /* only contact the lrmd if we don't already have a cached rsc info */ rsc = ((lrmd_t *) lrm_state->conn)->cmds->get_rsc_info(lrm_state->conn, rsc_id, options); if (rsc == NULL) { return NULL; } /* cache the result */ g_hash_table_insert(lrm_state->rsc_info_cache, rsc->id, rsc); } return lrmd_copy_rsc_info(rsc); } /*! * \internal * \brief Initiate a resource agent action * * \param[in,out] lrm_state Executor state object * \param[in] rsc_id ID of resource for action * \param[in] action Action to execute * \param[in] userdata String to copy and pass to execution callback * \param[in] interval_ms Action interval (in milliseconds) * \param[in] timeout_ms Action timeout (in milliseconds) * \param[in] start_delay_ms Delay (in ms) before initiating action * \param[in] parameters Hash table of resource parameters * \param[out] call_id Where to store call ID on success * * \return Standard Pacemaker return code */ int controld_execute_resource_agent(lrm_state_t *lrm_state, const char *rsc_id, const char *action, const char *userdata, guint interval_ms, int timeout_ms, int start_delay_ms, GHashTable *parameters, int *call_id) { int rc = pcmk_rc_ok; lrmd_key_value_t *params = NULL; if (lrm_state->conn == NULL) { return ENOTCONN; } // Convert parameters from hash table to list if (parameters != NULL) { const char *key = NULL; const char *value = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, parameters); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { params = lrmd_key_value_add(params, key, value); } } if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { rc = controld_execute_remote_agent(lrm_state, rsc_id, action, userdata, interval_ms, timeout_ms, start_delay_ms, params, call_id); } else { rc = ((lrmd_t *) lrm_state->conn)->cmds->exec(lrm_state->conn, rsc_id, action, userdata, interval_ms, timeout_ms, start_delay_ms, lrmd_opt_notify_changes_only, params); if (rc < 0) { rc = pcmk_legacy2rc(rc); } else { *call_id = rc; rc = pcmk_rc_ok; } } return rc; } int lrm_state_register_rsc(lrm_state_t * lrm_state, const char *rsc_id, const char *class, const char *provider, const char *agent, enum lrmd_call_options options) { lrmd_t *conn = (lrmd_t *) lrm_state->conn; if (conn == NULL) { return -ENOTCONN; } if (is_remote_lrmd_ra(agent, provider, NULL)) { return lrm_state_find_or_create(rsc_id)? pcmk_ok : -EINVAL; } /* @TODO Implement an asynchronous version of this (currently a blocking * call to the lrmd). */ return conn->cmds->register_rsc(lrm_state->conn, rsc_id, class, provider, agent, options); } int lrm_state_unregister_rsc(lrm_state_t * lrm_state, const char *rsc_id, enum lrmd_call_options options) { if (!lrm_state->conn) { return -ENOTCONN; } if (is_remote_lrmd_ra(NULL, NULL, rsc_id)) { g_hash_table_remove(lrm_state_table, rsc_id); return pcmk_ok; } g_hash_table_remove(lrm_state->rsc_info_cache, rsc_id); /* @TODO Optimize this ... this function is a blocking round trip from * client to daemon. The controld_execd_state.c code path that uses this * function should always treat it as an async operation. The executor API * should make an async version available. */ return ((lrmd_t *) lrm_state->conn)->cmds->unregister_rsc(lrm_state->conn, rsc_id, options); } diff --git a/daemons/controld/controld_join_dc.c b/daemons/controld/controld_join_dc.c index a7fd4f5274..5c876bd5e5 100644 --- a/daemons/controld/controld_join_dc.c +++ b/daemons/controld/controld_join_dc.c @@ -1,1034 +1,1034 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include static char *max_generation_from = NULL; static xmlNodePtr max_generation_xml = NULL; /*! * \internal * \brief Nodes from which a CIB sync has failed since the peer joined * * This table is of the form (node_name -> join_id). \p node_name is * the name of a client node from which a CIB \p sync_from() call has failed in * \p do_dc_join_finalize() since the client joined the cluster as a peer. * \p join_id is the ID of the join round in which the \p sync_from() failed, * and is intended for use in nack log messages. */ static GHashTable *failed_sync_nodes = NULL; void finalize_join_for(gpointer key, gpointer value, gpointer user_data); void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data); gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); /* Numeric counter used to identify join rounds (an unsigned int would be * appropriate, except we get and set it in XML as int) */ static int current_join_id = 0; /*! * \internal * \brief Destroy the hash table containing failed sync nodes */ void controld_destroy_failed_sync_table(void) { if (failed_sync_nodes != NULL) { g_hash_table_destroy(failed_sync_nodes); failed_sync_nodes = NULL; } } /*! * \internal * \brief Remove a node from the failed sync nodes table if present * * \param[in] node_name Node name to remove */ void controld_remove_failed_sync_node(const char *node_name) { if (failed_sync_nodes != NULL) { g_hash_table_remove(failed_sync_nodes, (gchar *) node_name); } } /*! * \internal * \brief Add to a hash table a node whose CIB failed to sync * * \param[in] node_name Name of node whose CIB failed to sync * \param[in] join_id Join round when the failure occurred */ static void record_failed_sync_node(const char *node_name, gint join_id) { if (failed_sync_nodes == NULL) { failed_sync_nodes = pcmk__strikey_table(g_free, NULL); } /* If the node is already in the table then we failed to nack it during the * filter offer step */ CRM_LOG_ASSERT(g_hash_table_insert(failed_sync_nodes, g_strdup(node_name), GINT_TO_POINTER(join_id))); } /*! * \internal * \brief Look up a node name in the failed sync table * * \param[in] node_name Name of node to look up * \param[out] join_id Where to store the join ID of when the sync failed * * \return Standard Pacemaker return code. Specifically, \p pcmk_rc_ok if the * node name was found, or \p pcmk_rc_node_unknown otherwise. * \note \p *join_id is set to -1 if the node is not found. */ static int lookup_failed_sync_node(const char *node_name, gint *join_id) { *join_id = -1; if (failed_sync_nodes != NULL) { gpointer result = g_hash_table_lookup(failed_sync_nodes, (gchar *) node_name); if (result != NULL) { *join_id = GPOINTER_TO_INT(result); return pcmk_rc_ok; } } return pcmk_rc_node_unknown; } void crm_update_peer_join(const char *source, crm_node_t * node, enum crm_join_phase phase) { enum crm_join_phase last = 0; CRM_CHECK(node != NULL, return); /* Remote nodes do not participate in joins */ if (pcmk_is_set(node->flags, crm_remote_node)) { return; } last = node->join; if(phase == last) { crm_trace("Node %s join-%d phase is still %s " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(last), node->id, source); } else if ((phase <= crm_join_none) || (phase == (last + 1))) { node->join = phase; crm_trace("Node %s join-%d phase is now %s (was %s) " CRM_XS " nodeid=%u source=%s", node->uname, current_join_id, crm_join_phase_str(phase), crm_join_phase_str(last), node->id, source); } else { crm_warn("Rejecting join-%d phase update for node %s because " "can't go from %s to %s " CRM_XS " nodeid=%u source=%s", current_join_id, node->uname, crm_join_phase_str(last), crm_join_phase_str(phase), node->id, source); } } static void start_join_round(void) { GHashTableIter iter; crm_node_t *peer = NULL; crm_debug("Starting new join round join-%d", current_join_id); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { crm_update_peer_join(__func__, peer, crm_join_none); } if (max_generation_from != NULL) { free(max_generation_from); max_generation_from = NULL; } if (max_generation_xml != NULL) { free_xml(max_generation_xml); max_generation_xml = NULL; } controld_clear_fsa_input_flags(R_HAVE_CIB); } /*! * \internal * \brief Create a join message from the DC * * \param[in] join_op Join operation name * \param[in] host_to Recipient of message */ static xmlNode * create_dc_message(const char *join_op, const char *host_to) { xmlNode *msg = create_request(join_op, NULL, host_to, CRM_SYSTEM_CRMD, CRM_SYSTEM_DC, NULL); /* Identify which election this is a part of */ crm_xml_add_int(msg, F_CRM_JOIN_ID, current_join_id); /* Add a field specifying whether the DC is shutting down. This keeps the * joining node from fencing the old DC if it becomes the new DC. */ pcmk__xe_set_bool_attr(msg, F_CRM_DC_LEAVING, pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)); return msg; } static void join_make_offer(gpointer key, gpointer value, gpointer user_data) { xmlNode *offer = NULL; crm_node_t *member = (crm_node_t *)value; CRM_ASSERT(member != NULL); if (crm_is_peer_active(member) == FALSE) { crm_info("Not making join-%d offer to inactive node %s", current_join_id, (member->uname? member->uname : "with unknown name")); if(member->expected == NULL && pcmk__str_eq(member->state, CRM_NODE_LOST, pcmk__str_casei)) { /* You would think this unsafe, but in fact this plus an * active resource is what causes it to be fenced. * * Yes, this does mean that any node that dies at the same * time as the old DC and is not running resource (still) * won't be fenced. * * I'm not happy about this either. */ pcmk__update_peer_expected(__func__, member, CRMD_JOINSTATE_DOWN); } return; } if (member->uname == NULL) { crm_info("Not making join-%d offer to node uuid %s with unknown name", current_join_id, member->uuid); return; } if (controld_globals.membership_id != crm_peer_seq) { controld_globals.membership_id = crm_peer_seq; crm_info("Making join-%d offers based on membership event %llu", current_join_id, crm_peer_seq); } if(user_data && member->join > crm_join_none) { crm_info("Not making join-%d offer to already known node %s (%s)", current_join_id, member->uname, crm_join_phase_str(member->join)); return; } crm_update_peer_join(__func__, (crm_node_t*)member, crm_join_none); offer = create_dc_message(CRM_OP_JOIN_OFFER, member->uname); // Advertise our feature set so the joining node can bail if not compatible crm_xml_add(offer, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET); crm_info("Sending join-%d offer to %s", current_join_id, member->uname); send_cluster_message(member, crm_msg_crmd, offer, TRUE); free_xml(offer); crm_update_peer_join(__func__, member, crm_join_welcomed); } /* A_DC_JOIN_OFFER_ALL */ void do_dc_join_offer_all(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { int count; /* Reset everyone's status back to down or in_ccm in the CIB. * Any nodes that are active in the CIB but not in the cluster membership * will be seen as offline by the scheduler anyway. */ current_join_id++; start_join_round(); update_dc(NULL); if (cause == C_HA_MESSAGE && current_input == I_NODE_JOIN) { crm_info("A new node joined the cluster"); } g_hash_table_foreach(crm_peer_cache, join_make_offer, NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } /* A_DC_JOIN_OFFER_ONE */ void do_dc_join_offer_one(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { crm_node_t *member; ha_msg_input_t *welcome = NULL; int count; const char *join_to = NULL; if (msg_data->data == NULL) { crm_info("Making join-%d offers to any unconfirmed nodes " "because an unknown node joined", current_join_id); g_hash_table_foreach(crm_peer_cache, join_make_offer, &member); check_join_state(cur_state, __func__); return; } welcome = fsa_typed_data(fsa_dt_ha_msg); if (welcome == NULL) { // fsa_typed_data() already logged an error return; } join_to = crm_element_value(welcome->msg, F_CRM_HOST_FROM); if (join_to == NULL) { crm_err("Can't make join-%d offer to unknown node", current_join_id); return; } member = crm_get_peer(0, join_to); /* It is possible that a node will have been sick or starting up when the * original offer was made. However, it will either re-announce itself in * due course, or we can re-store the original offer on the client. */ crm_update_peer_join(__func__, member, crm_join_none); join_make_offer(NULL, member, NULL); /* If the offer isn't to the local node, make an offer to the local node as * well, to ensure the correct value for max_generation_from. */ if (strcasecmp(join_to, controld_globals.our_nodename) != 0) { member = crm_get_peer(0, controld_globals.our_nodename); join_make_offer(NULL, member, NULL); } /* This was a genuine join request; cancel any existing transition and * invoke the scheduler. */ abort_transition(INFINITY, pcmk__graph_restart, "Node join", NULL); count = crmd_join_phase_count(crm_join_welcomed); crm_info("Waiting on join-%d requests from %d outstanding node%s", current_join_id, count, pcmk__plural_s(count)); // Don't waste time by invoking the scheduler yet } static int compare_int_fields(xmlNode * left, xmlNode * right, const char *field) { const char *elem_l = crm_element_value(left, field); const char *elem_r = crm_element_value(right, field); long long int_elem_l; long long int_elem_r; pcmk__scan_ll(elem_l, &int_elem_l, -1LL); pcmk__scan_ll(elem_r, &int_elem_r, -1LL); if (int_elem_l < int_elem_r) { return -1; } else if (int_elem_l > int_elem_r) { return 1; } return 0; } /* A_DC_JOIN_PROCESS_REQ */ void do_dc_join_filter_offer(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { xmlNode *generation = NULL; int cmp = 0; int join_id = -1; int count = 0; gint value = 0; gboolean ack_nack_bool = TRUE; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *join_from = crm_element_value(join_ack->msg, F_CRM_HOST_FROM); const char *ref = crm_element_value(join_ack->msg, F_CRM_REFERENCE); const char *join_version = crm_element_value(join_ack->msg, PCMK_XA_CRM_FEATURE_SET); crm_node_t *join_node = NULL; if (join_from == NULL) { crm_err("Ignoring invalid join request without node name"); return; } join_node = crm_get_peer(0, join_from); crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id); if (join_id != current_join_id) { crm_debug("Ignoring join-%d request from %s because we are on join-%d", join_id, join_from, current_join_id); check_join_state(cur_state, __func__); return; } generation = join_ack->xml; if (max_generation_xml != NULL && generation != NULL) { int lpc = 0; const char *attributes[] = { PCMK_XA_ADMIN_EPOCH, PCMK_XA_EPOCH, PCMK_XA_NUM_UPDATES, }; for (lpc = 0; cmp == 0 && lpc < PCMK__NELEM(attributes); lpc++) { cmp = compare_int_fields(max_generation_xml, generation, attributes[lpc]); } } if (ref == NULL) { ref = "none"; // for logging only } if (lookup_failed_sync_node(join_from, &value) == pcmk_rc_ok) { crm_err("Rejecting join-%d request from node %s because we failed to " "sync its CIB in join-%d " CRM_XS " ref=%s", join_id, join_from, value, ref); ack_nack_bool = FALSE; } else if (!crm_is_peer_active(join_node)) { if (match_down_event(join_from) != NULL) { /* The join request was received after the node was fenced or * otherwise shutdown in a way that we're aware of. No need to log * an error in this rare occurrence; we know the client was recently * shut down, and receiving a lingering in-flight request is not * cause for alarm. */ crm_debug("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } else { crm_err("Rejecting join-%d request from inactive node %s " CRM_XS " ref=%s", join_id, join_from, ref); } ack_nack_bool = FALSE; } else if (generation == NULL) { crm_err("Rejecting invalid join-%d request from node %s " "missing CIB generation " CRM_XS " ref=%s", join_id, join_from, ref); ack_nack_bool = FALSE; } else if ((join_version == NULL) || !feature_set_compatible(CRM_FEATURE_SET, join_version)) { crm_err("Rejecting join-%d request from node %s because feature set %s" " is incompatible with ours (%s) " CRM_XS " ref=%s", join_id, join_from, (join_version? join_version : "pre-3.1.0"), CRM_FEATURE_SET, ref); ack_nack_bool = FALSE; } else if (max_generation_xml == NULL) { const char *validation = crm_element_value(generation, PCMK_XA_VALIDATE_WITH); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with first CIB " "generation) due to unknown schema version %s " CRM_XS " ref=%s", join_id, join_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with first CIB " "generation) " CRM_XS " ref=%s", join_id, join_from, ref); max_generation_xml = copy_xml(generation); pcmk__str_update(&max_generation_from, join_from); } } else if ((cmp < 0) || ((cmp == 0) && pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei))) { const char *validation = crm_element_value(generation, PCMK_XA_VALIDATE_WITH); if (get_schema_version(validation) < 0) { crm_err("Rejecting join-%d request from %s (with better CIB " "generation than current best from %s) due to unknown " "schema version %s " CRM_XS " ref=%s", join_id, join_from, max_generation_from, validation, ref); ack_nack_bool = FALSE; } else { crm_debug("Accepting join-%d request from %s (with better CIB " "generation than current best from %s) " CRM_XS " ref=%s", join_id, join_from, max_generation_from, ref); crm_log_xml_debug(max_generation_xml, "Old max generation"); crm_log_xml_debug(generation, "New max generation"); free_xml(max_generation_xml); max_generation_xml = copy_xml(join_ack->xml); pcmk__str_update(&max_generation_from, join_from); } } else { crm_debug("Accepting join-%d request from %s " CRM_XS " ref=%s", join_id, join_from, ref); } if (!ack_nack_bool) { if (compare_version(join_version, "3.17.0") < 0) { /* Clients with CRM_FEATURE_SET < 3.17.0 may respawn infinitely * after a nack message, don't send one */ crm_update_peer_join(__func__, join_node, crm_join_nack_quiet); } else { crm_update_peer_join(__func__, join_node, crm_join_nack); } pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_NACK); } else { crm_update_peer_join(__func__, join_node, crm_join_integrated); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); } count = crmd_join_phase_count(crm_join_integrated); crm_debug("%d node%s currently integrated in join-%d", count, pcmk__plural_s(count), join_id); if (check_join_state(cur_state, __func__) == FALSE) { // Don't waste time by invoking the scheduler yet count = crmd_join_phase_count(crm_join_welcomed); crm_debug("Waiting on join-%d requests from %d outstanding node%s", join_id, count, pcmk__plural_s(count)); } } /* A_DC_JOIN_FINALIZE */ void do_dc_join_finalize(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { char *sync_from = NULL; int rc = pcmk_ok; int count_welcomed = crmd_join_phase_count(crm_join_welcomed); int count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); /* This we can do straight away and avoid clients timing us out * while we compute the latest CIB */ if (count_welcomed != 0) { crm_debug("Waiting on join-%d requests from %d outstanding node%s " "before finalizing join", current_join_id, count_welcomed, pcmk__plural_s(count_welcomed)); crmd_join_phase_log(LOG_DEBUG); /* crmd_fsa_stall(FALSE); Needed? */ return; } else if (count_finalizable == 0) { crm_debug("Finalization not needed for join-%d at the current time", current_join_id); crmd_join_phase_log(LOG_DEBUG); check_join_state(controld_globals.fsa_state, __func__); return; } controld_clear_fsa_input_flags(R_HAVE_CIB); if (pcmk__str_eq(max_generation_from, controld_globals.our_nodename, pcmk__str_null_matches|pcmk__str_casei)) { controld_set_fsa_input_flags(R_HAVE_CIB); } if (!controld_globals.transition_graph->complete) { crm_warn("Delaying join-%d finalization while transition in progress", current_join_id); crmd_join_phase_log(LOG_DEBUG); crmd_fsa_stall(FALSE); return; } if (pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { // Send our CIB out to everyone pcmk__str_update(&sync_from, controld_globals.our_nodename); crm_debug("Finalizing join-%d for %d node%s (sync'ing from local CIB)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable)); crm_log_xml_debug(max_generation_xml, "Requested CIB version"); } else { // Ask for the agreed best CIB pcmk__str_update(&sync_from, max_generation_from); crm_notice("Finalizing join-%d for %d node%s (sync'ing CIB from %s)", current_join_id, count_finalizable, pcmk__plural_s(count_finalizable), sync_from); crm_log_xml_notice(max_generation_xml, "Requested CIB version"); } crmd_join_phase_log(LOG_DEBUG); rc = controld_globals.cib_conn->cmds->sync_from(controld_globals.cib_conn, sync_from, NULL, cib_none); fsa_register_cib_callback(rc, sync_from, finalize_sync_callback); } void free_max_generation(void) { free(max_generation_from); max_generation_from = NULL; free_xml(max_generation_xml); max_generation_xml = NULL; } void finalize_sync_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { CRM_LOG_ASSERT(-EPERM != rc); if (rc != pcmk_ok) { const char *sync_from = (const char *) user_data; do_crm_log(((rc == -pcmk_err_old_data)? LOG_WARNING : LOG_ERR), "Could not sync CIB from %s in join-%d: %s", sync_from, current_join_id, pcmk_strerror(rc)); if (rc != -pcmk_err_old_data) { record_failed_sync_node(sync_from, current_join_id); } /* restart the whole join process */ register_fsa_error_adv(C_FSA_INTERNAL, I_ELECTION_DC, NULL, NULL, __func__); } else if (!AM_I_DC) { crm_debug("Sync'ed CIB for join-%d but no longer DC", current_join_id); } else if (controld_globals.fsa_state != S_FINALIZE_JOIN) { crm_debug("Sync'ed CIB for join-%d but no longer in S_FINALIZE_JOIN " "(%s)", current_join_id, fsa_state2string(controld_globals.fsa_state)); } else { controld_set_fsa_input_flags(R_HAVE_CIB); /* make sure dc_uuid is re-set to us */ if (!check_join_state(controld_globals.fsa_state, __func__)) { int count_finalizable = 0; count_finalizable = crmd_join_phase_count(crm_join_integrated) + crmd_join_phase_count(crm_join_nack) + crmd_join_phase_count(crm_join_nack_quiet); crm_debug("Notifying %d node%s of join-%d results", count_finalizable, pcmk__plural_s(count_finalizable), current_join_id); g_hash_table_foreach(crm_peer_cache, finalize_join_for, NULL); } } } static void join_node_state_commit_callback(xmlNode *msg, int call_id, int rc, xmlNode *output, void *user_data) { const char *node = user_data; if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; // for register_fsa_error() macro crm_crit("join-%d node history update (via CIB call %d) for node %s " "failed: %s", current_join_id, call_id, node, pcmk_strerror(rc)); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } crm_debug("join-%d node history update (via CIB call %d) for node %s " "complete", current_join_id, call_id, node); check_join_state(controld_globals.fsa_state, __func__); } /* A_DC_JOIN_PROCESS_ACK */ void do_dc_join_ack(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { int join_id = -1; ha_msg_input_t *join_ack = fsa_typed_data(fsa_dt_ha_msg); const char *op = crm_element_value(join_ack->msg, F_CRM_TASK); char *join_from = crm_element_value_copy(join_ack->msg, F_CRM_HOST_FROM); crm_node_t *peer = NULL; enum controld_section_e section = controld_section_lrm; char *xpath = NULL; xmlNode *state = join_ack->xml; xmlNode *execd_state = NULL; cib_t *cib = controld_globals.cib_conn; int rc = pcmk_ok; // Sanity checks if (join_from == NULL) { crm_warn("Ignoring message received without node identification"); goto done; } if (op == NULL) { crm_warn("Ignoring message received from %s without task", join_from); goto done; } if (strcmp(op, CRM_OP_JOIN_CONFIRM)) { crm_debug("Ignoring '%s' message from %s while waiting for '%s'", op, join_from, CRM_OP_JOIN_CONFIRM); goto done; } if (crm_element_value_int(join_ack->msg, F_CRM_JOIN_ID, &join_id) != 0) { crm_warn("Ignoring join confirmation from %s without valid join ID", join_from); goto done; } peer = crm_get_peer(0, join_from); if (peer->join != crm_join_finalized) { crm_info("Ignoring out-of-sequence join-%d confirmation from %s " "(currently %s not %s)", join_id, join_from, crm_join_phase_str(peer->join), crm_join_phase_str(crm_join_finalized)); goto done; } if (join_id != current_join_id) { crm_err("Rejecting join-%d confirmation from %s " "because currently on join-%d", join_id, join_from, current_join_id); crm_update_peer_join(__func__, peer, crm_join_nack); goto done; } crm_update_peer_join(__func__, peer, crm_join_confirmed); /* Update CIB with node's current executor state. A new transition will be * triggered later, when the CIB manager notifies us of the change. * * The delete and modify requests are part of an atomic transaction. */ rc = cib->cmds->init_transaction(cib); if (rc != pcmk_ok) { goto done; } // Delete relevant parts of node's current executor state from CIB if (pcmk_is_set(controld_globals.flags, controld_shutdown_lock_enabled)) { section = controld_section_lrm_unlocked; } controld_node_state_deletion_strings(join_from, section, &xpath, NULL); rc = cib->cmds->remove(cib, xpath, NULL, cib_scope_local |cib_xpath |cib_multiple |cib_transaction); if (rc != pcmk_ok) { goto done; } // Update CIB with node's latest known executor state if (pcmk__str_eq(join_from, controld_globals.our_nodename, pcmk__str_casei)) { // Use the latest possible state if processing our own join ack execd_state = controld_query_executor_state(); if (execd_state != NULL) { crm_debug("Updating local node history for join-%d from query " "result", current_join_id); state = execd_state; } else { crm_warn("Updating local node history from join-%d confirmation " "because query failed", current_join_id); } } else { crm_debug("Updating node history for %s from join-%d confirmation", join_from, current_join_id); } rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, state, cib_scope_local|cib_can_create|cib_transaction); free_xml(execd_state); if (rc != pcmk_ok) { goto done; } // Commit the transaction rc = cib->cmds->end_transaction(cib, true, cib_scope_local); fsa_register_cib_callback(rc, join_from, join_node_state_commit_callback); if (rc > 0) { // join_from will be freed after callback join_from = NULL; rc = pcmk_ok; } done: if (rc != pcmk_ok) { crm_crit("join-%d node history update for node %s failed: %s", current_join_id, join_from, pcmk_strerror(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } free(join_from); free(xpath); } void finalize_join_for(gpointer key, gpointer value, gpointer user_data) { xmlNode *acknak = NULL; xmlNode *tmp1 = NULL; crm_node_t *join_node = value; const char *join_to = join_node->uname; bool integrated = false; switch (join_node->join) { case crm_join_integrated: integrated = true; break; case crm_join_nack: case crm_join_nack_quiet: break; default: crm_trace("Not updating non-integrated and non-nacked node %s (%s) " "for join-%d", join_to, crm_join_phase_str(join_node->join), current_join_id); return; } /* Update the element with the node's name and UUID, in case they * weren't known before */ crm_trace("Updating node name and UUID in CIB for %s", join_to); tmp1 = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(tmp1, PCMK_XA_ID, crm_peer_uuid(join_node)); - crm_xml_add(tmp1, XML_ATTR_UNAME, join_to); + crm_xml_add(tmp1, PCMK_XA_UNAME, join_to); fsa_cib_anon_update(XML_CIB_TAG_NODES, tmp1); free_xml(tmp1); if (join_node->join == crm_join_nack_quiet) { crm_trace("Not sending nack message to node %s with feature set older " "than 3.17.0", join_to); return; } join_node = crm_get_peer(0, join_to); if (!crm_is_peer_active(join_node)) { /* * NACK'ing nodes that the membership layer doesn't know about yet * simply creates more churn * * Better to leave them waiting and let the join restart when * the new membership event comes in * * All other NACKs (due to versions etc) should still be processed */ pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_PENDING); return; } // Acknowledge or nack node's join request crm_debug("%sing join-%d request from %s", integrated? "Acknowledg" : "Nack", current_join_id, join_to); acknak = create_dc_message(CRM_OP_JOIN_ACKNAK, join_to); pcmk__xe_set_bool_attr(acknak, CRM_OP_JOIN_ACKNAK, integrated); if (integrated) { // No change needed for a nacked node crm_update_peer_join(__func__, join_node, crm_join_finalized); pcmk__update_peer_expected(__func__, join_node, CRMD_JOINSTATE_MEMBER); /* Iterate through the remote peer cache and add information on which * node hosts each to the ACK message. This keeps new controllers in * sync with what has already happened. */ if (crm_remote_peer_cache_size() != 0) { GHashTableIter iter; crm_node_t *node = NULL; xmlNode *remotes = create_xml_node(acknak, XML_CIB_TAG_NODES); g_hash_table_iter_init(&iter, crm_remote_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *remote = NULL; if (!node->conn_host) { continue; } remote = create_xml_node(remotes, XML_CIB_TAG_NODE); pcmk__xe_set_props(remote, PCMK_XA_ID, node->uname, XML_CIB_TAG_STATE, node->state, PCMK__XA_CONN_HOST, node->conn_host, NULL); } } } send_cluster_message(join_node, crm_msg_crmd, acknak, TRUE); free_xml(acknak); return; } gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source) { static unsigned long long highest_seq = 0; if (controld_globals.membership_id != crm_peer_seq) { crm_debug("join-%d: Membership changed from %llu to %llu " CRM_XS " highest=%llu state=%s for=%s", current_join_id, controld_globals.membership_id, crm_peer_seq, highest_seq, fsa_state2string(cur_state), source); if(highest_seq < crm_peer_seq) { /* Don't spam the FSA with duplicates */ highest_seq = crm_peer_seq; register_fsa_input_before(C_FSA_INTERNAL, I_NODE_JOIN, NULL); } } else if (cur_state == S_INTEGRATION) { if (crmd_join_phase_count(crm_join_welcomed) == 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Integration of %d peer%s complete " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); register_fsa_input_before(C_FSA_INTERNAL, I_INTEGRATED, NULL); return TRUE; } } else if (cur_state == S_FINALIZE_JOIN) { if (!pcmk_is_set(controld_globals.fsa_input_register, R_HAVE_CIB)) { crm_debug("join-%d: Delaying finalization until we have CIB " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); return TRUE; } else if (crmd_join_phase_count(crm_join_welcomed) != 0) { int count = crmd_join_phase_count(crm_join_welcomed); crm_debug("join-%d: Still waiting on %d welcomed node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_integrated) != 0) { int count = crmd_join_phase_count(crm_join_integrated); crm_debug("join-%d: Still waiting on %d integrated node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else if (crmd_join_phase_count(crm_join_finalized) != 0) { int count = crmd_join_phase_count(crm_join_finalized); crm_debug("join-%d: Still waiting on %d finalized node%s " CRM_XS " state=%s for=%s", current_join_id, count, pcmk__plural_s(count), fsa_state2string(cur_state), source); crmd_join_phase_log(LOG_DEBUG); } else { crm_debug("join-%d: Complete " CRM_XS " state=%s for=%s", current_join_id, fsa_state2string(cur_state), source); register_fsa_input_later(C_FSA_INTERNAL, I_FINALIZED, NULL); return TRUE; } } return FALSE; } void do_dc_join_final(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { crm_debug("Ensuring DC, quorum and node attributes are up-to-date"); crm_update_quorum(crm_have_quorum, TRUE); } int crmd_join_phase_count(enum crm_join_phase phase) { int count = 0; crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { if(peer->join == phase) { count++; } } return count; } void crmd_join_phase_log(int level) { crm_node_t *peer; GHashTableIter iter; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &peer)) { do_crm_log(level, "join-%d: %s=%s", current_join_id, peer->uname, crm_join_phase_str(peer->join)); } } diff --git a/daemons/controld/controld_membership.c b/daemons/controld/controld_membership.c index 2171b4b7a5..a0c0ddf98d 100644 --- a/daemons/controld/controld_membership.c +++ b/daemons/controld/controld_membership.c @@ -1,469 +1,469 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ /* put these first so that uuid_t is defined without conflicts */ #include #include #include #include #include #include #include #include void post_cache_update(int instance); extern gboolean check_join_state(enum crmd_fsa_state cur_state, const char *source); static void reap_dead_nodes(gpointer key, gpointer value, gpointer user_data) { crm_node_t *node = value; if (crm_is_peer_active(node) == FALSE) { crm_update_peer_join(__func__, node, crm_join_none); if(node && node->uname) { if (pcmk__str_eq(controld_globals.our_nodename, node->uname, pcmk__str_casei)) { crm_err("We're not part of the cluster anymore"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); } else if (!AM_I_DC && pcmk__str_eq(node->uname, controld_globals.dc_name, pcmk__str_casei)) { crm_warn("Our DC node (%s) left the cluster", node->uname); register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL); } } if ((controld_globals.fsa_state == S_INTEGRATION) || (controld_globals.fsa_state == S_FINALIZE_JOIN)) { check_join_state(controld_globals.fsa_state, __func__); } if ((node != NULL) && (node->uuid != NULL)) { fail_incompletable_actions(controld_globals.transition_graph, node->uuid); } } } void post_cache_update(int instance) { xmlNode *no_op = NULL; crm_peer_seq = instance; crm_debug("Updated cache after membership event %d.", instance); g_hash_table_foreach(crm_peer_cache, reap_dead_nodes, NULL); controld_set_fsa_input_flags(R_MEMBERSHIP); if (AM_I_DC) { populate_cib_nodes(node_update_quick | node_update_cluster | node_update_peer | node_update_expected, __func__); } /* * If we lost nodes, we should re-check the election status * Safe to call outside of an election */ controld_set_fsa_action_flags(A_ELECTION_CHECK); controld_trigger_fsa(); /* Membership changed, remind everyone we're here. * This will aid detection of duplicate DCs */ no_op = create_request(CRM_OP_NOOP, NULL, NULL, CRM_SYSTEM_CRMD, AM_I_DC ? CRM_SYSTEM_DC : CRM_SYSTEM_CRMD, NULL); send_cluster_message(NULL, crm_msg_crmd, no_op, FALSE); free_xml(no_op); } static void crmd_node_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if (rc == pcmk_ok) { crm_trace("Node update %d complete", call_id); } else if(call_id < pcmk_ok) { crm_err("Node update failed: %s (%d)", pcmk_strerror(call_id), call_id); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } else { crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } /*! * \internal * \brief Create an XML node state tag with updates * * \param[in,out] node Node whose state will be used for update * \param[in] flags Bitmask of node_update_flags indicating what to update * \param[in,out] parent XML node to contain update (or NULL) * \param[in] source Who requested the update (only used for logging) * * \return Pointer to created node state tag */ xmlNode * create_node_state_update(crm_node_t *node, int flags, xmlNode *parent, const char *source) { const char *value = NULL; xmlNode *node_state; if (!node->state) { crm_info("Node update for %s cancelled: no state, not seen yet", node->uname); return NULL; } node_state = create_xml_node(parent, XML_CIB_TAG_STATE); if (pcmk_is_set(node->flags, crm_remote_node)) { pcmk__xe_set_bool_attr(node_state, XML_NODE_IS_REMOTE, true); } if (crm_xml_add(node_state, PCMK_XA_ID, crm_peer_uuid(node)) == NULL) { crm_info("Node update for %s cancelled: no ID", node->uname); free_xml(node_state); return NULL; } - crm_xml_add(node_state, XML_ATTR_UNAME, node->uname); + crm_xml_add(node_state, PCMK_XA_UNAME, node->uname); if ((flags & node_update_cluster) && node->state) { if (compare_version(controld_globals.dc_version, "3.18.0") >= 0) { // A value 0 means the node is not a cluster member. crm_xml_add_ll(node_state, PCMK__XA_IN_CCM, node->when_member); } else { pcmk__xe_set_bool_attr(node_state, PCMK__XA_IN_CCM, pcmk__str_eq(node->state, CRM_NODE_MEMBER, pcmk__str_casei)); } } if (!pcmk_is_set(node->flags, crm_remote_node)) { if (flags & node_update_peer) { if (compare_version(controld_globals.dc_version, "3.18.0") >= 0) { // A value 0 means the peer is offline in CPG. crm_xml_add_ll(node_state, PCMK__XA_CRMD, node->when_online); } else { // @COMPAT DCs < 2.1.7 use online/offline rather than timestamp value = OFFLINESTATUS; if (pcmk_is_set(node->processes, crm_get_cluster_proc())) { value = ONLINESTATUS; } crm_xml_add(node_state, PCMK__XA_CRMD, value); } } if (flags & node_update_join) { if (node->join <= crm_join_none) { value = CRMD_JOINSTATE_DOWN; } else { value = CRMD_JOINSTATE_MEMBER; } crm_xml_add(node_state, PCMK__XA_JOIN, value); } if (flags & node_update_expected) { crm_xml_add(node_state, PCMK__XA_EXPECTED, node->expected); } } crm_xml_add(node_state, PCMK_XA_CRM_DEBUG_ORIGIN, source); return node_state; } static void remove_conflicting_node_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *node_uuid = user_data; do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE, "Deletion of the unknown conflicting node \"%s\": %s (rc=%d)", node_uuid, pcmk_strerror(rc), rc); } static void search_conflicting_node_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { char *new_node_uuid = user_data; xmlNode *node_xml = NULL; if (rc != pcmk_ok) { if (rc != -ENXIO) { crm_notice("Searching conflicting nodes for %s failed: %s (%d)", new_node_uuid, pcmk_strerror(rc), rc); } return; } else if (output == NULL) { return; } if (pcmk__xe_is(output, XML_CIB_TAG_NODE)) { node_xml = output; } else { node_xml = pcmk__xml_first_child(output); } for (; node_xml != NULL; node_xml = pcmk__xml_next(node_xml)) { const char *node_uuid = NULL; const char *node_uname = NULL; GHashTableIter iter; crm_node_t *node = NULL; gboolean known = FALSE; if (!pcmk__xe_is(node_xml, XML_CIB_TAG_NODE)) { continue; } node_uuid = crm_element_value(node_xml, PCMK_XA_ID); - node_uname = crm_element_value(node_xml, XML_ATTR_UNAME); + node_uname = crm_element_value(node_xml, PCMK_XA_UNAME); if (node_uuid == NULL || node_uname == NULL) { continue; } g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (node->uuid && pcmk__str_eq(node->uuid, node_uuid, pcmk__str_casei) && node->uname && pcmk__str_eq(node->uname, node_uname, pcmk__str_casei)) { known = TRUE; break; } } if (known == FALSE) { cib_t *cib_conn = controld_globals.cib_conn; int delete_call_id = 0; xmlNode *node_state_xml = NULL; crm_notice("Deleting unknown node %s/%s which has conflicting uname with %s", node_uuid, node_uname, new_node_uuid); delete_call_id = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_NODES, node_xml, cib_scope_local); fsa_register_cib_callback(delete_call_id, strdup(node_uuid), remove_conflicting_node_callback); node_state_xml = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(node_state_xml, PCMK_XA_ID, node_uuid); - crm_xml_add(node_state_xml, XML_ATTR_UNAME, node_uname); + crm_xml_add(node_state_xml, PCMK_XA_UNAME, node_uname); delete_call_id = cib_conn->cmds->remove(cib_conn, XML_CIB_TAG_STATUS, node_state_xml, cib_scope_local); fsa_register_cib_callback(delete_call_id, strdup(node_uuid), remove_conflicting_node_callback); free_xml(node_state_xml); } } } static void node_list_update_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if(call_id < pcmk_ok) { crm_err("Node list update failed: %s (%d)", pcmk_strerror(call_id), call_id); crm_log_xml_debug(msg, "update:failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } else if(rc < pcmk_ok) { crm_err("Node update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "update:failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } void populate_cib_nodes(enum node_update_flags flags, const char *source) { cib_t *cib_conn = controld_globals.cib_conn; int call_id = 0; gboolean from_hashtable = TRUE; xmlNode *node_list = create_xml_node(NULL, XML_CIB_TAG_NODES); #if SUPPORT_COROSYNC if (!pcmk_is_set(flags, node_update_quick) && is_corosync_cluster()) { from_hashtable = pcmk__corosync_add_nodes(node_list); } #endif if (from_hashtable) { GHashTableIter iter; crm_node_t *node = NULL; GString *xpath = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *new_node = NULL; if ((node->uuid != NULL) && (node->uname != NULL)) { crm_trace("Creating node entry for %s/%s", node->uname, node->uuid); if (xpath == NULL) { xpath = g_string_sized_new(512); } else { g_string_truncate(xpath, 0); } /* We need both to be valid */ new_node = create_xml_node(node_list, XML_CIB_TAG_NODE); crm_xml_add(new_node, PCMK_XA_ID, node->uuid); - crm_xml_add(new_node, XML_ATTR_UNAME, node->uname); + crm_xml_add(new_node, PCMK_XA_UNAME, node->uname); /* Search and remove unknown nodes with the conflicting uname from CIB */ pcmk__g_strcat(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE - "[@" XML_ATTR_UNAME "='", node->uname, "']" + "[@" PCMK_XA_UNAME "='", node->uname, "']" "[@" PCMK_XA_ID "!='", node->uuid, "']", NULL); call_id = cib_conn->cmds->query(cib_conn, (const char *) xpath->str, NULL, cib_scope_local|cib_xpath); fsa_register_cib_callback(call_id, strdup(node->uuid), search_conflicting_node_callback); } } if (xpath != NULL) { g_string_free(xpath, TRUE); } } crm_trace("Populating section from %s", from_hashtable ? "hashtable" : "cluster"); if ((controld_update_cib(XML_CIB_TAG_NODES, node_list, cib_scope_local, node_list_update_callback) == pcmk_rc_ok) && (crm_peer_cache != NULL) && AM_I_DC) { /* * There is no need to update the local CIB with our values if * we've not seen valid membership data */ GHashTableIter iter; crm_node_t *node = NULL; free_xml(node_list); node_list = create_xml_node(NULL, XML_CIB_TAG_STATUS); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { create_node_state_update(node, flags, node_list, source); } if (crm_remote_peer_cache) { g_hash_table_iter_init(&iter, crm_remote_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { create_node_state_update(node, flags, node_list, source); } } controld_update_cib(XML_CIB_TAG_STATUS, node_list, cib_scope_local, crmd_node_update_complete); } free_xml(node_list); } static void cib_quorum_update_complete(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { fsa_data_t *msg_data = NULL; if (rc == pcmk_ok) { crm_trace("Quorum update %d complete", call_id); } else { crm_err("Quorum update %d failed: %s (%d)", call_id, pcmk_strerror(rc), rc); crm_log_xml_debug(msg, "failed"); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } void crm_update_quorum(gboolean quorum, gboolean force_update) { bool has_quorum = pcmk_is_set(controld_globals.flags, controld_has_quorum); if (quorum) { controld_set_global_flags(controld_ever_had_quorum); } else if (pcmk_all_flags_set(controld_globals.flags, controld_ever_had_quorum |controld_no_quorum_suicide)) { pcmk__panic(__func__); } if (AM_I_DC && ((has_quorum && !quorum) || (!has_quorum && quorum) || force_update)) { xmlNode *update = NULL; update = create_xml_node(NULL, XML_TAG_CIB); crm_xml_add_int(update, PCMK_XA_HAVE_QUORUM, quorum); crm_xml_add(update, PCMK_XA_DC_UUID, controld_globals.our_uuid); crm_debug("Updating quorum status to %s", pcmk__btoa(quorum)); controld_update_cib(XML_TAG_CIB, update, cib_scope_local, cib_quorum_update_complete); free_xml(update); /* Quorum changes usually cause a new transition via other activity: * quorum gained via a node joining will abort via the node join, * and quorum lost via a node leaving will usually abort via resource * activity and/or fencing. * * However, it is possible that nothing else causes a transition (e.g. * someone forces quorum via corosync-cmaptcl, or quorum is lost due to * a node in standby shutting down cleanly), so here ensure a new * transition is triggered. */ if (quorum) { /* If quorum was gained, abort after a short delay, in case multiple * nodes are joining around the same time, so the one that brings us * to quorum doesn't cause all the remaining ones to be fenced. */ abort_after_delay(INFINITY, pcmk__graph_restart, "Quorum gained", 5000); } else { abort_transition(INFINITY, pcmk__graph_restart, "Quorum lost", NULL); } } if (quorum) { controld_set_global_flags(controld_has_quorum); } else { controld_clear_global_flags(controld_has_quorum); } } diff --git a/daemons/controld/controld_messages.c b/daemons/controld/controld_messages.c index 9e94e35a88..1928aa125f 100644 --- a/daemons/controld/controld_messages.c +++ b/daemons/controld/controld_messages.c @@ -1,1337 +1,1337 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include extern void crm_shutdown(int nsig); static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause); static void handle_response(xmlNode *stored_msg); static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause); static enum crmd_fsa_input handle_shutdown_request(xmlNode *stored_msg); static void send_msg_via_ipc(xmlNode * msg, const char *sys); /* debug only, can wrap all it likes */ static int last_data_id = 0; void register_fsa_error_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, fsa_data_t * cur_data, void *new_data, const char *raised_from) { /* save the current actions if any */ if (controld_globals.fsa_actions != A_NOTHING) { register_fsa_input_adv(cur_data ? cur_data->fsa_cause : C_FSA_INTERNAL, I_NULL, cur_data ? cur_data->data : NULL, controld_globals.fsa_actions, TRUE, __func__); } /* reset the action list */ crm_info("Resetting the current action list"); fsa_dump_actions(controld_globals.fsa_actions, "Drop"); controld_globals.fsa_actions = A_NOTHING; /* register the error */ register_fsa_input_adv(cause, input, new_data, A_NOTHING, TRUE, raised_from); } void register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input, void *data, uint64_t with_actions, gboolean prepend, const char *raised_from) { unsigned old_len = g_list_length(controld_globals.fsa_message_queue); fsa_data_t *fsa_data = NULL; if (raised_from == NULL) { raised_from = ""; } if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) { /* no point doing anything */ crm_err("Cannot add entry to queue: no input and no action"); return; } if (input == I_WAIT_FOR_EVENT) { controld_set_global_flags(controld_fsa_is_stalled); crm_debug("Stalling the FSA pending further input: source=%s cause=%s data=%p queue=%d", raised_from, fsa_cause2string(cause), data, old_len); if (old_len > 0) { fsa_dump_queue(LOG_TRACE); prepend = FALSE; } if (data == NULL) { controld_set_fsa_action_flags(with_actions); fsa_dump_actions(with_actions, "Restored"); return; } /* Store everything in the new event and reset * controld_globals.fsa_actions */ with_actions |= controld_globals.fsa_actions; controld_globals.fsa_actions = A_NOTHING; } last_data_id++; crm_trace("%s %s FSA input %d (%s) due to %s, %s data", raised_from, (prepend? "prepended" : "appended"), last_data_id, fsa_input2string(input), fsa_cause2string(cause), (data? "with" : "without")); fsa_data = calloc(1, sizeof(fsa_data_t)); fsa_data->id = last_data_id; fsa_data->fsa_input = input; fsa_data->fsa_cause = cause; fsa_data->origin = raised_from; fsa_data->data = NULL; fsa_data->data_type = fsa_dt_none; fsa_data->actions = with_actions; if (with_actions != A_NOTHING) { crm_trace("Adding actions %.16llx to input", (unsigned long long) with_actions); } if (data != NULL) { switch (cause) { case C_FSA_INTERNAL: case C_CRMD_STATUS_CALLBACK: case C_IPC_MESSAGE: case C_HA_MESSAGE: CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL, crm_err("Bogus data from %s", raised_from)); crm_trace("Copying %s data from %s as cluster message data", fsa_cause2string(cause), raised_from); fsa_data->data = copy_ha_msg_input(data); fsa_data->data_type = fsa_dt_ha_msg; break; case C_LRM_OP_CALLBACK: crm_trace("Copying %s data from %s as lrmd_event_data_t", fsa_cause2string(cause), raised_from); fsa_data->data = lrmd_copy_event((lrmd_event_data_t *) data); fsa_data->data_type = fsa_dt_lrm; break; case C_TIMER_POPPED: case C_SHUTDOWN: case C_UNKNOWN: case C_STARTUP: crm_crit("Copying %s data (from %s) is not yet implemented", fsa_cause2string(cause), raised_from); crmd_exit(CRM_EX_SOFTWARE); break; } } /* make sure to free it properly later */ if (prepend) { controld_globals.fsa_message_queue = g_list_prepend(controld_globals.fsa_message_queue, fsa_data); } else { controld_globals.fsa_message_queue = g_list_append(controld_globals.fsa_message_queue, fsa_data); } crm_trace("FSA message queue length is %d", g_list_length(controld_globals.fsa_message_queue)); /* fsa_dump_queue(LOG_TRACE); */ if (old_len == g_list_length(controld_globals.fsa_message_queue)) { crm_err("Couldn't add message to the queue"); } if (input != I_WAIT_FOR_EVENT) { controld_trigger_fsa(); } } void fsa_dump_queue(int log_level) { int offset = 0; for (GList *iter = controld_globals.fsa_message_queue; iter != NULL; iter = iter->next) { fsa_data_t *data = (fsa_data_t *) iter->data; do_crm_log_unlikely(log_level, "queue[%d.%d]: input %s raised by %s(%p.%d)\t(cause=%s)", offset++, data->id, fsa_input2string(data->fsa_input), data->origin, data->data, data->data_type, fsa_cause2string(data->fsa_cause)); } } ha_msg_input_t * copy_ha_msg_input(ha_msg_input_t * orig) { ha_msg_input_t *copy = calloc(1, sizeof(ha_msg_input_t)); CRM_ASSERT(copy != NULL); copy->msg = (orig && orig->msg)? copy_xml(orig->msg) : NULL; copy->xml = get_message_xml(copy->msg, F_CRM_DATA); return copy; } void delete_fsa_input(fsa_data_t * fsa_data) { lrmd_event_data_t *op = NULL; xmlNode *foo = NULL; if (fsa_data == NULL) { return; } crm_trace("About to free %s data", fsa_cause2string(fsa_data->fsa_cause)); if (fsa_data->data != NULL) { switch (fsa_data->data_type) { case fsa_dt_ha_msg: delete_ha_msg_input(fsa_data->data); break; case fsa_dt_xml: foo = fsa_data->data; free_xml(foo); break; case fsa_dt_lrm: op = (lrmd_event_data_t *) fsa_data->data; lrmd_free_event(op); break; case fsa_dt_none: if (fsa_data->data != NULL) { crm_err("Don't know how to free %s data from %s", fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin); crmd_exit(CRM_EX_SOFTWARE); } break; } crm_trace("%s data freed", fsa_cause2string(fsa_data->fsa_cause)); } free(fsa_data); } /* returns the next message */ fsa_data_t * get_message(void) { fsa_data_t *message = (fsa_data_t *) controld_globals.fsa_message_queue->data; controld_globals.fsa_message_queue = g_list_remove(controld_globals.fsa_message_queue, message); crm_trace("Processing input %d", message->id); return message; } void * fsa_typed_data_adv(fsa_data_t * fsa_data, enum fsa_data_type a_type, const char *caller) { void *ret_val = NULL; if (fsa_data == NULL) { crm_err("%s: No FSA data available", caller); } else if (fsa_data->data == NULL) { crm_err("%s: No message data available. Origin: %s", caller, fsa_data->origin); } else if (fsa_data->data_type != a_type) { crm_crit("%s: Message data was the wrong type! %d vs. requested=%d. Origin: %s", caller, fsa_data->data_type, a_type, fsa_data->origin); CRM_ASSERT(fsa_data->data_type == a_type); } else { ret_val = fsa_data->data; } return ret_val; } /* A_MSG_ROUTE */ void do_msg_route(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); route_message(msg_data->fsa_cause, input->msg); } void route_message(enum crmd_fsa_cause cause, xmlNode * input) { ha_msg_input_t fsa_input; enum crmd_fsa_input result = I_NULL; fsa_input.msg = input; CRM_CHECK(cause == C_IPC_MESSAGE || cause == C_HA_MESSAGE, return); /* try passing the buck first */ if (relay_message(input, cause == C_IPC_MESSAGE)) { return; } /* handle locally */ result = handle_message(input, cause); /* done or process later? */ switch (result) { case I_NULL: case I_CIB_OP: case I_ROUTER: case I_NODE_JOIN: case I_JOIN_REQUEST: case I_JOIN_RESULT: break; default: /* Defering local processing of message */ register_fsa_input_later(cause, result, &fsa_input); return; } if (result != I_NULL) { /* add to the front of the queue */ register_fsa_input(cause, result, &fsa_input); } } gboolean relay_message(xmlNode * msg, gboolean originated_locally) { enum crm_ais_msg_types dest = crm_msg_ais; bool is_for_dc = false; bool is_for_dcib = false; bool is_for_te = false; bool is_for_crm = false; bool is_for_cib = false; bool is_local = false; bool broadcast = false; const char *host_to = NULL; const char *sys_to = NULL; const char *sys_from = NULL; const char *type = NULL; const char *task = NULL; const char *ref = NULL; crm_node_t *node_to = NULL; CRM_CHECK(msg != NULL, return TRUE); host_to = crm_element_value(msg, F_CRM_HOST_TO); sys_to = crm_element_value(msg, F_CRM_SYS_TO); sys_from = crm_element_value(msg, F_CRM_SYS_FROM); type = crm_element_value(msg, F_TYPE); task = crm_element_value(msg, F_CRM_TASK); ref = crm_element_value(msg, XML_ATTR_REFERENCE); broadcast = pcmk__str_empty(host_to); if (ref == NULL) { ref = "without reference ID"; } if (pcmk__str_eq(task, CRM_OP_HELLO, pcmk__str_casei)) { crm_trace("Received hello %s from %s (no processing needed)", ref, pcmk__s(sys_from, "unidentified source")); crm_log_xml_trace(msg, "hello"); return TRUE; } // Require message type (set by create_request()) if (!pcmk__str_eq(type, T_CRM, pcmk__str_casei)) { crm_warn("Ignoring invalid message %s with type '%s' (not '" T_CRM "')", ref, pcmk__s(type, "")); crm_log_xml_trace(msg, "ignored"); return TRUE; } // Require a destination subsystem (also set by create_request()) if (sys_to == NULL) { crm_warn("Ignoring invalid message %s with no " F_CRM_SYS_TO, ref); crm_log_xml_trace(msg, "ignored"); return TRUE; } // Get the message type appropriate to the destination subsystem if (is_corosync_cluster()) { dest = text2msg_type(sys_to); if ((dest < crm_msg_ais) || (dest > crm_msg_stonith_ng)) { /* Unrecognized value, use a sane default * * @TODO Maybe we should bail instead */ dest = crm_msg_crmd; } } is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0); is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0); is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0); is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0); is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0); // Check whether message should be processed locally is_local = false; if (broadcast) { if (is_for_dc || is_for_te) { is_local = false; } else if (is_for_crm) { if (pcmk__strcase_any_of(task, CRM_OP_NODE_INFO, PCMK__CONTROLD_CMD_NODES, NULL)) { /* Node info requests do not specify a host, which is normally * treated as "all hosts", because the whole point is that the * client may not know the local node name. Always handle these * requests locally. */ is_local = true; } else { is_local = !originated_locally; } } else { is_local = true; } } else if (pcmk__str_eq(controld_globals.our_nodename, host_to, pcmk__str_casei)) { is_local = true; } else if (is_for_crm && pcmk__str_eq(task, CRM_OP_LRM_DELETE, pcmk__str_casei)) { xmlNode *msg_data = get_message_xml(msg, F_CRM_DATA); const char *mode = crm_element_value(msg_data, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_casei)) { // Local delete of an offline node's resource history is_local = true; } } // Check whether message should be relayed if (is_for_dc || is_for_dcib || is_for_te) { if (AM_I_DC) { if (is_for_te) { crm_trace("Route message %s locally as transition request", ref); crm_log_xml_trace(msg, sys_to); send_msg_via_ipc(msg, sys_to); return TRUE; // No further processing of message is needed } crm_trace("Route message %s locally as DC request", ref); return FALSE; // More to be done by caller } if (originated_locally && !pcmk__strcase_any_of(sys_from, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { crm_trace("Relay message %s to DC (via %s)", ref, pcmk__s(host_to, "broadcast")); crm_log_xml_trace(msg, "relayed"); if (!broadcast) { node_to = crm_get_peer(0, host_to); } send_cluster_message(node_to, dest, msg, TRUE); return TRUE; } /* Transition engine and scheduler messages are sent only to the DC on * the same node. If we are no longer the DC, discard this message. */ crm_trace("Ignoring message %s because we are no longer DC", ref); crm_log_xml_trace(msg, "ignored"); return TRUE; // No further processing of message is needed } if (is_local) { if (is_for_crm || is_for_cib) { crm_trace("Route message %s locally as controller request", ref); return FALSE; // More to be done by caller } crm_trace("Relay message %s locally to %s", ref, sys_to); crm_log_xml_trace(msg, "IPC-relay"); send_msg_via_ipc(msg, sys_to); return TRUE; } if (!broadcast) { node_to = pcmk__search_cluster_node_cache(0, host_to, NULL); if (node_to == NULL) { crm_warn("Ignoring message %s because node %s is unknown", ref, host_to); crm_log_xml_trace(msg, "ignored"); return TRUE; } } crm_trace("Relay message %s to %s", ref, pcmk__s(host_to, "all peers")); crm_log_xml_trace(msg, "relayed"); send_cluster_message(node_to, dest, msg, TRUE); return TRUE; } // Return true if field contains a positive integer static bool authorize_version(xmlNode *message_data, const char *field, const char *client_name, const char *ref, const char *uuid) { const char *version = crm_element_value(message_data, field); long long version_num; if ((pcmk__scan_ll(version, &version_num, -1LL) != pcmk_rc_ok) || (version_num < 0LL)) { crm_warn("Rejected IPC hello from %s: '%s' is not a valid protocol %s " CRM_XS " ref=%s uuid=%s", client_name, ((version == NULL)? "" : version), field, (ref? ref : "none"), uuid); return false; } return true; } /*! * \internal * \brief Check whether a client IPC message is acceptable * * If a given client IPC message is a hello, "authorize" it by ensuring it has * valid information such as a protocol version, and return false indicating * that nothing further needs to be done with the message. If the message is not * a hello, just return true to indicate it needs further processing. * * \param[in] client_msg XML of IPC message * \param[in,out] curr_client If IPC is not proxied, client that sent message * \param[in] proxy_session If IPC is proxied, the session ID * * \return true if message needs further processing, false if it doesn't */ bool controld_authorize_ipc_message(const xmlNode *client_msg, pcmk__client_t *curr_client, const char *proxy_session) { xmlNode *message_data = NULL; const char *client_name = NULL; const char *op = crm_element_value(client_msg, F_CRM_TASK); const char *ref = crm_element_value(client_msg, XML_ATTR_REFERENCE); const char *uuid = (curr_client? curr_client->id : proxy_session); if (uuid == NULL) { crm_warn("IPC message from client rejected: No client identifier " CRM_XS " ref=%s", (ref? ref : "none")); goto rejected; } if (!pcmk__str_eq(CRM_OP_HELLO, op, pcmk__str_casei)) { // Only hello messages need to be authorized return true; } message_data = get_message_xml(client_msg, F_CRM_DATA); client_name = crm_element_value(message_data, "client_name"); if (pcmk__str_empty(client_name)) { crm_warn("IPC hello from client rejected: No client name", CRM_XS " ref=%s uuid=%s", (ref? ref : "none"), uuid); goto rejected; } if (!authorize_version(message_data, "major_version", client_name, ref, uuid)) { goto rejected; } if (!authorize_version(message_data, "minor_version", client_name, ref, uuid)) { goto rejected; } crm_trace("Validated IPC hello from client %s", client_name); crm_log_xml_trace(client_msg, "hello"); if (curr_client) { curr_client->userdata = strdup(client_name); } controld_trigger_fsa(); return false; rejected: crm_log_xml_trace(client_msg, "rejected"); if (curr_client) { qb_ipcs_disconnect(curr_client->ipcs); } return false; } static enum crmd_fsa_input handle_message(xmlNode *msg, enum crmd_fsa_cause cause) { const char *type = NULL; CRM_CHECK(msg != NULL, return I_NULL); type = crm_element_value(msg, F_CRM_MSG_TYPE); if (pcmk__str_eq(type, PCMK__VALUE_REQUEST, pcmk__str_none)) { return handle_request(msg, cause); } if (pcmk__str_eq(type, PCMK__VALUE_RESPONSE, pcmk__str_none)) { handle_response(msg); return I_NULL; } crm_warn("Ignoring message with unknown " F_CRM_MSG_TYPE " '%s'", pcmk__s(type, "")); crm_log_xml_trace(msg, "bad"); return I_NULL; } static enum crmd_fsa_input handle_failcount_op(xmlNode * stored_msg) { const char *rsc = NULL; const char *uname = NULL; const char *op = NULL; char *interval_spec = NULL; guint interval_ms = 0; gboolean is_remote_node = FALSE; xmlNode *xml_op = get_message_xml(stored_msg, F_CRM_DATA); if (xml_op) { xmlNode *xml_rsc = first_named_child(xml_op, XML_CIB_TAG_RESOURCE); xmlNode *xml_attrs = first_named_child(xml_op, XML_TAG_ATTRS); if (xml_rsc) { rsc = ID(xml_rsc); } if (xml_attrs) { op = crm_element_value(xml_attrs, CRM_META "_" PCMK__META_CLEAR_FAILURE_OP); crm_element_value_ms(xml_attrs, CRM_META "_" PCMK__META_CLEAR_FAILURE_INTERVAL, &interval_ms); } } uname = crm_element_value(xml_op, XML_LRM_ATTR_TARGET); if ((rsc == NULL) || (uname == NULL)) { crm_log_xml_warn(stored_msg, "invalid failcount op"); return I_NULL; } if (crm_element_value(xml_op, XML_LRM_ATTR_ROUTER_NODE)) { is_remote_node = TRUE; } crm_debug("Clearing failures for %s-interval %s on %s " "from attribute manager, CIB, and executor state", pcmk__readable_interval(interval_ms), rsc, uname); if (interval_ms) { interval_spec = crm_strdup_printf("%ums", interval_ms); } update_attrd_clear_failures(uname, rsc, op, interval_spec, is_remote_node); free(interval_spec); controld_cib_delete_last_failure(rsc, uname, op, interval_ms); lrm_clear_last_failure(rsc, uname, op, interval_ms); return I_NULL; } static enum crmd_fsa_input handle_lrm_delete(xmlNode *stored_msg) { const char *mode = NULL; xmlNode *msg_data = get_message_xml(stored_msg, F_CRM_DATA); CRM_CHECK(msg_data != NULL, return I_NULL); /* CRM_OP_LRM_DELETE has two distinct modes. The default behavior is to * relay the operation to the affected node, which will unregister the * resource from the local executor, clear the resource's history from the * CIB, and do some bookkeeping in the controller. * * However, if the affected node is offline, the client will specify * mode="cib" which means the controller receiving the operation should * clear the resource's history from the CIB and nothing else. This is used * to clear shutdown locks. */ mode = crm_element_value(msg_data, PCMK__XA_MODE); if ((mode == NULL) || strcmp(mode, XML_TAG_CIB)) { // Relay to affected node crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else { // Delete CIB history locally (compare with do_lrm_delete()) const char *from_sys = NULL; const char *user_name = NULL; const char *rsc_id = NULL; const char *node = NULL; xmlNode *rsc_xml = NULL; int rc = pcmk_rc_ok; rsc_xml = first_named_child(msg_data, XML_CIB_TAG_RESOURCE); CRM_CHECK(rsc_xml != NULL, return I_NULL); rsc_id = ID(rsc_xml); from_sys = crm_element_value(stored_msg, F_CRM_SYS_FROM); node = crm_element_value(msg_data, XML_LRM_ATTR_TARGET); user_name = pcmk__update_acl_user(stored_msg, F_CRM_USER, NULL); crm_debug("Handling " CRM_OP_LRM_DELETE " for %s on %s locally%s%s " "(clearing CIB resource history only)", rsc_id, node, (user_name? " for user " : ""), (user_name? user_name : "")); rc = controld_delete_resource_history(rsc_id, node, user_name, cib_dryrun|cib_sync_call); if (rc == pcmk_rc_ok) { rc = controld_delete_resource_history(rsc_id, node, user_name, crmd_cib_smart_opt()); } //Notify client and tengine.(Only notify tengine if mode = "cib" and CRM_OP_LRM_DELETE.) if (from_sys) { lrmd_event_data_t *op = NULL; const char *from_host = crm_element_value(stored_msg, F_CRM_HOST_FROM); const char *transition; if (strcmp(from_sys, CRM_SYSTEM_TENGINE)) { transition = crm_element_value(msg_data, XML_ATTR_TRANSITION_KEY); } else { transition = crm_element_value(stored_msg, XML_ATTR_TRANSITION_KEY); } crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "local node"), rsc_id, ((rc == pcmk_rc_ok)? "" : " not")); op = lrmd_new_event(rsc_id, PCMK_ACTION_DELETE, 0); op->type = lrmd_event_exec_complete; op->user_data = strdup(transition? transition : FAKE_TE_ID); op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(PCMK_XA_CRM_FEATURE_SET), strdup(CRM_FEATURE_SET)); controld_rc2event(op, rc); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); controld_trigger_delete_refresh(from_sys, rsc_id); } return I_NULL; } } /*! * \brief Handle a CRM_OP_REMOTE_STATE message by updating remote peer cache * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_remote_state(const xmlNode *msg) { const char *conn_host = NULL; const char *remote_uname = ID(msg); crm_node_t *remote_peer; bool remote_is_up = false; int rc = pcmk_rc_ok; rc = pcmk__xe_get_bool_attr(msg, PCMK__XA_IN_CCM, &remote_is_up); CRM_CHECK(remote_uname && rc == pcmk_rc_ok, return I_NULL); remote_peer = crm_remote_peer_get(remote_uname); CRM_CHECK(remote_peer, return I_NULL); pcmk__update_peer_state(__func__, remote_peer, remote_is_up ? CRM_NODE_MEMBER : CRM_NODE_LOST, 0); conn_host = crm_element_value(msg, PCMK__XA_CONN_HOST); if (conn_host) { pcmk__str_update(&remote_peer->conn_host, conn_host); } else if (remote_peer->conn_host) { free(remote_peer->conn_host); remote_peer->conn_host = NULL; } return I_NULL; } /*! * \brief Handle a CRM_OP_PING message * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_ping(const xmlNode *msg) { const char *value = NULL; xmlNode *ping = NULL; xmlNode *reply = NULL; // Build reply ping = create_xml_node(NULL, XML_CRM_TAG_PING); value = crm_element_value(msg, F_CRM_SYS_TO); crm_xml_add(ping, XML_PING_ATTR_SYSFROM, value); // Add controller state value = fsa_state2string(controld_globals.fsa_state); crm_xml_add(ping, XML_PING_ATTR_CRMDSTATE, value); crm_notice("Current ping state: %s", value); // CTS needs this // Add controller health // @TODO maybe do some checks to determine meaningful status crm_xml_add(ping, XML_PING_ATTR_STATUS, "ok"); // Send reply reply = create_reply(msg, ping); free_xml(ping); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a PCMK__CONTROLD_CMD_NODES message * * \param[in] request Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_list(const xmlNode *request) { GHashTableIter iter; crm_node_t *node = NULL; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Create message data for reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODES); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & node)) { xmlNode *xml = create_xml_node(reply_data, XML_CIB_TAG_NODE); crm_xml_add_ll(xml, PCMK_XA_ID, (long long) node->id); // uint32_t - crm_xml_add(xml, XML_ATTR_UNAME, node->uname); + crm_xml_add(xml, PCMK_XA_UNAME, node->uname); crm_xml_add(xml, PCMK__XA_IN_CCM, node->state); } // Create and send reply reply = create_reply(request, reply_data); free_xml(reply_data); if (reply) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } /*! * \brief Handle a CRM_OP_NODE_INFO request * * \param[in] msg Message XML * * \return Next FSA input */ static enum crmd_fsa_input handle_node_info_request(const xmlNode *msg) { const char *value = NULL; crm_node_t *node = NULL; int node_id = 0; xmlNode *reply = NULL; xmlNode *reply_data = NULL; // Build reply reply_data = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(reply_data, XML_PING_ATTR_SYSFROM, CRM_SYSTEM_CRMD); // Add whether current partition has quorum pcmk__xe_set_bool_attr(reply_data, PCMK_XA_HAVE_QUORUM, pcmk_is_set(controld_globals.flags, controld_has_quorum)); // Check whether client requested node info by ID and/or name crm_element_value_int(msg, PCMK_XA_ID, &node_id); if (node_id < 0) { node_id = 0; } - value = crm_element_value(msg, XML_ATTR_UNAME); + value = crm_element_value(msg, PCMK_XA_UNAME); // Default to local node if none given if ((node_id == 0) && (value == NULL)) { value = controld_globals.our_nodename; } node = pcmk__search_node_caches(node_id, value, CRM_GET_PEER_ANY); if (node) { crm_xml_add(reply_data, PCMK_XA_ID, node->uuid); - crm_xml_add(reply_data, XML_ATTR_UNAME, node->uname); + crm_xml_add(reply_data, PCMK_XA_UNAME, node->uname); crm_xml_add(reply_data, PCMK__XA_CRMD, node->state); pcmk__xe_set_bool_attr(reply_data, XML_NODE_IS_REMOTE, pcmk_is_set(node->flags, crm_remote_node)); } // Send reply reply = create_reply(msg, reply_data); free_xml(reply_data); if (reply != NULL) { (void) relay_message(reply, TRUE); free_xml(reply); } // Nothing further to do return I_NULL; } static void verify_feature_set(xmlNode *msg) { const char *dc_version = crm_element_value(msg, PCMK_XA_CRM_FEATURE_SET); if (dc_version == NULL) { /* All we really know is that the DC feature set is older than 3.1.0, * but that's also all that really matters. */ dc_version = "3.0.14"; } if (feature_set_compatible(dc_version, CRM_FEATURE_SET)) { crm_trace("Local feature set (%s) is compatible with DC's (%s)", CRM_FEATURE_SET, dc_version); } else { crm_err("Local feature set (%s) is incompatible with DC's (%s)", CRM_FEATURE_SET, dc_version); // Nothing is likely to improve without administrator involvement controld_set_fsa_input_flags(R_STAYDOWN); crmd_exit(CRM_EX_FATAL); } } // DC gets own shutdown all-clear static enum crmd_fsa_input handle_shutdown_self_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { // The expected case -- we initiated own shutdown sequence crm_info("Shutting down controller"); return I_STOP; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_casei)) { // Must be logic error -- DC confirming its own unrequested shutdown crm_err("Shutting down controller immediately due to " "unexpected shutdown confirmation"); return I_TERMINATE; } if (controld_globals.fsa_state != S_STOPPING) { // Shouldn't happen -- non-DC confirming unrequested shutdown crm_err("Starting new DC election because %s is " "confirming shutdown we did not request", (host_from? host_from : "another node")); return I_ELECTION; } // Shouldn't happen, but we are already stopping anyway crm_debug("Ignoring unexpected shutdown confirmation from %s", (host_from? host_from : "another node")); return I_NULL; } // Non-DC gets shutdown all-clear from DC static enum crmd_fsa_input handle_shutdown_ack(xmlNode *stored_msg) { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { crm_warn("Ignoring shutdown request without origin specified"); return I_NULL; } if (pcmk__str_eq(host_from, controld_globals.dc_name, pcmk__str_null_matches|pcmk__str_casei)) { if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { crm_info("Shutting down controller after confirmation from %s", host_from); } else { crm_err("Shutting down controller after unexpected " "shutdown request from %s", host_from); controld_set_fsa_input_flags(R_STAYDOWN); } return I_STOP; } crm_warn("Ignoring shutdown request from %s because DC is %s", host_from, controld_globals.dc_name); return I_NULL; } static enum crmd_fsa_input handle_request(xmlNode *stored_msg, enum crmd_fsa_cause cause) { xmlNode *msg = NULL; const char *op = crm_element_value(stored_msg, F_CRM_TASK); /* Optimize this for the DC - it has the most to do */ crm_log_xml_trace(stored_msg, "request"); if (op == NULL) { crm_warn("Ignoring request without " F_CRM_TASK); return I_NULL; } if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { const char *from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_node_t *node = pcmk__search_cluster_node_cache(0, from, NULL); pcmk__update_peer_expected(__func__, node, CRMD_JOINSTATE_DOWN); if(AM_I_DC == FALSE) { return I_NULL; /* Done */ } } /*========== DC-Only Actions ==========*/ if (AM_I_DC) { if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) { return I_NODE_JOIN; } else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) { return I_JOIN_REQUEST; } else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) { return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_self_ack(stored_msg); } else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) { // Another controller wants to shut down its node return handle_shutdown_request(stored_msg); } } /*========== common actions ==========*/ if (strcmp(op, CRM_OP_NOVOTE) == 0) { ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); } else if (strcmp(op, CRM_OP_REMOTE_STATE) == 0) { /* a remote connection host is letting us know the node state */ return handle_remote_state(stored_msg); } else if (strcmp(op, CRM_OP_THROTTLE) == 0) { throttle_update(stored_msg); if (AM_I_DC && (controld_globals.transition_graph != NULL) && !controld_globals.transition_graph->complete) { crm_debug("The throttle changed. Trigger a graph."); trigger_graph(); } return I_NULL; } else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) { return handle_failcount_op(stored_msg); } else if (strcmp(op, CRM_OP_VOTE) == 0) { /* count the vote and decide what to do after that */ ha_msg_input_t fsa_input; fsa_input.msg = stored_msg; register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input, A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __func__); /* Sometimes we _must_ go into S_ELECTION */ if (controld_globals.fsa_state == S_HALT) { crm_debug("Forcing an election from S_HALT"); return I_ELECTION; } } else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) { verify_feature_set(stored_msg); crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_OFFER; } else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) { crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID)); return I_JOIN_RESULT; } else if (strcmp(op, CRM_OP_LRM_DELETE) == 0) { return handle_lrm_delete(stored_msg); } else if ((strcmp(op, CRM_OP_LRM_FAIL) == 0) || (strcmp(op, CRM_OP_LRM_REFRESH) == 0) // @COMPAT || (strcmp(op, CRM_OP_REPROBE) == 0)) { crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD); return I_ROUTER; } else if (strcmp(op, CRM_OP_NOOP) == 0) { return I_NULL; } else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) { crm_shutdown(SIGTERM); /*return I_SHUTDOWN; */ return I_NULL; } else if (strcmp(op, CRM_OP_PING) == 0) { return handle_ping(stored_msg); } else if (strcmp(op, CRM_OP_NODE_INFO) == 0) { return handle_node_info_request(stored_msg); } else if (strcmp(op, CRM_OP_RM_NODE_CACHE) == 0) { int id = 0; const char *name = NULL; crm_element_value_int(stored_msg, PCMK_XA_ID, &id); - name = crm_element_value(stored_msg, XML_ATTR_UNAME); + name = crm_element_value(stored_msg, PCMK_XA_UNAME); if(cause == C_IPC_MESSAGE) { msg = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) { crm_err("Could not instruct peers to remove references to node %s/%u", name, id); } else { crm_notice("Instructing peers to remove references to node %s/%u", name, id); } free_xml(msg); } else { reap_crm_member(id, name); /* If we're forgetting this node, also forget any failures to fence * it, so we don't carry that over to any node added later with the * same name. */ st_fail_count_reset(name); } } else if (strcmp(op, CRM_OP_MAINTENANCE_NODES) == 0) { xmlNode *xml = get_message_xml(stored_msg, F_CRM_DATA); remote_ra_process_maintenance_nodes(xml); } else if (strcmp(op, PCMK__CONTROLD_CMD_NODES) == 0) { return handle_node_list(stored_msg); /*========== (NOT_DC)-Only Actions ==========*/ } else if (!AM_I_DC) { if (strcmp(op, CRM_OP_SHUTDOWN) == 0) { return handle_shutdown_ack(stored_msg); } } else { crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node"); crm_log_xml_err(stored_msg, "Unexpected"); } return I_NULL; } static void handle_response(xmlNode *stored_msg) { const char *op = crm_element_value(stored_msg, F_CRM_TASK); crm_log_xml_trace(stored_msg, "reply"); if (op == NULL) { crm_warn("Ignoring reply without " F_CRM_TASK); } else if (AM_I_DC && strcmp(op, CRM_OP_PECALC) == 0) { // Check whether scheduler answer been superseded by subsequent request const char *msg_ref = crm_element_value(stored_msg, XML_ATTR_REFERENCE); if (msg_ref == NULL) { crm_err("%s - Ignoring calculation with no reference", op); } else if (pcmk__str_eq(msg_ref, controld_globals.fsa_pe_ref, pcmk__str_none)) { ha_msg_input_t fsa_input; controld_stop_sched_timer(); fsa_input.msg = stored_msg; register_fsa_input_later(C_IPC_MESSAGE, I_PE_SUCCESS, &fsa_input); } else { crm_info("%s calculation %s is obsolete", op, msg_ref); } } else if (strcmp(op, CRM_OP_VOTE) == 0 || strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0 || strcmp(op, CRM_OP_SHUTDOWN) == 0) { } else { const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); crm_err("Unexpected response (op=%s, src=%s) sent to the %s", op, host_from, AM_I_DC ? "DC" : "controller"); } } static enum crmd_fsa_input handle_shutdown_request(xmlNode * stored_msg) { /* handle here to avoid potential version issues * where the shutdown message/procedure may have * been changed in later versions. * * This way the DC is always in control of the shutdown */ char *now_s = NULL; const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM); if (host_from == NULL) { /* we're shutting down and the DC */ host_from = controld_globals.our_nodename; } crm_info("Creating shutdown request for %s (state=%s)", host_from, fsa_state2string(controld_globals.fsa_state)); crm_log_xml_trace(stored_msg, "message"); now_s = pcmk__ttoa(time(NULL)); update_attrd(host_from, XML_CIB_ATTR_SHUTDOWN, now_s, NULL, FALSE); free(now_s); /* will be picked up by the TE as long as its running */ return I_NULL; } static void send_msg_via_ipc(xmlNode * msg, const char *sys) { pcmk__client_t *client_channel = NULL; CRM_CHECK(sys != NULL, return); client_channel = pcmk__find_client_by_id(sys); if (crm_element_value(msg, F_CRM_HOST_FROM) == NULL) { crm_xml_add(msg, F_CRM_HOST_FROM, controld_globals.our_nodename); } if (client_channel != NULL) { /* Transient clients such as crmadmin */ pcmk__ipc_send_xml(client_channel, 0, msg, crm_ipc_server_event); } else if (pcmk__str_eq(sys, CRM_SYSTEM_TENGINE, pcmk__str_none)) { xmlNode *data = get_message_xml(msg, F_CRM_DATA); process_te_message(msg, data); } else if (pcmk__str_eq(sys, CRM_SYSTEM_LRMD, pcmk__str_none)) { fsa_data_t fsa_data; ha_msg_input_t fsa_input; fsa_input.msg = msg; fsa_input.xml = get_message_xml(msg, F_CRM_DATA); fsa_data.id = 0; fsa_data.actions = 0; fsa_data.data = &fsa_input; fsa_data.fsa_input = I_MESSAGE; fsa_data.fsa_cause = C_IPC_MESSAGE; fsa_data.origin = __func__; fsa_data.data_type = fsa_dt_ha_msg; do_lrm_invoke(A_LRM_INVOKE, C_IPC_MESSAGE, controld_globals.fsa_state, I_MESSAGE, &fsa_data); } else if (crmd_is_proxy_session(sys)) { crmd_proxy_send(sys, msg); } else { crm_info("Received invalid request: unknown subsystem '%s'", sys); } } void delete_ha_msg_input(ha_msg_input_t * orig) { if (orig == NULL) { return; } free_xml(orig->msg); free(orig); } /*! * \internal * \brief Notify the cluster of a remote node state change * * \param[in] node_name Node's name * \param[in] node_up true if node is up, false if down */ void broadcast_remote_state_message(const char *node_name, bool node_up) { xmlNode *msg = create_request(CRM_OP_REMOTE_STATE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL); crm_info("Notifying cluster of Pacemaker Remote node %s %s", node_name, node_up? "coming up" : "going down"); crm_xml_add(msg, PCMK_XA_ID, node_name); pcmk__xe_set_bool_attr(msg, PCMK__XA_IN_CCM, node_up); if (node_up) { crm_xml_add(msg, PCMK__XA_CONN_HOST, controld_globals.our_nodename); } send_cluster_message(NULL, crm_msg_crmd, msg, TRUE); free_xml(msg); } diff --git a/daemons/controld/controld_te_actions.c b/daemons/controld/controld_te_actions.c index 82f4cc5d39..b14907b70a 100644 --- a/daemons/controld/controld_te_actions.c +++ b/daemons/controld/controld_te_actions.c @@ -1,749 +1,749 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include // lrmd_event_data_t, lrmd_free_event() #include #include #include #include #include static GHashTable *te_targets = NULL; void send_rsc_command(pcmk__graph_action_t *action); static void te_update_job_count(pcmk__graph_action_t *action, int offset); static void te_start_action_timer(const pcmk__graph_t *graph, pcmk__graph_action_t *action) { action->timer = g_timeout_add(action->timeout + graph->network_delay, action_timer_callback, (void *) action); CRM_ASSERT(action->timer != 0); } /*! * \internal * \brief Execute a graph pseudo-action * * \param[in,out] graph Transition graph being executed * \param[in,out] pseudo Pseudo-action to execute * * \return Standard Pacemaker return code */ static int execute_pseudo_action(pcmk__graph_t *graph, pcmk__graph_action_t *pseudo) { const char *task = crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK); /* send to peers as well? */ if (pcmk__str_eq(task, PCMK_ACTION_MAINTENANCE_NODES, pcmk__str_casei)) { GHashTableIter iter; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { xmlNode *cmd = NULL; if (pcmk__str_eq(controld_globals.our_nodename, node->uname, pcmk__str_casei)) { continue; } cmd = create_request(task, pseudo->xml, node->uname, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); send_cluster_message(node, crm_msg_crmd, cmd, FALSE); free_xml(cmd); } remote_ra_process_maintenance_nodes(pseudo->xml); } else { /* Check action for Pacemaker Remote node side effects */ remote_ra_process_pseudo(pseudo->xml); } crm_debug("Pseudo-action %d (%s) fired and confirmed", pseudo->id, crm_element_value(pseudo->xml, XML_LRM_ATTR_TASK_KEY)); te_action_confirmed(pseudo, graph); return pcmk_rc_ok; } static int get_target_rc(pcmk__graph_action_t *action) { int exit_status; pcmk__scan_min_int(crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC), &exit_status, 0); return exit_status; } /*! * \internal * \brief Execute a cluster action from a transition graph * * \param[in,out] graph Transition graph being executed * \param[in,out] action Cluster action to execute * * \return Standard Pacemaker return code */ static int execute_cluster_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { char *counter = NULL; xmlNode *cmd = NULL; gboolean is_local = FALSE; const char *id = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; id = ID(action->xml); CRM_CHECK(!pcmk__str_empty(id), return EPROTO); task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); CRM_CHECK(!pcmk__str_empty(task), return EPROTO); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(!pcmk__str_empty(on_node), return pcmk_rc_node_unknown); router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (router_node == NULL) { router_node = on_node; if (pcmk__str_eq(task, PCMK_ACTION_LRM_DELETE, pcmk__str_none)) { const char *mode = crm_element_value(action->xml, PCMK__XA_MODE); if (pcmk__str_eq(mode, XML_TAG_CIB, pcmk__str_none)) { router_node = controld_globals.our_nodename; } } } if (pcmk__str_eq(router_node, controld_globals.our_nodename, pcmk__str_casei)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_info("Handling controller request '%s' (%s on %s)%s%s", id, task, on_node, (is_local? " locally" : ""), (no_wait? " without waiting" : "")); if (is_local && pcmk__str_eq(task, PCMK_ACTION_DO_SHUTDOWN, pcmk__str_none)) { /* defer until everything else completes */ crm_info("Controller request '%s' is a local shutdown", id); graph->completion_action = pcmk__graph_shutdown; graph->abort_reason = "local shutdown"; te_action_confirmed(action, graph); return pcmk_rc_ok; } else if (pcmk__str_eq(task, PCMK_ACTION_DO_SHUTDOWN, pcmk__str_none)) { crm_node_t *peer = crm_get_peer(0, router_node); pcmk__update_peer_expected(__func__, peer, CRMD_JOINSTATE_DOWN); } cmd = create_request(task, action->xml, router_node, CRM_SYSTEM_CRMD, CRM_SYSTEM_TENGINE, NULL); counter = pcmk__transition_key(controld_globals.transition_graph->id, action->id, get_target_rc(action), controld_globals.te_uuid); crm_xml_add(cmd, XML_ATTR_TRANSITION_KEY, counter); rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_crmd, cmd, TRUE); free(counter); free_xml(cmd); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return ECOMM; } else if (no_wait) { te_action_confirmed(action, graph); } else { if (action->timeout <= 0) { crm_err("Action %d: %s on %s had an invalid timeout (%dms). Using %ums instead", action->id, task, on_node, action->timeout, graph->network_delay); action->timeout = (int) graph->network_delay; } te_start_action_timer(graph, action); } return pcmk_rc_ok; } /*! * \internal * \brief Synthesize an executor event for a resource action timeout * * \param[in] action Resource action that timed out * \param[in] target_rc Expected result of action that timed out * * Synthesize an executor event for a resource action timeout. (If the executor * gets a timeout while waiting for a resource action to complete, that will be * reported via the usual callback. This timeout means we didn't hear from the * executor itself or the controller that relayed the action to the executor.) * * \return Newly created executor event for result of \p action * \note The caller is responsible for freeing the return value using * lrmd_free_event(). */ static lrmd_event_data_t * synthesize_timeout_event(const pcmk__graph_action_t *action, int target_rc) { lrmd_event_data_t *op = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *reason = NULL; char *dynamic_reason = NULL; if (pcmk__str_eq(target, get_local_node_name(), pcmk__str_casei)) { reason = "Local executor did not return result in time"; } else { const char *router_node = NULL; router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if (router_node == NULL) { router_node = target; } dynamic_reason = crm_strdup_printf("Controller on %s did not return " "result in time", router_node); reason = dynamic_reason; } op = pcmk__event_from_graph_action(NULL, action, PCMK_EXEC_TIMEOUT, PCMK_OCF_UNKNOWN_ERROR, reason); op->call_id = -1; op->user_data = pcmk__transition_key(controld_globals.transition_graph->id, action->id, target_rc, controld_globals.te_uuid); free(dynamic_reason); return op; } static void controld_record_action_event(pcmk__graph_action_t *action, lrmd_event_data_t *op) { cib_t *cib_conn = controld_globals.cib_conn; xmlNode *state = NULL; xmlNode *rsc = NULL; xmlNode *action_rsc = NULL; int rc = pcmk_ok; const char *rsc_id = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); const char *target_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID); int target_rc = get_target_rc(action); action_rsc = find_xml_node(action->xml, XML_CIB_TAG_RESOURCE, TRUE); if (action_rsc == NULL) { return; } rsc_id = ID(action_rsc); CRM_CHECK(rsc_id != NULL, crm_log_xml_err(action->xml, "Bad:action"); return); /* update the CIB */ state = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(state, PCMK_XA_ID, target_uuid); - crm_xml_add(state, XML_ATTR_UNAME, target); + crm_xml_add(state, PCMK_XA_UNAME, target); rsc = create_xml_node(state, XML_CIB_TAG_LRM); crm_xml_add(rsc, PCMK_XA_ID, target_uuid); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCES); rsc = create_xml_node(rsc, XML_LRM_TAG_RESOURCE); crm_xml_add(rsc, PCMK_XA_ID, rsc_id); crm_copy_xml_element(action_rsc, rsc, PCMK_XA_TYPE); crm_copy_xml_element(action_rsc, rsc, PCMK_XA_CLASS); crm_copy_xml_element(action_rsc, rsc, PCMK_XA_PROVIDER); pcmk__create_history_xml(rsc, op, CRM_FEATURE_SET, target_rc, target, __func__); rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_STATUS, state, cib_scope_local); fsa_register_cib_callback(rc, NULL, cib_action_updated); free_xml(state); crm_trace("Sent CIB update (call ID %d) for synthesized event of action %d (%s on %s)", rc, action->id, task_uuid, target); pcmk__set_graph_action_flags(action, pcmk__graph_action_sent_update); } void controld_record_action_timeout(pcmk__graph_action_t *action) { lrmd_event_data_t *op = NULL; const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); int target_rc = get_target_rc(action); crm_warn("%s %d: %s on %s timed out", action->xml->name, action->id, task_uuid, target); op = synthesize_timeout_event(action, target_rc); controld_record_action_event(action, op); lrmd_free_event(op); } /*! * \internal * \brief Execute a resource action from a transition graph * * \param[in,out] graph Transition graph being executed * \param[in,out] action Resource action to execute * * \return Standard Pacemaker return code */ static int execute_rsc_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { /* never overwrite stop actions in the CIB with * anything other than completed results * * Writing pending stops makes it look like the * resource is running again */ xmlNode *cmd = NULL; xmlNode *rsc_op = NULL; gboolean rc = TRUE; gboolean no_wait = FALSE; gboolean is_local = FALSE; char *counter = NULL; const char *task = NULL; const char *value = NULL; const char *on_node = NULL; const char *router_node = NULL; const char *task_uuid = NULL; CRM_ASSERT(action != NULL); CRM_ASSERT(action->xml != NULL); pcmk__clear_graph_action_flags(action, pcmk__graph_action_executed); on_node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); CRM_CHECK(!pcmk__str_empty(on_node), crm_err("Corrupted command(id=%s) %s: no node", ID(action->xml), pcmk__s(task, "without task")); return pcmk_rc_node_unknown); rsc_op = action->xml; task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); task_uuid = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); router_node = crm_element_value(rsc_op, XML_LRM_ATTR_ROUTER_NODE); if (!router_node) { router_node = on_node; } counter = pcmk__transition_key(controld_globals.transition_graph->id, action->id, get_target_rc(action), controld_globals.te_uuid); crm_xml_add(rsc_op, XML_ATTR_TRANSITION_KEY, counter); if (pcmk__str_eq(router_node, controld_globals.our_nodename, pcmk__str_casei)) { is_local = TRUE; } value = crm_meta_value(action->params, XML_ATTR_TE_NOWAIT); if (crm_is_true(value)) { no_wait = TRUE; } crm_notice("Initiating %s operation %s%s on %s%s "CRM_XS" action %d", task, task_uuid, (is_local? " locally" : ""), on_node, (no_wait? " without waiting" : ""), action->id); cmd = create_request(CRM_OP_INVOKE_LRM, rsc_op, router_node, CRM_SYSTEM_LRMD, CRM_SYSTEM_TENGINE, NULL); if (is_local) { /* shortcut local resource commands */ ha_msg_input_t data = { .msg = cmd, .xml = rsc_op, }; fsa_data_t msg = { .id = 0, .data = &data, .data_type = fsa_dt_ha_msg, .fsa_input = I_NULL, .fsa_cause = C_FSA_INTERNAL, .actions = A_LRM_INVOKE, .origin = __func__, }; do_lrm_invoke(A_LRM_INVOKE, C_FSA_INTERNAL, controld_globals.fsa_state, I_NULL, &msg); } else { rc = send_cluster_message(crm_get_peer(0, router_node), crm_msg_lrmd, cmd, TRUE); } free(counter); free_xml(cmd); pcmk__set_graph_action_flags(action, pcmk__graph_action_executed); if (rc == FALSE) { crm_err("Action %d failed: send", action->id); return ECOMM; } else if (no_wait) { /* Just mark confirmed. Don't bump the job count only to immediately * decrement it. */ crm_info("Action %d confirmed - no wait", action->id); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(controld_globals.transition_graph, action); trigger_graph(); } else if (pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { crm_debug("Action %d: %s %s on %s(timeout %dms) was already confirmed.", action->id, task, task_uuid, on_node, action->timeout); } else { if (action->timeout <= 0) { crm_err("Action %d: %s %s on %s had an invalid timeout (%dms). Using %ums instead", action->id, task, task_uuid, on_node, action->timeout, graph->network_delay); action->timeout = (int) graph->network_delay; } te_update_job_count(action, 1); te_start_action_timer(graph, action); } return pcmk_rc_ok; } struct te_peer_s { char *name; int jobs; int migrate_jobs; }; static void te_peer_free(gpointer p) { struct te_peer_s *peer = p; free(peer->name); free(peer); } void te_reset_job_counts(void) { GHashTableIter iter; struct te_peer_s *peer = NULL; if(te_targets == NULL) { te_targets = pcmk__strkey_table(NULL, te_peer_free); } g_hash_table_iter_init(&iter, te_targets); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) & peer)) { peer->jobs = 0; peer->migrate_jobs = 0; } } static void te_update_job_count_on(const char *target, int offset, bool migrate) { struct te_peer_s *r = NULL; if(target == NULL || te_targets == NULL) { return; } r = g_hash_table_lookup(te_targets, target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } r->jobs += offset; if(migrate) { r->migrate_jobs += offset; } crm_trace("jobs[%s] = %d", target, r->jobs); } static void te_update_job_count(pcmk__graph_action_t *action, int offset) { const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); if ((action->type != pcmk__rsc_graph_action) || (target == NULL)) { /* No limit on these */ return; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && pcmk__strcase_any_of(task, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL)) { const char *t1 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); const char *t2 = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); te_update_job_count_on(t1, offset, TRUE); te_update_job_count_on(t2, offset, TRUE); return; } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } te_update_job_count_on(target, offset, FALSE); } /*! * \internal * \brief Check whether a graph action is allowed to be executed on a node * * \param[in] graph Transition graph being executed * \param[in] action Graph action being executed * \param[in] target Name of node where action should be executed * * \return true if action is allowed, otherwise false */ static bool allowed_on_node(const pcmk__graph_t *graph, const pcmk__graph_action_t *action, const char *target) { int limit = 0; struct te_peer_s *r = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); const char *id = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); if(target == NULL) { /* No limit on these */ return true; } else if(te_targets == NULL) { return false; } r = g_hash_table_lookup(te_targets, target); limit = throttle_get_job_limit(target); if(r == NULL) { r = calloc(1, sizeof(struct te_peer_s)); r->name = strdup(target); g_hash_table_insert(te_targets, r->name, r); } if(limit <= r->jobs) { crm_trace("Peer %s is over their job limit of %d (%d): deferring %s", target, limit, r->jobs, id); return false; } else if(graph->migration_limit > 0 && r->migrate_jobs >= graph->migration_limit) { if (pcmk__strcase_any_of(task, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL)) { crm_trace("Peer %s is over their migration job limit of %d (%d): deferring %s", target, graph->migration_limit, r->migrate_jobs, id); return false; } } crm_trace("Peer %s has not hit their limit yet. current jobs = %d limit= %d limit", target, r->jobs, limit); return true; } /*! * \internal * \brief Check whether a graph action is allowed to be executed * * \param[in] graph Transition graph being executed * \param[in] action Graph action being executed * * \return true if action is allowed, otherwise false */ static bool graph_action_allowed(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *target = NULL; const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); if (action->type != pcmk__rsc_graph_action) { /* No limit on these */ return true; } /* if we have a router node, this means the action is performing * on a remote node. For now, we count all actions occurring on a * remote node against the job list on the cluster node hosting * the connection resources */ target = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); if ((target == NULL) && pcmk__strcase_any_of(task, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL)) { target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_SOURCE); if (!allowed_on_node(graph, action, target)) { return false; } target = crm_meta_value(action->params, XML_LRM_ATTR_MIGRATE_TARGET); } else if (target == NULL) { target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); } return allowed_on_node(graph, action, target); } /*! * \brief Confirm a graph action (and optionally update graph) * * \param[in,out] action Action to confirm * \param[in,out] graph Update and trigger this graph (if non-NULL) */ void te_action_confirmed(pcmk__graph_action_t *action, pcmk__graph_t *graph) { if (!pcmk_is_set(action->flags, pcmk__graph_action_confirmed)) { if ((action->type == pcmk__rsc_graph_action) && (crm_element_value(action->xml, XML_LRM_ATTR_TARGET) != NULL)) { te_update_job_count(action, -1); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); } if (graph) { pcmk__update_graph(graph, action); trigger_graph(); } } static pcmk__graph_functions_t te_graph_fns = { execute_pseudo_action, execute_rsc_action, execute_cluster_action, controld_execute_fence_action, graph_action_allowed, }; /* * \internal * \brief Register the transitioner's graph functions with \p libpacemaker */ void controld_register_graph_functions(void) { pcmk__set_graph_functions(&te_graph_fns); } void notify_crmd(pcmk__graph_t *graph) { const char *type = "unknown"; enum crmd_fsa_input event = I_NULL; crm_debug("Processing transition completion in state %s", fsa_state2string(controld_globals.fsa_state)); CRM_CHECK(graph->complete, graph->complete = true); switch (graph->completion_action) { case pcmk__graph_wait: type = "stop"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case pcmk__graph_done: type = "done"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { event = I_TE_SUCCESS; } break; case pcmk__graph_restart: type = "restart"; if (controld_globals.fsa_state == S_TRANSITION_ENGINE) { if (controld_get_period_transition_timer() > 0) { controld_stop_transition_timer(); controld_start_transition_timer(); } else { event = I_PE_CALC; } } else if (controld_globals.fsa_state == S_POLICY_ENGINE) { controld_set_fsa_action_flags(A_PE_INVOKE); controld_trigger_fsa(); } break; case pcmk__graph_shutdown: type = "shutdown"; if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) { event = I_STOP; } else { crm_err("We didn't ask to be shut down, yet the scheduler is telling us to"); event = I_TERMINATE; } } crm_debug("Transition %d status: %s - %s", graph->id, type, pcmk__s(graph->abort_reason, "unspecified reason")); graph->abort_reason = NULL; graph->completion_action = pcmk__graph_done; if (event != I_NULL) { register_fsa_input(C_FSA_INTERNAL, event, NULL); } else { controld_trigger_fsa(); } } diff --git a/daemons/fenced/fenced_cib.c b/daemons/fenced/fenced_cib.c index 087621656a..b4d8741ced 100644 --- a/daemons/fenced/fenced_cib.c +++ b/daemons/fenced/fenced_cib.c @@ -1,738 +1,738 @@ /* * Copyright 2009-2024 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 static xmlNode *local_cib = NULL; static cib_t *cib_api = NULL; static bool have_cib_devices = FALSE; /*! * \internal * \brief Check whether a node has a specific attribute name/value * * \param[in] node Name of node to check * \param[in] name Name of an attribute to look for * \param[in] value The value the named attribute needs to be set to in order to be considered a match * * \return TRUE if the locally cached CIB has the specified node attribute */ gboolean node_has_attr(const char *node, const char *name, const char *value) { GString *xpath = NULL; xmlNode *match; CRM_CHECK((local_cib != NULL) && (node != NULL) && (name != NULL) && (value != NULL), return FALSE); /* Search for the node's attributes in the CIB. While the schema allows * multiple sets of instance attributes, and allows instance attributes to * use id-ref to reference values elsewhere, that is intended for resources, * so we ignore that here. */ xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, "//" XML_CIB_TAG_NODES "/" XML_CIB_TAG_NODE - "[@" XML_ATTR_UNAME "='", node, "']/" XML_TAG_ATTR_SETS + "[@" PCMK_XA_UNAME "='", node, "']/" XML_TAG_ATTR_SETS "/" XML_CIB_TAG_NVPAIR "[@" PCMK_XA_NAME "='", name, "' " "and @" PCMK_XA_VALUE "='", value, "']", NULL); match = get_xpath_object((const char *) xpath->str, local_cib, LOG_NEVER); g_string_free(xpath, TRUE); return (match != NULL); } static void add_topology_level(xmlNode *match) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; CRM_CHECK(match != NULL, return); fenced_register_level(match, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_ADD, &result, desc); pcmk__reset_result(&result); free(desc); } static void topology_remove_helper(const char *node, int level) { char *desc = NULL; pcmk__action_result_t result = PCMK__UNKNOWN_RESULT; xmlNode *data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL); crm_xml_add(data, F_STONITH_ORIGIN, __func__); crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level); crm_xml_add(data, XML_ATTR_STONITH_TARGET, node); fenced_unregister_level(data, &desc, &result); fenced_send_level_notification(STONITH_OP_LEVEL_DEL, &result, desc); pcmk__reset_result(&result); free_xml(data); free(desc); } static void remove_topology_level(xmlNode *match) { int index = 0; char *key = NULL; CRM_CHECK(match != NULL, return); key = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); topology_remove_helper(key, index); free(key); } static void register_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); remove_topology_level(match); add_topology_level(match); } } /* Fencing */ void fencing_topology_init(void) { xmlXPathObjectPtr xpathObj = NULL; const char *xpath = "//" XML_TAG_FENCING_LEVEL; crm_trace("Full topology refresh"); free_topology_list(); init_topology_list(); /* Grab everything */ xpathObj = xpath_search(local_cib, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } static void remove_cib_device(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if(match != NULL) { standard = crm_element_value(match, PCMK_XA_CLASS); } if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } rsc_id = crm_element_value(match, PCMK_XA_ID); stonith_device_remove(rsc_id, true); } } #define XPATH_WATCHDOG_TIMEOUT "//" XML_CIB_TAG_NVPAIR \ "[@" PCMK_XA_NAME "='" \ PCMK_OPT_STONITH_WATCHDOG_TIMEOUT "']" static void update_stonith_watchdog_timeout_ms(xmlNode *cib) { long timeout_ms = 0; xmlNode *stonith_watchdog_xml = NULL; const char *value = NULL; stonith_watchdog_xml = get_xpath_object(XPATH_WATCHDOG_TIMEOUT, cib, LOG_NEVER); if (stonith_watchdog_xml) { value = crm_element_value(stonith_watchdog_xml, PCMK_XA_VALUE); } if (value) { timeout_ms = crm_get_msec(value); } if (timeout_ms < 0) { timeout_ms = pcmk__auto_watchdog_timeout(); } stonith_watchdog_timeout_ms = timeout_ms; } /*! * \internal * \brief Update all STONITH device definitions based on current CIB */ static void cib_devices_update(void) { GHashTableIter iter; stonith_device_t *device = NULL; crm_info("Updating devices to version %s.%s.%s", crm_element_value(local_cib, PCMK_XA_ADMIN_EPOCH), crm_element_value(local_cib, PCMK_XA_EPOCH), crm_element_value(local_cib, PCMK_XA_NUM_UPDATES)); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->cib_registered) { device->dirty = TRUE; } } /* have list repopulated if cib has a watchdog-fencing-resource TODO: keep a cached list for queries happening while we are refreshing */ g_list_free_full(stonith_watchdog_targets, free); stonith_watchdog_targets = NULL; fenced_scheduler_run(local_cib); g_hash_table_iter_init(&iter, device_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&device)) { if (device->dirty) { g_hash_table_iter_remove(&iter); } } } static void update_cib_stonith_devices_v1(const char *event, xmlNode * msg) { const char *reason = "none"; gboolean needs_update = FALSE; xmlXPathObjectPtr xpath_obj = NULL; /* process new constraints */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_CONS_TAG_RSC_LOCATION); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; /* Safest and simplest to always recompute */ needs_update = TRUE; reason = "new location constraint"; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpath_obj, lpc); crm_log_xml_trace(match, "new constraint"); } } freeXpathObject(xpath_obj); /* process deletions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { remove_cib_device(xpath_obj); } freeXpathObject(xpath_obj); /* process additions */ xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_RESOURCE); if (numXpathResults(xpath_obj) > 0) { int max = numXpathResults(xpath_obj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { const char *rsc_id = NULL; const char *standard = NULL; xmlNode *match = getXpathResult(xpath_obj, lpc); rsc_id = crm_element_value(match, PCMK_XA_ID); standard = crm_element_value(match, PCMK_XA_CLASS); if (!pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_STONITH, pcmk__str_casei)) { continue; } crm_trace("Fencing resource %s was added or modified", rsc_id); reason = "new resource"; needs_update = TRUE; } } freeXpathObject(xpath_obj); if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } } static void update_cib_stonith_devices_v2(const char *event, xmlNode * msg) { xmlNode *change = NULL; char *reason = NULL; bool needs_update = FALSE; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); const char *shortpath = NULL; if ((op == NULL) || (strcmp(op, "move") == 0) || strstr(xpath, "/"XML_CIB_TAG_STATUS)) { continue; } else if (pcmk__str_eq(op, "delete", pcmk__str_casei) && strstr(xpath, "/"XML_CIB_TAG_RESOURCE)) { const char *rsc_id = NULL; char *search = NULL; char *mutable = NULL; if (strstr(xpath, XML_TAG_ATTR_SETS) || strstr(xpath, XML_TAG_META_SETS)) { needs_update = TRUE; pcmk__str_update(&reason, "(meta) attribute deleted from resource"); break; } pcmk__str_update(&mutable, xpath); rsc_id = strstr(mutable, "primitive[@" PCMK_XA_ID "=\'"); if (rsc_id != NULL) { rsc_id += strlen("primitive[@" PCMK_XA_ID "=\'"); search = strchr(rsc_id, '\''); } if (search != NULL) { *search = 0; stonith_device_remove(rsc_id, true); /* watchdog_device_update called afterwards to fall back to implicit definition if needed */ } else { crm_warn("Ignoring malformed CIB update (resource deletion)"); } free(mutable); } else if (strstr(xpath, "/"XML_CIB_TAG_RESOURCES) || strstr(xpath, "/"XML_CIB_TAG_CONSTRAINTS) || strstr(xpath, "/"XML_CIB_TAG_RSCCONFIG)) { shortpath = strrchr(xpath, '/'); CRM_ASSERT(shortpath); reason = crm_strdup_printf("%s %s", op, shortpath+1); needs_update = TRUE; break; } } if(needs_update) { crm_info("Updating device list from CIB: %s", reason); cib_devices_update(); } else { crm_trace("No updates for device list found in CIB"); } free(reason); } static void update_cib_stonith_devices(const char *event, xmlNode * msg) { int format = 1; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, PCMK_XA_FORMAT, &format); switch(format) { case 1: update_cib_stonith_devices_v1(event, msg); break; case 2: update_cib_stonith_devices_v2(event, msg); break; default: crm_warn("Unknown patch format: %d", format); } } static void watchdog_device_update(void) { if (stonith_watchdog_timeout_ms > 0) { if (!g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) && !stonith_watchdog_targets) { /* getting here watchdog-fencing enabled, no device there yet and reason isn't stonith_watchdog_targets preventing that */ int rc; xmlNode *xml; xml = create_device_registration_xml( STONITH_WATCHDOG_ID, st_namespace_internal, STONITH_WATCHDOG_AGENT, NULL, /* stonith_device_register will add our own name as PCMK_STONITH_HOST_LIST param so we can skip that here */ NULL); rc = stonith_device_register(xml, TRUE); free_xml(xml); if (rc != pcmk_ok) { rc = pcmk_legacy2rc(rc); exit_code = CRM_EX_FATAL; crm_crit("Cannot register watchdog pseudo fence agent: %s", pcmk_rc_str(rc)); stonith_shutdown(0); } } } else if (g_hash_table_lookup(device_list, STONITH_WATCHDOG_ID) != NULL) { /* be silent if no device - todo parameter to stonith_device_remove */ stonith_device_remove(STONITH_WATCHDOG_ID, true); } } /*! * \internal * \brief Query the full CIB * * \return Standard Pacemaker return code */ static int fenced_query_cib(void) { int rc = pcmk_ok; crm_trace("Re-requesting full CIB"); rc = cib_api->cmds->query(cib_api, NULL, &local_cib, cib_scope_local|cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { CRM_ASSERT(local_cib != NULL); } else { crm_err("Couldn't retrieve the CIB: %s " CRM_XS " rc=%d", pcmk_rc_str(rc), rc); } return rc; } static void remove_fencing_topology(xmlXPathObjectPtr xpathObj) { int max = numXpathResults(xpathObj), lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *match = getXpathResult(xpathObj, lpc); CRM_LOG_ASSERT(match != NULL); if (match && crm_element_value(match, XML_DIFF_MARKER)) { /* Deletion */ int index = 0; char *target = stonith_level_key(match, fenced_target_by_unknown); crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index); if (target == NULL) { crm_err("Invalid fencing target in element %s", ID(match)); } else if (index <= 0) { crm_err("Invalid level for %s in element %s", target, ID(match)); } else { topology_remove_helper(target, index); } /* } else { Deal with modifications during the 'addition' stage */ } } } static void update_fencing_topology(const char *event, xmlNode * msg) { int format = 1; const char *xpath; xmlXPathObjectPtr xpathObj = NULL; xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); CRM_ASSERT(patchset); crm_element_value_int(patchset, PCMK_XA_FORMAT, &format); if(format == 1) { /* Process deletions (only) */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); remove_fencing_topology(xpathObj); freeXpathObject(xpathObj); /* Process additions and changes */ xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_FENCING_LEVEL; xpathObj = xpath_search(msg, xpath); register_fencing_topology(xpathObj); freeXpathObject(xpathObj); } else if(format == 2) { xmlNode *change = NULL; int add[] = { 0, 0, 0 }; int del[] = { 0, 0, 0 }; xml_patch_versions(patchset, add, del); for (change = pcmk__xml_first_child(patchset); change != NULL; change = pcmk__xml_next(change)) { const char *op = crm_element_value(change, XML_DIFF_OP); const char *xpath = crm_element_value(change, XML_DIFF_PATH); if(op == NULL) { continue; } else if(strstr(xpath, "/" XML_TAG_FENCING_LEVEL) != NULL) { /* Change to a specific entry */ crm_trace("Handling %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); if(strcmp(op, "move") == 0) { continue; } else if(strcmp(op, "create") == 0) { add_topology_level(change->children); } else if(strcmp(op, "modify") == 0) { xmlNode *match = first_named_child(change, XML_DIFF_RESULT); if(match) { remove_topology_level(match->children); add_topology_level(match->children); } } else if(strcmp(op, "delete") == 0) { /* Nuclear option, all we have is the path and an id... not enough to remove a specific entry */ crm_info("Re-initializing fencing topology after %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else if (strstr(xpath, "/" XML_TAG_FENCING_TOPOLOGY) != NULL) { /* Change to the topology in general */ crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } else if (strstr(xpath, "/" XML_CIB_TAG_CONFIGURATION)) { /* Changes to the whole config section, possibly including the topology as a whild */ if(first_named_child(change, XML_TAG_FENCING_TOPOLOGY) == NULL) { crm_trace("Nothing for us in %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); } else if(strcmp(op, "delete") == 0 || strcmp(op, "create") == 0) { crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s.", op, add[0], add[1], add[2], xpath); fencing_topology_init(); return; } } else { crm_trace("Nothing for us in %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath); } } } else { crm_warn("Unknown patch format: %d", format); } } static void update_cib_cache_cb(const char *event, xmlNode * msg) { long timeout_ms_saved = stonith_watchdog_timeout_ms; bool need_full_refresh = false; if(!have_cib_devices) { crm_trace("Skipping updates until we get a full dump"); return; } else if(msg == NULL) { crm_trace("Missing %s update", event); return; } /* Maintain a local copy of the CIB so that we have full access * to device definitions, location constraints, and node attributes */ if (local_cib != NULL) { int rc = pcmk_ok; xmlNode *patchset = NULL; crm_element_value_int(msg, F_CIB_RC, &rc); if (rc != pcmk_ok) { return; } patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT); rc = xml_apply_patchset(local_cib, patchset, TRUE); switch (rc) { case pcmk_ok: case -pcmk_err_old_data: break; case -pcmk_err_diff_resync: case -pcmk_err_diff_failed: crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; break; default: crm_warn("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc); free_xml(local_cib); local_cib = NULL; } } if (local_cib == NULL) { if (fenced_query_cib() != pcmk_rc_ok) { return; } need_full_refresh = true; } pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); if (timeout_ms_saved != stonith_watchdog_timeout_ms) { need_full_refresh = true; } if (need_full_refresh) { fencing_topology_init(); cib_devices_update(); } else { // Partial refresh update_fencing_topology(event, msg); update_cib_stonith_devices(event, msg); } watchdog_device_update(); } static void init_cib_cache_cb(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { crm_info("Updating device list from CIB"); have_cib_devices = TRUE; local_cib = copy_xml(output); pcmk__refresh_node_caches_from_cib(local_cib); update_stonith_watchdog_timeout_ms(local_cib); fencing_topology_init(); cib_devices_update(); watchdog_device_update(); } static void cib_connection_destroy(gpointer user_data) { if (stonith_shutdown_flag) { crm_info("Connection to the CIB manager closed"); return; } else { crm_crit("Lost connection to the CIB manager, shutting down"); } if (cib_api) { cib_api->cmds->signoff(cib_api); } stonith_shutdown(0); } /*! * \internal * \brief Disconnect from CIB manager */ void fenced_cib_cleanup(void) { if (cib_api != NULL) { cib_api->cmds->del_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb); cib__clean_up_connection(&cib_api); } free_xml(local_cib); local_cib = NULL; } void setup_cib(void) { int rc, retries = 0; cib_api = cib_new(); if (cib_api == NULL) { crm_err("No connection to the CIB manager"); return; } do { sleep(retries); rc = cib_api->cmds->signon(cib_api, CRM_SYSTEM_STONITHD, cib_command); } while (rc == -ENOTCONN && ++retries < 5); if (rc != pcmk_ok) { crm_err("Could not connect to the CIB manager: %s (%d)", pcmk_strerror(rc), rc); } else if (pcmk_ok != cib_api->cmds->add_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb)) { crm_err("Could not set CIB notification callback"); } else { rc = cib_api->cmds->query(cib_api, NULL, NULL, cib_scope_local); cib_api->cmds->register_callback(cib_api, rc, 120, FALSE, NULL, "init_cib_cache_cb", init_cib_cache_cb); cib_api->cmds->set_connection_dnotify(cib_api, cib_connection_destroy); crm_info("Watching for fencing topology changes"); } } diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index bf4556c7bb..42229b146b 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,3683 +1,3683 @@ /* * Copyright 2009-2024 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; static 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; /* Whether to perform searches that support the action */ uint32_t support_action_only; }; 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 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(const 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 (-1 means disable static/random fencing delays) int delay_id; char *op; char *origin; char *client; char *client_name; char *remote_op_id; char *target; uint32_t target_nodeid; char *action; char *device; GList *device_list; GList *next_device_iter; // device_list entry for next device to execute 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(const async_command_t *cmd, const pcmk__action_result_t *result); static gboolean is_action_required(const char *action, const stonith_device_t *device) { return (device != NULL) && device->automatic_unfencing && pcmk__str_eq(action, PCMK_ACTION_ON, pcmk__str_none); } static int get_action_delay_max(const stonith_device_t *device, const char *action) { const char *value = NULL; guint delay_max = 0U; if (!pcmk__is_fencing_action(action)) { return 0; } value = g_hash_table_lookup(device->params, PCMK_STONITH_DELAY_MAX); if (value) { pcmk_parse_interval_spec(value, &delay_max); delay_max /= 1000; } return (int) delay_max; } static int get_action_delay_base(const stonith_device_t *device, const char *action, const char *target) { char *hash_value = NULL; guint delay_base = 0U; 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 (target != NULL) { 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(target, val, (size_t)(mapval - val)) == 0) { value = mapval + 1; crm_debug("pcmk_delay_base mapped to %s for %s", value, target); break; } } } if (strchr(value, ':') == 0) { pcmk_parse_interval_spec(value, &delay_base); delay_base /= 1000; } free(valptr); } return (int) 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(const 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, PCMK_ACTION_REBOOT, pcmk__str_none) && !pcmk_is_set(device->flags, st_device_supports_reboot)) { crm_trace("%s doesn't support reboot, using timeout for off instead", device->id); action = PCMK_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; } /*! * \internal * \brief Get the currently executing device for a fencing operation * * \param[in] cmd Fencing operation to check * * \return Currently executing device for \p cmd if any, otherwise NULL */ static stonith_device_t * cmd_device(const async_command_t *cmd) { if ((cmd == NULL) || (cmd->device == NULL) || (device_list == NULL)) { return NULL; } return g_hash_table_lookup(device_list, cmd->device); } /*! * \internal * \brief Return the configured reboot action for a given device * * \param[in] device_id Device ID * * \return Configured reboot action for \p device_id */ const char * fenced_device_reboot_action(const char *device_id) { const char *action = NULL; if ((device_list != NULL) && (device_id != NULL)) { stonith_device_t *device = g_hash_table_lookup(device_list, device_id); if ((device != NULL) && (device->params != NULL)) { action = g_hash_table_lookup(device->params, "pcmk_reboot_action"); } } return pcmk__s(action, PCMK_ACTION_REBOOT); } /*! * \internal * \brief Check whether a given device supports the "on" action * * \param[in] device_id Device ID * * \return true if \p device_id supports "on", otherwise false */ bool fenced_device_supports_on(const char *device_id) { if ((device_list != NULL) && (device_id != NULL)) { stonith_device_t *device = g_hash_table_lookup(device_list, device_id); if (device != NULL) { return pcmk_is_set(device->flags, st_device_supports_on); } } return false; } 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->target); free(cmd->remote_op_id); free(cmd->client); free(cmd->client_name); free(cmd->origin); free(cmd->op); free(cmd); } /*! * \internal * \brief Create a new asynchronous fencing operation from request XML * * \param[in] msg Fencing request XML (from IPC or CPG) * * \return Newly allocated fencing operation on success, otherwise NULL * * \note This asserts on memory errors, so a NULL return indicates an * unparseable message. */ static async_command_t * create_async_command(xmlNode *msg) { xmlNode *op = NULL; async_command_t *cmd = NULL; if (msg == NULL) { return NULL; } op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR); if (op == NULL) { return NULL; } cmd = calloc(1, sizeof(async_command_t)); CRM_ASSERT(cmd != NULL); // All messages must include these cmd->action = crm_element_value_copy(op, F_STONITH_ACTION); cmd->op = crm_element_value_copy(msg, F_STONITH_OPERATION); cmd->client = crm_element_value_copy(msg, F_STONITH_CLIENTID); if ((cmd->action == NULL) || (cmd->op == NULL) || (cmd->client == NULL)) { free_async_command(cmd); return NULL; } 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_DELAY, &(cmd->start_delay)); crm_element_value_int(msg, F_STONITH_TIMEOUT, &(cmd->default_timeout)); cmd->timeout = cmd->default_timeout; cmd->origin = crm_element_value_copy(msg, F_ORIG); cmd->remote_op_id = crm_element_value_copy(msg, F_STONITH_REMOTE_OP_ID); cmd->client_name = crm_element_value_copy(msg, F_STONITH_CLIENTNAME); cmd->target = crm_element_value_copy(op, F_STONITH_TARGET); cmd->device = crm_element_value_copy(op, F_STONITH_DEVICE); cmd->done_cb = st_child_done; // Track in global command list 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->target == NULL)? "" : " targeting "), pcmk__s(cmd->target, ""), 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,out] 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->target == NULL)? "" : " targeting "), pcmk__s(pending_op->target, ""), 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, PCMK_ACTION_STOP, pcmk__str_none)) { 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, PCMK_ACTION_REBOOT, pcmk__str_none) && !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->target == NULL)? "" : " targeting "), pcmk__s(cmd->target, ""), device->id, device->agent); action_str = PCMK_ACTION_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->target, cmd->target_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__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 = cmd_device(cmd); cmd->delay_id = 0; 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->target != NULL)) { crm_node_t *node = crm_get_peer(0, cmd->target); cmd->target_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->target == NULL)? "" : " targeting ", pcmk__s(cmd->target, ""), 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->target == NULL)? "" : " targeting ", pcmk__s(cmd->target, ""), 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->target); 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[dontcall] It doesn't matter here if rand() is predictable 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->target == NULL)? "" : " targeting ", pcmk__s(cmd->target, ""), 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); if (device->on_target_actions != NULL) { g_string_free(device->on_target_actions, TRUE); } 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; } 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, PCMK_XA_NAME); if (pcmk__str_eq(action, PCMK_ACTION_LIST, pcmk__str_none)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_list); } else if (pcmk__str_eq(action, PCMK_ACTION_STATUS, pcmk__str_none)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_status); } else if (pcmk__str_eq(action, PCMK_ACTION_REBOOT, pcmk__str_none)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_reboot); } else if (pcmk__str_eq(action, PCMK_ACTION_ON, pcmk__str_none)) { /* "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; } stonith__set_device_flags(device->flags, device->id, st_device_supports_on); } if ((action != NULL) && pcmk__xe_attr_is_true(match, "on_target")) { pcmk__add_word(&(device->on_target_actions), 64, 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] dev XML containing device parameters */ static GHashTable * xml2device_params(const char *name, const 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, PCMK_ACTION_REBOOT) == 0) { crm_warn("Ignoring %s='reboot' (see " PCMK_OPT_STONITH_ACTION " cluster property instead)", STONITH_ATTR_ACTION_OP); } else if (strcmp(value, PCMK_ACTION_OFF) == 0) { map_action(params, PCMK_ACTION_REBOOT, value); } else { map_action(params, PCMK_ACTION_OFF, value); map_action(params, PCMK_ACTION_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, PCMK_XA_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, PCMK__VALUE_UNFENCING, pcmk__str_casei)) { device->automatic_unfencing = TRUE; } if (is_action_required(PCMK_ACTION_ON, device)) { crm_info("Fencing device '%s' requires unfencing", device->id); } if (device->on_target_actions != NULL) { crm_info("Fencing device '%s' requires actions (%s) to be executed " "on target", device->id, (const char *) device->on_target_actions->str); } 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 *target, 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->target, target); 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(cmd); 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(cmd); 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, PCMK_XA_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(const 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 " PCMK_OPT_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(const 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(const 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 enum fenced_target_by unpack_level_kind(const 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__xe_is(xml, XML_TAG_FENCING_LEVEL)) { 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((msg != NULL) && (result != NULL), return); level = unpack_level_request(msg, &mode, &target, &id, desc); if (level == NULL) { fenced_set_protocol_error(result); return; } // Ensure an ID was given (even the client API adds an ID) if (pcmk__str_empty(ID(level))) { crm_warn("Ignoring registration for topology level without ID"); free(target); crm_log_xml_trace(level, "Bad level"); pcmk__format_result(result, CRM_EX_INVALID_PARAM, PCMK_EXEC_INVALID, "Topology level is invalid without ID"); return; } // Ensure a valid target was specified if (mode == fenced_target_by_unknown) { crm_warn("Ignoring registration for topology level '%s' " "without valid target", 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'", 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), ""), 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), ""), // Client API doesn't add ID to unregistration XML 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, PCMK_ACTION_LIST, pcmk__str_none)) { 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, PCMK_ACTION_MONITOR, pcmk__str_none)) { 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, PCMK_ACTION_MONITOR) == 0)) { // 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) { crm_log_xml_warn(msg, "invalid"); 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) { if (search->support_action_only != st_device_supports_none) { stonith_device_t *dev = g_hash_table_lookup(device_list, device); if (dev && !pcmk_is_set(dev->flags, search->support_action_only)) { return; } } 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 != NULL) && (action != NULL) && (device->on_target_actions != NULL) && (strstr((const char*) device->on_target_actions->str, action) != NULL)) { 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; } /*! * \internal * \brief Check if local node is allowed to execute (possibly remapped) action * * \param[in] device Fence device to check * \param[in] action Fence action to check * \param[in] target Node name of fence target * \param[in] allow_self Whether self-fencing is allowed for this operation * * \return true if local node is allowed to execute \p action or any actions it * might be remapped to, otherwise false */ static bool localhost_is_eligible_with_remap(const stonith_device_t *device, const char *action, const char *target, gboolean allow_self) { // Check exact action if (localhost_is_eligible(device, action, target, allow_self)) { return true; } // Check potential remaps if (pcmk__str_eq(action, PCMK_ACTION_REBOOT, pcmk__str_none)) { /* "reboot" might get remapped to "off" then "on", so even if reboot is * disallowed, return true if either of those is allowed. We'll report * the disallowed actions with the results. We never allow self-fencing * for remapped "on" actions because the target is off at that point. */ if (localhost_is_eligible(device, PCMK_ACTION_OFF, target, allow_self) || localhost_is_eligible(device, PCMK_ACTION_ON, target, FALSE)) { return true; } } return false; } static void can_fence_host_with_device(stonith_device_t *dev, struct device_search_s *search) { gboolean can = FALSE; const char *check_type = "Internal bug"; const char *target = NULL; const char *alias = NULL; const char *dev_id = "Unspecified device"; const char *action = (search == NULL)? NULL : search->action; CRM_CHECK((dev != NULL) && (action != NULL), goto search_report_results); if (dev->id != NULL) { dev_id = dev->id; } target = search->host; if (target == NULL) { can = TRUE; check_type = "No target"; goto search_report_results; } /* Answer immediately if the device does not support the action * or the local node is not allowed to perform it */ if (pcmk__str_eq(action, PCMK_ACTION_ON, pcmk__str_none) && !pcmk_is_set(dev->flags, st_device_supports_on)) { check_type = "Agent does not support 'on'"; goto search_report_results; } else if (!localhost_is_eligible_with_remap(dev, action, target, search->allow_suicide)) { check_type = "This node is not allowed to execute action"; goto search_report_results; } // Check eligibility as specified by pcmk_host_check check_type = target_list_type(dev); alias = g_hash_table_lookup(dev->aliases, target); 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)) { if (pcmk__str_in_list(target, 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, target)) { 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, PCMK_ACTION_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 " PCMK_OPT_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, target, action); schedule_internal_command(__func__, dev, PCMK_ACTION_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 == NULL)? target : 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 " PCMK_OPT_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, target, action); schedule_internal_command(__func__, dev, PCMK_ACTION_STATUS, target, 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; } search_report_results: crm_info("%s is%s eligible to fence (%s) %s%s%s%s: %s", dev_id, (can? "" : " not"), pcmk__s(action, "unspecified action"), pcmk__s(target, "unspecified target"), (alias == NULL)? "" : " (as '", pcmk__s(alias, ""), (alias == NULL)? "" : "')", check_type); search_devices_record_result(search, ((dev == NULL)? NULL : dev_id), 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), uint32_t support_action_only) { 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; search->support_action_only = support_action_only; /* 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, const 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 %ds using %s", action, delay_max, device->id); crm_xml_add_int(xml, F_STONITH_DELAY_MAX, delay_max); } delay_base = get_action_delay_base(device, action, target); if (delay_base > 0) { crm_xml_add_int(xml, F_STONITH_DELAY_BASE, delay_base); } if ((delay_max > 0) && (delay_base == 0)) { crm_trace("Action '%s' has maximum random delay %ds using %s", action, delay_max, device->id); } else if ((delay_max == 0) && (delay_base > 0)) { crm_trace("Action '%s' has a static delay of %ds using %s", action, delay_base, device->id); } else if ((delay_max > 0) && (delay_base > 0)) { crm_trace("Action '%s' has a minimum delay of %ds and a randomly chosen " "maximum delay of %ds 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, const 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, const stonith_device_t *device, const char *target, gboolean allow_suicide) { xmlNode *child = create_xml_node(xml, F_STONITH_ACTION); crm_xml_add(child, PCMK_XA_ID, action); add_action_specific_attributes(child, action, device, target); add_disallowed(child, action, device, target, allow_suicide); } /*! * \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 * \param[in] remote_peer If not NULL, name of peer node to send CPG reply * \param[in,out] client If not NULL, client to send IPC reply */ static void stonith_send_reply(const 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 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, PCMK_XA_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); crm_xml_add_int(dev, F_STONITH_DEVICE_SUPPORT_FLAGS, device->flags); /* 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, PCMK_ACTION_REBOOT, pcmk__str_none)) { crm_trace("%s doesn't support reboot, using values for off instead", device->id); action = PCMK_ACTION_OFF; } /* Add action-specific values if available */ add_action_specific_attributes(dev, action, device, query->target); if (pcmk__str_eq(query->action, PCMK_ACTION_REBOOT, pcmk__str_none)) { /* 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, PCMK_ACTION_OFF, device, query->target, pcmk_is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, PCMK_ACTION_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(const 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->next_device_iter); 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->target == NULL)? LOG_DEBUG : LOG_NOTICE; if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_none)) { output_log_level = LOG_DEBUG; } next = NULL; } else { log_level = (cmd->target == NULL)? LOG_NOTICE : LOG_ERR; if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_none)) { output_log_level = LOG_WARNING; } } // Build the log message piece by piece pcmk__g_strcat(msg, "Operation '", cmd->action, "' ", NULL); if (pid != 0) { g_string_append_printf(msg, "[%d] ", pid); } if (cmd->target != NULL) { pcmk__g_strcat(msg, "targeting ", cmd->target, " ", NULL); } if (cmd->device != NULL) { pcmk__g_strcat(msg, "using ", cmd->device, " ", NULL); } // 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 { pcmk__g_strcat(msg, "could not be executed: ", pcmk_exec_status_str(result->execution_status), NULL); } // Add exit reason and next device if appropriate if (result->exit_reason != NULL) { pcmk__g_strcat(msg, " (", result->exit_reason, ")", NULL); } if (next != NULL) { pcmk__g_strcat(msg, ", retrying with ", next, NULL); } 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(const 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->target, 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->target); 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->target); 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 = cmd_device(cmd); if (device) { crm_trace("Cancel scheduled '%s' action using %s", cmd->action, device->id); device->pending_ops = g_list_remove(device->pending_ops, cmd); } } /*! * \internal * \brief Cancel and reply to any duplicates of a just-completed operation * * Check whether any fencing operations are scheduled to do the same thing as * one that just succeeded. If so, rather than performing the same operation * twice, return the result of this operation for all matching pending commands. * * \param[in,out] cmd Fencing operation that just succeeded * \param[in] result Result of \p cmd * \param[in] pid If nonzero, process ID of agent invocation (for logs) * * \note 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. */ static void reply_to_duplicates(async_command_t *cmd, const pcmk__action_result_t *result, int pid) { GList *next = NULL; for (GList *iter = cmd_list; iter != NULL; iter = next) { async_command_t *cmd_other = iter->data; next = iter->next; // We might delete this entry, so grab next now if (cmd == cmd_other) { continue; } /* A pending operation matches if: * 1. The client connections are different. * 2. The target 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->target, cmd_other->target, pcmk__str_casei) || !pcmk__str_eq(cmd->action, cmd_other->action, pcmk__str_none) || !pcmk__str_eq(cmd->device, cmd_other->device, pcmk__str_casei)) { continue; } crm_notice("Merging fencing action '%s'%s%s originating from " "client %s with identical fencing request from client %s", cmd_other->action, (cmd_other->target == NULL)? "" : " targeting ", pcmk__s(cmd_other->target, ""), cmd_other->client_name, cmd->client_name); // Stop tracking the duplicate, send its result, and cancel it cmd_list = g_list_remove_link(cmd_list, iter); send_async_reply(cmd_other, result, pid, true); cancel_stonith_command(cmd_other); free_async_command(cmd_other); g_list_free_1(iter); } } /*! * \internal * \brief Return the next required device (if any) for an operation * * \param[in,out] cmd Fencing operation that just succeeded * * \return Next device required for action if any, otherwise NULL */ static stonith_device_t * next_required_device(async_command_t *cmd) { for (GList *iter = cmd->next_device_iter; iter != NULL; iter = iter->next) { stonith_device_t *next_device = g_hash_table_lookup(device_list, iter->data); if (is_action_required(cmd->action, next_device)) { /* This is only called for successful actions, so it's OK to skip * non-required devices. */ cmd->next_device_iter = iter->next; return next_device; } } return NULL; } static void st_child_done(int pid, const pcmk__action_result_t *result, void *user_data) { async_command_t *cmd = user_data; stonith_device_t *device = NULL; stonith_device_t *next_device = NULL; CRM_CHECK(cmd != NULL, return); device = cmd_device(cmd); cmd->active_on = NULL; /* The device is ready to do something else now */ if (device) { if (!device->verified && pcmk__result_ok(result) && pcmk__strcase_any_of(cmd->action, PCMK_ACTION_LIST, PCMK_ACTION_MONITOR, PCMK_ACTION_STATUS, NULL)) { device->verified = TRUE; } mainloop_set_trigger(device->work); } if (pcmk__result_ok(result)) { next_device = next_required_device(cmd); } else if ((cmd->next_device_iter != 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->next_device_iter->data); cmd->next_device_iter = cmd->next_device_iter->next; } if (next_device == NULL) { send_async_reply(cmd, result, pid, false); if (pcmk__result_ok(result)) { reply_to_duplicates(cmd, result, pid); } free_async_command(cmd); } else { // This operation requires more fencing log_async_result(cmd, result, pid, next_device->id, false); schedule_stonith_command(cmd, next_device); } } 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->target); 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->target); 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->next_device_iter = 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 = NULL; xmlNode *dev = NULL; CRM_CHECK((msg != NULL) && (result != NULL), return); dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR); cmd = create_async_command(msg); if (cmd == NULL) { crm_log_xml_warn(msg, "invalid"); 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, fenced_support_flag(cmd->action)); } 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(const xmlNode *request, xmlNode *data, const 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(const 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->target); 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) { if (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)) { if (fencing_peer_active(entry) && !pcmk__str_eq(entry->uname, target, pcmk__str_casei)) { crm_notice("Forwarding self-fencing request to %s", entry->uname); return entry->uname; } } crm_warn("Will handle own fencing because no peer can"); } return NULL; } 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'%s%s for %s), " "replaced by op %s ('%s'%s%s for %s)", relay_op->id, relay_op->action, (relay_op->target == NULL)? "" : " targeting ", pcmk__s(relay_op->target, ""), relay_op->client_name, op_id, relay_op->action, (target == NULL)? "" : " targeting ", pcmk__s(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(const 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, st_device_supports_none); 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", NULL, 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); } alternate_host = check_alternate_host(target); } else { crm_notice("Peer %s wants to fence (%s) '%s' with device '%s'", request->peer, action, target, (device == NULL)? "(any)" : device); } if (alternate_host != NULL) { const char *client_id = NULL; remote_fencing_op_t *op = NULL; 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, PCMK_XA_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, PCMK_XA_ID, &node_id); - name = crm_element_value(request->xml, XML_ATTR_UNAME); + name = crm_element_value(request->xml, PCMK_XA_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) { crm_err("Unknown IPC request %s from %s %s", request->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?)", request->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,out] 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,out] 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_copy(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_request(&request); } } diff --git a/include/crm/msg_xml.h b/include/crm/msg_xml.h index 8bd6409748..44672a1583 100644 --- a/include/crm/msg_xml.h +++ b/include/crm/msg_xml.h @@ -1,416 +1,417 @@ /* * Copyright 2004-2024 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__CRM_MSG_XML__H # define PCMK__CRM_MSG_XML__H # include #if !defined(PCMK_ALLOW_DEPRECATED) || (PCMK_ALLOW_DEPRECATED == 1) #include #endif #ifdef __cplusplus extern "C" { #endif /* This file defines constants for various XML syntax (mainly element and * attribute names). * * For consistency, new constants should start with "PCMK_", followed by "XE" * for XML element names, "XA" for XML attribute names, and "META" for meta * attribute names. Old names that don't follow this policy should eventually be * deprecated and replaced with names that do. * * Symbols should be public if the user may specify them somewhere (especially * the CIB). They should be internal if they're used only internally to * Pacemaker (such as daemon IPC/CPG message XML). * * For meta-attributes that can be specified as either XML attributes or nvpair * names, use "META" unless using both "XA" and "META" constants adds clarity. */ /* * XML elements */ #define PCMK_XE_DATE_EXPRESSION "date_expression" #define PCMK_XE_OP "op" #define PCMK_XE_OP_EXPRESSION "op_expression" #define PCMK_XE_RSC_EXPRESSION "rsc_expression" /* This has been deprecated as a CIB element (an alias for with * PCMK_META_PROMOTABLE set to "true") since 2.0.0. */ #define PCMK_XE_PROMOTABLE_LEGACY "master" /* * XML attributes */ #define PCMK_XA_ADMIN_EPOCH "admin_epoch" #define PCMK_XA_CIB_LAST_WRITTEN "cib-last-written" #define PCMK_XA_CLASS "class" #define PCMK_XA_CRM_DEBUG_ORIGIN "crm-debug-origin" #define PCMK_XA_CRM_FEATURE_SET "crm_feature_set" #define PCMK_XA_CRM_TIMESTAMP "crm-timestamp" #define PCMK_XA_DC_UUID "dc-uuid" #define PCMK_XA_DESCRIPTION "description" #define PCMK_XA_EPOCH "epoch" #define PCMK_XA_FORMAT "format" #define PCMK_XA_HAVE_QUORUM "have-quorum" #define PCMK_XA_ID "id" #define PCMK_XA_ID_REF "id-ref" #define PCMK_XA_NAME "name" #define PCMK_XA_NO_QUORUM_PANIC "no-quorum-panic" #define PCMK_XA_NUM_UPDATES "num_updates" #define PCMK_XA_OP "op" #define PCMK_XA_PROVIDER "provider" #define PCMK_XA_REQUEST "request" #define PCMK_XA_TYPE "type" +#define PCMK_XA_UNAME "uname" #define PCMK_XA_UPDATE_CLIENT "update-client" #define PCMK_XA_UPDATE_ORIGIN "update-origin" #define PCMK_XA_UPDATE_USER "update-user" #define PCMK_XA_VALIDATE_WITH "validate-with" #define PCMK_XA_VALUE "value" #define PCMK_XA_VERSION "version" /* * Older constants that don't follow current naming */ # ifndef F_ORIG # define F_ORIG "src" # endif # ifndef F_SEQ # define F_SEQ "seq" # endif # ifndef F_SUBTYPE # define F_SUBTYPE "subt" # endif # ifndef F_TYPE # define F_TYPE "t" # endif # ifndef F_CLIENTNAME # define F_CLIENTNAME "cn" # endif # ifndef F_XML_TAGNAME # define F_XML_TAGNAME "__name__" # endif # ifndef T_CRM # define T_CRM "crmd" # endif # ifndef T_ATTRD # define T_ATTRD "attrd" # endif # define CIB_OPTIONS_FIRST "cib-bootstrap-options" # define F_CRM_DATA "crm_xml" # define F_CRM_TASK "crm_task" # define F_CRM_HOST_TO "crm_host_to" # define F_CRM_MSG_TYPE F_SUBTYPE # define F_CRM_SYS_TO "crm_sys_to" # define F_CRM_SYS_FROM "crm_sys_from" # define F_CRM_HOST_FROM F_ORIG # define F_CRM_REFERENCE XML_ATTR_REFERENCE # define F_CRM_VERSION PCMK_XA_VERSION # define F_CRM_ORIGIN "origin" # define F_CRM_USER "crm_user" # define F_CRM_JOIN_ID "join_id" # define F_CRM_DC_LEAVING "dc-leaving" # define F_CRM_ELECTION_ID "election-id" # define F_CRM_ELECTION_AGE_S "election-age-sec" # define F_CRM_ELECTION_AGE_US "election-age-nano-sec" # define F_CRM_ELECTION_OWNER "election-owner" # define F_CRM_TGRAPH "crm-tgraph-file" # define F_CRM_TGRAPH_INPUT "crm-tgraph-in" # define F_CRM_THROTTLE_MODE "crm-limit-mode" # define F_CRM_THROTTLE_MAX "crm-limit-max" /*---- Common tags/attrs */ # define XML_DIFF_MARKER "__crm_diff_marker__" # define XML_TAG_CIB "cib" # define XML_TAG_FAILED "failed" # define XML_ATTR_TIMEOUT "timeout" # define XML_BOOLEAN_TRUE "true" # define XML_BOOLEAN_FALSE "false" # define XML_TAG_OPTIONS "options" /*---- top level tags/attrs */ # define XML_ATTR_UNAME "uname" # define XML_ATTR_REFERENCE "reference" # define XML_CRM_TAG_PING "ping_response" # define XML_PING_ATTR_STATUS "result" # define XML_PING_ATTR_SYSFROM "crm_subsystem" # define XML_PING_ATTR_CRMDSTATE "crmd_state" # define XML_PING_ATTR_PACEMAKERDSTATE "pacemakerd_state" # define XML_PING_ATTR_PACEMAKERDSTATE_INIT "init" # define XML_PING_ATTR_PACEMAKERDSTATE_STARTINGDAEMONS "starting_daemons" # define XML_PING_ATTR_PACEMAKERDSTATE_WAITPING "wait_for_ping" # define XML_PING_ATTR_PACEMAKERDSTATE_RUNNING "running" # define XML_PING_ATTR_PACEMAKERDSTATE_SHUTTINGDOWN "shutting_down" # define XML_PING_ATTR_PACEMAKERDSTATE_SHUTDOWNCOMPLETE "shutdown_complete" # define XML_PING_ATTR_PACEMAKERDSTATE_REMOTE "remote" # define XML_FAIL_TAG_CIB "failed_update" # define XML_FAILCIB_ATTR_OBJTYPE "object_type" # define XML_FAILCIB_ATTR_OP "operation" # define XML_FAILCIB_ATTR_REASON "reason" /*---- CIB specific tags/attrs */ # define XML_CIB_TAG_SECTION_ALL "all" # define XML_CIB_TAG_CONFIGURATION "configuration" # define XML_CIB_TAG_STATUS "status" # define XML_CIB_TAG_RESOURCES "resources" # define XML_CIB_TAG_NODES "nodes" # define XML_CIB_TAG_CONSTRAINTS "constraints" # define XML_CIB_TAG_CRMCONFIG "crm_config" # define XML_CIB_TAG_OPCONFIG "op_defaults" # define XML_CIB_TAG_RSCCONFIG "rsc_defaults" # define XML_CIB_TAG_ACLS "acls" # define XML_CIB_TAG_ALERTS "alerts" # define XML_CIB_TAG_ALERT "alert" # define XML_CIB_TAG_ALERT_RECIPIENT "recipient" # define XML_CIB_TAG_ALERT_SELECT "select" # define XML_CIB_TAG_ALERT_ATTRIBUTES "select_attributes" # define XML_CIB_TAG_ALERT_FENCING "select_fencing" # define XML_CIB_TAG_ALERT_NODES "select_nodes" # define XML_CIB_TAG_ALERT_RESOURCES "select_resources" # define XML_CIB_TAG_ALERT_ATTR "attribute" # define XML_CIB_TAG_STATE "node_state" # define XML_CIB_TAG_NODE "node" # define XML_CIB_TAG_NVPAIR "nvpair" # define XML_CIB_TAG_PROPSET "cluster_property_set" # define XML_TAG_ATTR_SETS "instance_attributes" # define XML_TAG_META_SETS "meta_attributes" # define XML_TAG_ATTRS "attributes" # define XML_TAG_PARAMS "parameters" # define XML_TAG_PARAM "param" # define XML_TAG_UTILIZATION "utilization" # define XML_TAG_RESOURCE_REF "resource_ref" # define XML_CIB_TAG_RESOURCE "primitive" # define XML_CIB_TAG_GROUP "group" # define XML_CIB_TAG_INCARNATION "clone" # define XML_CIB_TAG_CONTAINER "bundle" # define XML_CIB_TAG_RSC_TEMPLATE "template" # define XML_OP_ATTR_ON_FAIL "on-fail" # define XML_OP_ATTR_START_DELAY "start-delay" # define XML_OP_ATTR_ORIGIN "interval-origin" # define XML_OP_ATTR_PENDING "record-pending" # define XML_OP_ATTR_DIGESTS_ALL "digests-all" # define XML_OP_ATTR_DIGESTS_SECURE "digests-secure" # define XML_CIB_TAG_LRM "lrm" # define XML_LRM_TAG_RESOURCES "lrm_resources" # define XML_LRM_TAG_RESOURCE "lrm_resource" # define XML_LRM_TAG_RSC_OP "lrm_rsc_op" //! \deprecated Do not use (will be removed in a future release) # define XML_CIB_ATTR_REPLACE "replace" # define XML_CIB_ATTR_PRIORITY "priority" # define XML_NODE_IS_REMOTE "remote_node" # define XML_NODE_IS_FENCED "node_fenced" # define XML_NODE_IS_MAINTENANCE "node_in_maintenance" # define XML_CIB_ATTR_SHUTDOWN "shutdown" /* Aside from being an old name for the executor, LRM is a misnomer here because * the controller and scheduler use these to track actions, which are not always * executor operations. */ // XML attribute that takes interval specification (user-facing configuration) # define XML_LRM_ATTR_INTERVAL "interval" // XML attribute that takes interval in milliseconds (daemon APIs) // (identical value as above, but different constant allows clearer code intent) # define XML_LRM_ATTR_INTERVAL_MS XML_LRM_ATTR_INTERVAL # define XML_LRM_ATTR_TASK "operation" # define XML_LRM_ATTR_TASK_KEY "operation_key" # define XML_LRM_ATTR_TARGET "on_node" # define XML_LRM_ATTR_TARGET_UUID "on_node_uuid" /*! Actions to be executed on Pacemaker Remote nodes are routed through the * controller on the cluster node hosting the remote connection. That cluster * node is considered the router node for the action. */ # define XML_LRM_ATTR_ROUTER_NODE "router_node" # define XML_LRM_ATTR_RSCID "rsc-id" # define XML_LRM_ATTR_OPSTATUS "op-status" # define XML_LRM_ATTR_RC "rc-code" # define XML_LRM_ATTR_CALLID "call-id" # define XML_LRM_ATTR_OP_DIGEST "op-digest" # define XML_LRM_ATTR_OP_RESTART "op-force-restart" # define XML_LRM_ATTR_OP_SECURE "op-secure-params" # define XML_LRM_ATTR_RESTART_DIGEST "op-restart-digest" # define XML_LRM_ATTR_SECURE_DIGEST "op-secure-digest" # define XML_LRM_ATTR_EXIT_REASON "exit-reason" # define XML_RSC_OP_LAST_CHANGE "last-rc-change" # define XML_RSC_OP_T_EXEC "exec-time" # define XML_RSC_OP_T_QUEUE "queue-time" # define XML_LRM_ATTR_MIGRATE_SOURCE "migrate_source" # define XML_LRM_ATTR_MIGRATE_TARGET "migrate_target" # define XML_TAG_GRAPH "transition_graph" # define XML_GRAPH_TAG_RSC_OP "rsc_op" # define XML_GRAPH_TAG_PSEUDO_EVENT "pseudo_event" # define XML_GRAPH_TAG_CRM_EVENT "crm_event" # define XML_GRAPH_TAG_DOWNED "downed" # define XML_GRAPH_TAG_MAINTENANCE "maintenance" # define XML_TAG_RULE "rule" # define XML_RULE_ATTR_SCORE "score" # define XML_RULE_ATTR_SCORE_ATTRIBUTE "score-attribute" # define XML_RULE_ATTR_ROLE "role" # define XML_RULE_ATTR_BOOLEAN_OP "boolean-op" # define XML_TAG_EXPRESSION "expression" # define XML_EXPR_ATTR_ATTRIBUTE "attribute" # define XML_EXPR_ATTR_OPERATION "operation" # define XML_EXPR_ATTR_VALUE_SOURCE "value-source" # define XML_CONS_TAG_RSC_DEPEND "rsc_colocation" # define XML_CONS_TAG_RSC_ORDER "rsc_order" # define XML_CONS_TAG_RSC_LOCATION "rsc_location" # define XML_CONS_TAG_RSC_TICKET "rsc_ticket" # define XML_CONS_TAG_RSC_SET "resource_set" # define XML_CONS_ATTR_SYMMETRICAL "symmetrical" # define XML_LOCATION_ATTR_DISCOVERY "resource-discovery" # define XML_COLOC_ATTR_SOURCE "rsc" # define XML_COLOC_ATTR_SOURCE_ROLE "rsc-role" # define XML_COLOC_ATTR_TARGET "with-rsc" # define XML_COLOC_ATTR_TARGET_ROLE "with-rsc-role" # define XML_COLOC_ATTR_NODE_ATTR "node-attribute" # define XML_COLOC_ATTR_INFLUENCE "influence" //! \deprecated Deprecated since 2.1.5 # define XML_COLOC_ATTR_SOURCE_INSTANCE "rsc-instance" //! \deprecated Deprecated since 2.1.5 # define XML_COLOC_ATTR_TARGET_INSTANCE "with-rsc-instance" # define XML_LOC_ATTR_SOURCE "rsc" # define XML_LOC_ATTR_SOURCE_PATTERN "rsc-pattern" # define XML_ORDER_ATTR_FIRST "first" # define XML_ORDER_ATTR_THEN "then" # define XML_ORDER_ATTR_FIRST_ACTION "first-action" # define XML_ORDER_ATTR_THEN_ACTION "then-action" # define XML_ORDER_ATTR_KIND "kind" //! \deprecated Deprecated since 2.1.5 # define XML_ORDER_ATTR_FIRST_INSTANCE "first-instance" //! \deprecated Deprecated since 2.1.5 # define XML_ORDER_ATTR_THEN_INSTANCE "then-instance" # define XML_TICKET_ATTR_TICKET "ticket" # define XML_TICKET_ATTR_LOSS_POLICY "loss-policy" # define XML_NODE_ATTR_RSC_DISCOVERY "resource-discovery-enabled" # define XML_ALERT_ATTR_PATH "path" # define XML_ALERT_ATTR_TIMEOUT "timeout" # define XML_ALERT_ATTR_TSTAMP_FORMAT "timestamp-format" # define XML_CIB_TAG_GENERATION_TUPPLE "generation_tuple" # define XML_ATTR_TRANSITION_MAGIC "transition-magic" # define XML_ATTR_TRANSITION_KEY "transition-key" # define XML_ATTR_TE_NOWAIT "op_no_wait" # define XML_ATTR_TE_TARGET_RC "op_target_rc" # define XML_TAG_TRANSIENT_NODEATTRS "transient_attributes" //! \deprecated Do not use (will be removed in a future release) # define XML_TAG_DIFF_ADDED "diff-added" //! \deprecated Do not use (will be removed in a future release) # define XML_TAG_DIFF_REMOVED "diff-removed" # define XML_ACL_TAG_USER "acl_target" # define XML_ACL_TAG_USERv1 "acl_user" # define XML_ACL_TAG_GROUP "acl_group" # define XML_ACL_TAG_ROLE "acl_role" # define XML_ACL_TAG_PERMISSION "acl_permission" # define XML_ACL_TAG_ROLE_REF "role" # define XML_ACL_TAG_ROLE_REFv1 "role_ref" # define XML_ACL_ATTR_KIND "kind" # define XML_ACL_TAG_READ "read" # define XML_ACL_TAG_WRITE "write" # define XML_ACL_TAG_DENY "deny" # define XML_ACL_ATTR_REF "reference" # define XML_ACL_ATTR_REFv1 "ref" # define XML_ACL_ATTR_TAG "object-type" # define XML_ACL_ATTR_TAGv1 "tag" # define XML_ACL_ATTR_XPATH "xpath" # define XML_ACL_ATTR_ATTRIBUTE "attribute" # define XML_CIB_TAG_TICKETS "tickets" # define XML_CIB_TAG_TICKET_STATE "ticket_state" # define XML_CIB_TAG_TAGS "tags" # define XML_CIB_TAG_TAG "tag" # define XML_CIB_TAG_OBJ_REF "obj_ref" # define XML_TAG_FENCING_TOPOLOGY "fencing-topology" # define XML_TAG_FENCING_LEVEL "fencing-level" # define XML_ATTR_STONITH_INDEX "index" # define XML_ATTR_STONITH_TARGET "target" # define XML_ATTR_STONITH_TARGET_VALUE "target-value" # define XML_ATTR_STONITH_TARGET_PATTERN "target-pattern" # define XML_ATTR_STONITH_TARGET_ATTRIBUTE "target-attribute" # define XML_ATTR_STONITH_DEVICES "devices" # define XML_TAG_DIFF "diff" # define XML_DIFF_VERSION "version" # define XML_DIFF_VSOURCE "source" # define XML_DIFF_VTARGET "target" # define XML_DIFF_CHANGE "change" # define XML_DIFF_LIST "change-list" # define XML_DIFF_ATTR "change-attr" # define XML_DIFF_RESULT "change-result" # define XML_DIFF_OP "operation" # define XML_DIFF_PATH "path" # define XML_DIFF_POSITION "position" # define ID(x) crm_element_value(x, PCMK_XA_ID) #ifdef __cplusplus } #endif #endif diff --git a/lib/cib/cib_attrs.c b/lib/cib/cib_attrs.c index 08aa9926fa..bef810f0b2 100644 --- a/lib/cib/cib_attrs.c +++ b/lib/cib/cib_attrs.c @@ -1,735 +1,735 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static pcmk__output_t * new_output_object(const char *ty) { int rc = pcmk_rc_ok; pcmk__output_t *out = NULL; const char* argv[] = { "", NULL }; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_LOG, PCMK__SUPPORTED_FORMAT_TEXT, { NULL, NULL, NULL } }; pcmk__register_formats(NULL, formats); rc = pcmk__output_new(&out, ty, NULL, (char**)argv); if ((rc != pcmk_rc_ok) || (out == NULL)) { crm_err("Can't out due to internal error: %s", pcmk_rc_str(rc)); return NULL; } return out; } static int find_attr(cib_t *cib, const char *section, const char *node_uuid, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *user_name, xmlNode **result) { int rc = pcmk_rc_ok; const char *xpath_base = NULL; GString *xpath = NULL; xmlNode *xml_search = NULL; const char *set_type = NULL; const char *node_type = NULL; if (attr_set_type) { set_type = attr_set_type; } else { set_type = XML_TAG_ATTR_SETS; } if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { node_uuid = NULL; set_type = XML_CIB_TAG_PROPSET; } else if (pcmk__strcase_any_of(section, XML_CIB_TAG_OPCONFIG, XML_CIB_TAG_RSCCONFIG, NULL)) { node_uuid = NULL; set_type = XML_TAG_META_SETS; } else if (pcmk__str_eq(section, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { node_uuid = NULL; section = XML_CIB_TAG_STATUS; node_type = XML_CIB_TAG_TICKETS; } else if (node_uuid == NULL) { return EINVAL; } xpath_base = pcmk_cib_xpath_for(section); if (xpath_base == NULL) { crm_warn("%s CIB section not known", section); return ENOMSG; } xpath = g_string_sized_new(1024); g_string_append(xpath, xpath_base); if (pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { pcmk__g_strcat(xpath, "//", node_type, NULL); } else if (node_uuid) { const char *node_type = XML_CIB_TAG_NODE; if (pcmk__str_eq(section, XML_CIB_TAG_STATUS, pcmk__str_casei)) { node_type = XML_CIB_TAG_STATE; set_type = XML_TAG_TRANSIENT_NODEATTRS; } pcmk__g_strcat(xpath, "//", node_type, "[@" PCMK_XA_ID "='", node_uuid, "']", NULL); } pcmk__g_strcat(xpath, "//", set_type, NULL); if (set_name) { pcmk__g_strcat(xpath, "[@" PCMK_XA_ID "='", set_name, "']", NULL); } g_string_append(xpath, "//nvpair"); if (attr_id && attr_name) { pcmk__g_strcat(xpath, "[@" PCMK_XA_ID "='", attr_id, "' " "and @" PCMK_XA_NAME "='", attr_name, "']", NULL); } else if (attr_id) { pcmk__g_strcat(xpath, "[@" PCMK_XA_ID "='", attr_id, "']", NULL); } else if (attr_name) { pcmk__g_strcat(xpath, "[@" PCMK_XA_NAME "='", attr_name, "']", NULL); } rc = cib_internal_op(cib, PCMK__CIB_REQUEST_QUERY, NULL, (const char *) xpath->str, NULL, &xml_search, cib_sync_call|cib_scope_local|cib_xpath, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); crm_trace("Query failed for attribute %s (section=%s, node=%s, set=%s, xpath=%s): %s", attr_name, section, pcmk__s(node_uuid, ""), pcmk__s(set_name, ""), (const char *) xpath->str, pcmk_rc_str(rc)); } else { rc = pcmk_rc_ok; crm_log_xml_debug(xml_search, "Match"); } g_string_free(xpath, TRUE); *result = xml_search; return rc; } static int handle_multiples(pcmk__output_t *out, xmlNode *search, const char *attr_name) { if ((search != NULL) && (search->children != NULL)) { xmlNode *child = NULL; out->info(out, "Multiple attributes match name=%s", attr_name); for (child = pcmk__xml_first_child(search); child != NULL; child = pcmk__xml_next(child)) { out->info(out, " Value: %s \t(id=%s)", crm_element_value(child, PCMK_XA_VALUE), ID(child)); } return ENOTUNIQ; } else { return pcmk_rc_ok; } } int cib__update_node_attr(pcmk__output_t *out, cib_t *cib, int call_options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, const char *user_name, const char *node_type) { const char *tag = NULL; int rc = pcmk_rc_ok; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; xmlNode *xml_search = NULL; char *local_attr_id = NULL; char *local_set_name = NULL; CRM_CHECK((out != NULL) && (cib != NULL) && (section != NULL) && ((attr_id != NULL) || (attr_name != NULL)) && (attr_value != NULL), return EINVAL); rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc == pcmk_rc_ok) { if (handle_multiples(out, xml_search, attr_name) == ENOTUNIQ) { free_xml(xml_search); return ENOTUNIQ; } else { pcmk__str_update(&local_attr_id, crm_element_value(xml_search, PCMK_XA_ID)); attr_id = local_attr_id; free_xml(xml_search); goto do_modify; } } else if (rc != ENXIO) { free_xml(xml_search); return rc; /* } else if(attr_id == NULL) { */ /* return EINVAL; */ } else { free_xml(xml_search); crm_trace("%s does not exist, create it", attr_name); if (pcmk__str_eq(section, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { node_uuid = NULL; section = XML_CIB_TAG_STATUS; node_type = XML_CIB_TAG_TICKETS; xml_top = create_xml_node(xml_obj, XML_CIB_TAG_STATUS); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_TICKETS); } else if (pcmk__str_eq(section, XML_CIB_TAG_NODES, pcmk__str_casei)) { if (node_uuid == NULL) { return EINVAL; } if (pcmk__str_eq(node_type, "remote", pcmk__str_casei)) { xml_top = create_xml_node(xml_obj, XML_CIB_TAG_NODES); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_NODE); crm_xml_add(xml_obj, PCMK_XA_TYPE, "remote"); crm_xml_add(xml_obj, PCMK_XA_ID, node_uuid); - crm_xml_add(xml_obj, XML_ATTR_UNAME, node_uuid); + crm_xml_add(xml_obj, PCMK_XA_UNAME, node_uuid); } else { tag = XML_CIB_TAG_NODE; } } else if (pcmk__str_eq(section, XML_CIB_TAG_STATUS, pcmk__str_casei)) { tag = XML_TAG_TRANSIENT_NODEATTRS; if (node_uuid == NULL) { return EINVAL; } xml_top = create_xml_node(xml_obj, XML_CIB_TAG_STATE); crm_xml_add(xml_top, PCMK_XA_ID, node_uuid); xml_obj = xml_top; } else { tag = section; node_uuid = NULL; } if (set_name == NULL) { if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { local_set_name = strdup(CIB_OPTIONS_FIRST); } else if (pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { local_set_name = crm_strdup_printf("%s-%s", section, XML_CIB_TAG_TICKETS); } else if (node_uuid) { local_set_name = crm_strdup_printf("%s-%s", section, node_uuid); if (set_type) { char *tmp_set_name = local_set_name; local_set_name = crm_strdup_printf("%s-%s", tmp_set_name, set_type); free(tmp_set_name); } } else { local_set_name = crm_strdup_printf("%s-options", section); } set_name = local_set_name; } if (attr_id == NULL) { local_attr_id = crm_strdup_printf("%s-%s", set_name, attr_name); crm_xml_sanitize_id(local_attr_id); attr_id = local_attr_id; } else if (attr_name == NULL) { attr_name = attr_id; } crm_trace("Creating %s/%s", section, tag); if (tag != NULL) { xml_obj = create_xml_node(xml_obj, tag); crm_xml_add(xml_obj, PCMK_XA_ID, node_uuid); if (xml_top == NULL) { xml_top = xml_obj; } } if (node_uuid == NULL && !pcmk__str_eq(node_type, XML_CIB_TAG_TICKETS, pcmk__str_casei)) { if (pcmk__str_eq(section, XML_CIB_TAG_CRMCONFIG, pcmk__str_casei)) { xml_obj = create_xml_node(xml_obj, XML_CIB_TAG_PROPSET); } else { xml_obj = create_xml_node(xml_obj, XML_TAG_META_SETS); } } else if (set_type) { xml_obj = create_xml_node(xml_obj, set_type); } else { xml_obj = create_xml_node(xml_obj, XML_TAG_ATTR_SETS); } crm_xml_add(xml_obj, PCMK_XA_ID, set_name); if (xml_top == NULL) { xml_top = xml_obj; } } do_modify: xml_obj = crm_create_nvpair_xml(xml_obj, attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_trace(xml_top, "update_attr"); rc = cib_internal_op(cib, PCMK__CIB_REQUEST_MODIFY, NULL, section, xml_top, NULL, call_options, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); out->err(out, "Error setting %s=%s (section=%s, set=%s): %s", attr_name, attr_value, section, pcmk__s(set_name, ""), pcmk_rc_str(rc)); crm_log_xml_info(xml_top, "Update"); } else { rc = pcmk_rc_ok; } free(local_set_name); free(local_attr_id); free_xml(xml_top); return rc; } int cib__get_node_attrs(pcmk__output_t *out, cib_t *cib, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *user_name, xmlNode **result) { int rc = pcmk_rc_ok; CRM_ASSERT(result != NULL); CRM_CHECK(section != NULL, return EINVAL); *result = NULL; rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, result); if (rc != pcmk_rc_ok) { crm_trace("Query failed for attribute %s (section=%s node=%s set=%s): %s", pcmk__s(attr_name, "with unspecified name"), section, pcmk__s(set_name, ""), pcmk__s(node_uuid, ""), pcmk_rc_str(rc)); } return rc; } int cib__delete_node_attr(pcmk__output_t *out, cib_t *cib, int options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, const char *user_name) { int rc = pcmk_rc_ok; xmlNode *xml_obj = NULL; xmlNode *xml_search = NULL; char *local_attr_id = NULL; CRM_CHECK(section != NULL, return EINVAL); CRM_CHECK(attr_name != NULL || attr_id != NULL, return EINVAL); if (attr_id == NULL) { rc = find_attr(cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc != pcmk_rc_ok || handle_multiples(out, xml_search, attr_name) == ENOTUNIQ) { free_xml(xml_search); return rc; } else { pcmk__str_update(&local_attr_id, crm_element_value(xml_search, PCMK_XA_ID)); attr_id = local_attr_id; free_xml(xml_search); } } xml_obj = crm_create_nvpair_xml(NULL, attr_id, attr_name, attr_value); rc = cib_internal_op(cib, PCMK__CIB_REQUEST_DELETE, NULL, section, xml_obj, NULL, options, user_name); if (rc < 0) { rc = pcmk_legacy2rc(rc); } else { rc = pcmk_rc_ok; out->info(out, "Deleted %s %s: id=%s%s%s%s%s", section, node_uuid ? "attribute" : "option", local_attr_id, set_name ? " set=" : "", set_name ? set_name : "", attr_name ? " name=" : "", attr_name ? attr_name : ""); } free(local_attr_id); free_xml(xml_obj); return rc; } int find_nvpair_attr_delegate(cib_t *cib, const char *attr, const char *section, const char *node_uuid, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, gboolean to_console, char **value, const char *user_name) { pcmk__output_t *out = NULL; xmlNode *xml_search = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = find_attr(cib, section, node_uuid, attr_set_type, set_name, attr_id, attr_name, user_name, &xml_search); if (rc == pcmk_rc_ok) { rc = handle_multiples(out, xml_search, attr_name); if (rc == pcmk_rc_ok) { pcmk__str_update(value, crm_element_value(xml_search, attr)); } } out->finish(out, CRM_EX_OK, true, NULL); free_xml(xml_search); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int update_attr_delegate(cib_t *cib, int call_options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, gboolean to_console, const char *user_name, const char *node_type) { pcmk__output_t *out = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__update_node_attr(out, cib, call_options, section, node_uuid, set_type, set_name, attr_id, attr_name, attr_value, user_name, node_type); out->finish(out, CRM_EX_OK, true, NULL); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int read_attr_delegate(cib_t *cib, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, char **attr_value, gboolean to_console, const char *user_name) { pcmk__output_t *out = NULL; xmlNode *result = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__get_node_attrs(out, cib, section, node_uuid, set_type, set_name, attr_id, attr_name, user_name, &result); if (rc == pcmk_rc_ok) { if (result->children == NULL) { pcmk__str_update(attr_value, crm_element_value(result, PCMK_XA_VALUE)); } else { rc = ENOTUNIQ; } } out->finish(out, CRM_EX_OK, true, NULL); free_xml(result); pcmk__output_free(out); return pcmk_rc2legacy(rc); } int delete_attr_delegate(cib_t *cib, int options, const char *section, const char *node_uuid, const char *set_type, const char *set_name, const char *attr_id, const char *attr_name, const char *attr_value, gboolean to_console, const char *user_name) { pcmk__output_t *out = NULL; int rc = pcmk_ok; out = new_output_object(to_console ? "text" : "log"); if (out == NULL) { return pcmk_err_generic; } rc = cib__delete_node_attr(out, cib, options, section, node_uuid, set_type, set_name, attr_id, attr_name, attr_value, user_name); out->finish(out, CRM_EX_OK, true, NULL); pcmk__output_free(out); return pcmk_rc2legacy(rc); } /*! * \internal * \brief Parse node UUID from search result * * \param[in] result XML search result * \param[out] uuid If non-NULL, where to store parsed UUID * \param[out] is_remote If non-NULL, set TRUE if result is remote node * * \return pcmk_ok if UUID was successfully parsed, -ENXIO otherwise */ static int get_uuid_from_result(const xmlNode *result, char **uuid, int *is_remote) { int rc = -ENXIO; const char *tag; const char *parsed_uuid = NULL; int parsed_is_remote = FALSE; if (result == NULL) { return rc; } /* If there are multiple results, the first is sufficient */ tag = (const char *) (result->name); if (pcmk__str_eq(tag, "xpath-query", pcmk__str_casei)) { result = pcmk__xml_first_child(result); CRM_CHECK(result != NULL, return rc); tag = (const char *) (result->name); } if (pcmk__str_eq(tag, XML_CIB_TAG_NODE, pcmk__str_casei)) { /* Result is tag from section */ if (pcmk__str_eq(crm_element_value(result, PCMK_XA_TYPE), "remote", pcmk__str_casei)) { - parsed_uuid = crm_element_value(result, XML_ATTR_UNAME); + parsed_uuid = crm_element_value(result, PCMK_XA_UNAME); parsed_is_remote = TRUE; } else { parsed_uuid = ID(result); parsed_is_remote = FALSE; } } else if (pcmk__str_eq(tag, XML_CIB_TAG_RESOURCE, pcmk__str_casei)) { /* Result is for ocf:pacemaker:remote resource */ parsed_uuid = ID(result); parsed_is_remote = TRUE; } else if (pcmk__str_eq(tag, XML_CIB_TAG_NVPAIR, pcmk__str_casei)) { /* Result is PCMK_META_REMOTE_NODE parameter of for guest * node */ parsed_uuid = crm_element_value(result, PCMK_XA_VALUE); parsed_is_remote = TRUE; } else if (pcmk__str_eq(tag, XML_CIB_TAG_STATE, pcmk__str_casei)) { /* Result is tag from section */ - parsed_uuid = crm_element_value(result, XML_ATTR_UNAME); + parsed_uuid = crm_element_value(result, PCMK_XA_UNAME); if (pcmk__xe_attr_is_true(result, XML_NODE_IS_REMOTE)) { parsed_is_remote = TRUE; } } if (parsed_uuid) { if (uuid) { *uuid = strdup(parsed_uuid); } if (is_remote) { *is_remote = parsed_is_remote; } rc = pcmk_ok; } return rc; } /* Search string to find a node by name, as: * - cluster or remote node in nodes section * - remote node in resources section * - guest node in resources section * - orphaned remote node or bundle guest node in status section */ #define XPATH_UPPER_TRANS "ABCDEFGHIJKLMNOPQRSTUVWXYZ" #define XPATH_LOWER_TRANS "abcdefghijklmnopqrstuvwxyz" #define XPATH_NODE \ "/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_NODES \ - "/" XML_CIB_TAG_NODE "[translate(@" XML_ATTR_UNAME ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ + "/" XML_CIB_TAG_NODE "[translate(@" PCMK_XA_UNAME ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "/" XML_CIB_TAG_RESOURCE \ "[@class='ocf'][@provider='pacemaker'][@type='remote'][translate(@id,'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \ "/" XML_CIB_TAG_RESOURCE "/" XML_TAG_META_SETS "/" XML_CIB_TAG_NVPAIR \ "[@name='" PCMK_META_REMOTE_NODE "'][translate(@value,'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" \ "|/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS "/" XML_CIB_TAG_STATE \ "[@" XML_NODE_IS_REMOTE "='true'][translate(@" PCMK_XA_ID ",'" XPATH_UPPER_TRANS "','" XPATH_LOWER_TRANS "') ='%s']" int query_node_uuid(cib_t * the_cib, const char *uname, char **uuid, int *is_remote_node) { int rc = pcmk_ok; char *xpath_string; xmlNode *xml_search = NULL; char *host_lowercase = NULL; CRM_ASSERT(uname != NULL); host_lowercase = g_ascii_strdown(uname, -1); if (uuid) { *uuid = NULL; } if (is_remote_node) { *is_remote_node = FALSE; } xpath_string = crm_strdup_printf(XPATH_NODE, host_lowercase, host_lowercase, host_lowercase, host_lowercase); if (cib_internal_op(the_cib, PCMK__CIB_REQUEST_QUERY, NULL, xpath_string, NULL, &xml_search, cib_sync_call|cib_scope_local|cib_xpath, NULL) == pcmk_ok) { rc = get_uuid_from_result(xml_search, uuid, is_remote_node); } else { rc = -ENXIO; } free(xpath_string); free_xml(xml_search); g_free(host_lowercase); if (rc != pcmk_ok) { crm_debug("Could not map node name '%s' to a UUID: %s", uname, pcmk_strerror(rc)); } else { crm_info("Mapped node name '%s' to UUID %s", uname, (uuid? *uuid : "")); } return rc; } int query_node_uname(cib_t * the_cib, const char *uuid, char **uname) { int rc = pcmk_ok; xmlNode *a_child = NULL; xmlNode *xml_obj = NULL; xmlNode *fragment = NULL; const char *child_name = NULL; CRM_ASSERT(uname != NULL); CRM_ASSERT(uuid != NULL); rc = the_cib->cmds->query(the_cib, XML_CIB_TAG_NODES, &fragment, cib_sync_call | cib_scope_local); if (rc != pcmk_ok) { return rc; } xml_obj = fragment; CRM_CHECK(pcmk__xe_is(xml_obj, XML_CIB_TAG_NODES), return -ENOMSG); crm_log_xml_trace(xml_obj, "Result section"); rc = -ENXIO; *uname = NULL; for (a_child = pcmk__xml_first_child(xml_obj); a_child != NULL; a_child = pcmk__xml_next(a_child)) { if (pcmk__str_eq((const char *)a_child->name, XML_CIB_TAG_NODE, pcmk__str_none)) { child_name = ID(a_child); if (pcmk__str_eq(uuid, child_name, pcmk__str_casei)) { - child_name = crm_element_value(a_child, XML_ATTR_UNAME); + child_name = crm_element_value(a_child, PCMK_XA_UNAME); if (child_name != NULL) { *uname = strdup(child_name); rc = pcmk_ok; } break; } } } free_xml(fragment); return rc; } int set_standby(cib_t * the_cib, const char *uuid, const char *scope, const char *standby_value) { int rc = pcmk_ok; char *attr_id = NULL; CRM_CHECK(uuid != NULL, return -EINVAL); CRM_CHECK(standby_value != NULL, return -EINVAL); if (pcmk__strcase_any_of(scope, "reboot", XML_CIB_TAG_STATUS, NULL)) { scope = XML_CIB_TAG_STATUS; attr_id = crm_strdup_printf("transient-standby-%.256s", uuid); } else { scope = XML_CIB_TAG_NODES; attr_id = crm_strdup_printf("standby-%.256s", uuid); } rc = update_attr_delegate(the_cib, cib_sync_call, scope, uuid, NULL, NULL, attr_id, "standby", standby_value, TRUE, NULL, NULL); free(attr_id); return rc; } diff --git a/lib/cluster/corosync.c b/lib/cluster/corosync.c index 08280cef7a..372dfe8b78 100644 --- a/lib/cluster/corosync.c +++ b/lib/cluster/corosync.c @@ -1,814 +1,814 @@ /* - * Copyright 2004-2022 the Pacemaker project contributors + * Copyright 2004-2024 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 // PRIu64 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* PCMK__SPECIAL_PID* */ #include "crmcluster_private.h" static quorum_handle_t pcmk_quorum_handle = 0; static gboolean (*quorum_app_callback)(unsigned long long seq, gboolean quorate) = NULL; /*! * \internal * \brief Get the Corosync UUID associated with a Pacemaker node * * \param[in] node Pacemaker node * * \return Newly allocated string with node's Corosync UUID, or NULL if unknown * \note It is the caller's responsibility to free the result with free(). */ char * pcmk__corosync_uuid(const crm_node_t *node) { if ((node != NULL) && is_corosync_cluster()) { if (node->id > 0) { return crm_strdup_printf("%u", node->id); } else { crm_info("Node %s is not yet known by Corosync", node->uname); } } return NULL; } static bool node_name_is_valid(const char *key, const char *name) { int octet; if (name == NULL) { crm_trace("%s is empty", key); return false; } else if (sscanf(name, "%d.%d.%d.%d", &octet, &octet, &octet, &octet) == 4) { crm_trace("%s contains an IPv4 address (%s), ignoring", key, name); return false; } else if (strstr(name, ":") != NULL) { crm_trace("%s contains an IPv6 address (%s), ignoring", key, name); return false; } crm_trace("'%s: %s' is valid", key, name); return true; } /* * \internal * \brief Get Corosync node name corresponding to a node ID * * \param[in] cmap_handle Connection to Corosync CMAP * \param[in] nodeid Node ID to check * * \return Newly allocated string with name or (if no name) IP address * associated with first address assigned to a Corosync node ID (or NULL * if unknown) * \note It is the caller's responsibility to free the result with free(). */ char * pcmk__corosync_name(uint64_t /*cmap_handle_t */ cmap_handle, uint32_t nodeid) { // Originally based on corosync-quorumtool.c:node_name() int lpc = 0; cs_error_t rc = CS_OK; int retries = 0; char *name = NULL; cmap_handle_t local_handle = 0; int fd = -1; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; if (nodeid == 0) { nodeid = get_local_nodeid(0); } if (cmap_handle == 0 && local_handle == 0) { retries = 0; crm_trace("Initializing CMAP connection"); do { rc = pcmk__init_cmap(&local_handle); if (rc != CS_OK) { retries++; crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } } while (retries < 5 && rc != CS_OK); if (rc != CS_OK) { crm_warn("Could not connect to Cluster Configuration Database API, error %s", cs_strerror(rc)); local_handle = 0; } } if (cmap_handle == 0) { cmap_handle = local_handle; rc = cmap_fd_get(cmap_handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the CMAP API connection: %s (%d)", cs_strerror(rc), rc); goto bail; } /* CMAP provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("CMAP provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of CMAP provider: %s (%d)", strerror(-rv), -rv); goto bail; } } while (name == NULL && cmap_handle != 0) { uint32_t id = 0; char *key = NULL; key = crm_strdup_printf("nodelist.node.%d.nodeid", lpc); rc = cmap_get_uint32(cmap_handle, key, &id); crm_trace("Checking %u vs %u from %s", nodeid, id, key); free(key); if (rc != CS_OK) { break; } if (nodeid == id) { crm_trace("Searching for node name for %u in nodelist.node.%d %s", nodeid, lpc, pcmk__s(name, "")); if (name == NULL) { key = crm_strdup_printf("nodelist.node.%d.name", lpc); cmap_get_string(cmap_handle, key, &name); crm_trace("%s = %s", key, pcmk__s(name, "")); free(key); } if (name == NULL) { key = crm_strdup_printf("nodelist.node.%d.ring0_addr", lpc); cmap_get_string(cmap_handle, key, &name); crm_trace("%s = %s", key, pcmk__s(name, "")); if (!node_name_is_valid(key, name)) { free(name); name = NULL; } free(key); } break; } lpc++; } bail: if(local_handle) { cmap_finalize(local_handle); } if (name == NULL) { crm_info("Unable to get node name for nodeid %u", nodeid); } return name; } /*! * \internal * \brief Disconnect from Corosync cluster * * \param[in,out] cluster Cluster connection to disconnect */ void pcmk__corosync_disconnect(crm_cluster_t *cluster) { cluster_disconnect_cpg(cluster); if (pcmk_quorum_handle) { quorum_finalize(pcmk_quorum_handle); pcmk_quorum_handle = 0; } crm_notice("Disconnected from Corosync"); } /*! * \internal * \brief Dispatch function for quorum connection file descriptor * * \param[in] user_data Ignored * * \return 0 on success, -1 on error (per mainloop_io_t interface) */ static int quorum_dispatch_cb(gpointer user_data) { int rc = quorum_dispatch(pcmk_quorum_handle, CS_DISPATCH_ALL); if (rc < 0) { crm_err("Connection to the Quorum API failed: %d", rc); quorum_finalize(pcmk_quorum_handle); pcmk_quorum_handle = 0; return -1; } return 0; } /*! * \internal * \brief Notification callback for Corosync quorum connection * * \param[in] handle Corosync quorum connection * \param[in] quorate Whether cluster is quorate * \param[in] ring_id Corosync ring ID * \param[in] view_list_entries Number of entries in \p view_list * \param[in] view_list Corosync node IDs in membership */ static void quorum_notification_cb(quorum_handle_t handle, uint32_t quorate, uint64_t ring_id, uint32_t view_list_entries, uint32_t *view_list) { int i; GHashTableIter iter; crm_node_t *node = NULL; static gboolean init_phase = TRUE; if (quorate != crm_have_quorum) { if (quorate) { crm_notice("Quorum acquired " CRM_XS " membership=%" PRIu64 " members=%lu", ring_id, (long unsigned int)view_list_entries); } else { crm_warn("Quorum lost " CRM_XS " membership=%" PRIu64 " members=%lu", ring_id, (long unsigned int)view_list_entries); } crm_have_quorum = quorate; } else { crm_info("Quorum %s " CRM_XS " membership=%" PRIu64 " members=%lu", (quorate? "retained" : "still lost"), ring_id, (long unsigned int)view_list_entries); } if (view_list_entries == 0 && init_phase) { crm_info("Corosync membership is still forming, ignoring"); return; } init_phase = FALSE; /* Reset last_seen for all cached nodes so we can tell which ones aren't * in the view list */ g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { node->last_seen = 0; } /* Update the peer cache for each node in view list */ for (i = 0; i < view_list_entries; i++) { uint32_t id = view_list[i]; crm_debug("Member[%d] %u ", i, id); /* Get this node's peer cache entry (adding one if not already there) */ node = crm_get_peer(id, NULL); if (node->uname == NULL) { char *name = pcmk__corosync_name(0, id); crm_info("Obtaining name for new node %u", id); node = crm_get_peer(id, name); free(name); } /* Update the node state (including updating last_seen to ring_id) */ pcmk__update_peer_state(__func__, node, CRM_NODE_MEMBER, ring_id); } /* Remove any peer cache entries we didn't update */ pcmk__reap_unseen_nodes(ring_id); if (quorum_app_callback) { quorum_app_callback(ring_id, quorate); } } /*! * \internal * \brief Connect to Corosync quorum service * * \param[in] dispatch Connection dispatch callback * \param[in] destroy Connection destroy callback */ void pcmk__corosync_quorum_connect(gboolean (*dispatch)(unsigned long long, gboolean), void (*destroy)(gpointer)) { cs_error_t rc; int fd = 0; int quorate = 0; uint32_t quorum_type = 0; struct mainloop_fd_callbacks quorum_fd_callbacks; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; quorum_fd_callbacks.dispatch = quorum_dispatch_cb; quorum_fd_callbacks.destroy = destroy; crm_debug("Configuring Pacemaker to obtain quorum from Corosync"); { #if 0 // New way but not supported by all Corosync 2 versions quorum_model_v0_data_t quorum_model_data = { .model = QUORUM_MODEL_V0, .quorum_notify_fn = quorum_notification_cb, }; rc = quorum_model_initialize(&pcmk_quorum_handle, QUORUM_MODEL_V0, (quorum_model_data_t *) &quorum_model_data, &quorum_type, NULL); #else quorum_callbacks_t quorum_callbacks = { .quorum_notify_fn = quorum_notification_cb, }; rc = quorum_initialize(&pcmk_quorum_handle, &quorum_callbacks, &quorum_type); #endif } if (rc != CS_OK) { crm_err("Could not connect to the Quorum API: %s (%d)", cs_strerror(rc), rc); goto bail; } else if (quorum_type != QUORUM_SET) { crm_err("Corosync quorum is not configured"); goto bail; } rc = quorum_fd_get(pcmk_quorum_handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the Quorum API connection: %s (%d)", strerror(rc), rc); goto bail; } /* Quorum provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("Quorum provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); rc = CS_ERR_ACCESS; goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of Quorum provider: %s (%d)", strerror(-rv), -rv); rc = CS_ERR_ACCESS; goto bail; } rc = quorum_getquorate(pcmk_quorum_handle, &quorate); if (rc != CS_OK) { crm_err("Could not obtain the current Quorum API state: %d", rc); goto bail; } if (quorate) { crm_notice("Quorum acquired"); } else { crm_warn("No quorum"); } quorum_app_callback = dispatch; crm_have_quorum = quorate; rc = quorum_trackstart(pcmk_quorum_handle, CS_TRACK_CHANGES | CS_TRACK_CURRENT); if (rc != CS_OK) { crm_err("Could not setup Quorum API notifications: %d", rc); goto bail; } mainloop_add_fd("quorum", G_PRIORITY_HIGH, fd, dispatch, &quorum_fd_callbacks); pcmk__corosync_add_nodes(NULL); bail: if (rc != CS_OK) { quorum_finalize(pcmk_quorum_handle); } } /*! * \internal * \brief Connect to Corosync cluster layer * * \param[in,out] cluster Initialized cluster object to connect */ gboolean pcmk__corosync_connect(crm_cluster_t *cluster) { crm_node_t *peer = NULL; enum cluster_type_e stack = get_cluster_type(); crm_peer_init(); if (stack != pcmk_cluster_corosync) { crm_err("Invalid cluster type: %s " CRM_XS " stack=%d", name_for_cluster_type(stack), stack); return FALSE; } if (!cluster_connect_cpg(cluster)) { // Error message was logged by cluster_connect_cpg() return FALSE; } crm_info("Connection to %s established", name_for_cluster_type(stack)); cluster->nodeid = get_local_nodeid(0); if (cluster->nodeid == 0) { crm_err("Could not determine local node ID"); return FALSE; } cluster->uname = get_node_name(0); if (cluster->uname == NULL) { crm_err("Could not determine local node name"); return FALSE; } // Ensure local node always exists in peer cache peer = crm_get_peer(cluster->nodeid, cluster->uname); cluster->uuid = pcmk__corosync_uuid(peer); return TRUE; } /*! * \internal * \brief Check whether a Corosync cluster is active * * \return pcmk_cluster_corosync if Corosync is found, else pcmk_cluster_unknown */ enum cluster_type_e pcmk__corosync_detect(void) { int rc = CS_OK; cmap_handle_t handle; rc = pcmk__init_cmap(&handle); switch(rc) { case CS_OK: break; case CS_ERR_SECURITY: crm_debug("Failed to initialize the cmap API: Permission denied (%d)", rc); /* It's there, we just can't talk to it. * Good enough for us to identify as 'corosync' */ return pcmk_cluster_corosync; default: crm_info("Failed to initialize the cmap API: %s (%d)", pcmk__cs_err_str(rc), rc); return pcmk_cluster_unknown; } cmap_finalize(handle); return pcmk_cluster_corosync; } /*! * \brief Check whether a Corosync cluster peer is active * * \param[in] node Node to check * * \return TRUE if \p node is an active Corosync peer, otherwise FALSE */ gboolean crm_is_corosync_peer_active(const crm_node_t *node) { if (node == NULL) { crm_trace("Corosync peer inactive: NULL"); return FALSE; } else if (!pcmk__str_eq(node->state, CRM_NODE_MEMBER, pcmk__str_casei)) { crm_trace("Corosync peer %s inactive: state=%s", node->uname, node->state); return FALSE; } else if (!pcmk_is_set(node->processes, crm_proc_cpg)) { crm_trace("Corosync peer %s inactive: processes=%.16x", node->uname, node->processes); return FALSE; } return TRUE; } /*! * \internal * \brief Load Corosync node list (via CMAP) into peer cache and optionally XML * * \param[in,out] xml_parent If not NULL, add entry here for each node * * \return true if any nodes were found, false otherwise */ bool pcmk__corosync_add_nodes(xmlNode *xml_parent) { int lpc = 0; cs_error_t rc = CS_OK; int retries = 0; bool any = false; cmap_handle_t cmap_handle; int fd = -1; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; do { rc = pcmk__init_cmap(&cmap_handle); if (rc != CS_OK) { retries++; crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries); sleep(retries); } } while (retries < 5 && rc != CS_OK); if (rc != CS_OK) { crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc); return false; } rc = cmap_fd_get(cmap_handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the CMAP API connection: %s (%d)", cs_strerror(rc), rc); goto bail; } /* CMAP provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("CMAP provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of CMAP provider: %s (%d)", strerror(-rv), -rv); goto bail; } crm_peer_init(); crm_trace("Initializing Corosync node list"); for (lpc = 0; TRUE; lpc++) { uint32_t nodeid = 0; char *name = NULL; char *key = NULL; key = crm_strdup_printf("nodelist.node.%d.nodeid", lpc); rc = cmap_get_uint32(cmap_handle, key, &nodeid); free(key); if (rc != CS_OK) { break; } name = pcmk__corosync_name(cmap_handle, nodeid); if (name != NULL) { GHashTableIter iter; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node && node->uname && strcasecmp(node->uname, name) == 0) { if (node->id && node->id != nodeid) { crm_crit("Nodes %u and %u share the same name '%s': shutting down", node->id, nodeid, name); crm_exit(CRM_EX_FATAL); } } } } if (nodeid > 0 || name != NULL) { crm_trace("Initializing node[%d] %u = %s", lpc, nodeid, name); crm_get_peer(nodeid, name); } if (nodeid > 0 && name != NULL) { any = true; if (xml_parent) { xmlNode *node = create_xml_node(xml_parent, XML_CIB_TAG_NODE); crm_xml_set_id(node, "%u", nodeid); - crm_xml_add(node, XML_ATTR_UNAME, name); + crm_xml_add(node, PCMK_XA_UNAME, name); } } free(name); } bail: cmap_finalize(cmap_handle); return any; } /*! * \internal * \brief Get cluster name from Corosync configuration (via CMAP) * * \return Newly allocated string with cluster name if configured, or NULL */ char * pcmk__corosync_cluster_name(void) { cmap_handle_t handle; char *cluster_name = NULL; cs_error_t rc = CS_OK; int fd = -1; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rv; rc = pcmk__init_cmap(&handle); if (rc != CS_OK) { crm_info("Failed to initialize the cmap API: %s (%d)", cs_strerror(rc), rc); return NULL; } rc = cmap_fd_get(handle, &fd); if (rc != CS_OK) { crm_err("Could not obtain the CMAP API connection: %s (%d)", cs_strerror(rc), rc); goto bail; } /* CMAP provider run as root (in given user namespace, anyway)? */ if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid, &found_uid, &found_gid))) { crm_err("CMAP provider is not authentic:" " process %lld (uid: %lld, gid: %lld)", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); goto bail; } else if (rv < 0) { crm_err("Could not verify authenticity of CMAP provider: %s (%d)", strerror(-rv), -rv); goto bail; } rc = cmap_get_string(handle, "totem.cluster_name", &cluster_name); if (rc != CS_OK) { crm_info("Cannot get totem.cluster_name: %s (%d)", cs_strerror(rc), rc); } else { crm_debug("cmap totem.cluster_name = '%s'", cluster_name); } bail: cmap_finalize(handle); return cluster_name; } /*! * \internal * \brief Check (via CMAP) whether Corosync configuration has a node list * * \return true if Corosync has node list, otherwise false */ bool pcmk__corosync_has_nodelist(void) { cs_error_t cs_rc = CS_OK; int retries = 0; cmap_handle_t cmap_handle; cmap_iter_handle_t iter_handle; char key_name[CMAP_KEYNAME_MAXLEN + 1]; int fd = -1; uid_t found_uid = 0; gid_t found_gid = 0; pid_t found_pid = 0; int rc = pcmk_ok; static bool got_result = false; static bool result = false; if (got_result) { return result; } // Connect to CMAP do { cs_rc = pcmk__init_cmap(&cmap_handle); if (cs_rc != CS_OK) { retries++; crm_debug("CMAP connection failed: %s (rc=%d, retrying in %ds)", cs_strerror(cs_rc), cs_rc, retries); sleep(retries); } } while ((retries < 5) && (cs_rc != CS_OK)); if (cs_rc != CS_OK) { crm_warn("Assuming Corosync does not have node list: " "CMAP connection failed (%s) " CRM_XS " rc=%d", cs_strerror(cs_rc), cs_rc); return false; } // Get CMAP connection file descriptor cs_rc = cmap_fd_get(cmap_handle, &fd); if (cs_rc != CS_OK) { crm_warn("Assuming Corosync does not have node list: " "CMAP unusable (%s) " CRM_XS " rc=%d", cs_strerror(cs_rc), cs_rc); goto bail; } // Check whether CMAP connection is authentic (i.e. provided by root) rc = crm_ipc_is_authentic_process(fd, (uid_t) 0, (gid_t) 0, &found_pid, &found_uid, &found_gid); if (rc == 0) { crm_warn("Assuming Corosync does not have node list: " "CMAP provider is inauthentic " CRM_XS " pid=%lld uid=%lld gid=%lld", (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); goto bail; } else if (rc < 0) { crm_warn("Assuming Corosync does not have node list: " "Could not verify CMAP authenticity (%s) " CRM_XS " rc=%d", pcmk_strerror(rc), rc); goto bail; } // Check whether nodelist section is presetn cs_rc = cmap_iter_init(cmap_handle, "nodelist", &iter_handle); if (cs_rc != CS_OK) { crm_warn("Assuming Corosync does not have node list: " "CMAP not readable (%s) " CRM_XS " rc=%d", cs_strerror(cs_rc), cs_rc); goto bail; } cs_rc = cmap_iter_next(cmap_handle, iter_handle, key_name, NULL, NULL); if (cs_rc == CS_OK) { result = true; } cmap_iter_finalize(cmap_handle, iter_handle); got_result = true; crm_debug("Corosync %s node list", (result? "has" : "does not have")); bail: cmap_finalize(cmap_handle); return result; } diff --git a/lib/cluster/membership.c b/lib/cluster/membership.c index 55c494d7f0..f0176de8c2 100644 --- a/lib/cluster/membership.c +++ b/lib/cluster/membership.c @@ -1,1448 +1,1448 @@ /* * Copyright 2004-2024 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #include #include "crmcluster_private.h" /* The peer cache remembers cluster nodes that have been seen. * This is managed mostly automatically by libcluster, based on * cluster membership events. * * Because cluster nodes can have conflicting names or UUIDs, * the hash table key is a uniquely generated ID. */ GHashTable *crm_peer_cache = NULL; /* * The remote peer cache tracks pacemaker_remote nodes. While the * value has the same type as the peer cache's, it is tracked separately for * three reasons: pacemaker_remote nodes can't have conflicting names or UUIDs, * so the name (which is also the UUID) is used as the hash table key; there * is no equivalent of membership events, so management is not automatic; and * most users of the peer cache need to exclude pacemaker_remote nodes. * * That said, using a single cache would be more logical and less error-prone, * so it would be a good idea to merge them one day. * * libcluster provides two avenues for populating the cache: * crm_remote_peer_get() and crm_remote_peer_cache_remove() directly manage it, * while crm_remote_peer_cache_refresh() populates it via the CIB. */ GHashTable *crm_remote_peer_cache = NULL; /* * The known node cache tracks cluster and remote nodes that have been seen in * the CIB. It is useful mainly when a caller needs to know about a node that * may no longer be in the membership, but doesn't want to add the node to the * main peer cache tables. */ static GHashTable *known_node_cache = NULL; unsigned long long crm_peer_seq = 0; gboolean crm_have_quorum = FALSE; static gboolean crm_autoreap = TRUE; // Flag setting and clearing for crm_node_t:flags #define set_peer_flags(peer, flags_to_set) do { \ (peer)->flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Peer", (peer)->uname, \ (peer)->flags, (flags_to_set), \ #flags_to_set); \ } while (0) #define clear_peer_flags(peer, flags_to_clear) do { \ (peer)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Peer", (peer)->uname, \ (peer)->flags, (flags_to_clear), \ #flags_to_clear); \ } while (0) static void update_peer_uname(crm_node_t *node, const char *uname); int crm_remote_peer_cache_size(void) { if (crm_remote_peer_cache == NULL) { return 0; } return g_hash_table_size(crm_remote_peer_cache); } /*! * \brief Get a remote node peer cache entry, creating it if necessary * * \param[in] node_name Name of remote node * * \return Cache entry for node on success, NULL (and set errno) otherwise * * \note When creating a new entry, this will leave the node state undetermined, * so the caller should also call pcmk__update_peer_state() if the state * is known. * \note Because this can add and remove cache entries, callers should not * assume any previously obtained cache entry pointers remain valid. */ crm_node_t * crm_remote_peer_get(const char *node_name) { crm_node_t *node; char *node_name_copy = NULL; if (node_name == NULL) { errno = EINVAL; return NULL; } /* It's theoretically possible that the node was added to the cluster peer * cache before it was known to be a Pacemaker Remote node. Remove that * entry unless it has a node ID, which means the name actually is * associated with a cluster node. (@TODO return an error in that case?) */ node = pcmk__search_cluster_node_cache(0, node_name, NULL); if ((node != NULL) && (node->uuid == NULL)) { /* node_name could be a pointer into the cache entry being removed, so * reassign it to a copy before the original gets freed */ node_name_copy = strdup(node_name); if (node_name_copy == NULL) { errno = ENOMEM; return NULL; } node_name = node_name_copy; reap_crm_member(0, node_name); } /* Return existing cache entry if one exists */ node = g_hash_table_lookup(crm_remote_peer_cache, node_name); if (node) { free(node_name_copy); return node; } /* Allocate a new entry */ node = calloc(1, sizeof(crm_node_t)); if (node == NULL) { free(node_name_copy); return NULL; } /* Populate the essential information */ set_peer_flags(node, crm_remote_node); node->uuid = strdup(node_name); if (node->uuid == NULL) { free(node); errno = ENOMEM; free(node_name_copy); return NULL; } /* Add the new entry to the cache */ g_hash_table_replace(crm_remote_peer_cache, node->uuid, node); crm_trace("added %s to remote cache", node_name); /* Update the entry's uname, ensuring peer status callbacks are called */ update_peer_uname(node, node_name); free(node_name_copy); return node; } /*! * \brief Remove a node from the Pacemaker Remote node cache * * \param[in] node_name Name of node to remove from cache * * \note The caller must be careful not to use \p node_name after calling this * function if it might be a pointer into the cache entry being removed. */ void crm_remote_peer_cache_remove(const char *node_name) { /* Do a lookup first, because node_name could be a pointer within the entry * being removed -- we can't log it *after* removing it. */ if (g_hash_table_lookup(crm_remote_peer_cache, node_name) != NULL) { crm_trace("Removing %s from Pacemaker Remote node cache", node_name); g_hash_table_remove(crm_remote_peer_cache, node_name); } } /*! * \internal * \brief Return node status based on a CIB status entry * * \param[in] node_state XML of node state * * \return CRM_NODE_LOST if PCMK__XA_IN_CCM is false in node_state, * CRM_NODE_MEMBER otherwise * \note Unlike most boolean XML attributes, this one defaults to true, for * backward compatibility with older controllers that don't set it. */ static const char * remote_state_from_cib(const xmlNode *node_state) { bool status = false; if ((pcmk__xe_get_bool_attr(node_state, PCMK__XA_IN_CCM, &status) == pcmk_rc_ok) && !status) { return CRM_NODE_LOST; } else { return CRM_NODE_MEMBER; } } /* user data for looping through remote node xpath searches */ struct refresh_data { const char *field; /* XML attribute to check for node name */ gboolean has_state; /* whether to update node state based on XML */ }; /*! * \internal * \brief Process one pacemaker_remote node xpath search result * * \param[in] result XML search result * \param[in] user_data what to look for in the XML */ static void remote_cache_refresh_helper(xmlNode *result, void *user_data) { const struct refresh_data *data = user_data; const char *remote = crm_element_value(result, data->field); const char *state = NULL; crm_node_t *node; CRM_CHECK(remote != NULL, return); /* Determine node's state, if the result has it */ if (data->has_state) { state = remote_state_from_cib(result); } /* Check whether cache already has entry for node */ node = g_hash_table_lookup(crm_remote_peer_cache, remote); if (node == NULL) { /* Node is not in cache, so add a new entry for it */ node = crm_remote_peer_get(remote); CRM_ASSERT(node); if (state) { pcmk__update_peer_state(__func__, node, state, 0); } } else if (pcmk_is_set(node->flags, crm_node_dirty)) { /* Node is in cache and hasn't been updated already, so mark it clean */ clear_peer_flags(node, crm_node_dirty); if (state) { pcmk__update_peer_state(__func__, node, state, 0); } } } static void mark_dirty(gpointer key, gpointer value, gpointer user_data) { set_peer_flags((crm_node_t *) value, crm_node_dirty); } static gboolean is_dirty(gpointer key, gpointer value, gpointer user_data) { return pcmk_is_set(((crm_node_t*)value)->flags, crm_node_dirty); } /*! * \brief Repopulate the remote peer cache based on CIB XML * * \param[in] xmlNode CIB XML to parse */ void crm_remote_peer_cache_refresh(xmlNode *cib) { struct refresh_data data; crm_peer_init(); /* First, we mark all existing cache entries as dirty, * so that later we can remove any that weren't in the CIB. * We don't empty the cache, because we need to detect changes in state. */ g_hash_table_foreach(crm_remote_peer_cache, mark_dirty, NULL); /* Look for guest nodes and remote nodes in the status section */ data.field = PCMK_XA_ID; data.has_state = TRUE; crm_foreach_xpath_result(cib, PCMK__XP_REMOTE_NODE_STATUS, remote_cache_refresh_helper, &data); /* Look for guest nodes and remote nodes in the configuration section, * because they may have just been added and not have a status entry yet. * In that case, the cached node state will be left NULL, so that the * peer status callback isn't called until we're sure the node started * successfully. */ data.field = PCMK_XA_VALUE; data.has_state = FALSE; crm_foreach_xpath_result(cib, PCMK__XP_GUEST_NODE_CONFIG, remote_cache_refresh_helper, &data); data.field = PCMK_XA_ID; data.has_state = FALSE; crm_foreach_xpath_result(cib, PCMK__XP_REMOTE_NODE_CONFIG, remote_cache_refresh_helper, &data); /* Remove all old cache entries that weren't seen in the CIB */ g_hash_table_foreach_remove(crm_remote_peer_cache, is_dirty, NULL); } gboolean crm_is_peer_active(const crm_node_t * node) { if(node == NULL) { return FALSE; } if (pcmk_is_set(node->flags, crm_remote_node)) { /* remote nodes are never considered active members. This * guarantees they will never be considered for DC membership.*/ return FALSE; } #if SUPPORT_COROSYNC if (is_corosync_cluster()) { return crm_is_corosync_peer_active(node); } #endif crm_err("Unhandled cluster type: %s", name_for_cluster_type(get_cluster_type())); return FALSE; } static gboolean crm_reap_dead_member(gpointer key, gpointer value, gpointer user_data) { crm_node_t *node = value; crm_node_t *search = user_data; if (search == NULL) { return FALSE; } else if (search->id && node->id != search->id) { return FALSE; } else if (search->id == 0 && !pcmk__str_eq(node->uname, search->uname, pcmk__str_casei)) { return FALSE; } else if (crm_is_peer_active(value) == FALSE) { crm_info("Removing node with name %s and id %u from membership cache", (node->uname? node->uname : "unknown"), node->id); return TRUE; } return FALSE; } /*! * \brief Remove all peer cache entries matching a node ID and/or uname * * \param[in] id ID of node to remove (or 0 to ignore) * \param[in] name Uname of node to remove (or NULL to ignore) * * \return Number of cache entries removed * * \note The caller must be careful not to use \p name after calling this * function if it might be a pointer into the cache entry being removed. */ guint reap_crm_member(uint32_t id, const char *name) { int matches = 0; crm_node_t search = { 0, }; if (crm_peer_cache == NULL) { crm_trace("Membership cache not initialized, ignoring purge request"); return 0; } search.id = id; pcmk__str_update(&search.uname, name); matches = g_hash_table_foreach_remove(crm_peer_cache, crm_reap_dead_member, &search); if(matches) { crm_notice("Purged %d peer%s with id=%u%s%s from the membership cache", matches, pcmk__plural_s(matches), search.id, (search.uname? " and/or uname=" : ""), (search.uname? search.uname : "")); } else { crm_info("No peers with id=%u%s%s to purge from the membership cache", search.id, (search.uname? " and/or uname=" : ""), (search.uname? search.uname : "")); } free(search.uname); return matches; } static void count_peer(gpointer key, gpointer value, gpointer user_data) { guint *count = user_data; crm_node_t *node = value; if (crm_is_peer_active(node)) { *count = *count + 1; } } guint crm_active_peers(void) { guint count = 0; if (crm_peer_cache) { g_hash_table_foreach(crm_peer_cache, count_peer, &count); } return count; } static void destroy_crm_node(gpointer data) { crm_node_t *node = data; crm_trace("Destroying entry for node %u: %s", node->id, node->uname); free(node->uname); free(node->state); free(node->uuid); free(node->expected); free(node->conn_host); free(node); } void crm_peer_init(void) { if (crm_peer_cache == NULL) { crm_peer_cache = pcmk__strikey_table(free, destroy_crm_node); } if (crm_remote_peer_cache == NULL) { crm_remote_peer_cache = pcmk__strikey_table(NULL, destroy_crm_node); } if (known_node_cache == NULL) { known_node_cache = pcmk__strikey_table(free, destroy_crm_node); } } void crm_peer_destroy(void) { if (crm_peer_cache != NULL) { crm_trace("Destroying peer cache with %d members", g_hash_table_size(crm_peer_cache)); g_hash_table_destroy(crm_peer_cache); crm_peer_cache = NULL; } if (crm_remote_peer_cache != NULL) { crm_trace("Destroying remote peer cache with %d members", g_hash_table_size(crm_remote_peer_cache)); g_hash_table_destroy(crm_remote_peer_cache); crm_remote_peer_cache = NULL; } if (known_node_cache != NULL) { crm_trace("Destroying known node cache with %d members", g_hash_table_size(known_node_cache)); g_hash_table_destroy(known_node_cache); known_node_cache = NULL; } } static void (*peer_status_callback)(enum crm_status_type, crm_node_t *, const void *) = NULL; /*! * \brief Set a client function that will be called after peer status changes * * \param[in] dispatch Pointer to function to use as callback * * \note Previously, client callbacks were responsible for peer cache * management. This is no longer the case, and client callbacks should do * only client-specific handling. Callbacks MUST NOT add or remove entries * in the peer caches. */ void crm_set_status_callback(void (*dispatch) (enum crm_status_type, crm_node_t *, const void *)) { peer_status_callback = dispatch; } /*! * \brief Tell the library whether to automatically reap lost nodes * * If TRUE (the default), calling crm_update_peer_proc() will also update the * peer state to CRM_NODE_MEMBER or CRM_NODE_LOST, and pcmk__update_peer_state() * will reap peers whose state changes to anything other than CRM_NODE_MEMBER. * Callers should leave this enabled unless they plan to manage the cache * separately on their own. * * \param[in] autoreap TRUE to enable automatic reaping, FALSE to disable */ void crm_set_autoreap(gboolean autoreap) { crm_autoreap = autoreap; } static void dump_peer_hash(int level, const char *caller) { GHashTableIter iter; const char *id = NULL; crm_node_t *node = NULL; g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, (gpointer *) &id, (gpointer *) &node)) { do_crm_log(level, "%s: Node %u/%s = %p - %s", caller, node->id, node->uname, node, id); } } static gboolean hash_find_by_data(gpointer key, gpointer value, gpointer user_data) { return value == user_data; } /*! * \internal * \brief Search caches for a node (cluster or Pacemaker Remote) * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] flags Bitmask of enum crm_get_peer_flags * * \return Node cache entry if found, otherwise NULL */ crm_node_t * pcmk__search_node_caches(unsigned int id, const char *uname, uint32_t flags) { crm_node_t *node = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); if ((uname != NULL) && pcmk_is_set(flags, CRM_GET_PEER_REMOTE)) { node = g_hash_table_lookup(crm_remote_peer_cache, uname); } if ((node == NULL) && pcmk_is_set(flags, CRM_GET_PEER_CLUSTER)) { node = pcmk__search_cluster_node_cache(id, uname, NULL); } return node; } /*! * \brief Get a node cache entry (cluster or Pacemaker Remote) * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] uuid If not NULL while id is 0, node UUID instead of cluster * node ID to search for * \param[in] flags Bitmask of enum crm_get_peer_flags * * \return (Possibly newly created) node cache entry */ crm_node_t * pcmk__get_peer_full(unsigned int id, const char *uname, const char *uuid, int flags) { crm_node_t *node = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); if (pcmk_is_set(flags, CRM_GET_PEER_REMOTE)) { node = g_hash_table_lookup(crm_remote_peer_cache, uname); } if ((node == NULL) && pcmk_is_set(flags, CRM_GET_PEER_CLUSTER)) { node = pcmk__get_peer(id, uname, uuid); } return node; } /*! * \internal * \brief Purge a node from cache (both cluster and Pacemaker Remote) * * \param[in] node_name If not NULL, purge only nodes with this name * \param[in] node_id If not 0, purge cluster nodes only if they have this ID * * \note If \p node_name is NULL and \p node_id is 0, no nodes will be purged. * If \p node_name is not NULL and \p node_id is not 0, Pacemaker Remote * nodes that match \p node_name will be purged, and cluster nodes that * match both \p node_name and \p node_id will be purged. * \note The caller must be careful not to use \p node_name after calling this * function if it might be a pointer into a cache entry being removed. */ void pcmk__purge_node_from_cache(const char *node_name, uint32_t node_id) { char *node_name_copy = NULL; if ((node_name == NULL) && (node_id == 0U)) { return; } // Purge from Pacemaker Remote node cache if ((node_name != NULL) && (g_hash_table_lookup(crm_remote_peer_cache, node_name) != NULL)) { /* node_name could be a pointer into the cache entry being purged, * so reassign it to a copy before the original gets freed */ node_name_copy = strdup(node_name); CRM_ASSERT(node_name_copy != NULL); node_name = node_name_copy; crm_trace("Purging %s from Pacemaker Remote node cache", node_name); g_hash_table_remove(crm_remote_peer_cache, node_name); } reap_crm_member(node_id, node_name); free(node_name_copy); } /*! * \brief Get a node cache entry (cluster or Pacemaker Remote) * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] flags Bitmask of enum crm_get_peer_flags * * \return (Possibly newly created) node cache entry */ crm_node_t * crm_get_peer_full(unsigned int id, const char *uname, int flags) { return pcmk__get_peer_full(id, uname, NULL, flags); } /*! * \internal * \brief Search cluster node cache * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] uuid If not NULL while id is 0, node UUID instead of cluster * node ID to search for * * \return Cluster node cache entry if found, otherwise NULL */ crm_node_t * pcmk__search_cluster_node_cache(unsigned int id, const char *uname, const char *uuid) { GHashTableIter iter; crm_node_t *node = NULL; crm_node_t *by_id = NULL; crm_node_t *by_name = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); if (uname != NULL) { g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node->uname && strcasecmp(node->uname, uname) == 0) { crm_trace("Name match: %s = %p", node->uname, node); by_name = node; break; } } } if (id > 0) { g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(node->id == id) { crm_trace("ID match: %u = %p", node->id, node); by_id = node; break; } } } else if (uuid != NULL) { g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (pcmk__str_eq(node->uuid, uuid, pcmk__str_casei)) { crm_trace("UUID match: %s = %p", node->uuid, node); by_id = node; break; } } } node = by_id; /* Good default */ if(by_id == by_name) { /* Nothing to do if they match (both NULL counts) */ crm_trace("Consistent: %p for %u/%s", by_id, id, uname); } else if(by_id == NULL && by_name) { crm_trace("Only one: %p for %u/%s", by_name, id, uname); if(id && by_name->id) { dump_peer_hash(LOG_WARNING, __func__); crm_crit("Node %u and %u share the same name '%s'", id, by_name->id, uname); node = NULL; /* Create a new one */ } else { node = by_name; } } else if(by_name == NULL && by_id) { crm_trace("Only one: %p for %u/%s", by_id, id, uname); if(uname && by_id->uname) { dump_peer_hash(LOG_WARNING, __func__); crm_crit("Node '%s' and '%s' share the same cluster nodeid %u: assuming '%s' is correct", uname, by_id->uname, id, uname); } } else if(uname && by_id->uname) { if(pcmk__str_eq(uname, by_id->uname, pcmk__str_casei)) { crm_notice("Node '%s' has changed its ID from %u to %u", by_id->uname, by_name->id, by_id->id); g_hash_table_foreach_remove(crm_peer_cache, hash_find_by_data, by_name); } else { crm_warn("Node '%s' and '%s' share the same cluster nodeid: %u %s", by_id->uname, by_name->uname, id, uname); dump_peer_hash(LOG_INFO, __func__); crm_abort(__FILE__, __func__, __LINE__, "member weirdness", TRUE, TRUE); } } else if(id && by_name->id) { crm_warn("Node %u and %u share the same name: '%s'", by_id->id, by_name->id, uname); } else { /* Simple merge */ /* Only corosync-based clusters use node IDs. The functions that call * pcmk__update_peer_state() and crm_update_peer_proc() only know * nodeid, so 'by_id' is authoritative when merging. */ dump_peer_hash(LOG_DEBUG, __func__); crm_info("Merging %p into %p", by_name, by_id); g_hash_table_foreach_remove(crm_peer_cache, hash_find_by_data, by_name); } return node; } #if SUPPORT_COROSYNC static guint remove_conflicting_peer(crm_node_t *node) { int matches = 0; GHashTableIter iter; crm_node_t *existing_node = NULL; if (node->id == 0 || node->uname == NULL) { return 0; } if (!pcmk__corosync_has_nodelist()) { return 0; } g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &existing_node)) { if (existing_node->id > 0 && existing_node->id != node->id && existing_node->uname != NULL && strcasecmp(existing_node->uname, node->uname) == 0) { if (crm_is_peer_active(existing_node)) { continue; } crm_warn("Removing cached offline node %u/%s which has conflicting uname with %u", existing_node->id, existing_node->uname, node->id); g_hash_table_iter_remove(&iter); matches++; } } return matches; } #endif /*! * \brief Get a cluster node cache entry * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] uuid If not NULL while id is 0, node UUID instead of cluster * node ID to search for * * \return (Possibly newly created) cluster node cache entry */ /* coverity[-alloc] Memory is referenced in one or both hashtables */ crm_node_t * pcmk__get_peer(unsigned int id, const char *uname, const char *uuid) { crm_node_t *node = NULL; char *uname_lookup = NULL; CRM_ASSERT(id > 0 || uname != NULL); crm_peer_init(); node = pcmk__search_cluster_node_cache(id, uname, uuid); /* if uname wasn't provided, and find_peer did not turn up a uname based on id. * we need to do a lookup of the node name using the id in the cluster membership. */ if ((node == NULL || node->uname == NULL) && (uname == NULL)) { uname_lookup = get_node_name(id); } if (uname_lookup) { uname = uname_lookup; crm_trace("Inferred a name of '%s' for node %u", uname, id); /* try to turn up the node one more time now that we know the uname. */ if (node == NULL) { node = pcmk__search_cluster_node_cache(id, uname, uuid); } } if (node == NULL) { char *uniqueid = crm_generate_uuid(); node = calloc(1, sizeof(crm_node_t)); CRM_ASSERT(node); crm_info("Created entry %s/%p for node %s/%u (%d total)", uniqueid, node, uname, id, 1 + g_hash_table_size(crm_peer_cache)); g_hash_table_replace(crm_peer_cache, uniqueid, node); } if(id > 0 && uname && (node->id == 0 || node->uname == NULL)) { crm_info("Node %u is now known as %s", id, uname); } if(id > 0 && node->id == 0) { node->id = id; } if (uname && (node->uname == NULL)) { update_peer_uname(node, uname); } if(node->uuid == NULL) { if (uuid == NULL) { uuid = crm_peer_uuid(node); } if (uuid) { crm_info("Node %u has uuid %s", id, uuid); } else { crm_info("Cannot obtain a UUID for node %u/%s", id, node->uname); } } free(uname_lookup); return node; } /*! * \brief Get a cluster node cache entry * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * * \return (Possibly newly created) cluster node cache entry */ /* coverity[-alloc] Memory is referenced in one or both hashtables */ crm_node_t * crm_get_peer(unsigned int id, const char *uname) { return pcmk__get_peer(id, uname, NULL); } /*! * \internal * \brief Update a node's uname * * \param[in,out] node Node object to update * \param[in] uname New name to set * * \note This function should not be called within a peer cache iteration, * because in some cases it can remove conflicting cache entries, * which would invalidate the iterator. */ static void update_peer_uname(crm_node_t *node, const char *uname) { CRM_CHECK(uname != NULL, crm_err("Bug: can't update node name without name"); return); CRM_CHECK(node != NULL, crm_err("Bug: can't update node name to %s without node", uname); return); if (pcmk__str_eq(uname, node->uname, pcmk__str_casei)) { crm_debug("Node uname '%s' did not change", uname); return; } for (const char *c = uname; *c; ++c) { if ((*c >= 'A') && (*c <= 'Z')) { crm_warn("Node names with capitals are discouraged, consider changing '%s'", uname); break; } } pcmk__str_update(&node->uname, uname); if (peer_status_callback != NULL) { peer_status_callback(crm_status_uname, node, NULL); } #if SUPPORT_COROSYNC if (is_corosync_cluster() && !pcmk_is_set(node->flags, crm_remote_node)) { remove_conflicting_peer(node); } #endif } /*! * \internal * \brief Get log-friendly string equivalent of a process flag * * \param[in] proc Process flag * * \return Log-friendly string equivalent of \p proc */ static inline const char * proc2text(enum crm_proc_flag proc) { const char *text = "unknown"; switch (proc) { case crm_proc_none: text = "none"; break; case crm_proc_based: text = "pacemaker-based"; break; case crm_proc_controld: text = "pacemaker-controld"; break; case crm_proc_schedulerd: text = "pacemaker-schedulerd"; break; case crm_proc_execd: text = "pacemaker-execd"; break; case crm_proc_attrd: text = "pacemaker-attrd"; break; case crm_proc_fenced: text = "pacemaker-fenced"; break; case crm_proc_cpg: text = "corosync-cpg"; break; } return text; } /*! * \internal * \brief Update a node's process information (and potentially state) * * \param[in] source Caller's function name (for log messages) * \param[in,out] node Node object to update * \param[in] flag Bitmask of new process information * \param[in] status node status (online, offline, etc.) * * \return NULL if any node was reaped from peer caches, value of node otherwise * * \note If this function returns NULL, the supplied node object was likely * freed and should not be used again. This function should not be * called within a cache iteration if reaping is possible, otherwise * reaping could invalidate the iterator. */ crm_node_t * crm_update_peer_proc(const char *source, crm_node_t * node, uint32_t flag, const char *status) { uint32_t last = 0; gboolean changed = FALSE; CRM_CHECK(node != NULL, crm_err("%s: Could not set %s to %s for NULL", source, proc2text(flag), status); return NULL); /* Pacemaker doesn't spawn processes on remote nodes */ if (pcmk_is_set(node->flags, crm_remote_node)) { return node; } last = node->processes; if (status == NULL) { node->processes = flag; if (node->processes != last) { changed = TRUE; } } else if (pcmk__str_eq(status, ONLINESTATUS, pcmk__str_casei)) { if ((node->processes & flag) != flag) { node->processes = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, "Peer process", node->uname, node->processes, flag, "processes"); changed = TRUE; } } else if (node->processes & flag) { node->processes = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, "Peer process", node->uname, node->processes, flag, "processes"); changed = TRUE; } if (changed) { if (status == NULL && flag <= crm_proc_none) { crm_info("%s: Node %s[%u] - all processes are now offline", source, node->uname, node->id); } else { crm_info("%s: Node %s[%u] - %s is now %s", source, node->uname, node->id, proc2text(flag), status); } if (pcmk_is_set(node->processes, crm_get_cluster_proc())) { node->when_online = time(NULL); } else { node->when_online = 0; } /* Call the client callback first, then update the peer state, * in case the node will be reaped */ if (peer_status_callback != NULL) { peer_status_callback(crm_status_processes, node, &last); } /* The client callback shouldn't touch the peer caches, * but as a safety net, bail if the peer cache was destroyed. */ if (crm_peer_cache == NULL) { return NULL; } if (crm_autoreap) { const char *peer_state = NULL; if (pcmk_is_set(node->processes, crm_get_cluster_proc())) { peer_state = CRM_NODE_MEMBER; } else { peer_state = CRM_NODE_LOST; } node = pcmk__update_peer_state(__func__, node, peer_state, 0); } } else { crm_trace("%s: Node %s[%u] - %s is unchanged (%s)", source, node->uname, node->id, proc2text(flag), status); } return node; } /*! * \internal * \brief Update a cluster node cache entry's expected join state * * \param[in] source Caller's function name (for logging) * \param[in,out] node Node to update * \param[in] expected Node's new join state */ void pcmk__update_peer_expected(const char *source, crm_node_t *node, const char *expected) { char *last = NULL; gboolean changed = FALSE; CRM_CHECK(node != NULL, crm_err("%s: Could not set 'expected' to %s", source, expected); return); /* Remote nodes don't participate in joins */ if (pcmk_is_set(node->flags, crm_remote_node)) { return; } last = node->expected; if (expected != NULL && !pcmk__str_eq(node->expected, expected, pcmk__str_casei)) { node->expected = strdup(expected); changed = TRUE; } if (changed) { crm_info("%s: Node %s[%u] - expected state is now %s (was %s)", source, node->uname, node->id, expected, last); free(last); } else { crm_trace("%s: Node %s[%u] - expected state is unchanged (%s)", source, node->uname, node->id, expected); } } /*! * \internal * \brief Update a node's state and membership information * * \param[in] source Caller's function name (for log messages) * \param[in,out] node Node object to update * \param[in] state Node's new state * \param[in] membership Node's new membership ID * \param[in,out] iter If not NULL, pointer to node's peer cache iterator * * \return NULL if any node was reaped, value of node otherwise * * \note If this function returns NULL, the supplied node object was likely * freed and should not be used again. This function may be called from * within a peer cache iteration if the iterator is supplied. */ static crm_node_t * update_peer_state_iter(const char *source, crm_node_t *node, const char *state, uint64_t membership, GHashTableIter *iter) { gboolean is_member; CRM_CHECK(node != NULL, crm_err("Could not set state for unknown host to %s" CRM_XS " source=%s", state, source); return NULL); is_member = pcmk__str_eq(state, CRM_NODE_MEMBER, pcmk__str_casei); if (is_member) { node->when_lost = 0; if (membership) { node->last_seen = membership; } } if (state && !pcmk__str_eq(node->state, state, pcmk__str_casei)) { char *last = node->state; if (is_member) { node->when_member = time(NULL); } else { node->when_member = 0; } node->state = strdup(state); crm_notice("Node %s state is now %s " CRM_XS " nodeid=%u previous=%s source=%s", node->uname, state, node->id, (last? last : "unknown"), source); if (peer_status_callback != NULL) { peer_status_callback(crm_status_nstate, node, last); } free(last); if (crm_autoreap && !is_member && !pcmk_is_set(node->flags, crm_remote_node)) { /* We only autoreap from the peer cache, not the remote peer cache, * because the latter should be managed only by * crm_remote_peer_cache_refresh(). */ if(iter) { crm_notice("Purged 1 peer with id=%u and/or uname=%s from the membership cache", node->id, node->uname); g_hash_table_iter_remove(iter); } else { reap_crm_member(node->id, node->uname); } node = NULL; } } else { crm_trace("Node %s state is unchanged (%s) " CRM_XS " nodeid=%u source=%s", node->uname, state, node->id, source); } return node; } /*! * \brief Update a node's state and membership information * * \param[in] source Caller's function name (for log messages) * \param[in,out] node Node object to update * \param[in] state Node's new state * \param[in] membership Node's new membership ID * * \return NULL if any node was reaped, value of node otherwise * * \note If this function returns NULL, the supplied node object was likely * freed and should not be used again. This function should not be * called within a cache iteration if reaping is possible, * otherwise reaping could invalidate the iterator. */ crm_node_t * pcmk__update_peer_state(const char *source, crm_node_t *node, const char *state, uint64_t membership) { return update_peer_state_iter(source, node, state, membership, NULL); } /*! * \internal * \brief Reap all nodes from cache whose membership information does not match * * \param[in] membership Membership ID of nodes to keep */ void pcmk__reap_unseen_nodes(uint64_t membership) { GHashTableIter iter; crm_node_t *node = NULL; crm_trace("Reaping unseen nodes..."); g_hash_table_iter_init(&iter, crm_peer_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *)&node)) { if (node->last_seen != membership) { if (node->state) { /* * Calling update_peer_state_iter() allows us to * remove the node from crm_peer_cache without * invalidating our iterator */ update_peer_state_iter(__func__, node, CRM_NODE_LOST, membership, &iter); } else { crm_info("State of node %s[%u] is still unknown", node->uname, node->id); } } } } static crm_node_t * find_known_node(const char *id, const char *uname) { GHashTableIter iter; crm_node_t *node = NULL; crm_node_t *by_id = NULL; crm_node_t *by_name = NULL; if (uname) { g_hash_table_iter_init(&iter, known_node_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if (node->uname && strcasecmp(node->uname, uname) == 0) { crm_trace("Name match: %s = %p", node->uname, node); by_name = node; break; } } } if (id) { g_hash_table_iter_init(&iter, known_node_cache); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { if(strcasecmp(node->uuid, id) == 0) { crm_trace("ID match: %s= %p", id, node); by_id = node; break; } } } node = by_id; /* Good default */ if (by_id == by_name) { /* Nothing to do if they match (both NULL counts) */ crm_trace("Consistent: %p for %s/%s", by_id, id, uname); } else if (by_id == NULL && by_name) { crm_trace("Only one: %p for %s/%s", by_name, id, uname); if (id) { node = NULL; } else { node = by_name; } } else if (by_name == NULL && by_id) { crm_trace("Only one: %p for %s/%s", by_id, id, uname); if (uname) { node = NULL; } } else if (uname && by_id->uname && pcmk__str_eq(uname, by_id->uname, pcmk__str_casei)) { /* Multiple nodes have the same uname in the CIB. * Return by_id. */ } else if (id && by_name->uuid && pcmk__str_eq(id, by_name->uuid, pcmk__str_casei)) { /* Multiple nodes have the same id in the CIB. * Return by_name. */ node = by_name; } else { node = NULL; } if (node == NULL) { crm_debug("Couldn't find node%s%s%s%s", id? " " : "", id? id : "", uname? " with name " : "", uname? uname : ""); } return node; } static void known_node_cache_refresh_helper(xmlNode *xml_node, void *user_data) { const char *id = crm_element_value(xml_node, PCMK_XA_ID); - const char *uname = crm_element_value(xml_node, XML_ATTR_UNAME); + const char *uname = crm_element_value(xml_node, PCMK_XA_UNAME); crm_node_t * node = NULL; CRM_CHECK(id != NULL && uname !=NULL, return); node = find_known_node(id, uname); if (node == NULL) { char *uniqueid = crm_generate_uuid(); node = calloc(1, sizeof(crm_node_t)); CRM_ASSERT(node != NULL); node->uname = strdup(uname); CRM_ASSERT(node->uname != NULL); node->uuid = strdup(id); CRM_ASSERT(node->uuid != NULL); g_hash_table_replace(known_node_cache, uniqueid, node); } else if (pcmk_is_set(node->flags, crm_node_dirty)) { pcmk__str_update(&node->uname, uname); /* Node is in cache and hasn't been updated already, so mark it clean */ clear_peer_flags(node, crm_node_dirty); } } static void refresh_known_node_cache(xmlNode *cib) { crm_peer_init(); g_hash_table_foreach(known_node_cache, mark_dirty, NULL); crm_foreach_xpath_result(cib, PCMK__XP_MEMBER_NODE_CONFIG, known_node_cache_refresh_helper, NULL); /* Remove all old cache entries that weren't seen in the CIB */ g_hash_table_foreach_remove(known_node_cache, is_dirty, NULL); } void pcmk__refresh_node_caches_from_cib(xmlNode *cib) { crm_remote_peer_cache_refresh(cib); refresh_known_node_cache(cib); } /*! * \internal * \brief Search known node cache * * \param[in] id If not 0, cluster node ID to search for * \param[in] uname If not NULL, node name to search for * \param[in] flags Bitmask of enum crm_get_peer_flags * * \return Known node cache entry if found, otherwise NULL */ crm_node_t * pcmk__search_known_node_cache(unsigned int id, const char *uname, uint32_t flags) { crm_node_t *node = NULL; char *id_str = NULL; CRM_ASSERT(id > 0 || uname != NULL); node = pcmk__search_node_caches(id, uname, flags); if (node || !(flags & CRM_GET_PEER_CLUSTER)) { return node; } if (id > 0) { id_str = crm_strdup_printf("%u", id); } node = find_known_node(id_str, uname); free(id_str); return node; } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START #include int crm_terminate_member(int nodeid, const char *uname, void *unused) { return stonith_api_kick(nodeid, uname, 120, TRUE); } int crm_terminate_member_no_mainloop(int nodeid, const char *uname, int *connection) { return stonith_api_kick(nodeid, uname, 120, TRUE); } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/common/ipc_client.c b/lib/common/ipc_client.c index 5e64e23241..fb5c40c81d 100644 --- a/lib/common/ipc_client.c +++ b/lib/common/ipc_client.c @@ -1,1690 +1,1690 @@ /* - * Copyright 2004-2023 the Pacemaker project contributors + * Copyright 2004-2024 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 #if defined(HAVE_UCRED) || defined(HAVE_SOCKPEERCRED) # ifdef HAVE_UCRED # ifndef _GNU_SOURCE # define _GNU_SOURCE # endif # endif # include #elif defined(HAVE_GETPEERUCRED) # include #endif #include #include #include #include #include /* indirectly: pcmk_err_generic */ #include #include #include #include "crmcommon_private.h" static int is_ipc_provider_expected(qb_ipcc_connection_t *qb_ipc, int sock, uid_t refuid, gid_t refgid, pid_t *gotpid, uid_t *gotuid, gid_t *gotgid); /*! * \brief Create a new object for using Pacemaker daemon IPC * * \param[out] api Where to store new IPC object * \param[in] server Which Pacemaker daemon the object is for * * \return Standard Pacemaker result code * * \note The caller is responsible for freeing *api using pcmk_free_ipc_api(). * \note This is intended to supersede crm_ipc_new() but currently only * supports the controller, pacemakerd, and schedulerd IPC API. */ int pcmk_new_ipc_api(pcmk_ipc_api_t **api, enum pcmk_ipc_server server) { if (api == NULL) { return EINVAL; } *api = calloc(1, sizeof(pcmk_ipc_api_t)); if (*api == NULL) { return errno; } (*api)->server = server; if (pcmk_ipc_name(*api, false) == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return EOPNOTSUPP; } (*api)->ipc_size_max = 0; // Set server methods and max_size (if not default) switch (server) { case pcmk_ipc_attrd: (*api)->cmds = pcmk__attrd_api_methods(); break; case pcmk_ipc_based: (*api)->ipc_size_max = 512 * 1024; // 512KB break; case pcmk_ipc_controld: (*api)->cmds = pcmk__controld_api_methods(); break; case pcmk_ipc_execd: break; case pcmk_ipc_fenced: break; case pcmk_ipc_pacemakerd: (*api)->cmds = pcmk__pacemakerd_api_methods(); break; case pcmk_ipc_schedulerd: (*api)->cmds = pcmk__schedulerd_api_methods(); // @TODO max_size could vary by client, maybe take as argument? (*api)->ipc_size_max = 5 * 1024 * 1024; // 5MB break; } if ((*api)->cmds == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } (*api)->ipc = crm_ipc_new(pcmk_ipc_name(*api, false), (*api)->ipc_size_max); if ((*api)->ipc == NULL) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } // If daemon API has its own data to track, allocate it if ((*api)->cmds->new_data != NULL) { if ((*api)->cmds->new_data(*api) != pcmk_rc_ok) { pcmk_free_ipc_api(*api); *api = NULL; return ENOMEM; } } crm_trace("Created %s API IPC object", pcmk_ipc_name(*api, true)); return pcmk_rc_ok; } static void free_daemon_specific_data(pcmk_ipc_api_t *api) { if ((api != NULL) && (api->cmds != NULL)) { if ((api->cmds->free_data != NULL) && (api->api_data != NULL)) { api->cmds->free_data(api->api_data); api->api_data = NULL; } free(api->cmds); api->cmds = NULL; } } /*! * \internal * \brief Call an IPC API event callback, if one is registed * * \param[in,out] api IPC API connection * \param[in] event_type The type of event that occurred * \param[in] status Event status * \param[in,out] event_data Event-specific data */ void pcmk__call_ipc_callback(pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data) { if ((api != NULL) && (api->cb != NULL)) { api->cb(api, event_type, status, event_data, api->user_data); } } /*! * \internal * \brief Clean up after an IPC disconnect * * \param[in,out] user_data IPC API connection that disconnected * * \note This function can be used as a main loop IPC destroy callback. */ static void ipc_post_disconnect(gpointer user_data) { pcmk_ipc_api_t *api = user_data; crm_info("Disconnected from %s", pcmk_ipc_name(api, true)); // Perform any daemon-specific handling needed if ((api->cmds != NULL) && (api->cmds->post_disconnect != NULL)) { api->cmds->post_disconnect(api); } // Call client's registered event callback pcmk__call_ipc_callback(api, pcmk_ipc_event_disconnect, CRM_EX_DISCONNECT, NULL); /* If this is being called from a running main loop, mainloop_gio_destroy() * will free ipc and mainloop_io immediately after calling this function. * If this is called from a stopped main loop, these will leak, so the best * practice is to close the connection before stopping the main loop. */ api->ipc = NULL; api->mainloop_io = NULL; if (api->free_on_disconnect) { /* pcmk_free_ipc_api() has already been called, but did not free api * or api->cmds because this function needed them. Do that now. */ free_daemon_specific_data(api); crm_trace("Freeing IPC API object after disconnect"); free(api); } } /*! * \brief Free the contents of an IPC API object * * \param[in,out] api IPC API object to free */ void pcmk_free_ipc_api(pcmk_ipc_api_t *api) { bool free_on_disconnect = false; if (api == NULL) { return; } crm_debug("Releasing %s IPC API", pcmk_ipc_name(api, true)); if (api->ipc != NULL) { if (api->mainloop_io != NULL) { /* We need to keep the api pointer itself around, because it is the * user data for the IPC client destroy callback. That will be * triggered by the pcmk_disconnect_ipc() call below, but it might * happen later in the main loop (if still running). * * This flag tells the destroy callback to free the object. It can't * do that unconditionally, because the application might call this * function after a disconnect that happened by other means. */ free_on_disconnect = api->free_on_disconnect = true; } pcmk_disconnect_ipc(api); // Frees api if free_on_disconnect is true } if (!free_on_disconnect) { free_daemon_specific_data(api); crm_trace("Freeing IPC API object"); free(api); } } /*! * \brief Get the IPC name used with an IPC API connection * * \param[in] api IPC API connection * \param[in] for_log If true, return human-friendly name instead of IPC name * * \return IPC API's human-friendly or connection name, or if none is available, * "Pacemaker" if for_log is true and NULL if for_log is false */ const char * pcmk_ipc_name(const pcmk_ipc_api_t *api, bool for_log) { if (api == NULL) { return for_log? "Pacemaker" : NULL; } switch (api->server) { case pcmk_ipc_attrd: return for_log? "attribute manager" : T_ATTRD; case pcmk_ipc_based: return for_log? "CIB manager" : NULL /* PCMK__SERVER_BASED_RW */; case pcmk_ipc_controld: return for_log? "controller" : CRM_SYSTEM_CRMD; case pcmk_ipc_execd: return for_log? "executor" : NULL /* CRM_SYSTEM_LRMD */; case pcmk_ipc_fenced: return for_log? "fencer" : NULL /* "stonith-ng" */; case pcmk_ipc_pacemakerd: return for_log? "launcher" : CRM_SYSTEM_MCP; case pcmk_ipc_schedulerd: return for_log? "scheduler" : CRM_SYSTEM_PENGINE; default: return for_log? "Pacemaker" : NULL; } } /*! * \brief Check whether an IPC API connection is active * * \param[in,out] api IPC API connection * * \return true if IPC is connected, false otherwise */ bool pcmk_ipc_is_connected(pcmk_ipc_api_t *api) { return (api != NULL) && crm_ipc_connected(api->ipc); } /*! * \internal * \brief Call the daemon-specific API's dispatch function * * Perform daemon-specific handling of IPC reply dispatch. It is the daemon * method's responsibility to call the client's registered event callback, as * well as allocate and free any event data. * * \param[in,out] api IPC API connection * \param[in,out] message IPC reply XML to dispatch */ static bool call_api_dispatch(pcmk_ipc_api_t *api, xmlNode *message) { crm_log_xml_trace(message, "ipc-received"); if ((api->cmds != NULL) && (api->cmds->dispatch != NULL)) { return api->cmds->dispatch(api, message); } return false; } /*! * \internal * \brief Dispatch previously read IPC data * * \param[in] buffer Data read from IPC * \param[in,out] api IPC object * * \return Standard Pacemaker return code. In particular: * * pcmk_rc_ok: There are no more messages expected from the server. Quit * reading. * EINPROGRESS: There are more messages expected from the server. Keep reading. * * All other values indicate an error. */ static int dispatch_ipc_data(const char *buffer, pcmk_ipc_api_t *api) { bool more = false; xmlNode *msg; if (buffer == NULL) { crm_warn("Empty message received from %s IPC", pcmk_ipc_name(api, true)); return ENOMSG; } msg = string2xml(buffer); if (msg == NULL) { crm_warn("Malformed message received from %s IPC", pcmk_ipc_name(api, true)); return EPROTO; } more = call_api_dispatch(api, msg); free_xml(msg); if (more) { return EINPROGRESS; } else { return pcmk_rc_ok; } } /*! * \internal * \brief Dispatch data read from IPC source * * \param[in] buffer Data read from IPC * \param[in] length Number of bytes of data in buffer (ignored) * \param[in,out] user_data IPC object * * \return Always 0 (meaning connection is still required) * * \note This function can be used as a main loop IPC dispatch callback. */ static int dispatch_ipc_source_data(const char *buffer, ssize_t length, gpointer user_data) { pcmk_ipc_api_t *api = user_data; CRM_CHECK(api != NULL, return 0); dispatch_ipc_data(buffer, api); return 0; } /*! * \brief Check whether an IPC connection has data available (without main loop) * * \param[in] api IPC API connection * \param[in] timeout_ms If less than 0, poll indefinitely; if 0, poll once * and return immediately; otherwise, poll for up to * this many milliseconds * * \return Standard Pacemaker return code * * \note Callers of pcmk_connect_ipc() using pcmk_ipc_dispatch_poll should call * this function to check whether IPC data is available. Return values of * interest include pcmk_rc_ok meaning data is available, and EAGAIN * meaning no data is available; all other values indicate errors. * \todo This does not allow the caller to poll multiple file descriptors at * once. If there is demand for that, we could add a wrapper for * pcmk__ipc_fd(api->ipc), so the caller can call poll() themselves. */ int pcmk_poll_ipc(const pcmk_ipc_api_t *api, int timeout_ms) { int rc; struct pollfd pollfd = { 0, }; if ((api == NULL) || (api->dispatch_type != pcmk_ipc_dispatch_poll)) { return EINVAL; } rc = pcmk__ipc_fd(api->ipc, &(pollfd.fd)); if (rc != pcmk_rc_ok) { crm_debug("Could not obtain file descriptor for %s IPC: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); return rc; } pollfd.events = POLLIN; rc = poll(&pollfd, 1, timeout_ms); if (rc < 0) { /* Some UNIX systems return negative and set EAGAIN for failure to * allocate memory; standardize the return code in that case */ return (errno == EAGAIN)? ENOMEM : errno; } else if (rc == 0) { return EAGAIN; } return pcmk_rc_ok; } /*! * \brief Dispatch available messages on an IPC connection (without main loop) * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code * * \note Callers of pcmk_connect_ipc() using pcmk_ipc_dispatch_poll should call * this function when IPC data is available. */ void pcmk_dispatch_ipc(pcmk_ipc_api_t *api) { if (api == NULL) { return; } while (crm_ipc_ready(api->ipc) > 0) { if (crm_ipc_read(api->ipc) > 0) { dispatch_ipc_data(crm_ipc_buffer(api->ipc), api); } } } // \return Standard Pacemaker return code static int connect_with_main_loop(pcmk_ipc_api_t *api) { int rc; struct ipc_client_callbacks callbacks = { .dispatch = dispatch_ipc_source_data, .destroy = ipc_post_disconnect, }; rc = pcmk__add_mainloop_ipc(api->ipc, G_PRIORITY_DEFAULT, api, &callbacks, &(api->mainloop_io)); if (rc != pcmk_rc_ok) { return rc; } crm_debug("Connected to %s IPC (attached to main loop)", pcmk_ipc_name(api, true)); /* After this point, api->mainloop_io owns api->ipc, so api->ipc * should not be explicitly freed. */ return pcmk_rc_ok; } // \return Standard Pacemaker return code static int connect_without_main_loop(pcmk_ipc_api_t *api) { int rc = pcmk__connect_generic_ipc(api->ipc); if (rc != pcmk_rc_ok) { crm_ipc_close(api->ipc); } else { crm_debug("Connected to %s IPC (without main loop)", pcmk_ipc_name(api, true)); } return rc; } /*! * \internal * \brief Connect to a Pacemaker daemon via IPC (retrying after soft errors) * * \param[in,out] api IPC API instance * \param[in] dispatch_type How IPC replies should be dispatched * \param[in] attempts How many times to try (in case of soft error) * * \return Standard Pacemaker return code */ int pcmk__connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type, int attempts) { int rc = pcmk_rc_ok; if ((api == NULL) || (attempts < 1)) { return EINVAL; } if (api->ipc == NULL) { api->ipc = crm_ipc_new(pcmk_ipc_name(api, false), api->ipc_size_max); if (api->ipc == NULL) { return ENOMEM; } } if (crm_ipc_connected(api->ipc)) { crm_trace("Already connected to %s", pcmk_ipc_name(api, true)); return pcmk_rc_ok; } api->dispatch_type = dispatch_type; crm_debug("Attempting connection to %s (up to %d time%s)", pcmk_ipc_name(api, true), attempts, pcmk__plural_s(attempts)); for (int remaining = attempts - 1; remaining >= 0; --remaining) { switch (dispatch_type) { case pcmk_ipc_dispatch_main: rc = connect_with_main_loop(api); break; case pcmk_ipc_dispatch_sync: case pcmk_ipc_dispatch_poll: rc = connect_without_main_loop(api); break; } if ((remaining == 0) || ((rc != EAGAIN) && (rc != EALREADY))) { break; // Result is final } // Retry after soft error (interrupted by signal, etc.) pcmk__sleep_ms((attempts - remaining) * 500); crm_debug("Re-attempting connection to %s (%d attempt%s remaining)", pcmk_ipc_name(api, true), remaining, pcmk__plural_s(remaining)); } if (rc != pcmk_rc_ok) { return rc; } if ((api->cmds != NULL) && (api->cmds->post_connect != NULL)) { rc = api->cmds->post_connect(api); if (rc != pcmk_rc_ok) { crm_ipc_close(api->ipc); } } return rc; } /*! * \brief Connect to a Pacemaker daemon via IPC * * \param[in,out] api IPC API instance * \param[in] dispatch_type How IPC replies should be dispatched * * \return Standard Pacemaker return code */ int pcmk_connect_ipc(pcmk_ipc_api_t *api, enum pcmk_ipc_dispatch dispatch_type) { int rc = pcmk__connect_ipc(api, dispatch_type, 2); if (rc != pcmk_rc_ok) { crm_err("Connection to %s failed: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); } return rc; } /*! * \brief Disconnect an IPC API instance * * \param[in,out] api IPC API connection * * \return Standard Pacemaker return code * * \note If the connection is attached to a main loop, this function should be * called before quitting the main loop, to ensure that all memory is * freed. */ void pcmk_disconnect_ipc(pcmk_ipc_api_t *api) { if ((api == NULL) || (api->ipc == NULL)) { return; } switch (api->dispatch_type) { case pcmk_ipc_dispatch_main: { mainloop_io_t *mainloop_io = api->mainloop_io; // Make sure no code with access to api can use these again api->mainloop_io = NULL; api->ipc = NULL; mainloop_del_ipc_client(mainloop_io); // After this point api might have already been freed } break; case pcmk_ipc_dispatch_poll: case pcmk_ipc_dispatch_sync: { crm_ipc_t *ipc = api->ipc; // Make sure no code with access to api can use ipc again api->ipc = NULL; // This should always be the case already, but to be safe api->free_on_disconnect = false; crm_ipc_close(ipc); crm_ipc_destroy(ipc); ipc_post_disconnect(api); } break; } } /*! * \brief Register a callback for IPC API events * * \param[in,out] api IPC API connection * \param[in] callback Callback to register * \param[in] userdata Caller data to pass to callback * * \note This function may be called multiple times to update the callback * and/or user data. The caller remains responsible for freeing * userdata in any case (after the IPC is disconnected, if the * user data is still registered with the IPC). */ void pcmk_register_ipc_callback(pcmk_ipc_api_t *api, pcmk_ipc_callback_t cb, void *user_data) { if (api == NULL) { return; } api->cb = cb; api->user_data = user_data; } /*! * \internal * \brief Send an XML request across an IPC API connection * * \param[in,out] api IPC API connection * \param[in] request XML request to send * * \return Standard Pacemaker return code * * \note Daemon-specific IPC API functions should call this function to send * requests, because it handles different dispatch types appropriately. */ int pcmk__send_ipc_request(pcmk_ipc_api_t *api, const xmlNode *request) { int rc; xmlNode *reply = NULL; enum crm_ipc_flags flags = crm_ipc_flags_none; if ((api == NULL) || (api->ipc == NULL) || (request == NULL)) { return EINVAL; } crm_log_xml_trace(request, "ipc-sent"); // Synchronous dispatch requires waiting for a reply if ((api->dispatch_type == pcmk_ipc_dispatch_sync) && (api->cmds != NULL) && (api->cmds->reply_expected != NULL) && (api->cmds->reply_expected(api, request))) { flags = crm_ipc_client_response; } // The 0 here means a default timeout of 5 seconds rc = crm_ipc_send(api->ipc, request, flags, 0, &reply); if (rc < 0) { return pcmk_legacy2rc(rc); } else if (rc == 0) { return ENODATA; } // With synchronous dispatch, we dispatch any reply now if (reply != NULL) { bool more = call_api_dispatch(api, reply); free_xml(reply); while (more) { rc = crm_ipc_read(api->ipc); if (rc == -EAGAIN) { continue; } else if (rc == -ENOMSG || rc == pcmk_ok) { return pcmk_rc_ok; } else if (rc < 0) { return -rc; } rc = dispatch_ipc_data(crm_ipc_buffer(api->ipc), api); if (rc == pcmk_rc_ok) { more = false; } else if (rc == EINPROGRESS) { more = true; } else { continue; } } } return pcmk_rc_ok; } /*! * \internal * \brief Create the XML for an IPC request to purge a node from the peer cache * * \param[in] api IPC API connection * \param[in] node_name If not NULL, name of node to purge * \param[in] nodeid If not 0, node ID of node to purge * * \return Newly allocated IPC request XML * * \note The controller, fencer, and pacemakerd use the same request syntax, but * the attribute manager uses a different one. The CIB manager doesn't * have any syntax for it. The executor and scheduler don't connect to the * cluster layer and thus don't have or need any syntax for it. * * \todo Modify the attribute manager to accept the common syntax (as well * as its current one, for compatibility with older clients). Modify * the CIB manager to accept and honor the common syntax. Modify the * executor and scheduler to accept the syntax (immediately returning * success), just for consistency. Modify this function to use the * common syntax with all daemons if their version supports it. */ static xmlNode * create_purge_node_request(const pcmk_ipc_api_t *api, const char *node_name, uint32_t nodeid) { xmlNode *request = NULL; const char *client = crm_system_name? crm_system_name : "client"; switch (api->server) { case pcmk_ipc_attrd: request = create_xml_node(NULL, __func__); crm_xml_add(request, F_TYPE, T_ATTRD); crm_xml_add(request, F_ORIG, crm_system_name); crm_xml_add(request, PCMK__XA_TASK, PCMK__ATTRD_CMD_PEER_REMOVE); pcmk__xe_set_bool_attr(request, PCMK__XA_REAP, true); pcmk__xe_add_node(request, node_name, nodeid); break; case pcmk_ipc_controld: case pcmk_ipc_fenced: case pcmk_ipc_pacemakerd: request = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, pcmk_ipc_name(api, false), client, NULL); if (nodeid > 0) { crm_xml_set_id(request, "%lu", (unsigned long) nodeid); } - crm_xml_add(request, XML_ATTR_UNAME, node_name); + crm_xml_add(request, PCMK_XA_UNAME, node_name); break; case pcmk_ipc_based: case pcmk_ipc_execd: case pcmk_ipc_schedulerd: break; } return request; } /*! * \brief Ask a Pacemaker daemon to purge a node from its peer cache * * \param[in,out] api IPC API connection * \param[in] node_name If not NULL, name of node to purge * \param[in] nodeid If not 0, node ID of node to purge * * \return Standard Pacemaker return code * * \note At least one of node_name or nodeid must be specified. */ int pcmk_ipc_purge_node(pcmk_ipc_api_t *api, const char *node_name, uint32_t nodeid) { int rc = 0; xmlNode *request = NULL; if (api == NULL) { return EINVAL; } if ((node_name == NULL) && (nodeid == 0)) { return EINVAL; } request = create_purge_node_request(api, node_name, nodeid); if (request == NULL) { return EOPNOTSUPP; } rc = pcmk__send_ipc_request(api, request); free_xml(request); crm_debug("%s peer cache purge of node %s[%lu]: rc=%d", pcmk_ipc_name(api, true), node_name, (unsigned long) nodeid, rc); return rc; } /* * Generic IPC API (to eventually be deprecated as public API and made internal) */ struct crm_ipc_s { struct pollfd pfd; unsigned int max_buf_size; // maximum bytes we can send or receive over IPC unsigned int buf_size; // size of allocated buffer int msg_size; int need_reply; char *buffer; char *server_name; // server IPC name being connected to qb_ipcc_connection_t *ipc; }; /*! * \brief Create a new (legacy) object for using Pacemaker daemon IPC * * \param[in] name IPC system name to connect to * \param[in] max_size Use a maximum IPC buffer size of at least this size * * \return Newly allocated IPC object on success, NULL otherwise * * \note The caller is responsible for freeing the result using * crm_ipc_destroy(). * \note This should be considered deprecated for use with daemons supported by * pcmk_new_ipc_api(). */ crm_ipc_t * crm_ipc_new(const char *name, size_t max_size) { crm_ipc_t *client = NULL; client = calloc(1, sizeof(crm_ipc_t)); if (client == NULL) { crm_err("Could not create IPC connection: %s", strerror(errno)); return NULL; } client->server_name = strdup(name); if (client->server_name == NULL) { crm_err("Could not create %s IPC connection: %s", name, strerror(errno)); free(client); return NULL; } client->buf_size = pcmk__ipc_buffer_size(max_size); client->buffer = malloc(client->buf_size); if (client->buffer == NULL) { crm_err("Could not create %s IPC connection: %s", name, strerror(errno)); free(client->server_name); free(client); return NULL; } /* Clients initiating connection pick the max buf size */ client->max_buf_size = client->buf_size; client->pfd.fd = -1; client->pfd.events = POLLIN; client->pfd.revents = 0; return client; } /*! * \internal * \brief Connect a generic (not daemon-specific) IPC object * * \param[in,out] ipc Generic IPC object to connect * * \return Standard Pacemaker return code */ int pcmk__connect_generic_ipc(crm_ipc_t *ipc) { uid_t cl_uid = 0; gid_t cl_gid = 0; pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; int rc = pcmk_rc_ok; if (ipc == NULL) { return EINVAL; } ipc->need_reply = FALSE; ipc->ipc = qb_ipcc_connect(ipc->server_name, ipc->buf_size); if (ipc->ipc == NULL) { return errno; } rc = qb_ipcc_fd_get(ipc->ipc, &ipc->pfd.fd); if (rc < 0) { // -errno crm_ipc_close(ipc); return -rc; } rc = pcmk_daemon_user(&cl_uid, &cl_gid); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { crm_ipc_close(ipc); return rc; } rc = is_ipc_provider_expected(ipc->ipc, ipc->pfd.fd, cl_uid, cl_gid, &found_pid, &found_uid, &found_gid); if (rc != pcmk_rc_ok) { if (rc == pcmk_rc_ipc_unauthorized) { crm_info("%s IPC provider authentication failed: process %lld has " "uid %lld (expected %lld) and gid %lld (expected %lld)", ipc->server_name, (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) cl_uid, (long long) found_gid, (long long) cl_gid); } crm_ipc_close(ipc); return rc; } ipc->max_buf_size = qb_ipcc_get_buffer_size(ipc->ipc); if (ipc->max_buf_size > ipc->buf_size) { free(ipc->buffer); ipc->buffer = calloc(ipc->max_buf_size, sizeof(char)); if (ipc->buffer == NULL) { rc = errno; crm_ipc_close(ipc); return rc; } ipc->buf_size = ipc->max_buf_size; } return pcmk_rc_ok; } /*! * \brief Establish an IPC connection to a Pacemaker component * * \param[in,out] client Connection instance obtained from crm_ipc_new() * * \return true on success, false otherwise (in which case errno will be set; * specifically, in case of discovering the remote side is not * authentic, its value is set to ECONNABORTED). */ bool crm_ipc_connect(crm_ipc_t *client) { int rc = pcmk__connect_generic_ipc(client); if (rc == pcmk_rc_ok) { return true; } if ((client != NULL) && (client->ipc == NULL)) { errno = (rc > 0)? rc : ENOTCONN; crm_debug("Could not establish %s IPC connection: %s (%d)", client->server_name, pcmk_rc_str(errno), errno); } else if (rc == pcmk_rc_ipc_unauthorized) { crm_err("%s IPC provider authentication failed", (client == NULL)? "Pacemaker" : client->server_name); errno = ECONNABORTED; } else { crm_perror(LOG_ERR, "Could not verify authenticity of %s IPC provider", (client == NULL)? "Pacemaker" : client->server_name); errno = ENOTCONN; } return false; } void crm_ipc_close(crm_ipc_t * client) { if (client) { if (client->ipc) { qb_ipcc_connection_t *ipc = client->ipc; client->ipc = NULL; qb_ipcc_disconnect(ipc); } } } void crm_ipc_destroy(crm_ipc_t * client) { if (client) { if (client->ipc && qb_ipcc_is_connected(client->ipc)) { crm_notice("Destroying active %s IPC connection", client->server_name); /* The next line is basically unsafe * * If this connection was attached to mainloop and mainloop is active, * the 'disconnected' callback will end up back here and we'll end * up free'ing the memory twice - something that can still happen * even without this if we destroy a connection and it closes before * we call exit */ /* crm_ipc_close(client); */ } else { crm_trace("Destroying inactive %s IPC connection", client->server_name); } free(client->buffer); free(client->server_name); free(client); } } /*! * \internal * \brief Get the file descriptor for a generic IPC object * * \param[in,out] ipc Generic IPC object to get file descriptor for * \param[out] fd Where to store file descriptor * * \return Standard Pacemaker return code */ int pcmk__ipc_fd(crm_ipc_t *ipc, int *fd) { if ((ipc == NULL) || (fd == NULL)) { return EINVAL; } if ((ipc->ipc == NULL) || (ipc->pfd.fd < 0)) { return ENOTCONN; } *fd = ipc->pfd.fd; return pcmk_rc_ok; } int crm_ipc_get_fd(crm_ipc_t * client) { int fd = -1; if (pcmk__ipc_fd(client, &fd) != pcmk_rc_ok) { crm_err("Could not obtain file descriptor for %s IPC", ((client == NULL)? "unspecified" : client->server_name)); errno = EINVAL; return -EINVAL; } return fd; } bool crm_ipc_connected(crm_ipc_t * client) { bool rc = FALSE; if (client == NULL) { crm_trace("No client"); return FALSE; } else if (client->ipc == NULL) { crm_trace("No connection"); return FALSE; } else if (client->pfd.fd < 0) { crm_trace("Bad descriptor"); return FALSE; } rc = qb_ipcc_is_connected(client->ipc); if (rc == FALSE) { client->pfd.fd = -EINVAL; } return rc; } /*! * \brief Check whether an IPC connection is ready to be read * * \param[in,out] client Connection to check * * \return Positive value if ready to be read, 0 if not ready, -errno on error */ int crm_ipc_ready(crm_ipc_t *client) { int rc; CRM_ASSERT(client != NULL); if (!crm_ipc_connected(client)) { return -ENOTCONN; } client->pfd.revents = 0; rc = poll(&(client->pfd), 1, 0); return (rc < 0)? -errno : rc; } // \return Standard Pacemaker return code static int crm_ipc_decompress(crm_ipc_t * client) { pcmk__ipc_header_t *header = (pcmk__ipc_header_t *)(void*)client->buffer; if (header->size_compressed) { int rc = 0; unsigned int size_u = 1 + header->size_uncompressed; /* never let buf size fall below our max size required for ipc reads. */ unsigned int new_buf_size = QB_MAX((sizeof(pcmk__ipc_header_t) + size_u), client->max_buf_size); char *uncompressed = calloc(1, new_buf_size); crm_trace("Decompressing message data %u bytes into %u bytes", header->size_compressed, size_u); rc = BZ2_bzBuffToBuffDecompress(uncompressed + sizeof(pcmk__ipc_header_t), &size_u, client->buffer + sizeof(pcmk__ipc_header_t), header->size_compressed, 1, 0); rc = pcmk__bzlib2rc(rc); if (rc != pcmk_rc_ok) { crm_err("Decompression failed: %s " CRM_XS " rc=%d", pcmk_rc_str(rc), rc); free(uncompressed); return rc; } /* * This assert no longer holds true. For an identical msg, some clients may * require compression, and others may not. If that same msg (event) is sent * to multiple clients, it could result in some clients receiving a compressed * msg even though compression was not explicitly required for them. * * CRM_ASSERT((header->size_uncompressed + sizeof(pcmk__ipc_header_t)) >= ipc_buffer_max); */ CRM_ASSERT(size_u == header->size_uncompressed); memcpy(uncompressed, client->buffer, sizeof(pcmk__ipc_header_t)); /* Preserve the header */ header = (pcmk__ipc_header_t *)(void*)uncompressed; free(client->buffer); client->buf_size = new_buf_size; client->buffer = uncompressed; } CRM_ASSERT(client->buffer[sizeof(pcmk__ipc_header_t) + header->size_uncompressed - 1] == 0); return pcmk_rc_ok; } long crm_ipc_read(crm_ipc_t * client) { pcmk__ipc_header_t *header = NULL; CRM_ASSERT(client != NULL); CRM_ASSERT(client->ipc != NULL); CRM_ASSERT(client->buffer != NULL); client->buffer[0] = 0; client->msg_size = qb_ipcc_event_recv(client->ipc, client->buffer, client->buf_size, 0); if (client->msg_size >= 0) { int rc = crm_ipc_decompress(client); if (rc != pcmk_rc_ok) { return pcmk_rc2legacy(rc); } header = (pcmk__ipc_header_t *)(void*)client->buffer; if (!pcmk__valid_ipc_header(header)) { return -EBADMSG; } crm_trace("Received %s IPC event %d size=%u rc=%d text='%.100s'", client->server_name, header->qb.id, header->qb.size, client->msg_size, client->buffer + sizeof(pcmk__ipc_header_t)); } else { crm_trace("No message received from %s IPC: %s", client->server_name, pcmk_strerror(client->msg_size)); if (client->msg_size == -EAGAIN) { return -EAGAIN; } } if (!crm_ipc_connected(client) || client->msg_size == -ENOTCONN) { crm_err("Connection to %s IPC failed", client->server_name); } if (header) { /* Data excluding the header */ return header->size_uncompressed; } return -ENOMSG; } const char * crm_ipc_buffer(crm_ipc_t * client) { CRM_ASSERT(client != NULL); return client->buffer + sizeof(pcmk__ipc_header_t); } uint32_t crm_ipc_buffer_flags(crm_ipc_t * client) { pcmk__ipc_header_t *header = NULL; CRM_ASSERT(client != NULL); if (client->buffer == NULL) { return 0; } header = (pcmk__ipc_header_t *)(void*)client->buffer; return header->flags; } const char * crm_ipc_name(crm_ipc_t * client) { CRM_ASSERT(client != NULL); return client->server_name; } // \return Standard Pacemaker return code static int internal_ipc_get_reply(crm_ipc_t *client, int request_id, int ms_timeout, ssize_t *bytes) { time_t timeout = time(NULL) + 1 + (ms_timeout / 1000); int rc = pcmk_rc_ok; /* get the reply */ crm_trace("Waiting on reply to %s IPC message %d", client->server_name, request_id); do { *bytes = qb_ipcc_recv(client->ipc, client->buffer, client->buf_size, 1000); if (*bytes > 0) { pcmk__ipc_header_t *hdr = NULL; rc = crm_ipc_decompress(client); if (rc != pcmk_rc_ok) { return rc; } hdr = (pcmk__ipc_header_t *)(void*)client->buffer; if (hdr->qb.id == request_id) { /* Got it */ break; } else if (hdr->qb.id < request_id) { xmlNode *bad = string2xml(crm_ipc_buffer(client)); crm_err("Discarding old reply %d (need %d)", hdr->qb.id, request_id); crm_log_xml_notice(bad, "OldIpcReply"); } else { xmlNode *bad = string2xml(crm_ipc_buffer(client)); crm_err("Discarding newer reply %d (need %d)", hdr->qb.id, request_id); crm_log_xml_notice(bad, "ImpossibleReply"); CRM_ASSERT(hdr->qb.id <= request_id); } } else if (!crm_ipc_connected(client)) { crm_err("%s IPC provider disconnected while waiting for message %d", client->server_name, request_id); break; } } while (time(NULL) < timeout); if (*bytes < 0) { rc = (int) -*bytes; // System errno } return rc; } /*! * \brief Send an IPC XML message * * \param[in,out] client Connection to IPC server * \param[in] message XML message to send * \param[in] flags Bitmask of crm_ipc_flags * \param[in] ms_timeout Give up if not sent within this much time * (5 seconds if 0, or no timeout if negative) * \param[out] reply Reply from server (or NULL if none) * * \return Negative errno on error, otherwise size of reply received in bytes * if reply was needed, otherwise number of bytes sent */ int crm_ipc_send(crm_ipc_t *client, const xmlNode *message, enum crm_ipc_flags flags, int32_t ms_timeout, xmlNode **reply) { int rc = 0; ssize_t qb_rc = 0; ssize_t bytes = 0; struct iovec *iov; static uint32_t id = 0; static int factor = 8; pcmk__ipc_header_t *header; if (client == NULL) { crm_notice("Can't send IPC request without connection (bug?): %.100s", message); return -ENOTCONN; } else if (!crm_ipc_connected(client)) { /* Don't even bother */ crm_notice("Can't send %s IPC requests: Connection closed", client->server_name); return -ENOTCONN; } if (ms_timeout == 0) { ms_timeout = 5000; } if (client->need_reply) { qb_rc = qb_ipcc_recv(client->ipc, client->buffer, client->buf_size, ms_timeout); if (qb_rc < 0) { crm_warn("Sending %s IPC disabled until pending reply received", client->server_name); return -EALREADY; } else { crm_notice("Sending %s IPC re-enabled after pending reply received", client->server_name); client->need_reply = FALSE; } } id++; CRM_LOG_ASSERT(id != 0); /* Crude wrap-around detection */ rc = pcmk__ipc_prepare_iov(id, message, client->max_buf_size, &iov, &bytes); if (rc != pcmk_rc_ok) { crm_warn("Couldn't prepare %s IPC request: %s " CRM_XS " rc=%d", client->server_name, pcmk_rc_str(rc), rc); return pcmk_rc2legacy(rc); } header = iov[0].iov_base; pcmk__set_ipc_flags(header->flags, client->server_name, flags); if (pcmk_is_set(flags, crm_ipc_proxied)) { /* Don't look for a synchronous response */ pcmk__clear_ipc_flags(flags, "client", crm_ipc_client_response); } if(header->size_compressed) { if(factor < 10 && (client->max_buf_size / 10) < (bytes / factor)) { crm_notice("Compressed message exceeds %d0%% of configured IPC " "limit (%u bytes); consider setting PCMK_ipc_buffer to " "%u or higher", factor, client->max_buf_size, 2 * client->max_buf_size); factor++; } } crm_trace("Sending %s IPC request %d of %u bytes using %dms timeout", client->server_name, header->qb.id, header->qb.size, ms_timeout); if ((ms_timeout > 0) || !pcmk_is_set(flags, crm_ipc_client_response)) { time_t timeout = time(NULL) + 1 + (ms_timeout / 1000); do { /* @TODO Is this check really needed? Won't qb_ipcc_sendv() return * an error if it's not connected? */ if (!crm_ipc_connected(client)) { goto send_cleanup; } qb_rc = qb_ipcc_sendv(client->ipc, iov, 2); } while ((qb_rc == -EAGAIN) && (time(NULL) < timeout)); rc = (int) qb_rc; // Negative of system errno, or bytes sent if (qb_rc <= 0) { goto send_cleanup; } else if (!pcmk_is_set(flags, crm_ipc_client_response)) { crm_trace("Not waiting for reply to %s IPC request %d", client->server_name, header->qb.id); goto send_cleanup; } rc = internal_ipc_get_reply(client, header->qb.id, ms_timeout, &bytes); if (rc != pcmk_rc_ok) { /* We didn't get the reply in time, so disable future sends for now. * The only alternative would be to close the connection since we * don't know how to detect and discard out-of-sequence replies. * * @TODO Implement out-of-sequence detection */ client->need_reply = TRUE; } rc = (int) bytes; // Negative system errno, or size of reply received } else { // No timeout, and client response needed do { qb_rc = qb_ipcc_sendv_recv(client->ipc, iov, 2, client->buffer, client->buf_size, -1); } while ((qb_rc == -EAGAIN) && crm_ipc_connected(client)); rc = (int) qb_rc; // Negative system errno, or size of reply received } if (rc > 0) { pcmk__ipc_header_t *hdr = (pcmk__ipc_header_t *)(void*)client->buffer; crm_trace("Received %d-byte reply %d to %s IPC %d: %.100s", rc, hdr->qb.id, client->server_name, header->qb.id, crm_ipc_buffer(client)); if (reply) { *reply = string2xml(crm_ipc_buffer(client)); } } else { crm_trace("No reply to %s IPC %d: rc=%d", client->server_name, header->qb.id, rc); } send_cleanup: if (!crm_ipc_connected(client)) { crm_notice("Couldn't send %s IPC request %d: Connection closed " CRM_XS " rc=%d", client->server_name, header->qb.id, rc); } else if (rc == -ETIMEDOUT) { crm_warn("%s IPC request %d failed: %s after %dms " CRM_XS " rc=%d", client->server_name, header->qb.id, pcmk_strerror(rc), ms_timeout, rc); crm_write_blackbox(0, NULL); } else if (rc <= 0) { crm_warn("%s IPC request %d failed: %s " CRM_XS " rc=%d", client->server_name, header->qb.id, ((rc == 0)? "No bytes sent" : pcmk_strerror(rc)), rc); } pcmk_free_ipc_event(iov); return rc; } /*! * \brief Ensure an IPC provider has expected user or group * * \param[in] qb_ipc libqb client connection if available * \param[in] sock Connected Unix socket for IPC * \param[in] refuid Expected user ID * \param[in] refgid Expected group ID * \param[out] gotpid If not NULL, where to store provider's actual process ID * (or 1 on platforms where ID is not available) * \param[out] gotuid If not NULL, where to store provider's actual user ID * \param[out] gotgid If not NULL, where to store provider's actual group ID * * \return Standard Pacemaker return code * \note An actual user ID of 0 (root) will always be considered authorized, * regardless of the expected values provided. The caller can use the * output arguments to be stricter than this function. */ static int is_ipc_provider_expected(qb_ipcc_connection_t *qb_ipc, int sock, uid_t refuid, gid_t refgid, pid_t *gotpid, uid_t *gotuid, gid_t *gotgid) { int rc = EOPNOTSUPP; pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; #ifdef HAVE_QB_IPCC_AUTH_GET if (qb_ipc != NULL) { rc = qb_ipcc_auth_get(qb_ipc, &found_pid, &found_uid, &found_gid); rc = -rc; // libqb returns 0 or -errno if (rc == pcmk_rc_ok) { goto found; } } #endif #ifdef HAVE_UCRED { struct ucred ucred; socklen_t ucred_len = sizeof(ucred); if (getsockopt(sock, SOL_SOCKET, SO_PEERCRED, &ucred, &ucred_len) < 0) { rc = errno; } else if (ucred_len != sizeof(ucred)) { rc = EOPNOTSUPP; } else { found_pid = ucred.pid; found_uid = ucred.uid; found_gid = ucred.gid; goto found; } } #endif #ifdef HAVE_SOCKPEERCRED { struct sockpeercred sockpeercred; socklen_t sockpeercred_len = sizeof(sockpeercred); if (getsockopt(sock, SOL_SOCKET, SO_PEERCRED, &sockpeercred, &sockpeercred_len) < 0) { rc = errno; } else if (sockpeercred_len != sizeof(sockpeercred)) { rc = EOPNOTSUPP; } else { found_pid = sockpeercred.pid; found_uid = sockpeercred.uid; found_gid = sockpeercred.gid; goto found; } } #endif #ifdef HAVE_GETPEEREID // For example, FreeBSD if (getpeereid(sock, &found_uid, &found_gid) < 0) { rc = errno; } else { found_pid = PCMK__SPECIAL_PID; goto found; } #endif #ifdef HAVE_GETPEERUCRED { ucred_t *ucred = NULL; if (getpeerucred(sock, &ucred) < 0) { rc = errno; } else { found_pid = ucred_getpid(ucred); found_uid = ucred_geteuid(ucred); found_gid = ucred_getegid(ucred); ucred_free(ucred); goto found; } } #endif return rc; // If we get here, nothing succeeded found: if (gotpid != NULL) { *gotpid = found_pid; } if (gotuid != NULL) { *gotuid = found_uid; } if (gotgid != NULL) { *gotgid = found_gid; } if ((found_uid != 0) && (found_uid != refuid) && (found_gid != refgid)) { return pcmk_rc_ipc_unauthorized; } return pcmk_rc_ok; } int crm_ipc_is_authentic_process(int sock, uid_t refuid, gid_t refgid, pid_t *gotpid, uid_t *gotuid, gid_t *gotgid) { int ret = is_ipc_provider_expected(NULL, sock, refuid, refgid, gotpid, gotuid, gotgid); /* The old function had some very odd return codes*/ if (ret == 0) { return 1; } else if (ret == pcmk_rc_ipc_unauthorized) { return 0; } else { return pcmk_rc2legacy(ret); } } int pcmk__ipc_is_authentic_process_active(const char *name, uid_t refuid, gid_t refgid, pid_t *gotpid) { static char last_asked_name[PATH_MAX / 2] = ""; /* log spam prevention */ int fd; int rc = pcmk_rc_ipc_unresponsive; int auth_rc = 0; int32_t qb_rc; pid_t found_pid = 0; uid_t found_uid = 0; gid_t found_gid = 0; qb_ipcc_connection_t *c; #ifdef HAVE_QB_IPCC_CONNECT_ASYNC struct pollfd pollfd = { 0, }; int poll_rc; c = qb_ipcc_connect_async(name, 0, &(pollfd.fd)); #else c = qb_ipcc_connect(name, 0); #endif if (c == NULL) { crm_info("Could not connect to %s IPC: %s", name, strerror(errno)); rc = pcmk_rc_ipc_unresponsive; goto bail; } #ifdef HAVE_QB_IPCC_CONNECT_ASYNC pollfd.events = POLLIN; do { poll_rc = poll(&pollfd, 1, 2000); } while ((poll_rc == -1) && (errno == EINTR)); if ((poll_rc <= 0) || (qb_ipcc_connect_continue(c) != 0)) { crm_info("Could not connect to %s IPC: %s", name, (poll_rc == 0)?"timeout":strerror(errno)); rc = pcmk_rc_ipc_unresponsive; if (poll_rc > 0) { c = NULL; // qb_ipcc_connect_continue cleaned up for us } goto bail; } #endif qb_rc = qb_ipcc_fd_get(c, &fd); if (qb_rc != 0) { rc = (int) -qb_rc; // System errno crm_err("Could not get fd from %s IPC: %s " CRM_XS " rc=%d", name, pcmk_rc_str(rc), rc); goto bail; } auth_rc = is_ipc_provider_expected(c, fd, refuid, refgid, &found_pid, &found_uid, &found_gid); if (auth_rc == pcmk_rc_ipc_unauthorized) { crm_err("Daemon (IPC %s) effectively blocked with unauthorized" " process %lld (uid: %lld, gid: %lld)", name, (long long) PCMK__SPECIAL_PID_AS_0(found_pid), (long long) found_uid, (long long) found_gid); rc = pcmk_rc_ipc_unauthorized; goto bail; } if (auth_rc != pcmk_rc_ok) { rc = auth_rc; crm_err("Could not get peer credentials from %s IPC: %s " CRM_XS " rc=%d", name, pcmk_rc_str(rc), rc); goto bail; } if (gotpid != NULL) { *gotpid = found_pid; } rc = pcmk_rc_ok; if ((found_uid != refuid || found_gid != refgid) && strncmp(last_asked_name, name, sizeof(last_asked_name))) { if ((found_uid == 0) && (refuid != 0)) { crm_warn("Daemon (IPC %s) runs as root, whereas the expected" " credentials are %lld:%lld, hazard of violating" " the least privilege principle", name, (long long) refuid, (long long) refgid); } else { crm_notice("Daemon (IPC %s) runs as %lld:%lld, whereas the" " expected credentials are %lld:%lld, which may" " mean a different set of privileges than expected", name, (long long) found_uid, (long long) found_gid, (long long) refuid, (long long) refgid); } memccpy(last_asked_name, name, '\0', sizeof(last_asked_name)); } bail: if (c != NULL) { qb_ipcc_disconnect(c); } return rc; } diff --git a/lib/common/ipc_controld.c b/lib/common/ipc_controld.c index 7f5a491b5a..524bb443e8 100644 --- a/lib/common/ipc_controld.c +++ b/lib/common/ipc_controld.c @@ -1,657 +1,657 @@ /* * Copyright 2020-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include "crmcommon_private.h" struct controld_api_private_s { char *client_uuid; unsigned int replies_expected; }; /*! * \internal * \brief Get a string representation of a controller API reply type * * \param[in] reply Controller API reply type * * \return String representation of a controller API reply type */ const char * pcmk__controld_api_reply2str(enum pcmk_controld_api_reply reply) { switch (reply) { case pcmk_controld_reply_reprobe: return "reprobe"; case pcmk_controld_reply_info: return "info"; case pcmk_controld_reply_resource: return "resource"; case pcmk_controld_reply_ping: return "ping"; case pcmk_controld_reply_nodes: return "nodes"; default: return "unknown"; } } // \return Standard Pacemaker return code static int new_data(pcmk_ipc_api_t *api) { struct controld_api_private_s *private = NULL; api->api_data = calloc(1, sizeof(struct controld_api_private_s)); if (api->api_data == NULL) { return errno; } private = api->api_data; /* This is set to the PID because that's how it was always done, but PIDs * are not unique because clients can be remote. The value appears to be * unused other than as part of F_CRM_SYS_FROM in IPC requests, which is * only compared against the internal system names (CRM_SYSTEM_TENGINE, * etc.), so it shouldn't be a problem. */ private->client_uuid = pcmk__getpid_s(); /* @TODO Implement a call ID model similar to the CIB, executor, and fencer * IPC APIs, so that requests and replies can be matched, and * duplicate replies can be discarded. */ return pcmk_rc_ok; } static void free_data(void *data) { free(((struct controld_api_private_s *) data)->client_uuid); free(data); } // \return Standard Pacemaker return code static int post_connect(pcmk_ipc_api_t *api) { /* The controller currently requires clients to register via a hello * request, but does not reply back. */ struct controld_api_private_s *private = api->api_data; const char *client_name = crm_system_name? crm_system_name : "client"; xmlNode *hello; int rc; hello = create_hello_message(private->client_uuid, client_name, PCMK__CONTROLD_API_MAJOR, PCMK__CONTROLD_API_MINOR); rc = pcmk__send_ipc_request(api, hello); free_xml(hello); if (rc != pcmk_rc_ok) { crm_info("Could not send IPC hello to %s: %s " CRM_XS " rc=%s", pcmk_ipc_name(api, true), pcmk_rc_str(rc), rc); } else { crm_debug("Sent IPC hello to %s", pcmk_ipc_name(api, true)); } return rc; } static void set_node_info_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_info; if (msg_data == NULL) { return; } data->data.node_info.have_quorum = pcmk__xe_attr_is_true(msg_data, PCMK_XA_HAVE_QUORUM); data->data.node_info.is_remote = pcmk__xe_attr_is_true(msg_data, XML_NODE_IS_REMOTE); /* Integer node_info.id is currently valid only for Corosync nodes. * * @TODO: Improve handling after crm_node_t is refactored to handle layer- * specific data better. */ crm_element_value_int(msg_data, PCMK_XA_ID, &(data->data.node_info.id)); data->data.node_info.uuid = crm_element_value(msg_data, PCMK_XA_ID); - data->data.node_info.uname = crm_element_value(msg_data, XML_ATTR_UNAME); + data->data.node_info.uname = crm_element_value(msg_data, PCMK_XA_UNAME); data->data.node_info.state = crm_element_value(msg_data, PCMK__XA_CRMD); } static void set_ping_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { data->reply_type = pcmk_controld_reply_ping; if (msg_data == NULL) { return; } data->data.ping.sys_from = crm_element_value(msg_data, XML_PING_ATTR_SYSFROM); data->data.ping.fsa_state = crm_element_value(msg_data, XML_PING_ATTR_CRMDSTATE); data->data.ping.result = crm_element_value(msg_data, XML_PING_ATTR_STATUS); } static void set_nodes_data(pcmk_controld_api_reply_t *data, xmlNode *msg_data) { pcmk_controld_api_node_t *node_info; data->reply_type = pcmk_controld_reply_nodes; for (xmlNode *node = first_named_child(msg_data, XML_CIB_TAG_NODE); node != NULL; node = crm_next_same_xml(node)) { long long id_ll = 0; node_info = calloc(1, sizeof(pcmk_controld_api_node_t)); crm_element_value_ll(node, PCMK_XA_ID, &id_ll); if (id_ll > 0) { node_info->id = id_ll; } - node_info->uname = crm_element_value(node, XML_ATTR_UNAME); + node_info->uname = crm_element_value(node, PCMK_XA_UNAME); node_info->state = crm_element_value(node, PCMK__XA_IN_CCM); data->data.nodes = g_list_prepend(data->data.nodes, node_info); } } static bool reply_expected(pcmk_ipc_api_t *api, const xmlNode *request) { // We only need to handle commands that API functions can send return pcmk__str_any_of(crm_element_value(request, F_CRM_TASK), PCMK__CONTROLD_CMD_NODES, CRM_OP_LRM_DELETE, CRM_OP_LRM_FAIL, CRM_OP_NODE_INFO, CRM_OP_PING, CRM_OP_REPROBE, CRM_OP_RM_NODE_CACHE, NULL); } static bool dispatch(pcmk_ipc_api_t *api, xmlNode *reply) { struct controld_api_private_s *private = api->api_data; crm_exit_t status = CRM_EX_OK; xmlNode *msg_data = NULL; const char *value = NULL; pcmk_controld_api_reply_t reply_data = { pcmk_controld_reply_unknown, NULL, NULL, }; if (pcmk__xe_is(reply, "ack")) { /* ACKs are trivial responses that do not count toward expected replies, * and do not have all the fields that validation requires, so skip that * processing. */ return private->replies_expected > 0; } if (private->replies_expected > 0) { private->replies_expected--; } // Do some basic validation of the reply /* @TODO We should be able to verify that value is always a response, but * currently the controller doesn't always properly set the type. Even * if we fix the controller, we'll still need to handle replies from * old versions (feature set could be used to differentiate). */ value = crm_element_value(reply, F_CRM_MSG_TYPE); if (!pcmk__str_any_of(value, PCMK__VALUE_REQUEST, PCMK__VALUE_RESPONSE, NULL)) { crm_info("Unrecognizable message from controller: " "invalid message type '%s'", pcmk__s(value, "")); status = CRM_EX_PROTOCOL; goto done; } if (pcmk__str_empty(crm_element_value(reply, XML_ATTR_REFERENCE))) { crm_info("Unrecognizable message from controller: no reference"); status = CRM_EX_PROTOCOL; goto done; } value = crm_element_value(reply, F_CRM_TASK); if (pcmk__str_empty(value)) { crm_info("Unrecognizable message from controller: no command name"); status = CRM_EX_PROTOCOL; goto done; } // Parse useful info from reply reply_data.feature_set = crm_element_value(reply, PCMK_XA_VERSION); reply_data.host_from = crm_element_value(reply, F_CRM_HOST_FROM); msg_data = get_message_xml(reply, F_CRM_DATA); if (!strcmp(value, CRM_OP_REPROBE)) { reply_data.reply_type = pcmk_controld_reply_reprobe; } else if (!strcmp(value, CRM_OP_NODE_INFO)) { set_node_info_data(&reply_data, msg_data); } else if (!strcmp(value, CRM_OP_INVOKE_LRM)) { reply_data.reply_type = pcmk_controld_reply_resource; reply_data.data.resource.node_state = msg_data; } else if (!strcmp(value, CRM_OP_PING)) { set_ping_data(&reply_data, msg_data); } else if (!strcmp(value, PCMK__CONTROLD_CMD_NODES)) { set_nodes_data(&reply_data, msg_data); } else { crm_info("Unrecognizable message from controller: unknown command '%s'", value); status = CRM_EX_PROTOCOL; } done: pcmk__call_ipc_callback(api, pcmk_ipc_event_reply, status, &reply_data); // Free any reply data that was allocated if (pcmk__str_eq(value, PCMK__CONTROLD_CMD_NODES, pcmk__str_casei)) { g_list_free_full(reply_data.data.nodes, free); } return false; // No further replies needed } pcmk__ipc_methods_t * pcmk__controld_api_methods(void) { pcmk__ipc_methods_t *cmds = calloc(1, sizeof(pcmk__ipc_methods_t)); if (cmds != NULL) { cmds->new_data = new_data; cmds->free_data = free_data; cmds->post_connect = post_connect; cmds->reply_expected = reply_expected; cmds->dispatch = dispatch; } return cmds; } /*! * \internal * \brief Create XML for a controller IPC request * * \param[in] api Controller connection * \param[in] op Controller IPC command name * \param[in] node Node name to set as destination host * \param[in] msg_data XML to attach to request as message data * * \return Newly allocated XML for request */ static xmlNode * create_controller_request(const pcmk_ipc_api_t *api, const char *op, const char *node, xmlNode *msg_data) { struct controld_api_private_s *private = NULL; const char *sys_to = NULL; if (api == NULL) { return NULL; } private = api->api_data; if ((node == NULL) && !strcmp(op, CRM_OP_PING)) { sys_to = CRM_SYSTEM_DC; } else { sys_to = CRM_SYSTEM_CRMD; } return create_request(op, msg_data, node, sys_to, (crm_system_name? crm_system_name : "client"), private->client_uuid); } // \return Standard Pacemaker return code static int send_controller_request(pcmk_ipc_api_t *api, const xmlNode *request, bool reply_is_expected) { if (crm_element_value(request, XML_ATTR_REFERENCE) == NULL) { return EINVAL; } if (reply_is_expected) { struct controld_api_private_s *private = api->api_data; private->replies_expected++; } return pcmk__send_ipc_request(api, request); } static xmlNode * create_reprobe_message_data(const char *target_node, const char *router_node) { xmlNode *msg_data; msg_data = create_xml_node(NULL, "data_for_" CRM_OP_REPROBE); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if ((router_node != NULL) && !pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } return msg_data; } /*! * \brief Send a reprobe controller operation * * \param[in,out] api Controller connection * \param[in] target_node Name of node to reprobe * \param[in] router_node Router node for host * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_reprobe. */ int pcmk_controld_api_reprobe(pcmk_ipc_api_t *api, const char *target_node, const char *router_node) { xmlNode *request; xmlNode *msg_data; int rc = pcmk_rc_ok; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } crm_debug("Sending %s IPC request to reprobe %s via %s", pcmk_ipc_name(api, true), pcmk__s(target_node, "local node"), pcmk__s(router_node, "local node")); msg_data = create_reprobe_message_data(target_node, router_node); request = create_controller_request(api, CRM_OP_REPROBE, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Send a "node info" controller operation * * \param[in,out] api Controller connection * \param[in] nodeid ID of node to get info for (or 0 for local node) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_info. */ int pcmk_controld_api_node_info(pcmk_ipc_api_t *api, uint32_t nodeid) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_NODE_INFO, NULL, NULL); if (request == NULL) { return EINVAL; } if (nodeid > 0) { crm_xml_set_id(request, "%lu", (unsigned long) nodeid); } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for status * * \param[in,out] api Controller connection * \param[in] node_name Name of node whose status is desired (NULL for DC) * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_ping. */ int pcmk_controld_api_ping(pcmk_ipc_api_t *api, const char *node_name) { xmlNode *request; int rc = pcmk_rc_ok; request = create_controller_request(api, CRM_OP_PING, node_name, NULL); if (request == NULL) { return EINVAL; } rc = send_controller_request(api, request, true); free_xml(request); return rc; } /*! * \brief Ask the controller for cluster information * * \param[in,out] api Controller connection * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_nodes. */ int pcmk_controld_api_list_nodes(pcmk_ipc_api_t *api) { xmlNode *request; int rc = EINVAL; request = create_controller_request(api, PCMK__CONTROLD_CMD_NODES, NULL, NULL); if (request != NULL) { rc = send_controller_request(api, request, true); free_xml(request); } return rc; } // \return Standard Pacemaker return code static int controller_resource_op(pcmk_ipc_api_t *api, const char *op, const char *target_node, const char *router_node, bool cib_only, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { int rc = pcmk_rc_ok; char *key; xmlNode *request, *msg_data, *xml_rsc, *params; if (api == NULL) { return EINVAL; } if (router_node == NULL) { router_node = target_node; } msg_data = create_xml_node(NULL, XML_GRAPH_TAG_RSC_OP); /* The controller logs the transition key from resource op requests, so we * need to have *something* for it. * @TODO don't use "crm-resource" */ key = pcmk__transition_key(0, getpid(), 0, "xxxxxxxx-xrsc-opxx-xcrm-resourcexxxx"); crm_xml_add(msg_data, XML_ATTR_TRANSITION_KEY, key); free(key); crm_xml_add(msg_data, XML_LRM_ATTR_TARGET, target_node); if (!pcmk__str_eq(router_node, target_node, pcmk__str_casei)) { crm_xml_add(msg_data, XML_LRM_ATTR_ROUTER_NODE, router_node); } if (cib_only) { // Indicate that only the CIB needs to be cleaned crm_xml_add(msg_data, PCMK__XA_MODE, XML_TAG_CIB); } xml_rsc = create_xml_node(msg_data, XML_CIB_TAG_RESOURCE); crm_xml_add(xml_rsc, PCMK_XA_ID, rsc_id); crm_xml_add(xml_rsc, PCMK__XA_LONG_ID, rsc_long_id); crm_xml_add(xml_rsc, PCMK_XA_CLASS, standard); crm_xml_add(xml_rsc, PCMK_XA_PROVIDER, provider); crm_xml_add(xml_rsc, PCMK_XA_TYPE, type); params = create_xml_node(msg_data, XML_TAG_ATTRS); crm_xml_add(params, PCMK_XA_CRM_FEATURE_SET, CRM_FEATURE_SET); // The controller parses the timeout from the request key = crm_meta_name(XML_ATTR_TIMEOUT); crm_xml_add(params, key, "60000"); /* 1 minute */ //@TODO pass as arg free(key); request = create_controller_request(api, op, router_node, msg_data); rc = send_controller_request(api, request, true); free_xml(msg_data); free_xml(request); return rc; } /*! * \brief Ask the controller to fail a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to fail * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource to fail * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_fail(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type) { crm_debug("Sending %s IPC request to fail %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_FAIL, target_node, router_node, false, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Ask the controller to refresh a resource * * \param[in,out] api Controller connection * \param[in] target_node Name of node resource is on * \param[in] router_node Router node for target * \param[in] rsc_id ID of resource to refresh * \param[in] rsc_long_id Long ID of resource (if any) * \param[in] standard Standard of resource * \param[in] provider Provider of resource (if any) * \param[in] type Type of resource * \param[in] cib_only If true, clean resource from CIB only * * \return Standard Pacemaker return code * \note Event callback will get a reply of type pcmk_controld_reply_resource. */ int pcmk_controld_api_refresh(pcmk_ipc_api_t *api, const char *target_node, const char *router_node, const char *rsc_id, const char *rsc_long_id, const char *standard, const char *provider, const char *type, bool cib_only) { crm_debug("Sending %s IPC request to refresh %s (a.k.a. %s) on %s via %s", pcmk_ipc_name(api, true), pcmk__s(rsc_id, "unknown resource"), pcmk__s(rsc_long_id, "no other names"), pcmk__s(target_node, "unspecified node"), pcmk__s(router_node, "unspecified node")); return controller_resource_op(api, CRM_OP_LRM_DELETE, target_node, router_node, cib_only, rsc_id, rsc_long_id, standard, provider, type); } /*! * \brief Get the number of IPC replies currently expected from the controller * * \param[in] api Controller IPC API connection * * \return Number of replies expected */ unsigned int pcmk_controld_api_replies_expected(const pcmk_ipc_api_t *api) { struct controld_api_private_s *private = api->api_data; return private->replies_expected; } /*! * \brief Create XML for a controller IPC "hello" message * * \deprecated This function is deprecated as part of the public C API. */ // \todo make this static to this file when breaking API backward compatibility xmlNode * create_hello_message(const char *uuid, const char *client_name, const char *major_version, const char *minor_version) { xmlNode *hello_node = NULL; xmlNode *hello = NULL; if (pcmk__str_empty(uuid) || pcmk__str_empty(client_name) || pcmk__str_empty(major_version) || pcmk__str_empty(minor_version)) { crm_err("Could not create IPC hello message from %s (UUID %s): " "missing information", client_name? client_name : "unknown client", uuid? uuid : "unknown"); return NULL; } hello_node = create_xml_node(NULL, XML_TAG_OPTIONS); if (hello_node == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Message data creation failed", client_name, uuid); return NULL; } crm_xml_add(hello_node, "major_version", major_version); crm_xml_add(hello_node, "minor_version", minor_version); crm_xml_add(hello_node, "client_name", client_name); crm_xml_add(hello_node, "client_uuid", uuid); hello = create_request(CRM_OP_HELLO, hello_node, NULL, NULL, client_name, uuid); if (hello == NULL) { crm_err("Could not create IPC hello message from %s (UUID %s): " "Request creation failed", client_name, uuid); return NULL; } free_xml(hello_node); crm_trace("Created hello message from %s (UUID %s)", client_name, uuid); return hello; } diff --git a/lib/pacemaker/pcmk_cluster_queries.c b/lib/pacemaker/pcmk_cluster_queries.c index 800fcd4817..c3ab1f51c0 100644 --- a/lib/pacemaker/pcmk_cluster_queries.c +++ b/lib/pacemaker/pcmk_cluster_queries.c @@ -1,899 +1,899 @@ /* * Copyright 2020-2024 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 // xmlNode #include #include #include #include #include #include #include #include #include #include #include #include //! Object to store node info from the controller API typedef struct { /* Adapted from pcmk_controld_api_reply_t:data:node_info. * (char **) are convenient here for use within callbacks: we can skip * copying strings unless the caller passes a non-NULL value. */ uint32_t id; char **node_name; char **uuid; char **state; bool have_quorum; bool is_remote; } node_info_t; //! Object to store API results, a timeout, and an output object typedef struct { pcmk__output_t *out; bool show_output; int rc; unsigned int message_timeout_ms; enum pcmk_pacemakerd_state pcmkd_state; node_info_t node_info; } data_t; /*! * \internal * \brief Validate that an IPC API event is a good reply * * \param[in,out] data API results and options * \param[in] api IPC API connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * * \return Standard Pacemaker return code */ static int validate_reply_event(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status) { pcmk__output_t *out = data->out; switch (event_type) { case pcmk_ipc_event_reply: break; case pcmk_ipc_event_disconnect: if (data->rc == ECONNRESET) { // Unexpected out->err(out, "error: Lost connection to %s", pcmk_ipc_name(api, true)); } // Nothing bad but not the reply we're looking for return ENOTSUP; default: // Ditto return ENOTSUP; } if (status != CRM_EX_OK) { out->err(out, "error: Bad reply from %s: %s", pcmk_ipc_name(api, true), crm_exit_str(status)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a controller API event is a good reply of expected type * * \param[in,out] data API results and options * \param[in] api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_controld_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_controld_api_reply expected_type) { pcmk__output_t *out = data->out; int rc = pcmk_rc_ok; const pcmk_controld_api_reply_t *reply = NULL; rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from controller", pcmk__controld_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Validate that a \p pacemakerd API event is a good reply of expected * type * * \param[in,out] data API results and options * \param[in] api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in] event_data Event-specific data * \param[in] expected_type Expected reply type * * \return Standard Pacemaker return code */ static int validate_pcmkd_reply(data_t *data, const pcmk_ipc_api_t *api, enum pcmk_ipc_event event_type, crm_exit_t status, const void *event_data, enum pcmk_pacemakerd_api_reply expected_type) { pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_reply_event(data, api, event_type, status); if (rc != pcmk_rc_ok) { return rc; } reply = (const pcmk_pacemakerd_api_reply_t *) event_data; if (reply->reply_type != expected_type) { out->err(out, "error: Unexpected reply type '%s' from pacemakerd", pcmk__pcmkd_api_reply2str(reply->reply_type)); data->rc = EBADMSG; return data->rc; } return pcmk_rc_ok; } /*! * \internal * \brief Process a controller status IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void controller_status_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "health", reply->data.ping.sys_from, reply->host_from, reply->data.ping.fsa_state, reply->data.ping.result); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a designated controller IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void designated_controller_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_ping); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; out->message(out, "dc", reply->host_from); data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a node info IPC event * * \param[in,out] controld_api Controller connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_controld_api_reply_t object containing * event-specific data * \param[in,out] user_data \p data_t object for API results and options */ static void node_info_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { data_t *data = (data_t *) user_data; pcmk__output_t *out = data->out; const pcmk_controld_api_reply_t *reply = NULL; int rc = validate_controld_reply(data, controld_api, event_type, status, event_data, pcmk_controld_reply_info); if (rc != pcmk_rc_ok) { return; } reply = (const pcmk_controld_api_reply_t *) event_data; if (reply->data.node_info.uname == NULL) { out->err(out, "Node is not known to cluster"); data->rc = pcmk_rc_node_unknown; return; } data->node_info.have_quorum = reply->data.node_info.have_quorum; data->node_info.is_remote = reply->data.node_info.is_remote; data->node_info.id = (uint32_t) reply->data.node_info.id; pcmk__str_update(data->node_info.node_name, reply->data.node_info.uname); pcmk__str_update(data->node_info.uuid, reply->data.node_info.uuid); pcmk__str_update(data->node_info.state, reply->data.node_info.state); if (data->show_output) { out->message(out, "node-info", reply->data.node_info.id, reply->data.node_info.uname, reply->data.node_info.uuid, reply->data.node_info.state, reply->data.node_info.have_quorum, reply->data.node_info.is_remote); } data->rc = pcmk_rc_ok; } /*! * \internal * \brief Process a \p pacemakerd status IPC event * * \param[in,out] pacemakerd_api \p pacemakerd connection * \param[in] event_type Type of event that occurred * \param[in] status Event status * \param[in,out] event_data \p pcmk_pacemakerd_api_reply_t object * containing event-specific data * \param[in,out] user_data \p data_t object for API results and options */ 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) { data_t *data = user_data; pcmk__output_t *out = data->out; const pcmk_pacemakerd_api_reply_t *reply = NULL; int rc = validate_pcmkd_reply(data, pacemakerd_api, event_type, status, event_data, pcmk_pacemakerd_reply_ping); if (rc != pcmk_rc_ok) { return; } // Parse desired information from reply reply = (const pcmk_pacemakerd_api_reply_t *) event_data; data->pcmkd_state = reply->data.ping.state; data->rc = pcmk_rc_ok; if (!data->show_output) { return; } if (reply->data.ping.status == pcmk_rc_ok) { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, NULL, reply->data.ping.last_good); } else { out->message(out, "pacemakerd-health", reply->data.ping.sys_from, reply->data.ping.state, "query failed", time(NULL)); } } static pcmk_ipc_api_t * ipc_connect(data_t *data, enum pcmk_ipc_server server, pcmk_ipc_callback_t cb, enum pcmk_ipc_dispatch dispatch_type, bool eremoteio_ok) { int rc; pcmk__output_t *out = data->out; pcmk_ipc_api_t *api = NULL; rc = pcmk_new_ipc_api(&api, server); if (api == NULL) { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); data->rc = rc; return NULL; } if (cb != NULL) { pcmk_register_ipc_callback(api, cb, data); } rc = pcmk__connect_ipc(api, dispatch_type, 5); if (rc != pcmk_rc_ok) { if (rc == EREMOTEIO) { data->pcmkd_state = pcmk_pacemakerd_state_remote; if (eremoteio_ok) { /* EREMOTEIO may be expected and acceptable for some callers * on a Pacemaker Remote node */ crm_debug("Ignoring %s connection failure: No " "Pacemaker Remote connection", pcmk_ipc_name(api, true)); rc = pcmk_rc_ok; } else { out->err(out, "error: Could not connect to %s: %s", pcmk_ipc_name(api, true), pcmk_rc_str(rc)); } } data->rc = rc; pcmk_free_ipc_api(api); return NULL; } return api; } /*! * \internal * \brief Poll an IPC API connection until timeout or a reply is received * * \param[in,out] data API results and options * \param[in,out] api IPC API connection * \param[in] on_node If not \p NULL, name of the node to poll (used only * for logging) * * \note Sets the \p rc member of \p data on error */ static void poll_until_reply(data_t *data, pcmk_ipc_api_t *api, const char *on_node) { pcmk__output_t *out = data->out; uint64_t start_nsec = qb_util_nano_current_get(); uint64_t end_nsec = 0; uint64_t elapsed_ms = 0; uint64_t remaining_ms = data->message_timeout_ms; while (remaining_ms > 0) { int rc = pcmk_poll_ipc(api, remaining_ms); if (rc == EAGAIN) { // Poll timed out break; } if (rc != pcmk_rc_ok) { out->err(out, "error: Failed to poll %s API%s%s: %s", pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, ""), pcmk_rc_str(rc)); data->rc = rc; return; } pcmk_dispatch_ipc(api); if (data->rc != EAGAIN) { // Received a reply return; } end_nsec = qb_util_nano_current_get(); elapsed_ms = (end_nsec - start_nsec) / QB_TIME_NS_IN_MSEC; remaining_ms = data->message_timeout_ms - elapsed_ms; } out->err(out, "error: Timed out after %ums waiting for reply from %s API%s%s", data->message_timeout_ms, pcmk_ipc_name(api, true), (on_node != NULL)? " on " : "", pcmk__s(on_node, "")); data->rc = EAGAIN; } /*! * \internal * \brief Get and output controller status * * \param[in,out] out Output object * \param[in] node_name Name of node whose status is desired * (\p NULL for DC) * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__controller_status(pcmk__output_t *out, const char *node_name, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, controller_status_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, node_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on %s: %s", pcmk__s(node_name, "DC"), pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, pcmk__s(node_name, "DC")); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_controller_status(xmlNodePtr *xml, const char *node_name, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__controller_status(out, node_name, message_timeout_ms); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /*! * \internal * \brief Get and output designated controller node name * * \param[in,out] out Output object * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code */ int pcmk__designated_controller(pcmk__output_t *out, unsigned int message_timeout_ms) { data_t data = { .out = out, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, designated_controller_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_ping(controld_api, NULL); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping controller API on DC: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "DC"); } pcmk_free_ipc_api(controld_api); } return data.rc; } // Documented in header int pcmk_designated_controller(xmlNodePtr *xml, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__designated_controller(out, message_timeout_ms); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /*! * \internal * \brief Get and optionally output node info corresponding to a node ID from * the controller * * \param[in,out] out Output object * \param[in,out] node_id ID of node whose name to get. If \p NULL * or 0, get the local node name. If not * \p NULL, store the true node ID here on * success. * \param[out] node_name If not \p NULL, where to store the node * name * \param[out] uuid If not \p NULL, where to store the node * UUID * \param[out] state If not \p NULL, where to store the * membership state * \param[out] is_remote If not \p NULL, where to store whether the * node is a Pacemaker Remote node * \param[out] have_quorum If not \p NULL, where to store whether the * node has quorum * \param[in] show_output Whether to show the node info * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemaker-controld API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * * \return Standard Pacemaker return code * * \note The caller is responsible for freeing \p *node_name, \p *uuid, and * \p *state using \p free(). */ int pcmk__query_node_info(pcmk__output_t *out, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .node_info = { .id = (node_id == NULL)? 0 : *node_id, .node_name = node_name, .uuid = uuid, .state = state, }, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *controld_api = NULL; if (node_name != NULL) { *node_name = NULL; } if (uuid != NULL) { *uuid = NULL; } if (state != NULL) { *state = NULL; } if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } controld_api = ipc_connect(&data, pcmk_ipc_controld, node_info_event_cb, dispatch_type, false); if (controld_api != NULL) { int rc = pcmk_controld_api_node_info(controld_api, (node_id != NULL)? *node_id : 0); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not send request to controller API on local " "node: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, controld_api, "local node"); } pcmk_free_ipc_api(controld_api); } if (data.rc != pcmk_rc_ok) { return data.rc; } // String outputs are set in callback if (node_id != NULL) { *node_id = data.node_info.id; } if (have_quorum != NULL) { *have_quorum = data.node_info.have_quorum; } if (is_remote != NULL) { *is_remote = data.node_info.is_remote; } return data.rc; } // Documented in header int pcmk_query_node_info(xmlNodePtr *xml, uint32_t *node_id, char **node_name, char **uuid, char **state, bool *have_quorum, bool *is_remote, bool show_output, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; CRM_ASSERT(node_name != NULL); rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__query_node_info(out, node_id, node_name, uuid, state, have_quorum, is_remote, show_output, message_timeout_ms); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /*! * \internal * \brief Get and optionally output \p pacemakerd status * * \param[in,out] out Output object * \param[in] ipc_name IPC name for request * \param[in] message_timeout_ms How long to wait for a reply from the * \p pacemakerd API. If 0, * \p pcmk_ipc_dispatch_sync will be used. * Otherwise, \p pcmk_ipc_dispatch_poll will * be used. * \param[in] show_output Whether to output the \p pacemakerd state * \param[out] state Where to store the \p pacemakerd state, if * not \p NULL * * \return Standard Pacemaker return code * * \note This function sets \p state to \p pcmk_pacemakerd_state_remote and * returns \p pcmk_rc_ok if the IPC connection attempt returns * \p EREMOTEIO. That code indicates that this is a Pacemaker Remote node * with \p pacemaker-remoted running. The node may be connected to the * cluster. */ int pcmk__pacemakerd_status(pcmk__output_t *out, const char *ipc_name, unsigned int message_timeout_ms, bool show_output, enum pcmk_pacemakerd_state *state) { data_t data = { .out = out, .show_output = show_output, .rc = EAGAIN, .message_timeout_ms = message_timeout_ms, .pcmkd_state = pcmk_pacemakerd_state_invalid, }; enum pcmk_ipc_dispatch dispatch_type = pcmk_ipc_dispatch_poll; pcmk_ipc_api_t *pacemakerd_api = NULL; if (message_timeout_ms == 0) { dispatch_type = pcmk_ipc_dispatch_sync; } pacemakerd_api = ipc_connect(&data, pcmk_ipc_pacemakerd, pacemakerd_event_cb, dispatch_type, true); if (pacemakerd_api != NULL) { int rc = pcmk_pacemakerd_api_ping(pacemakerd_api, ipc_name); if (rc != pcmk_rc_ok) { out->err(out, "error: Could not ping launcher API: %s", pcmk_rc_str(rc)); data.rc = rc; } if (dispatch_type == pcmk_ipc_dispatch_poll) { poll_until_reply(&data, pacemakerd_api, NULL); } pcmk_free_ipc_api(pacemakerd_api); } else if ((data.pcmkd_state == pcmk_pacemakerd_state_remote) && show_output) { // No API connection so the callback wasn't run out->message(out, "pacemakerd-health", NULL, data.pcmkd_state, NULL, time(NULL)); } if (state != NULL) { *state = data.pcmkd_state; } return data.rc; } // Documented in header int pcmk_pacemakerd_status(xmlNodePtr *xml, const char *ipc_name, unsigned int message_timeout_ms) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__pacemakerd_status(out, ipc_name, message_timeout_ms, true, NULL); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /* user data for looping through remote node xpath searches */ struct node_data { pcmk__output_t *out; int found; const char *field; /* XML attribute to check for node name */ const char *type; bool bash_export; }; static void remote_node_print_helper(xmlNode *result, void *user_data) { struct node_data *data = user_data; pcmk__output_t *out = data->out; - const char *name = crm_element_value(result, XML_ATTR_UNAME); + const char *name = crm_element_value(result, PCMK_XA_UNAME); const char *id = crm_element_value(result, data->field); // node name and node id are the same for remote/guest nodes out->message(out, "crmadmin-node", data->type, pcmk__s(name, id), id, data->bash_export); data->found++; } // \return Standard Pacemaker return code int pcmk__list_nodes(pcmk__output_t *out, const char *node_types, bool bash_export) { xmlNode *xml_node = NULL; int rc; rc = cib__signon_query(out, NULL, &xml_node); if (rc == pcmk_rc_ok) { struct node_data data = { .out = out, .found = 0, .bash_export = bash_export }; out->begin_list(out, NULL, NULL, "nodes"); if (!pcmk__str_empty(node_types) && strstr(node_types, "all")) { node_types = NULL; } if (pcmk__str_empty(node_types) || strstr(node_types, "cluster")) { data.field = PCMK_XA_ID; data.type = "cluster"; crm_foreach_xpath_result(xml_node, PCMK__XP_MEMBER_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || strstr(node_types, "guest")) { data.field = PCMK_XA_VALUE; data.type = "guest"; crm_foreach_xpath_result(xml_node, PCMK__XP_GUEST_NODE_CONFIG, remote_node_print_helper, &data); } if (pcmk__str_empty(node_types) || pcmk__str_eq(node_types, ",|^remote", pcmk__str_regex)) { data.field = PCMK_XA_ID; data.type = "remote"; crm_foreach_xpath_result(xml_node, PCMK__XP_REMOTE_NODE_CONFIG, remote_node_print_helper, &data); } out->end_list(out); if (data.found == 0) { out->info(out, "No nodes configured"); } free_xml(xml_node); } return rc; } int pcmk_list_nodes(xmlNodePtr *xml, const char *node_types) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__list_nodes(out, node_types, FALSE); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } diff --git a/lib/pacemaker/pcmk_injections.c b/lib/pacemaker/pcmk_injections.c index 994a038674..560e9d2a2c 100644 --- a/lib/pacemaker/pcmk_injections.c +++ b/lib/pacemaker/pcmk_injections.c @@ -1,788 +1,788 @@ /* * Copyright 2009-2024 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 // lrmd_event_data_t, etc. #include #include #include #include "libpacemaker_private.h" bool pcmk__simulate_node_config = false; -#define XPATH_NODE_CONFIG "//" XML_CIB_TAG_NODE "[@" XML_ATTR_UNAME "='%s']" -#define XPATH_NODE_STATE "//" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" +#define XPATH_NODE_CONFIG "//" XML_CIB_TAG_NODE "[@" PCMK_XA_UNAME "='%s']" +#define XPATH_NODE_STATE "//" XML_CIB_TAG_STATE "[@" PCMK_XA_UNAME "='%s']" #define XPATH_NODE_STATE_BY_ID "//" XML_CIB_TAG_STATE "[@" PCMK_XA_ID "='%s']" #define XPATH_RSC_HISTORY XPATH_NODE_STATE \ "//" XML_LRM_TAG_RESOURCE "[@" PCMK_XA_ID "='%s']" /*! * \internal * \brief Inject a fictitious transient node attribute into scheduler input * * \param[in,out] out Output object for displaying error messages * \param[in,out] cib_node node_state XML to inject attribute into * \param[in] name Transient node attribute name to inject * \param[in] value Transient node attribute value to inject */ static void inject_transient_attr(pcmk__output_t *out, xmlNode *cib_node, const char *name, const char *value) { xmlNode *attrs = NULL; xmlNode *instance_attrs = NULL; const char *node_uuid = ID(cib_node); out->message(out, "inject-attr", name, value, cib_node); attrs = first_named_child(cib_node, XML_TAG_TRANSIENT_NODEATTRS); if (attrs == NULL) { attrs = create_xml_node(cib_node, XML_TAG_TRANSIENT_NODEATTRS); crm_xml_add(attrs, PCMK_XA_ID, node_uuid); } instance_attrs = first_named_child(attrs, XML_TAG_ATTR_SETS); if (instance_attrs == NULL) { instance_attrs = create_xml_node(attrs, XML_TAG_ATTR_SETS); crm_xml_add(instance_attrs, PCMK_XA_ID, node_uuid); } crm_create_nvpair_xml(instance_attrs, NULL, name, value); } /*! * \internal * \brief Inject a fictitious fail count into a scheduler input * * \param[in,out] out Output object for displaying error messages * \param[in,out] cib_node Node state XML to inject into * \param[in] resource ID of resource for fail count to inject * \param[in] task Action name for fail count to inject * \param[in] interval_ms Action interval (in milliseconds) for fail count * \param[in] rc Action result for fail count to inject (if 0, or * 7 when interval_ms is 0, inject nothing) */ void pcmk__inject_failcount(pcmk__output_t *out, xmlNode *cib_node, const char *resource, const char *task, guint interval_ms, int rc) { if (rc == 0) { return; } else if ((rc == 7) && (interval_ms == 0)) { return; } else { char *name = NULL; char *now = pcmk__ttoa(time(NULL)); name = pcmk__failcount_name(resource, task, interval_ms); inject_transient_attr(out, cib_node, name, "value++"); free(name); name = pcmk__lastfailure_name(resource, task, interval_ms); inject_transient_attr(out, cib_node, name, now); free(name); free(now); } } /*! * \internal * \brief Create a CIB configuration entry for a fictitious node * * \param[in,out] cib_conn CIB object to use * \param[in] node Node name to use */ static void create_node_entry(cib_t *cib_conn, const char *node) { int rc = pcmk_ok; char *xpath = crm_strdup_printf(XPATH_NODE_CONFIG, node); rc = cib_conn->cmds->query(cib_conn, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); if (rc == -ENXIO) { // Only add if not already existing xmlNode *cib_object = create_xml_node(NULL, XML_CIB_TAG_NODE); crm_xml_add(cib_object, PCMK_XA_ID, node); // Use node name as ID - crm_xml_add(cib_object, XML_ATTR_UNAME, node); + crm_xml_add(cib_object, PCMK_XA_UNAME, node); cib_conn->cmds->create(cib_conn, XML_CIB_TAG_NODES, cib_object, cib_sync_call|cib_scope_local); /* Not bothering with subsequent query to see if it exists, we'll bomb out later in the call to query_node_uuid()... */ free_xml(cib_object); } free(xpath); } /*! * \internal * \brief Synthesize a fake executor event for an action * * \param[in] cib_resource XML for any existing resource action history * \param[in] task Name of action to synthesize * \param[in] interval_ms Interval of action to synthesize * \param[in] outcome Result of action to synthesize * * \return Newly allocated executor event * \note It is the caller's responsibility to free the result with * lrmd_free_event(). */ static lrmd_event_data_t * create_op(const xmlNode *cib_resource, const char *task, guint interval_ms, int outcome) { lrmd_event_data_t *op = NULL; xmlNode *xop = NULL; op = lrmd_new_event(ID(cib_resource), task, interval_ms); lrmd__set_result(op, outcome, PCMK_EXEC_DONE, "Simulated action result"); op->params = NULL; // Not needed for simulation purposes op->t_run = (unsigned int) time(NULL); op->t_rcchange = op->t_run; // Use a call ID higher than any existing history entries op->call_id = 0; for (xop = pcmk__xe_first_child(cib_resource); xop != NULL; xop = pcmk__xe_next(xop)) { int tmp = 0; crm_element_value_int(xop, XML_LRM_ATTR_CALLID, &tmp); if (tmp > op->call_id) { op->call_id = tmp; } } op->call_id++; return op; } /*! * \internal * \brief Inject a fictitious resource history entry into a scheduler input * * \param[in,out] cib_resource Resource history XML to inject entry into * \param[in,out] op Action result to inject * \param[in] target_rc Expected result for action to inject * * \return XML of injected resource history entry */ xmlNode * pcmk__inject_action_result(xmlNode *cib_resource, lrmd_event_data_t *op, int target_rc) { return pcmk__create_history_xml(cib_resource, op, CRM_FEATURE_SET, target_rc, NULL, crm_system_name); } /*! * \internal * \brief Inject a fictitious node into a scheduler input * * \param[in,out] cib_conn Scheduler input CIB to inject node into * \param[in] node Name of node to inject * \param[in] uuid UUID of node to inject * * \return XML of node_state entry for new node * \note If the global pcmk__simulate_node_config has been set to true, a * node entry in the configuration section will be added, as well as a * node state entry in the status section. */ xmlNode * pcmk__inject_node(cib_t *cib_conn, const char *node, const char *uuid) { int rc = pcmk_ok; xmlNode *cib_object = NULL; char *xpath = crm_strdup_printf(XPATH_NODE_STATE, node); bool duplicate = false; char *found_uuid = NULL; if (pcmk__simulate_node_config) { create_node_entry(cib_conn, node); } rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object, cib_xpath|cib_sync_call|cib_scope_local); if ((cib_object != NULL) && (ID(cib_object) == NULL)) { crm_err("Detected multiple node_state entries for xpath=%s, bailing", xpath); duplicate = true; goto done; } if (rc == -ENXIO) { if (uuid == NULL) { query_node_uuid(cib_conn, node, &found_uuid, NULL); } else { found_uuid = strdup(uuid); } if (found_uuid) { char *xpath_by_uuid = crm_strdup_printf(XPATH_NODE_STATE_BY_ID, found_uuid); // It's possible that a node_state entry doesn't have an uname yet. rc = cib_conn->cmds->query(cib_conn, xpath_by_uuid, &cib_object, cib_xpath|cib_sync_call|cib_scope_local); if ((cib_object != NULL) && (ID(cib_object) == NULL)) { crm_err("Can't inject node state for %s because multiple " "state entries found for ID %s", node, found_uuid); duplicate = true; free(xpath_by_uuid); goto done; } else if (cib_object != NULL) { - crm_xml_add(cib_object, XML_ATTR_UNAME, node); + crm_xml_add(cib_object, PCMK_XA_UNAME, node); rc = cib_conn->cmds->modify(cib_conn, XML_CIB_TAG_STATUS, cib_object, cib_sync_call|cib_scope_local); } free(xpath_by_uuid); } } if (rc == -ENXIO) { cib_object = create_xml_node(NULL, XML_CIB_TAG_STATE); crm_xml_add(cib_object, PCMK_XA_ID, found_uuid); - crm_xml_add(cib_object, XML_ATTR_UNAME, node); + crm_xml_add(cib_object, PCMK_XA_UNAME, node); cib_conn->cmds->create(cib_conn, XML_CIB_TAG_STATUS, cib_object, cib_sync_call|cib_scope_local); free_xml(cib_object); rc = cib_conn->cmds->query(cib_conn, xpath, &cib_object, cib_xpath|cib_sync_call|cib_scope_local); crm_trace("Injecting node state for %s (rc=%d)", node, rc); } done: free(found_uuid); free(xpath); if (duplicate) { crm_log_xml_warn(cib_object, "Duplicates"); crm_exit(CRM_EX_SOFTWARE); return NULL; // not reached, but makes static analysis happy } CRM_ASSERT(rc == pcmk_ok); return cib_object; } /*! * \internal * \brief Inject a fictitious node state change into a scheduler input * * \param[in,out] cib_conn Scheduler input CIB to inject into * \param[in] node Name of node to inject change for * \param[in] up If true, change state to online, otherwise offline * * \return XML of changed (or added) node state entry */ xmlNode * pcmk__inject_node_state_change(cib_t *cib_conn, const char *node, bool up) { xmlNode *cib_node = pcmk__inject_node(cib_conn, node, NULL); if (up) { pcmk__xe_set_props(cib_node, PCMK__XA_IN_CCM, XML_BOOLEAN_TRUE, PCMK__XA_CRMD, ONLINESTATUS, PCMK__XA_JOIN, CRMD_JOINSTATE_MEMBER, PCMK__XA_EXPECTED, CRMD_JOINSTATE_MEMBER, NULL); } else { pcmk__xe_set_props(cib_node, PCMK__XA_IN_CCM, XML_BOOLEAN_FALSE, PCMK__XA_CRMD, OFFLINESTATUS, PCMK__XA_JOIN, CRMD_JOINSTATE_DOWN, PCMK__XA_EXPECTED, CRMD_JOINSTATE_DOWN, NULL); } crm_xml_add(cib_node, PCMK_XA_CRM_DEBUG_ORIGIN, crm_system_name); return cib_node; } /*! * \internal * \brief Check whether a node has history for a given resource * * \param[in,out] cib_node Node state XML to check * \param[in] resource Resource name to check for * * \return Resource's lrm_resource XML entry beneath \p cib_node if found, * otherwise NULL */ static xmlNode * find_resource_xml(xmlNode *cib_node, const char *resource) { - const char *node = crm_element_value(cib_node, XML_ATTR_UNAME); + const char *node = crm_element_value(cib_node, PCMK_XA_UNAME); char *xpath = crm_strdup_printf(XPATH_RSC_HISTORY, node, resource); xmlNode *match = get_xpath_object(xpath, cib_node, LOG_TRACE); free(xpath); return match; } /*! * \internal * \brief Inject a resource history element into a scheduler input * * \param[in,out] out Output object for displaying error messages * \param[in,out] cib_node Node state XML to inject resource history entry into * \param[in] resource ID (in configuration) of resource to inject * \param[in] lrm_name ID as used in history (could be clone instance) * \param[in] rclass Resource agent class of resource to inject * \param[in] rtype Resource agent type of resource to inject * \param[in] rprovider Resource agent provider of resource to inject * * \return XML of injected resource history element * \note If a history element already exists under either \p resource or * \p lrm_name, this will return it rather than injecting a new one. */ xmlNode * pcmk__inject_resource_history(pcmk__output_t *out, xmlNode *cib_node, const char *resource, const char *lrm_name, const char *rclass, const char *rtype, const char *rprovider) { xmlNode *lrm = NULL; xmlNode *container = NULL; xmlNode *cib_resource = NULL; cib_resource = find_resource_xml(cib_node, resource); if (cib_resource != NULL) { /* If an existing LRM history entry uses the resource name, * continue using it, even if lrm_name is different. */ return cib_resource; } // Check for history entry under preferred name if (strcmp(resource, lrm_name) != 0) { cib_resource = find_resource_xml(cib_node, lrm_name); if (cib_resource != NULL) { return cib_resource; } } if ((rclass == NULL) || (rtype == NULL)) { // @TODO query configuration for class, provider, type out->err(out, "Resource %s not found in the status section of %s " "(supply class and type to continue)", resource, ID(cib_node)); return NULL; } else if (!pcmk__strcase_any_of(rclass, PCMK_RESOURCE_CLASS_OCF, PCMK_RESOURCE_CLASS_STONITH, PCMK_RESOURCE_CLASS_SERVICE, PCMK_RESOURCE_CLASS_UPSTART, PCMK_RESOURCE_CLASS_SYSTEMD, PCMK_RESOURCE_CLASS_LSB, NULL)) { out->err(out, "Invalid class for %s: %s", resource, rclass); return NULL; } else if (pcmk_is_set(pcmk_get_ra_caps(rclass), pcmk_ra_cap_provider) && (rprovider == NULL)) { // @TODO query configuration for provider out->err(out, "Please specify the provider for resource %s", resource); return NULL; } crm_info("Injecting new resource %s into node state '%s'", lrm_name, ID(cib_node)); lrm = first_named_child(cib_node, XML_CIB_TAG_LRM); if (lrm == NULL) { const char *node_uuid = ID(cib_node); lrm = create_xml_node(cib_node, XML_CIB_TAG_LRM); crm_xml_add(lrm, PCMK_XA_ID, node_uuid); } container = first_named_child(lrm, XML_LRM_TAG_RESOURCES); if (container == NULL) { container = create_xml_node(lrm, XML_LRM_TAG_RESOURCES); } cib_resource = create_xml_node(container, XML_LRM_TAG_RESOURCE); // If we're creating a new entry, use the preferred name crm_xml_add(cib_resource, PCMK_XA_ID, lrm_name); crm_xml_add(cib_resource, PCMK_XA_CLASS, rclass); crm_xml_add(cib_resource, PCMK_XA_PROVIDER, rprovider); crm_xml_add(cib_resource, PCMK_XA_TYPE, rtype); return cib_resource; } static int find_ticket_state(pcmk__output_t *out, cib_t *the_cib, const char *ticket_id, xmlNode **ticket_state_xml) { int rc = pcmk_ok; xmlNode *xml_search = NULL; GString *xpath = g_string_sized_new(256); CRM_ASSERT(ticket_state_xml != NULL); *ticket_state_xml = NULL; g_string_append(xpath, "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS "/" XML_CIB_TAG_TICKETS); if (ticket_id) { pcmk__g_strcat(xpath, "/" XML_CIB_TAG_TICKET_STATE "[@" PCMK_XA_ID "=\"", ticket_id, "\"]", NULL); } rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search, cib_sync_call|cib_scope_local|cib_xpath); g_string_free(xpath, TRUE); if (rc != pcmk_ok) { return rc; } crm_log_xml_debug(xml_search, "Match"); if ((xml_search->children != NULL) && (ticket_id != NULL)) { out->err(out, "Multiple ticket_states match ticket_id=%s", ticket_id); } *ticket_state_xml = xml_search; return rc; } /*! * \internal * \brief Inject a ticket attribute into ticket state * * \param[in,out] out Output object for displaying error messages * \param[in] ticket_id Ticket whose state should be changed * \param[in] attr_name Ticket attribute name to inject * \param[in] attr_value Boolean value of ticket attribute to inject * \param[in,out] cib CIB object to use * * \return Standard Pacemaker return code */ static int set_ticket_state_attr(pcmk__output_t *out, const char *ticket_id, const char *attr_name, bool attr_value, cib_t *cib) { int rc = pcmk_rc_ok; xmlNode *xml_top = NULL; xmlNode *ticket_state_xml = NULL; // Check for an existing ticket state entry rc = find_ticket_state(out, cib, ticket_id, &ticket_state_xml); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { // Ticket state found, use it crm_debug("Injecting attribute into existing ticket state %s", ticket_id); xml_top = ticket_state_xml; } else if (rc == ENXIO) { // No ticket state, create it xmlNode *xml_obj = NULL; xml_top = create_xml_node(NULL, XML_CIB_TAG_STATUS); xml_obj = create_xml_node(xml_top, XML_CIB_TAG_TICKETS); ticket_state_xml = create_xml_node(xml_obj, XML_CIB_TAG_TICKET_STATE); crm_xml_add(ticket_state_xml, PCMK_XA_ID, ticket_id); } else { // Error return rc; } // Add the attribute to the ticket state pcmk__xe_set_bool_attr(ticket_state_xml, attr_name, attr_value); crm_log_xml_debug(xml_top, "Update"); // Commit the change to the CIB rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, xml_top, cib_sync_call|cib_scope_local); rc = pcmk_legacy2rc(rc); free_xml(xml_top); return rc; } /*! * \internal * \brief Inject a fictitious action into the cluster * * \param[in,out] out Output object for displaying error messages * \param[in] spec Action specification to inject * \param[in,out] cib CIB object for scheduler input * \param[in] scheduler Scheduler data */ static void inject_action(pcmk__output_t *out, const char *spec, cib_t *cib, const pcmk_scheduler_t *scheduler) { int rc; int outcome = PCMK_OCF_OK; guint interval_ms = 0; char *key = NULL; char *node = NULL; char *task = NULL; char *resource = NULL; const char *rtype = NULL; const char *rclass = NULL; const char *rprovider = NULL; xmlNode *cib_op = NULL; xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; const pcmk_resource_t *rsc = NULL; lrmd_event_data_t *op = NULL; out->message(out, "inject-spec", spec); key = calloc(1, strlen(spec) + 1); node = calloc(1, strlen(spec) + 1); rc = sscanf(spec, "%[^@]@%[^=]=%d", key, node, &outcome); if (rc != 3) { out->err(out, "Invalid operation spec: %s. Only found %d fields", spec, rc); goto done; } parse_op_key(key, &resource, &task, &interval_ms); rsc = pe_find_resource(scheduler->resources, resource); if (rsc == NULL) { out->err(out, "Invalid resource name: %s", resource); goto done; } rclass = crm_element_value(rsc->xml, PCMK_XA_CLASS); rtype = crm_element_value(rsc->xml, PCMK_XA_TYPE); rprovider = crm_element_value(rsc->xml, PCMK_XA_PROVIDER); cib_node = pcmk__inject_node(cib, node, NULL); CRM_ASSERT(cib_node != NULL); pcmk__inject_failcount(out, cib_node, resource, task, interval_ms, outcome); cib_resource = pcmk__inject_resource_history(out, cib_node, resource, resource, rclass, rtype, rprovider); CRM_ASSERT(cib_resource != NULL); op = create_op(cib_resource, task, interval_ms, outcome); CRM_ASSERT(op != NULL); cib_op = pcmk__inject_action_result(cib_resource, op, 0); CRM_ASSERT(cib_op != NULL); lrmd_free_event(op); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); done: free(task); free(node); free(key); } /*! * \internal * \brief Inject fictitious scheduler inputs * * \param[in,out] scheduler Scheduler data * \param[in,out] cib CIB object for scheduler input to modify * \param[in] injections Injections to apply */ void pcmk__inject_scheduler_input(pcmk_scheduler_t *scheduler, cib_t *cib, const pcmk_injections_t *injections) { int rc = pcmk_ok; const GList *iter = NULL; xmlNode *cib_node = NULL; pcmk__output_t *out = scheduler->priv; out->message(out, "inject-modify-config", injections->quorum, injections->watchdog); if (injections->quorum != NULL) { xmlNode *top = create_xml_node(NULL, XML_TAG_CIB); /* crm_xml_add(top, PCMK_XA_DC_UUID, dc_uuid); */ crm_xml_add(top, PCMK_XA_HAVE_QUORUM, injections->quorum); rc = cib->cmds->modify(cib, NULL, top, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); } if (injections->watchdog != NULL) { rc = cib__update_node_attr(out, cib, cib_sync_call|cib_scope_local, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, PCMK_OPT_HAVE_WATCHDOG, injections->watchdog, NULL, NULL); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->node_up; iter != NULL; iter = iter->next) { const char *node = (const char *) iter->data; out->message(out, "inject-modify-node", "Online", node); cib_node = pcmk__inject_node_state_change(cib, node, true); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); } for (iter = injections->node_down; iter != NULL; iter = iter->next) { const char *node = (const char *) iter->data; char *xpath = NULL; out->message(out, "inject-modify-node", "Offline", node); cib_node = pcmk__inject_node_state_change(cib, node, false); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); xpath = crm_strdup_printf("//node_state[@uname='%s']/%s", node, XML_CIB_TAG_LRM); cib->cmds->remove(cib, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); free(xpath); xpath = crm_strdup_printf("//node_state[@uname='%s']/%s", node, XML_TAG_TRANSIENT_NODEATTRS); cib->cmds->remove(cib, xpath, NULL, cib_xpath|cib_sync_call|cib_scope_local); free(xpath); } for (iter = injections->node_fail; iter != NULL; iter = iter->next) { const char *node = (const char *) iter->data; out->message(out, "inject-modify-node", "Failing", node); cib_node = pcmk__inject_node_state_change(cib, node, true); crm_xml_add(cib_node, PCMK__XA_IN_CCM, XML_BOOLEAN_FALSE); CRM_ASSERT(cib_node != NULL); rc = cib->cmds->modify(cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); free_xml(cib_node); } for (iter = injections->ticket_grant; iter != NULL; iter = iter->next) { const char *ticket_id = (const char *) iter->data; out->message(out, "inject-modify-ticket", "Granting", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "granted", true, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_revoke; iter != NULL; iter = iter->next) { const char *ticket_id = (const char *) iter->data; out->message(out, "inject-modify-ticket", "Revoking", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "granted", false, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_standby; iter != NULL; iter = iter->next) { const char *ticket_id = (const char *) iter->data; out->message(out, "inject-modify-ticket", "Standby", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "standby", true, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->ticket_activate; iter != NULL; iter = iter->next) { const char *ticket_id = (const char *) iter->data; out->message(out, "inject-modify-ticket", "Activating", ticket_id); rc = set_ticket_state_attr(out, ticket_id, "standby", false, cib); CRM_ASSERT(rc == pcmk_rc_ok); } for (iter = injections->op_inject; iter != NULL; iter = iter->next) { inject_action(out, (const char *) iter->data, cib, scheduler); } if (!out->is_quiet(out)) { out->end_list(out); } } void pcmk_free_injections(pcmk_injections_t *injections) { if (injections == NULL) { return; } g_list_free_full(injections->node_up, g_free); g_list_free_full(injections->node_down, g_free); g_list_free_full(injections->node_fail, g_free); g_list_free_full(injections->op_fail, g_free); g_list_free_full(injections->op_inject, g_free); g_list_free_full(injections->ticket_grant, g_free); g_list_free_full(injections->ticket_revoke, g_free); g_list_free_full(injections->ticket_standby, g_free); g_list_free_full(injections->ticket_activate, g_free); free(injections->quorum); free(injections->watchdog); free(injections); } diff --git a/lib/pacemaker/pcmk_output.c b/lib/pacemaker/pcmk_output.c index 9e5bcc137c..2173063464 100644 --- a/lib/pacemaker/pcmk_output.c +++ b/lib/pacemaker/pcmk_output.c @@ -1,2405 +1,2405 @@ /* * Copyright 2019-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include static char * colocations_header(pcmk_resource_t *rsc, pcmk__colocation_t *cons, bool dependents) { char *retval = NULL; if (cons->primary_role > pcmk_role_started) { retval = crm_strdup_printf("%s (score=%s, %s role=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), (dependents? "needs" : "with"), role2text(cons->primary_role), cons->id); } else { retval = crm_strdup_printf("%s (score=%s, id=%s)", rsc->id, pcmk_readable_score(cons->score), cons->id); } return retval; } static void colocations_xml_node(pcmk__output_t *out, pcmk_resource_t *rsc, pcmk__colocation_t *cons) { xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_DEPEND, PCMK_XA_ID, cons->id, "rsc", cons->dependent->id, "with-rsc", cons->primary->id, "score", pcmk_readable_score(cons->score), NULL); if (cons->node_attribute) { xmlSetProp(node, (pcmkXmlStr) "node-attribute", (pcmkXmlStr) cons->node_attribute); } if (cons->dependent_role != pcmk_role_unknown) { xmlSetProp(node, (pcmkXmlStr) "rsc-role", (pcmkXmlStr) role2text(cons->dependent_role)); } if (cons->primary_role != pcmk_role_unknown) { xmlSetProp(node, (pcmkXmlStr) "with-rsc-role", (pcmkXmlStr) role2text(cons->primary_role)); } } static int do_locations_list_xml(pcmk__output_t *out, pcmk_resource_t *rsc, bool add_header) { GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pcmk__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->nodes; lpc2 != NULL; lpc2 = lpc2->next) { pcmk_node_t *node = (pcmk_node_t *) lpc2->data; if (add_header) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "locations"); } pcmk__output_create_xml_node(out, XML_CONS_TAG_RSC_LOCATION, "node", node->details->uname, "rsc", rsc->id, PCMK_XA_ID, cons->id, "score", pcmk_readable_score(node->weight), NULL); } } if (add_header) { PCMK__OUTPUT_LIST_FOOTER(out, rc); } return rc; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pcmk_resource_t *", "pcmk_node_t *", "pcmk_node_t *", "pcmk_action_t *", "pcmk_action_t *") static int rsc_action_item(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *origin = va_arg(args, pcmk_node_t *); pcmk_node_t *destination = va_arg(args, pcmk_node_t *); pcmk_action_t *action = va_arg(args, pcmk_action_t *); pcmk_action_t *source = va_arg(args, pcmk_action_t *); int len = 0; char *reason = NULL; char *details = NULL; bool same_host = false; bool same_role = false; bool need_role = false; static int rsc_width = 5; static int detail_width = 5; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if (source == NULL) { source = action; } len = strlen(rsc->id); if (len > rsc_width) { rsc_width = len + 2; } if ((rsc->role > pcmk_role_started) || (rsc->next_role > pcmk_role_unpromoted)) { need_role = true; } if (pe__same_node(origin, destination)) { same_host = true; } if (rsc->role == rsc->next_role) { same_role = true; } if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(destination)); } else if (origin == NULL) { /* Starting a resource */ details = crm_strdup_printf("%s", pe__node_name(destination)); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (destination == NULL) { /* Stopping a resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ details = crm_strdup_printf("%s %s", role2text(rsc->role), pe__node_name(origin)); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ details = crm_strdup_printf("%s", pe__node_name(origin)); } else if (need_role && same_role) { /* Moving a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", pe__node_name(origin), pe__node_name(destination), role2text(rsc->role)); } else if (same_role) { /* Moving a normal resource */ details = crm_strdup_printf("%s -> %s", pe__node_name(origin), pe__node_name(destination)); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), pe__node_name(origin)); } else { /* Moving and promoting/demoting */ details = crm_strdup_printf("%s %s -> %s %s", role2text(rsc->role), pe__node_name(origin), role2text(rsc->next_role), pe__node_name(destination)); } len = strlen(details); if (len > detail_width) { detail_width = len; } if ((source->reason != NULL) && !pcmk_is_set(action->flags, pcmk_action_runnable)) { reason = crm_strdup_printf("due to %s (blocked)", source->reason); } else if (source->reason) { reason = crm_strdup_printf("due to %s", source->reason); } else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) { reason = strdup("blocked"); } out->list_item(out, NULL, "%-8s %-*s ( %*s )%s%s", change, rsc_width, rsc->id, detail_width, details, ((reason == NULL)? "" : " "), pcmk__s(reason, "")); free(details); free(reason); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-action-item", "const char *", "pcmk_resource_t *", "pcmk_node_t *", "pcmk_node_t *", "pcmk_action_t *", "pcmk_action_t *") static int rsc_action_item_xml(pcmk__output_t *out, va_list args) { const char *change = va_arg(args, const char *); pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *origin = va_arg(args, pcmk_node_t *); pcmk_node_t *destination = va_arg(args, pcmk_node_t *); pcmk_action_t *action = va_arg(args, pcmk_action_t *); pcmk_action_t *source = va_arg(args, pcmk_action_t *); char *change_str = NULL; bool same_host = false; bool same_role = false; bool need_role = false; xmlNode *xml = NULL; CRM_ASSERT(action); CRM_ASSERT(destination != NULL || origin != NULL); if (source == NULL) { source = action; } if ((rsc->role > pcmk_role_started) || (rsc->next_role > pcmk_role_unpromoted)) { need_role = true; } if (pe__same_node(origin, destination)) { same_host = true; } if (rsc->role == rsc->next_role) { same_role = true; } change_str = g_ascii_strdown(change, -1); xml = pcmk__output_create_xml_node(out, "rsc_action", "action", change_str, "resource", rsc->id, NULL); g_free(change_str); if (need_role && (origin == NULL)) { /* Starting and promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } else if (origin == NULL) { /* Starting a resource */ crm_xml_add(xml, "node", destination->details->uname); } else if (need_role && (destination == NULL)) { /* Stopping a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "node", origin->details->uname, NULL); } else if (destination == NULL) { /* Stopping a resource */ crm_xml_add(xml, "node", origin->details->uname); } else if (need_role && same_role && same_host) { /* Recovering, restarting or re-promoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, NULL); } else if (same_role && same_host) { /* Recovering or Restarting a normal resource */ crm_xml_add(xml, "source", origin->details->uname); } else if (need_role && same_role) { /* Moving a promotable clone instance */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, "role", role2text(rsc->role), NULL); } else if (same_role) { /* Moving a normal resource */ pcmk__xe_set_props(xml, "source", origin->details->uname, "dest", destination->details->uname, NULL); } else if (same_host) { /* Promoting or demoting a promotable clone instance */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "next-role", role2text(rsc->next_role), "source", origin->details->uname, NULL); } else { /* Moving and promoting/demoting */ pcmk__xe_set_props(xml, "role", role2text(rsc->role), "source", origin->details->uname, "next-role", role2text(rsc->next_role), "dest", destination->details->uname, NULL); } if ((source->reason != NULL) && !pcmk_is_set(action->flags, pcmk_action_runnable)) { pcmk__xe_set_props(xml, "reason", source->reason, "blocked", "true", NULL); } else if (source->reason != NULL) { crm_xml_add(xml, "reason", source->reason); } else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) { pcmk__xe_set_bool_attr(xml, "blocked", true); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pcmk_resource_t *", "bool") static int rsc_is_colocated_with_list(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons * directly rather than rsc->cmds->this_with_colocations(). */ pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources %s is colocated with", rsc->id); if (pcmk_is_set(cons->primary->flags, pcmk_rsc_detect_loop)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->primary->id, cons->id); continue; } hdr = colocations_header(cons->primary, cons, false); out->list_item(out, NULL, "%s", hdr); free(hdr); // Empty list header for indentation of information about this resource out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->primary); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rsc-is-colocated-with-list", "pcmk_resource_t *", "bool") static int rsc_is_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use rsc->rsc_cons * directly rather than rsc->cmds->this_with_colocations(). */ pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->primary->flags, pcmk_rsc_detect_loop)) { colocations_xml_node(out, cons->primary, cons); continue; } colocations_xml_node(out, cons->primary, cons); do_locations_list_xml(out, cons->primary, false); if (recursive) { out->message(out, "rsc-is-colocated-with-list", cons->primary, recursive); } } return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pcmk_resource_t *", "bool") static int rscs_colocated_with_list(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use * rsc->rsc_cons_lhs directly rather than * rsc->cmds->with_this_colocations(). */ pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; char *hdr = NULL; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources colocated with %s", rsc->id); if (pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop)) { out->list_item(out, NULL, "%s (id=%s - loop)", cons->dependent->id, cons->id); continue; } hdr = colocations_header(cons->dependent, cons, true); out->list_item(out, NULL, "%s", hdr); free(hdr); // Empty list header for indentation of information about this resource out->begin_list(out, NULL, NULL, NULL); out->message(out, "locations-list", cons->dependent); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("rscs-colocated-with-list", "pcmk_resource_t *", "bool") static int rscs_colocated_with_list_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); int rc = pcmk_rc_no_output; if (pcmk_is_set(rsc->flags, pcmk_rsc_detect_loop)) { return rc; } /* We're listing constraints explicitly involving rsc, so use * rsc->rsc_cons_lhs directly rather than * rsc->cmds->with_this_colocations(). */ pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop); for (GList *lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; if (pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop)) { colocations_xml_node(out, cons->dependent, cons); continue; } colocations_xml_node(out, cons->dependent, cons); do_locations_list_xml(out, cons->dependent, false); if (recursive) { out->message(out, "rscs-colocated-with-list", cons->dependent, recursive); } } return rc; } PCMK__OUTPUT_ARGS("locations-list", "pcmk_resource_t *") static int locations_list(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); GList *lpc = NULL; GList *list = rsc->rsc_location; int rc = pcmk_rc_no_output; for (lpc = list; lpc != NULL; lpc = lpc->next) { pcmk__location_t *cons = lpc->data; GList *lpc2 = NULL; for (lpc2 = cons->nodes; lpc2 != NULL; lpc2 = lpc2->next) { pcmk_node_t *node = (pcmk_node_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Locations"); out->list_item(out, NULL, "Node %s (score=%s, id=%s, rsc=%s)", pe__node_name(node), pcmk_readable_score(node->weight), cons->id, rsc->id); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("locations-list", "pcmk_resource_t *") static int locations_list_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); return do_locations_list_xml(out, rsc, true); } PCMK__OUTPUT_ARGS("locations-and-colocations", "pcmk_resource_t *", "bool", "bool") static int locations_and_colocations(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(rsc->cluster); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } out->message(out, "locations-list", rsc); pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("locations-and-colocations", "pcmk_resource_t *", "bool", "bool") static int locations_and_colocations_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); bool recursive = va_arg(args, int); bool force = va_arg(args, int); pcmk__unpack_constraints(rsc->cluster); // Constraints apply to group/clone, not member/instance if (!force) { rsc = uber_parent(rsc); } pcmk__output_xml_create_parent(out, "constraints", NULL); do_locations_list_xml(out, rsc, false); pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop); out->message(out, "rscs-colocated-with-list", rsc, recursive); pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop); out->message(out, "rsc-is-colocated-with-list", rsc, recursive); pcmk__output_xml_pop_parent(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health(pcmk__output_t *out, va_list args) { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); return out->info(out, "Controller on %s in state %s: %s", pcmk__s(host_from, "unknown node"), pcmk__s(fsa_state, "unknown"), pcmk__s(result, "unknown result")); } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); const char *host_from G_GNUC_UNUSED = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result G_GNUC_UNUSED = va_arg(args, const char *); if (fsa_state != NULL) { pcmk__formatted_printf(out, "%s\n", fsa_state); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("health", "const char *", "const char *", "const char *", "const char *") static int health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); const char *host_from = va_arg(args, const char *); const char *fsa_state = va_arg(args, const char *); const char *result = va_arg(args, const char *); pcmk__output_create_xml_node(out, pcmk__s(sys_from, ""), "node_name", pcmk__s(host_from, ""), "state", pcmk__s(fsa_state, ""), "result", pcmk__s(result, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; int rc = pcmk_rc_ok; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } rc = out->info(out, "Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); free(last_updated_s); return rc; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_html(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; char *msg = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk__pcmkd_state_enum2friendly(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } msg = crm_strdup_printf("Status of %s: '%s' (last updated %s)", sys_from, state_s, pcmk__s(last_updated_s, "at unknown time")); pcmk__output_create_html_node(out, "li", NULL, NULL, msg); free(msg); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return pacemakerd_health(out, args); } else { const char *sys_from G_GNUC_UNUSED = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated G_GNUC_UNUSED = va_arg(args, time_t); if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } pcmk__formatted_printf(out, "%s\n", state_s); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("pacemakerd-health", "const char *", "enum pcmk_pacemakerd_state", "const char *", "time_t") static int pacemakerd_health_xml(pcmk__output_t *out, va_list args) { const char *sys_from = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = va_arg(args, const char *); time_t last_updated = va_arg(args, time_t); char *last_updated_s = NULL; if (sys_from == NULL) { if (state == pcmk_pacemakerd_state_remote) { sys_from = "pacemaker-remoted"; } else { sys_from = CRM_SYSTEM_MCP; } } if (state_s == NULL) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } if (last_updated != 0) { last_updated_s = pcmk__epoch2str(&last_updated, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); } pcmk__output_create_xml_node(out, "pacemakerd", "sys_from", sys_from, "state", state_s, "last_updated", last_updated_s, NULL); free(last_updated_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_default(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); out->list_item(out, NULL, "Testing %s ... %.2f secs", xml_file, (end - start) / (float) CLOCKS_PER_SEC); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("profile", "const char *", "clock_t", "clock_t") static int profile_xml(pcmk__output_t *out, va_list args) { const char *xml_file = va_arg(args, const char *); clock_t start = va_arg(args, clock_t); clock_t end = va_arg(args, clock_t); char *duration = pcmk__ftoa((end - start) / (float) CLOCKS_PER_SEC); pcmk__output_create_xml_node(out, "timing", "file", xml_file, "duration", duration, NULL); free(duration); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); return out->info(out, "Designated Controller is: %s", pcmk__s(dc, "not yet elected")); } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return dc(out, args); } else { const char *dc = va_arg(args, const char *); if (dc != NULL) { pcmk__formatted_printf(out, "%s\n", pcmk__s(dc, "")); return pcmk_rc_ok; } } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("dc", "const char *") static int dc_xml(pcmk__output_t *out, va_list args) { const char *dc = va_arg(args, const char *); pcmk__output_create_xml_node(out, "dc", "node_name", pcmk__s(dc, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export = va_arg(args, int); if (bash_export) { return out->info(out, "export %s=%s", pcmk__s(name, ""), pcmk__s(id, "")); } else { return out->info(out, "%s node: %s (%s)", type ? type : "cluster", pcmk__s(name, ""), pcmk__s(id, "")); } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_text(pcmk__output_t *out, va_list args) { if (!out->is_quiet(out)) { return crmadmin_node(out, args); } else { const char *type G_GNUC_UNUSED = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id G_GNUC_UNUSED = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__formatted_printf(out, "%s\n", pcmk__s(name, "")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("crmadmin-node", "const char *", "const char *", "const char *", "bool") static int crmadmin_node_xml(pcmk__output_t *out, va_list args) { const char *type = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *id = va_arg(args, const char *); bool bash_export G_GNUC_UNUSED = va_arg(args, int); pcmk__output_create_xml_node(out, "node", PCMK_XA_TYPE, pcmk__s(type, "cluster"), PCMK_XA_NAME, pcmk__s(name, ""), PCMK_XA_ID, pcmk__s(id, ""), NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("digests", "const pcmk_resource_t *", "const pcmk_node_t *", "const char *", "guint", "const pcmk__op_digest_t *") static int digests_text(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const pcmk_node_t *node = va_arg(args, const pcmk_node_t *); const char *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); const pcmk__op_digest_t *digests = va_arg(args, const pcmk__op_digest_t *); char *action_desc = NULL; const char *rsc_desc = "unknown resource"; const char *node_desc = "unknown node"; if (interval_ms != 0) { action_desc = crm_strdup_printf("%ums-interval %s action", interval_ms, ((task == NULL)? "unknown" : task)); } else if (pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_none)) { action_desc = strdup("probe action"); } else { action_desc = crm_strdup_printf("%s action", ((task == NULL)? "unknown" : task)); } if ((rsc != NULL) && (rsc->id != NULL)) { rsc_desc = rsc->id; } if ((node != NULL) && (node->details->uname != NULL)) { node_desc = node->details->uname; } out->begin_list(out, NULL, NULL, "Digests for %s %s on %s", rsc_desc, action_desc, node_desc); free(action_desc); if (digests == NULL) { out->list_item(out, NULL, "none"); out->end_list(out); return pcmk_rc_ok; } if (digests->digest_all_calc != NULL) { out->list_item(out, NULL, "%s (all parameters)", digests->digest_all_calc); } if (digests->digest_secure_calc != NULL) { out->list_item(out, NULL, "%s (non-private parameters)", digests->digest_secure_calc); } if (digests->digest_restart_calc != NULL) { out->list_item(out, NULL, "%s (non-reloadable parameters)", digests->digest_restart_calc); } out->end_list(out); return pcmk_rc_ok; } static void add_digest_xml(xmlNode *parent, const char *type, const char *digest, xmlNode *digest_source) { if (digest != NULL) { xmlNodePtr digest_xml = create_xml_node(parent, "digest"); crm_xml_add(digest_xml, PCMK_XA_TYPE, pcmk__s(type, "unspecified")); crm_xml_add(digest_xml, "hash", digest); if (digest_source != NULL) { add_node_copy(digest_xml, digest_source); } } } PCMK__OUTPUT_ARGS("digests", "const pcmk_resource_t *", "const pcmk_node_t *", "const char *", "guint", "const pcmk__op_digest_t *") static int digests_xml(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const pcmk_node_t *node = va_arg(args, const pcmk_node_t *); const char *task = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); const pcmk__op_digest_t *digests = va_arg(args, const pcmk__op_digest_t *); char *interval_s = crm_strdup_printf("%ums", interval_ms); xmlNode *xml = NULL; xml = pcmk__output_create_xml_node(out, "digests", "resource", pcmk__s(rsc->id, ""), "node", pcmk__s(node->details->uname, ""), "task", pcmk__s(task, ""), "interval", interval_s, NULL); free(interval_s); if (digests != NULL) { add_digest_xml(xml, "all", digests->digest_all_calc, digests->params_all); add_digest_xml(xml, "nonprivate", digests->digest_secure_calc, digests->params_secure); add_digest_xml(xml, "nonreloadable", digests->digest_restart_calc, digests->params_restart); } return pcmk_rc_ok; } #define STOP_SANITY_ASSERT(lineno) do { \ if ((current != NULL) && current->details->unclean) { \ /* It will be a pseudo op */ \ } else if (stop == NULL) { \ crm_err("%s:%d: No stop action exists for %s", \ __func__, lineno, rsc->id); \ CRM_ASSERT(stop != NULL); \ } else if (pcmk_is_set(stop->flags, pcmk_action_optional)) { \ crm_err("%s:%d: Action %s is still optional", \ __func__, lineno, stop->uuid); \ CRM_ASSERT(!pcmk_is_set(stop->flags, pcmk_action_optional));\ } \ } while (0) PCMK__OUTPUT_ARGS("rsc-action", "pcmk_resource_t *", "pcmk_node_t *", "pcmk_node_t *") static int rsc_action_default(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *current = va_arg(args, pcmk_node_t *); pcmk_node_t *next = va_arg(args, pcmk_node_t *); GList *possible_matches = NULL; char *key = NULL; int rc = pcmk_rc_no_output; bool moving = false; pcmk_node_t *start_node = NULL; pcmk_action_t *start = NULL; pcmk_action_t *stop = NULL; pcmk_action_t *promote = NULL; pcmk_action_t *demote = NULL; pcmk_action_t *reason_op = NULL; if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed) || (current == NULL && next == NULL)) { const bool managed = pcmk_is_set(rsc->flags, pcmk_rsc_managed); pcmk__rsc_info(rsc, "Leave %s\t(%s%s)", rsc->id, role2text(rsc->role), (managed? "" : " unmanaged")); return rc; } moving = (current != NULL) && (next != NULL) && !pe__same_node(current, next); possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_START, false); if (possible_matches) { start = possible_matches->data; g_list_free(possible_matches); } if ((start == NULL) || !pcmk_is_set(start->flags, pcmk_action_runnable)) { start_node = NULL; } else { start_node = current; } possible_matches = pe__resource_actions(rsc, start_node, PCMK_ACTION_STOP, false); if (possible_matches) { stop = possible_matches->data; g_list_free(possible_matches); } else if (pcmk_is_set(rsc->flags, pcmk_rsc_stop_unexpected)) { /* The resource is multiply active with PCMK_META_MULTIPLE_ACTIVE set to * stop_unexpected, and not stopping on its current node, but it should * be stopping elsewhere. */ possible_matches = pe__resource_actions(rsc, NULL, PCMK_ACTION_STOP, false); if (possible_matches != NULL) { stop = possible_matches->data; g_list_free(possible_matches); } } possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_PROMOTE, false); if (possible_matches) { promote = possible_matches->data; g_list_free(possible_matches); } possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_DEMOTE, false); if (possible_matches) { demote = possible_matches->data; g_list_free(possible_matches); } if (rsc->role == rsc->next_role) { pcmk_action_t *migrate_op = NULL; CRM_CHECK(next != NULL, return rc); possible_matches = pe__resource_actions(rsc, next, PCMK_ACTION_MIGRATE_FROM, false); if (possible_matches) { migrate_op = possible_matches->data; } if ((migrate_op != NULL) && (current != NULL) && pcmk_is_set(migrate_op->flags, pcmk_action_runnable)) { rc = out->message(out, "rsc-action-item", "Migrate", rsc, current, next, start, NULL); } else if (pcmk_is_set(rsc->flags, pcmk_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if ((start == NULL) || pcmk_is_set(start->flags, pcmk_action_optional)) { if ((demote != NULL) && (promote != NULL) && !pcmk_is_set(demote->flags, pcmk_action_optional) && !pcmk_is_set(promote->flags, pcmk_action_optional)) { rc = out->message(out, "rsc-action-item", "Re-promote", rsc, current, next, promote, demote); } else { pcmk__rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id, role2text(rsc->role), pe__node_name(next)); } } else if (!pcmk_is_set(start->flags, pcmk_action_runnable)) { if ((stop == NULL) || (stop->reason == NULL)) { reason_op = start; } else { reason_op = stop; } rc = out->message(out, "rsc-action-item", "Stop", rsc, current, NULL, stop, reason_op); STOP_SANITY_ASSERT(__LINE__); } else if (moving && current) { const bool failed = pcmk_is_set(rsc->flags, pcmk_rsc_failed); rc = out->message(out, "rsc-action-item", (failed? "Recover" : "Move"), rsc, current, next, stop, NULL); } else if (pcmk_is_set(rsc->flags, pcmk_rsc_failed)) { rc = out->message(out, "rsc-action-item", "Recover", rsc, current, NULL, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); #if 0 /* @TODO This can be reached in situations that should really be * "Start" (see for example the migrate-fail-7 regression test) */ STOP_SANITY_ASSERT(__LINE__); #endif } g_list_free(possible_matches); return rc; } if ((stop != NULL) && ((rsc->next_role == pcmk_role_stopped) || ((start != NULL) && !pcmk_is_set(start->flags, pcmk_action_runnable)))) { key = stop_key(rsc); for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) { pcmk_node_t *node = iter->data; pcmk_action_t *stop_op = NULL; reason_op = start; possible_matches = find_actions(rsc->actions, key, node); if (possible_matches) { stop_op = possible_matches->data; g_list_free(possible_matches); } if (stop_op != NULL) { if (pcmk_is_set(stop_op->flags, pcmk_action_runnable)) { STOP_SANITY_ASSERT(__LINE__); } if (stop_op->reason != NULL) { reason_op = stop_op; } } if (out->message(out, "rsc-action-item", "Stop", rsc, node, NULL, stop_op, reason_op) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } free(key); } else if ((stop != NULL) && pcmk_all_flags_set(rsc->flags, pcmk_rsc_failed|pcmk_rsc_stop_if_failed)) { /* 'stop' may be NULL if the failure was ignored */ rc = out->message(out, "rsc-action-item", "Recover", rsc, current, next, stop, start); STOP_SANITY_ASSERT(__LINE__); } else if (moving) { rc = out->message(out, "rsc-action-item", "Move", rsc, current, next, stop, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (pcmk_is_set(rsc->flags, pcmk_rsc_reload)) { rc = out->message(out, "rsc-action-item", "Reload", rsc, current, next, start, NULL); } else if ((stop != NULL) && !pcmk_is_set(stop->flags, pcmk_action_optional)) { rc = out->message(out, "rsc-action-item", "Restart", rsc, current, next, start, NULL); STOP_SANITY_ASSERT(__LINE__); } else if (rsc->role == pcmk_role_promoted) { CRM_LOG_ASSERT(current != NULL); rc = out->message(out, "rsc-action-item", "Demote", rsc, current, next, demote, NULL); } else if (rsc->next_role == pcmk_role_promoted) { CRM_LOG_ASSERT(next); rc = out->message(out, "rsc-action-item", "Promote", rsc, current, next, promote, NULL); } else if ((rsc->role == pcmk_role_stopped) && (rsc->next_role > pcmk_role_stopped)) { rc = out->message(out, "rsc-action-item", "Start", rsc, current, next, start, NULL); } return rc; } PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *") static int node_action(pcmk__output_t *out, va_list args) { const char *task = va_arg(args, const char *); const char *node_name = va_arg(args, const char *); const char *reason = va_arg(args, const char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { out->list_item(out, NULL, "%s %s '%s'", task, node_name, reason); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-action", "const char *", "const char *", "const char *") static int node_action_xml(pcmk__output_t *out, va_list args) { const char *task = va_arg(args, const char *); const char *node_name = va_arg(args, const char *); const char *reason = va_arg(args, const char *); if (task == NULL) { return pcmk_rc_no_output; } else if (reason) { pcmk__output_create_xml_node(out, "node_action", "task", task, "node", node_name, "reason", reason, NULL); } else { crm_notice(" * %s %s", task, node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "const char *", "const char *", "bool", "bool") static int node_info_default(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); return out->info(out, "Node %" PRIu32 ": %s " "(uuid=%s, state=%s, have_quorum=%s, is_remote=%s)", node_id, pcmk__s(node_name, "unknown"), pcmk__s(uuid, "unknown"), pcmk__s(state, "unknown"), pcmk__btoa(have_quorum), pcmk__btoa(is_remote)); } PCMK__OUTPUT_ARGS("node-info", "uint32_t", "const char *", "const char *", "const char *", "bool", "bool") static int node_info_xml(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); const char *uuid = va_arg(args, const char *); const char *state = va_arg(args, const char *); bool have_quorum = (bool) va_arg(args, int); bool is_remote = (bool) va_arg(args, int); char *id_s = crm_strdup_printf("%" PRIu32, node_id); pcmk__output_create_xml_node(out, "node-info", "nodeid", id_s, - XML_ATTR_UNAME, node_name, + PCMK_XA_UNAME, node_name, PCMK_XA_ID, uuid, PCMK__XA_CRMD, state, PCMK_XA_HAVE_QUORUM, pcmk__btoa(have_quorum), XML_NODE_IS_REMOTE, pcmk__btoa(is_remote), NULL); free(id_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (rsc != NULL) { out->list_item(out, NULL, "Cluster action: %s for %s on %s", task, ID(rsc), node); } else { out->list_item(out, NULL, "Cluster action: %s on %s", task, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-cluster-action", "const char *", "const char *", "xmlNodePtr") static int inject_cluster_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr rsc = va_arg(args, xmlNodePtr); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "cluster_action", "task", task, "node", node, NULL); if (rsc) { crm_xml_add(xml_node, PCMK_XA_ID, ID(rsc)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *") static int inject_fencing_action(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Fencing %s (%s)", target, op); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-fencing-action", "const char *", "const char *") static int inject_fencing_action_xml(pcmk__output_t *out, va_list args) { const char *target = va_arg(args, const char *); const char *op = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "fencing_action", "target", target, PCMK_XA_OP, op, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); out->list_item(out, NULL, "Injecting attribute %s=%s into %s '%s'", name, value, node_path, ID(cib_node)); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-attr", "const char *", "const char *", "xmlNodePtr") static int inject_attr_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); xmlNodePtr cib_node = va_arg(args, xmlNodePtr); xmlChar *node_path = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node_path = xmlGetNodePath(cib_node); pcmk__output_create_xml_node(out, "inject_attr", PCMK_XA_NAME, name, PCMK_XA_VALUE, value, "node_path", node_path, "cib_node", ID(cib_node), NULL); free(node_path); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Injecting %s into the configuration", spec); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-spec", "const char *") static int inject_spec_xml(pcmk__output_t *out, va_list args) { const char *spec = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "inject_spec", "spec", spec, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *") static int inject_modify_config(pcmk__output_t *out, va_list args) { const char *quorum = va_arg(args, const char *); const char *watchdog = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->begin_list(out, NULL, NULL, "Performing Requested Modifications"); if (quorum) { out->list_item(out, NULL, "Setting quorum: %s", quorum); } if (watchdog) { out->list_item(out, NULL, "Setting watchdog: %s", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-config", "const char *", "const char *") static int inject_modify_config_xml(pcmk__output_t *out, va_list args) { const char *quorum = va_arg(args, const char *); const char *watchdog = va_arg(args, const char *); xmlNodePtr node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } node = pcmk__output_xml_create_parent(out, "modifications", NULL); if (quorum) { crm_xml_add(node, "quorum", quorum); } if (watchdog) { crm_xml_add(node, "watchdog", watchdog); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *") static int inject_modify_node(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *node = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Online", pcmk__str_none)) { out->list_item(out, NULL, "Bringing node %s online", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Offline", pcmk__str_none)) { out->list_item(out, NULL, "Taking node %s offline", node); return pcmk_rc_ok; } else if (pcmk__str_eq(action, "Failing", pcmk__str_none)) { out->list_item(out, NULL, "Failing node %s", node); return pcmk_rc_ok; } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("inject-modify-node", "const char *", "const char *") static int inject_modify_node_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *node = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_node", "action", action, "node", node, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *") static int inject_modify_ticket(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *ticket = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (pcmk__str_eq(action, "Standby", pcmk__str_none)) { out->list_item(out, NULL, "Making ticket %s standby", ticket); } else { out->list_item(out, NULL, "%s ticket %s", action, ticket); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-modify-ticket", "const char *", "const char *") static int inject_modify_ticket_xml(pcmk__output_t *out, va_list args) { const char *action = va_arg(args, const char *); const char *ticket = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } pcmk__output_create_xml_node(out, "modify_ticket", "action", action, "ticket", ticket, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); if (out->is_quiet(out)) { return pcmk_rc_no_output; } out->list_item(out, NULL, "Pseudo action: %s%s%s", task, ((node == NULL)? "" : " on "), pcmk__s(node, "")); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-pseudo-action", "const char *", "const char *") static int inject_pseudo_action_xml(pcmk__output_t *out, va_list args) { const char *node = va_arg(args, const char *); const char *task = va_arg(args, const char *); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "pseudo_action", "task", task, NULL); if (node) { crm_xml_add(xml_node, "node", node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "const char *", "guint") static int inject_rsc_action(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); const char *node = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); if (out->is_quiet(out)) { return pcmk_rc_no_output; } if (interval_ms) { out->list_item(out, NULL, "Resource action: %-15s %s=%u on %s", rsc, operation, interval_ms, node); } else { out->list_item(out, NULL, "Resource action: %-15s %s on %s", rsc, operation, node); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("inject-rsc-action", "const char *", "const char *", "const char *", "guint") static int inject_rsc_action_xml(pcmk__output_t *out, va_list args) { const char *rsc = va_arg(args, const char *); const char *operation = va_arg(args, const char *); const char *node = va_arg(args, const char *); guint interval_ms = va_arg(args, guint); xmlNodePtr xml_node = NULL; if (out->is_quiet(out)) { return pcmk_rc_no_output; } xml_node = pcmk__output_create_xml_node(out, "rsc_action", "resource", rsc, PCMK_XA_OP, operation, "node", node, NULL); if (interval_ms) { char *interval_s = pcmk__itoa(interval_ms); crm_xml_add(xml_node, "interval", interval_s); free(interval_s); } return pcmk_rc_ok; } #define CHECK_RC(retcode, retval) \ if (retval == pcmk_rc_ok) { \ retcode = pcmk_rc_ok; \ } PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") int pcmk__cluster_status_text(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); int rc = pcmk_rc_no_output; bool already_printed_failure = false; CHECK_RC(rc, out->message(out, "cluster-summary", scheduler, pcmkd_state, section_opts, show_opts)); if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { CHECK_RC(rc, out->message(out, "node-list", scheduler->nodes, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { CHECK_RC(rc, out->message(out, "resource-list", scheduler, show_opts, true, unames, resources, rc == pcmk_rc_ok)); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { CHECK_RC(rc, out->message(out, "node-attribute-list", scheduler, show_opts, (rc == pcmk_rc_ok), unames, resources)); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations|pcmk_section_failcounts)) { CHECK_RC(rc, out->message(out, "node-summary", scheduler, unames, resources, section_opts, show_opts, (rc == pcmk_rc_ok))); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && (scheduler->failed != NULL) && (scheduler->failed->children != NULL)) { CHECK_RC(rc, out->message(out, "failed-action-list", scheduler, unames, resources, show_opts, rc == pcmk_rc_ok)); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); already_printed_failure = true; } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { CHECK_RC(rc, out->message(out, "ticket-list", scheduler, (rc == pcmk_rc_ok))); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { CHECK_RC(rc, out->message(out, "ban-list", scheduler, prefix, resources, show_opts, rc == pcmk_rc_ok)); } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { PCMK__OUTPUT_SPACER_IF(out, rc == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { CHECK_RC(rc, out->message(out, "fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { CHECK_RC(rc, out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, rc == pcmk_rc_ok)); } } } return rc; } PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_xml(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); out->message(out, "cluster-summary", scheduler, pcmkd_state, section_opts, show_opts); /*** NODES ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes)) { out->message(out, "node-list", scheduler->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { /* XML output always displays full details. */ uint32_t full_show_opts = show_opts & ~pcmk_show_brief; out->message(out, "resource-list", scheduler, full_show_opts, false, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", scheduler, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations|pcmk_section_failcounts)) { out->message(out, "node-summary", scheduler, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && (scheduler->failed != NULL) && (scheduler->failed->children != NULL)) { out->message(out, "failed-action-list", scheduler, unames, resources, show_opts, false); } /* Print stonith history */ if (pcmk_is_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { out->message(out, "full-fencing-list", history_rc, stonith_history, unames, section_opts, show_opts, false); } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", scheduler, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", scheduler, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-status", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "crm_exit_t", "stonith_history_t *", "enum pcmk__fence_history", "uint32_t", "uint32_t", "const char *", "GList *", "GList *") static int cluster_status_html(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); crm_exit_t history_rc = va_arg(args, crm_exit_t); stonith_history_t *stonith_history = va_arg(args, stonith_history_t *); enum pcmk__fence_history fence_history = va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); const char *prefix = va_arg(args, const char *); GList *unames = va_arg(args, GList *); GList *resources = va_arg(args, GList *); bool already_printed_failure = false; out->message(out, "cluster-summary", scheduler, pcmkd_state, section_opts, show_opts); /*** NODE LIST ***/ if (pcmk_is_set(section_opts, pcmk_section_nodes) && unames) { out->message(out, "node-list", scheduler->nodes, unames, resources, show_opts, false); } /* Print resources section, if needed */ if (pcmk_is_set(section_opts, pcmk_section_resources)) { out->message(out, "resource-list", scheduler, show_opts, true, unames, resources, false); } /* print Node Attributes section if requested */ if (pcmk_is_set(section_opts, pcmk_section_attributes)) { out->message(out, "node-attribute-list", scheduler, show_opts, false, unames, resources); } /* If requested, print resource operations (which includes failcounts) * or just failcounts */ if (pcmk_any_flags_set(section_opts, pcmk_section_operations|pcmk_section_failcounts)) { out->message(out, "node-summary", scheduler, unames, resources, section_opts, show_opts, false); } /* If there were any failed actions, print them */ if (pcmk_is_set(section_opts, pcmk_section_failures) && (scheduler->failed != NULL) && (scheduler->failed->children != NULL)) { out->message(out, "failed-action-list", scheduler, unames, resources, show_opts, false); } /* Print failed stonith actions */ if (pcmk_is_set(section_opts, pcmk_section_fence_failed) && fence_history != pcmk__fence_history_none) { if (history_rc == 0) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_eq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "failed-fencing-list", stonith_history, unames, section_opts, show_opts, false); } } else { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } /* Print stonith history */ if (pcmk_any_flags_set(section_opts, pcmk_section_fencing_all) && fence_history != pcmk__fence_history_none) { if (history_rc != 0) { if (!already_printed_failure) { out->begin_list(out, NULL, NULL, "Failed Fencing Actions"); out->list_item(out, NULL, "Failed to get fencing history: %s", crm_exit_str(history_rc)); out->end_list(out); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_worked)) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_neq, GINT_TO_POINTER(st_failed)); if (hp) { out->message(out, "fencing-list", hp, unames, section_opts, show_opts, false); } } else if (pcmk_is_set(section_opts, pcmk_section_fence_pending)) { stonith_history_t *hp = NULL; hp = stonith__first_matching_event(stonith_history, stonith__event_state_pending, NULL); if (hp) { out->message(out, "pending-fencing-list", hp, unames, section_opts, show_opts, false); } } } /* Print tickets if requested */ if (pcmk_is_set(section_opts, pcmk_section_tickets)) { out->message(out, "ticket-list", scheduler, false); } /* Print negative location constraints if requested */ if (pcmk_is_set(section_opts, pcmk_section_bans)) { out->message(out, "ban-list", scheduler, prefix, resources, show_opts, false); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_default(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); GString *s = g_string_sized_new(50); if (!pcmk__str_empty(scope)) { pcmk__g_strcat(s, "scope=\"", scope, "\" ", NULL); } if (!pcmk__str_empty(instance)) { pcmk__g_strcat(s, "id=\"", instance, "\" ", NULL); } pcmk__g_strcat(s, "name=\"", pcmk__s(name, ""), "\" ", NULL); if (!pcmk__str_empty(host)) { pcmk__g_strcat(s, "host=\"", host, "\" ", NULL); } pcmk__g_strcat(s, "value=\"", pcmk__s(value, ""), "\"", NULL); out->info(out, "%s", s->str); g_string_free(s, TRUE); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("attribute", "const char *", "const char *", "const char *", "const char *", "const char *") static int attribute_xml(pcmk__output_t *out, va_list args) { const char *scope = va_arg(args, const char *); const char *instance = va_arg(args, const char *); const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); const char *host = va_arg(args, const char *); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, "attribute", PCMK_XA_NAME, name, PCMK_XA_VALUE, pcmk__s(value, ""), NULL); if (!pcmk__str_empty(scope)) { crm_xml_add(node, "scope", scope); } if (!pcmk__str_empty(instance)) { crm_xml_add(node, PCMK_XA_ID, instance); } if (!pcmk__str_empty(host)) { crm_xml_add(node, "host", host); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_default(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); switch (result) { case pcmk_rc_within_range: return out->info(out, "Rule %s is still in effect", rule_id); case pcmk_rc_ok: return out->info(out, "Rule %s satisfies conditions", rule_id); case pcmk_rc_after_range: return out->info(out, "Rule %s is expired", rule_id); case pcmk_rc_before_range: return out->info(out, "Rule %s has not yet taken effect", rule_id); case pcmk_rc_op_unsatisfied: return out->info(out, "Rule %s does not satisfy conditions", rule_id); default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("rule-check", "const char *", "int", "const char *") static int rule_check_xml(pcmk__output_t *out, va_list args) { const char *rule_id = va_arg(args, const char *); int result = va_arg(args, int); const char *error = va_arg(args, const char *); char *rc_str = pcmk__itoa(pcmk_rc2exitc(result)); pcmk__output_create_xml_node(out, "rule-check", "rule-id", rule_id, "rc", rc_str, NULL); free(rc_str); switch (result) { case pcmk_rc_within_range: case pcmk_rc_ok: case pcmk_rc_after_range: case pcmk_rc_before_range: case pcmk_rc_op_unsatisfied: return pcmk_rc_ok; default: out->err(out, "Could not determine whether rule %s is in effect: %s", rule_id, ((error != NULL)? error : "unexpected error")); return pcmk_rc_ok; } } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_none(pcmk__output_t *out, va_list args) { return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_text(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); static int code_width = 0; if (out->is_quiet(out)) { /* If out->is_quiet(), don't print the code. Print name and/or desc in a * compact format for text output, or print nothing at all for none-type * output. */ if ((name != NULL) && (desc != NULL)) { pcmk__formatted_printf(out, "%s - %s\n", name, desc); } else if ((name != NULL) || (desc != NULL)) { pcmk__formatted_printf(out, "%s\n", ((name != NULL)? name : desc)); } return pcmk_rc_ok; } /* Get length of longest (most negative) standard Pacemaker return code * This should be longer than all the values of any other type of return * code. */ if (code_width == 0) { long long most_negative = pcmk_rc_error - (long long) pcmk__n_rc + 1; code_width = (int) snprintf(NULL, 0, "%lld", most_negative); } if ((name != NULL) && (desc != NULL)) { static int name_width = 0; if (name_width == 0) { // Get length of longest standard Pacemaker return code name for (int lpc = 0; lpc < pcmk__n_rc; lpc++) { int len = (int) strlen(pcmk_rc_name(pcmk_rc_error - lpc)); name_width = QB_MAX(name_width, len); } } return out->info(out, "% *d: %-*s %s", code_width, code, name_width, name, desc); } if ((name != NULL) || (desc != NULL)) { return out->info(out, "% *d: %s", code_width, code, ((name != NULL)? name : desc)); } return out->info(out, "% *d", code_width, code); } PCMK__OUTPUT_ARGS("result-code", "int", "const char *", "const char *") static int result_code_xml(pcmk__output_t *out, va_list args) { int code = va_arg(args, int); const char *name = va_arg(args, const char *); const char *desc = va_arg(args, const char *); char *code_str = pcmk__itoa(code); pcmk__output_create_xml_node(out, "result-code", "code", code_str, PCMK_XA_NAME, name, PCMK_XA_DESCRIPTION, desc, NULL); free(code_str); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "attribute", "default", attribute_default }, { "attribute", "xml", attribute_xml }, { "cluster-status", "default", pcmk__cluster_status_text }, { "cluster-status", "html", cluster_status_html }, { "cluster-status", "xml", cluster_status_xml }, { "crmadmin-node", "default", crmadmin_node }, { "crmadmin-node", "text", crmadmin_node_text }, { "crmadmin-node", "xml", crmadmin_node_xml }, { "dc", "default", dc }, { "dc", "text", dc_text }, { "dc", "xml", dc_xml }, { "digests", "default", digests_text }, { "digests", "xml", digests_xml }, { "health", "default", health }, { "health", "text", health_text }, { "health", "xml", health_xml }, { "inject-attr", "default", inject_attr }, { "inject-attr", "xml", inject_attr_xml }, { "inject-cluster-action", "default", inject_cluster_action }, { "inject-cluster-action", "xml", inject_cluster_action_xml }, { "inject-fencing-action", "default", inject_fencing_action }, { "inject-fencing-action", "xml", inject_fencing_action_xml }, { "inject-modify-config", "default", inject_modify_config }, { "inject-modify-config", "xml", inject_modify_config_xml }, { "inject-modify-node", "default", inject_modify_node }, { "inject-modify-node", "xml", inject_modify_node_xml }, { "inject-modify-ticket", "default", inject_modify_ticket }, { "inject-modify-ticket", "xml", inject_modify_ticket_xml }, { "inject-pseudo-action", "default", inject_pseudo_action }, { "inject-pseudo-action", "xml", inject_pseudo_action_xml }, { "inject-rsc-action", "default", inject_rsc_action }, { "inject-rsc-action", "xml", inject_rsc_action_xml }, { "inject-spec", "default", inject_spec }, { "inject-spec", "xml", inject_spec_xml }, { "locations-list", "default", locations_list }, { "locations-list", "xml", locations_list_xml }, { "node-action", "default", node_action }, { "node-action", "xml", node_action_xml }, { "node-info", "default", node_info_default }, { "node-info", "xml", node_info_xml }, { "pacemakerd-health", "default", pacemakerd_health }, { "pacemakerd-health", "html", pacemakerd_health_html }, { "pacemakerd-health", "text", pacemakerd_health_text }, { "pacemakerd-health", "xml", pacemakerd_health_xml }, { "profile", "default", profile_default, }, { "profile", "xml", profile_xml }, { "result-code", "none", result_code_none }, { "result-code", "text", result_code_text }, { "result-code", "xml", result_code_xml }, { "rsc-action", "default", rsc_action_default }, { "rsc-action-item", "default", rsc_action_item }, { "rsc-action-item", "xml", rsc_action_item_xml }, { "rsc-is-colocated-with-list", "default", rsc_is_colocated_with_list }, { "rsc-is-colocated-with-list", "xml", rsc_is_colocated_with_list_xml }, { "rscs-colocated-with-list", "default", rscs_colocated_with_list }, { "rscs-colocated-with-list", "xml", rscs_colocated_with_list_xml }, { "rule-check", "default", rule_check_default }, { "rule-check", "xml", rule_check_xml }, { "locations-and-colocations", "default", locations_and_colocations }, { "locations-and-colocations", "xml", locations_and_colocations_xml }, { NULL, NULL, NULL } }; void pcmk__register_lib_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } diff --git a/lib/pacemaker/pcmk_resource.c b/lib/pacemaker/pcmk_resource.c index 71d5aa2b21..fe3605aef5 100644 --- a/lib/pacemaker/pcmk_resource.c +++ b/lib/pacemaker/pcmk_resource.c @@ -1,248 +1,248 @@ /* * Copyright 2021-2024 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 // Search path for resource operation history (takes node name and resource ID) #define XPATH_OP_HISTORY "//" XML_CIB_TAG_STATUS \ - "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ + "/" XML_CIB_TAG_STATE "[@" PCMK_XA_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" PCMK_XA_ID "='%s']" static xmlNode * best_op(const pcmk_resource_t *rsc, const pcmk_node_t *node) { char *xpath = NULL; xmlNode *history = NULL; xmlNode *best = NULL; bool best_effective_op = false; guint best_interval = 0; bool best_failure = false; const char *best_digest = NULL; // Find node's resource history xpath = crm_strdup_printf(XPATH_OP_HISTORY, node->details->uname, rsc->id); history = get_xpath_object(xpath, rsc->cluster->input, LOG_NEVER); free(xpath); // Examine each history entry for (xmlNode *lrm_rsc_op = first_named_child(history, XML_LRM_TAG_RSC_OP); lrm_rsc_op != NULL; lrm_rsc_op = crm_next_same_xml(lrm_rsc_op)) { const char *digest = crm_element_value(lrm_rsc_op, XML_LRM_ATTR_RESTART_DIGEST); guint interval_ms = 0; const char *task = crm_element_value(lrm_rsc_op, XML_LRM_ATTR_TASK); bool effective_op = false; bool failure = pcmk__ends_with(ID(lrm_rsc_op), "_last_failure_0"); crm_element_value_ms(lrm_rsc_op, XML_LRM_ATTR_INTERVAL, &interval_ms); effective_op = interval_ms == 0 && pcmk__strcase_any_of(task, PCMK_ACTION_MONITOR, PCMK_ACTION_START, PCMK_ACTION_PROMOTE, PCMK_ACTION_MIGRATE_FROM, NULL); if (best == NULL) { goto is_best; } if (best_effective_op) { // Do not use an ineffective op if there's an effective one. if (!effective_op) { continue; } // Do not use an ineffective non-recurring op if there's a recurring one } else if (best_interval != 0 && !effective_op && interval_ms == 0) { continue; } // Do not use last failure if there's a successful one. if (!best_failure && failure) { continue; } // Do not use an op without a restart digest if there's one with. if (best_digest != NULL && digest == NULL) { continue; } // Do not use an older op if there's a newer one. if (pe__is_newer_op(best, lrm_rsc_op, true) > 0) { continue; } is_best: best = lrm_rsc_op; best_effective_op = effective_op; best_interval = interval_ms; best_failure = failure; best_digest = digest; } return best; } /*! * \internal * \brief Remove a resource * * \param[in,out] cib An open connection to the CIB * \param[in] cib_opts Options to use in the CIB operation call * \param[in] rsc_id Resource to remove * \param[in] rsc_type Type of the resource ("primitive", "group", etc.) * * \return Standard Pacemaker return code */ int pcmk__resource_delete(cib_t *cib, uint32_t cib_opts, const char *rsc_id, const char *rsc_type) { int rc = pcmk_rc_ok; xmlNode *msg_data = NULL; if (cib == NULL) { return ENOTCONN; } if (rsc_id == NULL || rsc_type == NULL) { return EINVAL; } msg_data = create_xml_node(NULL, rsc_type); crm_xml_add(msg_data, PCMK_XA_ID, rsc_id); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, msg_data, cib_opts); rc = pcmk_legacy2rc(rc); free_xml(msg_data); return rc; } int pcmk_resource_delete(xmlNodePtr *xml, const char *rsc_id, const char *rsc_type) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; uint32_t cib_opts = cib_sync_call; cib_t *cib = NULL; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } cib = cib_new(); if (cib == NULL) { rc = pcmk_rc_cib_corrupt; goto done; } rc = cib->cmds->signon(cib, crm_system_name, cib_command); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { goto done; } rc = pcmk__resource_delete(cib, cib_opts, rsc_id, rsc_type); done: if (cib != NULL) { cib__clean_up_connection(&cib); } pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } /*! * \internal * \brief Calculate and output resource operation digests * * \param[in,out] out Output object * \param[in,out] rsc Resource to calculate digests for * \param[in] node Node whose operation history should be used * \param[in] overrides Hash table of configuration parameters to override * * \return Standard Pacemaker return code */ int pcmk__resource_digests(pcmk__output_t *out, pcmk_resource_t *rsc, const pcmk_node_t *node, GHashTable *overrides) { const char *task = NULL; xmlNode *xml_op = NULL; pcmk__op_digest_t *digests = NULL; guint interval_ms = 0; int rc = pcmk_rc_ok; if ((out == NULL) || (rsc == NULL) || (node == NULL)) { return EINVAL; } if (rsc->variant != pcmk_rsc_variant_primitive) { // Only primitives get operation digests return EOPNOTSUPP; } // Find XML of operation history to use xml_op = best_op(rsc, node); // Generate an operation key if (xml_op != NULL) { task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); } if (task == NULL) { // Assume start if no history is available task = PCMK_ACTION_START; interval_ms = 0; } // Calculate and show digests digests = pe__calculate_digests(rsc, task, &interval_ms, node, xml_op, overrides, true, rsc->cluster); rc = out->message(out, "digests", rsc, node, task, interval_ms, digests); pe__free_digests(digests); return rc; } // @COMPAT The scheduler parameter is unused and can be removed at the next break int pcmk_resource_digests(xmlNodePtr *xml, pcmk_resource_t *rsc, const pcmk_node_t *node, GHashTable *overrides, pcmk_scheduler_t *scheduler) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pcmk__register_lib_messages(out); rc = pcmk__resource_digests(out, rsc, node, overrides); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } diff --git a/lib/pacemaker/pcmk_sched_actions.c b/lib/pacemaker/pcmk_sched_actions.c index 8cce32d6a9..15d4044ca0 100644 --- a/lib/pacemaker/pcmk_sched_actions.c +++ b/lib/pacemaker/pcmk_sched_actions.c @@ -1,1932 +1,1932 @@ /* * Copyright 2004-2024 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 "libpacemaker_private.h" /*! * \internal * \brief Get the action flags relevant to ordering constraints * * \param[in,out] action Action to check * \param[in] node Node that *other* action in the ordering is on * (used only for clone resource actions) * * \return Action flags that should be used for orderings */ static uint32_t action_flags_for_ordering(pcmk_action_t *action, const pcmk_node_t *node) { bool runnable = false; uint32_t flags; // For non-resource actions, return the action flags if (action->rsc == NULL) { return action->flags; } /* For non-clone resources, or a clone action not assigned to a node, * return the flags as determined by the resource method without a node * specified. */ flags = action->rsc->cmds->action_flags(action, NULL); if ((node == NULL) || !pe_rsc_is_clone(action->rsc)) { return flags; } /* Otherwise (i.e., for clone resource actions on a specific node), first * remember whether the non-node-specific action is runnable. */ runnable = pcmk_is_set(flags, pcmk_action_runnable); // Then recheck the resource method with the node flags = action->rsc->cmds->action_flags(action, node); /* For clones in ordering constraints, the node-specific "runnable" doesn't * matter, just the non-node-specific setting (i.e., is the action runnable * anywhere). * * This applies only to runnable, and only for ordering constraints. This * function shouldn't be used for other types of constraints without * changes. Not very satisfying, but it's logical and appears to work well. */ if (runnable && !pcmk_is_set(flags, pcmk_action_runnable)) { pcmk__set_raw_action_flags(flags, action->rsc->id, pcmk_action_runnable); } return flags; } /*! * \internal * \brief Get action UUID that should be used with a resource ordering * * When an action is ordered relative to an action for a collective resource * (clone, group, or bundle), it actually needs to be ordered after all * instances of the collective have completed the relevant action (for example, * given "start CLONE then start RSC", RSC must wait until all instances of * CLONE have started). Given the UUID and resource of the first action in an * ordering, this returns the UUID of the action that should actually be used * for ordering (for example, "CLONE_started_0" instead of "CLONE_start_0"). * * \param[in] first_uuid UUID of first action in ordering * \param[in] first_rsc Resource of first action in ordering * * \return Newly allocated copy of UUID to use with ordering * \note It is the caller's responsibility to free the return value. */ static char * action_uuid_for_ordering(const char *first_uuid, const pcmk_resource_t *first_rsc) { guint interval_ms = 0; char *uuid = NULL; char *rid = NULL; char *first_task_str = NULL; enum action_tasks first_task = pcmk_action_unspecified; enum action_tasks remapped_task = pcmk_action_unspecified; // Only non-notify actions for collective resources need remapping if ((strstr(first_uuid, PCMK_ACTION_NOTIFY) != NULL) || (first_rsc->variant < pcmk_rsc_variant_group)) { goto done; } // Only non-recurring actions need remapping CRM_ASSERT(parse_op_key(first_uuid, &rid, &first_task_str, &interval_ms)); if (interval_ms > 0) { goto done; } first_task = text2task(first_task_str); switch (first_task) { case pcmk_action_stop: case pcmk_action_start: case pcmk_action_notify: case pcmk_action_promote: case pcmk_action_demote: remapped_task = first_task + 1; break; case pcmk_action_stopped: case pcmk_action_started: case pcmk_action_notified: case pcmk_action_promoted: case pcmk_action_demoted: remapped_task = first_task; break; case pcmk_action_monitor: case pcmk_action_shutdown: case pcmk_action_fence: break; default: crm_err("Unknown action '%s' in ordering", first_task_str); break; } if (remapped_task != pcmk_action_unspecified) { /* If a clone or bundle has notifications enabled, the ordering will be * relative to when notifications have been sent for the remapped task. */ if (pcmk_is_set(first_rsc->flags, pcmk_rsc_notify) && (pe_rsc_is_clone(first_rsc) || pe_rsc_is_bundled(first_rsc))) { uuid = pcmk__notify_key(rid, "confirmed-post", task2text(remapped_task)); } else { uuid = pcmk__op_key(rid, task2text(remapped_task), 0); } pcmk__rsc_trace(first_rsc, "Remapped action UUID %s to %s for ordering purposes", first_uuid, uuid); } done: if (uuid == NULL) { uuid = strdup(first_uuid); CRM_ASSERT(uuid != NULL); } free(first_task_str); free(rid); return uuid; } /*! * \internal * \brief Get actual action that should be used with an ordering * * When an action is ordered relative to an action for a collective resource * (clone, group, or bundle), it actually needs to be ordered after all * instances of the collective have completed the relevant action (for example, * given "start CLONE then start RSC", RSC must wait until all instances of * CLONE have started). Given the first action in an ordering, this returns the * the action that should actually be used for ordering (for example, the * started action instead of the start action). * * \param[in] action First action in an ordering * * \return Actual action that should be used for the ordering */ static pcmk_action_t * action_for_ordering(pcmk_action_t *action) { pcmk_action_t *result = action; pcmk_resource_t *rsc = action->rsc; if ((rsc != NULL) && (rsc->variant >= pcmk_rsc_variant_group) && (action->uuid != NULL)) { char *uuid = action_uuid_for_ordering(action->uuid, rsc); result = find_first_action(rsc->actions, uuid, NULL, NULL); if (result == NULL) { crm_warn("Not remapping %s to %s because %s does not have " "remapped action", action->uuid, uuid, rsc->id); result = action; } free(uuid); } return result; } /*! * \internal * \brief Wrapper for update_ordered_actions() method for readability * * \param[in,out] rsc Resource to call method for * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this * node (only used when interleaving instances) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates * (may include pcmk_action_optional to affect only * mandatory actions, and pe_action_runnable to * affect only runnable actions) * \param[in] type Group of enum pcmk__action_relation_flags to apply * \param[in,out] scheduler Scheduler data * * \return Group of enum pcmk__updated flags indicating what was updated */ static inline uint32_t update(pcmk_resource_t *rsc, pcmk_action_t *first, pcmk_action_t *then, const pcmk_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pcmk_scheduler_t *scheduler) { return rsc->cmds->update_ordered_actions(first, then, node, flags, filter, type, scheduler); } /*! * \internal * \brief Update flags for ordering's actions appropriately for ordering's flags * * \param[in,out] first First action in an ordering * \param[in,out] then Then action in an ordering * \param[in] first_flags Action flags for \p first for ordering purposes * \param[in] then_flags Action flags for \p then for ordering purposes * \param[in,out] order Action wrapper for \p first in ordering * \param[in,out] scheduler Scheduler data * * \return Group of enum pcmk__updated flags */ static uint32_t update_action_for_ordering_flags(pcmk_action_t *first, pcmk_action_t *then, uint32_t first_flags, uint32_t then_flags, pcmk__related_action_t *order, pcmk_scheduler_t *scheduler) { uint32_t changed = pcmk__updated_none; /* The node will only be used for clones. If interleaved, node will be NULL, * otherwise the ordering scope will be limited to the node. Normally, the * whole 'then' clone should restart if 'first' is restarted, so then->node * is needed. */ pcmk_node_t *node = then->node; if (pcmk_is_set(order->type, pcmk__ar_first_implies_same_node_then)) { /* For unfencing, only instances of 'then' on the same node as 'first' * (the unfencing operation) should restart, so reset node to * first->node, at which point this case is handled like a normal * pcmk__ar_first_implies_then. */ pcmk__clear_relation_flags(order->type, pcmk__ar_first_implies_same_node_then); pcmk__set_relation_flags(order->type, pcmk__ar_first_implies_then); node = first->node; pcmk__rsc_trace(then->rsc, "%s then %s: mapped " "pcmk__ar_first_implies_same_node_then to " "pcmk__ar_first_implies_then on %s", first->uuid, then->uuid, pe__node_name(node)); } if (pcmk_is_set(order->type, pcmk__ar_first_implies_then)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags & pcmk_action_optional, pcmk_action_optional, pcmk__ar_first_implies_then, scheduler); } else if (!pcmk_is_set(first_flags, pcmk_action_optional) && pcmk_is_set(then->flags, pcmk_action_optional)) { pcmk__clear_action_flags(then, pcmk_action_optional); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_first_implies_then", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_intermediate_stop) && (then->rsc != NULL)) { enum pe_action_flags restart = pcmk_action_optional |pcmk_action_runnable; changed |= update(then->rsc, first, then, node, first_flags, restart, pcmk__ar_intermediate_stop, scheduler); pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_intermediate_stop", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_then_implies_first)) { if (first->rsc != NULL) { changed |= update(first->rsc, first, then, node, first_flags, pcmk_action_optional, pcmk__ar_then_implies_first, scheduler); } else if (!pcmk_is_set(first_flags, pcmk_action_optional) && pcmk_is_set(first->flags, pcmk_action_runnable)) { pcmk__clear_action_flags(first, pcmk_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_first); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_then_implies_first", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_promoted_then_implies_first)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags & pcmk_action_optional, pcmk_action_optional, pcmk__ar_promoted_then_implies_first, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after " "pcmk__ar_promoted_then_implies_first", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_min_runnable)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_runnable, pcmk__ar_min_runnable, scheduler); } else if (pcmk_is_set(first_flags, pcmk_action_runnable)) { // We have another runnable instance of "first" then->runnable_before++; /* Mark "then" as runnable if it requires a certain number of * "before" instances to be runnable, and they now are. */ if ((then->runnable_before >= then->required_runnable_before) && !pcmk_is_set(then->flags, pcmk_action_runnable)) { pcmk__set_action_flags(then, pcmk_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_min_runnable", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_nested_remote_probe) && (then->rsc != NULL)) { if (!pcmk_is_set(first_flags, pcmk_action_runnable) && (first->rsc != NULL) && (first->rsc->running_on != NULL)) { pcmk__rsc_trace(then->rsc, "%s then %s: ignoring because first is stopping", first->uuid, then->uuid); order->type = (enum pe_ordering) pcmk__ar_none; } else { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_runnable, pcmk__ar_unrunnable_first_blocks, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_nested_remote_probe", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_unrunnable_first_blocks)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_runnable, pcmk__ar_unrunnable_first_blocks, scheduler); } else if (!pcmk_is_set(first_flags, pcmk_action_runnable) && pcmk_is_set(then->flags, pcmk_action_runnable)) { pcmk__clear_action_flags(then, pcmk_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_unrunnable_first_blocks", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_unmigratable_then_blocks)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_optional, pcmk__ar_unmigratable_then_blocks, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after " "pcmk__ar_unmigratable_then_blocks", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_first_else_then)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_optional, pcmk__ar_first_else_then, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_first_else_then", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_ordered)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_runnable, pcmk__ar_ordered, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_ordered", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(order->type, pcmk__ar_asymmetric)) { if (then->rsc != NULL) { changed |= update(then->rsc, first, then, node, first_flags, pcmk_action_runnable, pcmk__ar_asymmetric, scheduler); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after pcmk__ar_asymmetric", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } if (pcmk_is_set(first->flags, pcmk_action_runnable) && pcmk_is_set(order->type, pcmk__ar_first_implies_then_graphed) && !pcmk_is_set(first_flags, pcmk_action_optional)) { pcmk__rsc_trace(then->rsc, "%s will be in graph because %s is required", then->uuid, first->uuid); pcmk__set_action_flags(then, pcmk_action_always_in_graph); // Don't bother marking 'then' as changed just for this } if (pcmk_is_set(order->type, pcmk__ar_then_implies_first_graphed) && !pcmk_is_set(then_flags, pcmk_action_optional)) { pcmk__rsc_trace(then->rsc, "%s will be in graph because %s is required", first->uuid, then->uuid); pcmk__set_action_flags(first, pcmk_action_always_in_graph); // Don't bother marking 'first' as changed just for this } if (pcmk_any_flags_set(order->type, pcmk__ar_first_implies_then |pcmk__ar_then_implies_first |pcmk__ar_intermediate_stop) && (first->rsc != NULL) && !pcmk_is_set(first->rsc->flags, pcmk_rsc_managed) && pcmk_is_set(first->rsc->flags, pcmk_rsc_blocked) && !pcmk_is_set(first->flags, pcmk_action_runnable) && pcmk__str_eq(first->task, PCMK_ACTION_STOP, pcmk__str_none)) { if (pcmk_is_set(then->flags, pcmk_action_runnable)) { pcmk__clear_action_flags(then, pcmk_action_runnable); pcmk__set_updated_flags(changed, first, pcmk__updated_then); } pcmk__rsc_trace(then->rsc, "%s then %s: %s after checking whether first " "is blocked, unmanaged, unrunnable stop", first->uuid, then->uuid, (changed? "changed" : "unchanged")); } return changed; } // Convenience macros for logging action properties #define action_type_str(flags) \ (pcmk_is_set((flags), pcmk_action_pseudo)? "pseudo-action" : "action") #define action_optional_str(flags) \ (pcmk_is_set((flags), pcmk_action_optional)? "optional" : "required") #define action_runnable_str(flags) \ (pcmk_is_set((flags), pcmk_action_runnable)? "runnable" : "unrunnable") #define action_node_str(a) \ (((a)->node == NULL)? "no node" : (a)->node->details->uname) /*! * \internal * \brief Update an action's flags for all orderings where it is "then" * * \param[in,out] then Action to update * \param[in,out] scheduler Scheduler data */ void pcmk__update_action_for_orderings(pcmk_action_t *then, pcmk_scheduler_t *scheduler) { GList *lpc = NULL; uint32_t changed = pcmk__updated_none; int last_flags = then->flags; pcmk__rsc_trace(then->rsc, "Updating %s %s (%s %s) on %s", action_type_str(then->flags), then->uuid, action_optional_str(then->flags), action_runnable_str(then->flags), action_node_str(then)); if (pcmk_is_set(then->flags, pcmk_action_min_runnable)) { /* Initialize current known "runnable before" actions. As * update_action_for_ordering_flags() is called for each of then's * before actions, this number will increment as runnable 'first' * actions are encountered. */ then->runnable_before = 0; if (then->required_runnable_before == 0) { /* @COMPAT This ordering constraint uses the deprecated * "require-all=false" attribute. Treat it like * PCMK_META_CLONE_MIN=1. */ then->required_runnable_before = 1; } /* The pcmk__ar_min_runnable clause of * update_action_for_ordering_flags() (called below) * will reset runnable if appropriate. */ pcmk__clear_action_flags(then, pcmk_action_runnable); } for (lpc = then->actions_before; lpc != NULL; lpc = lpc->next) { pcmk__related_action_t *other = lpc->data; pcmk_action_t *first = other->action; pcmk_node_t *then_node = then->node; pcmk_node_t *first_node = first->node; if ((first->rsc != NULL) && (first->rsc->variant == pcmk_rsc_variant_group) && pcmk__str_eq(first->task, PCMK_ACTION_START, pcmk__str_none)) { first_node = first->rsc->fns->location(first->rsc, NULL, FALSE); if (first_node != NULL) { pcmk__rsc_trace(first->rsc, "Found %s for 'first' %s", pe__node_name(first_node), first->uuid); } } if ((then->rsc != NULL) && (then->rsc->variant == pcmk_rsc_variant_group) && pcmk__str_eq(then->task, PCMK_ACTION_START, pcmk__str_none)) { then_node = then->rsc->fns->location(then->rsc, NULL, FALSE); if (then_node != NULL) { pcmk__rsc_trace(then->rsc, "Found %s for 'then' %s", pe__node_name(then_node), then->uuid); } } // Disable constraint if it only applies when on same node, but isn't if (pcmk_is_set(other->type, pcmk__ar_if_on_same_node) && (first_node != NULL) && (then_node != NULL) && !pe__same_node(first_node, then_node)) { pcmk__rsc_trace(then->rsc, "Disabled ordering %s on %s then %s on %s: " "not same node", other->action->uuid, pe__node_name(first_node), then->uuid, pe__node_name(then_node)); other->type = (enum pe_ordering) pcmk__ar_none; continue; } pcmk__clear_updated_flags(changed, then, pcmk__updated_first); if ((first->rsc != NULL) && pcmk_is_set(other->type, pcmk__ar_then_cancels_first) && !pcmk_is_set(then->flags, pcmk_action_optional)) { /* 'then' is required, so we must abandon 'first' * (e.g. a required stop cancels any agent reload). */ pcmk__set_action_flags(other->action, pcmk_action_optional); if (!strcmp(first->task, PCMK_ACTION_RELOAD_AGENT)) { pcmk__clear_rsc_flags(first->rsc, pcmk_rsc_reload); } } if ((first->rsc != NULL) && (then->rsc != NULL) && (first->rsc != then->rsc) && !is_parent(then->rsc, first->rsc)) { first = action_for_ordering(first); } if (first != other->action) { pcmk__rsc_trace(then->rsc, "Ordering %s after %s instead of %s", then->uuid, first->uuid, other->action->uuid); } pcmk__rsc_trace(then->rsc, "%s (%#.6x) then %s (%#.6x): type=%#.6x node=%s", first->uuid, first->flags, then->uuid, then->flags, other->type, action_node_str(first)); if (first == other->action) { /* 'first' was not remapped (e.g. from 'start' to 'running'), which * could mean it is a non-resource action, a primitive resource * action, or already expanded. */ uint32_t first_flags, then_flags; first_flags = action_flags_for_ordering(first, then_node); then_flags = action_flags_for_ordering(then, first_node); changed |= update_action_for_ordering_flags(first, then, first_flags, then_flags, other, scheduler); /* 'first' was for a complex resource (clone, group, etc), * create a new dependency if necessary */ } else if (order_actions(first, then, other->type)) { /* This was the first time 'first' and 'then' were associated, * start again to get the new actions_before list */ pcmk__set_updated_flags(changed, then, pcmk__updated_then); pcmk__rsc_trace(then->rsc, "Disabled ordering %s then %s in favor of %s " "then %s", other->action->uuid, then->uuid, first->uuid, then->uuid); other->type = (enum pe_ordering) pcmk__ar_none; } if (pcmk_is_set(changed, pcmk__updated_first)) { crm_trace("Re-processing %s and its 'after' actions " "because it changed", first->uuid); for (GList *lpc2 = first->actions_after; lpc2 != NULL; lpc2 = lpc2->next) { pcmk__related_action_t *other = lpc2->data; pcmk__update_action_for_orderings(other->action, scheduler); } pcmk__update_action_for_orderings(first, scheduler); } } if (pcmk_is_set(then->flags, pcmk_action_min_runnable)) { if (last_flags == then->flags) { pcmk__clear_updated_flags(changed, then, pcmk__updated_then); } else { pcmk__set_updated_flags(changed, then, pcmk__updated_then); } } if (pcmk_is_set(changed, pcmk__updated_then)) { crm_trace("Re-processing %s and its 'after' actions because it changed", then->uuid); if (pcmk_is_set(last_flags, pcmk_action_runnable) && !pcmk_is_set(then->flags, pcmk_action_runnable)) { pcmk__block_colocation_dependents(then); } pcmk__update_action_for_orderings(then, scheduler); for (lpc = then->actions_after; lpc != NULL; lpc = lpc->next) { pcmk__related_action_t *other = lpc->data; pcmk__update_action_for_orderings(other->action, scheduler); } } } static inline bool is_primitive_action(const pcmk_action_t *action) { return (action != NULL) && (action->rsc != NULL) && (action->rsc->variant == pcmk_rsc_variant_primitive); } /*! * \internal * \brief Clear a single action flag and set reason text * * \param[in,out] action Action whose flag should be cleared * \param[in] flag Action flag that should be cleared * \param[in] reason Action that is the reason why flag is being cleared */ #define clear_action_flag_because(action, flag, reason) do { \ if (pcmk_is_set((action)->flags, (flag))) { \ pcmk__clear_action_flags(action, flag); \ if ((action)->rsc != (reason)->rsc) { \ char *reason_text = pe__action2reason((reason), (flag)); \ pe_action_set_reason((action), reason_text, false); \ free(reason_text); \ } \ } \ } while (0) /*! * \internal * \brief Update actions in an asymmetric ordering * * If the "first" action in an asymmetric ordering is unrunnable, make the * "second" action unrunnable as well, if appropriate. * * \param[in] first 'First' action in an asymmetric ordering * \param[in,out] then 'Then' action in an asymmetric ordering */ static void handle_asymmetric_ordering(const pcmk_action_t *first, pcmk_action_t *then) { /* Only resource actions after an unrunnable 'first' action need updates for * asymmetric ordering. */ if ((then->rsc == NULL) || pcmk_is_set(first->flags, pcmk_action_runnable)) { return; } // Certain optional 'then' actions are unaffected by unrunnable 'first' if (pcmk_is_set(then->flags, pcmk_action_optional)) { enum rsc_role_e then_rsc_role = then->rsc->fns->state(then->rsc, TRUE); if ((then_rsc_role == pcmk_role_stopped) && pcmk__str_eq(then->task, PCMK_ACTION_STOP, pcmk__str_none)) { /* If 'then' should stop after 'first' but is already stopped, the * ordering is irrelevant. */ return; } else if ((then_rsc_role >= pcmk_role_started) && pcmk__str_eq(then->task, PCMK_ACTION_START, pcmk__str_none) && pe__rsc_running_on_only(then->rsc, then->node)) { /* Similarly if 'then' should start after 'first' but is already * started on a single node. */ return; } } // 'First' can't run, so 'then' can't either clear_action_flag_because(then, pcmk_action_optional, first); clear_action_flag_because(then, pcmk_action_runnable, first); } /*! * \internal * \brief Set action bits appropriately when pe_restart_order is used * * \param[in,out] first 'First' action in an ordering with pe_restart_order * \param[in,out] then 'Then' action in an ordering with pe_restart_order * \param[in] filter What action flags to care about * * \note pe_restart_order is set for "stop resource before starting it" and * "stop later group member before stopping earlier group member" */ static void handle_restart_ordering(pcmk_action_t *first, pcmk_action_t *then, uint32_t filter) { const char *reason = NULL; CRM_ASSERT(is_primitive_action(first)); CRM_ASSERT(is_primitive_action(then)); // We need to update the action in two cases: // ... if 'then' is required if (pcmk_is_set(filter, pcmk_action_optional) && !pcmk_is_set(then->flags, pcmk_action_optional)) { reason = "restart"; } /* ... if 'then' is unrunnable action on same resource (if a resource * should restart but can't start, we still want to stop) */ if (pcmk_is_set(filter, pcmk_action_runnable) && !pcmk_is_set(then->flags, pcmk_action_runnable) && pcmk_is_set(then->rsc->flags, pcmk_rsc_managed) && (first->rsc == then->rsc)) { reason = "stop"; } if (reason == NULL) { return; } pcmk__rsc_trace(first->rsc, "Handling %s -> %s for %s", first->uuid, then->uuid, reason); // Make 'first' required if it is runnable if (pcmk_is_set(first->flags, pcmk_action_runnable)) { clear_action_flag_because(first, pcmk_action_optional, then); } // Make 'first' required if 'then' is required if (!pcmk_is_set(then->flags, pcmk_action_optional)) { clear_action_flag_because(first, pcmk_action_optional, then); } // Make 'first' unmigratable if 'then' is unmigratable if (!pcmk_is_set(then->flags, pcmk_action_migratable)) { clear_action_flag_because(first, pcmk_action_migratable, then); } // Make 'then' unrunnable if 'first' is required but unrunnable if (!pcmk_is_set(first->flags, pcmk_action_optional) && !pcmk_is_set(first->flags, pcmk_action_runnable)) { clear_action_flag_because(then, pcmk_action_runnable, first); } } /*! * \internal * \brief Update two actions according to an ordering between them * * Given information about an ordering of two actions, update the actions' flags * (and runnable_before members if appropriate) as appropriate for the ordering. * Effects may cascade to other orderings involving the actions as well. * * \param[in,out] first 'First' action in an ordering * \param[in,out] then 'Then' action in an ordering * \param[in] node If not NULL, limit scope of ordering to this node * (ignored) * \param[in] flags Action flags for \p first for ordering purposes * \param[in] filter Action flags to limit scope of certain updates (may * include pcmk_action_optional to affect only * mandatory actions, and pcmk_action_runnable to * affect only runnable actions) * \param[in] type Group of enum pcmk__action_relation_flags to apply * \param[in,out] scheduler Scheduler data * * \return Group of enum pcmk__updated flags indicating what was updated */ uint32_t pcmk__update_ordered_actions(pcmk_action_t *first, pcmk_action_t *then, const pcmk_node_t *node, uint32_t flags, uint32_t filter, uint32_t type, pcmk_scheduler_t *scheduler) { uint32_t changed = pcmk__updated_none; uint32_t then_flags = 0U; uint32_t first_flags = 0U; CRM_ASSERT((first != NULL) && (then != NULL) && (scheduler != NULL)); then_flags = then->flags; first_flags = first->flags; if (pcmk_is_set(type, pcmk__ar_asymmetric)) { handle_asymmetric_ordering(first, then); } if (pcmk_is_set(type, pcmk__ar_then_implies_first) && !pcmk_is_set(then_flags, pcmk_action_optional)) { // Then is required, and implies first should be, too if (pcmk_is_set(filter, pcmk_action_optional) && !pcmk_is_set(flags, pcmk_action_optional) && pcmk_is_set(first_flags, pcmk_action_optional)) { clear_action_flag_because(first, pcmk_action_optional, then); } if (pcmk_is_set(flags, pcmk_action_migratable) && !pcmk_is_set(then->flags, pcmk_action_migratable)) { clear_action_flag_because(first, pcmk_action_migratable, then); } } if (pcmk_is_set(type, pcmk__ar_promoted_then_implies_first) && (then->rsc != NULL) && (then->rsc->role == pcmk_role_promoted) && pcmk_is_set(filter, pcmk_action_optional) && !pcmk_is_set(then->flags, pcmk_action_optional)) { clear_action_flag_because(first, pcmk_action_optional, then); if (pcmk_is_set(first->flags, pcmk_action_migratable) && !pcmk_is_set(then->flags, pcmk_action_migratable)) { clear_action_flag_because(first, pcmk_action_migratable, then); } } if (pcmk_is_set(type, pcmk__ar_unmigratable_then_blocks) && pcmk_is_set(filter, pcmk_action_optional)) { if (!pcmk_all_flags_set(then->flags, pcmk_action_migratable |pcmk_action_runnable)) { clear_action_flag_because(first, pcmk_action_runnable, then); } if (!pcmk_is_set(then->flags, pcmk_action_optional)) { clear_action_flag_because(first, pcmk_action_optional, then); } } if (pcmk_is_set(type, pcmk__ar_first_else_then) && pcmk_is_set(filter, pcmk_action_optional) && !pcmk_is_set(first->flags, pcmk_action_runnable)) { clear_action_flag_because(then, pcmk_action_migratable, first); pcmk__clear_action_flags(then, pcmk_action_pseudo); } if (pcmk_is_set(type, pcmk__ar_unrunnable_first_blocks) && pcmk_is_set(filter, pcmk_action_runnable) && pcmk_is_set(then->flags, pcmk_action_runnable) && !pcmk_is_set(flags, pcmk_action_runnable)) { clear_action_flag_because(then, pcmk_action_runnable, first); clear_action_flag_because(then, pcmk_action_migratable, first); } if (pcmk_is_set(type, pcmk__ar_first_implies_then) && pcmk_is_set(filter, pcmk_action_optional) && pcmk_is_set(then->flags, pcmk_action_optional) && !pcmk_is_set(flags, pcmk_action_optional) && !pcmk_is_set(first->flags, pcmk_action_migratable)) { clear_action_flag_because(then, pcmk_action_optional, first); } if (pcmk_is_set(type, pcmk__ar_intermediate_stop)) { handle_restart_ordering(first, then, filter); } if (then_flags != then->flags) { pcmk__set_updated_flags(changed, first, pcmk__updated_then); pcmk__rsc_trace(then->rsc, "%s on %s: flags are now %#.6x (was %#.6x) " "because of 'first' %s (%#.6x)", then->uuid, pe__node_name(then->node), then->flags, then_flags, first->uuid, first->flags); if ((then->rsc != NULL) && (then->rsc->parent != NULL)) { // Required to handle "X_stop then X_start" for cloned groups pcmk__update_action_for_orderings(then, scheduler); } } if (first_flags != first->flags) { pcmk__set_updated_flags(changed, first, pcmk__updated_first); pcmk__rsc_trace(first->rsc, "%s on %s: flags are now %#.6x (was %#.6x) " "because of 'then' %s (%#.6x)", first->uuid, pe__node_name(first->node), first->flags, first_flags, then->uuid, then->flags); } return changed; } /*! * \internal * \brief Trace-log an action (optionally with its dependent actions) * * \param[in] pre_text If not NULL, prefix the log with this plus ": " * \param[in] action Action to log * \param[in] details If true, recursively log dependent actions */ void pcmk__log_action(const char *pre_text, const pcmk_action_t *action, bool details) { const char *node_uname = NULL; const char *node_uuid = NULL; const char *desc = NULL; CRM_CHECK(action != NULL, return); if (!pcmk_is_set(action->flags, pcmk_action_pseudo)) { if (action->node != NULL) { node_uname = action->node->details->uname; node_uuid = action->node->details->id; } else { node_uname = ""; } } switch (text2task(action->task)) { case pcmk_action_fence: case pcmk_action_shutdown: if (pcmk_is_set(action->flags, pcmk_action_pseudo)) { desc = "Pseudo "; } else if (pcmk_is_set(action->flags, pcmk_action_optional)) { desc = "Optional "; } else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) { desc = "!!Non-Startable!! "; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; default: if (pcmk_is_set(action->flags, pcmk_action_optional)) { desc = "Optional "; } else if (pcmk_is_set(action->flags, pcmk_action_pseudo)) { desc = "Pseudo "; } else if (!pcmk_is_set(action->flags, pcmk_action_runnable)) { desc = "!!Non-Startable!! "; } else { desc = "(Provisional) "; } crm_trace("%s%s%sAction %d: %s %s%s%s%s%s%s", ((pre_text == NULL)? "" : pre_text), ((pre_text == NULL)? "" : ": "), desc, action->id, action->uuid, (action->rsc? action->rsc->id : ""), (node_uname? "\ton " : ""), (node_uname? node_uname : ""), (node_uuid? "\t\t(" : ""), (node_uuid? node_uuid : ""), (node_uuid? ")" : "")); break; } if (details) { const GList *iter = NULL; const pcmk__related_action_t *other = NULL; crm_trace("\t\t====== Preceding Actions"); for (iter = action->actions_before; iter != NULL; iter = iter->next) { other = (const pcmk__related_action_t *) iter->data; pcmk__log_action("\t\t", other->action, false); } crm_trace("\t\t====== Subsequent Actions"); for (iter = action->actions_after; iter != NULL; iter = iter->next) { other = (const pcmk__related_action_t *) iter->data; pcmk__log_action("\t\t", other->action, false); } crm_trace("\t\t====== End"); } else { crm_trace("\t\t(before=%d, after=%d)", g_list_length(action->actions_before), g_list_length(action->actions_after)); } } /*! * \internal * \brief Create a new shutdown action for a node * * \param[in,out] node Node being shut down * * \return Newly created shutdown action for \p node */ pcmk_action_t * pcmk__new_shutdown_action(pcmk_node_t *node) { char *shutdown_id = NULL; pcmk_action_t *shutdown_op = NULL; CRM_ASSERT(node != NULL); shutdown_id = crm_strdup_printf("%s-%s", PCMK_ACTION_DO_SHUTDOWN, node->details->uname); shutdown_op = custom_action(NULL, shutdown_id, PCMK_ACTION_DO_SHUTDOWN, node, FALSE, node->details->data_set); pcmk__order_stops_before_shutdown(node, shutdown_op); add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); return shutdown_op; } /*! * \internal * \brief Calculate and add an operation digest to XML * * Calculate an operation digest, which enables us to later determine when a * restart is needed due to the resource's parameters being changed, and add it * to given XML. * * \param[in] op Operation result from executor * \param[in,out] update XML to add digest to */ static void add_op_digest_to_xml(const lrmd_event_data_t *op, xmlNode *update) { char *digest = NULL; xmlNode *args_xml = NULL; if (op->params == NULL) { return; } args_xml = create_xml_node(NULL, XML_TAG_PARAMS); g_hash_table_foreach(op->params, hash2field, args_xml); pcmk__filter_op_for_digest(args_xml); digest = calculate_operation_digest(args_xml, NULL); crm_xml_add(update, XML_LRM_ATTR_OP_DIGEST, digest); free_xml(args_xml); free(digest); } #define FAKE_TE_ID "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx" /*! * \internal * \brief Create XML for resource operation history update * * \param[in,out] parent Parent XML node to add to * \param[in,out] op Operation event data * \param[in] caller_version DC feature set * \param[in] target_rc Expected result of operation * \param[in] node Name of node on which operation was performed * \param[in] origin Arbitrary description of update source * * \return Newly created XML node for history update */ xmlNode * pcmk__create_history_xml(xmlNode *parent, lrmd_event_data_t *op, const char *caller_version, int target_rc, const char *node, const char *origin) { char *key = NULL; char *magic = NULL; char *op_id = NULL; char *op_id_additional = NULL; char *local_user_data = NULL; const char *exit_reason = NULL; xmlNode *xml_op = NULL; const char *task = NULL; CRM_CHECK(op != NULL, return NULL); crm_trace("Creating history XML for %s-interval %s action for %s on %s " "(DC version: %s, origin: %s)", pcmk__readable_interval(op->interval_ms), op->op_type, op->rsc_id, ((node == NULL)? "no node" : node), caller_version, origin); task = op->op_type; /* Record a successful agent reload as a start, and a failed one as a * monitor, to make life easier for the scheduler when determining the * current state. * * @COMPAT We should check "reload" here only if the operation was for a * pre-OCF-1.1 resource agent, but we don't know that here, and we should * only ever get results for actions scheduled by us, so we can reasonably * assume any "reload" is actually a pre-1.1 agent reload. */ if (pcmk__str_any_of(task, PCMK_ACTION_RELOAD, PCMK_ACTION_RELOAD_AGENT, NULL)) { if (op->op_status == PCMK_EXEC_DONE) { task = PCMK_ACTION_START; } else { task = PCMK_ACTION_MONITOR; } } key = pcmk__op_key(op->rsc_id, task, op->interval_ms); if (pcmk__str_eq(task, PCMK_ACTION_NOTIFY, pcmk__str_none)) { const char *n_type = crm_meta_value(op->params, "notify_type"); const char *n_task = crm_meta_value(op->params, "notify_operation"); CRM_LOG_ASSERT(n_type != NULL); CRM_LOG_ASSERT(n_task != NULL); op_id = pcmk__notify_key(op->rsc_id, n_type, n_task); if (op->op_status != PCMK_EXEC_PENDING) { /* Ignore notify errors. * * @TODO It might be better to keep the correct result here, and * ignore it in process_graph_event(). */ lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } /* Migration history is preserved separately, which usually matters for * multiple nodes and is important for future cluster transitions. */ } else if (pcmk__str_any_of(op->op_type, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL)) { op_id = strdup(key); } else if (did_rsc_op_fail(op, target_rc)) { op_id = pcmk__op_key(op->rsc_id, "last_failure", 0); if (op->interval_ms == 0) { // Ensure 'last' gets updated, in case record-pending is true op_id_additional = pcmk__op_key(op->rsc_id, "last", 0); } exit_reason = op->exit_reason; } else if (op->interval_ms > 0) { op_id = strdup(key); } else { op_id = pcmk__op_key(op->rsc_id, "last", 0); } again: xml_op = pcmk__xe_match(parent, XML_LRM_TAG_RSC_OP, PCMK_XA_ID, op_id); if (xml_op == NULL) { xml_op = create_xml_node(parent, XML_LRM_TAG_RSC_OP); } if (op->user_data == NULL) { crm_debug("Generating fake transition key for: " PCMK__OP_FMT " %d from %s", op->rsc_id, op->op_type, op->interval_ms, op->call_id, origin); local_user_data = pcmk__transition_key(-1, op->call_id, target_rc, FAKE_TE_ID); op->user_data = local_user_data; } if (magic == NULL) { magic = crm_strdup_printf("%d:%d;%s", op->op_status, op->rc, (const char *) op->user_data); } crm_xml_add(xml_op, PCMK_XA_ID, op_id); crm_xml_add(xml_op, XML_LRM_ATTR_TASK_KEY, key); crm_xml_add(xml_op, XML_LRM_ATTR_TASK, task); crm_xml_add(xml_op, PCMK_XA_CRM_DEBUG_ORIGIN, origin); crm_xml_add(xml_op, PCMK_XA_CRM_FEATURE_SET, caller_version); crm_xml_add(xml_op, XML_ATTR_TRANSITION_KEY, op->user_data); crm_xml_add(xml_op, XML_ATTR_TRANSITION_MAGIC, magic); crm_xml_add(xml_op, XML_LRM_ATTR_EXIT_REASON, pcmk__s(exit_reason, "")); crm_xml_add(xml_op, XML_LRM_ATTR_TARGET, node); // For context during triage crm_xml_add_int(xml_op, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add_int(xml_op, XML_LRM_ATTR_RC, op->rc); crm_xml_add_int(xml_op, XML_LRM_ATTR_OPSTATUS, op->op_status); crm_xml_add_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, op->interval_ms); if (compare_version("2.1", caller_version) <= 0) { if (op->t_run || op->t_rcchange || op->exec_time || op->queue_time) { crm_trace("Timing data (" PCMK__OP_FMT "): last=%u change=%u exec=%u queue=%u", op->rsc_id, op->op_type, op->interval_ms, op->t_run, op->t_rcchange, op->exec_time, op->queue_time); if ((op->interval_ms != 0) && (op->t_rcchange != 0)) { // Recurring ops may have changed rc after initial run crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_rcchange); } else { crm_xml_add_ll(xml_op, XML_RSC_OP_LAST_CHANGE, (long long) op->t_run); } crm_xml_add_int(xml_op, XML_RSC_OP_T_EXEC, op->exec_time); crm_xml_add_int(xml_op, XML_RSC_OP_T_QUEUE, op->queue_time); } } if (pcmk__str_any_of(op->op_type, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL)) { /* * Record migrate_source and migrate_target always for migrate ops. */ const char *name = XML_LRM_ATTR_MIGRATE_SOURCE; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); name = XML_LRM_ATTR_MIGRATE_TARGET; crm_xml_add(xml_op, name, crm_meta_value(op->params, name)); } add_op_digest_to_xml(op, xml_op); if (op_id_additional) { free(op_id); op_id = op_id_additional; op_id_additional = NULL; goto again; } if (local_user_data) { free(local_user_data); op->user_data = NULL; } free(magic); free(op_id); free(key); return xml_op; } /*! * \internal * \brief Check whether an action shutdown-locks a resource to a node * * If the PCMK_OPT_SHUTDOWN_LOCK cluster property is set, resources will not be * recovered on a different node if cleanly stopped, and may start only on that * same node. This function checks whether that applies to a given action, so * that the transition graph can be marked appropriately. * * \param[in] action Action to check * * \return true if \p action locks its resource to the action's node, * otherwise false */ bool pcmk__action_locks_rsc_to_node(const pcmk_action_t *action) { // Only resource actions taking place on resource's lock node are locked if ((action == NULL) || (action->rsc == NULL) || !pe__same_node(action->node, action->rsc->lock_node)) { return false; } /* During shutdown, only stops are locked (otherwise, another action such as * a demote would cause the controller to clear the lock) */ if (action->node->details->shutdown && (action->task != NULL) && (strcmp(action->task, PCMK_ACTION_STOP) != 0)) { return false; } return true; } /* lowest to highest */ static gint sort_action_id(gconstpointer a, gconstpointer b) { const pcmk__related_action_t *action_wrapper2 = a; const pcmk__related_action_t *action_wrapper1 = b; if (a == NULL) { return 1; } if (b == NULL) { return -1; } if (action_wrapper1->action->id < action_wrapper2->action->id) { return 1; } if (action_wrapper1->action->id > action_wrapper2->action->id) { return -1; } return 0; } /*! * \internal * \brief Remove any duplicate action inputs, merging action flags * * \param[in,out] action Action whose inputs should be checked */ void pcmk__deduplicate_action_inputs(pcmk_action_t *action) { GList *item = NULL; GList *next = NULL; pcmk__related_action_t *last_input = NULL; action->actions_before = g_list_sort(action->actions_before, sort_action_id); for (item = action->actions_before; item != NULL; item = next) { pcmk__related_action_t *input = item->data; next = item->next; if ((last_input != NULL) && (input->action->id == last_input->action->id)) { crm_trace("Input %s (%d) duplicate skipped for action %s (%d)", input->action->uuid, input->action->id, action->uuid, action->id); /* For the purposes of scheduling, the ordering flags no longer * matter, but crm_simulate looks at certain ones when creating a * dot graph. Combining the flags is sufficient for that purpose. */ last_input->type |= input->type; if (input->state == pe_link_dumped) { last_input->state = pe_link_dumped; } free(item->data); action->actions_before = g_list_delete_link(action->actions_before, item); } else { last_input = input; input->state = pe_link_not_dumped; } } } /*! * \internal * \brief Output all scheduled actions * * \param[in,out] scheduler Scheduler data */ void pcmk__output_actions(pcmk_scheduler_t *scheduler) { pcmk__output_t *out = scheduler->priv; // Output node (non-resource) actions for (GList *iter = scheduler->actions; iter != NULL; iter = iter->next) { char *node_name = NULL; char *task = NULL; pcmk_action_t *action = (pcmk_action_t *) iter->data; if (action->rsc != NULL) { continue; // Resource actions will be output later } else if (pcmk_is_set(action->flags, pcmk_action_optional)) { continue; // This action was not scheduled } if (pcmk__str_eq(action->task, PCMK_ACTION_DO_SHUTDOWN, pcmk__str_none)) { task = strdup("Shutdown"); } else if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH, pcmk__str_none)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); task = crm_strdup_printf("Fence (%s)", op); } else { continue; // Don't display other node action types } if (pe__is_guest_node(action->node)) { const pcmk_resource_t *remote = action->node->details->remote_rsc; node_name = crm_strdup_printf("%s (resource: %s)", pe__node_name(action->node), remote->container->id); } else if (action->node != NULL) { node_name = crm_strdup_printf("%s", pe__node_name(action->node)); } out->message(out, "node-action", task, node_name, action->reason); free(node_name); free(task); } // Output resource actions for (GList *iter = scheduler->resources; iter != NULL; iter = iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data; rsc->cmds->output_actions(rsc); } } /*! * \internal * \brief Get action name needed to compare digest for configuration changes * * \param[in] task Action name from history * \param[in] interval_ms Action interval (in milliseconds) * * \return Action name whose digest should be compared */ static const char * task_for_digest(const char *task, guint interval_ms) { /* Certain actions need to be compared against the parameters used to start * the resource. */ if ((interval_ms == 0) && pcmk__str_any_of(task, PCMK_ACTION_MONITOR, PCMK_ACTION_MIGRATE_FROM, PCMK_ACTION_PROMOTE, NULL)) { task = PCMK_ACTION_START; } return task; } /*! * \internal * \brief Check whether only sanitized parameters to an action changed * * When collecting CIB files for troubleshooting, crm_report will mask * sensitive resource parameters. If simulations were run using that, affected * resources would appear to need a restart, which would complicate * troubleshooting. To avoid that, we save a "secure digest" of non-sensitive * parameters. This function used that digest to check whether only masked * parameters are different. * * \param[in] xml_op Resource history entry with secure digest * \param[in] digest_data Operation digest information being compared * \param[in] scheduler Scheduler data * * \return true if only sanitized parameters changed, otherwise false */ static bool only_sanitized_changed(const xmlNode *xml_op, const pcmk__op_digest_t *digest_data, const pcmk_scheduler_t *scheduler) { const char *digest_secure = NULL; if (!pcmk_is_set(scheduler->flags, pcmk_sched_sanitized)) { // The scheduler is not being run as a simulation return false; } digest_secure = crm_element_value(xml_op, XML_LRM_ATTR_SECURE_DIGEST); return (digest_data->rc != pcmk__digest_match) && (digest_secure != NULL) && (digest_data->digest_secure_calc != NULL) && (strcmp(digest_data->digest_secure_calc, digest_secure) == 0); } /*! * \internal * \brief Force a restart due to a configuration change * * \param[in,out] rsc Resource that action is for * \param[in] task Name of action whose configuration changed * \param[in] interval_ms Action interval (in milliseconds) * \param[in,out] node Node where resource should be restarted */ static void force_restart(pcmk_resource_t *rsc, const char *task, guint interval_ms, pcmk_node_t *node) { char *key = pcmk__op_key(rsc->id, task, interval_ms); pcmk_action_t *required = custom_action(rsc, key, task, NULL, FALSE, rsc->cluster); pe_action_set_reason(required, "resource definition change", true); trigger_unfencing(rsc, node, "Device parameters changed", NULL, rsc->cluster); } /*! * \internal * \brief Schedule a reload of a resource on a node * * \param[in,out] data Resource to reload * \param[in] user_data Where resource should be reloaded */ static void schedule_reload(gpointer data, gpointer user_data) { pcmk_resource_t *rsc = data; const pcmk_node_t *node = user_data; pcmk_action_t *reload = NULL; // For collective resources, just call recursively for children if (rsc->variant > pcmk_rsc_variant_primitive) { g_list_foreach(rsc->children, schedule_reload, user_data); return; } // Skip the reload in certain situations if ((node == NULL) || !pcmk_is_set(rsc->flags, pcmk_rsc_managed) || pcmk_is_set(rsc->flags, pcmk_rsc_failed)) { pcmk__rsc_trace(rsc, "Skip reload of %s:%s%s %s", rsc->id, pcmk_is_set(rsc->flags, pcmk_rsc_managed)? "" : " unmanaged", pcmk_is_set(rsc->flags, pcmk_rsc_failed)? " failed" : "", (node == NULL)? "inactive" : node->details->uname); return; } /* If a resource's configuration changed while a start was pending, * force a full restart instead of a reload. */ if (pcmk_is_set(rsc->flags, pcmk_rsc_start_pending)) { pcmk__rsc_trace(rsc, "%s: preventing agent reload because start pending", rsc->id); custom_action(rsc, stop_key(rsc), PCMK_ACTION_STOP, node, FALSE, rsc->cluster); return; } // Schedule the reload pcmk__set_rsc_flags(rsc, pcmk_rsc_reload); reload = custom_action(rsc, reload_key(rsc), PCMK_ACTION_RELOAD_AGENT, node, FALSE, rsc->cluster); pe_action_set_reason(reload, "resource definition change", FALSE); // Set orderings so that a required stop or demote cancels the reload pcmk__new_ordering(NULL, NULL, reload, rsc, stop_key(rsc), NULL, pcmk__ar_ordered|pcmk__ar_then_cancels_first, rsc->cluster); pcmk__new_ordering(NULL, NULL, reload, rsc, demote_key(rsc), NULL, pcmk__ar_ordered|pcmk__ar_then_cancels_first, rsc->cluster); } /*! * \internal * \brief Handle any configuration change for an action * * Given an action from resource history, if the resource's configuration * changed since the action was done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, etc.). * * \param[in,out] rsc Resource that action is for * \param[in,out] node Node that action was on * \param[in] xml_op Action XML from resource history * * \return true if action configuration changed, otherwise false */ bool pcmk__check_action_config(pcmk_resource_t *rsc, pcmk_node_t *node, const xmlNode *xml_op) { guint interval_ms = 0; const char *task = NULL; const pcmk__op_digest_t *digest_data = NULL; CRM_CHECK((rsc != NULL) && (node != NULL) && (xml_op != NULL), return false); task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); CRM_CHECK(task != NULL, return false); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); // If this is a recurring action, check whether it has been orphaned if (interval_ms > 0) { if (pcmk__find_action_config(rsc, task, interval_ms, false) != NULL) { pcmk__rsc_trace(rsc, "%s-interval %s for %s on %s is in configuration", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); } else if (pcmk_is_set(rsc->cluster->flags, pcmk_sched_cancel_removed_actions)) { pcmk__schedule_cancel(rsc, crm_element_value(xml_op, XML_LRM_ATTR_CALLID), task, interval_ms, node, "orphan"); return true; } else { pcmk__rsc_debug(rsc, "%s-interval %s for %s on %s is orphaned", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); return true; } } crm_trace("Checking %s-interval %s for %s on %s for configuration changes", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node)); task = task_for_digest(task, interval_ms); digest_data = rsc_action_digest_cmp(rsc, xml_op, node, rsc->cluster); if (only_sanitized_changed(xml_op, digest_data, rsc->cluster)) { if (!pcmk__is_daemon && (rsc->cluster->priv != NULL)) { pcmk__output_t *out = rsc->cluster->priv; out->info(out, "Only 'private' parameters to %s-interval %s for %s " "on %s changed: %s", pcmk__readable_interval(interval_ms), task, rsc->id, pe__node_name(node), crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC)); } return false; } switch (digest_data->rc) { case pcmk__digest_restart: crm_log_xml_debug(digest_data->params_restart, "params:restart"); force_restart(rsc, task, interval_ms, node); return true; case pcmk__digest_unknown: case pcmk__digest_mismatch: // Changes that can potentially be handled by an agent reload if (interval_ms > 0) { /* Recurring actions aren't reloaded per se, they are just * re-scheduled so the next run uses the new parameters. * The old instance will be cancelled automatically. */ crm_log_xml_debug(digest_data->params_all, "params:reschedule"); pcmk__reschedule_recurring(rsc, task, interval_ms, node); } else if (crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) { // Agent supports reload, so use it trigger_unfencing(rsc, node, "Device parameters changed (reload)", NULL, rsc->cluster); crm_log_xml_debug(digest_data->params_all, "params:reload"); schedule_reload((gpointer) rsc, (gpointer) node); } else { pcmk__rsc_trace(rsc, "Restarting %s " "because agent doesn't support reload", rsc->id); crm_log_xml_debug(digest_data->params_restart, "params:restart"); force_restart(rsc, task, interval_ms, node); } return true; default: break; } return false; } /*! * \internal * \brief Create a list of resource's action history entries, sorted by call ID * * \param[in] rsc_entry Resource's status XML * \param[out] start_index Where to store index of start-like action, if any * \param[out] stop_index Where to store index of stop action, if any */ static GList * rsc_history_as_list(const xmlNode *rsc_entry, int *start_index, int *stop_index) { GList *ops = NULL; for (xmlNode *rsc_op = first_named_child(rsc_entry, XML_LRM_TAG_RSC_OP); rsc_op != NULL; rsc_op = crm_next_same_xml(rsc_op)) { ops = g_list_prepend(ops, rsc_op); } ops = g_list_sort(ops, sort_op_by_callid); calculate_active_ops(ops, start_index, stop_index); return ops; } /*! * \internal * \brief Process a resource's action history from the CIB status * * Given a resource's action history, if the resource's configuration * changed since the actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in] rsc_entry Resource's status XML * \param[in,out] rsc Resource whose history is being processed * \param[in,out] node Node whose history is being processed */ static void process_rsc_history(const xmlNode *rsc_entry, pcmk_resource_t *rsc, pcmk_node_t *node) { int offset = -1; int stop_index = 0; int start_index = 0; GList *sorted_op_list = NULL; if (pcmk_is_set(rsc->flags, pcmk_rsc_removed)) { if (pe_rsc_is_anon_clone(pe__const_top_resource(rsc, false))) { pcmk__rsc_trace(rsc, "Skipping configuration check " "for orphaned clone instance %s", rsc->id); } else { pcmk__rsc_trace(rsc, "Skipping configuration check and scheduling " "clean-up for orphaned resource %s", rsc->id); pcmk__schedule_cleanup(rsc, node, false); } return; } if (pe_find_node_id(rsc->running_on, node->details->id) == NULL) { if (pcmk__rsc_agent_changed(rsc, node, rsc_entry, false)) { pcmk__schedule_cleanup(rsc, node, false); } pcmk__rsc_trace(rsc, "Skipping configuration check for %s " "because no longer active on %s", rsc->id, pe__node_name(node)); return; } pcmk__rsc_trace(rsc, "Checking for configuration changes for %s on %s", rsc->id, pe__node_name(node)); if (pcmk__rsc_agent_changed(rsc, node, rsc_entry, true)) { pcmk__schedule_cleanup(rsc, node, false); } sorted_op_list = rsc_history_as_list(rsc_entry, &start_index, &stop_index); if (start_index < stop_index) { return; // Resource is stopped } for (GList *iter = sorted_op_list; iter != NULL; iter = iter->next) { xmlNode *rsc_op = (xmlNode *) iter->data; const char *task = NULL; guint interval_ms = 0; if (++offset < start_index) { // Skip actions that happened before a start continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if ((interval_ms > 0) && (pcmk_is_set(rsc->flags, pcmk_rsc_maintenance) || node->details->maintenance)) { // Maintenance mode cancels recurring operations pcmk__schedule_cancel(rsc, crm_element_value(rsc_op, XML_LRM_ATTR_CALLID), task, interval_ms, node, "maintenance mode"); } else if ((interval_ms > 0) || pcmk__strcase_any_of(task, PCMK_ACTION_MONITOR, PCMK_ACTION_START, PCMK_ACTION_PROMOTE, PCMK_ACTION_MIGRATE_FROM, NULL)) { /* If a resource operation failed, and the operation's definition * has changed, clear any fail count so they can be retried fresh. */ if (pe__bundle_needs_remote_name(rsc)) { /* We haven't assigned resources to nodes yet, so if the * REMOTE_CONTAINER_HACK is used, we may calculate the digest * based on the literal "#uname" value rather than the properly * substituted value. That would mistakenly make the action * definition appear to have been changed. Defer the check until * later in this case. */ pe__add_param_check(rsc_op, rsc, node, pcmk__check_active, rsc->cluster); } else if (pcmk__check_action_config(rsc, node, rsc_op) && (pe_get_failcount(node, rsc, NULL, pcmk__fc_effective, NULL) != 0)) { pe__clear_failcount(rsc, node, "action definition changed", rsc->cluster); } } } g_list_free(sorted_op_list); } /*! * \internal * \brief Process a node's action history from the CIB status * * Given a node's resource history, if the resource's configuration changed * since the actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in,out] node Node whose history is being processed * \param[in] lrm_rscs Node's from CIB status XML */ static void process_node_history(pcmk_node_t *node, const xmlNode *lrm_rscs) { crm_trace("Processing node history for %s", pe__node_name(node)); for (const xmlNode *rsc_entry = first_named_child(lrm_rscs, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { if (rsc_entry->children != NULL) { GList *result = pcmk__rscs_matching_id(ID(rsc_entry), node->details->data_set); for (GList *iter = result; iter != NULL; iter = iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) iter->data; if (rsc->variant == pcmk_rsc_variant_primitive) { process_rsc_history(rsc_entry, rsc, node); } } g_list_free(result); } } } // XPath to find a node's resource history #define XPATH_NODE_HISTORY "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ - "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ + "/" XML_CIB_TAG_STATE "[@" PCMK_XA_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES /*! * \internal * \brief Process any resource configuration changes in the CIB status * * Go through all nodes' resource history, and if a resource's configuration * changed since its actions were done, schedule any actions needed (restart, * reload, unfencing, rescheduling recurring actions, clean-up, etc.). * (This also cancels recurring actions for maintenance mode, which is not * entirely related but convenient to do here.) * * \param[in,out] scheduler Scheduler data */ void pcmk__handle_rsc_config_changes(pcmk_scheduler_t *scheduler) { crm_trace("Check resource and action configuration for changes"); /* Rather than iterate through the status section, iterate through the nodes * and search for the appropriate status subsection for each. This skips * orphaned nodes and lets us eliminate some cases before searching the XML. */ for (GList *iter = scheduler->nodes; iter != NULL; iter = iter->next) { pcmk_node_t *node = (pcmk_node_t *) iter->data; /* Don't bother checking actions for a node that can't run actions ... * unless it's in maintenance mode, in which case we still need to * cancel any existing recurring monitors. */ if (node->details->maintenance || pcmk__node_available(node, false, false)) { char *xpath = NULL; xmlNode *history = NULL; xpath = crm_strdup_printf(XPATH_NODE_HISTORY, node->details->uname); history = get_xpath_object(xpath, scheduler->input, LOG_NEVER); free(xpath); process_node_history(node, history); } } } diff --git a/lib/pacemaker/pcmk_simulate.c b/lib/pacemaker/pcmk_simulate.c index 5591523f3c..95708694e1 100644 --- a/lib/pacemaker/pcmk_simulate.c +++ b/lib/pacemaker/pcmk_simulate.c @@ -1,1006 +1,1006 @@ /* * Copyright 2021-2024 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 "libpacemaker_private.h" static pcmk__output_t *out = NULL; static cib_t *fake_cib = NULL; static GList *fake_resource_list = NULL; static const GList *fake_op_fail_list = NULL; static void set_effective_date(pcmk_scheduler_t *scheduler, bool print_original, const char *use_date); /*! * \internal * \brief Create an action name for use in a dot graph * * \param[in] action Action to create name for * \param[in] verbose If true, add action ID to name * * \return Newly allocated string with action name * \note It is the caller's responsibility to free the result. */ static char * create_action_name(const pcmk_action_t *action, bool verbose) { char *action_name = NULL; const char *prefix = ""; const char *action_host = NULL; const char *clone_name = NULL; const char *task = action->task; if (action->node != NULL) { action_host = action->node->details->uname; } else if (!pcmk_is_set(action->flags, pcmk_action_pseudo)) { action_host = ""; } if (pcmk__str_eq(action->task, PCMK_ACTION_CANCEL, pcmk__str_none)) { prefix = "Cancel "; task = action->cancel_task; } if (action->rsc != NULL) { clone_name = action->rsc->clone_name; } if (clone_name != NULL) { char *key = NULL; guint interval_ms = 0; if (pcmk__guint_from_hash(action->meta, XML_LRM_ATTR_INTERVAL_MS, 0, &interval_ms) != pcmk_rc_ok) { interval_ms = 0; } if (pcmk__strcase_any_of(action->task, PCMK_ACTION_NOTIFY, PCMK_ACTION_NOTIFIED, NULL)) { const char *n_type = g_hash_table_lookup(action->meta, "notify_key_type"); const char *n_task = g_hash_table_lookup(action->meta, "notify_key_operation"); CRM_ASSERT(n_type != NULL); CRM_ASSERT(n_task != NULL); key = pcmk__notify_key(clone_name, n_type, n_task); } else { key = pcmk__op_key(clone_name, task, interval_ms); } if (action_host != NULL) { action_name = crm_strdup_printf("%s%s %s", prefix, key, action_host); } else { action_name = crm_strdup_printf("%s%s", prefix, key); } free(key); } else if (pcmk__str_eq(action->task, PCMK_ACTION_STONITH, pcmk__str_none)) { const char *op = g_hash_table_lookup(action->meta, "stonith_action"); action_name = crm_strdup_printf("%s%s '%s' %s", prefix, action->task, op, action_host); } else if (action->rsc && action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->uuid, action_host); } else if (action_host) { action_name = crm_strdup_printf("%s%s %s", prefix, action->task, action_host); } else { action_name = crm_strdup_printf("%s", action->uuid); } if (verbose) { char *with_id = crm_strdup_printf("%s (%d)", action_name, action->id); free(action_name); action_name = with_id; } return action_name; } /*! * \internal * \brief Display the status of a cluster * * \param[in,out] scheduler Scheduler data * \param[in] show_opts How to modify display (as pcmk_show_opt_e flags) * \param[in] section_opts Sections to display (as pcmk_section_e flags) * \param[in] title What to use as list title * \param[in] print_spacer Whether to display a spacer first */ static void print_cluster_status(pcmk_scheduler_t *scheduler, uint32_t show_opts, uint32_t section_opts, const char *title, bool print_spacer) { pcmk__output_t *out = scheduler->priv; GList *all = NULL; crm_exit_t stonith_rc = 0; enum pcmk_pacemakerd_state state = pcmk_pacemakerd_state_invalid; section_opts |= pcmk_section_nodes | pcmk_section_resources; show_opts |= pcmk_show_inactive_rscs | pcmk_show_failed_detail; all = g_list_prepend(all, (gpointer) "*"); PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "%s", title); out->message(out, "cluster-status", scheduler, state, stonith_rc, NULL, false, section_opts, show_opts, NULL, all, all); out->end_list(out); g_list_free(all); } /*! * \internal * \brief Display a summary of all actions scheduled in a transition * * \param[in,out] scheduler Scheduler data (fully scheduled) * \param[in] print_spacer Whether to display a spacer first */ static void print_transition_summary(pcmk_scheduler_t *scheduler, bool print_spacer) { pcmk__output_t *out = scheduler->priv; PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "Transition Summary"); pcmk__output_actions(scheduler); out->end_list(out); } /*! * \internal * \brief Reset scheduler input, output, date, and flags * * \param[in,out] scheduler Scheduler data * \param[in] input What to set as cluster input * \param[in] out What to set as cluster output object * \param[in] use_date What to set as cluster's current timestamp * \param[in] flags Group of enum pcmk_scheduler_flags to set */ static void reset(pcmk_scheduler_t *scheduler, xmlNodePtr input, pcmk__output_t *out, const char *use_date, unsigned int flags) { scheduler->input = input; scheduler->priv = out; set_effective_date(scheduler, true, use_date); if (pcmk_is_set(flags, pcmk_sim_sanitized)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_sanitized); } if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_output_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_show_utilization); } } /*! * \brief Write out a file in dot(1) format describing the actions that will * be taken by the scheduler in response to an input CIB file. * * \param[in,out] scheduler Scheduler data * \param[in] dot_file The filename to write * \param[in] all_actions Write all actions, even those that are optional * or are on unmanaged resources * \param[in] verbose Add extra information, such as action IDs, to the * output * * \return Standard Pacemaker return code */ static int write_sim_dotfile(pcmk_scheduler_t *scheduler, const char *dot_file, bool all_actions, bool verbose) { GList *iter = NULL; FILE *dot_strm = fopen(dot_file, "w"); if (dot_strm == NULL) { return errno; } fprintf(dot_strm, " digraph \"g\" {\n"); for (iter = scheduler->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = (pcmk_action_t *) iter->data; const char *style = "dashed"; const char *font = "black"; const char *color = "black"; char *action_name = create_action_name(action, verbose); if (pcmk_is_set(action->flags, pcmk_action_pseudo)) { font = "orange"; } if (pcmk_is_set(action->flags, pcmk_action_added_to_graph)) { style = "bold"; color = "green"; } else if ((action->rsc != NULL) && !pcmk_is_set(action->rsc->flags, pcmk_rsc_managed)) { color = "red"; font = "purple"; if (!all_actions) { goto do_not_write; } } else if (pcmk_is_set(action->flags, pcmk_action_optional)) { color = "blue"; if (!all_actions) { goto do_not_write; } } else { color = "red"; CRM_LOG_ASSERT(!pcmk_is_set(action->flags, pcmk_action_runnable)); } pcmk__set_action_flags(action, pcmk_action_added_to_graph); fprintf(dot_strm, "\"%s\" [ style=%s color=\"%s\" fontcolor=\"%s\"]\n", action_name, style, color, font); do_not_write: free(action_name); } for (iter = scheduler->actions; iter != NULL; iter = iter->next) { pcmk_action_t *action = (pcmk_action_t *) iter->data; for (GList *before_iter = action->actions_before; before_iter != NULL; before_iter = before_iter->next) { pcmk__related_action_t *before = before_iter->data; char *before_name = NULL; char *after_name = NULL; const char *style = "dashed"; bool optional = true; if (before->state == pe_link_dumped) { optional = false; style = "bold"; } else if ((uint32_t) before->type == pcmk__ar_none) { continue; } else if (pcmk_is_set(before->action->flags, pcmk_action_added_to_graph) && pcmk_is_set(action->flags, pcmk_action_added_to_graph) && (uint32_t) before->type != pcmk__ar_if_on_same_node_or_target) { optional = false; } if (all_actions || !optional) { before_name = create_action_name(before->action, verbose); after_name = create_action_name(action, verbose); fprintf(dot_strm, "\"%s\" -> \"%s\" [ style = %s]\n", before_name, after_name, style); free(before_name); free(after_name); } } } fprintf(dot_strm, "}\n"); fflush(dot_strm); fclose(dot_strm); return pcmk_rc_ok; } /*! * \brief Profile the configuration updates and scheduler actions in a single * CIB file, printing the profiling timings. * * \note \p scheduler->priv must have been set to a valid \p pcmk__output_t * object before this function is called. * * \param[in] xml_file The CIB file to profile * \param[in] repeat Number of times to run * \param[in,out] scheduler Scheduler data * \param[in] use_date The date to set the cluster's time to (may be NULL) */ static void profile_file(const char *xml_file, long long repeat, pcmk_scheduler_t *scheduler, const char *use_date) { pcmk__output_t *out = scheduler->priv; xmlNode *cib_object = NULL; clock_t start = 0; clock_t end; unsigned long long scheduler_flags = pcmk_sched_no_compat; CRM_ASSERT(out != NULL); cib_object = filename2xml(xml_file); start = clock(); if (pcmk_find_cib_element(cib_object, XML_CIB_TAG_STATUS) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return; } if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) { scheduler_flags |= pcmk_sched_output_scores; } if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) { scheduler_flags |= pcmk_sched_show_utilization; } for (int i = 0; i < repeat; ++i) { xmlNode *input = (repeat == 1)? cib_object : copy_xml(cib_object); scheduler->input = input; set_effective_date(scheduler, false, use_date); pcmk__schedule_actions(input, scheduler_flags, scheduler); pe_reset_working_set(scheduler); } end = clock(); out->message(out, "profile", xml_file, start, end); } void pcmk__profile_dir(const char *dir, long long repeat, pcmk_scheduler_t *scheduler, const char *use_date) { pcmk__output_t *out = scheduler->priv; struct dirent **namelist; int file_num = scandir(dir, &namelist, 0, alphasort); CRM_ASSERT(out != NULL); if (file_num > 0) { struct stat prop; char buffer[FILENAME_MAX]; out->begin_list(out, NULL, NULL, "Timings"); while (file_num--) { if ('.' == namelist[file_num]->d_name[0]) { free(namelist[file_num]); continue; } else if (!pcmk__ends_with_ext(namelist[file_num]->d_name, ".xml")) { free(namelist[file_num]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", dir, namelist[file_num]->d_name); if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) { profile_file(buffer, repeat, scheduler, use_date); } free(namelist[file_num]); } free(namelist); out->end_list(out); } } /*! * \brief Set the date of the cluster, either to the value given by * \p use_date, or to the "execution-date" value in the CIB. * * \note \p scheduler->priv must have been set to a valid \p pcmk__output_t * object before this function is called. * * \param[in,out] scheduler Scheduler data * \param[in] print_original If \p true, the "execution-date" should * also be printed * \param[in] use_date The date to set the cluster's time to * (may be NULL) */ static void set_effective_date(pcmk_scheduler_t *scheduler, bool print_original, const char *use_date) { pcmk__output_t *out = scheduler->priv; time_t original_date = 0; CRM_ASSERT(out != NULL); crm_element_value_epoch(scheduler->input, "execution-date", &original_date); if (use_date) { scheduler->now = crm_time_new(use_date); out->info(out, "Setting effective cluster time: %s", use_date); crm_time_log(LOG_NOTICE, "Pretending 'now' is", scheduler->now, crm_time_log_date | crm_time_log_timeofday); } else if (original_date != 0) { scheduler->now = pcmk__copy_timet(original_date); if (print_original) { char *when = crm_time_as_string(scheduler->now, crm_time_log_date|crm_time_log_timeofday); out->info(out, "Using the original execution date of: %s", when); free(when); } } } /*! * \internal * \brief Simulate successfully executing a pseudo-action in a graph * * \param[in,out] graph Graph to update with pseudo-action result * \param[in,out] action Pseudo-action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_pseudo_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK_KEY); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-pseudo-action", node, task); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate executing a resource action in a graph * * \param[in,out] graph Graph to update with resource action result * \param[in,out] action Resource action to simulate executing * * \return Standard Pacemaker return code */ static int simulate_resource_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { int rc; lrmd_event_data_t *op = NULL; int target_outcome = PCMK_OCF_OK; const char *rtype = NULL; const char *rclass = NULL; const char *resource = NULL; const char *rprovider = NULL; const char *resource_config_name = NULL; const char *operation = crm_element_value(action->xml, "operation"); const char *target_rc_s = crm_meta_value(action->params, XML_ATTR_TE_TARGET_RC); xmlNode *cib_node = NULL; xmlNode *cib_resource = NULL; xmlNode *action_rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE); char *node = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET); char *uuid = NULL; const char *router_node = crm_element_value(action->xml, XML_LRM_ATTR_ROUTER_NODE); // Certain actions don't need to be displayed or history entries if (pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_none)) { crm_debug("No history injection for %s op on %s", operation, node); goto done; // Confirm action and update graph } if (action_rsc == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "Bad"); free(node); return EPROTO; } /* A resource might be known by different names in the configuration and in * the action (for example, a clone instance). Grab the configuration name * (which is preferred when writing history), and if necessary, the instance * name. */ resource_config_name = crm_element_value(action_rsc, PCMK_XA_ID); if (resource_config_name == NULL) { // Shouldn't be possible crm_log_xml_err(action->xml, "No ID"); free(node); return EPROTO; } resource = resource_config_name; if (pe_find_resource(fake_resource_list, resource) == NULL) { const char *longname = crm_element_value(action_rsc, PCMK__XA_LONG_ID); if ((longname != NULL) && (pe_find_resource(fake_resource_list, longname) != NULL)) { resource = longname; } } // Certain actions need to be displayed but don't need history entries if (pcmk__strcase_any_of(operation, PCMK_ACTION_DELETE, PCMK_ACTION_META_DATA, NULL)) { out->message(out, "inject-rsc-action", resource, operation, node, (guint) 0); goto done; // Confirm action and update graph } rclass = crm_element_value(action_rsc, PCMK_XA_CLASS); rtype = crm_element_value(action_rsc, PCMK_XA_TYPE); rprovider = crm_element_value(action_rsc, PCMK_XA_PROVIDER); pcmk__scan_min_int(target_rc_s, &target_outcome, 0); CRM_ASSERT(fake_cib->cmds->query(fake_cib, NULL, NULL, cib_sync_call|cib_scope_local) == pcmk_ok); // Ensure the action node is in the CIB uuid = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET_UUID); cib_node = pcmk__inject_node(fake_cib, node, ((router_node == NULL)? uuid: node)); free(uuid); CRM_ASSERT(cib_node != NULL); // Add a history entry for the action cib_resource = pcmk__inject_resource_history(out, cib_node, resource, resource_config_name, rclass, rtype, rprovider); if (cib_resource == NULL) { crm_err("Could not simulate action %d history for resource %s", action->id, resource); free(node); free_xml(cib_node); return EINVAL; } // Simulate and display an executor event for the action result op = pcmk__event_from_graph_action(cib_resource, action, PCMK_EXEC_DONE, target_outcome, "User-injected result"); out->message(out, "inject-rsc-action", resource, op->op_type, node, op->interval_ms); // Check whether action is in a list of desired simulated failures for (const GList *iter = fake_op_fail_list; iter != NULL; iter = iter->next) { const char *spec = (const char *) iter->data; char *key = NULL; const char *match_name = NULL; // Allow user to specify anonymous clone with or without instance number key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource; } free(key); // If not found, try the resource's name in the configuration if ((match_name == NULL) && (strcmp(resource, resource_config_name) != 0)) { key = crm_strdup_printf(PCMK__OP_FMT "@%s=", resource_config_name, op->op_type, op->interval_ms, node); if (strncasecmp(key, spec, strlen(key)) == 0) { match_name = resource_config_name; } free(key); } if (match_name == NULL) { continue; // This failed action entry doesn't match } // ${match_name}_${task}_${interval_in_ms}@${node}=${rc} rc = sscanf(spec, "%*[^=]=%d", (int *) &op->rc); if (rc != 1) { out->err(out, "Invalid failed operation '%s' " "(result code must be integer)", spec); continue; // Keep checking other list entries } out->info(out, "Pretending action %d failed with rc=%d", action->id, op->rc); pcmk__set_graph_action_flags(action, pcmk__graph_action_failed); graph->abort_priority = INFINITY; pcmk__inject_failcount(out, cib_node, match_name, op->op_type, op->interval_ms, op->rc); break; } pcmk__inject_action_result(cib_resource, op, target_outcome); lrmd_free_event(op); rc = fake_cib->cmds->modify(fake_cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); done: free(node); free_xml(cib_node); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a cluster action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Cluster action to simulate * * \return Standard Pacemaker return code */ static int simulate_cluster_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *node = crm_element_value(action->xml, XML_LRM_ATTR_TARGET); const char *task = crm_element_value(action->xml, XML_LRM_ATTR_TASK); xmlNode *rsc = first_named_child(action->xml, XML_CIB_TAG_RESOURCE); pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); out->message(out, "inject-cluster-action", node, task, rsc); pcmk__update_graph(graph, action); return pcmk_rc_ok; } /*! * \internal * \brief Simulate successfully executing a fencing action * * \param[in,out] graph Graph to update with action result * \param[in,out] action Fencing action to simulate * * \return Standard Pacemaker return code */ static int simulate_fencing_action(pcmk__graph_t *graph, pcmk__graph_action_t *action) { const char *op = crm_meta_value(action->params, "stonith_action"); char *target = crm_element_value_copy(action->xml, XML_LRM_ATTR_TARGET); out->message(out, "inject-fencing-action", target, op); if (!pcmk__str_eq(op, PCMK_ACTION_ON, pcmk__str_casei)) { int rc = pcmk_ok; GString *xpath = g_string_sized_new(512); // Set node state to offline xmlNode *cib_node = pcmk__inject_node_state_change(fake_cib, target, false); CRM_ASSERT(cib_node != NULL); crm_xml_add(cib_node, PCMK_XA_CRM_DEBUG_ORIGIN, __func__); rc = fake_cib->cmds->replace(fake_cib, XML_CIB_TAG_STATUS, cib_node, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); // Simulate controller clearing node's resource history and attributes pcmk__g_strcat(xpath, "//" XML_CIB_TAG_STATE - "[@" XML_ATTR_UNAME "='", target, "']/" XML_CIB_TAG_LRM, + "[@" PCMK_XA_UNAME "='", target, "']/" XML_CIB_TAG_LRM, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call|cib_scope_local); g_string_truncate(xpath, 0); pcmk__g_strcat(xpath, "//" XML_CIB_TAG_STATE - "[@" XML_ATTR_UNAME "='", target, "']" + "[@" PCMK_XA_UNAME "='", target, "']" "/" XML_TAG_TRANSIENT_NODEATTRS, NULL); fake_cib->cmds->remove(fake_cib, (const char *) xpath->str, NULL, cib_xpath|cib_sync_call|cib_scope_local); free_xml(cib_node); g_string_free(xpath, TRUE); } pcmk__set_graph_action_flags(action, pcmk__graph_action_confirmed); pcmk__update_graph(graph, action); free(target); return pcmk_rc_ok; } enum pcmk__graph_status pcmk__simulate_transition(pcmk_scheduler_t *scheduler, cib_t *cib, const GList *op_fail_list) { pcmk__graph_t *transition = NULL; enum pcmk__graph_status graph_rc; pcmk__graph_functions_t simulation_fns = { simulate_pseudo_action, simulate_resource_action, simulate_cluster_action, simulate_fencing_action, }; out = scheduler->priv; fake_cib = cib; fake_op_fail_list = op_fail_list; if (!out->is_quiet(out)) { out->begin_list(out, NULL, NULL, "Executing Cluster Transition"); } pcmk__set_graph_functions(&simulation_fns); transition = pcmk__unpack_graph(scheduler->graph, crm_system_name); pcmk__log_graph(LOG_DEBUG, transition); fake_resource_list = scheduler->resources; do { graph_rc = pcmk__execute_graph(transition); } while (graph_rc == pcmk__graph_active); fake_resource_list = NULL; if (graph_rc != pcmk__graph_complete) { out->err(out, "Transition failed: %s", pcmk__graph_status2text(graph_rc)); pcmk__log_graph(LOG_ERR, transition); out->err(out, "An invalid transition was produced"); } pcmk__free_graph(transition); if (!out->is_quiet(out)) { // If not quiet, we'll need the resulting CIB for later display xmlNode *cib_object = NULL; int rc = fake_cib->cmds->query(fake_cib, NULL, &cib_object, cib_sync_call|cib_scope_local); CRM_ASSERT(rc == pcmk_ok); pe_reset_working_set(scheduler); scheduler->input = cib_object; out->end_list(out); } return graph_rc; } int pcmk__simulate(pcmk_scheduler_t *scheduler, pcmk__output_t *out, const pcmk_injections_t *injections, unsigned int flags, uint32_t section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { int printed = pcmk_rc_no_output; int rc = pcmk_rc_ok; xmlNodePtr input = NULL; cib_t *cib = NULL; rc = cib__signon_query(out, &cib, &input); if (rc != pcmk_rc_ok) { goto simulate_done; } reset(scheduler, input, out, use_date, flags); cluster_status(scheduler); if ((cib->variant == cib_native) && pcmk_is_set(section_opts, pcmk_section_times)) { if (pcmk__our_nodename == NULL) { // Currently used only in the times section pcmk__query_node_name(out, 0, &pcmk__our_nodename, 0); } scheduler->localhost = pcmk__our_nodename; } if (!out->is_quiet(out)) { const bool show_pending = pcmk_is_set(flags, pcmk_sim_show_pending); if (pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) { printed = out->message(out, "maint-mode", scheduler->flags); } if (scheduler->disabled_resources || scheduler->blocked_resources) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); printed = out->info(out, "%d of %d resource instances DISABLED and " "%d BLOCKED from further action due to failure", scheduler->disabled_resources, scheduler->ninstances, scheduler->blocked_resources); } /* Most formatted output headers use caps for each word, but this one * only has the first word capitalized for compatibility with pcs. */ print_cluster_status(scheduler, (show_pending? pcmk_show_pending : 0), section_opts, "Current cluster status", (printed == pcmk_rc_ok)); printed = pcmk_rc_ok; } // If the user requested any injections, handle them if ((injections->node_down != NULL) || (injections->node_fail != NULL) || (injections->node_up != NULL) || (injections->op_inject != NULL) || (injections->ticket_activate != NULL) || (injections->ticket_grant != NULL) || (injections->ticket_revoke != NULL) || (injections->ticket_standby != NULL) || (injections->watchdog != NULL)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); pcmk__inject_scheduler_input(scheduler, cib, injections); printed = pcmk_rc_ok; rc = cib->cmds->query(cib, NULL, &input, cib_sync_call); if (rc != pcmk_rc_ok) { rc = pcmk_legacy2rc(rc); goto simulate_done; } cleanup_calculations(scheduler); reset(scheduler, input, out, use_date, flags); cluster_status(scheduler); } if (input_file != NULL) { rc = write_xml_file(input, input_file, FALSE); if (rc < 0) { rc = pcmk_legacy2rc(rc); goto simulate_done; } } if (pcmk_any_flags_set(flags, pcmk_sim_process | pcmk_sim_simulate)) { pcmk__output_t *logger_out = NULL; unsigned long long scheduler_flags = pcmk_sched_no_compat; if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) { scheduler_flags |= pcmk_sched_output_scores; } if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) { scheduler_flags |= pcmk_sched_show_utilization; } if (pcmk_all_flags_set(scheduler->flags, pcmk_sched_output_scores |pcmk_sched_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Assignment Scores and Utilization Information"); printed = pcmk_rc_ok; } else if (pcmk_is_set(scheduler->flags, pcmk_sched_output_scores)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Assignment Scores"); printed = pcmk_rc_ok; } else if (pcmk_is_set(scheduler->flags, pcmk_sched_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Utilization Information"); printed = pcmk_rc_ok; } else { rc = pcmk__log_output_new(&logger_out); if (rc != pcmk_rc_ok) { goto simulate_done; } pe__register_messages(logger_out); pcmk__register_lib_messages(logger_out); scheduler->priv = logger_out; } pcmk__schedule_actions(input, scheduler_flags, scheduler); if (logger_out == NULL) { out->end_list(out); } else { logger_out->finish(logger_out, CRM_EX_OK, true, NULL); pcmk__output_free(logger_out); scheduler->priv = out; } input = NULL; /* Don't try and free it twice */ if (graph_file != NULL) { rc = write_xml_file(scheduler->graph, graph_file, FALSE); if (rc < 0) { rc = pcmk_rc_graph_error; goto simulate_done; } } if (dot_file != NULL) { rc = write_sim_dotfile(scheduler, dot_file, pcmk_is_set(flags, pcmk_sim_all_actions), pcmk_is_set(flags, pcmk_sim_verbose)); if (rc != pcmk_rc_ok) { rc = pcmk_rc_dot_error; goto simulate_done; } } if (!out->is_quiet(out)) { print_transition_summary(scheduler, printed == pcmk_rc_ok); } } rc = pcmk_rc_ok; if (!pcmk_is_set(flags, pcmk_sim_simulate)) { goto simulate_done; } PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); if (pcmk__simulate_transition(scheduler, cib, injections->op_fail) != pcmk__graph_complete) { rc = pcmk_rc_invalid_transition; } if (out->is_quiet(out)) { goto simulate_done; } set_effective_date(scheduler, true, use_date); if (pcmk_is_set(flags, pcmk_sim_show_scores)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_output_scores); } if (pcmk_is_set(flags, pcmk_sim_show_utilization)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_show_utilization); } cluster_status(scheduler); print_cluster_status(scheduler, 0, section_opts, "Revised Cluster Status", true); simulate_done: cib__clean_up_connection(&cib); return rc; } int pcmk_simulate(xmlNodePtr *xml, pcmk_scheduler_t *scheduler, const pcmk_injections_t *injections, unsigned int flags, unsigned int section_opts, const char *use_date, const char *input_file, const char *graph_file, const char *dot_file) { pcmk__output_t *out = NULL; int rc = pcmk_rc_ok; rc = pcmk__xml_output_new(&out, xml); if (rc != pcmk_rc_ok) { return rc; } pe__register_messages(out); pcmk__register_lib_messages(out); rc = pcmk__simulate(scheduler, out, injections, flags, section_opts, use_date, input_file, graph_file, dot_file); pcmk__xml_output_finish(out, pcmk_rc2exitc(rc), xml); return rc; } diff --git a/lib/pengine/failcounts.c b/lib/pengine/failcounts.c index 5c0e112db6..2bf700b1a7 100644 --- a/lib/pengine/failcounts.c +++ b/lib/pengine/failcounts.c @@ -1,471 +1,471 @@ /* * Copyright 2008-2024 the Pacemaker project contributors * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include static gboolean is_matched_failure(const char *rsc_id, const xmlNode *conf_op_xml, const xmlNode *lrm_op_xml) { gboolean matched = FALSE; const char *conf_op_name = NULL; const char *lrm_op_task = NULL; const char *conf_op_interval_spec = NULL; guint conf_op_interval_ms = 0; guint lrm_op_interval_ms = 0; const char *lrm_op_id = NULL; char *last_failure_key = NULL; if (rsc_id == NULL || conf_op_xml == NULL || lrm_op_xml == NULL) { return FALSE; } // Get name and interval from configured op conf_op_name = crm_element_value(conf_op_xml, PCMK_XA_NAME); conf_op_interval_spec = crm_element_value(conf_op_xml, XML_LRM_ATTR_INTERVAL); pcmk_parse_interval_spec(conf_op_interval_spec, &conf_op_interval_ms); // Get name and interval from op history entry lrm_op_task = crm_element_value(lrm_op_xml, XML_LRM_ATTR_TASK); crm_element_value_ms(lrm_op_xml, XML_LRM_ATTR_INTERVAL_MS, &lrm_op_interval_ms); if ((conf_op_interval_ms != lrm_op_interval_ms) || !pcmk__str_eq(conf_op_name, lrm_op_task, pcmk__str_casei)) { return FALSE; } lrm_op_id = ID(lrm_op_xml); last_failure_key = pcmk__op_key(rsc_id, "last_failure", 0); if (pcmk__str_eq(last_failure_key, lrm_op_id, pcmk__str_casei)) { matched = TRUE; } else { char *expected_op_key = pcmk__op_key(rsc_id, conf_op_name, conf_op_interval_ms); if (pcmk__str_eq(expected_op_key, lrm_op_id, pcmk__str_casei)) { int rc = 0; int target_rc = pe__target_rc_from_xml(lrm_op_xml); crm_element_value_int(lrm_op_xml, XML_LRM_ATTR_RC, &rc); if (rc != target_rc) { matched = TRUE; } } free(expected_op_key); } free(last_failure_key); return matched; } static gboolean block_failure(const pcmk_node_t *node, pcmk_resource_t *rsc, const xmlNode *xml_op) { char *xml_name = clone_strip(rsc->id); /* @TODO This xpath search occurs after template expansion, but it is unable * to properly detect on-fail in id-ref, operation meta-attributes, or * op_defaults, or evaluate rules. * * Also, on-fail defaults to block (in unpack_operation()) for stop actions * when stonith is disabled. * * Ideally, we'd unpack the operation before this point, and pass in a * meta-attributes table that takes all that into consideration. */ char *xpath = crm_strdup_printf("//" XML_CIB_TAG_RESOURCE "[@" PCMK_XA_ID "='%s']" "//" PCMK_XE_OP "[@" XML_OP_ATTR_ON_FAIL "='block']", xml_name); xmlXPathObject *xpathObj = xpath_search(rsc->xml, xpath); gboolean should_block = FALSE; free(xpath); if (xpathObj) { int max = numXpathResults(xpathObj); int lpc = 0; for (lpc = 0; lpc < max; lpc++) { xmlNode *pref = getXpathResult(xpathObj, lpc); if (xml_op) { should_block = is_matched_failure(xml_name, pref, xml_op); if (should_block) { break; } } else { const char *conf_op_name = NULL; const char *conf_op_interval_spec = NULL; guint conf_op_interval_ms = 0; char *lrm_op_xpath = NULL; xmlXPathObject *lrm_op_xpathObj = NULL; // Get name and interval from configured op conf_op_name = crm_element_value(pref, PCMK_XA_NAME); conf_op_interval_spec = crm_element_value(pref, XML_LRM_ATTR_INTERVAL); pcmk_parse_interval_spec(conf_op_interval_spec, &conf_op_interval_ms); -#define XPATH_FMT "//" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ +#define XPATH_FMT "//" XML_CIB_TAG_STATE "[@" PCMK_XA_UNAME "='%s']" \ "//" XML_LRM_TAG_RESOURCE "[@" PCMK_XA_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP "[@" XML_LRM_ATTR_TASK "='%s']" \ "[@" XML_LRM_ATTR_INTERVAL "='%u']" lrm_op_xpath = crm_strdup_printf(XPATH_FMT, node->details->uname, xml_name, conf_op_name, conf_op_interval_ms); lrm_op_xpathObj = xpath_search(rsc->cluster->input, lrm_op_xpath); free(lrm_op_xpath); if (lrm_op_xpathObj) { int max2 = numXpathResults(lrm_op_xpathObj); int lpc2 = 0; for (lpc2 = 0; lpc2 < max2; lpc2++) { xmlNode *lrm_op_xml = getXpathResult(lrm_op_xpathObj, lpc2); should_block = is_matched_failure(xml_name, pref, lrm_op_xml); if (should_block) { break; } } } freeXpathObject(lrm_op_xpathObj); if (should_block) { break; } } } } free(xml_name); freeXpathObject(xpathObj); return should_block; } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(const pcmk_resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); return pcmk_is_set(rsc->flags, pcmk_rsc_unique)? strdup(name) : clone_strip(name); } /*! * \internal * \brief Compile regular expression to match a failure-related node attribute * * \param[in] prefix Attribute prefix to match * \param[in] rsc_name Resource name to match as used in failure attributes * \param[in] is_legacy Whether DC uses per-resource fail counts * \param[in] is_unique Whether the resource is a globally unique clone * \param[out] re Where to store resulting regular expression * * \return Standard Pacemaker return code * \note Fail attributes are named like PREFIX-RESOURCE#OP_INTERVAL. * The caller is responsible for freeing re with regfree(). */ static int generate_fail_regex(const char *prefix, const char *rsc_name, gboolean is_legacy, gboolean is_unique, regex_t *re) { char *pattern; /* @COMPAT DC < 1.1.17: Fail counts used to be per-resource rather than * per-operation. */ const char *op_pattern = (is_legacy? "" : "#.+_[0-9]+"); /* Ignore instance numbers for anything other than globally unique clones. * Anonymous clone fail counts could contain an instance number if the * clone was initially unique, failed, then was converted to anonymous. * @COMPAT Also, before 1.1.8, anonymous clone fail counts always contained * clone instance numbers. */ const char *instance_pattern = (is_unique? "" : "(:[0-9]+)?"); pattern = crm_strdup_printf("^%s-%s%s%s$", prefix, rsc_name, instance_pattern, op_pattern); if (regcomp(re, pattern, REG_EXTENDED|REG_NOSUB) != 0) { free(pattern); return EINVAL; } free(pattern); return pcmk_rc_ok; } /*! * \internal * \brief Compile regular expressions to match failure-related node attributes * * \param[in] rsc Resource being checked for failures * \param[out] failcount_re Storage for regular expression for fail count * \param[out] lastfailure_re Storage for regular expression for last failure * * \return Standard Pacemaker return code * \note On success, the caller is responsible for freeing the expressions with * regfree(). */ static int generate_fail_regexes(const pcmk_resource_t *rsc, regex_t *failcount_re, regex_t *lastfailure_re) { int rc = pcmk_rc_ok; char *rsc_name = rsc_fail_name(rsc); const char *version = crm_element_value(rsc->cluster->input, PCMK_XA_CRM_FEATURE_SET); // @COMPAT Pacemaker <= 1.1.16 used a single fail count per resource gboolean is_legacy = (compare_version(version, "3.0.13") < 0); if (generate_fail_regex(PCMK__FAIL_COUNT_PREFIX, rsc_name, is_legacy, pcmk_is_set(rsc->flags, pcmk_rsc_unique), failcount_re) != pcmk_rc_ok) { rc = EINVAL; } else if (generate_fail_regex(PCMK__LAST_FAILURE_PREFIX, rsc_name, is_legacy, pcmk_is_set(rsc->flags, pcmk_rsc_unique), lastfailure_re) != pcmk_rc_ok) { rc = EINVAL; regfree(failcount_re); } free(rsc_name); return rc; } // Data for fail-count-related iterators struct failcount_data { const pcmk_node_t *node;// Node to check for fail count pcmk_resource_t *rsc; // Resource to check for fail count uint32_t flags; // Fail count flags const xmlNode *xml_op; // History entry for expiration purposes (or NULL) regex_t failcount_re; // Fail count regular expression to match regex_t lastfailure_re; // Last failure regular expression to match int failcount; // Fail count so far time_t last_failure; // Time of most recent failure so far }; /*! * \internal * \brief Update fail count and last failure appropriately for a node attribute * * \param[in] key Node attribute name * \param[in] value Node attribute value * \param[in] user_data Fail count data to update */ static void update_failcount_for_attr(gpointer key, gpointer value, gpointer user_data) { struct failcount_data *fc_data = user_data; // If this is a matching fail count attribute, update fail count if (regexec(&(fc_data->failcount_re), (const char *) key, 0, NULL, 0) == 0) { fc_data->failcount = pcmk__add_scores(fc_data->failcount, char2score(value)); pcmk__rsc_trace(fc_data->rsc, "Added %s (%s) to %s fail count (now %s)", (const char *) key, (const char *) value, fc_data->rsc->id, pcmk_readable_score(fc_data->failcount)); return; } // If this is a matching last failure attribute, update last failure if (regexec(&(fc_data->lastfailure_re), (const char *) key, 0, NULL, 0) == 0) { long long last_ll; if (pcmk__scan_ll(value, &last_ll, 0LL) == pcmk_rc_ok) { fc_data->last_failure = (time_t) QB_MAX(fc_data->last_failure, last_ll); } } } /*! * \internal * \brief Update fail count and last failure appropriately for a filler resource * * \param[in] data Filler resource * \param[in] user_data Fail count data to update */ static void update_failcount_for_filler(gpointer data, gpointer user_data) { pcmk_resource_t *filler = data; struct failcount_data *fc_data = user_data; time_t filler_last_failure = 0; fc_data->failcount += pe_get_failcount(fc_data->node, filler, &filler_last_failure, fc_data->flags, fc_data->xml_op); fc_data->last_failure = QB_MAX(fc_data->last_failure, filler_last_failure); } /*! * \internal * \brief Get a resource's fail count on a node * * \param[in] node Node to check * \param[in,out] rsc Resource to check * \param[out] last_failure If not NULL, where to set time of most recent * failure of \p rsc on \p node * \param[in] flags Group of enum pcmk__fc_flags * \param[in] xml_op If not NULL, consider only the action in this * history entry when determining whether on-fail * is configured as "blocked", otherwise consider * all actions configured for \p rsc * * \return Fail count for \p rsc on \p node according to \p flags */ int pe_get_failcount(const pcmk_node_t *node, pcmk_resource_t *rsc, time_t *last_failure, uint32_t flags, const xmlNode *xml_op) { struct failcount_data fc_data = { .node = node, .rsc = rsc, .flags = flags, .xml_op = xml_op, .failcount = 0, .last_failure = (time_t) 0, }; // Calculate resource failcount as sum of all matching operation failcounts CRM_CHECK(generate_fail_regexes(rsc, &fc_data.failcount_re, &fc_data.lastfailure_re) == pcmk_rc_ok, return 0); g_hash_table_foreach(node->details->attrs, update_failcount_for_attr, &fc_data); regfree(&(fc_data.failcount_re)); regfree(&(fc_data.lastfailure_re)); // If failure blocks the resource, disregard any failure timeout if ((fc_data.failcount > 0) && (rsc->failure_timeout > 0) && block_failure(node, rsc, xml_op)) { pcmk__config_warn("Ignoring failure timeout %d for %s " "because it conflicts with on-fail=block", rsc->failure_timeout, rsc->id); rsc->failure_timeout = 0; } // If all failures have expired, ignore fail count if (pcmk_is_set(flags, pcmk__fc_effective) && (fc_data.failcount > 0) && (fc_data.last_failure > 0) && (rsc->failure_timeout != 0)) { time_t now = get_effective_time(rsc->cluster); if (now > (fc_data.last_failure + rsc->failure_timeout)) { pcmk__rsc_debug(rsc, "Failcount for %s on %s expired after %ds", rsc->id, pe__node_name(node), rsc->failure_timeout); fc_data.failcount = 0; } } /* Add the fail count of any filler resources, except that we never want the * fail counts of a bundle container's fillers to count towards the * container's fail count. * * Most importantly, a Pacemaker Remote connection to a bundle container * is a filler of the container, but can reside on a different node than the * container itself. Counting its fail count on its node towards the * container's fail count on that node could lead to attempting to stop the * container on the wrong node. */ if (pcmk_is_set(flags, pcmk__fc_fillers) && (rsc->fillers != NULL) && !pe_rsc_is_bundled(rsc)) { g_list_foreach(rsc->fillers, update_failcount_for_filler, &fc_data); if (fc_data.failcount > 0) { pcmk__rsc_info(rsc, "Container %s and the resources within it " "have failed %s time%s on %s", rsc->id, pcmk_readable_score(fc_data.failcount), pcmk__plural_s(fc_data.failcount), pe__node_name(node)); } } else if (fc_data.failcount > 0) { pcmk__rsc_info(rsc, "%s has failed %s time%s on %s", rsc->id, pcmk_readable_score(fc_data.failcount), pcmk__plural_s(fc_data.failcount), pe__node_name(node)); } if (last_failure != NULL) { if ((fc_data.failcount > 0) && (fc_data.last_failure > 0)) { *last_failure = fc_data.last_failure; } else { *last_failure = 0; } } return fc_data.failcount; } /*! * \brief Schedule a controller operation to clear a fail count * * \param[in,out] rsc Resource with failure * \param[in] node Node failure occurred on * \param[in] reason Readable description why needed (for logging) * \param[in,out] scheduler Scheduler data cluster * * \return Scheduled action */ pcmk_action_t * pe__clear_failcount(pcmk_resource_t *rsc, const pcmk_node_t *node, const char *reason, pcmk_scheduler_t *scheduler) { char *key = NULL; pcmk_action_t *clear = NULL; CRM_CHECK(rsc && node && reason && scheduler, return NULL); key = pcmk__op_key(rsc->id, PCMK_ACTION_CLEAR_FAILCOUNT, 0); clear = custom_action(rsc, key, PCMK_ACTION_CLEAR_FAILCOUNT, node, FALSE, scheduler); add_hash_param(clear->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); crm_notice("Clearing failure of %s on %s because %s " CRM_XS " %s", rsc->id, pe__node_name(node), reason, clear->uuid); return clear; } diff --git a/lib/pengine/pe_output.c b/lib/pengine/pe_output.c index 6847c634fa..b8dfa4be70 100644 --- a/lib/pengine/pe_output.c +++ b/lib/pengine/pe_output.c @@ -1,3179 +1,3182 @@ /* * Copyright 2019-2024 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 const char * pe__resource_description(const pcmk_resource_t *rsc, uint32_t show_opts) { const char * desc = NULL; // User-supplied description if (pcmk_any_flags_set(show_opts, pcmk_show_rsc_only|pcmk_show_description)) { desc = crm_element_value(rsc->xml, PCMK_XA_DESCRIPTION); } return desc; } /* Never display node attributes whose name starts with one of these prefixes */ #define FILTER_STR { PCMK__FAIL_COUNT_PREFIX, PCMK__LAST_FAILURE_PREFIX, \ "shutdown", PCMK_NODE_ATTR_TERMINATE, "standby", "#", \ NULL } static int compare_attribute(gconstpointer a, gconstpointer b) { int rc; rc = strcmp((const char *)a, (const char *)b); return rc; } /*! * \internal * \brief Determine whether extended information about an attribute should be added. * * \param[in] node Node that ran this resource * \param[in,out] rsc_list List of resources for this node * \param[in,out] scheduler Scheduler data * \param[in] attrname Attribute to find * \param[out] expected_score Expected value for this attribute * * \return true if extended information should be printed, false otherwise * \note Currently, extended information is only supported for ping/pingd * resources, for which a message will be printed if connectivity is lost * or degraded. */ static bool add_extra_info(const pcmk_node_t *node, GList *rsc_list, pcmk_scheduler_t *scheduler, const char *attrname, int *expected_score) { GList *gIter = NULL; for (gIter = rsc_list; gIter != NULL; gIter = gIter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter->data; const char *type = g_hash_table_lookup(rsc->meta, PCMK_XA_TYPE); const char *name = NULL; GHashTable *params = NULL; if (rsc->children != NULL) { if (add_extra_info(node, rsc->children, scheduler, attrname, expected_score)) { return true; } } if (!pcmk__strcase_any_of(type, "ping", "pingd", NULL)) { continue; } params = pe_rsc_params(rsc, node, scheduler); name = g_hash_table_lookup(params, PCMK_XA_NAME); if (name == NULL) { name = "pingd"; } /* To identify the resource with the attribute name. */ if (pcmk__str_eq(name, attrname, pcmk__str_casei)) { int host_list_num = 0; const char *hosts = g_hash_table_lookup(params, "host_list"); const char *multiplier = g_hash_table_lookup(params, "multiplier"); int multiplier_i; if (hosts) { char **host_list = g_strsplit(hosts, " ", 0); host_list_num = g_strv_length(host_list); g_strfreev(host_list); } if ((multiplier == NULL) || (pcmk__scan_min_int(multiplier, &multiplier_i, INT_MIN) != pcmk_rc_ok)) { /* The ocf:pacemaker:ping resource agent defaults multiplier to * 1. The agent currently does not handle invalid text, but it * should, and this would be a reasonable choice ... */ multiplier_i = 1; } *expected_score = host_list_num * multiplier_i; return true; } } return false; } static GList * filter_attr_list(GList *attr_list, char *name) { int i; const char *filt_str[] = FILTER_STR; CRM_CHECK(name != NULL, return attr_list); /* filtering automatic attributes */ for (i = 0; filt_str[i] != NULL; i++) { if (g_str_has_prefix(name, filt_str[i])) { return attr_list; } } return g_list_insert_sorted(attr_list, name, compare_attribute); } static GList * get_operation_list(xmlNode *rsc_entry) { GList *op_list = NULL; xmlNode *rsc_op = NULL; for (rsc_op = pcmk__xe_first_child(rsc_entry); rsc_op != NULL; rsc_op = pcmk__xe_next(rsc_op)) { const char *task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(rsc_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC); int op_rc_i; pcmk__scan_min_int(op_rc, &op_rc_i, 0); /* Display 0-interval monitors as "probe" */ if (pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei) && pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches | pcmk__str_casei)) { task = "probe"; } /* Ignore notifies and some probes */ if (pcmk__str_eq(task, PCMK_ACTION_NOTIFY, pcmk__str_none) || (pcmk__str_eq(task, "probe", pcmk__str_none) && (op_rc_i == CRM_EX_NOT_RUNNING))) { continue; } if (pcmk__str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, pcmk__str_none)) { op_list = g_list_append(op_list, rsc_op); } } op_list = g_list_sort(op_list, sort_op_by_callid); return op_list; } static void add_dump_node(gpointer key, gpointer value, gpointer user_data) { xmlNodePtr node = user_data; pcmk_create_xml_text_node(node, (const char *) key, (const char *) value); } static void append_dump_text(gpointer key, gpointer value, gpointer user_data) { char **dump_text = user_data; char *new_text = crm_strdup_printf("%s %s=%s", *dump_text, (char *)key, (char *)value); free(*dump_text); *dump_text = new_text; } #define XPATH_STACK "//" XML_CIB_TAG_NVPAIR \ "[@" PCMK_XA_NAME "='" \ PCMK_OPT_CLUSTER_INFRASTRUCTURE "']" static const char * get_cluster_stack(pcmk_scheduler_t *scheduler) { xmlNode *stack = get_xpath_object(XPATH_STACK, scheduler->input, LOG_DEBUG); return (stack != NULL)? crm_element_value(stack, PCMK_XA_VALUE) : "unknown"; } static char * last_changed_string(const char *last_written, const char *user, const char *client, const char *origin) { if (last_written != NULL || user != NULL || client != NULL || origin != NULL) { return crm_strdup_printf("%s%s%s%s%s%s%s", last_written ? last_written : "", user ? " by " : "", user ? user : "", client ? " via " : "", client ? client : "", origin ? " on " : "", origin ? origin : ""); } else { return strdup(""); } } static char * op_history_string(xmlNode *xml_op, const char *task, const char *interval_ms_s, int rc, bool print_timing) { const char *call = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); char *interval_str = NULL; char *buf = NULL; if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *pair = pcmk__format_nvpair("interval", interval_ms_s, "ms"); interval_str = crm_strdup_printf(" %s", pair); free(pair); } if (print_timing) { char *last_change_str = NULL; char *exec_str = NULL; char *queue_str = NULL; const char *value = NULL; time_t epoch = 0; if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { char *epoch_str = pcmk__epoch2str(&epoch, 0); last_change_str = crm_strdup_printf(" %s=\"%s\"", XML_RSC_OP_LAST_CHANGE, pcmk__s(epoch_str, "")); free(epoch_str); } value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); if (value) { char *pair = pcmk__format_nvpair(XML_RSC_OP_T_EXEC, value, "ms"); exec_str = crm_strdup_printf(" %s", pair); free(pair); } value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); if (value) { char *pair = pcmk__format_nvpair(XML_RSC_OP_T_QUEUE, value, "ms"); queue_str = crm_strdup_printf(" %s", pair); free(pair); } buf = crm_strdup_printf("(%s) %s:%s%s%s%s rc=%d (%s)", call, task, interval_str ? interval_str : "", last_change_str ? last_change_str : "", exec_str ? exec_str : "", queue_str ? queue_str : "", rc, services_ocf_exitcode_str(rc)); if (last_change_str) { free(last_change_str); } if (exec_str) { free(exec_str); } if (queue_str) { free(queue_str); } } else { buf = crm_strdup_printf("(%s) %s%s%s", call, task, interval_str ? ":" : "", interval_str ? interval_str : ""); } if (interval_str) { free(interval_str); } return buf; } static char * resource_history_string(pcmk_resource_t *rsc, const char *rsc_id, bool all, int failcount, time_t last_failure) { char *buf = NULL; if (rsc == NULL) { buf = crm_strdup_printf("%s: orphan", rsc_id); } else if (all || failcount || last_failure > 0) { char *failcount_s = NULL; char *lastfail_s = NULL; if (failcount > 0) { failcount_s = crm_strdup_printf(" %s=%d", PCMK__FAIL_COUNT_PREFIX, failcount); } else { failcount_s = strdup(""); } if (last_failure > 0) { buf = pcmk__epoch2str(&last_failure, 0); lastfail_s = crm_strdup_printf(" %s='%s'", PCMK__LAST_FAILURE_PREFIX, buf); free(buf); } buf = crm_strdup_printf("%s: " PCMK_META_MIGRATION_THRESHOLD "=%d%s%s", rsc_id, rsc->migration_threshold, failcount_s, lastfail_s? lastfail_s : ""); free(failcount_s); free(lastfail_s); } else { buf = crm_strdup_printf("%s:", rsc_id); } return buf; } /*! * \internal * \brief Get a node's feature set for status display purposes * * \param[in] node Node to check * * \return String representation of feature set if the node is fully up (using * "<3.15.1" for older nodes that don't set the #feature-set attribute), * otherwise NULL */ static const char * get_node_feature_set(const pcmk_node_t *node) { if (node->details->online && node->details->expected_up && !pe__is_guest_or_remote_node(node)) { const char *feature_set = g_hash_table_lookup(node->details->attrs, CRM_ATTR_FEATURE_SET); /* The feature set attribute is present since 3.15.1. If it is missing, * then the node must be running an earlier version. */ return pcmk__s(feature_set, "<3.15.1"); } return NULL; } static bool is_mixed_version(pcmk_scheduler_t *scheduler) { const char *feature_set = NULL; for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = gIter->data; const char *node_feature_set = get_node_feature_set(node); if (node_feature_set != NULL) { if (feature_set == NULL) { feature_set = node_feature_set; } else if (strcmp(feature_set, node_feature_set) != 0) { return true; } } } return false; } static char * formatted_xml_buf(const pcmk_resource_t *rsc, bool raw) { if (raw) { return dump_xml_formatted(rsc->orig_xml ? rsc->orig_xml : rsc->xml); } else { return dump_xml_formatted(rsc->xml); } } #define XPATH_DC_VERSION "//" XML_CIB_TAG_NVPAIR \ "[@" PCMK_XA_NAME "='" PCMK_OPT_DC_VERSION "']" PCMK__OUTPUT_ARGS("cluster-summary", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "uint32_t", "uint32_t") static int cluster_summary(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(scheduler); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s, pcmkd_state); } if (pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = get_xpath_object(XPATH_DC_VERSION, scheduler->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, PCMK_XA_VALUE) : NULL; const char *quorum = crm_element_value(scheduler->input, PCMK_XA_HAVE_QUORUM); char *dc_name = scheduler->dc_node? pe__node_display_name(scheduler->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; bool mixed_version = is_mixed_version(scheduler); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-dc", scheduler->dc_node, quorum, dc_version_s, dc_name, mixed_version); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = crm_element_value(scheduler->input, PCMK_XA_CIB_LAST_WRITTEN); const char *user = crm_element_value(scheduler->input, PCMK_XA_UPDATE_USER); const char *client = crm_element_value(scheduler->input, PCMK_XA_UPDATE_CLIENT); const char *origin = crm_element_value(scheduler->input, PCMK_XA_UPDATE_ORIGIN); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-times", scheduler->localhost, last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(scheduler->nodes), scheduler->ninstances, scheduler->disabled_resources, scheduler->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-options", scheduler); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", scheduler->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } PCMK__OUTPUT_ARGS("cluster-summary", "pcmk_scheduler_t *", "enum pcmk_pacemakerd_state", "uint32_t", "uint32_t") static int cluster_summary_html(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); int rc = pcmk_rc_no_output; const char *stack_s = get_cluster_stack(scheduler); if (pcmk_is_set(section_opts, pcmk_section_stack)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-stack", stack_s, pcmkd_state); } /* Always print DC if none, even if not requested */ if ((scheduler->dc_node == NULL) || pcmk_is_set(section_opts, pcmk_section_dc)) { xmlNode *dc_version = get_xpath_object(XPATH_DC_VERSION, scheduler->input, LOG_DEBUG); const char *dc_version_s = dc_version? crm_element_value(dc_version, PCMK_XA_VALUE) : NULL; const char *quorum = crm_element_value(scheduler->input, PCMK_XA_HAVE_QUORUM); char *dc_name = scheduler->dc_node? pe__node_display_name(scheduler->dc_node, pcmk_is_set(show_opts, pcmk_show_node_id)) : NULL; bool mixed_version = is_mixed_version(scheduler); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-dc", scheduler->dc_node, quorum, dc_version_s, dc_name, mixed_version); free(dc_name); } if (pcmk_is_set(section_opts, pcmk_section_times)) { const char *last_written = crm_element_value(scheduler->input, PCMK_XA_CIB_LAST_WRITTEN); const char *user = crm_element_value(scheduler->input, PCMK_XA_UPDATE_USER); const char *client = crm_element_value(scheduler->input, PCMK_XA_UPDATE_CLIENT); const char *origin = crm_element_value(scheduler->input, PCMK_XA_UPDATE_ORIGIN); PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-times", scheduler->localhost, last_written, user, client, origin); } if (pcmk_is_set(section_opts, pcmk_section_counts)) { PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Cluster Summary"); out->message(out, "cluster-counts", g_list_length(scheduler->nodes), scheduler->ninstances, scheduler->disabled_resources, scheduler->blocked_resources); } if (pcmk_is_set(section_opts, pcmk_section_options)) { /* Kind of a hack - close the list we may have opened earlier in this * function so we can put all the options into their own list. We * only want to do this on HTML output, though. */ PCMK__OUTPUT_LIST_FOOTER(out, rc); out->begin_list(out, NULL, NULL, "Config Options"); out->message(out, "cluster-options", scheduler); } PCMK__OUTPUT_LIST_FOOTER(out, rc); if (pcmk_is_set(section_opts, pcmk_section_maint_mode)) { if (out->message(out, "maint-mode", scheduler->flags) == pcmk_rc_ok) { rc = pcmk_rc_ok; } } return rc; } char * pe__node_display_name(pcmk_node_t *node, bool print_detail) { char *node_name; const char *node_host = NULL; const char *node_id = NULL; int name_len; CRM_ASSERT((node != NULL) && (node->details != NULL) && (node->details->uname != NULL)); /* Host is displayed only if this is a guest node and detail is requested */ if (print_detail && pe__is_guest_node(node)) { const pcmk_resource_t *container = node->details->remote_rsc->container; const pcmk_node_t *host_node = pe__current_node(container); if (host_node && host_node->details) { node_host = host_node->details->uname; } if (node_host == NULL) { node_host = ""; /* so we at least get "uname@" to indicate guest */ } } /* Node ID is displayed if different from uname and detail is requested */ if (print_detail && !pcmk__str_eq(node->details->uname, node->details->id, pcmk__str_casei)) { node_id = node->details->id; } /* Determine name length */ name_len = strlen(node->details->uname) + 1; if (node_host) { name_len += strlen(node_host) + 1; /* "@node_host" */ } if (node_id) { name_len += strlen(node_id) + 3; /* + " (node_id)" */ } /* Allocate and populate display name */ node_name = malloc(name_len); CRM_ASSERT(node_name != NULL); strcpy(node_name, node->details->uname); if (node_host) { strcat(node_name, "@"); strcat(node_name, node_host); } if (node_id) { strcat(node_name, " ("); strcat(node_name, node_id); strcat(node_name, ")"); } return node_name; } int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name , size_t pairs_count, ...) { xmlNodePtr xml_node = NULL; va_list args; CRM_ASSERT(tag_name != NULL); xml_node = pcmk__output_xml_peek_parent(out); CRM_ASSERT(xml_node != NULL); xml_node = create_xml_node(xml_node, tag_name); va_start(args, pairs_count); while(pairs_count--) { const char *param_name = va_arg(args, const char *); const char *param_value = va_arg(args, const char *); if (param_name && param_value) { crm_xml_add(xml_node, param_name, param_value); } }; va_end(args); if (is_list) { pcmk__output_xml_push_parent(out, xml_node); } return pcmk_rc_ok; } static const char * role_desc(enum rsc_role_e role) { if (role == pcmk_role_promoted) { #ifdef PCMK__COMPAT_2_0 return "as " PCMK__ROLE_PROMOTED_LEGACY " "; #else return "in " PCMK__ROLE_PROMOTED " role "; #endif } return ""; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_html(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts = va_arg(args, uint32_t); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); char *buf = crm_strdup_printf("%s\tprevents %s from running %son %s", location->id, location->rsc->id, role_desc(location->role_filter), node_name); pcmk__output_create_html_node(out, "li", NULL, NULL, buf); free(node_name); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_text(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts = va_arg(args, uint32_t); char *node_name = pe__node_display_name(pe_node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->list_item(out, NULL, "%s\tprevents %s from running %son %s", location->id, location->rsc->id, role_desc(location->role_filter), node_name); free(node_name); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban", "pcmk_node_t *", "pcmk__location_t *", "uint32_t") static int ban_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *pe_node = va_arg(args, pcmk_node_t *); pcmk__location_t *location = va_arg(args, pcmk__location_t *); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); const char *promoted_only = pcmk__btoa(location->role_filter == pcmk_role_promoted); char *weight_s = pcmk__itoa(pe_node->weight); pcmk__output_create_xml_node(out, "ban", PCMK_XA_ID, location->id, "resource", location->rsc->id, "node", pe_node->details->uname, "weight", weight_s, "promoted-only", promoted_only, /* This is a deprecated alias for * promoted_only. Removing it will break * backward compatibility of the API schema, * which will require an API schema major * version bump. */ "master_only", promoted_only, NULL); free(weight_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ban-list", "pcmk_scheduler_t *", "const char *", "GList *", "uint32_t", "bool") static int ban_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); const char *prefix = va_arg(args, const char *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); GList *gIter, *gIter2; int rc = pcmk_rc_no_output; /* Print each ban */ for (gIter = scheduler->placement_constraints; gIter != NULL; gIter = gIter->next) { pcmk__location_t *location = gIter->data; const pcmk_resource_t *rsc = location->rsc; if (prefix != NULL && !g_str_has_prefix(location->id, prefix)) { continue; } if (!pcmk__str_in_list(rsc_printable_id(rsc), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(pe__const_top_resource(rsc, false)), only_rsc, pcmk__str_star_matches)) { continue; } for (gIter2 = location->nodes; gIter2 != NULL; gIter2 = gIter2->next) { pcmk_node_t *node = (pcmk_node_t *) gIter2->data; if (node->weight < 0) { PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Negative Location Constraints"); out->message(out, "ban", node, location, show_opts); } } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_html(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "li", NULL); char *nnodes_str = crm_strdup_printf("%d node%s configured", nnodes, pcmk__plural_s(nnodes)); pcmk_create_html_node(nodes_node, "span", NULL, NULL, nnodes_str); free(nnodes_str); if (ndisabled && nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED"); s = crm_strdup_printf(", %d ", nblocked); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, " from further action due to failure)"); } else if (ndisabled && !nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), ndisabled); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "DISABLED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, ")"); } else if (!ndisabled && nblocked) { char *s = crm_strdup_printf("%d resource instance%s configured (%d ", nresources, pcmk__plural_s(nresources), nblocked); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); pcmk_create_html_node(resources_node, "span", NULL, "bold", "BLOCKED"); pcmk_create_html_node(resources_node, "span", NULL, NULL, " from further action due to failure)"); } else { char *s = crm_strdup_printf("%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); pcmk_create_html_node(resources_node, "span", NULL, NULL, s); free(s); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_text(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); out->list_item(out, NULL, "%d node%s configured", nnodes, pcmk__plural_s(nnodes)); if (ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED, %d BLOCKED from " "further action due to failure)", nresources, pcmk__plural_s(nresources), ndisabled, nblocked); } else if (ndisabled && !nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d DISABLED)", nresources, pcmk__plural_s(nresources), ndisabled); } else if (!ndisabled && nblocked) { out->list_item(out, NULL, "%d resource instance%s configured " "(%d BLOCKED from further action " "due to failure)", nresources, pcmk__plural_s(nresources), nblocked); } else { out->list_item(out, NULL, "%d resource instance%s configured", nresources, pcmk__plural_s(nresources)); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-counts", "unsigned int", "int", "int", "int") static int cluster_counts_xml(pcmk__output_t *out, va_list args) { unsigned int nnodes = va_arg(args, unsigned int); int nresources = va_arg(args, int); int ndisabled = va_arg(args, int); int nblocked = va_arg(args, int); xmlNodePtr nodes_node = pcmk__output_create_xml_node(out, "nodes_configured", NULL); xmlNodePtr resources_node = pcmk__output_create_xml_node(out, "resources_configured", NULL); char *s = pcmk__itoa(nnodes); crm_xml_add(nodes_node, "number", s); free(s); s = pcmk__itoa(nresources); crm_xml_add(resources_node, "number", s); free(s); s = pcmk__itoa(ndisabled); crm_xml_add(resources_node, "disabled", s); free(s); s = pcmk__itoa(nblocked); crm_xml_add(resources_node, "blocked", s); free(s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_html(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); bool mixed_version = va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, "bold", "Current DC: "); if (dc) { char *buf = crm_strdup_printf("%s (version %s) -", dc_name, dc_version_s ? dc_version_s : "unknown"); pcmk_create_html_node(node, "span", NULL, NULL, buf); free(buf); if (mixed_version) { pcmk_create_html_node(node, "span", NULL, "warning", " MIXED-VERSION"); } pcmk_create_html_node(node, "span", NULL, NULL, " partition"); if (crm_is_true(quorum)) { pcmk_create_html_node(node, "span", NULL, NULL, " with"); } else { pcmk_create_html_node(node, "span", NULL, "warning", " WITHOUT"); } pcmk_create_html_node(node, "span", NULL, NULL, " quorum"); } else { pcmk_create_html_node(node, "span", NULL, "warning", "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_text(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name = va_arg(args, char *); bool mixed_version = va_arg(args, int); if (dc) { out->list_item(out, "Current DC", "%s (version %s) - %spartition %s quorum", dc_name, dc_version_s ? dc_version_s : "unknown", mixed_version ? "MIXED-VERSION " : "", crm_is_true(quorum) ? "with" : "WITHOUT"); } else { out->list_item(out, "Current DC", "NONE"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-dc", "pcmk_node_t *", "const char *", "const char *", "char *", "int") static int cluster_dc_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *dc = va_arg(args, pcmk_node_t *); const char *quorum = va_arg(args, const char *); const char *dc_version_s = va_arg(args, const char *); char *dc_name G_GNUC_UNUSED = va_arg(args, char *); bool mixed_version = va_arg(args, int); if (dc) { pcmk__output_create_xml_node(out, "current_dc", "present", "true", PCMK_XA_VERSION, pcmk__s(dc_version_s, ""), PCMK_XA_NAME, dc->details->uname, PCMK_XA_ID, dc->details->id, "with_quorum", pcmk__btoa(crm_is_true(quorum)), "mixed_version", pcmk__btoa(mixed_version), NULL); } else { pcmk__output_create_xml_node(out, "current_dc", "present", "false", NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("maint-mode", "unsigned long long int") static int cluster_maint_mode_text(pcmk__output_t *out, va_list args) { unsigned long long flags = va_arg(args, unsigned long long); if (pcmk_is_set(flags, pcmk_sched_in_maintenance)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will not attempt to start, stop or recover services\n"); return pcmk_rc_ok; } else if (pcmk_is_set(flags, pcmk_sched_stop_all)) { pcmk__formatted_printf(out, "\n *** Resource management is DISABLED ***\n"); pcmk__formatted_printf(out, " The cluster will keep all resources stopped\n"); return pcmk_rc_ok; } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_html(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) { out->list_item(out, NULL, "STONITH of failed nodes enabled"); } else { out->list_item(out, NULL, "STONITH of failed nodes disabled"); } if (pcmk_is_set(scheduler->flags, pcmk_sched_symmetric_cluster)) { out->list_item(out, NULL, "Cluster is symmetric"); } else { out->list_item(out, NULL, "Cluster is asymmetric"); } switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case pcmk_no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case pcmk_no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case pcmk_no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case pcmk_no_quorum_fence: out->list_item(out, NULL, "No quorum policy: Suicide"); break; } if (pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, NULL, "Resource management: "); pcmk_create_html_node(node, "span", NULL, "bold", "DISABLED"); pcmk_create_html_node(node, "span", NULL, NULL, " (the cluster will not attempt to start, stop, or recover services)"); } else if (pcmk_is_set(scheduler->flags, pcmk_sched_stop_all)) { xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, NULL, "Resource management: "); pcmk_create_html_node(node, "span", NULL, "bold", "STOPPED"); pcmk_create_html_node(node, "span", NULL, NULL, " (the cluster will keep all resources stopped)"); } else { out->list_item(out, NULL, "Resource management: enabled"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_log(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance)) { return out->info(out, "Resource management is DISABLED. The cluster will not attempt to start, stop or recover services."); } else if (pcmk_is_set(scheduler->flags, pcmk_sched_stop_all)) { return out->info(out, "Resource management is DISABLED. The cluster has stopped all resources."); } else { return pcmk_rc_no_output; } } PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_text(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) { out->list_item(out, NULL, "STONITH of failed nodes enabled"); } else { out->list_item(out, NULL, "STONITH of failed nodes disabled"); } if (pcmk_is_set(scheduler->flags, pcmk_sched_symmetric_cluster)) { out->list_item(out, NULL, "Cluster is symmetric"); } else { out->list_item(out, NULL, "Cluster is asymmetric"); } switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: out->list_item(out, NULL, "No quorum policy: Freeze resources"); break; case pcmk_no_quorum_stop: out->list_item(out, NULL, "No quorum policy: Stop ALL resources"); break; case pcmk_no_quorum_demote: out->list_item(out, NULL, "No quorum policy: Demote promotable " "resources and stop all other resources"); break; case pcmk_no_quorum_ignore: out->list_item(out, NULL, "No quorum policy: Ignore"); break; case pcmk_no_quorum_fence: out->list_item(out, NULL, "No quorum policy: Suicide"); break; } return pcmk_rc_ok; } #define bv(flag) pcmk__btoa(pcmk_is_set(scheduler->flags, (flag))) PCMK__OUTPUT_ARGS("cluster-options", "pcmk_scheduler_t *") static int cluster_options_xml(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); const char *no_quorum_policy = NULL; char *stonith_timeout_str = pcmk__itoa(scheduler->stonith_timeout); char *priority_fencing_delay_str = pcmk__itoa(scheduler->priority_fencing_delay * 1000); switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: no_quorum_policy = "freeze"; break; case pcmk_no_quorum_stop: no_quorum_policy = "stop"; break; case pcmk_no_quorum_demote: no_quorum_policy = "demote"; break; case pcmk_no_quorum_ignore: no_quorum_policy = "ignore"; break; case pcmk_no_quorum_fence: no_quorum_policy = "suicide"; break; } pcmk__output_create_xml_node(out, "cluster_options", PCMK_OPT_STONITH_ENABLED, bv(pcmk_sched_fencing_enabled), PCMK_OPT_SYMMETRIC_CLUSTER, bv(pcmk_sched_symmetric_cluster), PCMK_OPT_NO_QUORUM_POLICY, no_quorum_policy, PCMK_OPT_MAINTENANCE_MODE, bv(pcmk_sched_in_maintenance), PCMK_OPT_STOP_ALL_RESOURCES, bv(pcmk_sched_stop_all), PCMK_OPT_STONITH_TIMEOUT "-ms", stonith_timeout_str, PCMK_OPT_PRIORITY_FENCING_DELAY "-ms", priority_fencing_delay_str, NULL); free(stonith_timeout_str); free(priority_fencing_delay_str); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_html(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(node, "span", NULL, "bold", "Stack: "); pcmk_create_html_node(node, "span", NULL, NULL, stack_s); if (pcmkd_state != pcmk_pacemakerd_state_invalid) { pcmk_create_html_node(node, "span", NULL, NULL, " ("); pcmk_create_html_node(node, "span", NULL, NULL, pcmk__pcmkd_state_enum2friendly(pcmkd_state)); pcmk_create_html_node(node, "span", NULL, NULL, ")"); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_text(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); if (pcmkd_state != pcmk_pacemakerd_state_invalid) { out->list_item(out, "Stack", "%s (%s)", stack_s, pcmk__pcmkd_state_enum2friendly(pcmkd_state)); } else { out->list_item(out, "Stack", "%s", stack_s); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-stack", "const char *", "enum pcmk_pacemakerd_state") static int cluster_stack_xml(pcmk__output_t *out, va_list args) { const char *stack_s = va_arg(args, const char *); enum pcmk_pacemakerd_state pcmkd_state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = NULL; if (pcmkd_state != pcmk_pacemakerd_state_invalid) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(pcmkd_state); } pcmk__output_create_xml_node(out, "stack", PCMK_XA_TYPE, stack_s, "pacemakerd-state", state_s, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_html(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); xmlNodePtr updated_node = pcmk__output_create_xml_node(out, "li", NULL); xmlNodePtr changed_node = pcmk__output_create_xml_node(out, "li", NULL); char *time_s = pcmk__epoch2str(NULL, 0); pcmk_create_html_node(updated_node, "span", NULL, "bold", "Last updated: "); pcmk_create_html_node(updated_node, "span", NULL, NULL, time_s); if (our_nodename != NULL) { pcmk_create_html_node(updated_node, "span", NULL, NULL, " on "); pcmk_create_html_node(updated_node, "span", NULL, NULL, our_nodename); } free(time_s); time_s = last_changed_string(last_written, user, client, origin); pcmk_create_html_node(changed_node, "span", NULL, "bold", "Last change: "); pcmk_create_html_node(changed_node, "span", NULL, NULL, time_s); free(time_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_xml(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); char *time_s = pcmk__epoch2str(NULL, 0); pcmk__output_create_xml_node(out, "last_update", "time", time_s, "origin", our_nodename, NULL); pcmk__output_create_xml_node(out, "last_change", "time", last_written ? last_written : "", "user", user ? user : "", "client", client ? client : "", "origin", origin ? origin : "", NULL); free(time_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("cluster-times", "const char *", "const char *", "const char *", "const char *", "const char *") static int cluster_times_text(pcmk__output_t *out, va_list args) { const char *our_nodename = va_arg(args, const char *); const char *last_written = va_arg(args, const char *); const char *user = va_arg(args, const char *); const char *client = va_arg(args, const char *); const char *origin = va_arg(args, const char *); char *time_s = pcmk__epoch2str(NULL, 0); out->list_item(out, "Last updated", "%s%s%s", time_s, (our_nodename != NULL)? " on " : "", pcmk__s(our_nodename, "")); free(time_s); time_s = last_changed_string(last_written, user, client, origin); out->list_item(out, "Last change", " %s", time_s); free(time_s); return pcmk_rc_ok; } /*! * \internal * \brief Display a failed action in less-technical natural language * * \param[in,out] out Output object to use for display * \param[in] xml_op XML containing failed action * \param[in] op_key Operation key of failed action * \param[in] node_name Where failed action occurred * \param[in] rc OCF exit code of failed action * \param[in] status Execution status of failed action * \param[in] exit_reason Exit reason given for failed action * \param[in] exec_time String containing execution time in milliseconds */ static void failed_action_friendly(pcmk__output_t *out, const xmlNode *xml_op, const char *op_key, const char *node_name, int rc, int status, const char *exit_reason, const char *exec_time) { char *rsc_id = NULL; char *task = NULL; guint interval_ms = 0; time_t last_change_epoch = 0; GString *str = NULL; if (pcmk__str_empty(op_key) || !parse_op_key(op_key, &rsc_id, &task, &interval_ms)) { rsc_id = strdup("unknown resource"); task = strdup("unknown action"); interval_ms = 0; } CRM_ASSERT((rsc_id != NULL) && (task != NULL)); str = g_string_sized_new(256); // Should be sufficient for most messages pcmk__g_strcat(str, rsc_id, " ", NULL); if (interval_ms != 0) { pcmk__g_strcat(str, pcmk__readable_interval(interval_ms), "-interval ", NULL); } pcmk__g_strcat(str, pcmk__readable_action(task, interval_ms), " on ", node_name, NULL); if (status == PCMK_EXEC_DONE) { pcmk__g_strcat(str, " returned '", services_ocf_exitcode_str(rc), "'", NULL); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, " (", exit_reason, ")", NULL); } } else { pcmk__g_strcat(str, " could not be executed (", pcmk_exec_status_str(status), NULL); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, ": ", exit_reason, NULL); } g_string_append_c(str, ')'); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change_epoch) == pcmk_ok) { char *s = pcmk__epoch2str(&last_change_epoch, 0); pcmk__g_strcat(str, " at ", s, NULL); free(s); } if (!pcmk__str_empty(exec_time)) { int exec_time_ms = 0; if ((pcmk__scan_min_int(exec_time, &exec_time_ms, 0) == pcmk_rc_ok) && (exec_time_ms > 0)) { pcmk__g_strcat(str, " after ", pcmk__readable_interval(exec_time_ms), NULL); } } out->list_item(out, NULL, "%s", str->str); g_string_free(str, TRUE); free(rsc_id); free(task); } /*! * \internal * \brief Display a failed action with technical details * * \param[in,out] out Output object to use for display * \param[in] xml_op XML containing failed action * \param[in] op_key Operation key of failed action * \param[in] node_name Where failed action occurred * \param[in] rc OCF exit code of failed action * \param[in] status Execution status of failed action * \param[in] exit_reason Exit reason given for failed action * \param[in] exec_time String containing execution time in milliseconds */ static void failed_action_technical(pcmk__output_t *out, const xmlNode *xml_op, const char *op_key, const char *node_name, int rc, int status, const char *exit_reason, const char *exec_time) { const char *call_id = crm_element_value(xml_op, XML_LRM_ATTR_CALLID); const char *queue_time = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); const char *exit_status = services_ocf_exitcode_str(rc); const char *lrm_status = pcmk_exec_status_str(status); time_t last_change_epoch = 0; GString *str = NULL; if (pcmk__str_empty(op_key)) { op_key = "unknown operation"; } if (pcmk__str_empty(exit_status)) { exit_status = "unknown exit status"; } if (pcmk__str_empty(call_id)) { call_id = "unknown"; } str = g_string_sized_new(256); g_string_append_printf(str, "%s on %s '%s' (%d): call=%s, status='%s'", op_key, node_name, exit_status, rc, call_id, lrm_status); if (!pcmk__str_empty(exit_reason)) { pcmk__g_strcat(str, ", exitreason='", exit_reason, "'", NULL); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change_epoch) == pcmk_ok) { char *last_change_str = pcmk__epoch2str(&last_change_epoch, 0); pcmk__g_strcat(str, ", " XML_RSC_OP_LAST_CHANGE "=" "'", last_change_str, "'", NULL); free(last_change_str); } if (!pcmk__str_empty(queue_time)) { pcmk__g_strcat(str, ", queued=", queue_time, "ms", NULL); } if (!pcmk__str_empty(exec_time)) { pcmk__g_strcat(str, ", exec=", exec_time, "ms", NULL); } out->list_item(out, NULL, "%s", str->str); g_string_free(str, TRUE); } PCMK__OUTPUT_ARGS("failed-action", "xmlNodePtr", "uint32_t") static int failed_action_default(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); uint32_t show_opts = va_arg(args, uint32_t); const char *op_key = pe__xe_history_key(xml_op); - const char *node_name = crm_element_value(xml_op, XML_ATTR_UNAME); + const char *node_name = crm_element_value(xml_op, PCMK_XA_UNAME); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); const char *exec_time = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); int rc; int status; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), &rc, 0); pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); if (pcmk__str_empty(node_name)) { node_name = "unknown node"; } if (pcmk_is_set(show_opts, pcmk_show_failed_detail)) { failed_action_technical(out, xml_op, op_key, node_name, rc, status, exit_reason, exec_time); } else { failed_action_friendly(out, xml_op, op_key, node_name, rc, status, exit_reason, exec_time); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action", "xmlNodePtr", "uint32_t") static int failed_action_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); const char *op_key = pe__xe_history_key(xml_op); const char *op_key_name = "op_key"; int rc; int status; + const char *uname = crm_element_value(xml_op, PCMK_XA_UNAME); const char *exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON); time_t epoch = 0; char *rc_s = NULL; char *reason_s = crm_xml_escape(exit_reason ? exit_reason : "none"); xmlNodePtr node = NULL; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_RC), &rc, 0); pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, 0); rc_s = pcmk__itoa(rc); if (crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY) == NULL) { op_key_name = PCMK_XA_ID; } node = pcmk__output_create_xml_node(out, "failure", op_key_name, op_key, - "node", crm_element_value(xml_op, XML_ATTR_UNAME), + "node", uname, "exitstatus", services_ocf_exitcode_str(rc), "exitreason", pcmk__s(reason_s, ""), "exitcode", rc_s, "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "status", pcmk_exec_status_str(status), NULL); free(rc_s); if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { guint interval_ms = 0; char *interval_ms_s = NULL; char *rc_change = pcmk__epoch2str(&epoch, crm_time_log_date |crm_time_log_timeofday |crm_time_log_with_timezone); crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); interval_ms_s = crm_strdup_printf("%u", interval_ms); pcmk__xe_set_props(node, XML_RSC_OP_LAST_CHANGE, rc_change, "queued", crm_element_value(xml_op, XML_RSC_OP_T_QUEUE), "exec", crm_element_value(xml_op, XML_RSC_OP_T_EXEC), "interval", interval_ms_s, "task", crm_element_value(xml_op, XML_LRM_ATTR_TASK), NULL); free(interval_ms_s); free(rc_change); } free(reason_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("failed-action-list", "pcmk_scheduler_t *", "GList *", "GList *", "uint32_t", "bool") static int failed_action_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); xmlNode *xml_op = NULL; int rc = pcmk_rc_no_output; if (xmlChildElementCount(scheduler->failed) == 0) { return rc; } for (xml_op = pcmk__xml_first_child(scheduler->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { char *rsc = NULL; - if (!pcmk__str_in_list(crm_element_value(xml_op, XML_ATTR_UNAME), only_node, + if (!pcmk__str_in_list(crm_element_value(xml_op, PCMK_XA_UNAME), + only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } if (pcmk_xe_mask_probe_failure(xml_op)) { continue; } if (!parse_op_key(pe__xe_history_key(xml_op), &rsc, NULL, NULL)) { continue; } if (!pcmk__str_in_list(rsc, only_rsc, pcmk__str_star_matches)) { free(rsc); continue; } free(rsc); PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Failed Resource Actions"); out->message(out, "failed-action", xml_op, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } static void status_node(pcmk_node_t *node, xmlNodePtr parent, uint32_t show_opts) { int health = pe__node_health(node); // Cluster membership if (node->details->online) { pcmk_create_html_node(parent, "span", NULL, "online", " online"); } else { pcmk_create_html_node(parent, "span", NULL, "offline", " OFFLINE"); } // Standby mode if (node->details->standby_onfail && (node->details->running_rsc != NULL)) { pcmk_create_html_node(parent, "span", NULL, "standby", " (in standby due to on-fail," " with active resources)"); } else if (node->details->standby_onfail) { pcmk_create_html_node(parent, "span", NULL, "standby", " (in standby due to on-fail)"); } else if (node->details->standby && (node->details->running_rsc != NULL)) { pcmk_create_html_node(parent, "span", NULL, "standby", " (in standby, with active resources)"); } else if (node->details->standby) { pcmk_create_html_node(parent, "span", NULL, "standby", " (in standby)"); } // Maintenance mode if (node->details->maintenance) { pcmk_create_html_node(parent, "span", NULL, "maint", " (in maintenance mode)"); } // Node health if (health < 0) { pcmk_create_html_node(parent, "span", NULL, "health_red", " (health is RED)"); } else if (health == 0) { pcmk_create_html_node(parent, "span", NULL, "health_yellow", " (health is YELLOW)"); } // Feature set if (pcmk_is_set(show_opts, pcmk_show_feature_set)) { const char *feature_set = get_node_feature_set(node); if (feature_set != NULL) { char *buf = crm_strdup_printf(", feature set %s", feature_set); pcmk_create_html_node(parent, "span", NULL, NULL, buf); free(buf); } } } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_html(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (full) { xmlNodePtr item_node; if (pcmk_all_flags_set(show_opts, pcmk_show_brief | pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); out->begin_list(out, NULL, NULL, "%s:", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, NULL, "Status:"); status_node(node, item_node, show_opts); if (rscs != NULL) { uint32_t new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); } pcmk__output_xml_pop_parent(out); out->end_list(out); } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc2 = NULL; int rc = pcmk_rc_no_output; out->begin_list(out, NULL, NULL, "%s:", node_name); item_node = pcmk__output_xml_create_parent(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, NULL, "Status:"); status_node(node, item_node, show_opts); for (lpc2 = node->details->running_rsc; lpc2 != NULL; lpc2 = lpc2->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) lpc2->data; PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Resources"); show_opts |= pcmk_show_rsc_only; out->message(out, crm_map_element_name(rsc->xml), show_opts, rsc, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); pcmk__output_xml_pop_parent(out); out->end_list(out); } else { char *buf = crm_strdup_printf("%s:", node_name); item_node = pcmk__output_create_xml_node(out, "li", NULL); pcmk_create_html_node(item_node, "span", NULL, "bold", buf); status_node(node, item_node, show_opts); free(buf); } } else { out->begin_list(out, NULL, NULL, "%s:", node_name); } free(node_name); return pcmk_rc_ok; } /*! * \internal * \brief Get a human-friendly textual description of a node's status * * \param[in] node Node to check * * \return String representation of node's status */ static const char * node_text_status(const pcmk_node_t *node) { if (node->details->unclean) { if (node->details->online) { return "UNCLEAN (online)"; } else if (node->details->pending) { return "UNCLEAN (pending)"; } else { return "UNCLEAN (offline)"; } } else if (node->details->pending) { return "pending"; } else if (node->details->standby_onfail && node->details->online) { return "standby (on-fail)"; } else if (node->details->standby) { if (node->details->online) { if (node->details->running_rsc) { return "standby (with active resources)"; } else { return "standby"; } } else { return "OFFLINE (standby)"; } } else if (node->details->maintenance) { if (node->details->online) { return "maintenance"; } else { return "OFFLINE (maintenance)"; } } else if (node->details->online) { return "online"; } return "OFFLINE"; } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_text(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); GString *str = g_string_sized_new(64); int health = pe__node_health(node); // Create a summary line with node type, name, and status if (pe__is_guest_node(node)) { g_string_append(str, "GuestNode"); } else if (pe__is_remote_node(node)) { g_string_append(str, "RemoteNode"); } else { g_string_append(str, "Node"); } pcmk__g_strcat(str, " ", node_name, ": ", node_text_status(node), NULL); if (health < 0) { g_string_append(str, " (health is RED)"); } else if (health == 0) { g_string_append(str, " (health is YELLOW)"); } if (pcmk_is_set(show_opts, pcmk_show_feature_set)) { const char *feature_set = get_node_feature_set(node); if (feature_set != NULL) { pcmk__g_strcat(str, ", feature set ", feature_set, NULL); } } /* If we're grouping by node, print its resources */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (pcmk_is_set(show_opts, pcmk_show_brief)) { GList *rscs = pe__filter_rsc_list(node->details->running_rsc, only_rsc); if (rscs != NULL) { uint32_t new_show_opts = (show_opts | pcmk_show_rsc_only) & ~pcmk_show_inactive_rscs; out->begin_list(out, NULL, NULL, "%s", str->str); out->begin_list(out, NULL, NULL, "Resources"); pe__rscs_brief_output(out, rscs, new_show_opts); out->end_list(out); out->end_list(out); g_list_free(rscs); } } else { GList *gIter2 = NULL; out->begin_list(out, NULL, NULL, "%s", str->str); out->begin_list(out, NULL, NULL, "Resources"); for (gIter2 = node->details->running_rsc; gIter2 != NULL; gIter2 = gIter2->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter2->data; show_opts |= pcmk_show_rsc_only; out->message(out, crm_map_element_name(rsc->xml), show_opts, rsc, only_node, only_rsc); } out->end_list(out); out->end_list(out); } } else { out->list_item(out, NULL, "%s", str->str); } g_string_free(str, TRUE); free(node_name); } else { char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); out->begin_list(out, NULL, NULL, "Node: %s", node_name); free(node_name); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node", "pcmk_node_t *", "uint32_t", "bool", "GList *", "GList *") static int node_xml(pcmk__output_t *out, va_list args) { pcmk_node_t *node = va_arg(args, pcmk_node_t *); uint32_t show_opts G_GNUC_UNUSED = va_arg(args, uint32_t); bool full = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); if (full) { const char *node_type = "unknown"; char *length_s = pcmk__itoa(g_list_length(node->details->running_rsc)); int health = pe__node_health(node); const char *health_s = NULL; const char *feature_set; switch (node->details->type) { case pcmk_node_variant_cluster: node_type = "member"; break; case pcmk_node_variant_remote: node_type = "remote"; break; case node_ping: node_type = "ping"; break; } if (health < 0) { health_s = "red"; } else if (health == 0) { health_s = "yellow"; } else { health_s = "green"; } feature_set = get_node_feature_set(node); pe__name_and_nvpairs_xml(out, true, "node", 15, PCMK_XA_NAME, node->details->uname, PCMK_XA_ID, node->details->id, "online", pcmk__btoa(node->details->online), "standby", pcmk__btoa(node->details->standby), "standby_onfail", pcmk__btoa(node->details->standby_onfail), "maintenance", pcmk__btoa(node->details->maintenance), "pending", pcmk__btoa(node->details->pending), "unclean", pcmk__btoa(node->details->unclean), "health", health_s, "feature_set", feature_set, "shutdown", pcmk__btoa(node->details->shutdown), "expected_up", pcmk__btoa(node->details->expected_up), "is_dc", pcmk__btoa(node->details->is_dc), "resources_running", length_s, PCMK_XA_TYPE, node_type); if (pe__is_guest_node(node)) { xmlNodePtr xml_node = pcmk__output_xml_peek_parent(out); crm_xml_add(xml_node, "id_as_resource", node->details->remote_rsc->container->id); } if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *lpc = NULL; for (lpc = node->details->running_rsc; lpc != NULL; lpc = lpc->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) lpc->data; show_opts |= pcmk_show_rsc_only; out->message(out, crm_map_element_name(rsc->xml), show_opts, rsc, only_node, only_rsc); } } free(length_s); out->end_list(out); } else { pcmk__output_xml_create_parent(out, "node", PCMK_XA_NAME, node->details->uname, NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_text(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); if (add_extra) { int v; if (value == NULL) { v = 0; } else { pcmk__scan_min_int(value, &v, INT_MIN); } if (v <= 0) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is lost", name, value); } else if (v < expected_score) { out->list_item(out, NULL, "%-32s\t: %-10s\t: Connectivity is degraded (Expected=%d)", name, value, expected_score); } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } } else { out->list_item(out, NULL, "%-32s\t: %-10s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_html(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); if (add_extra) { int v; char *s = crm_strdup_printf("%s: %s", name, value); xmlNodePtr item_node = pcmk__output_create_xml_node(out, "li", NULL); if (value == NULL) { v = 0; } else { pcmk__scan_min_int(value, &v, INT_MIN); } pcmk_create_html_node(item_node, "span", NULL, NULL, s); free(s); if (v <= 0) { pcmk_create_html_node(item_node, "span", NULL, "bold", "(connectivity is lost)"); } else if (v < expected_score) { char *buf = crm_strdup_printf("(connectivity is degraded -- expected %d", expected_score); pcmk_create_html_node(item_node, "span", NULL, "bold", buf); free(buf); } } else { out->list_item(out, NULL, "%s: %s", name, value); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pcmk_scheduler_t *", "xmlNodePtr") static int node_and_op(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pcmk_resource_t *rsc = NULL; gchar *node_str = NULL; char *last_change_str = NULL; const char *op_rsc = crm_element_value(xml_op, "resource"); int status; time_t last_change = 0; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, PCMK_EXEC_UNKNOWN); rsc = pe_find_resource(scheduler->resources, op_rsc); if (rsc) { const pcmk_node_t *node = pe__current_node(rsc); const char *target_role = g_hash_table_lookup(rsc->meta, PCMK_META_TARGET_ROLE); uint32_t show_opts = pcmk_show_rsc_only | pcmk_show_pending; if (node == NULL) { node = rsc->pending_node; } node_str = pcmk__native_output_string(rsc, rsc_printable_id(rsc), node, show_opts, target_role, false); } else { node_str = crm_strdup_printf("Unknown resource %s", op_rsc); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change) == pcmk_ok) { last_change_str = crm_strdup_printf(", %s='%s', exec=%sms", XML_RSC_OP_LAST_CHANGE, pcmk__trim(ctime(&last_change)), crm_element_value(xml_op, XML_RSC_OP_T_EXEC)); } out->list_item(out, NULL, "%s: %s (node=%s, call=%s, rc=%s%s): %s", node_str, pe__xe_history_key(xml_op), - crm_element_value(xml_op, XML_ATTR_UNAME), + crm_element_value(xml_op, PCMK_XA_UNAME), crm_element_value(xml_op, XML_LRM_ATTR_CALLID), crm_element_value(xml_op, XML_LRM_ATTR_RC), last_change_str ? last_change_str : "", pcmk_exec_status_str(status)); g_free(node_str); free(last_change_str); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-and-op", "pcmk_scheduler_t *", "xmlNodePtr") static int node_and_op_xml(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); xmlNodePtr xml_op = va_arg(args, xmlNodePtr); pcmk_resource_t *rsc = NULL; + const char *uname = crm_element_value(xml_op, PCMK_XA_UNAME); const char *op_rsc = crm_element_value(xml_op, "resource"); int status; time_t last_change = 0; xmlNode *node = NULL; pcmk__scan_min_int(crm_element_value(xml_op, XML_LRM_ATTR_OPSTATUS), &status, PCMK_EXEC_UNKNOWN); node = pcmk__output_create_xml_node(out, "operation", PCMK_XA_OP, pe__xe_history_key(xml_op), - "node", crm_element_value(xml_op, XML_ATTR_UNAME), + "node", uname, "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "rc", crm_element_value(xml_op, XML_LRM_ATTR_RC), "status", pcmk_exec_status_str(status), NULL); rsc = pe_find_resource(scheduler->resources, op_rsc); if (rsc) { const char *class = crm_element_value(rsc->xml, PCMK_XA_CLASS); const char *provider = crm_element_value(rsc->xml, PCMK_XA_PROVIDER); const char *kind = crm_element_value(rsc->xml, PCMK_XA_TYPE); bool has_provider = pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider); char *agent_tuple = crm_strdup_printf("%s:%s:%s", class, (has_provider? provider : ""), kind); pcmk__xe_set_props(node, "rsc", rsc_printable_id(rsc), "agent", agent_tuple, NULL); free(agent_tuple); } if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &last_change) == pcmk_ok) { pcmk__xe_set_props(node, XML_RSC_OP_LAST_CHANGE, pcmk__trim(ctime(&last_change)), XML_RSC_OP_T_EXEC, crm_element_value(xml_op, XML_RSC_OP_T_EXEC), NULL); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute", "const char *", "const char *", "bool", "int") static int node_attribute_xml(pcmk__output_t *out, va_list args) { const char *name = va_arg(args, const char *); const char *value = va_arg(args, const char *); bool add_extra = va_arg(args, int); int expected_score = va_arg(args, int); xmlNodePtr node = pcmk__output_create_xml_node(out, "attribute", PCMK_XA_NAME, name, PCMK_XA_VALUE, value, NULL); if (add_extra) { char *buf = pcmk__itoa(expected_score); crm_xml_add(node, "expected", buf); free(buf); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-attribute-list", "pcmk_scheduler_t *", "uint32_t", "bool", "GList *", "GList *") static int node_attribute_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); int rc = pcmk_rc_no_output; /* Display each node's attributes */ for (GList *gIter = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = gIter->data; GList *attr_list = NULL; GHashTableIter iter; gpointer key; if (!node || !node->details || !node->details->online) { continue; } g_hash_table_iter_init(&iter, node->details->attrs); while (g_hash_table_iter_next (&iter, &key, NULL)) { attr_list = filter_attr_list(attr_list, key); } if (attr_list == NULL) { continue; } if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { g_list_free(attr_list); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Node Attributes"); out->message(out, "node", node, show_opts, false, only_node, only_rsc); for (GList *aIter = attr_list; aIter != NULL; aIter = aIter->next) { const char *name = aIter->data; const char *value = NULL; int expected_score = 0; bool add_extra = false; value = pe_node_attribute_raw(node, name); add_extra = add_extra_info(node, node->details->running_rsc, scheduler, name, &expected_score); /* Print attribute name and value */ out->message(out, "node-attribute", name, value, add_extra, expected_score); } g_list_free(attr_list); out->end_list(out); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-capacity", "const pcmk_node_t *", "const char *") static int node_capacity(pcmk__output_t *out, va_list args) { const pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *comment = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s capacity:", comment, pe__node_name(node)); g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-capacity", "const pcmk_node_t *", "const char *") static int node_capacity_xml(pcmk__output_t *out, va_list args) { const pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *comment = va_arg(args, const char *); xmlNodePtr xml_node = pcmk__output_create_xml_node(out, "capacity", "node", node->details->uname, "comment", comment, NULL); g_hash_table_foreach(node->details->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-history-list", "pcmk_scheduler_t *", "pcmk_node_t *", "xmlNodePtr", "GList *", "GList *", "uint32_t", "uint32_t") static int node_history_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); xmlNode *node_state = va_arg(args, xmlNode *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); xmlNode *lrm_rsc = NULL; xmlNode *rsc_entry = NULL; int rc = pcmk_rc_no_output; lrm_rsc = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE); lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE); /* Print history of each of the node's resources */ for (rsc_entry = first_named_child(lrm_rsc, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { const char *rsc_id = crm_element_value(rsc_entry, PCMK_XA_ID); pcmk_resource_t *rsc = pe_find_resource(scheduler->resources, rsc_id); const pcmk_resource_t *parent = pe__const_top_resource(rsc, false); /* We can't use is_filtered here to filter group resources. For is_filtered, * we have to decide whether to check the parent or not. If we check the * parent, all elements of a group will always be printed because that's how * is_filtered works for groups. If we do not check the parent, sometimes * this will filter everything out. * * For other resource types, is_filtered is okay. */ if (parent->variant == pcmk_rsc_variant_group) { if (!pcmk__str_in_list(rsc_printable_id(rsc), only_rsc, pcmk__str_star_matches) && !pcmk__str_in_list(rsc_printable_id(parent), only_rsc, pcmk__str_star_matches)) { continue; } } else { if (rsc->fns->is_filtered(rsc, only_rsc, TRUE)) { continue; } } if (!pcmk_is_set(section_opts, pcmk_section_operations)) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pcmk__fc_default, NULL); if (failcount <= 0) { continue; } if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, false, only_node, only_rsc); } out->message(out, "resource-history", rsc, rsc_id, false, failcount, last_failure, false); } else { GList *op_list = get_operation_list(rsc_entry); pcmk_resource_t *rsc = NULL; if (op_list == NULL) { continue; } rsc = pe_find_resource(scheduler->resources, crm_element_value(rsc_entry, PCMK_XA_ID)); if (rc == pcmk_rc_no_output) { rc = pcmk_rc_ok; out->message(out, "node", node, show_opts, false, only_node, only_rsc); } out->message(out, "resource-operation-list", scheduler, rsc, node, op_list, show_opts); } } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_html(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer G_GNUC_UNUSED = va_arg(args, int); int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, false, rc, "Node List"); out->message(out, "node", node, show_opts, true, only_node, only_rsc); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_text(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); /* space-separated lists of node names */ GString *online_nodes = NULL; GString *online_remote_nodes = NULL; GString *online_guest_nodes = NULL; GString *offline_nodes = NULL; GString *offline_remote_nodes = NULL; int rc = pcmk_rc_no_output; for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; char *node_name = pe__node_display_name(node, pcmk_is_set(show_opts, pcmk_show_node_id)); if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { free(node_name); continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, "Node List"); // Determine whether to display node individually or in a list if (node->details->unclean || node->details->pending || (node->details->standby_onfail && node->details->online) || node->details->standby || node->details->maintenance || pcmk_is_set(show_opts, pcmk_show_rscs_by_node) || pcmk_is_set(show_opts, pcmk_show_feature_set) || (pe__node_health(node) <= 0)) { // Display node individually } else if (node->details->online) { // Display online node in a list if (pe__is_guest_node(node)) { pcmk__add_word(&online_guest_nodes, 1024, node_name); } else if (pe__is_remote_node(node)) { pcmk__add_word(&online_remote_nodes, 1024, node_name); } else { pcmk__add_word(&online_nodes, 1024, node_name); } free(node_name); continue; } else { // Display offline node in a list if (pe__is_remote_node(node)) { pcmk__add_word(&offline_remote_nodes, 1024, node_name); } else if (pe__is_guest_node(node)) { /* ignore offline guest nodes */ } else { pcmk__add_word(&offline_nodes, 1024, node_name); } free(node_name); continue; } /* If we get here, node is in bad state, or we're grouping by node */ out->message(out, "node", node, show_opts, true, only_node, only_rsc); free(node_name); } /* If we're not grouping by node, summarize nodes by status */ if (online_nodes != NULL) { out->list_item(out, "Online", "[ %s ]", (const char *) online_nodes->str); g_string_free(online_nodes, TRUE); } if (offline_nodes != NULL) { out->list_item(out, "OFFLINE", "[ %s ]", (const char *) offline_nodes->str); g_string_free(offline_nodes, TRUE); } if (online_remote_nodes) { out->list_item(out, "RemoteOnline", "[ %s ]", (const char *) online_remote_nodes->str); g_string_free(online_remote_nodes, TRUE); } if (offline_remote_nodes) { out->list_item(out, "RemoteOFFLINE", "[ %s ]", (const char *) offline_remote_nodes->str); g_string_free(offline_remote_nodes, TRUE); } if (online_guest_nodes != NULL) { out->list_item(out, "GuestOnline", "[ %s ]", (const char *) online_guest_nodes->str); g_string_free(online_guest_nodes, TRUE); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-list", "GList *", "GList *", "GList *", "uint32_t", "bool") static int node_list_xml(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer G_GNUC_UNUSED = va_arg(args, int); out->begin_list(out, NULL, NULL, "nodes"); for (GList *gIter = nodes; gIter != NULL; gIter = gIter->next) { pcmk_node_t *node = (pcmk_node_t *) gIter->data; if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } out->message(out, "node", node, show_opts, true, only_node, only_rsc); } out->end_list(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-summary", "pcmk_scheduler_t *", "GList *", "GList *", "uint32_t", "uint32_t", "bool") static int node_summary(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); uint32_t section_opts = va_arg(args, uint32_t); uint32_t show_opts = va_arg(args, uint32_t); bool print_spacer = va_arg(args, int); xmlNode *node_state = NULL; xmlNode *cib_status = pcmk_find_cib_element(scheduler->input, XML_CIB_TAG_STATUS); int rc = pcmk_rc_no_output; if (xmlChildElementCount(cib_status) == 0) { return rc; } for (node_state = first_named_child(cib_status, XML_CIB_TAG_STATE); node_state != NULL; node_state = crm_next_same_xml(node_state)) { pcmk_node_t *node = pe_find_node_id(scheduler->nodes, ID(node_state)); if (!node || !node->details || !node->details->online) { continue; } if (!pcmk__str_in_list(node->details->uname, only_node, pcmk__str_star_matches|pcmk__str_casei)) { continue; } PCMK__OUTPUT_LIST_HEADER(out, print_spacer, rc, pcmk_is_set(section_opts, pcmk_section_operations) ? "Operations" : "Migration Summary"); out->message(out, "node-history-list", scheduler, node, node_state, only_node, only_rsc, section_opts, show_opts); } PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("node-weight", "const pcmk_resource_t *", "const char *", "const char *", "const char *") static int node_weight(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); const char *score = va_arg(args, const char *); if (rsc) { out->list_item(out, NULL, "%s: %s allocation score on %s: %s", prefix, rsc->id, uname, score); } else { out->list_item(out, NULL, "%s: %s = %s", prefix, uname, score); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-weight", "const pcmk_resource_t *", "const char *", "const char *", "const char *") static int node_weight_xml(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); const char *prefix = va_arg(args, const char *); const char *uname = va_arg(args, const char *); const char *score = va_arg(args, const char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "node_weight", "function", prefix, "node", uname, "score", score, NULL); if (rsc) { crm_xml_add(node, PCMK_XA_ID, rsc->id); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNodePtr", "const char *", "const char *", "int", "uint32_t") static int op_history_text(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); uint32_t show_opts = va_arg(args, uint32_t); char *buf = op_history_string(xml_op, task, interval_ms_s, rc, pcmk_is_set(show_opts, pcmk_show_timing)); out->list_item(out, NULL, "%s", buf); free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("op-history", "xmlNodePtr", "const char *", "const char *", "int", "uint32_t") static int op_history_xml(pcmk__output_t *out, va_list args) { xmlNodePtr xml_op = va_arg(args, xmlNodePtr); const char *task = va_arg(args, const char *); const char *interval_ms_s = va_arg(args, const char *); int rc = va_arg(args, int); uint32_t show_opts = va_arg(args, uint32_t); char *rc_s = pcmk__itoa(rc); xmlNodePtr node = pcmk__output_create_xml_node(out, "operation_history", "call", crm_element_value(xml_op, XML_LRM_ATTR_CALLID), "task", task, "rc", rc_s, "rc_text", services_ocf_exitcode_str(rc), NULL); free(rc_s); if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { char *s = crm_strdup_printf("%sms", interval_ms_s); crm_xml_add(node, "interval", s); free(s); } if (pcmk_is_set(show_opts, pcmk_show_timing)) { const char *value = NULL; time_t epoch = 0; if ((crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &epoch) == pcmk_ok) && (epoch > 0)) { char *s = pcmk__epoch2str(&epoch, 0); crm_xml_add(node, XML_RSC_OP_LAST_CHANGE, s); free(s); } value = crm_element_value(xml_op, XML_RSC_OP_T_EXEC); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, XML_RSC_OP_T_EXEC, s); free(s); } value = crm_element_value(xml_op, XML_RSC_OP_T_QUEUE); if (value) { char *s = crm_strdup_printf("%sms", value); crm_xml_add(node, XML_RSC_OP_T_QUEUE, s); free(s); } } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int promotion_score(pcmk__output_t *out, va_list args) { pcmk_resource_t *child_rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *chosen = va_arg(args, pcmk_node_t *); const char *score = va_arg(args, const char *); out->list_item(out, NULL, "%s promotion score on %s: %s", child_rsc->id, chosen? chosen->details->uname : "none", score); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("promotion-score", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int promotion_score_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *child_rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *chosen = va_arg(args, pcmk_node_t *); const char *score = va_arg(args, const char *); xmlNodePtr node = pcmk__output_create_xml_node(out, "promotion_score", PCMK_XA_ID, child_rsc->id, "score", score, NULL); if (chosen) { crm_xml_add(node, "node", chosen->details->uname); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-config", "const pcmk_resource_t *", "bool") static int resource_config(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); bool raw = va_arg(args, int); char *rsc_xml = formatted_xml_buf(rsc, raw); out->output_xml(out, "xml", rsc_xml); free(rsc_xml); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-config", "const pcmk_resource_t *", "bool") static int resource_config_text(pcmk__output_t *out, va_list args) { const pcmk_resource_t *rsc = va_arg(args, const pcmk_resource_t *); bool raw = va_arg(args, int); char *rsc_xml = formatted_xml_buf(rsc, raw); pcmk__formatted_printf(out, "Resource XML:\n"); out->output_xml(out, "xml", rsc_xml); free(rsc_xml); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-history", "pcmk_resource_t *", "const char *", "bool", "int", "time_t", "bool") static int resource_history_text(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); const char *rsc_id = va_arg(args, const char *); bool all = va_arg(args, int); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, time_t); bool as_header = va_arg(args, int); char *buf = resource_history_string(rsc, rsc_id, all, failcount, last_failure); if (as_header) { out->begin_list(out, NULL, NULL, "%s", buf); } else { out->list_item(out, NULL, "%s", buf); } free(buf); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-history", "pcmk_resource_t *", "const char *", "bool", "int", "time_t", "bool") static int resource_history_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); const char *rsc_id = va_arg(args, const char *); bool all = va_arg(args, int); int failcount = va_arg(args, int); time_t last_failure = va_arg(args, time_t); bool as_header = va_arg(args, int); xmlNodePtr node = pcmk__output_xml_create_parent(out, "resource_history", PCMK_XA_ID, rsc_id, NULL); if (rsc == NULL) { pcmk__xe_set_bool_attr(node, "orphan", true); } else if (all || failcount || last_failure > 0) { char *migration_s = pcmk__itoa(rsc->migration_threshold); pcmk__xe_set_props(node, "orphan", "false", PCMK_META_MIGRATION_THRESHOLD, migration_s, NULL); free(migration_s); if (failcount > 0) { char *s = pcmk__itoa(failcount); crm_xml_add(node, PCMK__FAIL_COUNT_PREFIX, s); free(s); } if (last_failure > 0) { char *s = pcmk__epoch2str(&last_failure, 0); crm_xml_add(node, PCMK__LAST_FAILURE_PREFIX, s); free(s); } } if (!as_header) { pcmk__output_xml_pop_parent(out); } return pcmk_rc_ok; } static void print_resource_header(pcmk__output_t *out, uint32_t show_opts) { if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { /* Active resources have already been printed by node */ out->begin_list(out, NULL, NULL, "Inactive Resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->begin_list(out, NULL, NULL, "Full List of Resources"); } else { out->begin_list(out, NULL, NULL, "Active Resources"); } } PCMK__OUTPUT_ARGS("resource-list", "pcmk_scheduler_t *", "uint32_t", "bool", "GList *", "GList *", "bool") static int resource_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); uint32_t show_opts = va_arg(args, uint32_t); bool print_summary = va_arg(args, int); GList *only_node = va_arg(args, GList *); GList *only_rsc = va_arg(args, GList *); bool print_spacer = va_arg(args, int); GList *rsc_iter; int rc = pcmk_rc_no_output; bool printed_header = false; /* If we already showed active resources by node, and * we're not showing inactive resources, we have nothing to do */ if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node) && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { return rc; } /* If we haven't already printed resources grouped by node, * and brief output was requested, print resource summary */ if (pcmk_is_set(show_opts, pcmk_show_brief) && !pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { GList *rscs = pe__filter_rsc_list(scheduler->resources, only_rsc); PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; rc = pe__rscs_brief_output(out, rscs, show_opts); g_list_free(rscs); } /* For each resource, display it if appropriate */ for (rsc_iter = scheduler->resources; rsc_iter != NULL; rsc_iter = rsc_iter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) rsc_iter->data; int x; /* Complex resources may have some sub-resources active and some inactive */ gboolean is_active = rsc->fns->active(rsc, TRUE); gboolean partially_active = rsc->fns->active(rsc, FALSE); /* Skip inactive orphans (deleted but still in CIB) */ if (pcmk_is_set(rsc->flags, pcmk_rsc_removed) && !is_active) { continue; /* Skip active resources if we already displayed them by node */ } else if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { if (is_active) { continue; } /* Skip primitives already counted in a brief summary */ } else if (pcmk_is_set(show_opts, pcmk_show_brief) && (rsc->variant == pcmk_rsc_variant_primitive)) { continue; /* Skip resources that aren't at least partially active, * unless we're displaying inactive resources */ } else if (!partially_active && !pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { continue; } else if (partially_active && !pe__rsc_running_on_any(rsc, only_node)) { continue; } if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } /* Print this resource */ x = out->message(out, crm_map_element_name(rsc->xml), show_opts, rsc, only_node, only_rsc); if (x == pcmk_rc_ok) { rc = pcmk_rc_ok; } } if (print_summary && rc != pcmk_rc_ok) { if (!printed_header) { PCMK__OUTPUT_SPACER_IF(out, print_spacer); print_resource_header(out, show_opts); printed_header = true; } if (pcmk_is_set(show_opts, pcmk_show_rscs_by_node)) { out->list_item(out, NULL, "No inactive resources"); } else if (pcmk_is_set(show_opts, pcmk_show_inactive_rscs)) { out->list_item(out, NULL, "No resources"); } else { out->list_item(out, NULL, "No active resources"); } } if (printed_header) { out->end_list(out); } return rc; } PCMK__OUTPUT_ARGS("resource-operation-list", "pcmk_scheduler_t *", "pcmk_resource_t *", "pcmk_node_t *", "GList *", "uint32_t") static int resource_operation_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler G_GNUC_UNUSED = va_arg(args, pcmk_scheduler_t *); pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); GList *op_list = va_arg(args, GList *); uint32_t show_opts = va_arg(args, uint32_t); GList *gIter = NULL; int rc = pcmk_rc_no_output; /* Print each operation */ for (gIter = op_list; gIter != NULL; gIter = gIter->next) { xmlNode *xml_op = (xmlNode *) gIter->data; const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); const char *interval_ms_s = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); const char *op_rc = crm_element_value(xml_op, XML_LRM_ATTR_RC); int op_rc_i; pcmk__scan_min_int(op_rc, &op_rc_i, 0); /* Display 0-interval monitors as "probe" */ if (pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei) && pcmk__str_eq(interval_ms_s, "0", pcmk__str_null_matches | pcmk__str_casei)) { task = "probe"; } /* If this is the first printed operation, print heading for resource */ if (rc == pcmk_rc_no_output) { time_t last_failure = 0; int failcount = pe_get_failcount(node, rsc, &last_failure, pcmk__fc_default, NULL); out->message(out, "resource-history", rsc, rsc_printable_id(rsc), true, failcount, last_failure, true); rc = pcmk_rc_ok; } /* Print the operation */ out->message(out, "op-history", xml_op, task, interval_ms_s, op_rc_i, show_opts); } /* Free the list we created (no need to free the individual items) */ g_list_free(op_list); PCMK__OUTPUT_LIST_FOOTER(out, rc); return rc; } PCMK__OUTPUT_ARGS("resource-util", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int resource_util(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *fn = va_arg(args, const char *); char *dump_text = crm_strdup_printf("%s: %s utilization on %s:", fn, rsc->id, pe__node_name(node)); g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text); out->list_item(out, NULL, "%s", dump_text); free(dump_text); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("resource-util", "pcmk_resource_t *", "pcmk_node_t *", "const char *") static int resource_util_xml(pcmk__output_t *out, va_list args) { pcmk_resource_t *rsc = va_arg(args, pcmk_resource_t *); pcmk_node_t *node = va_arg(args, pcmk_node_t *); const char *fn = va_arg(args, const char *); xmlNodePtr xml_node = pcmk__output_create_xml_node(out, "utilization", "resource", rsc->id, "node", node->details->uname, "function", fn, NULL); g_hash_table_foreach(rsc->utilization, add_dump_node, xml_node); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pcmk_ticket_t *") static int ticket_html(pcmk__output_t *out, va_list args) { pcmk_ticket_t *ticket = va_arg(args, pcmk_ticket_t *); if (ticket->last_granted > -1) { char *epoch_str = pcmk__epoch2str(&(ticket->last_granted), 0); out->list_item(out, NULL, "%s:\t%s%s %s=\"%s\"", ticket->id, ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : "", "last-granted", pcmk__s(epoch_str, "")); free(epoch_str); } else { out->list_item(out, NULL, "%s:\t%s%s", ticket->id, ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : ""); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pcmk_ticket_t *") static int ticket_text(pcmk__output_t *out, va_list args) { pcmk_ticket_t *ticket = va_arg(args, pcmk_ticket_t *); if (ticket->last_granted > -1) { char *epoch_str = pcmk__epoch2str(&(ticket->last_granted), 0); out->list_item(out, ticket->id, "%s%s %s=\"%s\"", ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : "", "last-granted", pcmk__s(epoch_str, "")); free(epoch_str); } else { out->list_item(out, ticket->id, "%s%s", ticket->granted ? "granted" : "revoked", ticket->standby ? " [standby]" : ""); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket", "pcmk_ticket_t *") static int ticket_xml(pcmk__output_t *out, va_list args) { pcmk_ticket_t *ticket = va_arg(args, pcmk_ticket_t *); xmlNodePtr node = NULL; node = pcmk__output_create_xml_node(out, "ticket", PCMK_XA_ID, ticket->id, "status", ticket->granted ? "granted" : "revoked", "standby", pcmk__btoa(ticket->standby), NULL); if (ticket->last_granted > -1) { char *buf = pcmk__epoch2str(&ticket->last_granted, 0); crm_xml_add(node, "last-granted", buf); free(buf); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("ticket-list", "pcmk_scheduler_t *", "bool") static int ticket_list(pcmk__output_t *out, va_list args) { pcmk_scheduler_t *scheduler = va_arg(args, pcmk_scheduler_t *); bool print_spacer = va_arg(args, int); GHashTableIter iter; gpointer key, value; if (g_hash_table_size(scheduler->tickets) == 0) { return pcmk_rc_no_output; } PCMK__OUTPUT_SPACER_IF(out, print_spacer); /* Print section heading */ out->begin_list(out, NULL, NULL, "Tickets"); /* Print each ticket */ g_hash_table_iter_init(&iter, scheduler->tickets); while (g_hash_table_iter_next(&iter, &key, &value)) { pcmk_ticket_t *ticket = (pcmk_ticket_t *) value; out->message(out, "ticket", ticket); } /* Close section */ out->end_list(out); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "ban", "default", ban_text }, { "ban", "html", ban_html }, { "ban", "xml", ban_xml }, { "ban-list", "default", ban_list }, { "bundle", "default", pe__bundle_text }, { "bundle", "xml", pe__bundle_xml }, { "bundle", "html", pe__bundle_html }, { "clone", "default", pe__clone_default }, { "clone", "xml", pe__clone_xml }, { "cluster-counts", "default", cluster_counts_text }, { "cluster-counts", "html", cluster_counts_html }, { "cluster-counts", "xml", cluster_counts_xml }, { "cluster-dc", "default", cluster_dc_text }, { "cluster-dc", "html", cluster_dc_html }, { "cluster-dc", "xml", cluster_dc_xml }, { "cluster-options", "default", cluster_options_text }, { "cluster-options", "html", cluster_options_html }, { "cluster-options", "log", cluster_options_log }, { "cluster-options", "xml", cluster_options_xml }, { "cluster-summary", "default", cluster_summary }, { "cluster-summary", "html", cluster_summary_html }, { "cluster-stack", "default", cluster_stack_text }, { "cluster-stack", "html", cluster_stack_html }, { "cluster-stack", "xml", cluster_stack_xml }, { "cluster-times", "default", cluster_times_text }, { "cluster-times", "html", cluster_times_html }, { "cluster-times", "xml", cluster_times_xml }, { "failed-action", "default", failed_action_default }, { "failed-action", "xml", failed_action_xml }, { "failed-action-list", "default", failed_action_list }, { "group", "default", pe__group_default}, { "group", "xml", pe__group_xml }, { "maint-mode", "text", cluster_maint_mode_text }, { "node", "default", node_text }, { "node", "html", node_html }, { "node", "xml", node_xml }, { "node-and-op", "default", node_and_op }, { "node-and-op", "xml", node_and_op_xml }, { "node-capacity", "default", node_capacity }, { "node-capacity", "xml", node_capacity_xml }, { "node-history-list", "default", node_history_list }, { "node-list", "default", node_list_text }, { "node-list", "html", node_list_html }, { "node-list", "xml", node_list_xml }, { "node-weight", "default", node_weight }, { "node-weight", "xml", node_weight_xml }, { "node-attribute", "default", node_attribute_text }, { "node-attribute", "html", node_attribute_html }, { "node-attribute", "xml", node_attribute_xml }, { "node-attribute-list", "default", node_attribute_list }, { "node-summary", "default", node_summary }, { "op-history", "default", op_history_text }, { "op-history", "xml", op_history_xml }, { "primitive", "default", pe__resource_text }, { "primitive", "xml", pe__resource_xml }, { "primitive", "html", pe__resource_html }, { "promotion-score", "default", promotion_score }, { "promotion-score", "xml", promotion_score_xml }, { "resource-config", "default", resource_config }, { "resource-config", "text", resource_config_text }, { "resource-history", "default", resource_history_text }, { "resource-history", "xml", resource_history_xml }, { "resource-list", "default", resource_list }, { "resource-operation-list", "default", resource_operation_list }, { "resource-util", "default", resource_util }, { "resource-util", "xml", resource_util_xml }, { "ticket", "default", ticket_text }, { "ticket", "html", ticket_html }, { "ticket", "xml", ticket_xml }, { "ticket-list", "default", ticket_list }, { NULL, NULL, NULL } }; void pe__register_messages(pcmk__output_t *out) { pcmk__register_messages(out, fmt_functions); } diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c index 18ee0a3cd3..e63e0a7b69 100644 --- a/lib/pengine/unpack.c +++ b/lib/pengine/unpack.c @@ -1,5128 +1,5129 @@ /* * Copyright 2004-2024 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); // A (parsed) resource action history entry struct action_history { pcmk_resource_t *rsc; // Resource that history is for pcmk_node_t *node; // Node that history is for xmlNode *xml; // History entry XML // Parsed from entry XML const char *id; // XML ID of history entry const char *key; // Operation key of action const char *task; // Action name const char *exit_reason; // Exit reason given for result guint interval_ms; // Action interval int call_id; // Call ID of action int expected_exit_status; // Expected exit status of action int exit_status; // Actual exit status of action int execution_status; // Execution status of action }; /* This uses pcmk__set_flags_as()/pcmk__clear_flags_as() directly rather than * use pcmk__set_scheduler_flags()/pcmk__clear_scheduler_flags() so that the * flag is stringified more readably in log messages. */ #define set_config_flag(scheduler, option, flag) do { \ const char *scf_value = pe_pref((scheduler)->config_hash, (option)); \ if (scf_value != NULL) { \ if (crm_is_true(scf_value)) { \ (scheduler)->flags = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, "Scheduler", \ crm_system_name, (scheduler)->flags, \ (flag), #flag); \ } else { \ (scheduler)->flags = pcmk__clear_flags_as(__func__, __LINE__, \ LOG_TRACE, "Scheduler", \ crm_system_name, (scheduler)->flags, \ (flag), #flag); \ } \ } \ } while(0) static void unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *failed); static void determine_remote_online_status(pcmk_scheduler_t *scheduler, pcmk_node_t *this_node); static void add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node, bool overwrite, pcmk_scheduler_t *scheduler); static void determine_online_status(const xmlNode *node_state, pcmk_node_t *this_node, pcmk_scheduler_t *scheduler); static void unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml, pcmk_scheduler_t *scheduler); static gboolean is_dangling_guest_node(pcmk_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, pcmk_rsc_removed_filler)) { return TRUE; } return FALSE; } /*! * \brief Schedule a fence action for a node * * \param[in,out] scheduler Scheduler data * \param[in,out] node Node to fence * \param[in] reason Text description of why fencing is needed * \param[in] priority_delay Whether to consider * \c PCMK_OPT_PRIORITY_FENCING_DELAY */ void pe_fence_node(pcmk_scheduler_t *scheduler, pcmk_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)) { pcmk_resource_t *rsc = node->details->remote_rsc->container; if (!pcmk_is_set(rsc->flags, pcmk_rsc_failed)) { if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)) { crm_notice("Not fencing guest node %s " "(otherwise would because %s): " "its guest resource %s is unmanaged", pe__node_name(node), reason, rsc->id); } else { pcmk__sched_warn("Guest node %s will be fenced " "(by recovering its guest resource %s): %s", pe__node_name(node), 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; pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); } } } 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", pe__node_name(node), reason); pcmk__set_rsc_flags(node->details->remote_rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); } else if (pe__is_remote_node(node)) { pcmk_resource_t *rsc = node->details->remote_rsc; if ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_managed)) { crm_notice("Not fencing remote node %s " "(otherwise would because %s): connection is unmanaged", pe__node_name(node), reason); } else if(node->details->remote_requires_reset == FALSE) { node->details->remote_requires_reset = TRUE; pcmk__sched_warn("Remote node %s %s: %s", pe__node_name(node), pe_can_fence(scheduler, node)? "will be fenced" : "is unclean", reason); } node->details->unclean = TRUE; // No need to apply PCMK_OPT_PRIORITY_FENCING_DELAY for remote nodes pe_fence_op(node, NULL, TRUE, reason, FALSE, scheduler); } else if (node->details->unclean) { crm_trace("Cluster node %s %s because %s", pe__node_name(node), pe_can_fence(scheduler, node)? "would also be fenced" : "also is unclean", reason); } else { pcmk__sched_warn("Cluster node %s %s: %s", pe__node_name(node), pe_can_fence(scheduler, node)? "will be fenced" : "is unclean", reason); node->details->unclean = TRUE; pe_fence_op(node, NULL, TRUE, reason, priority_delay, scheduler); } } // @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 \ "[(@" PCMK_XA_NAME "='" PCMK_STONITH_PROVIDES "'" \ "or @" PCMK_XA_NAME "='" PCMK_META_REQUIRES "') " \ "and @" PCMK_XA_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, pcmk_scheduler_t *scheduler) { xmlXPathObjectPtr result = NULL; if (!pcmk_is_set(scheduler->flags, flag)) { result = xpath_search(scheduler->input, xpath); if (result && (numXpathResults(result) > 0)) { pcmk__set_scheduler_flags(scheduler, flag); } freeXpathObject(result); } } gboolean unpack_config(xmlNode *config, pcmk_scheduler_t *scheduler) { const char *value = NULL; guint interval_ms = 0U; GHashTable *config_hash = pcmk__strkey_table(free, free); pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = pcmk_role_unknown, .now = scheduler->now, .match_data = NULL, .rsc_data = NULL, .op_data = NULL }; scheduler->config_hash = config_hash; pe__unpack_dataset_nvpairs(config, XML_CIB_TAG_PROPSET, &rule_data, config_hash, CIB_OPTIONS_FIRST, FALSE, scheduler); verify_pe_options(scheduler->config_hash); set_config_flag(scheduler, PCMK_OPT_ENABLE_STARTUP_PROBES, pcmk_sched_probe_resources); if (!pcmk_is_set(scheduler->flags, pcmk_sched_probe_resources)) { crm_info("Startup probes: disabled (dangerous)"); } value = pe_pref(scheduler->config_hash, PCMK_OPT_HAVE_WATCHDOG); if (value && crm_is_true(value)) { crm_info("Watchdog-based self-fencing will be performed via SBD if " "fencing is required and " PCMK_OPT_STONITH_WATCHDOG_TIMEOUT " is nonzero"); pcmk__set_scheduler_flags(scheduler, pcmk_sched_have_fencing); } /* Set certain flags via xpath here, so they can be used before the relevant * configuration sections are unpacked. */ set_if_xpath(pcmk_sched_enable_unfencing, XPATH_ENABLE_UNFENCING, scheduler); value = pe_pref(scheduler->config_hash, PCMK_OPT_STONITH_TIMEOUT); pcmk_parse_interval_spec(value, &interval_ms); if (interval_ms >= INT_MAX) { scheduler->stonith_timeout = INT_MAX; } else { scheduler->stonith_timeout = (int) interval_ms; } crm_debug("STONITH timeout: %d", scheduler->stonith_timeout); set_config_flag(scheduler, PCMK_OPT_STONITH_ENABLED, pcmk_sched_fencing_enabled); if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) { crm_debug("STONITH of failed nodes is enabled"); } else { crm_debug("STONITH of failed nodes is disabled"); } scheduler->stonith_action = pe_pref(scheduler->config_hash, PCMK_OPT_STONITH_ACTION); if (!strcmp(scheduler->stonith_action, "poweroff")) { pcmk__warn_once(pcmk__wo_poweroff, "Support for " PCMK_OPT_STONITH_ACTION " of " "'poweroff' is deprecated and will be removed in a " "future release (use 'off' instead)"); scheduler->stonith_action = PCMK_ACTION_OFF; } crm_trace("STONITH will %s nodes", scheduler->stonith_action); set_config_flag(scheduler, PCMK_OPT_CONCURRENT_FENCING, pcmk_sched_concurrent_fencing); if (pcmk_is_set(scheduler->flags, pcmk_sched_concurrent_fencing)) { crm_debug("Concurrent fencing is enabled"); } else { crm_debug("Concurrent fencing is disabled"); } value = pe_pref(scheduler->config_hash, PCMK_OPT_PRIORITY_FENCING_DELAY); if (value) { pcmk_parse_interval_spec(value, &interval_ms); scheduler->priority_fencing_delay = (int) (interval_ms / 1000); crm_trace("Priority fencing delay is %ds", scheduler->priority_fencing_delay); } set_config_flag(scheduler, PCMK_OPT_STOP_ALL_RESOURCES, pcmk_sched_stop_all); crm_debug("Stop all active resources: %s", pcmk__btoa(pcmk_is_set(scheduler->flags, pcmk_sched_stop_all))); set_config_flag(scheduler, PCMK_OPT_SYMMETRIC_CLUSTER, pcmk_sched_symmetric_cluster); if (pcmk_is_set(scheduler->flags, pcmk_sched_symmetric_cluster)) { crm_debug("Cluster is symmetric" " - resources can run anywhere by default"); } value = pe_pref(scheduler->config_hash, PCMK_OPT_NO_QUORUM_POLICY); if (pcmk__str_eq(value, "ignore", pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_ignore; } else if (pcmk__str_eq(value, "freeze", pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_freeze; } else if (pcmk__str_eq(value, "demote", pcmk__str_casei)) { scheduler->no_quorum_policy = pcmk_no_quorum_demote; } else if (pcmk__str_eq(value, "suicide", pcmk__str_casei)) { if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) { int do_panic = 0; crm_element_value_int(scheduler->input, PCMK_XA_NO_QUORUM_PANIC, &do_panic); if (do_panic || pcmk_is_set(scheduler->flags, pcmk_sched_quorate)) { scheduler->no_quorum_policy = pcmk_no_quorum_fence; } else { crm_notice("Resetting " PCMK_OPT_NO_QUORUM_POLICY " to 'stop': cluster has never had quorum"); scheduler->no_quorum_policy = pcmk_no_quorum_stop; } } else { pcmk__config_err("Resetting " PCMK_OPT_NO_QUORUM_POLICY " to 'stop' because fencing is disabled"); scheduler->no_quorum_policy = pcmk_no_quorum_stop; } } else { scheduler->no_quorum_policy = pcmk_no_quorum_stop; } switch (scheduler->no_quorum_policy) { case pcmk_no_quorum_freeze: crm_debug("On loss of quorum: Freeze resources"); break; case pcmk_no_quorum_stop: crm_debug("On loss of quorum: Stop ALL resources"); break; case pcmk_no_quorum_demote: crm_debug("On loss of quorum: " "Demote promotable resources and stop other resources"); break; case pcmk_no_quorum_fence: crm_notice("On loss of quorum: Fence all remaining nodes"); break; case pcmk_no_quorum_ignore: crm_notice("On loss of quorum: Ignore"); break; } set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_RESOURCES, pcmk_sched_stop_removed_resources); if (pcmk_is_set(scheduler->flags, pcmk_sched_stop_removed_resources)) { crm_trace("Orphan resources are stopped"); } else { crm_trace("Orphan resources are ignored"); } set_config_flag(scheduler, PCMK_OPT_STOP_ORPHAN_ACTIONS, pcmk_sched_cancel_removed_actions); if (pcmk_is_set(scheduler->flags, pcmk_sched_cancel_removed_actions)) { crm_trace("Orphan resource actions are stopped"); } else { crm_trace("Orphan resource actions are ignored"); } value = pe_pref(scheduler->config_hash, PCMK__OPT_REMOVE_AFTER_STOP); if (value != NULL) { if (crm_is_true(value)) { pcmk__set_scheduler_flags(scheduler, pcmk_sched_remove_after_stop); #ifndef PCMK__COMPAT_2_0 pcmk__warn_once(pcmk__wo_remove_after, "Support for the " PCMK__OPT_REMOVE_AFTER_STOP " cluster property is deprecated and will be " "removed in a future release"); #endif } else { pcmk__clear_scheduler_flags(scheduler, pcmk_sched_remove_after_stop); } } set_config_flag(scheduler, PCMK_OPT_MAINTENANCE_MODE, pcmk_sched_in_maintenance); crm_trace("Maintenance mode: %s", pcmk__btoa(pcmk_is_set(scheduler->flags, pcmk_sched_in_maintenance))); set_config_flag(scheduler, PCMK_OPT_START_FAILURE_IS_FATAL, pcmk_sched_start_failure_fatal); if (pcmk_is_set(scheduler->flags, pcmk_sched_start_failure_fatal)) { crm_trace("Start failures are always fatal"); } else { crm_trace("Start failures are handled by failcount"); } if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled)) { set_config_flag(scheduler, PCMK_OPT_STARTUP_FENCING, pcmk_sched_startup_fencing); } if (pcmk_is_set(scheduler->flags, pcmk_sched_startup_fencing)) { crm_trace("Unseen nodes will be fenced"); } else { pcmk__warn_once(pcmk__wo_blind, "Blind faith: not fencing unseen nodes"); } pe__unpack_node_health_scores(scheduler); scheduler->placement_strategy = pe_pref(scheduler->config_hash, PCMK_OPT_PLACEMENT_STRATEGY); crm_trace("Placement strategy: %s", scheduler->placement_strategy); set_config_flag(scheduler, PCMK_OPT_SHUTDOWN_LOCK, pcmk_sched_shutdown_lock); if (pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) { value = pe_pref(scheduler->config_hash, PCMK_OPT_SHUTDOWN_LOCK_LIMIT); pcmk_parse_interval_spec(value, &(scheduler->shutdown_lock)); scheduler->shutdown_lock /= 1000; crm_trace("Resources will be locked to nodes that were cleanly " "shut down (locks expire after %s)", pcmk__readable_interval(scheduler->shutdown_lock)); } else { crm_trace("Resources will not be locked to nodes that were cleanly " "shut down"); } value = pe_pref(scheduler->config_hash, PCMK_OPT_NODE_PENDING_TIMEOUT); pcmk_parse_interval_spec(value, &(scheduler->node_pending_timeout)); scheduler->node_pending_timeout /= 1000; if (scheduler->node_pending_timeout == 0) { crm_trace("Do not fence pending nodes"); } else { crm_trace("Fence pending nodes after %s", pcmk__readable_interval(scheduler->node_pending_timeout * 1000)); } return TRUE; } pcmk_node_t * pe_create_node(const char *id, const char *uname, const char *type, const char *score, pcmk_scheduler_t *scheduler) { pcmk_node_t *new_node = NULL; if (pe_find_node(scheduler->nodes, uname) != NULL) { pcmk__config_warn("More than one node entry has name '%s'", uname); } new_node = calloc(1, sizeof(pcmk_node_t)); if (new_node == NULL) { pcmk__sched_err("Could not allocate memory for node %s", uname); return NULL; } new_node->weight = char2score(score); new_node->details = calloc(1, sizeof(struct pe_node_shared_s)); if (new_node->details == NULL) { free(new_node); pcmk__sched_err("Could not allocate memory for node %s", uname); 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 = scheduler; if (pcmk__str_eq(type, "member", pcmk__str_null_matches | pcmk__str_casei)) { new_node->details->type = pcmk_node_variant_cluster; } else if (pcmk__str_eq(type, "remote", pcmk__str_casei)) { new_node->details->type = pcmk_node_variant_remote; pcmk__set_scheduler_flags(scheduler, pcmk_sched_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); } pcmk__warn_once(pcmk__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); scheduler->nodes = g_list_insert_sorted(scheduler->nodes, new_node, pe__cmp_node_name); return new_node; } static const char * expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pcmk_scheduler_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, PCMK_XA_VALUE); const char *name = crm_element_value(attr, PCMK_XA_NAME); if (name == NULL) { // Sanity continue; } if (strcmp(name, PCMK_META_REMOTE_NODE) == 0) { remote_name = value; } else if (strcmp(name, PCMK_META_REMOTE_ADDR) == 0) { remote_server = value; } else if (strcmp(name, PCMK_META_REMOTE_PORT) == 0) { remote_port = value; } else if (strcmp(name, PCMK_META_REMOTE_CONNECT_TIMEOUT) == 0) { connect_timeout = value; } else if (strcmp(name, PCMK_META_REMOTE_ALLOW_MIGRATE) == 0) { remote_allow_migrate = value; } else if (strcmp(name, PCMK_META_IS_MANAGED) == 0) { 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(pcmk_scheduler_t *scheduler, pcmk_node_t *new_node) { if ((new_node->details->type == pcmk_node_variant_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(scheduler->flags, pcmk_sched_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, pcmk_scheduler_t *scheduler) { xmlNode *xml_obj = NULL; pcmk_node_t *new_node = NULL; const char *id = NULL; const char *uname = NULL; const char *type = NULL; const char *score = 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, PCMK_XA_ID); - uname = crm_element_value(xml_obj, XML_ATTR_UNAME); + uname = crm_element_value(xml_obj, PCMK_XA_UNAME); type = crm_element_value(xml_obj, PCMK_XA_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, scheduler); if (new_node == NULL) { return FALSE; } handle_startup_fencing(scheduler, new_node); add_node_attrs(xml_obj, new_node, FALSE, scheduler); - crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME)); + crm_trace("Done with node %s", + crm_element_value(xml_obj, PCMK_XA_UNAME)); } } if (scheduler->localhost && (pe_find_node(scheduler->nodes, scheduler->localhost) == NULL)) { crm_info("Creating a fake local node"); pe_create_node(scheduler->localhost, scheduler->localhost, NULL, 0, scheduler); } return TRUE; } static void setup_container(pcmk_resource_t *rsc, pcmk_scheduler_t *scheduler) { const char *container_id = NULL; if (rsc->children) { g_list_foreach(rsc->children, (GFunc) setup_container, scheduler); return; } container_id = g_hash_table_lookup(rsc->meta, PCMK__META_CONTAINER); if (container_id && !pcmk__str_eq(container_id, rsc->id, pcmk__str_casei)) { pcmk_resource_t *container = pe_find_resource(scheduler->resources, container_id); if (container) { rsc->container = container; pcmk__set_rsc_flags(container, pcmk_rsc_has_filler); container->fillers = g_list_append(container->fillers, rsc); pcmk__rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id); } else { pcmk__config_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id); } } } gboolean unpack_remote_nodes(xmlNode *xml_resources, pcmk_scheduler_t *scheduler) { 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(scheduler->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, scheduler); } 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, scheduler); if (new_node_id && (pe_find_node(scheduler->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, scheduler); } 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, scheduler); if (new_node_id && (pe_find_node(scheduler->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, scheduler); } } } } 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(pcmk_scheduler_t *scheduler, pcmk_resource_t *new_rsc) { pcmk_node_t *remote_node = NULL; if (new_rsc->is_remote_node == FALSE) { return; } if (pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) { /* remote_nodes and remote_resources are not linked in quick location calculations */ return; } remote_node = pe_find_node(scheduler->nodes, new_rsc->id); CRM_CHECK(remote_node != NULL, return); pcmk__rsc_trace(new_rsc, "Linking remote connection resource %s to %s", new_rsc->id, pe__node_name(remote_node)); 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(scheduler, 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) { pcmk_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] scheduler Scheduler data * * \return TRUE * * \note unpack_remote_nodes() MUST be called before this, so that the nodes can * be used when pe__unpack_resource() calls resource_location() */ gboolean unpack_resources(const xmlNode *xml_resources, pcmk_scheduler_t *scheduler) { xmlNode *xml_obj = NULL; GList *gIter = NULL; scheduler->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)) { pcmk_resource_t *new_rsc = NULL; const char *id = ID(xml_obj); if (pcmk__str_empty(id)) { pcmk__config_err("Ignoring <%s> resource without ID", xml_obj->name); continue; } if (pcmk__str_eq((const char *) xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, pcmk__str_none)) { if (g_hash_table_lookup_extended(scheduler->template_rsc_sets, id, NULL, NULL) == FALSE) { /* Record the template's ID for the knowledge of its existence anyway. */ g_hash_table_insert(scheduler->template_rsc_sets, strdup(id), NULL); } continue; } crm_trace("Unpacking <%s " PCMK_XA_ID "='%s'>", xml_obj->name, id); if (pe__unpack_resource(xml_obj, &new_rsc, NULL, scheduler) == pcmk_rc_ok) { scheduler->resources = g_list_append(scheduler->resources, new_rsc); pcmk__rsc_trace(new_rsc, "Added resource %s", new_rsc->id); } else { pcmk__config_err("Ignoring <%s> resource '%s' " "because configuration is invalid", xml_obj->name, id); } } for (gIter = scheduler->resources; gIter != NULL; gIter = gIter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) gIter->data; setup_container(rsc, scheduler); link_rsc2remotenode(scheduler, rsc); } scheduler->resources = g_list_sort(scheduler->resources, pe__cmp_rsc_priority); if (pcmk_is_set(scheduler->flags, pcmk_sched_location_only)) { /* Ignore */ } else if (pcmk_is_set(scheduler->flags, pcmk_sched_fencing_enabled) && !pcmk_is_set(scheduler->flags, pcmk_sched_have_fencing)) { 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 " PCMK_OPT_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, pcmk_scheduler_t *scheduler) { xmlNode *xml_tag = NULL; scheduler->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 " PCMK_XA_ID, (const char *) xml_tag->name); 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 " PCMK_XA_ID, xml_obj_ref->name, tag_id); continue; } if (add_tag_ref(scheduler->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, pcmk_scheduler_t *scheduler) { const char *ticket_id = NULL; const char *granted = NULL; const char *last_granted = NULL; const char *standby = NULL; xmlAttrPtr xIter = NULL; pcmk_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(scheduler->tickets, ticket_id); if (ticket == NULL) { ticket = ticket_new(ticket_id, scheduler); 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 = pcmk__xml_attr_value(xIter); if (pcmk__str_eq(prop_name, PCMK_XA_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, pcmk_scheduler_t *scheduler) { 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, scheduler); } return TRUE; } static void unpack_handle_remote_attrs(pcmk_node_t *this_node, const xmlNode *state, pcmk_scheduler_t *scheduler) { const char *resource_discovery_enabled = NULL; const xmlNode *attrs = NULL; pcmk_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 Pacemaker Remote node %s", pe__node_name(this_node)); 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, scheduler); if (pe__shutdown_requested(this_node)) { crm_info("%s is shutting down", pe__node_name(this_node)); this_node->details->shutdown = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) { crm_info("%s is in standby mode", pe__node_name(this_node)); this_node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) || ((rsc != NULL) && !pcmk_is_set(rsc->flags, pcmk_rsc_managed))) { crm_info("%s is in maintenance mode", pe__node_name(this_node)); 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(scheduler->flags, pcmk_sched_fencing_enabled)) { pcmk__config_warn("Ignoring " XML_NODE_ATTR_RSC_DISCOVERY " attribute on Pacemaker Remote node %s" " because fencing is disabled", pe__node_name(this_node)); } 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("%s has resource discovery disabled", pe__node_name(this_node)); 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,out] node Cluster node whose attributes are being unpacked * \param[in,out] scheduler Scheduler data */ static void unpack_transient_attributes(const xmlNode *state, pcmk_node_t *node, pcmk_scheduler_t *scheduler) { const char *discovery = NULL; const xmlNode *attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE); add_node_attrs(attrs, node, TRUE, scheduler); if (crm_is_true(pe_node_attribute_raw(node, "standby"))) { crm_info("%s is in standby mode", pe__node_name(node)); node->details->standby = TRUE; } if (crm_is_true(pe_node_attribute_raw(node, "maintenance"))) { crm_info("%s is in maintenance mode", pe__node_name(node)); node->details->maintenance = TRUE; } discovery = pe_node_attribute_raw(node, XML_NODE_ATTR_RSC_DISCOVERY); if ((discovery != NULL) && !crm_is_true(discovery)) { pcmk__config_warn("Ignoring " XML_NODE_ATTR_RSC_DISCOVERY " attribute " "for %s because disabling resource discovery is " "not allowed for cluster nodes", pe__node_name(node)); } } /*! * \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,out] scheduler Scheduler data */ static void unpack_node_state(const xmlNode *state, pcmk_scheduler_t *scheduler) { const char *id = NULL; const char *uname = NULL; pcmk_node_t *this_node = NULL; id = crm_element_value(state, PCMK_XA_ID); if (id == NULL) { pcmk__config_err("Ignoring invalid " XML_CIB_TAG_STATE " entry without " PCMK_XA_ID); crm_log_xml_info(state, "missing-id"); return; } - uname = crm_element_value(state, XML_ATTR_UNAME); + uname = crm_element_value(state, PCMK_XA_UNAME); if (uname == NULL) { /* If a joining peer makes the cluster acquire the quorum from corosync * meanwhile it has not joined CPG membership of pacemaker-controld yet, * it's possible that the created node_state entry doesn't have an uname * yet. We should recognize the node as `pending` and wait for it to * join CPG. */ crm_trace("Handling " XML_CIB_TAG_STATE " entry with id=\"%s\" without " - XML_ATTR_UNAME, id); + PCMK_XA_UNAME, id); } this_node = pe_find_node_any(scheduler->nodes, id, uname); if (this_node == NULL) { pcmk__config_warn("Ignoring recorded node state for id=\"%s\" (%s) " "because it is no longer in the configuration", id, pcmk__s(uname, "uname unknown")); 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, scheduler); /* 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)", pe__node_name(this_node), id); determine_online_status(state, this_node, scheduler); if (!pcmk_is_set(scheduler->flags, pcmk_sched_quorate) && this_node->details->online && (scheduler->no_quorum_policy == pcmk_no_quorum_fence)) { /* Everything else should flow from this automatically * (at least until the scheduler becomes able to migrate off * healthy resources) */ pe_fence_node(scheduler, 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,out] scheduler Scheduler data * * \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(const xmlNode *status, bool fence, pcmk_scheduler_t *scheduler) { int rc = pcmk_rc_ok; // Loop through all node_state entries in CIB status for (const 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); + const char *uname = crm_element_value(state, PCMK_XA_UNAME); pcmk_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(scheduler->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. */ pcmk_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (rsc->role != pcmk_role_started) || (rsc->container->role != pcmk_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. */ pcmk_resource_t *rsc = this_node->details->remote_rsc; if ((rsc == NULL) || (!pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock) && (rsc->role != pcmk_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(scheduler->flags, pcmk_sched_fencing_enabled |pcmk_sched_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(scheduler, this_node); unpack_handle_remote_attrs(this_node, state, scheduler); } crm_trace("Unpacking resource history for %snode %s", (fence? "unseen " : ""), id); this_node->details->unpacked = TRUE; unpack_node_lrm(this_node, state, scheduler); 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, pcmk_scheduler_t *scheduler) { xmlNode *state = NULL; crm_trace("Beginning unpack"); if (scheduler->tickets == NULL) { scheduler->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, scheduler); } else if (pcmk__str_eq((const char *)state->name, XML_CIB_TAG_STATE, pcmk__str_none)) { unpack_node_state(state, scheduler); } } while (unpack_node_history(status, FALSE, scheduler) == 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(scheduler->flags, pcmk_sched_fencing_enabled), scheduler); /* Now that we know where resources are, we can schedule stops of containers * with failed bundle connections */ if (scheduler->stop_needed != NULL) { for (GList *item = scheduler->stop_needed; item; item = item->next) { pcmk_resource_t *container = item->data; pcmk_node_t *node = pe__current_node(container); if (node) { stop_action(container, node, FALSE); } } g_list_free(scheduler->stop_needed); scheduler->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 = scheduler->nodes; gIter != NULL; gIter = gIter->next) { pcmk_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, pcmk_role_stopped, "remote shutdown"); } if (!this_node->details->unpacked) { determine_remote_online_status(scheduler, this_node); } } return TRUE; } /*! * \internal * \brief Unpack node's time when it became a member at the cluster layer * * \param[in] node_state Node's node_state entry * \param[in,out] scheduler Scheduler data * * \return Epoch time when node became a cluster member * (or scheduler effective time for legacy entries) if a member, * 0 if not a member, or -1 if no valid information available */ static long long unpack_node_member(const xmlNode *node_state, pcmk_scheduler_t *scheduler) { const char *member_time = crm_element_value(node_state, PCMK__XA_IN_CCM); int member = 0; if (member_time == NULL) { return -1LL; } else if (crm_str_to_boolean(member_time, &member) == 1) { /* If in_ccm=0, we'll return 0 here. If in_ccm=1, either the entry was * recorded as a boolean for a DC < 2.1.7, or the node is pending * shutdown and has left the CPG, in which case it was set to 1 to avoid * fencing for PCMK_OPT_NODE_PENDING_TIMEOUT. * * We return the effective time for in_ccm=1 because what's important to * avoid fencing is that effective time minus this value is less than * the pending node timeout. */ return member? (long long) get_effective_time(scheduler) : 0LL; } else { long long when_member = 0LL; if ((pcmk__scan_ll(member_time, &when_member, 0LL) != pcmk_rc_ok) || (when_member < 0LL)) { crm_warn("Unrecognized value '%s' for " PCMK__XA_IN_CCM " in " XML_CIB_TAG_STATE " entry", member_time); return -1LL; } return when_member; } } /*! * \internal * \brief Unpack node's time when it became online in process group * * \param[in] node_state Node's node_state entry * * \return Epoch time when node became online in process group (or 0 if not * online, or 1 for legacy online entries) */ static long long unpack_node_online(const xmlNode *node_state) { const char *peer_time = crm_element_value(node_state, PCMK__XA_CRMD); // @COMPAT Entries recorded for DCs < 2.1.7 have "online" or "offline" if (pcmk__str_eq(peer_time, OFFLINESTATUS, pcmk__str_casei|pcmk__str_null_matches)) { return 0LL; } else if (pcmk__str_eq(peer_time, ONLINESTATUS, pcmk__str_casei)) { return 1LL; } else { long long when_online = 0LL; if ((pcmk__scan_ll(peer_time, &when_online, 0LL) != pcmk_rc_ok) || (when_online < 0)) { crm_warn("Unrecognized value '%s' for " PCMK__XA_CRMD " in " XML_CIB_TAG_STATE " entry, assuming offline", peer_time); return 0LL; } return when_online; } } /*! * \internal * \brief Unpack node attribute for user-requested fencing * * \param[in] node Node to check * \param[in] node_state Node's node_state entry in CIB status * * \return \c true if fencing has been requested for \p node, otherwise \c false */ static bool unpack_node_terminate(const pcmk_node_t *node, const xmlNode *node_state) { long long value = 0LL; int value_i = 0; const char *value_s = pe_node_attribute_raw(node, PCMK_NODE_ATTR_TERMINATE); // Value may be boolean or an epoch time if (crm_str_to_boolean(value_s, &value_i) == 1) { return (value_i != 0); } if (pcmk__scan_ll(value_s, &value, 0LL) == pcmk_rc_ok) { return (value > 0); } crm_warn("Ignoring unrecognized value '%s' for " PCMK_NODE_ATTR_TERMINATE "node attribute for %s", value_s, pe__node_name(node)); return false; } static gboolean determine_online_status_no_fencing(pcmk_scheduler_t *scheduler, const xmlNode *node_state, pcmk_node_t *this_node) { gboolean online = FALSE; const char *join = crm_element_value(node_state, PCMK__XA_JOIN); const char *exp_state = crm_element_value(node_state, PCMK__XA_EXPECTED); long long when_member = unpack_node_member(node_state, scheduler); long long when_online = unpack_node_online(node_state); if (when_member <= 0) { crm_trace("Node %s is %sdown", pe__node_name(this_node), ((when_member < 0)? "presumed " : "")); } else if (when_online > 0) { if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_casei)) { online = TRUE; } else { crm_debug("Node %s is not ready to run resources: %s", pe__node_name(this_node), join); } } else if (this_node->details->expected_up == FALSE) { crm_trace("Node %s controller is down: " "member@%lld online@%lld join=%s expected=%s", pe__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, "")); } else { /* mark it unclean */ pe_fence_node(scheduler, this_node, "peer is unexpectedly down", FALSE); crm_info("Node %s member@%lld online@%lld join=%s expected=%s", pe__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, "")); } return online; } /*! * \internal * \brief Check whether a node has taken too long to join controller group * * \param[in,out] scheduler Scheduler data * \param[in] node Node to check * \param[in] when_member Epoch time when node became a cluster member * \param[in] when_online Epoch time when node joined controller group * * \return true if node has been pending (on the way up) longer than * \c PCMK_OPT_NODE_PENDING_TIMEOUT, otherwise false * \note This will also update the cluster's recheck time if appropriate. */ static inline bool pending_too_long(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, long long when_member, long long when_online) { if ((scheduler->node_pending_timeout > 0) && (when_member > 0) && (when_online <= 0)) { // There is a timeout on pending nodes, and node is pending time_t timeout = when_member + scheduler->node_pending_timeout; if (get_effective_time(node->details->data_set) >= timeout) { return true; // Node has timed out } // Node is pending, but still has time pe__update_recheck_time(timeout, scheduler, "pending node timeout"); } return false; } static bool determine_online_status_fencing(pcmk_scheduler_t *scheduler, const xmlNode *node_state, pcmk_node_t *this_node) { bool termination_requested = unpack_node_terminate(this_node, node_state); const char *join = crm_element_value(node_state, PCMK__XA_JOIN); const char *exp_state = crm_element_value(node_state, PCMK__XA_EXPECTED); long long when_member = unpack_node_member(node_state, scheduler); long long when_online = unpack_node_online(node_state); /* - PCMK__XA_JOIN ::= member|down|pending|banned - PCMK__XA_EXPECTED ::= member|down @COMPAT with entries recorded for DCs < 2.1.7 - PCMK__XA_IN_CCM ::= true|false - PCMK__XA_CRMD ::= online|offline Since crm_feature_set 3.18.0 (pacemaker-2.1.7): - PCMK__XA_IN_CCM ::= |0 Since when node has been a cluster member. A value 0 of means the node is not a cluster member. - PCMK__XA_CRMD ::= |0 Since when peer has been online in CPG. A value 0 means the peer is offline in CPG. */ crm_trace("Node %s member@%lld online@%lld join=%s expected=%s%s", pe__node_name(this_node), when_member, when_online, pcmk__s(join, ""), pcmk__s(exp_state, ""), (termination_requested? " (termination requested)" : "")); if (this_node->details->shutdown) { crm_debug("%s is shutting down", pe__node_name(this_node)); /* Slightly different criteria since we can't shut down a dead peer */ return (when_online > 0); } if (when_member < 0) { pe_fence_node(scheduler, this_node, "peer has not been seen by the cluster", FALSE); return false; } if (pcmk__str_eq(join, CRMD_JOINSTATE_NACK, pcmk__str_none)) { pe_fence_node(scheduler, this_node, "peer failed Pacemaker membership criteria", FALSE); } else if (termination_requested) { if ((when_member <= 0) && (when_online <= 0) && pcmk__str_eq(join, CRMD_JOINSTATE_DOWN, pcmk__str_none)) { crm_info("%s was fenced as requested", pe__node_name(this_node)); return false; } pe_fence_node(scheduler, this_node, "fencing was requested", false); } else if (pcmk__str_eq(exp_state, CRMD_JOINSTATE_DOWN, pcmk__str_null_matches)) { if (pending_too_long(scheduler, this_node, when_member, when_online)) { pe_fence_node(scheduler, this_node, "peer pending timed out on joining the process group", FALSE); } else if ((when_member > 0) || (when_online > 0)) { crm_info("- %s is not ready to run resources", pe__node_name(this_node)); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { crm_trace("%s is down or still coming up", pe__node_name(this_node)); } } else if (when_member <= 0) { // Consider PCMK_OPT_PRIORITY_FENCING_DELAY for lost nodes pe_fence_node(scheduler, this_node, "peer is no longer part of the cluster", TRUE); } else if (when_online <= 0) { pe_fence_node(scheduler, this_node, "peer process is no longer available", FALSE); /* Everything is running at this point, now check join state */ } else if (pcmk__str_eq(join, CRMD_JOINSTATE_MEMBER, pcmk__str_none)) { crm_info("%s is active", pe__node_name(this_node)); } else if (pcmk__str_any_of(join, CRMD_JOINSTATE_PENDING, CRMD_JOINSTATE_DOWN, NULL)) { crm_info("%s is not ready to run resources", pe__node_name(this_node)); this_node->details->standby = TRUE; this_node->details->pending = TRUE; } else { pe_fence_node(scheduler, this_node, "peer was in an unknown state", FALSE); } return (when_member > 0); } static void determine_remote_online_status(pcmk_scheduler_t *scheduler, pcmk_node_t *this_node) { pcmk_resource_t *rsc = this_node->details->remote_rsc; pcmk_resource_t *container = NULL; pcmk_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 == pcmk_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 == pcmk_role_started) && (rsc->next_role == pcmk_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, pcmk_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, pcmk_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 == pcmk_role_stopped) || ((container != NULL) && (container->role == pcmk_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(const xmlNode *node_state, pcmk_node_t *this_node, pcmk_scheduler_t *scheduler) { gboolean online = FALSE; const char *exp_state = crm_element_value(node_state, PCMK__XA_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(scheduler->flags, pcmk_sched_fencing_enabled)) { online = determine_online_status_no_fencing(scheduler, node_state, this_node); } else { online = determine_online_status_fencing(scheduler, node_state, this_node); } if (online) { this_node->details->online = TRUE; } else { /* remove node from contention */ this_node->fixed = TRUE; // @COMPAT deprecated and unused this_node->weight = -INFINITY; } if (online && this_node->details->shutdown) { /* don't run resources here */ this_node->fixed = TRUE; // @COMPAT deprecated and unused this_node->weight = -INFINITY; } if (this_node->details->type == node_ping) { crm_info("%s is not a Pacemaker node", pe__node_name(this_node)); } else if (this_node->details->unclean) { pcmk__sched_warn("%s is unclean", pe__node_name(this_node)); } else if (this_node->details->online) { crm_info("%s is %s", pe__node_name(this_node), 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("%s is offline", pe__node_name(this_node)); } } /*! * \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 pcmk_resource_t * create_fake_resource(const char *rsc_id, const xmlNode *rsc_entry, pcmk_scheduler_t *scheduler) { pcmk_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, PCMK_XA_ID, rsc_id); crm_log_xml_debug(xml_rsc, "Orphan resource"); if (pe__unpack_resource(xml_rsc, &rsc, NULL, scheduler) != pcmk_rc_ok) { return NULL; } if (xml_contains_remote_node(xml_rsc)) { pcmk_node_t *node; crm_debug("Detected orphaned remote node %s", rsc_id); node = pe_find_node(scheduler->nodes, rsc_id); if (node == NULL) { node = pe_create_node(rsc_id, rsc_id, "remote", NULL, scheduler); } link_rsc2remotenode(scheduler, 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, PCMK__META_CONTAINER)) { /* This orphaned rsc needs to be mapped to a container. */ crm_trace("Detected orphaned container filler %s", rsc_id); pcmk__set_rsc_flags(rsc, pcmk_rsc_removed_filler); } pcmk__set_rsc_flags(rsc, pcmk_rsc_removed); scheduler->resources = g_list_append(scheduler->resources, rsc); return rsc; } /*! * \internal * \brief Create orphan instance for anonymous clone resource history * * \param[in,out] parent Clone resource that orphan will be added to * \param[in] rsc_id Orphan's resource ID * \param[in] node Where orphan is active (for logging only) * \param[in,out] scheduler Scheduler data * * \return Newly added orphaned instance of \p parent */ static pcmk_resource_t * create_anonymous_orphan(pcmk_resource_t *parent, const char *rsc_id, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { pcmk_resource_t *top = pe__create_clone_child(parent, scheduler); // find_rsc() because we might be a cloned group pcmk_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pcmk_rsc_match_clone_only); pcmk__rsc_debug(parent, "Created orphan %s for %s: %s on %s", top->id, parent->id, rsc_id, pe__node_name(node)); 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 \c PCMK_META_CLONE_MAX * instances); (3) a newly created orphan (that is, \c PCMK_META_CLONE_MAX * instances are already active). * * \param[in,out] scheduler Scheduler data * \param[in] node Node on which to check for instance * \param[in,out] parent Clone to check * \param[in] rsc_id Name of cloned resource in history (no instance) */ static pcmk_resource_t * find_anonymous_clone(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, pcmk_resource_t *parent, const char *rsc_id) { GList *rIter = NULL; pcmk_resource_t *rsc = NULL; pcmk_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, pcmk_rsc_unique)); // Check for active (or partially active, for cloned groups) instance pcmk__rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, pe__node_name(node), parent->id); for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) { GList *locations = NULL; pcmk_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 * PCMK_META_GLOBALLY_UNIQUE was flipped from true to false); and * (3) when we re-run calculations on the same scheduler data 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 (((pcmk_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, pcmk_rsc_match_clone_only); if (rsc) { /* If there are multiple instance history entries for an * anonymous clone in a single node's history (which can * happen if PCMK_META_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, pe__node_name(node)); skip_inactive = TRUE; rsc = NULL; } else { pcmk__rsc_trace(parent, "Resource %s, active", rsc->id); } } } g_list_free(locations); } else { pcmk__rsc_trace(parent, "Resource %s, skip inactive", child->id); if (!skip_inactive && !inactive_instance && !pcmk_is_set(child->flags, pcmk_rsc_blocked)) { // Remember one inactive instance in case we don't find active inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL, pcmk_rsc_match_clone_only); /* ... 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)) { pcmk__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, pcmk_rsc_needs_fencing) && (!node->details->online || node->details->unclean) && !pe__is_guest_node(node) && !pe__is_universal_clone(parent, scheduler)) { rsc = NULL; } if (rsc == NULL) { rsc = create_anonymous_orphan(parent, rsc_id, node, scheduler); pcmk__rsc_trace(parent, "Resource %s, orphan", rsc->id); } return rsc; } static pcmk_resource_t * unpack_find_resource(pcmk_scheduler_t *scheduler, const pcmk_node_t *node, const char *rsc_id) { pcmk_resource_t *rsc = NULL; pcmk_resource_t *parent = NULL; crm_trace("looking for %s", rsc_id); rsc = pe_find_resource(scheduler->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 PCMK_META_CLONE_MAX=0, * we create a single :0 orphan to match against here. */ char *clone0_id = clone_zero(rsc_id); pcmk_resource_t *clone0 = pe_find_resource(scheduler->resources, clone0_id); if (clone0 && !pcmk_is_set(clone0->flags, pcmk_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 > pcmk_rsc_variant_primitive) { 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(scheduler, 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); pcmk__rsc_debug(rsc, "Internally renamed %s on %s to %s%s", rsc_id, pe__node_name(node), rsc->id, pcmk_is_set(rsc->flags, pcmk_rsc_removed)? " (ORPHAN)" : ""); } return rsc; } static pcmk_resource_t * process_orphan_resource(const xmlNode *rsc_entry, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { pcmk_resource_t *rsc = NULL; const char *rsc_id = crm_element_value(rsc_entry, PCMK_XA_ID); crm_debug("Detected orphan resource %s on %s", rsc_id, pe__node_name(node)); rsc = create_fake_resource(rsc_id, rsc_entry, scheduler); if (rsc == NULL) { return NULL; } if (!pcmk_is_set(scheduler->flags, pcmk_sched_stop_removed_resources)) { pcmk__clear_rsc_flags(rsc, pcmk_rsc_managed); } else { CRM_CHECK(rsc != NULL, return NULL); pcmk__rsc_trace(rsc, "Added orphan %s", rsc->id); resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__", scheduler); } return rsc; } static void process_rsc_state(pcmk_resource_t *rsc, pcmk_node_t *node, enum action_fail_response on_fail) { pcmk_node_t *tmpnode = NULL; char *reason = NULL; enum action_fail_response save_on_fail = pcmk_on_fail_ignore; CRM_ASSERT(rsc); pcmk__rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s", rsc->id, role2text(rsc->role), pe__node_name(node), fail2text(on_fail)); /* process current state */ if (rsc->role != pcmk_role_unknown) { pcmk_resource_t *iter = rsc; while (iter) { if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) { pcmk_node_t *n = pe__copy_node(node); pcmk__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), pe__node_name(n)); g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n); } if (pcmk_is_set(iter->flags, pcmk_rsc_unique)) { break; } iter = iter->parent; } } /* If a managed resource is believed to be running, but node is down ... */ if ((rsc->role > pcmk_role_stopped) && node->details->online == FALSE && node->details->maintenance == FALSE && pcmk_is_set(rsc->flags, pcmk_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)) { pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); should_fence = TRUE; } else if (pcmk_is_set(rsc->cluster->flags, pcmk_sched_fencing_enabled)) { if (pe__is_remote_node(node) && node->details->remote_rsc && !pcmk_is_set(node->details->remote_rsc->flags, pcmk_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(rsc->cluster, 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 = pcmk_on_fail_ignore; } switch (on_fail) { case pcmk_on_fail_ignore: /* nothing to do */ break; case pcmk_on_fail_demote: pcmk__set_rsc_flags(rsc, pcmk_rsc_failed); demote_action(rsc, node, FALSE); break; case pcmk_on_fail_fence_node: /* 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(rsc->cluster, node, reason, FALSE); free(reason); break; case pcmk_on_fail_standby_node: node->details->standby = TRUE; node->details->standby_onfail = TRUE; break; case pcmk_on_fail_block: /* is_managed == FALSE will prevent any * actions being sent for the resource */ pcmk__clear_rsc_flags(rsc, pcmk_rsc_managed); pcmk__set_rsc_flags(rsc, pcmk_rsc_blocked); break; case pcmk_on_fail_ban: /* make sure it comes up somewhere else * or not at all */ resource_location(rsc, node, -INFINITY, "__action_migration_auto__", rsc->cluster); break; case pcmk_on_fail_stop: pe__set_next_role(rsc, pcmk_role_stopped, "on-fail=stop"); break; case pcmk_on_fail_restart: if ((rsc->role != pcmk_role_stopped) && (rsc->role != pcmk_role_unknown)) { pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); stop_action(rsc, node, FALSE); } break; case pcmk_on_fail_restart_container: pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); 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. */ rsc->cluster->stop_needed = g_list_prepend(rsc->cluster->stop_needed, rsc->container); } else if (rsc->container) { stop_action(rsc->container, node, FALSE); } else if ((rsc->role != pcmk_role_stopped) && (rsc->role != pcmk_role_unknown)) { stop_action(rsc, node, FALSE); } break; case pcmk_on_fail_reset_remote: pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); if (pcmk_is_set(rsc->cluster->flags, pcmk_sched_fencing_enabled)) { tmpnode = NULL; if (rsc->is_remote_node) { tmpnode = pe_find_node(rsc->cluster->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(rsc->cluster, tmpnode, "remote connection is unrecoverable", FALSE); } } /* require the stop action regardless if fencing is occurring or not. */ if (rsc->role > pcmk_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, pcmk_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, pcmk_rsc_failed) && rsc->is_remote_node) { tmpnode = pe_find_node(rsc->cluster->nodes, rsc->id); if (tmpnode && tmpnode->details->unclean) { tmpnode->details->unseen = FALSE; } } if ((rsc->role != pcmk_role_stopped) && (rsc->role != pcmk_role_unknown)) { if (pcmk_is_set(rsc->flags, pcmk_rsc_removed)) { if (pcmk_is_set(rsc->flags, pcmk_rsc_managed)) { pcmk__config_warn("Detected active orphan %s running on %s", rsc->id, pe__node_name(node)); } else { pcmk__config_warn("Resource '%s' must be stopped manually on " "%s because cluster is configured not to " "stop active orphans", rsc->id, pe__node_name(node)); } } native_add_running(rsc, node, rsc->cluster, (save_on_fail != pcmk_on_fail_ignore)); switch (on_fail) { case pcmk_on_fail_ignore: break; case pcmk_on_fail_demote: case pcmk_on_fail_block: pcmk__set_rsc_flags(rsc, pcmk_rsc_failed); break; default: pcmk__set_rsc_flags(rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); 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 */ pcmk__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, PCMK_ACTION_STOP, FALSE); GList *gIter = possible_matches; for (; gIter != NULL; gIter = gIter->next) { pcmk_action_t *stop = (pcmk_action_t *) gIter->data; pcmk__set_action_flags(stop, pcmk_action_optional); } g_list_free(possible_matches); } /* A successful stop after migrate_to on the migration source doesn't make * the partially migrated resource stopped on the migration target. */ if ((rsc->role == pcmk_role_stopped) && rsc->partial_migration_source && rsc->partial_migration_source->details == node->details && rsc->partial_migration_target && rsc->running_on) { rsc->role = pcmk_role_started; } } /* create active recurring operations as optional */ static void process_recurring(pcmk_node_t *node, pcmk_resource_t *rsc, int start_index, int stop_index, GList *sorted_op_list, pcmk_scheduler_t *scheduler) { int counter = -1; const char *task = NULL; const char *status = NULL; GList *gIter = sorted_op_list; CRM_ASSERT(rsc); pcmk__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) { pcmk__rsc_trace(rsc, "Skipping %s on %s: node is offline", rsc->id, pe__node_name(node)); break; /* Need to check if there's a monitor for role="Stopped" */ } else if (start_index < stop_index && counter <= stop_index) { pcmk__rsc_trace(rsc, "Skipping %s on %s: resource is not active", id, pe__node_name(node)); continue; } else if (counter < start_index) { pcmk__rsc_trace(rsc, "Skipping %s on %s: old %d", id, pe__node_name(node), counter); continue; } crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms == 0) { pcmk__rsc_trace(rsc, "Skipping %s on %s: non-recurring", id, pe__node_name(node)); continue; } status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS); if (pcmk__str_eq(status, "-1", pcmk__str_casei)) { pcmk__rsc_trace(rsc, "Skipping %s on %s: status", id, pe__node_name(node)); continue; } task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK); /* create the action */ key = pcmk__op_key(rsc->id, task, interval_ms); pcmk__rsc_trace(rsc, "Creating %s on %s", key, pe__node_name(node)); custom_action(rsc, key, task, node, TRUE, scheduler); } } void calculate_active_ops(const 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; *stop_index = -1; *start_index = -1; for (const GList *iter = sorted_op_list; iter != NULL; iter = iter->next) { const xmlNode *rsc_op = (const xmlNode *) iter->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, PCMK_ACTION_STOP, pcmk__str_casei) && pcmk__str_eq(status, "0", pcmk__str_casei)) { *stop_index = counter; } else if (pcmk__strcase_any_of(task, PCMK_ACTION_START, PCMK_ACTION_MIGRATE_FROM, NULL)) { *start_index = counter; } else if ((implied_monitor_start <= *stop_index) && pcmk__str_eq(task, PCMK_ACTION_MONITOR, 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, PCMK_ACTION_PROMOTE, PCMK_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(const xmlNode *rsc_entry, pcmk_resource_t *rsc, const pcmk_node_t *node, pcmk_scheduler_t *scheduler) { time_t lock_time = 0; // When lock started (i.e. node shutdown time) if ((crm_element_value_epoch(rsc_entry, PCMK_OPT_SHUTDOWN_LOCK, &lock_time) == pcmk_ok) && (lock_time != 0)) { if ((scheduler->shutdown_lock > 0) && (get_effective_time(scheduler) > (lock_time + scheduler->shutdown_lock))) { pcmk__rsc_info(rsc, "Shutdown lock for %s on %s expired", rsc->id, pe__node_name(node)); pe__clear_resource_history(rsc, node); } else { /* @COMPAT I don't like breaking const signatures, but * rsc->lock_node should really be const -- we just can't change it * until the next API compatibility break. */ rsc->lock_node = (pcmk_node_t *) node; rsc->lock_time = lock_time; } } } /*! * \internal * \brief Unpack one lrm_resource entry from a node's CIB status * * \param[in,out] node Node whose status is being unpacked * \param[in] rsc_entry lrm_resource XML being unpacked * \param[in,out] scheduler Scheduler data * * \return Resource corresponding to the entry, or NULL if no operation history */ static pcmk_resource_t * unpack_lrm_resource(pcmk_node_t *node, const xmlNode *lrm_resource, pcmk_scheduler_t *scheduler) { GList *gIter = NULL; int stop_index = -1; int start_index = -1; enum rsc_role_e req_role = pcmk_role_unknown; const char *rsc_id = ID(lrm_resource); pcmk_resource_t *rsc = NULL; GList *op_list = NULL; GList *sorted_op_list = NULL; xmlNode *rsc_op = NULL; xmlNode *last_failure = NULL; enum action_fail_response on_fail = pcmk_on_fail_ignore; enum rsc_role_e saved_role = pcmk_role_unknown; if (rsc_id == NULL) { pcmk__config_err("Ignoring invalid " XML_LRM_TAG_RESOURCE " entry: No " PCMK_XA_ID); crm_log_xml_info(lrm_resource, "missing-id"); return NULL; } crm_trace("Unpacking " XML_LRM_TAG_RESOURCE " for %s on %s", rsc_id, pe__node_name(node)); // 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(scheduler->flags, pcmk_sched_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(scheduler, node, rsc_id); 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, scheduler); } } CRM_ASSERT(rsc != NULL); // Check whether the resource is "shutdown-locked" to this node if (pcmk_is_set(scheduler->flags, pcmk_sched_shutdown_lock)) { unpack_shutdown_lock(lrm_resource, rsc, node, scheduler); } /* process operations */ saved_role = rsc->role; rsc->role = pcmk_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; unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail); } /* 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, scheduler); /* no need to free the contents */ g_list_free(sorted_op_list); process_rsc_state(rsc, node, on_fail); if (get_target_role(rsc, &req_role)) { if ((rsc->next_role == pcmk_role_unknown) || (req_role < rsc->next_role)) { pe__set_next_role(rsc, req_role, PCMK_META_TARGET_ROLE); } else if (req_role > rsc->next_role) { pcmk__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(const xmlNode *lrm_rsc_list, pcmk_scheduler_t *scheduler) { for (const xmlNode *rsc_entry = pcmk__xe_first_child(lrm_rsc_list); rsc_entry != NULL; rsc_entry = pcmk__xe_next(rsc_entry)) { pcmk_resource_t *rsc; pcmk_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, PCMK__META_CONTAINER); rsc_id = crm_element_value(rsc_entry, PCMK_XA_ID); if (container_id == NULL || rsc_id == NULL) { continue; } container = pe_find_resource(scheduler->resources, container_id); if (container == NULL) { continue; } rsc = pe_find_resource(scheduler->resources, rsc_id); if ((rsc == NULL) || (rsc->container != NULL) || !pcmk_is_set(rsc->flags, pcmk_rsc_removed_filler)) { continue; } pcmk__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,out] node Node whose status is being unpacked * \param[in] xml CIB node state XML * \param[in,out] scheduler Scheduler data */ static void unpack_node_lrm(pcmk_node_t *node, const xmlNode *xml, pcmk_scheduler_t *scheduler) { 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 (const xmlNode *rsc_entry = first_named_child(xml, XML_LRM_TAG_RESOURCE); rsc_entry != NULL; rsc_entry = crm_next_same_xml(rsc_entry)) { pcmk_resource_t *rsc = unpack_lrm_resource(node, rsc_entry, scheduler); if ((rsc != NULL) && pcmk_is_set(rsc->flags, pcmk_rsc_removed_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, scheduler); } } static void set_active(pcmk_resource_t *rsc) { const pcmk_resource_t *top = pe__const_top_resource(rsc, false); if (top && pcmk_is_set(top->flags, pcmk_rsc_promotable)) { rsc->role = pcmk_role_unpromoted; } else { rsc->role = pcmk_role_started; } } static void set_node_score(gpointer key, gpointer value, gpointer user_data) { pcmk_node_t *node = value; int *score = user_data; node->weight = *score; } #define XPATH_NODE_STATE "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE #define SUB_XPATH_LRM_RESOURCE "/" XML_CIB_TAG_LRM \ "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE #define SUB_XPATH_LRM_RSC_OP "/" XML_LRM_TAG_RSC_OP static xmlNode * find_lrm_op(const char *resource, const char *op, const char *node, const char *source, int target_rc, pcmk_scheduler_t *scheduler) { GString *xpath = NULL; xmlNode *xml = NULL; CRM_CHECK((resource != NULL) && (op != NULL) && (node != NULL), return NULL); xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, - XPATH_NODE_STATE "[@" XML_ATTR_UNAME "='", node, "']" + XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node, "']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", resource, "']" SUB_XPATH_LRM_RSC_OP "[@" XML_LRM_ATTR_TASK "='", op, "'", NULL); /* Need to check against transition_magic too? */ if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_TO) == 0)) { pcmk__g_strcat(xpath, " and @" XML_LRM_ATTR_MIGRATE_TARGET "='", source, "']", NULL); } else if ((source != NULL) && (strcmp(op, PCMK_ACTION_MIGRATE_FROM) == 0)) { pcmk__g_strcat(xpath, " and @" XML_LRM_ATTR_MIGRATE_SOURCE "='", source, "']", NULL); } else { g_string_append_c(xpath, ']'); } xml = get_xpath_object((const char *) xpath->str, scheduler->input, LOG_DEBUG); g_string_free(xpath, TRUE); 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, pcmk_scheduler_t *scheduler) { GString *xpath = NULL; xmlNode *xml = NULL; CRM_CHECK((rsc_id != NULL) && (node_name != NULL), return NULL); xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, - XPATH_NODE_STATE "[@" XML_ATTR_UNAME "='", node_name, "']" + XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node_name, "']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", rsc_id, "']", NULL); xml = get_xpath_object((const char *) xpath->str, scheduler->input, LOG_DEBUG); g_string_free(xpath, TRUE); return xml; } /*! * \internal * \brief Check whether a resource has no completed action history on a node * * \param[in,out] rsc Resource to check * \param[in] node_name Node to check * * \return true if \p rsc_id is unknown on \p node_name, otherwise false */ static bool unknown_on_node(pcmk_resource_t *rsc, const char *node_name) { bool result = false; xmlXPathObjectPtr search; GString *xpath = g_string_sized_new(256); pcmk__g_strcat(xpath, - XPATH_NODE_STATE "[@" XML_ATTR_UNAME "='", node_name, "']" + XPATH_NODE_STATE "[@" PCMK_XA_UNAME "='", node_name, "']" SUB_XPATH_LRM_RESOURCE "[@" PCMK_XA_ID "='", rsc->id, "']" SUB_XPATH_LRM_RSC_OP "[@" XML_LRM_ATTR_RC "!='193']", NULL); search = xpath_search(rsc->cluster->input, (const char *) xpath->str); result = (numXpathResults(search) == 0); freeXpathObject(search); g_string_free(xpath, TRUE); return result; } /*! * \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] same_node Whether the operations are on the same node * \param[in,out] scheduler Scheduler data * * \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, const xmlNode *xml_op, bool same_node, pcmk_scheduler_t *scheduler) { /* Any probe/monitor operation on the node indicating it was not running * there */ xmlNode *monitor = find_lrm_op(rsc_id, PCMK_ACTION_MONITOR, node_name, NULL, PCMK_OCF_NOT_RUNNING, scheduler); return (monitor && pe__is_newer_op(monitor, xml_op, same_node) > 0); } /*! * \brief Check whether any non-monitor operation 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 non-monitor is being compared to * \param[in] same_node Whether the operations are on the same node * \param[in,out] scheduler Scheduler data * * \return true if such a operation happened after event, false otherwise */ static bool non_monitor_after(const char *rsc_id, const char *node_name, const xmlNode *xml_op, bool same_node, pcmk_scheduler_t *scheduler) { xmlNode *lrm_resource = NULL; lrm_resource = find_lrm_resource(rsc_id, node_name, scheduler); if (lrm_resource == NULL) { return false; } for (xmlNode *op = first_named_child(lrm_resource, XML_LRM_TAG_RSC_OP); op != NULL; op = crm_next_same_xml(op)) { const char * task = NULL; if (op == xml_op) { continue; } task = crm_element_value(op, XML_LRM_ATTR_TASK); if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_STOP, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, NULL) && pe__is_newer_op(op, xml_op, same_node) > 0) { return true; } } return false; } /*! * \brief Check whether the resource has newer state on a node after a migration * attempt * * \param[in] rsc_id Resource being checked * \param[in] node_name Node being checked * \param[in] migrate_to Any migrate_to event that is being compared to * \param[in] migrate_from Any migrate_from event that is being compared to * \param[in,out] scheduler Scheduler data * * \return true if such a operation happened after event, false otherwise */ static bool newer_state_after_migrate(const char *rsc_id, const char *node_name, const xmlNode *migrate_to, const xmlNode *migrate_from, pcmk_scheduler_t *scheduler) { const xmlNode *xml_op = migrate_to; const char *source = NULL; const char *target = NULL; bool same_node = false; if (migrate_from) { xml_op = migrate_from; } source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE); target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET); /* It's preferred to compare to the migrate event on the same node if * existing, since call ids are more reliable. */ if (pcmk__str_eq(node_name, target, pcmk__str_casei)) { if (migrate_from) { xml_op = migrate_from; same_node = true; } else { xml_op = migrate_to; } } else if (pcmk__str_eq(node_name, source, pcmk__str_casei)) { if (migrate_to) { xml_op = migrate_to; same_node = true; } else { xml_op = migrate_from; } } /* If there's any newer non-monitor operation on the node, or any newer * probe/monitor operation on the node indicating it was not running there, * the migration events potentially no longer matter for the node. */ return non_monitor_after(rsc_id, node_name, xml_op, same_node, scheduler) || monitor_not_running_after(rsc_id, node_name, xml_op, same_node, scheduler); } /*! * \internal * \brief Parse migration source and target node names from history entry * * \param[in] entry Resource history entry for a migration action * \param[in] source_node If not NULL, source must match this node * \param[in] target_node If not NULL, target must match this node * \param[out] source_name Where to store migration source node name * \param[out] target_name Where to store migration target node name * * \return Standard Pacemaker return code */ static int get_migration_node_names(const xmlNode *entry, const pcmk_node_t *source_node, const pcmk_node_t *target_node, const char **source_name, const char **target_name) { *source_name = crm_element_value(entry, XML_LRM_ATTR_MIGRATE_SOURCE); *target_name = crm_element_value(entry, XML_LRM_ATTR_MIGRATE_TARGET); if ((*source_name == NULL) || (*target_name == NULL)) { pcmk__config_err("Ignoring resource history entry %s without " XML_LRM_ATTR_MIGRATE_SOURCE " and " XML_LRM_ATTR_MIGRATE_TARGET, ID(entry)); return pcmk_rc_unpack_error; } if ((source_node != NULL) && !pcmk__str_eq(*source_name, source_node->details->uname, pcmk__str_casei|pcmk__str_null_matches)) { pcmk__config_err("Ignoring resource history entry %s because " XML_LRM_ATTR_MIGRATE_SOURCE "='%s' does not match %s", ID(entry), *source_name, pe__node_name(source_node)); return pcmk_rc_unpack_error; } if ((target_node != NULL) && !pcmk__str_eq(*target_name, target_node->details->uname, pcmk__str_casei|pcmk__str_null_matches)) { pcmk__config_err("Ignoring resource history entry %s because " XML_LRM_ATTR_MIGRATE_TARGET "='%s' does not match %s", ID(entry), *target_name, pe__node_name(target_node)); return pcmk_rc_unpack_error; } return pcmk_rc_ok; } /* * \internal * \brief Add a migration source to a resource's list of dangling migrations * * If the migrate_to and migrate_from actions in a live migration both * succeeded, but there is no stop on the source, the migration is considered * "dangling." Add the source to the resource's dangling migration list, which * will be used to schedule a stop on the source without affecting the target. * * \param[in,out] rsc Resource involved in migration * \param[in] node Migration source */ static void add_dangling_migration(pcmk_resource_t *rsc, const pcmk_node_t *node) { pcmk__rsc_trace(rsc, "Dangling migration of %s requires stop on %s", rsc->id, pe__node_name(node)); rsc->role = pcmk_role_stopped; rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, (gpointer) node); } /*! * \internal * \brief Update resource role etc. after a successful migrate_to action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_to_success(struct action_history *history) { /* A complete migration sequence is: * 1. migrate_to on source node (which succeeded if we get to this function) * 2. migrate_from on target node * 3. stop on source node * * If no migrate_from has happened, the migration is considered to be * "partial". If the migrate_from succeeded but no stop has happened, the * migration is considered to be "dangling". * * If a successful migrate_to and stop have happened on the source node, we * still need to check for a partial migration, due to scenarios (easier to * produce with batch-limit=1) like: * * - A resource is migrating from node1 to node2, and a migrate_to is * initiated for it on node1. * * - node2 goes into standby mode while the migrate_to is pending, which * aborts the transition. * * - Upon completion of the migrate_to, a new transition schedules a stop * on both nodes and a start on node1. * * - If the new transition is aborted for any reason while the resource is * stopping on node1, the transition after that stop completes will see * the migrate_to and stop on the source, but it's still a partial * migration, and the resource must be stopped on node2 because it is * potentially active there due to the migrate_to. * * We also need to take into account that either node's history may be * cleared at any point in the migration process. */ int from_rc = PCMK_OCF_OK; int from_status = PCMK_EXEC_PENDING; pcmk_node_t *target_node = NULL; xmlNode *migrate_from = NULL; const char *source = NULL; const char *target = NULL; bool source_newer_op = false; bool target_newer_state = false; bool active_on_target = false; // Get source and target node names from XML if (get_migration_node_names(history->xml, history->node, NULL, &source, &target) != pcmk_rc_ok) { return; } // Check for newer state on the source source_newer_op = non_monitor_after(history->rsc->id, source, history->xml, true, history->rsc->cluster); // Check for a migrate_from action from this source on the target migrate_from = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_FROM, target, source, -1, history->rsc->cluster); if (migrate_from != NULL) { if (source_newer_op) { /* There's a newer non-monitor operation on the source and a * migrate_from on the target, so this migrate_to is irrelevant to * the resource's state. */ return; } crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc); crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status); } /* If the resource has newer state on both the source and target after the * migration events, this migrate_to is irrelevant to the resource's state. */ target_newer_state = newer_state_after_migrate(history->rsc->id, target, history->xml, migrate_from, history->rsc->cluster); if (source_newer_op && target_newer_state) { return; } /* Check for dangling migration (migrate_from succeeded but stop not done). * We know there's no stop because we already returned if the target has a * migrate_from and the source has any newer non-monitor operation. */ if ((from_rc == PCMK_OCF_OK) && (from_status == PCMK_EXEC_DONE)) { add_dangling_migration(history->rsc, history->node); return; } /* Without newer state, this migrate_to implies the resource is active. * (Clones are not allowed to migrate, so role can't be promoted.) */ history->rsc->role = pcmk_role_started; target_node = pe_find_node(history->rsc->cluster->nodes, target); active_on_target = !target_newer_state && (target_node != NULL) && target_node->details->online; if (from_status != PCMK_EXEC_PENDING) { // migrate_from failed on target if (active_on_target) { native_add_running(history->rsc, target_node, history->rsc->cluster, TRUE); } else { // Mark resource as failed, require recovery, and prevent migration pcmk__set_rsc_flags(history->rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_migratable); } return; } // The migrate_from is pending, complete but erased, or to be scheduled /* If there is no history at all for the resource on an online target, then * it was likely cleaned. Just return, and we'll schedule a probe. Once we * have the probe result, it will be reflected in target_newer_state. */ if ((target_node != NULL) && target_node->details->online && unknown_on_node(history->rsc, target)) { return; } if (active_on_target) { pcmk_node_t *source_node = pe_find_node(history->rsc->cluster->nodes, source); native_add_running(history->rsc, target_node, history->rsc->cluster, FALSE); if ((source_node != NULL) && 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. */ history->rsc->partial_migration_target = target_node; history->rsc->partial_migration_source = source_node; } } else if (!source_newer_op) { // Mark resource as failed, require recovery, and prevent migration pcmk__set_rsc_flags(history->rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_migratable); } } /*! * \internal * \brief Update resource role etc. after a failed migrate_to action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_to_failure(struct action_history *history) { xmlNode *target_migrate_from = NULL; const char *source = NULL; const char *target = NULL; // Get source and target node names from XML if (get_migration_node_names(history->xml, history->node, NULL, &source, &target) != pcmk_rc_ok) { 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. */ history->rsc->role = pcmk_role_started; // Check for migrate_from on the target target_migrate_from = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_FROM, target, source, PCMK_OCF_OK, history->rsc->cluster); 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(history->rsc, target) /* If the resource has newer state on the target after the migration * events, this migrate_to no longer matters for the target. */ && !newer_state_after_migrate(history->rsc->id, target, history->xml, target_migrate_from, history->rsc->cluster)) { /* The resource has no newer state on the target, so assume it's still * active there. * (if it is up). */ pcmk_node_t *target_node = pe_find_node(history->rsc->cluster->nodes, target); if (target_node && target_node->details->online) { native_add_running(history->rsc, target_node, history->rsc->cluster, FALSE); } } else if (!non_monitor_after(history->rsc->id, source, history->xml, true, history->rsc->cluster)) { /* We know the resource has newer state on the target, but this * migrate_to still matters for the source as long as there's no newer * non-monitor operation there. */ // Mark node as having dangling migration so we can force a stop later history->rsc->dangling_migrations = g_list_prepend(history->rsc->dangling_migrations, (gpointer) history->node); } } /*! * \internal * \brief Update resource role etc. after a failed migrate_from action * * \param[in,out] history Parsed action result history */ static void unpack_migrate_from_failure(struct action_history *history) { xmlNode *source_migrate_to = NULL; const char *source = NULL; const char *target = NULL; // Get source and target node names from XML if (get_migration_node_names(history->xml, NULL, history->node, &source, &target) != pcmk_rc_ok) { 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. */ history->rsc->role = pcmk_role_started; // Check for a migrate_to on the source source_migrate_to = find_lrm_op(history->rsc->id, PCMK_ACTION_MIGRATE_TO, source, target, PCMK_OCF_OK, history->rsc->cluster); if (/* If the resource state is unknown on the source, 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(history->rsc, source) /* If the resource has newer state on the source after the migration * events, this migrate_from no longer matters for the source. */ && !newer_state_after_migrate(history->rsc->id, source, source_migrate_to, history->xml, history->rsc->cluster)) { /* The resource has no newer state on the source, so assume it's still * active there (if it is up). */ pcmk_node_t *source_node = pe_find_node(history->rsc->cluster->nodes, source); if (source_node && source_node->details->online) { native_add_running(history->rsc, source_node, history->rsc->cluster, TRUE); } } } /*! * \internal * \brief Add an action to cluster's list of failed actions * * \param[in,out] history Parsed action result history */ static void record_failed_op(struct action_history *history) { if (!(history->node->details->online)) { return; } for (const xmlNode *xIter = history->rsc->cluster->failed->children; xIter != NULL; xIter = xIter->next) { const char *key = pe__xe_history_key(xIter); - const char *uname = crm_element_value(xIter, XML_ATTR_UNAME); + const char *uname = crm_element_value(xIter, PCMK_XA_UNAME); if (pcmk__str_eq(history->key, key, pcmk__str_none) && pcmk__str_eq(uname, history->node->details->uname, pcmk__str_casei)) { crm_trace("Skipping duplicate entry %s on %s", history->key, pe__node_name(history->node)); return; } } crm_trace("Adding entry for %s on %s to failed action list", history->key, pe__node_name(history->node)); - crm_xml_add(history->xml, XML_ATTR_UNAME, history->node->details->uname); + crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->details->uname); crm_xml_add(history->xml, XML_LRM_ATTR_RSCID, history->rsc->id); add_node_copy(history->rsc->cluster->failed, history->xml); } static char * last_change_str(const xmlNode *xml_op) { time_t when; char *result = NULL; if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE, &when) == pcmk_ok) { char *when_s = pcmk__epoch2str(&when, 0); const char *p = strchr(when_s, ' '); // Skip day of week to make message shorter if ((p != NULL) && (*(++p) != '\0')) { result = strdup(p); CRM_ASSERT(result != NULL); } free(when_s); } if (result == NULL) { result = strdup("unknown time"); CRM_ASSERT(result != NULL); } return result; } /*! * \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 pcmk_on_fail_demote: switch (second) { case pcmk_on_fail_ignore: return 1; case pcmk_on_fail_demote: return 0; default: return -1; } break; case pcmk_on_fail_reset_remote: switch (second) { case pcmk_on_fail_ignore: case pcmk_on_fail_demote: case pcmk_on_fail_restart: return 1; case pcmk_on_fail_reset_remote: return 0; default: return -1; } break; case pcmk_on_fail_restart_container: switch (second) { case pcmk_on_fail_ignore: case pcmk_on_fail_demote: case pcmk_on_fail_restart: case pcmk_on_fail_reset_remote: return 1; case pcmk_on_fail_restart_container: return 0; default: return -1; } break; default: break; } switch (second) { case pcmk_on_fail_demote: return (first == pcmk_on_fail_ignore)? -1 : 1; case pcmk_on_fail_reset_remote: switch (first) { case pcmk_on_fail_ignore: case pcmk_on_fail_demote: case pcmk_on_fail_restart: return -1; default: return 1; } break; case pcmk_on_fail_restart_container: switch (first) { case pcmk_on_fail_ignore: case pcmk_on_fail_demote: case pcmk_on_fail_restart: case pcmk_on_fail_reset_remote: return -1; default: return 1; } break; default: break; } return first - second; } /*! * \internal * \brief Ban a resource (or its clone if an anonymous instance) from all nodes * * \param[in,out] rsc Resource to ban */ static void ban_from_all_nodes(pcmk_resource_t *rsc) { int score = -INFINITY; pcmk_resource_t *fail_rsc = rsc; if (fail_rsc->parent != NULL) { pcmk_resource_t *parent = uber_parent(fail_rsc); if (pe_rsc_is_anon_clone(parent)) { /* For anonymous clones, if an operation with on-fail=stop fails for * any instance, the entire clone must stop. */ fail_rsc = parent; } } // Ban the resource from all nodes crm_notice("%s will not be started under current conditions", fail_rsc->id); if (fail_rsc->allowed_nodes != NULL) { g_hash_table_destroy(fail_rsc->allowed_nodes); } fail_rsc->allowed_nodes = pe__node_list2table(rsc->cluster->nodes); g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score); } /*! * \internal * \brief Get configured failure handling and role after failure for an action * * \param[in,out] history Unpacked action history entry * \param[out] on_fail Where to set configured failure handling * \param[out] fail_role Where to set to role after failure */ static void unpack_failure_handling(struct action_history *history, enum action_fail_response *on_fail, enum rsc_role_e *fail_role) { xmlNode *config = pcmk__find_action_config(history->rsc, history->task, history->interval_ms, true); GHashTable *meta = pcmk__unpack_action_meta(history->rsc, history->node, history->task, history->interval_ms, config); const char *on_fail_str = g_hash_table_lookup(meta, XML_OP_ATTR_ON_FAIL); *on_fail = pcmk__parse_on_fail(history->rsc, history->task, history->interval_ms, on_fail_str); *fail_role = pcmk__role_after_failure(history->rsc, history->task, *on_fail, meta); g_hash_table_destroy(meta); } /*! * \internal * \brief Update resource role, failure handling, etc., after a failed action * * \param[in,out] history Parsed action result history * \param[in] config_on_fail Action failure handling from configuration * \param[in] fail_role Resource's role after failure of this action * \param[out] last_failure This will be set to the history XML * \param[in,out] on_fail Actual handling of action result */ static void unpack_rsc_op_failure(struct action_history *history, enum action_fail_response config_on_fail, enum rsc_role_e fail_role, xmlNode **last_failure, enum action_fail_response *on_fail) { bool is_probe = false; char *last_change_s = NULL; *last_failure = history->xml; is_probe = pcmk_xe_is_probe(history->xml); last_change_s = last_change_str(history->xml); if (!pcmk_is_set(history->rsc->cluster->flags, pcmk_sched_symmetric_cluster) && (history->exit_status == PCMK_OCF_NOT_INSTALLED)) { crm_trace("Unexpected result (%s%s%s) was recorded for " "%s of %s on %s at %s " CRM_XS " exit-status=%d id=%s", services_ocf_exitcode_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), (is_probe? "probe" : history->task), history->rsc->id, pe__node_name(history->node), last_change_s, history->exit_status, history->id); } else { pcmk__sched_warn("Unexpected result (%s%s%s) was recorded for %s of " "%s on %s at %s " CRM_XS " exit-status=%d id=%s", services_ocf_exitcode_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), (is_probe? "probe" : history->task), history->rsc->id, pe__node_name(history->node), last_change_s, history->exit_status, history->id); if (is_probe && (history->exit_status != PCMK_OCF_OK) && (history->exit_status != PCMK_OCF_NOT_RUNNING) && (history->exit_status != 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", history->rsc->id, pe__node_name(history->node)); } record_failed_op(history); } free(last_change_s); if (cmp_on_fail(*on_fail, config_on_fail) < 0) { pcmk__rsc_trace(history->rsc, "on-fail %s -> %s for %s", fail2text(*on_fail), fail2text(config_on_fail), history->key); *on_fail = config_on_fail; } if (strcmp(history->task, PCMK_ACTION_STOP) == 0) { resource_location(history->rsc, history->node, -INFINITY, "__stop_fail__", history->rsc->cluster); } else if (strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0) { unpack_migrate_to_failure(history); } else if (strcmp(history->task, PCMK_ACTION_MIGRATE_FROM) == 0) { unpack_migrate_from_failure(history); } else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) { history->rsc->role = pcmk_role_promoted; } else if (strcmp(history->task, PCMK_ACTION_DEMOTE) == 0) { if (config_on_fail == pcmk_on_fail_block) { history->rsc->role = pcmk_role_promoted; pe__set_next_role(history->rsc, pcmk_role_stopped, "demote with on-fail=block"); } else if (history->exit_status == PCMK_OCF_NOT_RUNNING) { history->rsc->role = pcmk_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. */ history->rsc->role = pcmk_role_unpromoted; } } if (is_probe && (history->exit_status == PCMK_OCF_NOT_INSTALLED)) { /* leave stopped */ pcmk__rsc_trace(history->rsc, "Leaving %s stopped", history->rsc->id); history->rsc->role = pcmk_role_stopped; } else if (history->rsc->role < pcmk_role_started) { pcmk__rsc_trace(history->rsc, "Setting %s active", history->rsc->id); set_active(history->rsc); } pcmk__rsc_trace(history->rsc, "Resource %s: role=%s unclean=%s on_fail=%s fail_role=%s", history->rsc->id, role2text(history->rsc->role), pcmk__btoa(history->node->details->unclean), fail2text(config_on_fail), role2text(fail_role)); if ((fail_role != pcmk_role_started) && (history->rsc->next_role < fail_role)) { pe__set_next_role(history->rsc, fail_role, "failure"); } if (fail_role == pcmk_role_stopped) { ban_from_all_nodes(history->rsc); } } /*! * \internal * \brief Block a resource with a failed action if it cannot be recovered * * If resource action is a failed stop and fencing is not possible, mark the * resource as unmanaged and blocked, since recovery cannot be done. * * \param[in,out] history Parsed action history entry */ static void block_if_unrecoverable(struct action_history *history) { char *last_change_s = NULL; if (strcmp(history->task, PCMK_ACTION_STOP) != 0) { return; // All actions besides stop are always recoverable } if (pe_can_fence(history->node->details->data_set, history->node)) { return; // Failed stops are recoverable via fencing } last_change_s = last_change_str(history->xml); pcmk__sched_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", history->rsc->id, history->task, pe__node_name(history->node), services_ocf_exitcode_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, ""), last_change_s, history->exit_status, history->id); free(last_change_s); pcmk__clear_rsc_flags(history->rsc, pcmk_rsc_managed); pcmk__set_rsc_flags(history->rsc, pcmk_rsc_blocked); } /*! * \internal * \brief Update action history's execution status and why * * \param[in,out] history Parsed action history entry * \param[out] why Where to store reason for update * \param[in] value New value * \param[in] reason Description of why value was changed */ static inline void remap_because(struct action_history *history, const char **why, int value, const char *reason) { if (history->execution_status != value) { history->execution_status = value; *why = reason; } } /*! * \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,out] history Parsed action history entry * \param[in,out] on_fail What should be done about the result * \param[in] expired Whether result is expired * * \note If the result is remapped and the node is not shutting down or failed, * the operation will be recorded in the scheduler data'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(struct action_history *history, enum action_fail_response *on_fail, bool expired) { bool is_probe = false; int orig_exit_status = history->exit_status; int orig_exec_status = history->execution_status; const char *why = NULL; const char *task = history->task; // Remap degraded results to their successful counterparts history->exit_status = pcmk__effective_rc(history->exit_status); if (history->exit_status != orig_exit_status) { why = "degraded result"; if (!expired && (!history->node->details->shutdown || history->node->details->online)) { record_failed_op(history); } } if (!pe_rsc_is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml) && ((history->execution_status != PCMK_EXEC_DONE) || (history->exit_status != PCMK_OCF_NOT_RUNNING))) { history->execution_status = PCMK_EXEC_DONE; history->exit_status = PCMK_OCF_NOT_RUNNING; why = "equivalent probe result"; } /* If the executor reported an execution 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. */ switch (history->execution_status) { case PCMK_EXEC_DONE: case PCMK_EXEC_ERROR: break; // These should be treated as node-fatal case PCMK_EXEC_NO_FENCE_DEVICE: case PCMK_EXEC_NO_SECRETS: remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "node-fatal error"); goto remap_done; default: goto remap_done; } is_probe = pcmk_xe_is_probe(history->xml); if (is_probe) { task = "probe"; } if (history->expected_exit_status < 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 * expected exit status 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. */ remap_because(history, &why, PCMK_EXEC_ERROR, "obsolete history format"); pcmk__config_warn("Expected result not found for %s on %s " "(corrupt or obsolete CIB?)", history->key, pe__node_name(history->node)); } else if (history->exit_status == history->expected_exit_status) { remap_because(history, &why, PCMK_EXEC_DONE, "expected result"); } else { remap_because(history, &why, PCMK_EXEC_ERROR, "unexpected result"); pcmk__rsc_debug(history->rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)", history->key, pe__node_name(history->node), history->expected_exit_status, services_ocf_exitcode_str(history->expected_exit_status), history->exit_status, services_ocf_exitcode_str(history->exit_status), (pcmk__str_empty(history->exit_reason)? "" : ": "), pcmk__s(history->exit_reason, "")); } switch (history->exit_status) { case PCMK_OCF_OK: if (is_probe && (history->expected_exit_status == PCMK_OCF_NOT_RUNNING)) { char *last_change_s = last_change_str(history->xml); remap_because(history, &why, PCMK_EXEC_DONE, "probe"); pcmk__rsc_info(history->rsc, "Probe found %s active on %s at %s", history->rsc->id, pe__node_name(history->node), last_change_s); free(last_change_s); } break; case PCMK_OCF_NOT_RUNNING: if (is_probe || (history->expected_exit_status == history->exit_status) || !pcmk_is_set(history->rsc->flags, pcmk_rsc_managed)) { /* For probes, recurring monitors for the Stopped role, and * unmanaged resources, "not running" is not considered a * failure. */ remap_because(history, &why, PCMK_EXEC_DONE, "exit status"); history->rsc->role = pcmk_role_stopped; *on_fail = pcmk_on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "not running"); } break; case PCMK_OCF_RUNNING_PROMOTED: if (is_probe && (history->exit_status != history->expected_exit_status)) { char *last_change_s = last_change_str(history->xml); remap_because(history, &why, PCMK_EXEC_DONE, "probe"); pcmk__rsc_info(history->rsc, "Probe found %s active and promoted on %s at %s", history->rsc->id, pe__node_name(history->node), last_change_s); free(last_change_s); } if (!expired || (history->exit_status == history->expected_exit_status)) { history->rsc->role = pcmk_role_promoted; } break; case PCMK_OCF_FAILED_PROMOTED: if (!expired) { history->rsc->role = pcmk_role_promoted; } remap_because(history, &why, PCMK_EXEC_ERROR, "exit status"); break; case PCMK_OCF_NOT_CONFIGURED: remap_because(history, &why, PCMK_EXEC_ERROR_FATAL, "exit status"); break; case PCMK_OCF_UNIMPLEMENT_FEATURE: { guint interval_ms = 0; crm_element_value_ms(history->xml, XML_LRM_ATTR_INTERVAL_MS, &interval_ms); if (interval_ms == 0) { if (!expired) { block_if_unrecoverable(history); } remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "exit status"); } else { remap_because(history, &why, PCMK_EXEC_NOT_SUPPORTED, "exit status"); } } break; case PCMK_OCF_NOT_INSTALLED: case PCMK_OCF_INVALID_PARAM: case PCMK_OCF_INSUFFICIENT_PRIV: if (!expired) { block_if_unrecoverable(history); } remap_because(history, &why, PCMK_EXEC_ERROR_HARD, "exit status"); break; default: if (history->execution_status == PCMK_EXEC_DONE) { char *last_change_s = last_change_str(history->xml); crm_info("Treating unknown exit status %d from %s of %s " "on %s at %s as failure", history->exit_status, task, history->rsc->id, pe__node_name(history->node), last_change_s); remap_because(history, &why, PCMK_EXEC_ERROR, "unknown exit status"); free(last_change_s); } break; } remap_done: if (why != NULL) { pcmk__rsc_trace(history->rsc, "Remapped %s result from [%s: %s] to [%s: %s] " "because of %s", history->key, pcmk_exec_status_str(orig_exec_status), crm_exit_str(orig_exit_status), pcmk_exec_status_str(history->execution_status), crm_exit_str(history->exit_status), why); } } // return TRUE if start or monitor last failure but parameters changed static bool should_clear_for_param_change(const xmlNode *xml_op, const char *task, pcmk_resource_t *rsc, pcmk_node_t *node) { if (pcmk__str_any_of(task, PCMK_ACTION_START, PCMK_ACTION_MONITOR, NULL)) { if (pe__bundle_needs_remote_name(rsc)) { /* 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, pcmk__check_last_failure, rsc->cluster); } else { pcmk__op_digest_t *digest_data = NULL; digest_data = rsc_action_digest_cmp(rsc, xml_op, node, rsc->cluster); switch (digest_data->rc) { case pcmk__digest_unknown: crm_trace("Resource %s history entry %s on %s" " has no digest to compare", rsc->id, pe__xe_history_key(xml_op), node->details->id); break; case pcmk__digest_match: break; default: return TRUE; } } } return FALSE; } // Order action after fencing of remote node, given connection rsc static void order_after_remote_fencing(pcmk_action_t *action, pcmk_resource_t *remote_conn, pcmk_scheduler_t *scheduler) { pcmk_node_t *remote_node = pe_find_node(scheduler->nodes, remote_conn->id); if (remote_node) { pcmk_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL, FALSE, scheduler); order_actions(fence, action, pcmk__ar_first_implies_then); } } static bool should_ignore_failure_timeout(const pcmk_resource_t *rsc, const char *task, guint interval_ms, bool is_last_failure) { /* 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(rsc->cluster->flags, pcmk_sched_fencing_enabled) && (interval_ms != 0) && pcmk__str_eq(task, PCMK_ACTION_MONITOR, pcmk__str_casei)) { pcmk_node_t *remote_node = pe_find_node(rsc->cluster->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,out] history Parsed action result history * * \return true if operation history entry is expired, otherwise false */ static bool check_operation_expiry(struct action_history *history) { bool expired = false; bool is_last_failure = pcmk__ends_with(history->id, "_last_failure_0"); time_t last_run = 0; int unexpired_fail_count = 0; const char *clear_reason = NULL; if (history->execution_status == PCMK_EXEC_NOT_INSTALLED) { pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not expired: " "Not Installed does not expire", history->id, pe__node_name(history->node)); return false; // "Not installed" must always be cleared manually } if ((history->rsc->failure_timeout > 0) && (crm_element_value_epoch(history->xml, XML_RSC_OP_LAST_CHANGE, &last_run) == 0)) { /* Resource has a PCMK_META_FAILURE_TIMEOUT and history entry has a * timestamp */ time_t now = get_effective_time(history->rsc->cluster); time_t last_failure = 0; // Is this particular operation history older than the failure timeout? if ((now >= (last_run + history->rsc->failure_timeout)) && !should_ignore_failure_timeout(history->rsc, history->task, history->interval_ms, is_last_failure)) { expired = true; } // Does the resource as a whole have an unexpired fail count? unexpired_fail_count = pe_get_failcount(history->node, history->rsc, &last_failure, pcmk__fc_effective, history->xml); // Update scheduler recheck time according to *last* failure crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds" " last-failure@%lld", history->id, (long long) last_run, (expired? "" : "not "), (long long) now, unexpired_fail_count, history->rsc->failure_timeout, (long long) last_failure); last_failure += history->rsc->failure_timeout + 1; if (unexpired_fail_count && (now < last_failure)) { pe__update_recheck_time(last_failure, history->rsc->cluster, "fail count expiration"); } } if (expired) { if (pe_get_failcount(history->node, history->rsc, NULL, pcmk__fc_default, history->xml)) { // 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. */ pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not " "expired: Unexpired fail count", history->id, pe__node_name(history->node)); expired = false; } } else if (is_last_failure && (history->rsc->remote_reconnect_ms != 0)) { /* 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(history->xml, history->task, history->rsc, history->node)) { clear_reason = "resource parameters have changed"; } if (clear_reason != NULL) { pcmk_action_t *clear_op = NULL; // Schedule clearing of the fail count clear_op = pe__clear_failcount(history->rsc, history->node, clear_reason, history->rsc->cluster); if (pcmk_is_set(history->rsc->cluster->flags, pcmk_sched_fencing_enabled) && (history->rsc->remote_reconnect_ms != 0)) { /* 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", history->task, history->rsc->id); order_after_remote_fencing(clear_op, history->rsc, history->rsc->cluster); } } if (expired && (history->interval_ms == 0) && pcmk__str_eq(history->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { switch (history->exit_status) { 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 pcmk__rsc_trace(history->rsc, "Resource history entry %s on %s is not " "expired: Probe result", history->id, pe__node_name(history->node)); expired = false; break; } } return expired; } int pe__target_rc_from_xml(const 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; } /*! * \internal * \brief Update a resource's state for an action result * * \param[in,out] history Parsed action history entry * \param[in] exit_status Exit status to base new state on * \param[in] last_failure Resource's last_failure entry, if known * \param[in,out] on_fail Resource's current failure handling */ static void update_resource_state(struct action_history *history, int exit_status, const xmlNode *last_failure, enum action_fail_response *on_fail) { bool clear_past_failure = false; if ((exit_status == PCMK_OCF_NOT_INSTALLED) || (!pe_rsc_is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml))) { history->rsc->role = pcmk_role_stopped; } else if (exit_status == PCMK_OCF_NOT_RUNNING) { clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MONITOR, pcmk__str_none)) { if ((last_failure != NULL) && pcmk__str_eq(history->key, pe__xe_history_key(last_failure), pcmk__str_none)) { clear_past_failure = true; } if (history->rsc->role < pcmk_role_started) { set_active(history->rsc); } } else if (pcmk__str_eq(history->task, PCMK_ACTION_START, pcmk__str_none)) { history->rsc->role = pcmk_role_started; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_STOP, pcmk__str_none)) { history->rsc->role = pcmk_role_stopped; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_PROMOTE, pcmk__str_none)) { history->rsc->role = pcmk_role_promoted; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_DEMOTE, pcmk__str_none)) { if (*on_fail == pcmk_on_fail_demote) { // Demote clears an error only if on-fail=demote clear_past_failure = true; } history->rsc->role = pcmk_role_unpromoted; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_FROM, pcmk__str_none)) { history->rsc->role = pcmk_role_started; clear_past_failure = true; } else if (pcmk__str_eq(history->task, PCMK_ACTION_MIGRATE_TO, pcmk__str_none)) { unpack_migrate_to_success(history); } else if (history->rsc->role < pcmk_role_started) { pcmk__rsc_trace(history->rsc, "%s active on %s", history->rsc->id, pe__node_name(history->node)); set_active(history->rsc); } if (!clear_past_failure) { return; } switch (*on_fail) { case pcmk_on_fail_stop: case pcmk_on_fail_ban: case pcmk_on_fail_standby_node: case pcmk_on_fail_fence_node: pcmk__rsc_trace(history->rsc, "%s (%s) is not cleared by a completed %s", history->rsc->id, fail2text(*on_fail), history->task); break; case pcmk_on_fail_block: case pcmk_on_fail_ignore: case pcmk_on_fail_demote: case pcmk_on_fail_restart: case pcmk_on_fail_restart_container: *on_fail = pcmk_on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "clear past failures"); break; case pcmk_on_fail_reset_remote: if (history->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 = pcmk_on_fail_ignore; pe__set_next_role(history->rsc, pcmk_role_unknown, "clear past failures and reset remote"); } break; } } /*! * \internal * \brief Check whether a given history entry matters for resource state * * \param[in] history Parsed action history entry * * \return true if action can affect resource state, otherwise false */ static inline bool can_affect_state(struct action_history *history) { #if 0 /* @COMPAT It might be better to parse only actions we know we're interested * in, rather than exclude a couple we don't. However that would be a * behavioral change that should be done at a major or minor series release. * Currently, unknown operations can affect whether a resource is considered * active and/or failed. */ return pcmk__str_any_of(history->task, PCMK_ACTION_MONITOR, PCMK_ACTION_START, PCMK_ACTION_STOP, PCMK_ACTION_PROMOTE, PCMK_ACTION_DEMOTE, PCMK_ACTION_MIGRATE_TO, PCMK_ACTION_MIGRATE_FROM, "asyncmon", NULL); #else return !pcmk__str_any_of(history->task, PCMK_ACTION_NOTIFY, PCMK_ACTION_META_DATA, NULL); #endif } /*! * \internal * \brief Unpack execution/exit status and exit reason from a history entry * * \param[in,out] history Action history entry to unpack * * \return Standard Pacemaker return code */ static int unpack_action_result(struct action_history *history) { if ((crm_element_value_int(history->xml, XML_LRM_ATTR_OPSTATUS, &(history->execution_status)) < 0) || (history->execution_status < PCMK_EXEC_PENDING) || (history->execution_status > PCMK_EXEC_MAX) || (history->execution_status == PCMK_EXEC_CANCELLED)) { pcmk__config_err("Ignoring resource history entry %s for %s on %s " "with invalid " XML_LRM_ATTR_OPSTATUS " '%s'", history->id, history->rsc->id, pe__node_name(history->node), pcmk__s(crm_element_value(history->xml, XML_LRM_ATTR_OPSTATUS), "")); return pcmk_rc_unpack_error; } if ((crm_element_value_int(history->xml, XML_LRM_ATTR_RC, &(history->exit_status)) < 0) || (history->exit_status < 0) || (history->exit_status > CRM_EX_MAX)) { #if 0 /* @COMPAT We should ignore malformed entries, but since that would * change behavior, it should be done at a major or minor series * release. */ pcmk__config_err("Ignoring resource history entry %s for %s on %s " "with invalid " XML_LRM_ATTR_RC " '%s'", history->id, history->rsc->id, pe__node_name(history->node), pcmk__s(crm_element_value(history->xml, XML_LRM_ATTR_RC), "")); return pcmk_rc_unpack_error; #else history->exit_status = CRM_EX_ERROR; #endif } history->exit_reason = crm_element_value(history->xml, XML_LRM_ATTR_EXIT_REASON); return pcmk_rc_ok; } /*! * \internal * \brief Process an action history entry whose result expired * * \param[in,out] history Parsed action history entry * \param[in] orig_exit_status Action exit status before remapping * * \return Standard Pacemaker return code (in particular, pcmk_rc_ok means the * entry needs no further processing) */ static int process_expired_result(struct action_history *history, int orig_exit_status) { if (!pe_rsc_is_bundled(history->rsc) && pcmk_xe_mask_probe_failure(history->xml) && (orig_exit_status != history->expected_exit_status)) { if (history->rsc->role <= pcmk_role_stopped) { history->rsc->role = pcmk_role_unknown; } crm_trace("Ignoring resource history entry %s for probe of %s on %s: " "Masked failure expired", history->id, history->rsc->id, pe__node_name(history->node)); return pcmk_rc_ok; } if (history->exit_status == history->expected_exit_status) { return pcmk_rc_undetermined; // Only failures expire } if (history->interval_ms == 0) { crm_notice("Ignoring resource history entry %s for %s of %s on %s: " "Expired failure", history->id, history->task, history->rsc->id, pe__node_name(history->node)); return pcmk_rc_ok; } if (history->node->details->online && !history->node->details->unclean) { /* Reschedule the recurring action. 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-interval %s of %s on %s " "after failure expired", pcmk__readable_interval(history->interval_ms), history->task, history->rsc->id, pe__node_name(history->node)); crm_xml_add(history->xml, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout"); return pcmk_rc_ok; } return pcmk_rc_undetermined; } /*! * \internal * \brief Process a masked probe failure * * \param[in,out] history Parsed action history entry * \param[in] orig_exit_status Action exit status before remapping * \param[in] last_failure Resource's last_failure entry, if known * \param[in,out] on_fail Resource's current failure handling */ static void mask_probe_failure(struct action_history *history, int orig_exit_status, const xmlNode *last_failure, enum action_fail_response *on_fail) { pcmk_resource_t *ban_rsc = history->rsc; if (!pcmk_is_set(history->rsc->flags, pcmk_rsc_unique)) { ban_rsc = uber_parent(history->rsc); } crm_notice("Treating probe result '%s' for %s on %s as 'not running'", services_ocf_exitcode_str(orig_exit_status), history->rsc->id, pe__node_name(history->node)); update_resource_state(history, history->expected_exit_status, last_failure, on_fail); - crm_xml_add(history->xml, XML_ATTR_UNAME, history->node->details->uname); + crm_xml_add(history->xml, PCMK_XA_UNAME, history->node->details->uname); record_failed_op(history); resource_location(ban_rsc, history->node, -INFINITY, "masked-probe-failure", history->rsc->cluster); } /*! * \internal Check whether a given failure is for a given pending action * * \param[in] history Parsed history entry for pending action * \param[in] last_failure Resource's last_failure entry, if known * * \return true if \p last_failure is failure of pending action in \p history, * otherwise false * \note Both \p history and \p last_failure must come from the same * lrm_resource block, as node and resource are assumed to be the same. */ static bool failure_is_newer(const struct action_history *history, const xmlNode *last_failure) { guint failure_interval_ms = 0U; long long failure_change = 0LL; long long this_change = 0LL; if (last_failure == NULL) { return false; // Resource has no last_failure entry } if (!pcmk__str_eq(history->task, crm_element_value(last_failure, XML_LRM_ATTR_TASK), pcmk__str_none)) { return false; // last_failure is for different action } if ((crm_element_value_ms(last_failure, XML_LRM_ATTR_INTERVAL_MS, &failure_interval_ms) != pcmk_ok) || (history->interval_ms != failure_interval_ms)) { return false; // last_failure is for action with different interval } if ((pcmk__scan_ll(crm_element_value(history->xml, XML_RSC_OP_LAST_CHANGE), &this_change, 0LL) != pcmk_rc_ok) || (pcmk__scan_ll(crm_element_value(last_failure, XML_RSC_OP_LAST_CHANGE), &failure_change, 0LL) != pcmk_rc_ok) || (failure_change < this_change)) { return false; // Failure is not known to be newer } return true; } /*! * \internal * \brief Update a resource's role etc. for a pending action * * \param[in,out] history Parsed history entry for pending action * \param[in] last_failure Resource's last_failure entry, if known */ static void process_pending_action(struct action_history *history, const xmlNode *last_failure) { /* For recurring monitors, a failure is recorded only in RSC_last_failure_0, * and there might be a RSC_monitor_INTERVAL entry with the last successful * or pending result. * * If last_failure contains the failure of the pending recurring monitor * we're processing here, and is newer, the action is no longer pending. * (Pending results have call ID -1, which sorts last, so the last failure * if any should be known.) */ if (failure_is_newer(history, last_failure)) { return; } if (strcmp(history->task, PCMK_ACTION_START) == 0) { pcmk__set_rsc_flags(history->rsc, pcmk_rsc_start_pending); set_active(history->rsc); } else if (strcmp(history->task, PCMK_ACTION_PROMOTE) == 0) { history->rsc->role = pcmk_role_promoted; } else if ((strcmp(history->task, PCMK_ACTION_MIGRATE_TO) == 0) && history->node->details->unclean) { /* A migrate_to action is pending on a unclean source, so force a stop * on the target. */ const char *migrate_target = NULL; pcmk_node_t *target = NULL; migrate_target = crm_element_value(history->xml, XML_LRM_ATTR_MIGRATE_TARGET); target = pe_find_node(history->rsc->cluster->nodes, migrate_target); if (target != NULL) { stop_action(history->rsc, target, FALSE); } } if (history->rsc->pending_task != NULL) { /* There should never be multiple pending actions, but as a failsafe, * just remember the first one processed for display purposes. */ return; } if (pcmk_is_probe(history->task, history->interval_ms)) { /* Pending probes are currently never displayed, even if pending * operations are requested. If we ever want to change that, * enable the below and the corresponding part of * native.c:native_pending_task(). */ #if 0 history->rsc->pending_task = strdup("probe"); history->rsc->pending_node = history->node; #endif } else { history->rsc->pending_task = strdup(history->task); history->rsc->pending_node = history->node; } } static void unpack_rsc_op(pcmk_resource_t *rsc, pcmk_node_t *node, xmlNode *xml_op, xmlNode **last_failure, enum action_fail_response *on_fail) { int old_rc = 0; bool expired = false; pcmk_resource_t *parent = rsc; enum rsc_role_e fail_role = pcmk_role_unknown; enum action_fail_response failure_strategy = pcmk_on_fail_restart; struct action_history history = { .rsc = rsc, .node = node, .xml = xml_op, .execution_status = PCMK_EXEC_UNKNOWN, }; CRM_CHECK(rsc && node && xml_op, return); history.id = ID(xml_op); if (history.id == NULL) { pcmk__config_err("Ignoring resource history entry for %s on %s " "without ID", rsc->id, pe__node_name(node)); return; } // Task and interval history.task = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (history.task == NULL) { pcmk__config_err("Ignoring resource history entry %s for %s on %s " "without " XML_LRM_ATTR_TASK, history.id, rsc->id, pe__node_name(node)); return; } crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &(history.interval_ms)); if (!can_affect_state(&history)) { pcmk__rsc_trace(rsc, "Ignoring resource history entry %s for %s on %s " "with irrelevant action '%s'", history.id, rsc->id, pe__node_name(node), history.task); return; } if (unpack_action_result(&history) != pcmk_rc_ok) { return; // Error already logged } history.expected_exit_status = pe__target_rc_from_xml(xml_op); history.key = pe__xe_history_key(xml_op); crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &(history.call_id)); pcmk__rsc_trace(rsc, "Unpacking %s (%s call %d on %s): %s (%s)", history.id, history.task, history.call_id, pe__node_name(node), pcmk_exec_status_str(history.execution_status), crm_exit_str(history.exit_status)); if (node->details->unclean) { pcmk__rsc_trace(rsc, "%s is running on %s, which is unclean (further action " "depends on value of stop's on-fail attribute)", rsc->id, pe__node_name(node)); } expired = check_operation_expiry(&history); old_rc = history.exit_status; remap_operation(&history, on_fail, expired); if (expired && (process_expired_result(&history, old_rc) == pcmk_rc_ok)) { goto done; } if (!pe_rsc_is_bundled(rsc) && pcmk_xe_mask_probe_failure(xml_op)) { mask_probe_failure(&history, old_rc, *last_failure, on_fail); goto done; } if (!pcmk_is_set(rsc->flags, pcmk_rsc_unique)) { parent = uber_parent(rsc); } switch (history.execution_status) { case PCMK_EXEC_PENDING: process_pending_action(&history, *last_failure); goto done; case PCMK_EXEC_DONE: update_resource_state(&history, history.exit_status, *last_failure, on_fail); goto done; case PCMK_EXEC_NOT_INSTALLED: unpack_failure_handling(&history, &failure_strategy, &fail_role); if (failure_strategy == pcmk_on_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", history.task, rsc->id, pe__node_name(node), history.execution_status, history.exit_status, history.id); /* Also for printing it as "FAILED" by marking it as * pcmk_rsc_failed later */ *on_fail = pcmk_on_fail_ban; } resource_location(parent, node, -INFINITY, "hard-error", rsc->cluster); unpack_rsc_op_failure(&history, failure_strategy, fail_role, last_failure, on_fail); goto done; case PCMK_EXEC_NOT_CONNECTED: if (pe__is_guest_or_remote_node(node) && pcmk_is_set(node->details->remote_rsc->flags, pcmk_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. */ pcmk__set_rsc_flags(node->details->remote_rsc, pcmk_rsc_failed|pcmk_rsc_stop_if_failed); } 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 default: // No other value should be possible at this point break; } unpack_failure_handling(&history, &failure_strategy, &fail_role); if ((failure_strategy == pcmk_on_fail_ignore) || ((failure_strategy == pcmk_on_fail_restart_container) && (strcmp(history.task, PCMK_ACTION_STOP) == 0))) { char *last_change_s = last_change_str(xml_op); crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s succeeded " CRM_XS " %s", history.task, services_ocf_exitcode_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), rsc->id, pe__node_name(node), last_change_s, history.id); free(last_change_s); update_resource_state(&history, history.expected_exit_status, *last_failure, on_fail); - crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname); + crm_xml_add(xml_op, PCMK_XA_UNAME, node->details->uname); pcmk__set_rsc_flags(rsc, pcmk_rsc_ignore_failure); record_failed_op(&history); if ((failure_strategy == pcmk_on_fail_restart_container) && cmp_on_fail(*on_fail, pcmk_on_fail_restart) <= 0) { *on_fail = failure_strategy; } } else { unpack_rsc_op_failure(&history, failure_strategy, fail_role, last_failure, on_fail); if (history.execution_status == PCMK_EXEC_ERROR_HARD) { uint8_t log_level = LOG_ERR; if (history.exit_status == PCMK_OCF_NOT_INSTALLED) { log_level = LOG_NOTICE; } do_crm_log(log_level, "Preventing %s from restarting on %s because " "of hard failure (%s%s%s) " CRM_XS " %s", parent->id, pe__node_name(node), services_ocf_exitcode_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), history.id); resource_location(parent, node, -INFINITY, "hard-error", rsc->cluster); } else if (history.execution_status == PCMK_EXEC_ERROR_FATAL) { pcmk__sched_err("Preventing %s from restarting anywhere because " "of fatal failure (%s%s%s) " CRM_XS " %s", parent->id, services_ocf_exitcode_str(history.exit_status), (pcmk__str_empty(history.exit_reason)? "" : ": "), pcmk__s(history.exit_reason, ""), history.id); resource_location(parent, NULL, -INFINITY, "fatal-error", rsc->cluster); } } done: pcmk__rsc_trace(rsc, "%s role on %s after %s is %s (next %s)", rsc->id, pe__node_name(node), history.id, role2text(rsc->role), role2text(rsc->next_role)); } static void add_node_attrs(const xmlNode *xml_obj, pcmk_node_t *node, bool overwrite, pcmk_scheduler_t *scheduler) { const char *cluster_name = NULL; pe_rule_eval_data_t rule_data = { .node_hash = NULL, .role = pcmk_role_unknown, .now = scheduler->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, scheduler->dc_uuid, pcmk__str_casei)) { scheduler->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(scheduler->config_hash, PCMK_OPT_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, scheduler); pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_UTILIZATION, &rule_data, node->details->utilization, NULL, FALSE, scheduler); 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); + crm_xml_add(rsc_op, PCMK_XA_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, pcmk_scheduler_t *scheduler) { GList *output = NULL; GList *intermediate = NULL; xmlNode *tmp = NULL; xmlNode *status = find_xml_node(scheduler->input, XML_CIB_TAG_STATUS, TRUE); pcmk_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); + const char *uname = crm_element_value(node_state, PCMK_XA_UNAME); if (node != NULL && !pcmk__str_eq(uname, node, pcmk__str_casei)) { continue; } this_node = pe_find_node(scheduler->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(scheduler, this_node); } else { determine_online_status(node_state, this_node, scheduler); } if (this_node->details->online || pcmk_is_set(scheduler->flags, pcmk_sched_fencing_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, PCMK_XA_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/tools/crm_mon.c b/tools/crm_mon.c index c9a069186e..dde08195b5 100644 --- a/tools/crm_mon.c +++ b/tools/crm_mon.c @@ -1,2287 +1,2287 @@ /* * Copyright 2004-2024 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 enum pcmk_pacemakerd_state pcmkd_state = pcmk_pacemakerd_state_invalid; 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; 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 } }; PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_default(pcmk__output_t *out, va_list args) { return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_html(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); if (out->dest != stdout) { out->reset(out); } pcmk__output_create_xml_text_node(out, "span", "Not connected to CIB"); if (desc != NULL) { pcmk__output_create_xml_text_node(out, "span", ": "); pcmk__output_create_xml_text_node(out, "span", desc); } if (state != pcmk_pacemakerd_state_invalid) { const char *state_s = pcmk__pcmkd_state_enum2friendly(state); pcmk__output_create_xml_text_node(out, "span", " ("); pcmk__output_create_xml_text_node(out, "span", state_s); pcmk__output_create_xml_text_node(out, "span", ")"); } out->finish(out, CRM_EX_DISCONNECT, true, NULL); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_text(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); int rc = pcmk_rc_ok; if (out->dest != stdout) { out->reset(out); } if (state != pcmk_pacemakerd_state_invalid) { rc = out->info(out, "Not connected to CIB%s%s (%s)", (desc != NULL)? ": " : "", pcmk__s(desc, ""), pcmk__pcmkd_state_enum2friendly(state)); } else { rc = out->info(out, "Not connected to CIB%s%s", (desc != NULL)? ": " : "", pcmk__s(desc, "")); } out->finish(out, CRM_EX_DISCONNECT, true, NULL); return rc; } PCMK__OUTPUT_ARGS("crm-mon-disconnected", "const char *", "enum pcmk_pacemakerd_state") static int crm_mon_disconnected_xml(pcmk__output_t *out, va_list args) { const char *desc = va_arg(args, const char *); enum pcmk_pacemakerd_state state = (enum pcmk_pacemakerd_state) va_arg(args, int); const char *state_s = NULL; if (out->dest != stdout) { out->reset(out); } if (state != pcmk_pacemakerd_state_invalid) { state_s = pcmk_pacemakerd_api_daemon_state_enum2text(state); } pcmk__output_create_xml_node(out, "crm-mon-disconnected", PCMK_XA_DESCRIPTION, desc, "pacemakerd-state", state_s, NULL); out->finish(out, CRM_EX_DISCONNECT, true, NULL); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "crm-mon-disconnected", "default", crm_mon_disconnected_default }, { "crm-mon-disconnected", "html", crm_mon_disconnected_html }, { "crm-mon-disconnected", "text", crm_mon_disconnected_text }, { "crm-mon-disconnected", "xml", crm_mon_disconnected_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; enum mon_exec_mode exec_mode; gboolean fence_connect; 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 = { .reconnect_ms = RECONNECT_MSECS, .exec_mode = mon_exec_unset, .fence_connect = TRUE, }; 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 setup_cib_connection(void); static int setup_fencer_connection(void); static int setup_api_connections(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: case mon_output_html: case mon_output_cgi: return pcmk_section_summary |pcmk_section_nodes |pcmk_section_resources |pcmk_section_failures; case mon_output_xml: return all_includes(fmt); 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 }, { 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.exec_mode = mon_exec_one_shot; 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); // World-readable HTML 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.exec_mode = mon_exec_one_shot; 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", PCMK__VALUE_FENCING, data, err); case 2: options.fence_connect = TRUE; fence_history = pcmk__fence_history_full; 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", 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) { pcmk__str_update(&args->output_ty, "text"); 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_description_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { show_opts |= pcmk_show_description; 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 { pcmk_parse_interval_spec(optarg, &options.reconnect_ms); if (options.exec_mode != mon_exec_daemonized) { // Reconnect interval applies to daemonized too, so don't override options.exec_mode = mon_exec_update; } } return TRUE; } /*! * \internal * \brief Enable one-shot mode * * \param[in] option_name Name of option being parsed (ignored) * \param[in] optarg Value to be parsed (ignored) * \param[in] data User data (ignored) * \param[out] err Where to store error (ignored) */ static gboolean one_shot_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { options.exec_mode = mon_exec_one_shot; return TRUE; } /*! * \internal * \brief Enable daemonized mode * * \param[in] option_name Name of option being parsed (ignored) * \param[in] optarg Value to be parsed (ignored) * \param[in] data User data (ignored) * \param[out] err Where to store error (ignored) */ static gboolean daemonize_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) { options.exec_mode = mon_exec_daemonized; 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.exec_mode = mon_exec_one_shot; 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', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, one_shot_cb, "Display the cluster status once and exit", NULL }, { "daemonize", 'd', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, daemonize_cb, "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 }, { "show-description", 0, G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_description_cb, "Show resource descriptions", 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 }, { 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 }, { "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\n" INDENT "(suitable for nagios)", 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->transient(out, "Reconnecting..."); if (setup_api_connections() == pcmk_rc_ok) { // Trigger redrawing the screen (needs reconnect_timer == 0) reconnect_timer = 0; refresh_after_event(FALSE, TRUE); return G_SOURCE_REMOVE; } out->message(out, "crm-mon-disconnected", "Latest connection attempt failed", pcmkd_state); 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) { const char *msg = "Connection to the cluster lost"; pcmkd_state = pcmk_pacemakerd_state_invalid; /* No crm-mon-disconnected message for console; a working implementation * is not currently worth the effort */ out->transient(out, "%s", msg); out->message(out, "crm-mon-disconnected", msg, pcmkd_state); 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; } /* the client API won't properly reconnect notifications if they are still * in the table - so remove them */ stonith_api_delete(st); st = NULL; if (cib) { cib->cmds->signoff(cib); reconnect_timer = g_timeout_add(options.reconnect_ms, reconnect_after_timeout, NULL); } } /* 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 setup_fencer_connection(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 { stonith_api_delete(st); st = NULL; } return rc; } static int setup_cib_connection(void) { int rc = pcmk_rc_ok; CRM_CHECK(cib != NULL, return EINVAL); if (cib->state != cib_disconnected) { // Already connected with notifications registered for CIB updates return rc; } rc = cib__signon_query(out, &cib, ¤t_cib); if (rc == pcmk_rc_ok) { rc = pcmk_legacy2rc(cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy)); if (rc == EPROTONOSUPPORT) { out->err(out, "CIB client does not support connection loss " "notifications; crm_mon will be unable to reconnect after " "connection loss"); 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) { if (rc == EPROTONOSUPPORT) { out->err(out, "CIB client does not support CIB diff " "notifications"); } else { out->err(out, "CIB diff notification setup failed"); } out->err(out, "Cannot monitor CIB changes; exiting"); cib__clean_up_connection(&cib); stonith_api_delete(st); st = NULL; } } 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; } } static int setup_api_connections(void) { int rc = pcmk_rc_ok; CRM_CHECK(cib != NULL, return EINVAL); if (cib->state != cib_disconnected) { return rc; } if (cib->variant == cib_native) { rc = pcmk__pacemakerd_status(out, crm_system_name, options.reconnect_ms / 2, false, &pcmkd_state); if (rc != pcmk_rc_ok) { return rc; } switch (pcmkd_state) { case pcmk_pacemakerd_state_running: case pcmk_pacemakerd_state_remote: case pcmk_pacemakerd_state_shutting_down: /* Fencer and CIB may still be available while shutting down or * running on a Pacemaker Remote node */ break; default: // Fencer and CIB are definitely unavailable return ENOTCONN; } setup_fencer_connection(); } rc = setup_cib_connection(); return rc; } #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; clear(); refresh(); 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 // CURSES_ENABLED // 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 } }; #if CURSES_ENABLED const char *fmts = "console (default), html, text, xml, none"; #else const char *fmts = "text (default), html, xml, none"; #endif // CURSES_ENABLED const char *desc = NULL; desc = "Notes:\n\n" "If this program is called as crm_mon.cgi, --output-as=html and\n" "--html-cgi are automatically added to the command line\n" "arguments.\n\n" "Time Specification:\n\n" "The TIMESPEC in any command line option can be specified in many\n" "different formats. It can be an integer number of seconds, a\n" "number plus units (us/usec/ms/msec/s/sec/m/min/h/hr), or an ISO\n" "8601 period specification.\n\n" "Output Control:\n\n" "By default, a particular set of sections are written to the\n" "output destination. The default varies based on the output\n" "format: XML includes all sections by default, while other output\n" "formats include less. This set can be modified with the --include\n" "and --exclude command line options. Each option may be passed\n" "multiple times, and each can specify a comma-separated list of\n" "sections. The options are applied to the default set, in order\n" "from left to right as they are passed on the command line. For a\n" "list of valid sections, pass --include=list or --exclude=list.\n\n" "Interactive Use:\n\n" #if CURSES_ENABLED "When run interactively, crm_mon can be told to hide and show\n" "various sections of output. To see a help screen explaining the\n" "options, press '?'. Any key stroke aside from those listed will\n" "cause the screen to refresh.\n\n" #else "The local installation of Pacemaker was built without support for\n" "interactive (console) mode. A curses library must be available at\n" "build time to support interactive mode.\n\n" #endif // CURSES_ENABLED "Examples:\n\n" #if CURSES_ENABLED "Display the cluster status on the console with updates as they\n" "occur:\n\n" "\tcrm_mon\n\n" #endif // CURSES_ENABLED "Display the cluster status once and exit:\n\n" "\tcrm_mon -1\n\n" "Display the cluster status, group resources by node, and include\n" "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\n" "cluster status to an HTML file:\n\n" "\tcrm_mon --daemonize --output-as html " "--output-to /path/to/docroot/filename.html\n\n" "Display the cluster status as XML:\n\n" "\tcrm_mon --output-as xml\n\n"; context = pcmk__build_arg_context(args, fmts, group, NULL); pcmk__add_main_args(context, extra_prog_entries); g_option_context_set_description(context, desc); 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); } } } /*! * \internal * \brief Set output format based on \p --output-as arguments and mode arguments * * When the deprecated output format arguments (\p --as-cgi, \p --as-html, * \p --simple-status, \p --as-xml) are parsed, callback functions set * \p output_format (and the umask if appropriate). If none of the deprecated * arguments were specified, this function does the same based on the current * \p --output-as arguments and the \p --one-shot and \p --daemonize arguments. * * \param[in,out] args Command line arguments */ static void reconcile_output_format(pcmk__common_args_t *args) { if (output_format != mon_output_unset) { /* One of the deprecated arguments was used, and we're finished. Note * that this means the deprecated arguments take precedence. */ return; } if (pcmk__str_eq(args->output_ty, "none", pcmk__str_none)) { output_format = mon_output_none; } else if (pcmk__str_eq(args->output_ty, "html", pcmk__str_none)) { output_format = mon_output_html; umask(S_IWGRP | S_IWOTH); // World-readable HTML } else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) { output_format = mon_output_xml; #if CURSES_ENABLED } else if (pcmk__str_eq(args->output_ty, "console", pcmk__str_null_matches)) { /* Console is the default format if no conflicting options are given. * * Use text output instead if one of the following conditions is met: * * We've requested daemonized or one-shot mode (console output is * incompatible with modes other than mon_exec_update) * * We requested the version, which is effectively one-shot * * We specified a non-stdout output destination (console mode is * compatible only with stdout) */ if ((options.exec_mode == mon_exec_daemonized) || (options.exec_mode == mon_exec_one_shot) || args->version || !pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } else { pcmk__str_update(&args->output_ty, "console"); output_format = mon_output_console; crm_enable_stderr(FALSE); } #endif // CURSES_ENABLED } else if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches)) { /* Text output was explicitly requested, or it's the default because * curses is not enabled */ pcmk__str_update(&args->output_ty, "text"); output_format = mon_output_plain; } // Otherwise, invalid format. Let pcmk__output_new() throw an error. } /*! * \internal * \brief Set execution mode to the output format's default if appropriate * * \param[in,out] args Command line arguments */ static void set_default_exec_mode(const pcmk__common_args_t *args) { if (output_format == mon_output_console) { /* Update is the only valid mode for console, but set here instead of * reconcile_output_format() for isolation and consistency */ options.exec_mode = mon_exec_update; } else if (options.exec_mode == mon_exec_unset) { // Default to one-shot mode for all other formats options.exec_mode = mon_exec_one_shot; } else if ((options.exec_mode == mon_exec_update) && pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { // If not using console format, update mode cannot be used with stdout options.exec_mode = mon_exec_one_shot; } } 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 (pcmkd_state == pcmk_pacemakerd_state_remote) { 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, 0); 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.exec_mode = mon_exec_one_shot; } 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: // Everything (fencer, CIB, pcmkd status) should be available break; case cib_file: // Live fence history is not meaningful fence_history_cb("--fence-history", "0", NULL, NULL); /* Notifications are unsupported; nothing to monitor * @COMPAT: Let setup_cib_connection() handle this by exiting? */ options.exec_mode = mon_exec_one_shot; break; case cib_remote: // We won't receive any fencing updates fence_history_cb("--fence-history", "0", NULL, NULL); break; default: /* something is odd */ exit_on_invalid_cib(); break; } if ((options.exec_mode == mon_exec_daemonized) && !options.external_agent && pcmk__str_eq(args->output_dest, "-", pcmk__str_null_matches)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires at least one of --output-to " "(with value not set to '-') and --external-agent"); return clean_up(CRM_EX_USAGE); } } reconcile_output_format(args); set_default_exec_mode(args); add_output_args(); /* output_format MUST NOT BE CHANGED AFTER THIS POINT. */ 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 we had a valid format for pcmk__output_new(), output_format should be * set by now. */ CRM_ASSERT(output_format != mon_output_unset); if (options.exec_mode == mon_exec_daemonized) { if (!options.external_agent && (output_format == mon_output_none)) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_USAGE, "--daemonize requires --external-agent if used with " "--output-as=none"); 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); // Messages internal to this file, nothing curses-specific pcmk__register_messages(out, fmt_functions); 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.exec_mode == mon_exec_daemonized) { 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) { show_opts |= pcmk_show_inactive_rscs | pcmk_show_timing; } 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.exec_mode == mon_exec_one_shot) { one_shot(); } out->message(out, "crm-mon-disconnected", "Waiting for initial connection", pcmkd_state); do { out->transient(out, "Connecting to cluster..."); rc = setup_api_connections(); if (rc != pcmk_rc_ok) { if ((rc == ENOTCONN) || (rc == ECONNREFUSED)) { out->transient(out, "Connection failed. Retrying in %ums...", options.reconnect_ms); } // Give some time to view all output even if we won't retry pcmk__sleep_ms(options.reconnect_ms); #if CURSES_ENABLED if (output_format == mon_output_console) { clear(); refresh(); } #endif } } while ((rc == ENOTCONN) || (rc == ECONNREFUSED)); 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) { out->err(out, "notification fork() failed: %s", strerror(errno)); } 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 int handle_rsc_op(xmlNode *xml, void *userdata) { const char *node_id = (const char *) userdata; 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) { pcmk__xe_foreach_child(xml, NULL, handle_rsc_op, (void *) node_id); return pcmk_rc_ok; } id = pe__xe_history_key(rsc_op); magic = crm_element_value(rsc_op, XML_ATTR_TRANSITION_MAGIC); if (magic == NULL) { /* non-change */ return pcmk_rc_ok; } if (!decode_transition_magic(magic, NULL, NULL, NULL, &status, &rc, &target_rc)) { crm_err("Invalid event %s detected for %s", magic, id); return pcmk_rc_ok; } 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__xe_is(n, XML_CIB_TAG_STATE)) { n = n->parent; } if(node == NULL && n) { - node = crm_element_value(n, XML_ATTR_UNAME); + node = crm_element_value(n, PCMK_XA_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); return pcmk_rc_ok; } /* 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 int handle_op_for_node(xmlNode *xml, void *userdata) { - const char *node = crm_element_value(xml, XML_ATTR_UNAME); + const char *node = crm_element_value(xml, PCMK_XA_UNAME); if (node == NULL) { node = ID(xml); } handle_rsc_op(xml, (void *) node); return pcmk_rc_ok; } 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) { pcmk__xe_foreach_child(first_named_child(match, XML_CIB_TAG_STATUS), NULL, handle_op_for_node, NULL); } else if(strcmp(name, XML_CIB_TAG_STATUS) == 0) { pcmk__xe_foreach_child(match, NULL, handle_op_for_node, NULL); } else if(strcmp(name, XML_CIB_TAG_STATE) == 0) { - node = crm_element_value(match, XML_ATTR_UNAME); + node = crm_element_value(match, PCMK_XA_UNAME); if (node == NULL) { node = ID(match); } handle_rsc_op(match, (void *) node); } else if(strcmp(name, XML_CIB_TAG_LRM) == 0) { node = ID(match); handle_rsc_op(match, (void *) 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, PCMK_XA_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) { 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; } // Get an up-to-date pacemakerd status for the cluster summary if (cib->variant == cib_native) { pcmk__pacemakerd_status(out, crm_system_name, options.reconnect_ms / 2, false, &pcmkd_state); } if (out->dest != stdout) { out->reset(out); } rc = pcmk__output_cluster_status(out, st, cib, current_cib, pcmkd_state, 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 setup_fencer_connection() 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 */ setup_fencer_connection(); 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 setup_fencer_connection() 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); } } /* * 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); stonith_api_delete(st); 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 != NULL)) { 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.exec_mode != mon_exec_daemonized) { out->finish(out, exit_code, true, NULL); } pcmk__output_free(out); pcmk__unregister_formats(); } crm_exit(exit_code); } diff --git a/tools/crm_node.c b/tools/crm_node.c index af2e5f6c93..da46f326d8 100644 --- a/tools/crm_node.c +++ b/tools/crm_node.c @@ -1,882 +1,882 @@ /* * Copyright 2004-2024 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 #define SUMMARY "crm_node - Tool for displaying low-level node information" struct { gboolean corosync; gboolean dangerous_cmd; gboolean force_flag; char command; int nodeid; char *target_uname; } options = { .command = '\0', .force_flag = FALSE }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error); static GError *error = NULL; static GMainLoop *mainloop = NULL; static crm_exit_t exit_code = CRM_EX_OK; static pcmk__output_t *out = NULL; #define INDENT " " static GOptionEntry command_entries[] = { { "cluster-id", 'i', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display this node's cluster id", NULL }, { "list", 'l', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display all known members (past and present) of this cluster", NULL }, { "name", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the name used by the cluster for this node", NULL }, { "partition", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display the members of this partition", NULL }, { "quorum", 'q', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, command_cb, "Display a 1 if our partition has quorum, 0 if not", NULL }, { "name-for-id", 'N', 0, G_OPTION_ARG_CALLBACK, name_cb, "Display the name used by the cluster for the node with the specified ID", "ID" }, { "remove", 'R', 0, G_OPTION_ARG_CALLBACK, remove_cb, "(Advanced) Remove the (stopped) node with the specified name from Pacemaker's\n" INDENT "configuration and caches (the node must already have been removed from\n" INDENT "the underlying cluster stack configuration", "NAME" }, { NULL } }; static GOptionEntry addl_entries[] = { { "force", 'f', 0, G_OPTION_ARG_NONE, &options.force_flag, NULL, NULL }, #if SUPPORT_COROSYNC /* Unused and deprecated */ { "corosync", 'C', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_NONE, &options.corosync, NULL, NULL }, #endif // @TODO add timeout option for when IPC replies are needed { NULL } }; static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; gboolean command_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (pcmk__str_eq("-i", option_name, pcmk__str_casei) || pcmk__str_eq("--cluster-id", option_name, pcmk__str_casei)) { options.command = 'i'; } else if (pcmk__str_eq("-l", option_name, pcmk__str_casei) || pcmk__str_eq("--list", option_name, pcmk__str_casei)) { options.command = 'l'; } else if (pcmk__str_eq("-n", option_name, pcmk__str_casei) || pcmk__str_eq("--name", option_name, pcmk__str_casei)) { options.command = 'n'; } else if (pcmk__str_eq("-p", option_name, pcmk__str_casei) || pcmk__str_eq("--partition", option_name, pcmk__str_casei)) { options.command = 'p'; } else if (pcmk__str_eq("-q", option_name, pcmk__str_casei) || pcmk__str_eq("--quorum", option_name, pcmk__str_casei)) { options.command = 'q'; } else { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "Unknown param passed to command_cb: %s", option_name); return FALSE; } return TRUE; } gboolean name_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.command = 'N'; pcmk__scan_min_int(optarg, &(options.nodeid), 0); return TRUE; } gboolean remove_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (optarg == NULL) { g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_INVALID_PARAM, "-R option requires an argument"); return FALSE; } options.command = 'R'; options.dangerous_cmd = TRUE; pcmk__str_update(&options.target_uname, optarg); return TRUE; } PCMK__OUTPUT_ARGS("node-id", "uint32_t") static int node_id_default(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); out->info(out, "%" PRIu32, node_id); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-id", "uint32_t") static int node_id_xml(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); char *id_s = crm_strdup_printf("%" PRIu32, node_id); pcmk__output_create_xml_node(out, "node-info", "nodeid", id_s, NULL); free(id_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-list", "GList *") static int node_list_default(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); for (GList *node_iter = nodes; node_iter != NULL; node_iter = node_iter->next) { pcmk_controld_api_node_t *node = node_iter->data; out->info(out, "%" PRIu32 " %s %s", node->id, pcmk__s(node->uname, ""), pcmk__s(node->state, "")); } return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-list", "GList *") static int node_list_xml(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); out->begin_list(out, NULL, NULL, "nodes"); for (GList *node_iter = nodes; node_iter != NULL; node_iter = node_iter->next) { pcmk_controld_api_node_t *node = node_iter->data; char *id_s = crm_strdup_printf("%" PRIu32, node->id); pcmk__output_create_xml_node(out, "node", PCMK_XA_ID, id_s, PCMK_XA_NAME, node->uname, "state", node->state, NULL); free(id_s); } out->end_list(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-name", "uint32_t", "const char *") static int node_name_default(pcmk__output_t *out, va_list args) { uint32_t node_id G_GNUC_UNUSED = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); out->info(out, "%s", node_name); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("node-name", "uint32_t", "const char *") static int node_name_xml(pcmk__output_t *out, va_list args) { uint32_t node_id = va_arg(args, uint32_t); const char *node_name = va_arg(args, const char *); char *id_s = crm_strdup_printf("%" PRIu32, node_id); pcmk__output_create_xml_node(out, "node-info", "nodeid", id_s, - XML_ATTR_UNAME, node_name, + PCMK_XA_UNAME, node_name, NULL); free(id_s); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("partition-list", "GList *") static int partition_list_default(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); GString *buffer = NULL; for (GList *node_iter = nodes; node_iter != NULL; node_iter = node_iter->next) { pcmk_controld_api_node_t *node = node_iter->data; if (pcmk__str_eq(node->state, "member", pcmk__str_none)) { pcmk__add_separated_word(&buffer, 128, pcmk__s(node->uname, ""), " "); } } if (buffer != NULL) { out->info(out, "%s", buffer->str); g_string_free(buffer, TRUE); return pcmk_rc_ok; } return pcmk_rc_no_output; } PCMK__OUTPUT_ARGS("partition-list", "GList *") static int partition_list_xml(pcmk__output_t *out, va_list args) { GList *nodes = va_arg(args, GList *); out->begin_list(out, NULL, NULL, "nodes"); for (GList *node_iter = nodes; node_iter != NULL; node_iter = node_iter->next) { pcmk_controld_api_node_t *node = node_iter->data; if (pcmk__str_eq(node->state, "member", pcmk__str_none)) { char *id_s = crm_strdup_printf("%" PRIu32, node->id); pcmk__output_create_xml_node(out, "node", PCMK_XA_ID, id_s, PCMK_XA_NAME, node->uname, "state", node->state, NULL); free(id_s); } } out->end_list(out); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("quorum", "bool") static int quorum_default(pcmk__output_t *out, va_list args) { bool have_quorum = va_arg(args, int); out->info(out, "%d", have_quorum); return pcmk_rc_ok; } PCMK__OUTPUT_ARGS("quorum", "bool") static int quorum_xml(pcmk__output_t *out, va_list args) { bool have_quorum = va_arg(args, int); pcmk__output_create_xml_node(out, "cluster-info", "quorum", have_quorum ? "true" : "false", NULL); return pcmk_rc_ok; } static pcmk__message_entry_t fmt_functions[] = { { "node-id", "default", node_id_default }, { "node-id", "xml", node_id_xml }, { "node-list", "default", node_list_default }, { "node-list", "xml", node_list_xml }, { "node-name", "default", node_name_default }, { "node-name", "xml", node_name_xml }, { "quorum", "default", quorum_default }, { "quorum", "xml", quorum_xml }, { "partition-list", "default", partition_list_default }, { "partition-list", "xml", partition_list_xml }, { NULL, NULL, NULL } }; static gint sort_node(gconstpointer a, gconstpointer b) { const pcmk_controld_api_node_t *node_a = a; const pcmk_controld_api_node_t *node_b = b; return pcmk__numeric_strcasecmp((node_a->uname? node_a->uname : ""), (node_b->uname? node_b->uname : "")); } static void controller_event_cb(pcmk_ipc_api_t *controld_api, enum pcmk_ipc_event event_type, crm_exit_t status, void *event_data, void *user_data) { pcmk_controld_api_reply_t *reply = event_data; switch (event_type) { case pcmk_ipc_event_disconnect: if (exit_code == CRM_EX_DISCONNECT) { // Unexpected g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Lost connection to controller"); } goto done; break; case pcmk_ipc_event_reply: break; default: return; } if (status != CRM_EX_OK) { exit_code = status; g_set_error(&error, PCMK__EXITC_ERROR, status, "Bad reply from controller: %s", crm_exit_str(status)); goto done; } if (reply->reply_type != pcmk_controld_reply_nodes) { g_set_error(&error, PCMK__EXITC_ERROR, CRM_EX_INDETERMINATE, "Unknown reply type %d from controller", reply->reply_type); goto done; } reply->data.nodes = g_list_sort(reply->data.nodes, sort_node); if (options.command == 'p') { out->message(out, "partition-list", reply->data.nodes); } else if (options.command == 'l') { out->message(out, "node-list", reply->data.nodes); } // Success exit_code = CRM_EX_OK; done: pcmk_disconnect_ipc(controld_api); pcmk_quit_main_loop(mainloop, 10); } static void run_controller_mainloop(void) { pcmk_ipc_api_t *controld_api = NULL; int rc; // Set disconnect exit code to handle unexpected disconnects exit_code = CRM_EX_DISCONNECT; // Create controller IPC object rc = pcmk_new_ipc_api(&controld_api, pcmk_ipc_controld); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not connect to controller: %s", pcmk_rc_str(rc)); return; } pcmk_register_ipc_callback(controld_api, controller_event_cb, NULL); // Connect to controller rc = pcmk__connect_ipc(controld_api, pcmk_ipc_dispatch_main, 5); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to %s: %s", pcmk_ipc_name(controld_api, true), pcmk_rc_str(rc)); return; } rc = pcmk_controld_api_list_nodes(controld_api); if (rc != pcmk_rc_ok) { pcmk_disconnect_ipc(controld_api); exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not ping controller: %s", pcmk_rc_str(rc)); return; } // Run main loop to get controller reply via controller_event_cb() mainloop = g_main_loop_new(NULL, FALSE); g_main_loop_run(mainloop); g_main_loop_unref(mainloop); mainloop = NULL; pcmk_free_ipc_api(controld_api); } static void print_node_id(void) { uint32_t nodeid; int rc = pcmk__query_node_info(out, &nodeid, NULL, NULL, NULL, NULL, NULL, false, 0); if (rc != pcmk_rc_ok) { /* pcmk__query_node_info already sets an error message on the output object, * so there's no need to call g_set_error here. That would just create a * duplicate error message in the output. */ exit_code = pcmk_rc2exitc(rc); return; } rc = out->message(out, "node-id", nodeid); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not print node ID: %s", pcmk_rc_str(rc)); } exit_code = pcmk_rc2exitc(rc); } static void print_node_name(uint32_t nodeid) { int rc = pcmk_rc_ok; char *node_name = NULL; if (nodeid == 0) { // Check environment first (i.e. when called by resource agent) const char *name = getenv("OCF_RESKEY_" CRM_META "_" XML_LRM_ATTR_TARGET); if (name != NULL) { rc = out->message(out, "node-name", 0, name); goto done; } } // Otherwise ask the controller /* pcmk__query_node_name already sets an error message on the output object, * so there's no need to call g_set_error here. That would just create a * duplicate error message in the output. */ rc = pcmk__query_node_name(out, nodeid, &node_name, 0); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); return; } rc = out->message(out, "node-name", 0, node_name); done: if (node_name != NULL) { free(node_name); } if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not print node name: %s", pcmk_rc_str(rc)); } exit_code = pcmk_rc2exitc(rc); } static void print_quorum(void) { bool quorum; int rc = pcmk__query_node_info(out, NULL, NULL, NULL, NULL, &quorum, NULL, false, 0); if (rc != pcmk_rc_ok) { /* pcmk__query_node_info already sets an error message on the output object, * so there's no need to call g_set_error here. That would just create a * duplicate error message in the output. */ exit_code = pcmk_rc2exitc(rc); return; } rc = out->message(out, "quorum", quorum); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not print quorum status: %s", pcmk_rc_str(rc)); } exit_code = pcmk_rc2exitc(rc); } /*! * \internal * \brief Extend a transaction by removing a node from a CIB section * * \param[in,out] cib Active CIB connection * \param[in] element CIB element containing node name and/or ID * \param[in] section CIB section that \p element is in * \param[in] node_name Name of node to purge (NULL to leave unspecified) * \param[in] node_id Node ID of node to purge (0 to leave unspecified) * * \note At least one of node_name and node_id must be specified. * \return Standard Pacemaker return code */ static int remove_from_section(cib_t *cib, const char *element, const char *section, const char *node_name, long node_id) { xmlNode *xml = NULL; int rc = pcmk_rc_ok; xml = create_xml_node(NULL, element); if (xml == NULL) { return pcmk_rc_error; } - crm_xml_add(xml, XML_ATTR_UNAME, node_name); + crm_xml_add(xml, PCMK_XA_UNAME, node_name); if (node_id > 0) { crm_xml_set_id(xml, "%ld", node_id); } rc = cib->cmds->remove(cib, section, xml, cib_transaction); free_xml(xml); return (rc >= 0)? pcmk_rc_ok : pcmk_legacy2rc(rc); } /*! * \internal * \brief Purge a node from CIB * * \param[in] node_name Name of node to purge (or NULL to leave unspecified) * \param[in] node_id Node ID of node to purge (or 0 to leave unspecified) * * \note At least one of node_name and node_id must be specified. * \return Standard Pacemaker return code */ static int purge_node_from_cib(const char *node_name, long node_id) { int rc = pcmk_rc_ok; int commit_rc = pcmk_rc_ok; cib_t *cib = NULL; // Connect to CIB and start a transaction cib = cib_new(); if (cib == NULL) { return ENOTCONN; } rc = cib->cmds->signon(cib, crm_system_name, cib_command); if (rc == pcmk_ok) { rc = cib->cmds->init_transaction(cib); } if (rc != pcmk_ok) { rc = pcmk_legacy2rc(rc); cib__clean_up_connection(&cib); return rc; } // Remove from configuration and status rc = remove_from_section(cib, XML_CIB_TAG_NODE, XML_CIB_TAG_NODES, node_name, node_id); if (rc == pcmk_rc_ok) { rc = remove_from_section(cib, XML_CIB_TAG_STATE, XML_CIB_TAG_STATUS, node_name, node_id); } // Commit the transaction commit_rc = cib->cmds->end_transaction(cib, (rc == pcmk_rc_ok), cib_sync_call); cib__clean_up_connection(&cib); if ((rc == pcmk_rc_ok) && (commit_rc == pcmk_ok)) { crm_debug("Purged node %s (%ld) from CIB", pcmk__s(node_name, "by ID"), node_id); } return rc; } /*! * \internal * \brief Purge a node from a single server's peer cache * * \param[in] server IPC server to send request to * \param[in] node_name Name of node to purge (or NULL to leave unspecified) * \param[in] node_id Node ID of node to purge (or 0 to leave unspecified) * * \note At least one of node_name and node_id must be specified. * \return Standard Pacemaker return code */ static int purge_node_from(enum pcmk_ipc_server server, const char *node_name, long node_id) { pcmk_ipc_api_t *api = NULL; int rc; rc = pcmk_new_ipc_api(&api, server); if (rc != pcmk_rc_ok) { goto done; } rc = pcmk__connect_ipc(api, pcmk_ipc_dispatch_sync, 5); if (rc != pcmk_rc_ok) { goto done; } rc = pcmk_ipc_purge_node(api, node_name, node_id); done: if (rc != pcmk_rc_ok) { // Debug message already logged on success g_set_error(&error, PCMK__RC_ERROR, rc, "Could not purge node %s from %s: %s", pcmk__s(node_name, "by ID"), pcmk_ipc_name(api, true), pcmk_rc_str(rc)); } pcmk_free_ipc_api(api); return rc; } /*! * \internal * \brief Purge a node from the fencer's peer cache * * \param[in] node_name Name of node to purge (or NULL to leave unspecified) * \param[in] node_id Node ID of node to purge (or 0 to leave unspecified) * * \note At least one of node_name and node_id must be specified. * \return Standard Pacemaker return code */ static int purge_node_from_fencer(const char *node_name, long node_id) { int rc = pcmk_rc_ok; crm_ipc_t *conn = NULL; xmlNode *cmd = NULL; conn = crm_ipc_new("stonith-ng", 0); if (conn == NULL) { rc = ENOTCONN; exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to fencer to purge node %s", pcmk__s(node_name, "by ID")); return rc; } rc = pcmk__connect_generic_ipc(conn); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Could not connect to fencer to purge node %s: %s", pcmk__s(node_name, "by ID"), pcmk_rc_str(rc)); crm_ipc_destroy(conn); return rc; } cmd = create_request(CRM_OP_RM_NODE_CACHE, NULL, NULL, "stonith-ng", crm_system_name, NULL); if (node_id > 0) { crm_xml_set_id(cmd, "%ld", node_id); } - crm_xml_add(cmd, XML_ATTR_UNAME, node_name); + crm_xml_add(cmd, PCMK_XA_UNAME, node_name); rc = crm_ipc_send(conn, cmd, 0, 0, NULL); if (rc >= 0) { rc = pcmk_rc_ok; crm_debug("Purged node %s (%ld) from fencer", pcmk__s(node_name, "by ID"), node_id); } else { rc = pcmk_legacy2rc(rc); fprintf(stderr, "Could not purge node %s from fencer: %s\n", pcmk__s(node_name, "by ID"), pcmk_rc_str(rc)); } free_xml(cmd); crm_ipc_close(conn); crm_ipc_destroy(conn); return rc; } static void remove_node(const char *target_uname) { int rc = pcmk_rc_ok; long nodeid = 0; const char *node_name = NULL; char *endptr = NULL; const enum pcmk_ipc_server servers[] = { pcmk_ipc_controld, pcmk_ipc_attrd, }; // Check whether node was specified by name or numeric ID errno = 0; nodeid = strtol(target_uname, &endptr, 10); if ((errno != 0) || (endptr == target_uname) || (*endptr != '\0') || (nodeid <= 0)) { // It's not a positive integer, so assume it's a node name nodeid = 0; node_name = target_uname; } for (int i = 0; i < PCMK__NELEM(servers); ++i) { rc = purge_node_from(servers[i], node_name, nodeid); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); return; } } // The fencer hasn't been converted to pcmk_ipc_api_t yet rc = purge_node_from_fencer(node_name, nodeid); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); return; } // Lastly, purge the node from the CIB itself rc = purge_node_from_cib(node_name, nodeid); exit_code = pcmk_rc2exitc(rc); } 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 } }; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); /* Add the -q option, which cannot be part of the globally supported options * because some tools use that flag for something else. */ pcmk__add_main_args(context, extra_prog_entries); pcmk__add_arg_group(context, "commands", "Commands:", "Show command help", command_entries); pcmk__add_arg_group(context, "additional", "Additional Options:", "Show additional options", addl_entries); return context; } int main(int argc, char **argv) { int rc = pcmk_rc_ok; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "NR"); GOptionContext *context = build_arg_context(args, &output_group); pcmk__register_formats(output_group, formats); if (!g_option_context_parse_strv(context, &processed_args, &error)) { exit_code = CRM_EX_USAGE; goto done; } pcmk__cli_init_logging("crm_node", args->verbosity); rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "Error creating output format %s: %s", args->output_ty, pcmk_rc_str(rc)); goto done; } if (!pcmk__force_args(context, &error, "%s --xml-simple-list", g_get_prgname())) { exit_code = CRM_EX_SOFTWARE; goto done; } if (args->version) { out->version(out, false); goto done; } if (options.command == 0) { char *help = g_option_context_get_help(context, TRUE, NULL); out->err(out, "%s", help); g_free(help); exit_code = CRM_EX_USAGE; goto done; } if (options.dangerous_cmd && options.force_flag == FALSE) { exit_code = CRM_EX_USAGE; g_set_error(&error, PCMK__EXITC_ERROR, exit_code, "The supplied command is considered dangerous." " To prevent accidental destruction of the cluster," " the --force flag is required in order to proceed."); goto done; } pcmk__register_lib_messages(out); pcmk__register_messages(out, fmt_functions); switch (options.command) { case 'i': print_node_id(); break; case 'n': print_node_name(0); break; case 'q': print_quorum(); break; case 'N': print_node_name(options.nodeid); break; case 'R': remove_node(options.target_uname); break; case 'l': case 'p': run_controller_mainloop(); break; default: break; } done: g_strfreev(processed_args); pcmk__free_arg_context(context); pcmk__output_and_clear_error(&error, out); if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } pcmk__unregister_formats(); return crm_exit(exit_code); } diff --git a/tools/crm_resource_runtime.c b/tools/crm_resource_runtime.c index 00648dfa83..3fe238523f 100644 --- a/tools/crm_resource_runtime.c +++ b/tools/crm_resource_runtime.c @@ -1,2254 +1,2254 @@ /* * Copyright 2004-2024 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include static GList * build_node_info_list(const pcmk_resource_t *rsc) { GList *retval = NULL; for (const GList *iter = rsc->children; iter != NULL; iter = iter->next) { const pcmk_resource_t *child = (const pcmk_resource_t *) iter->data; for (const GList *iter2 = child->running_on; iter2 != NULL; iter2 = iter2->next) { const pcmk_node_t *node = (const pcmk_node_t *) iter2->data; node_info_t *ni = calloc(1, sizeof(node_info_t)); ni->node_name = node->details->uname; ni->promoted = pcmk_is_set(rsc->flags, pcmk_rsc_promotable) && child->fns->state(child, TRUE) == pcmk_role_promoted; retval = g_list_prepend(retval, ni); } } return retval; } GList * cli_resource_search(pcmk_resource_t *rsc, const char *requested_name, pcmk_scheduler_t *scheduler) { GList *retval = NULL; const pcmk_resource_t *parent = pe__const_top_resource(rsc, false); if (pe_rsc_is_clone(rsc)) { retval = build_node_info_list(rsc); /* The anonymous clone children's common ID is supplied */ } else if (pe_rsc_is_clone(parent) && !pcmk_is_set(rsc->flags, pcmk_rsc_unique) && rsc->clone_name && pcmk__str_eq(requested_name, rsc->clone_name, pcmk__str_casei) && !pcmk__str_eq(requested_name, rsc->id, pcmk__str_casei)) { retval = build_node_info_list(parent); } else if (rsc->running_on != NULL) { for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) { pcmk_node_t *node = (pcmk_node_t *) iter->data; node_info_t *ni = calloc(1, sizeof(node_info_t)); ni->node_name = node->details->uname; ni->promoted = (rsc->fns->state(rsc, TRUE) == pcmk_role_promoted); retval = g_list_prepend(retval, ni); } } return retval; } // \return Standard Pacemaker return code static int find_resource_attr(pcmk__output_t *out, cib_t * the_cib, const char *attr, const char *rsc, const char *attr_set_type, const char *set_name, const char *attr_id, const char *attr_name, char **value) { int rc = pcmk_rc_ok; xmlNode *xml_search = NULL; GString *xpath = NULL; const char *xpath_base = NULL; if(value) { *value = NULL; } if(the_cib == NULL) { return ENOTCONN; } xpath_base = pcmk_cib_xpath_for(XML_CIB_TAG_RESOURCES); if (xpath_base == NULL) { crm_err(XML_CIB_TAG_RESOURCES " CIB element not known (bug?)"); return ENOMSG; } xpath = g_string_sized_new(1024); pcmk__g_strcat(xpath, xpath_base, "//*[@" PCMK_XA_ID "=\"", rsc, "\"]", NULL); if (attr_set_type != NULL) { pcmk__g_strcat(xpath, "/", attr_set_type, NULL); if (set_name != NULL) { pcmk__g_strcat(xpath, "[@" PCMK_XA_ID "=\"", set_name, "\"]", NULL); } } g_string_append(xpath, "//" XML_CIB_TAG_NVPAIR "["); if (attr_id != NULL) { pcmk__g_strcat(xpath, "@" PCMK_XA_ID "=\"", attr_id, "\"", NULL); } if (attr_name != NULL) { if (attr_id != NULL) { g_string_append(xpath, " and "); } pcmk__g_strcat(xpath, "@" PCMK_XA_NAME "=\"", attr_name, "\"", NULL); } g_string_append_c(xpath, ']'); rc = the_cib->cmds->query(the_cib, (const char *) xpath->str, &xml_search, cib_sync_call | cib_scope_local | cib_xpath); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { goto done; } crm_log_xml_debug(xml_search, "Match"); if (xml_search->children != NULL) { xmlNode *child = NULL; rc = ENOTUNIQ; out->info(out, "Multiple attributes match name=%s", attr_name); for (child = pcmk__xml_first_child(xml_search); child != NULL; child = pcmk__xml_next(child)) { out->info(out, " Value: %s \t(id=%s)", crm_element_value(child, PCMK_XA_VALUE), ID(child)); } out->spacer(out); } else if(value) { pcmk__str_update(value, crm_element_value(xml_search, attr)); } done: g_string_free(xpath, TRUE); free_xml(xml_search); return rc; } /* PRIVATE. Use the find_matching_attr_resources instead. */ static void find_matching_attr_resources_recursive(pcmk__output_t *out, GList /* */ **result, pcmk_resource_t *rsc, const char *rsc_id, const char * attr_set, const char * attr_set_type, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, int depth) { int rc = pcmk_rc_ok; char *lookup_id = clone_strip(rsc->id); char *local_attr_id = NULL; /* visit the children */ for(GList *gIter = rsc->children; gIter; gIter = gIter->next) { find_matching_attr_resources_recursive(out, result, (pcmk_resource_t *) gIter->data, rsc_id, attr_set, attr_set_type, attr_id, attr_name, cib, cmd, depth+1); /* do it only once for clones */ if (rsc->variant == pcmk_rsc_variant_clone) { break; } } rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); /* Post-order traversal. * The root is always on the list and it is the last item. */ if((0 == depth) || (pcmk_rc_ok == rc)) { /* push the head */ *result = g_list_append(*result, rsc); } free(local_attr_id); free(lookup_id); } /* The result is a linearized pre-ordered tree of resources. */ static GList/**/ * find_matching_attr_resources(pcmk__output_t *out, pcmk_resource_t *rsc, const char * rsc_id, const char * attr_set, const char * attr_set_type, const char * attr_id, const char * attr_name, cib_t * cib, const char * cmd, gboolean force) { int rc = pcmk_rc_ok; char *lookup_id = NULL; char *local_attr_id = NULL; GList * result = NULL; /* If --force is used, update only the requested resource (clone or primitive). * Otherwise, if the primitive has the attribute, use that. * Otherwise use the clone. */ if(force == TRUE) { return g_list_append(result, rsc); } if ((rsc->parent != NULL) && (rsc->parent->variant == pcmk_rsc_variant_clone)) { int rc = pcmk_rc_ok; char *local_attr_id = NULL; rc = find_resource_attr(out, cib, PCMK_XA_ID, rsc_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); free(local_attr_id); if(rc != pcmk_rc_ok) { rsc = rsc->parent; out->info(out, "Performing %s of '%s' on '%s', the parent of '%s'", cmd, attr_name, rsc->id, rsc_id); } return g_list_append(result, rsc); } else if ((rsc->parent == NULL) && (rsc->children != NULL) && (rsc->variant == pcmk_rsc_variant_clone)) { pcmk_resource_t *child = rsc->children->data; if (child->variant == pcmk_rsc_variant_primitive) { lookup_id = clone_strip(child->id); /* Could be a cloned group! */ rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &local_attr_id); if(rc == pcmk_rc_ok) { rsc = child; out->info(out, "A value for '%s' already exists in child '%s', performing %s on that instead of '%s'", attr_name, lookup_id, cmd, rsc_id); } free(local_attr_id); free(lookup_id); } return g_list_append(result, rsc); } /* If the resource is a group ==> children inherit the attribute if defined. */ find_matching_attr_resources_recursive(out, &result, rsc, rsc_id, attr_set, attr_set_type, attr_id, attr_name, cib, cmd, 0); return result; } // \return Standard Pacemaker return code int cli_resource_update_attribute(pcmk_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_set_type, const char *attr_id, const char *attr_name, const char *attr_value, gboolean recursive, cib_t *cib, int cib_options, gboolean force) { pcmk__output_t *out = rsc->cluster->priv; int rc = pcmk_rc_ok; char *found_attr_id = NULL; GList/**/ *resources = NULL; const char *top_id = pe__const_top_resource(rsc, false)->id; if ((attr_id == NULL) && !force) { find_resource_attr(out, cib, PCMK_XA_ID, top_id, NULL, NULL, NULL, attr_name, NULL); } if (pcmk__str_eq(attr_set_type, XML_TAG_ATTR_SETS, pcmk__str_casei)) { if (!force) { rc = find_resource_attr(out, cib, PCMK_XA_ID, top_id, XML_TAG_META_SETS, attr_set, attr_id, attr_name, &found_attr_id); if ((rc == pcmk_rc_ok) && !out->is_quiet(out)) { out->err(out, "WARNING: There is already a meta attribute " "for '%s' called '%s' (id=%s)", top_id, attr_name, found_attr_id); out->err(out, " Delete '%s' first or use the force option " "to override", found_attr_id); } free(found_attr_id); if (rc == pcmk_rc_ok) { return ENOTUNIQ; } } resources = g_list_append(resources, rsc); } else if (pcmk__str_eq(attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) { crm_xml_add(rsc->xml, attr_name, attr_value); CRM_ASSERT(cib != NULL); rc = cib->cmds->replace(cib, XML_CIB_TAG_RESOURCES, rsc->xml, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { out->info(out, "Set attribute: name=%s value=%s", attr_name, attr_value); } return rc; } else { resources = find_matching_attr_resources(out, rsc, requested_name, attr_set, attr_set_type, attr_id, attr_name, cib, "update", force); } /* If the user specified attr_set or attr_id, the intent is to modify a * single resource, which will be the last item in the list. */ if ((attr_set != NULL) || (attr_id != NULL)) { GList *last = g_list_last(resources); resources = g_list_remove_link(resources, last); g_list_free(resources); resources = last; } for (GList *iter = resources; iter != NULL; iter = iter->next) { char *lookup_id = NULL; char *local_attr_set = NULL; const char *rsc_attr_id = attr_id; const char *rsc_attr_set = attr_set; xmlNode *xml_top = NULL; xmlNode *xml_obj = NULL; found_attr_id = NULL; rsc = (pcmk_resource_t *) iter->data; lookup_id = clone_strip(rsc->id); /* Could be a cloned group! */ rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &found_attr_id); switch (rc) { case pcmk_rc_ok: crm_debug("Found a match for name=%s: id=%s", attr_name, found_attr_id); rsc_attr_id = found_attr_id; break; case ENXIO: if (rsc_attr_set == NULL) { local_attr_set = crm_strdup_printf("%s-%s", lookup_id, attr_set_type); rsc_attr_set = local_attr_set; } if (rsc_attr_id == NULL) { found_attr_id = crm_strdup_printf("%s-%s", rsc_attr_set, attr_name); rsc_attr_id = found_attr_id; } xml_top = create_xml_node(NULL, (const char *) rsc->xml->name); crm_xml_add(xml_top, PCMK_XA_ID, lookup_id); xml_obj = create_xml_node(xml_top, attr_set_type); crm_xml_add(xml_obj, PCMK_XA_ID, rsc_attr_set); break; default: free(lookup_id); free(found_attr_id); g_list_free(resources); return rc; } xml_obj = crm_create_nvpair_xml(xml_obj, rsc_attr_id, attr_name, attr_value); if (xml_top == NULL) { xml_top = xml_obj; } crm_log_xml_debug(xml_top, "Update"); rc = cib->cmds->modify(cib, XML_CIB_TAG_RESOURCES, xml_top, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { out->info(out, "Set '%s' option: id=%s%s%s%s%s value=%s", lookup_id, found_attr_id, ((rsc_attr_set == NULL)? "" : " set="), pcmk__s(rsc_attr_set, ""), ((attr_name == NULL)? "" : " name="), pcmk__s(attr_name, ""), attr_value); } free_xml(xml_top); free(lookup_id); free(found_attr_id); free(local_attr_set); if (recursive && pcmk__str_eq(attr_set_type, XML_TAG_META_SETS, pcmk__str_casei)) { GList *lpc = NULL; static bool need_init = true; if (need_init) { need_init = false; pcmk__unpack_constraints(rsc->cluster); pe__clear_resource_flags_on_all(rsc->cluster, pcmk_rsc_detect_loop); } /* We want to set the attribute only on resources explicitly * colocated with this one, so we use rsc->rsc_cons_lhs directly * rather than the with_this_colocations() method. */ pcmk__set_rsc_flags(rsc, pcmk_rsc_detect_loop); for (lpc = rsc->rsc_cons_lhs; lpc != NULL; lpc = lpc->next) { pcmk__colocation_t *cons = (pcmk__colocation_t *) lpc->data; crm_debug("Checking %s %d", cons->id, cons->score); if (!pcmk_is_set(cons->dependent->flags, pcmk_rsc_detect_loop) && (cons->score > 0)) { crm_debug("Setting %s=%s for dependent resource %s", attr_name, attr_value, cons->dependent->id); cli_resource_update_attribute(cons->dependent, cons->dependent->id, NULL, attr_set_type, NULL, attr_name, attr_value, recursive, cib, cib_options, force); } } } } g_list_free(resources); return rc; } // \return Standard Pacemaker return code int cli_resource_delete_attribute(pcmk_resource_t *rsc, const char *requested_name, const char *attr_set, const char *attr_set_type, const char *attr_id, const char *attr_name, cib_t *cib, int cib_options, gboolean force) { pcmk__output_t *out = rsc->cluster->priv; int rc = pcmk_rc_ok; GList/**/ *resources = NULL; if ((attr_id == NULL) && !force) { find_resource_attr(out, cib, PCMK_XA_ID, pe__const_top_resource(rsc, false)->id, NULL, NULL, NULL, attr_name, NULL); } if (pcmk__str_eq(attr_set_type, XML_TAG_META_SETS, pcmk__str_casei)) { resources = find_matching_attr_resources(out, rsc, requested_name, attr_set, attr_set_type, attr_id, attr_name, cib, "delete", force); } else if (pcmk__str_eq(attr_set_type, ATTR_SET_ELEMENT, pcmk__str_none)) { xml_remove_prop(rsc->xml, attr_name); CRM_ASSERT(cib != NULL); rc = cib->cmds->replace(cib, XML_CIB_TAG_RESOURCES, rsc->xml, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { out->info(out, "Deleted attribute: %s", attr_name); } return rc; } else { resources = g_list_append(resources, rsc); } for (GList *iter = resources; iter != NULL; iter = iter->next) { char *lookup_id = NULL; xmlNode *xml_obj = NULL; char *found_attr_id = NULL; const char *rsc_attr_id = attr_id; rsc = (pcmk_resource_t *) iter->data; lookup_id = clone_strip(rsc->id); rc = find_resource_attr(out, cib, PCMK_XA_ID, lookup_id, attr_set_type, attr_set, attr_id, attr_name, &found_attr_id); switch (rc) { case pcmk_rc_ok: break; case ENXIO: free(lookup_id); rc = pcmk_rc_ok; continue; default: free(lookup_id); g_list_free(resources); return rc; } if (rsc_attr_id == NULL) { rsc_attr_id = found_attr_id; } xml_obj = crm_create_nvpair_xml(NULL, rsc_attr_id, attr_name, NULL); crm_log_xml_debug(xml_obj, "Delete"); CRM_ASSERT(cib); rc = cib->cmds->remove(cib, XML_CIB_TAG_RESOURCES, xml_obj, cib_options); rc = pcmk_legacy2rc(rc); if (rc == pcmk_rc_ok) { out->info(out, "Deleted '%s' option: id=%s%s%s%s%s", lookup_id, found_attr_id, ((attr_set == NULL)? "" : " set="), pcmk__s(attr_set, ""), ((attr_name == NULL)? "" : " name="), pcmk__s(attr_name, "")); } free(lookup_id); free_xml(xml_obj); free(found_attr_id); } g_list_free(resources); return rc; } // \return Standard Pacemaker return code static int send_lrm_rsc_op(pcmk_ipc_api_t *controld_api, bool do_fail_resource, const char *host_uname, const char *rsc_id, pcmk_scheduler_t *scheduler) { pcmk__output_t *out = scheduler->priv; const char *router_node = host_uname; const char *rsc_api_id = NULL; const char *rsc_long_id = NULL; const char *rsc_class = NULL; const char *rsc_provider = NULL; const char *rsc_type = NULL; bool cib_only = false; pcmk_resource_t *rsc = pe_find_resource(scheduler->resources, rsc_id); if (rsc == NULL) { out->err(out, "Resource %s not found", rsc_id); return ENXIO; } else if (rsc->variant != pcmk_rsc_variant_primitive) { out->err(out, "We can only process primitive resources, not %s", rsc_id); return EINVAL; } rsc_class = crm_element_value(rsc->xml, PCMK_XA_CLASS); rsc_provider = crm_element_value(rsc->xml, PCMK_XA_PROVIDER), rsc_type = crm_element_value(rsc->xml, PCMK_XA_TYPE); if ((rsc_class == NULL) || (rsc_type == NULL)) { out->err(out, "Resource %s does not have a class and type", rsc_id); return EINVAL; } { pcmk_node_t *node = pe_find_node(scheduler->nodes, host_uname); if (node == NULL) { out->err(out, "Node %s not found", host_uname); return pcmk_rc_node_unknown; } if (!(node->details->online)) { if (do_fail_resource) { out->err(out, "Node %s is not online", host_uname); return ENOTCONN; } else { cib_only = true; } } if (!cib_only && pe__is_guest_or_remote_node(node)) { node = pe__current_node(node->details->remote_rsc); if (node == NULL) { out->err(out, "No cluster connection to Pacemaker Remote node %s detected", host_uname); return ENOTCONN; } router_node = node->details->uname; } } if (rsc->clone_name) { rsc_api_id = rsc->clone_name; rsc_long_id = rsc->id; } else { rsc_api_id = rsc->id; } if (do_fail_resource) { return pcmk_controld_api_fail(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type); } else { return pcmk_controld_api_refresh(controld_api, host_uname, router_node, rsc_api_id, rsc_long_id, rsc_class, rsc_provider, rsc_type, cib_only); } } /*! * \internal * \brief Get resource name as used in failure-related node attributes * * \param[in] rsc Resource to check * * \return Newly allocated string containing resource's fail name * \note The caller is responsible for freeing the result. */ static inline char * rsc_fail_name(const pcmk_resource_t *rsc) { const char *name = (rsc->clone_name? rsc->clone_name : rsc->id); if (pcmk_is_set(rsc->flags, pcmk_rsc_unique)) { return strdup(name); } return clone_strip(name); } // \return Standard Pacemaker return code static int clear_rsc_history(pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pcmk_scheduler_t *scheduler) { int rc = pcmk_rc_ok; /* Erase the resource's entire LRM history in the CIB, even if we're only * clearing a single operation's fail count. If we erased only entries for a * single operation, we might wind up with a wrong idea of the current * resource state, and we might not re-probe the resource. */ rc = send_lrm_rsc_op(controld_api, false, host_uname, rsc_id, scheduler); if (rc != pcmk_rc_ok) { return rc; } crm_trace("Processing %d mainloop inputs", pcmk_controld_api_replies_expected(controld_api)); while (g_main_context_iteration(NULL, FALSE)) { crm_trace("Processed mainloop input, %d still remaining", pcmk_controld_api_replies_expected(controld_api)); } return rc; } // \return Standard Pacemaker return code static int clear_rsc_failures(pcmk__output_t *out, pcmk_ipc_api_t *controld_api, const char *node_name, const char *rsc_id, const char *operation, const char *interval_spec, pcmk_scheduler_t *scheduler) { int rc = pcmk_rc_ok; const char *failed_value = NULL; const char *failed_id = NULL; char *interval_ms_s = NULL; GHashTable *rscs = NULL; GHashTableIter iter; /* Create a hash table to use as a set of resources to clean. This lets us * clean each resource only once (per node) regardless of how many failed * operations it has. */ rscs = pcmk__strkey_table(NULL, NULL); // Normalize interval to milliseconds for comparison to history entry if (operation) { guint interval_ms = 0U; pcmk_parse_interval_spec(interval_spec, &interval_ms); interval_ms_s = crm_strdup_printf("%u", interval_ms); } for (xmlNode *xml_op = pcmk__xml_first_child(scheduler->failed); xml_op != NULL; xml_op = pcmk__xml_next(xml_op)) { failed_id = crm_element_value(xml_op, XML_LRM_ATTR_RSCID); if (failed_id == NULL) { // Malformed history entry, should never happen continue; } // No resource specified means all resources match if (rsc_id) { pcmk_resource_t *fail_rsc = NULL; fail_rsc = pe_find_resource_with_flags(scheduler->resources, failed_id, pcmk_rsc_match_history |pcmk_rsc_match_anon_basename); if (!fail_rsc || !pcmk__str_eq(rsc_id, fail_rsc->id, pcmk__str_casei)) { continue; } } // Host name should always have been provided by this point - failed_value = crm_element_value(xml_op, XML_ATTR_UNAME); + failed_value = crm_element_value(xml_op, PCMK_XA_UNAME); if (!pcmk__str_eq(node_name, failed_value, pcmk__str_casei)) { continue; } // No operation specified means all operations match if (operation) { failed_value = crm_element_value(xml_op, XML_LRM_ATTR_TASK); if (!pcmk__str_eq(operation, failed_value, pcmk__str_casei)) { continue; } // Interval (if operation was specified) defaults to 0 (not all) failed_value = crm_element_value(xml_op, XML_LRM_ATTR_INTERVAL_MS); if (!pcmk__str_eq(interval_ms_s, failed_value, pcmk__str_casei)) { continue; } } g_hash_table_add(rscs, (gpointer) failed_id); } free(interval_ms_s); g_hash_table_iter_init(&iter, rscs); while (g_hash_table_iter_next(&iter, (gpointer *) &failed_id, NULL)) { crm_debug("Erasing failures of %s on %s", failed_id, node_name); rc = clear_rsc_history(controld_api, node_name, failed_id, scheduler); if (rc != pcmk_rc_ok) { return rc; } } g_hash_table_destroy(rscs); return rc; } // \return Standard Pacemaker return code static int clear_rsc_fail_attrs(const pcmk_resource_t *rsc, const char *operation, const char *interval_spec, const pcmk_node_t *node) { int rc = pcmk_rc_ok; int attr_options = pcmk__node_attr_none; char *rsc_name = rsc_fail_name(rsc); if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } rc = pcmk__attrd_api_clear_failures(NULL, node->details->uname, rsc_name, operation, interval_spec, NULL, attr_options); free(rsc_name); return rc; } // \return Standard Pacemaker return code int cli_resource_delete(pcmk_ipc_api_t *controld_api, const char *host_uname, const pcmk_resource_t *rsc, const char *operation, const char *interval_spec, bool just_failures, pcmk_scheduler_t *scheduler, gboolean force) { pcmk__output_t *out = scheduler->priv; int rc = pcmk_rc_ok; pcmk_node_t *node = NULL; if (rsc == NULL) { return ENXIO; } else if (rsc->children) { for (const GList *lpc = rsc->children; lpc != NULL; lpc = lpc->next) { const pcmk_resource_t *child = (const pcmk_resource_t *) lpc->data; rc = cli_resource_delete(controld_api, host_uname, child, operation, interval_spec, just_failures, scheduler, force); if (rc != pcmk_rc_ok) { return rc; } } return pcmk_rc_ok; } else if (host_uname == NULL) { GList *lpc = NULL; GList *nodes = g_hash_table_get_values(rsc->known_on); if(nodes == NULL && force) { nodes = pcmk__copy_node_list(scheduler->nodes, false); } else if(nodes == NULL && rsc->exclusive_discover) { GHashTableIter iter; pcmk_node_t *node = NULL; g_hash_table_iter_init(&iter, rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void**)&node)) { if(node->weight >= 0) { nodes = g_list_prepend(nodes, node); } } } else if(nodes == NULL) { nodes = g_hash_table_get_values(rsc->allowed_nodes); } for (lpc = nodes; lpc != NULL; lpc = lpc->next) { node = (pcmk_node_t *) lpc->data; if (node->details->online) { rc = cli_resource_delete(controld_api, node->details->uname, rsc, operation, interval_spec, just_failures, scheduler, force); } if (rc != pcmk_rc_ok) { g_list_free(nodes); return rc; } } g_list_free(nodes); return pcmk_rc_ok; } node = pe_find_node(scheduler->nodes, host_uname); if (node == NULL) { out->err(out, "Unable to clean up %s because node %s not found", rsc->id, host_uname); return ENODEV; } if (!node->details->rsc_discovery_enabled) { out->err(out, "Unable to clean up %s because resource discovery disabled on %s", rsc->id, host_uname); return EOPNOTSUPP; } if (controld_api == NULL) { out->err(out, "Dry run: skipping clean-up of %s on %s due to CIB_file", rsc->id, host_uname); return pcmk_rc_ok; } rc = clear_rsc_fail_attrs(rsc, operation, interval_spec, node); if (rc != pcmk_rc_ok) { out->err(out, "Unable to clean up %s failures on %s: %s", rsc->id, host_uname, pcmk_rc_str(rc)); return rc; } if (just_failures) { rc = clear_rsc_failures(out, controld_api, host_uname, rsc->id, operation, interval_spec, scheduler); } else { rc = clear_rsc_history(controld_api, host_uname, rsc->id, scheduler); } if (rc != pcmk_rc_ok) { out->err(out, "Cleaned %s failures on %s, but unable to clean history: %s", rsc->id, host_uname, pcmk_rc_str(rc)); } else { out->info(out, "Cleaned up %s on %s", rsc->id, host_uname); } return rc; } // \return Standard Pacemaker return code int cli_cleanup_all(pcmk_ipc_api_t *controld_api, const char *node_name, const char *operation, const char *interval_spec, pcmk_scheduler_t *scheduler) { pcmk__output_t *out = scheduler->priv; int rc = pcmk_rc_ok; int attr_options = pcmk__node_attr_none; const char *display_name = node_name? node_name : "all nodes"; if (controld_api == NULL) { out->info(out, "Dry run: skipping clean-up of %s due to CIB_file", display_name); return rc; } if (node_name) { pcmk_node_t *node = pe_find_node(scheduler->nodes, node_name); if (node == NULL) { out->err(out, "Unknown node: %s", node_name); return ENXIO; } if (pe__is_guest_or_remote_node(node)) { attr_options |= pcmk__node_attr_remote; } } rc = pcmk__attrd_api_clear_failures(NULL, node_name, NULL, operation, interval_spec, NULL, attr_options); if (rc != pcmk_rc_ok) { out->err(out, "Unable to clean up all failures on %s: %s", display_name, pcmk_rc_str(rc)); return rc; } if (node_name) { rc = clear_rsc_failures(out, controld_api, node_name, NULL, operation, interval_spec, scheduler); if (rc != pcmk_rc_ok) { out->err(out, "Cleaned all resource failures on %s, but unable to clean history: %s", node_name, pcmk_rc_str(rc)); return rc; } } else { for (GList *iter = scheduler->nodes; iter; iter = iter->next) { pcmk_node_t *node = (pcmk_node_t *) iter->data; rc = clear_rsc_failures(out, controld_api, node->details->uname, NULL, operation, interval_spec, scheduler); if (rc != pcmk_rc_ok) { out->err(out, "Cleaned all resource failures on all nodes, but unable to clean history: %s", pcmk_rc_str(rc)); return rc; } } } out->info(out, "Cleaned up all resources on %s", display_name); return rc; } static void check_role(resource_checks_t *checks) { const char *role_s = g_hash_table_lookup(checks->rsc->meta, PCMK_META_TARGET_ROLE); if (role_s == NULL) { return; } switch (text2role(role_s)) { case pcmk_role_stopped: checks->flags |= rsc_remain_stopped; break; case pcmk_role_unpromoted: if (pcmk_is_set(pe__const_top_resource(checks->rsc, false)->flags, pcmk_rsc_promotable)) { checks->flags |= rsc_unpromotable; } break; default: break; } } static void check_managed(resource_checks_t *checks) { const char *managed_s = g_hash_table_lookup(checks->rsc->meta, PCMK_META_IS_MANAGED); if ((managed_s != NULL) && !crm_is_true(managed_s)) { checks->flags |= rsc_unmanaged; } } static void check_locked(resource_checks_t *checks) { if (checks->rsc->lock_node != NULL) { checks->flags |= rsc_locked; checks->lock_node = checks->rsc->lock_node->details->uname; } } static bool node_is_unhealthy(pcmk_node_t *node) { switch (pe__health_strategy(node->details->data_set)) { case pcmk__health_strategy_none: break; case pcmk__health_strategy_no_red: if (pe__node_health(node) < 0) { return true; } break; case pcmk__health_strategy_only_green: if (pe__node_health(node) <= 0) { return true; } break; case pcmk__health_strategy_progressive: case pcmk__health_strategy_custom: /* @TODO These are finite scores, possibly with rules, and possibly * combining with other scores, so attributing these as a cause is * nontrivial. */ break; } return false; } static void check_node_health(resource_checks_t *checks, pcmk_node_t *node) { if (node == NULL) { GHashTableIter iter; bool allowed = false; bool all_nodes_unhealthy = true; g_hash_table_iter_init(&iter, checks->rsc->allowed_nodes); while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) { allowed = true; if (!node_is_unhealthy(node)) { all_nodes_unhealthy = false; break; } } if (allowed && all_nodes_unhealthy) { checks->flags |= rsc_node_health; } } else if (node_is_unhealthy(node)) { checks->flags |= rsc_node_health; } } int cli_resource_check(pcmk__output_t *out, pcmk_resource_t *rsc, pcmk_node_t *node) { resource_checks_t checks = { .rsc = rsc }; check_role(&checks); check_managed(&checks); check_locked(&checks); check_node_health(&checks, node); return out->message(out, "resource-check-list", &checks); } // \return Standard Pacemaker return code int cli_resource_fail(pcmk_ipc_api_t *controld_api, const char *host_uname, const char *rsc_id, pcmk_scheduler_t *scheduler) { crm_notice("Failing %s on %s", rsc_id, host_uname); return send_lrm_rsc_op(controld_api, true, host_uname, rsc_id, scheduler); } static GHashTable * generate_resource_params(pcmk_resource_t *rsc, pcmk_node_t *node, pcmk_scheduler_t *scheduler) { GHashTable *params = NULL; GHashTable *meta = NULL; GHashTable *combined = NULL; GHashTableIter iter; char *key = NULL; char *value = NULL; combined = pcmk__strkey_table(free, free); params = pe_rsc_params(rsc, node, scheduler); if (params != NULL) { g_hash_table_iter_init(&iter, params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { g_hash_table_insert(combined, strdup(key), strdup(value)); } } meta = pcmk__strkey_table(free, free); get_meta_attributes(meta, rsc, node, scheduler); if (meta != NULL) { g_hash_table_iter_init(&iter, meta); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { char *crm_name = crm_meta_name(key); g_hash_table_insert(combined, crm_name, strdup(value)); } g_hash_table_destroy(meta); } return combined; } bool resource_is_running_on(pcmk_resource_t *rsc, const char *host) { bool found = true; GList *hIter = NULL; GList *hosts = NULL; if (rsc == NULL) { return false; } rsc->fns->location(rsc, &hosts, TRUE); for (hIter = hosts; host != NULL && hIter != NULL; hIter = hIter->next) { pcmk_node_t *node = (pcmk_node_t *) hIter->data; if (pcmk__strcase_any_of(host, node->details->uname, node->details->id, NULL)) { crm_trace("Resource %s is running on %s\n", rsc->id, host); goto done; } } if (host != NULL) { crm_trace("Resource %s is not running on: %s\n", rsc->id, host); found = false; } else if(host == NULL && hosts == NULL) { crm_trace("Resource %s is not running\n", rsc->id); found = false; } done: g_list_free(hosts); return found; } /*! * \internal * \brief Create a list of all resources active on host from a given list * * \param[in] host Name of host to check whether resources are active * \param[in] rsc_list List of resources to check * * \return New list of resources from list that are active on host */ static GList * get_active_resources(const char *host, GList *rsc_list) { GList *rIter = NULL; GList *active = NULL; for (rIter = rsc_list; rIter != NULL; rIter = rIter->next) { pcmk_resource_t *rsc = (pcmk_resource_t *) rIter->data; /* Expand groups to their members, because if we're restarting a member * other than the first, we can't otherwise tell which resources are * stopping and starting. */ if (rsc->variant == pcmk_rsc_variant_group) { active = g_list_concat(active, get_active_resources(host, rsc->children)); } else if (resource_is_running_on(rsc, host)) { active = g_list_append(active, strdup(rsc->id)); } } return active; } static void dump_list(GList *items, const char *tag) { int lpc = 0; GList *item = NULL; for (item = items; item != NULL; item = item->next) { crm_trace("%s[%d]: %s", tag, lpc, (char*)item->data); lpc++; } } static void display_list(pcmk__output_t *out, GList *items, const char *tag) { GList *item = NULL; for (item = items; item != NULL; item = item->next) { out->info(out, "%s%s", tag, (const char *)item->data); } } /*! * \internal * \brief Upgrade XML to latest schema version and use it as scheduler input * * This also updates the scheduler timestamp to the current time. * * \param[in,out] scheduler Scheduler data to update * \param[in,out] xml XML to use as input * * \return Standard Pacemaker return code * \note On success, caller is responsible for freeing memory allocated for * scheduler->now. * \todo This follows the example of other callers of cli_config_update() * and returns ENOKEY ("Required key not available") if that fails, * but perhaps pcmk_rc_schema_validation would be better in that case. */ int update_scheduler_input(pcmk_scheduler_t *scheduler, xmlNode **xml) { if (cli_config_update(xml, NULL, FALSE) == FALSE) { return ENOKEY; } scheduler->input = *xml; scheduler->now = crm_time_new(NULL); return pcmk_rc_ok; } /*! * \internal * \brief Update scheduler XML input based on a CIB query * * \param[in] scheduler Scheduler data to initialize * \param[in] cib Connection to the CIB manager * * \return Standard Pacemaker return code * \note On success, caller is responsible for freeing memory allocated for * scheduler->input and scheduler->now. */ static int update_scheduler_input_to_cib(pcmk__output_t *out, pcmk_scheduler_t *scheduler, cib_t *cib) { xmlNode *cib_xml_copy = NULL; int rc = pcmk_rc_ok; rc = cib->cmds->query(cib, NULL, &cib_xml_copy, cib_scope_local | cib_sync_call); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { out->err(out, "Could not obtain the current CIB: %s (%d)", pcmk_rc_str(rc), rc); return rc; } rc = update_scheduler_input(scheduler, &cib_xml_copy); if (rc != pcmk_rc_ok) { out->err(out, "Could not upgrade the current CIB XML"); free_xml(cib_xml_copy); return rc; } return rc; } // \return Standard Pacemaker return code static int update_dataset(cib_t *cib, pcmk_scheduler_t *scheduler, bool simulate) { char *pid = NULL; char *shadow_file = NULL; cib_t *shadow_cib = NULL; int rc = pcmk_rc_ok; pcmk__output_t *out = scheduler->priv; pe_reset_working_set(scheduler); pcmk__set_scheduler_flags(scheduler, pcmk_sched_no_counts|pcmk_sched_no_compat); rc = update_scheduler_input_to_cib(out, scheduler, cib); if (rc != pcmk_rc_ok) { return rc; } if(simulate) { bool prev_quiet = false; pid = pcmk__getpid_s(); shadow_cib = cib_shadow_new(pid); shadow_file = get_shadow_file(pid); if (shadow_cib == NULL) { out->err(out, "Could not create shadow cib: '%s'", pid); rc = ENXIO; goto done; } rc = write_xml_file(scheduler->input, shadow_file, FALSE); if (rc < 0) { out->err(out, "Could not populate shadow cib: %s (%d)", pcmk_strerror(rc), rc); goto done; } rc = shadow_cib->cmds->signon(shadow_cib, crm_system_name, cib_command); rc = pcmk_legacy2rc(rc); if (rc != pcmk_rc_ok) { out->err(out, "Could not connect to shadow cib: %s (%d)", pcmk_rc_str(rc), rc); goto done; } pcmk__schedule_actions(scheduler->input, pcmk_sched_no_counts|pcmk_sched_no_compat, scheduler); prev_quiet = out->is_quiet(out); out->quiet = true; pcmk__simulate_transition(scheduler, shadow_cib, NULL); out->quiet = prev_quiet; rc = update_dataset(shadow_cib, scheduler, false); } else { cluster_status(scheduler); } done: // Do not free scheduler->input here, we need rsc->xml to be valid later on cib_delete(shadow_cib); free(pid); if(shadow_file) { unlink(shadow_file); free(shadow_file); } return rc; } /*! * \internal * \brief Find the maximum stop timeout of a resource and its children (if any) * * \param[in,out] rsc Resource to get timeout for * * \return Maximum stop timeout for \p rsc (in milliseconds) */ static int max_rsc_stop_timeout(pcmk_resource_t *rsc) { long long result_ll; int max_delay = 0; xmlNode *config = NULL; GHashTable *meta = NULL; if (rsc == NULL) { return 0; } // If resource is collective, use maximum of its children's stop timeouts if (rsc->children != NULL) { for (GList *iter = rsc->children; iter; iter = iter->next) { pcmk_resource_t *child = iter->data; int delay = max_rsc_stop_timeout(child); if (delay > max_delay) { pcmk__rsc_trace(rsc, "Maximum stop timeout for %s is now %s " "due to %s", rsc->id, pcmk__readable_interval(delay), child->id); max_delay = delay; } } return max_delay; } // Get resource's stop action configuration from CIB config = pcmk__find_action_config(rsc, PCMK_ACTION_STOP, 0, true); /* Get configured timeout for stop action (fully evaluated for rules, * defaults, etc.). * * @TODO This currently ignores node (which might matter for rules) */ meta = pcmk__unpack_action_meta(rsc, NULL, PCMK_ACTION_STOP, 0, config); if ((pcmk__scan_ll(g_hash_table_lookup(meta, XML_ATTR_TIMEOUT), &result_ll, -1LL) == pcmk_rc_ok) && (result_ll >= 0) && (result_ll <= INT_MAX)) { max_delay = (int) result_ll; } g_hash_table_destroy(meta); return max_delay; } /*! * \internal * \brief Find a reasonable waiting time for stopping any one resource in a list * * \param[in,out] scheduler Scheduler data * \param[in] resources List of names of resources that will be stopped * * \return Rough estimate of a reasonable time to wait (in seconds) to stop any * one resource in \p resources * \note This estimate is very rough, simply the maximum stop timeout of all * given resources and their children, plus a small fudge factor. It does * not account for children that must be stopped in sequence, action * throttling, or any demotions needed. It checks the stop timeout, even * if the resources in question are actually being started. */ static int wait_time_estimate(pcmk_scheduler_t *scheduler, const GList *resources) { int max_delay = 0; // Find maximum stop timeout in milliseconds for (const GList *item = resources; item != NULL; item = item->next) { pcmk_resource_t *rsc = pe_find_resource(scheduler->resources, (const char *) item->data); int delay = max_rsc_stop_timeout(rsc); if (delay > max_delay) { pcmk__rsc_trace(rsc, "Wait time is now %s due to %s", pcmk__readable_interval(delay), rsc->id); max_delay = delay; } } return (max_delay / 1000) + 5; } #define waiting_for_starts(d, r, h) ((d != NULL) || \ (!resource_is_running_on((r), (h)))) /*! * \internal * \brief Restart a resource (on a particular host if requested). * * \param[in,out] out Output object * \param[in,out] rsc The resource to restart * \param[in] node Node to restart resource on (NULL for all) * \param[in] move_lifetime If not NULL, how long constraint should * remain in effect (as ISO 8601 string) * \param[in] timeout_ms Consider failed if actions do not complete * in this time (specified in milliseconds, * but a two-second granularity is actually * used; if 0, it will be calculated based on * the resource timeout) * \param[in,out] cib Connection to the CIB manager * \param[in] cib_options Group of enum cib_call_options flags to * use with CIB calls * \param[in] promoted_role_only If true, limit to promoted instances * \param[in] force If true, apply only to requested instance * if part of a collective resource * * \return Standard Pacemaker return code (exits on certain failures) */ int cli_resource_restart(pcmk__output_t *out, pcmk_resource_t *rsc, const pcmk_node_t *node, const char *move_lifetime, int timeout_ms, cib_t *cib, int cib_options, gboolean promoted_role_only, gboolean force) { int rc = pcmk_rc_ok; int lpc = 0; int before = 0; int step_timeout_s = 0; int sleep_interval = 2; int timeout = timeout_ms / 1000; bool stop_via_ban = false; char *rsc_id = NULL; char *lookup_id = NULL; char *orig_target_role = NULL; GList *list_delta = NULL; GList *target_active = NULL; GList *current_active = NULL; GList *restart_target_active = NULL; pcmk_scheduler_t *scheduler = NULL; pcmk_resource_t *parent = uber_parent(rsc); bool running = false; const char *id = rsc->clone_name ? rsc->clone_name : rsc->id; const char *host = node ? node->details->uname : NULL; /* If the implicit resource or primitive resource of a bundle is given, operate on the * bundle itself instead. */ if (pe_rsc_is_bundled(rsc)) { rsc = parent->parent; } running = resource_is_running_on(rsc, host); if (pe_rsc_is_clone(parent) && !running) { if (pe_rsc_is_unique_clone(parent)) { lookup_id = strdup(rsc->id); } else { lookup_id = clone_strip(rsc->id); } rsc = parent->fns->find_rsc(parent, lookup_id, node, pcmk_rsc_match_basename |pcmk_rsc_match_current_node); free(lookup_id); running = resource_is_running_on(rsc, host); } if (!running) { if (host) { out->err(out, "%s is not running on %s and so cannot be restarted", id, host); } else { out->err(out, "%s is not running anywhere and so cannot be restarted", id); } return ENXIO; } if (!pcmk_is_set(rsc->flags, pcmk_rsc_managed)) { out->err(out, "Unmanaged resources cannot be restarted."); return EAGAIN; } rsc_id = strdup(rsc->id); if (pe_rsc_is_unique_clone(parent)) { lookup_id = strdup(rsc->id); } else { lookup_id = clone_strip(rsc->id); } if (host) { if (pe_rsc_is_clone(rsc) || pe_bundle_replicas(rsc)) { stop_via_ban = true; } else if (pe_rsc_is_clone(parent)) { stop_via_ban = true; free(lookup_id); lookup_id = strdup(parent->id); } } /* grab full cib determine originally active resources disable or ban poll cib and watch for affected resources to get stopped without --timeout, calculate the stop timeout for each step and wait for that if we hit --timeout or the service timeout, re-enable or un-ban, report failure and indicate which resources we couldn't take down if everything stopped, re-enable or un-ban poll cib and watch for affected resources to get started without --timeout, calculate the start timeout for each step and wait for that if we hit --timeout or the service timeout, report (different) failure and indicate which resources we couldn't bring back up report success Optimizations: - use constraints to determine ordered list of affected resources - Allow a --no-deps option (aka. --force-restart) */ scheduler = pe_new_working_set(); if (scheduler == NULL) { rc = errno; out->err(out, "Could not allocate scheduler data: %s", pcmk_rc_str(rc)); goto done; } scheduler->priv = out; rc = update_dataset(cib, scheduler, false); if(rc != pcmk_rc_ok) { out->err(out, "Could not get new resource list: %s (%d)", pcmk_rc_str(rc), rc); goto done; } restart_target_active = get_active_resources(host, scheduler->resources); current_active = get_active_resources(host, scheduler->resources); dump_list(current_active, "Origin"); if (stop_via_ban) { /* Stop the clone or bundle instance by banning it from the host */ out->quiet = true; rc = cli_resource_ban(out, lookup_id, host, move_lifetime, cib, cib_options, promoted_role_only, PCMK__ROLE_PROMOTED); } else { /* Stop the resource by setting PCMK_META_TARGET_ROLE to Stopped. * Remember any existing PCMK_META_TARGET_ROLE so we can restore it * later (though it only makes any difference if it's Unpromoted). */ find_resource_attr(out, cib, PCMK_XA_VALUE, lookup_id, NULL, NULL, NULL, PCMK_META_TARGET_ROLE, &orig_target_role); rc = cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, PCMK_META_TARGET_ROLE, PCMK_ACTION_STOPPED, FALSE, cib, cib_options, force); } if(rc != pcmk_rc_ok) { out->err(out, "Could not set " PCMK_META_TARGET_ROLE " for %s: %s (%d)", rsc_id, pcmk_rc_str(rc), rc); if (current_active != NULL) { g_list_free_full(current_active, free); current_active = NULL; } if (restart_target_active != NULL) { g_list_free_full(restart_target_active, free); restart_target_active = NULL; } goto done; } rc = update_dataset(cib, scheduler, true); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources would be stopped"); goto failure; } target_active = get_active_resources(host, scheduler->resources); dump_list(target_active, "Target"); list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp); out->info(out, "Waiting for %d resources to stop:", g_list_length(list_delta)); display_list(out, list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (list_delta != NULL) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = wait_time_estimate(scheduler, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for(lpc = 0; (lpc < step_timeout_s) && (list_delta != NULL); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, scheduler, FALSE); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources were stopped"); goto failure; } if (current_active != NULL) { g_list_free_full(current_active, free); } current_active = get_active_resources(host, scheduler->resources); g_list_free(list_delta); list_delta = pcmk__subtract_lists(current_active, target_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } crm_trace("%d (was %d) resources remaining", g_list_length(list_delta), before); if(before == g_list_length(list_delta)) { /* aborted during stop phase, print the contents of list_delta */ out->err(out, "Could not complete shutdown of %s, %d resources remaining", rsc_id, g_list_length(list_delta)); display_list(out, list_delta, " * "); rc = ETIME; goto failure; } } if (stop_via_ban) { rc = cli_resource_clear(lookup_id, host, NULL, cib, cib_options, true, force); } else if (orig_target_role) { rc = cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, PCMK_META_TARGET_ROLE, orig_target_role, FALSE, cib, cib_options, force); free(orig_target_role); orig_target_role = NULL; } else { rc = cli_resource_delete_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, PCMK_META_TARGET_ROLE, cib, cib_options, force); } if(rc != pcmk_rc_ok) { out->err(out, "Could not unset " PCMK_META_TARGET_ROLE " for %s: %s (%d)", rsc_id, pcmk_rc_str(rc), rc); goto done; } if (target_active != NULL) { g_list_free_full(target_active, free); } target_active = restart_target_active; list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp); out->info(out, "Waiting for %d resources to start again:", g_list_length(list_delta)); display_list(out, list_delta, " * "); step_timeout_s = timeout / sleep_interval; while (waiting_for_starts(list_delta, rsc, host)) { before = g_list_length(list_delta); if(timeout_ms == 0) { step_timeout_s = wait_time_estimate(scheduler, list_delta) / sleep_interval; } /* We probably don't need the entire step timeout */ for (lpc = 0; (lpc < step_timeout_s) && waiting_for_starts(list_delta, rsc, host); lpc++) { sleep(sleep_interval); if(timeout) { timeout -= sleep_interval; crm_trace("%ds remaining", timeout); } rc = update_dataset(cib, scheduler, false); if(rc != pcmk_rc_ok) { out->err(out, "Could not determine which resources were started"); goto failure; } /* It's OK if dependent resources moved to a different node, * so we check active resources on all nodes. */ if (current_active != NULL) { g_list_free_full(current_active, free); } current_active = get_active_resources(NULL, scheduler->resources); g_list_free(list_delta); list_delta = pcmk__subtract_lists(target_active, current_active, (GCompareFunc) strcmp); dump_list(current_active, "Current"); dump_list(list_delta, "Delta"); } if(before == g_list_length(list_delta)) { /* aborted during start phase, print the contents of list_delta */ out->err(out, "Could not complete restart of %s, %d resources remaining", rsc_id, g_list_length(list_delta)); display_list(out, list_delta, " * "); rc = ETIME; goto failure; } } rc = pcmk_rc_ok; goto done; failure: if (stop_via_ban) { cli_resource_clear(lookup_id, host, NULL, cib, cib_options, true, force); } else if (orig_target_role) { cli_resource_update_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, PCMK_META_TARGET_ROLE, orig_target_role, FALSE, cib, cib_options, force); free(orig_target_role); } else { cli_resource_delete_attribute(rsc, rsc_id, NULL, XML_TAG_META_SETS, NULL, PCMK_META_TARGET_ROLE, cib, cib_options, force); } done: if (list_delta != NULL) { g_list_free(list_delta); } if (current_active != NULL) { g_list_free_full(current_active, free); } if (target_active != NULL && (target_active != restart_target_active)) { g_list_free_full(target_active, free); } if (restart_target_active != NULL) { g_list_free_full(restart_target_active, free); } free(rsc_id); free(lookup_id); pe_free_working_set(scheduler); return rc; } static inline bool action_is_pending(const pcmk_action_t *action) { if (pcmk_any_flags_set(action->flags, pcmk_action_optional|pcmk_action_pseudo) || !pcmk_is_set(action->flags, pcmk_action_runnable) || pcmk__str_eq(PCMK_ACTION_NOTIFY, action->task, pcmk__str_casei)) { return false; } return true; } /*! * \internal * \brief Check whether any actions in a list are pending * * \param[in] actions List of actions to check * * \return true if any actions in the list are pending, otherwise false */ static bool actions_are_pending(const GList *actions) { for (const GList *action = actions; action != NULL; action = action->next) { const pcmk_action_t *a = (const pcmk_action_t *) action->data; if (action_is_pending(a)) { crm_notice("Waiting for %s (flags=%#.8x)", a->uuid, a->flags); return true; } } return false; } static void print_pending_actions(pcmk__output_t *out, GList *actions) { GList *action; out->info(out, "Pending actions:"); for (action = actions; action != NULL; action = action->next) { pcmk_action_t *a = (pcmk_action_t *) action->data; if (!action_is_pending(a)) { continue; } if (a->node) { out->info(out, "\tAction %d: %s\ton %s", a->id, a->uuid, pe__node_name(a->node)); } else { out->info(out, "\tAction %d: %s", a->id, a->uuid); } } } /* For --wait, timeout (in seconds) to use if caller doesn't specify one */ #define WAIT_DEFAULT_TIMEOUT_S (60 * 60) /* For --wait, how long to sleep between cluster state checks */ #define WAIT_SLEEP_S (2) /*! * \internal * \brief Wait until all pending cluster actions are complete * * This waits until either the CIB's transition graph is idle or a timeout is * reached. * * \param[in,out] out Output object * \param[in] timeout_ms Consider failed if actions do not complete in * this time (specified in milliseconds, but * one-second granularity is actually used; if 0, a * default will be used) * \param[in,out] cib Connection to the CIB manager * * \return Standard Pacemaker return code */ int wait_till_stable(pcmk__output_t *out, int timeout_ms, cib_t * cib) { pcmk_scheduler_t *scheduler = NULL; xmlXPathObjectPtr search; int rc = pcmk_rc_ok; bool pending_unknown_state_resources; int timeout_s = timeout_ms? ((timeout_ms + 999) / 1000) : WAIT_DEFAULT_TIMEOUT_S; time_t expire_time = time(NULL) + timeout_s; time_t time_diff; bool printed_version_warning = out->is_quiet(out); // i.e. don't print if quiet scheduler = pe_new_working_set(); if (scheduler == NULL) { return ENOMEM; } do { /* Abort if timeout is reached */ time_diff = expire_time - time(NULL); if (time_diff > 0) { crm_info("Waiting up to %lld seconds for cluster actions to complete", (long long) time_diff); } else { print_pending_actions(out, scheduler->actions); pe_free_working_set(scheduler); return ETIME; } if (rc == pcmk_rc_ok) { /* this avoids sleep on first loop iteration */ sleep(WAIT_SLEEP_S); } /* Get latest transition graph */ pe_reset_working_set(scheduler); rc = update_scheduler_input_to_cib(out, scheduler, cib); if (rc != pcmk_rc_ok) { pe_free_working_set(scheduler); return rc; } pcmk__schedule_actions(scheduler->input, pcmk_sched_no_counts|pcmk_sched_no_compat, scheduler); if (!printed_version_warning) { /* If the DC has a different version than the local node, the two * could come to different conclusions about what actions need to be * done. Warn the user in this case. * * @TODO A possible long-term solution would be to reimplement the * wait as a new controller operation that would be forwarded to the * DC. However, that would have potential problems of its own. */ const char *dc_version = g_hash_table_lookup(scheduler->config_hash, PCMK_OPT_DC_VERSION); if (!pcmk__str_eq(dc_version, PACEMAKER_VERSION "-" BUILD_VERSION, pcmk__str_casei)) { out->info(out, "warning: wait option may not work properly in " "mixed-version cluster"); printed_version_warning = true; } } search = xpath_search(scheduler->input, "/cib/status/node_state/lrm/lrm_resources/lrm_resource/" XML_LRM_TAG_RSC_OP "[@" XML_LRM_ATTR_RC "='193']"); pending_unknown_state_resources = (numXpathResults(search) > 0); freeXpathObject(search); } while (actions_are_pending(scheduler->actions) || pending_unknown_state_resources); pe_free_working_set(scheduler); return rc; } static const char * get_action(const char *rsc_action) { const char *action = NULL; if (pcmk__str_eq(rsc_action, "validate", pcmk__str_casei)) { action = PCMK_ACTION_VALIDATE_ALL; } else if (pcmk__str_eq(rsc_action, "force-check", pcmk__str_casei)) { action = PCMK_ACTION_MONITOR; } else if (pcmk__strcase_any_of(rsc_action, "force-start", "force-stop", "force-demote", "force-promote", NULL)) { action = rsc_action+6; } else { action = rsc_action; } return action; } /*! * \brief Set up environment variables as expected by resource agents * * When the cluster executes resource agents, it adds certain environment * variables (directly or via resource meta-attributes) expected by some * resource agents. Add the essential ones that many resource agents expect, so * the behavior is the same for command-line execution. * * \param[in,out] params Resource parameters that will be passed to agent * \param[in] timeout_ms Action timeout (in milliseconds) * \param[in] check_level OCF check level * \param[in] verbosity Verbosity level */ static void set_agent_environment(GHashTable *params, int timeout_ms, int check_level, int verbosity) { g_hash_table_insert(params, strdup("CRM_meta_timeout"), crm_strdup_printf("%d", timeout_ms)); g_hash_table_insert(params, strdup(PCMK_XA_CRM_FEATURE_SET), strdup(CRM_FEATURE_SET)); if (check_level >= 0) { char *level = crm_strdup_printf("%d", check_level); setenv("OCF_CHECK_LEVEL", level, 1); free(level); } pcmk__set_env_option(PCMK__ENV_DEBUG, ((verbosity > 0)? "1" : "0"), true); if (verbosity > 1) { setenv("OCF_TRACE_RA", "1", 1); } /* A resource agent using the standard ocf-shellfuncs library will not print * messages to stderr if it doesn't have a controlling terminal (e.g. if * crm_resource is called via script or ssh). This forces it to do so. */ setenv("OCF_TRACE_FILE", "/dev/stderr", 0); } /*! * \internal * \brief Apply command-line overrides to resource parameters * * \param[in,out] params Parameters to be passed to agent * \param[in] overrides Parameters to override (or NULL if none) */ static void apply_overrides(GHashTable *params, GHashTable *overrides) { if (overrides != NULL) { GHashTableIter iter; char *name = NULL; char *value = NULL; g_hash_table_iter_init(&iter, overrides); while (g_hash_table_iter_next(&iter, (gpointer *) &name, (gpointer *) &value)) { g_hash_table_replace(params, strdup(name), strdup(value)); } } } crm_exit_t cli_resource_execute_from_params(pcmk__output_t *out, const char *rsc_name, const char *rsc_class, const char *rsc_prov, const char *rsc_type, const char *rsc_action, GHashTable *params, GHashTable *override_hash, int timeout_ms, int resource_verbose, gboolean force, int check_level) { const char *class = rsc_class; const char *action = get_action(rsc_action); crm_exit_t exit_code = CRM_EX_OK; svc_action_t *op = NULL; // If no timeout was provided, use the same default as the cluster if (timeout_ms == 0) { timeout_ms = PCMK_DEFAULT_ACTION_TIMEOUT_MS; } set_agent_environment(params, timeout_ms, check_level, resource_verbose); apply_overrides(params, override_hash); op = services__create_resource_action(rsc_name? rsc_name : "test", rsc_class, rsc_prov, rsc_type, action, 0, timeout_ms, params, 0); if (op == NULL) { out->err(out, "Could not execute %s using %s%s%s:%s: %s", action, rsc_class, (rsc_prov? ":" : ""), (rsc_prov? rsc_prov : ""), rsc_type, strerror(ENOMEM)); g_hash_table_destroy(params); return CRM_EX_OSERR; } if (pcmk__str_eq(rsc_class, PCMK_RESOURCE_CLASS_SERVICE, pcmk__str_casei)) { class = resources_find_service_class(rsc_type); } if (!pcmk__strcase_any_of(class, PCMK_RESOURCE_CLASS_OCF, PCMK_RESOURCE_CLASS_LSB, NULL)) { services__format_result(op, CRM_EX_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR, "Manual execution of the %s standard is " "unsupported", pcmk__s(class, "unspecified")); } if (op->rc != PCMK_OCF_UNKNOWN) { exit_code = op->rc; goto done; } services_action_sync(op); // Map results to OCF codes for consistent reporting to user { enum ocf_exitcode ocf_code = services_result2ocf(class, action, op->rc); // Cast variable instead of function return to keep compilers happy exit_code = (crm_exit_t) ocf_code; } done: out->message(out, "resource-agent-action", resource_verbose, rsc_class, rsc_prov, rsc_type, rsc_name, rsc_action, override_hash, exit_code, op->status, services__exit_reason(op), op->stdout_data, op->stderr_data); services_action_free(op); return exit_code; } crm_exit_t cli_resource_execute(pcmk_resource_t *rsc, const char *requested_name, const char *rsc_action, GHashTable *override_hash, int timeout_ms, cib_t *cib, pcmk_scheduler_t *scheduler, int resource_verbose, gboolean force, int check_level) { pcmk__output_t *out = scheduler->priv; crm_exit_t exit_code = CRM_EX_OK; const char *rid = NULL; const char *rtype = NULL; const char *rprov = NULL; const char *rclass = NULL; GHashTable *params = NULL; if (pcmk__strcase_any_of(rsc_action, "force-start", "force-demote", "force-promote", NULL)) { if(pe_rsc_is_clone(rsc)) { GList *nodes = cli_resource_search(rsc, requested_name, scheduler); if(nodes != NULL && force == FALSE) { out->err(out, "It is not safe to %s %s here: the cluster claims it is already active", rsc_action, rsc->id); out->err(out, "Try setting " PCMK_META_TARGET_ROLE "=" PCMK__ROLE_STOPPED " first or specifying the force option"); return CRM_EX_UNSAFE; } g_list_free_full(nodes, free); } } if(pe_rsc_is_clone(rsc)) { /* Grab the first child resource in the hope it's not a group */ rsc = rsc->children->data; } if (rsc->variant == pcmk_rsc_variant_group) { out->err(out, "Sorry, the %s option doesn't support group resources", rsc_action); return CRM_EX_UNIMPLEMENT_FEATURE; } else if (pe_rsc_is_bundled(rsc)) { out->err(out, "Sorry, the %s option doesn't support bundled resources", rsc_action); return CRM_EX_UNIMPLEMENT_FEATURE; } rclass = crm_element_value(rsc->xml, PCMK_XA_CLASS); rprov = crm_element_value(rsc->xml, PCMK_XA_PROVIDER); rtype = crm_element_value(rsc->xml, PCMK_XA_TYPE); params = generate_resource_params(rsc, NULL /* @TODO use local node */, scheduler); if (timeout_ms == 0) { timeout_ms = pe_get_configured_timeout(rsc, get_action(rsc_action), scheduler); } rid = pe_rsc_is_anon_clone(rsc->parent)? requested_name : rsc->id; exit_code = cli_resource_execute_from_params(out, rid, rclass, rprov, rtype, rsc_action, params, override_hash, timeout_ms, resource_verbose, force, check_level); return exit_code; } // \return Standard Pacemaker return code int cli_resource_move(const pcmk_resource_t *rsc, const char *rsc_id, const char *host_name, const char *move_lifetime, cib_t *cib, int cib_options, pcmk_scheduler_t *scheduler, gboolean promoted_role_only, gboolean force) { pcmk__output_t *out = scheduler->priv; int rc = pcmk_rc_ok; unsigned int count = 0; pcmk_node_t *current = NULL; pcmk_node_t *dest = pe_find_node(scheduler->nodes, host_name); bool cur_is_dest = false; if (dest == NULL) { return pcmk_rc_node_unknown; } if (promoted_role_only && !pcmk_is_set(rsc->flags, pcmk_rsc_promotable)) { const pcmk_resource_t *p = pe__const_top_resource(rsc, false); if (pcmk_is_set(p->flags, pcmk_rsc_promotable)) { out->info(out, "Using parent '%s' for move instead of '%s'.", rsc->id, rsc_id); rsc_id = p->id; rsc = p; } else { out->info(out, "Ignoring --promoted option: %s is not promotable", rsc_id); promoted_role_only = FALSE; } } current = pe__find_active_requires(rsc, &count); if (pcmk_is_set(rsc->flags, pcmk_rsc_promotable)) { unsigned int promoted_count = 0; pcmk_node_t *promoted_node = NULL; for (const GList *iter = rsc->children; iter; iter = iter->next) { const pcmk_resource_t *child = (const pcmk_resource_t *) iter->data; enum rsc_role_e child_role = child->fns->state(child, TRUE); if (child_role == pcmk_role_promoted) { rsc = child; promoted_node = pe__current_node(child); promoted_count++; } } if (promoted_role_only || (promoted_count != 0)) { count = promoted_count; current = promoted_node; } } if (count > 1) { if (pe_rsc_is_clone(rsc)) { current = NULL; } else { return pcmk_rc_multiple; } } if (current && (current->details == dest->details)) { cur_is_dest = true; if (force) { crm_info("%s is already %s on %s, reinforcing placement with location constraint.", rsc_id, promoted_role_only?"promoted":"active", pe__node_name(dest)); } else { return pcmk_rc_already; } } /* Clear any previous prefer constraints across all nodes. */ cli_resource_clear(rsc_id, NULL, scheduler->nodes, cib, cib_options, false, force); /* Clear any previous ban constraints on 'dest'. */ cli_resource_clear(rsc_id, dest->details->uname, scheduler->nodes, cib, cib_options, TRUE, force); /* Record an explicit preference for 'dest' */ rc = cli_resource_prefer(out, rsc_id, dest->details->uname, move_lifetime, cib, cib_options, promoted_role_only, PCMK__ROLE_PROMOTED); crm_trace("%s%s now prefers %s%s", rsc->id, (promoted_role_only? " (promoted)" : ""), pe__node_name(dest), force?"(forced)":""); /* only ban the previous location if current location != destination location. * it is possible to use -M to enforce a location without regard of where the * resource is currently located */ if (force && !cur_is_dest) { /* Ban the original location if possible */ if(current) { (void)cli_resource_ban(out, rsc_id, current->details->uname, move_lifetime, cib, cib_options, promoted_role_only, PCMK__ROLE_PROMOTED); } else if(count > 1) { out->info(out, "Resource '%s' is currently %s in %d locations. " "One may now move to %s", rsc_id, (promoted_role_only? "promoted" : "active"), count, pe__node_name(dest)); out->info(out, "To prevent '%s' from being %s at a specific location, " "specify a node.", rsc_id, (promoted_role_only? "promoted" : "active")); } else { crm_trace("Not banning %s from its current location: not active", rsc_id); } } return rc; }