diff --git a/daemons/controld/controld_cib.c b/daemons/controld/controld_cib.c index ee624c7bfa..122e5d8144 100644 --- a/daemons/controld/controld_cib.c +++ b/daemons/controld/controld_cib.c @@ -1,1055 +1,1060 @@ /* * 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 // 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 (pcmk__cib_element_in_patchset(patchset, PCMK_XE_ALERTS) || pcmk__cib_element_in_patchset(patchset, PCMK_XE_CRM_CONFIG)) { controld_trigger_config(); } if (!AM_I_DC) { // We're not in control of the join sequence return; } client_name = crm_element_value(msg, PCMK__XA_CIB_CLIENTNAME); if (!cib__client_triggers_refresh(client_name)) { // The CIB is still accurate return; } if (pcmk__cib_element_in_patchset(patchset, PCMK_XE_NODES) || pcmk__cib_element_in_patchset(patchset, PCMK_XE_STATUS)) { /* An unsafe client modified the PCMK_XE_NODES or PCMK_XE_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, PCMK__XA_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; pcmk__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, PCMK__VALUE_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_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; pcmk__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_name, 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_name, 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, PCMK__VALUE_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) { unsigned int calculated_timeout = 10U * (pcmk__cluster_num_active_nodes() + pcmk__cluster_num_remote_nodes() + 1U); calculated_timeout = QB_MAX(calculated_timeout, MIN_CIB_OP_TIMEOUT); crm_trace("Calculated timeout: %s", pcmk__readable_interval(calculated_timeout * 1000)); 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_none); } 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 " QB_XS " rc=%d", desc, call_id, pcmk_strerror(rc), rc); } } // Searches for various portions of PCMK__XE_NODE_STATE to delete // Match a particular node's PCMK__XE_NODE_STATE (takes node name 1x) #define XPATH_NODE_STATE "//" PCMK__XE_NODE_STATE "[@" PCMK_XA_UNAME "='%s']" // Node's lrm section (name 1x) #define XPATH_NODE_LRM XPATH_NODE_STATE "/" PCMK__XE_LRM /* Node's PCMK__XE_LRM_RSC_OP entries and PCMK__XE_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 "//" PCMK__XE_LRM_RSC_OP \ "|" XPATH_NODE_STATE \ "//" PCMK__XE_LRM_RESOURCE \ "[not(@" PCMK_OPT_SHUTDOWN_LOCK ") " \ "or " PCMK_OPT_SHUTDOWN_LOCK "<%lld]" // Node's PCMK__XE_TRANSIENT_ATTRIBUTES section (name 1x) #define XPATH_NODE_ATTRS XPATH_NODE_STATE "/" PCMK__XE_TRANSIENT_ATTRIBUTES // Everything under PCMK__XE_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 \c PCMK__XE_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 pcmk__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 \c PCMK__XE_NODE_STATE * * \param[in] uname Desired node * \param[in] section Subsection of \c PCMK__XE_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; pcmk__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) " QB_XS " xpath=%s", desc, cib_rc, xpath); // CIB library handles freeing desc free(xpath); } // Takes node name and resource ID #define XPATH_RESOURCE_HISTORY "//" PCMK__XE_NODE_STATE \ "[@" PCMK_XA_UNAME "='%s']/" \ PCMK__XE_LRM "/" PCMK__XE_LRM_RESOURCES \ "/" PCMK__XE_LRM_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 " QB_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) " QB_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 pcmk__xml_free(). */ 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 = pcmk__xe_create(NULL, PCMK_XE_PARAMETERS); /* 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); pcmk__xe_remove_attr(*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 PCMK__XA_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 * PCMK__XA_OP_FORCE_RESTART list. */ list = build_parameter_list(op, metadata, ra_param_unique, &restart); } else { // Resource does not support agent reloads return; } digest = pcmk__digest_operation(restart); /* Add PCMK__XA_OP_FORCE_RESTART and PCMK__XA_OP_RESTART_DIGEST to indicate * the resource supports reload, no matter if it actually supports any * reloadable parameters */ crm_xml_add(update, PCMK__XA_OP_FORCE_RESTART, (list == NULL)? "" : (const char *) list->str); crm_xml_add(update, PCMK__XA_OP_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); } pcmk__xml_free(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 PCMK__XA_OP_SECURE_PARAMS short, we want it to contain the secure * parameters but PCMK__XA_OP_SECURE_DIGEST to be based on the insecure ones */ list = build_parameter_list(op, metadata, ra_param_private, &secure); if (list != NULL) { digest = pcmk__digest_operation(secure); crm_xml_add(update, PCMK__XA_OP_SECURE_PARAMS, (const char *) list->str); crm_xml_add(update, PCMK__XA_OP_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); } pcmk__xml_free(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.cluster->priv->node_name, 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 = controld_get_executor_state(node_name, false); 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 PCMK_META_RECORD_PENDING meta-attribute (defaults to true) record_pending = crm_meta_value(op->params, PCMK_META_RECORD_PENDING); if ((record_pending != NULL) && !crm_is_true(record_pending)) { + pcmk__warn_once(pcmk__wo_record_pending, + "The " PCMK_META_RECORD_PENDING " option (for example, " + "for the %s resource's %s operation) is deprecated and " + "will be removed in a future release", + rsc->id, op->op_type); 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 " QB_XS " rc=%d", call_id, pcmk_strerror(rc), rc); } else { crm_warn("Resource history update failed: %s " QB_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; pcmk__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 = pcmk__xe_create(NULL, PCMK_XE_STATUS); // xml = pcmk__xe_create(update, PCMK__XE_NODE_STATE); if (controld_is_local_node(node_name)) { node_id = controld_globals.our_uuid; } else { node_id = node_name; pcmk__xe_set_bool_attr(xml, PCMK_XA_REMOTE_NODE, true); } crm_xml_add(xml, PCMK_XA_ID, node_id); crm_xml_add(xml, PCMK_XA_UNAME, node_name); crm_xml_add(xml, PCMK_XA_CRM_DEBUG_ORIGIN, __func__); // xml = pcmk__xe_create(xml, PCMK__XE_LRM); crm_xml_add(xml, PCMK_XA_ID, node_id); // xml = pcmk__xe_create(xml, PCMK__XE_LRM_RESOURCES); // xml = pcmk__xe_create(xml, PCMK__XE_LRM_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(PCMK_XE_STATUS, update, call_opt, cib_rsc_callback); pcmk__xml_free(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 = pcmk__xe_create(NULL, PCMK__XE_LRM_RSC_OP); crm_xml_add_int(xml_top, PCMK__XA_CALL_ID, op->call_id); crm_xml_add(xml_top, PCMK__XA_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, PCMK_XE_STATUS, xml_top, cib_none); crm_log_xml_trace(xml_top, "op:cancel"); pcmk__xml_free(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" PCMK_XE_CIB "/" PCMK_XE_STATUS \ "/" PCMK__XE_NODE_STATE "[@" PCMK_XA_UNAME "='%s']" \ "/" PCMK__XE_LRM "/" PCMK__XE_LRM_RESOURCES \ "/" PCMK__XE_LRM_RESOURCE "[@" PCMK_XA_ID "='%s']" \ "/" PCMK__XE_LRM_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 @" PCMK__XA_CALL_ID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" PCMK_XA_ID "='%s' and @" PCMK__XA_OPERATION_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/doc/sphinx/Pacemaker_Explained/operations.rst b/doc/sphinx/Pacemaker_Explained/operations.rst index b390f97ebd..b8a324b8ab 100644 --- a/doc/sphinx/Pacemaker_Explained/operations.rst +++ b/doc/sphinx/Pacemaker_Explained/operations.rst @@ -1,688 +1,688 @@ .. index:: single: resource; action single: resource; operation .. _operation: Resource Operations ------------------- *Operations* are actions the cluster can perform on a resource by calling the resource agent. Resource agents must support certain common operations such as start, stop, and monitor, and may implement any others. Operations may be explicitly configured for two purposes: to override defaults for options (such as timeout) that the cluster will use whenever it initiates the operation, and to run an operation on a recurring basis (for example, to monitor the resource for failure). .. topic:: An OCF resource with a non-default start timeout .. code-block:: xml Pacemaker identifies operations by a combination of name and interval, so this combination must be unique for each resource. That is, you should not configure two operations for the same resource with the same name and interval. .. _operation_properties: Operation Properties #################### The ``id``, ``name``, ``interval``, and ``role`` operation properties may be specified only as XML attributes of the ``op`` element. Other operation properties may be specified in any of the following ways, from highest precedence to lowest: * directly in the ``op`` element as an XML attribute * in an ``nvpair`` element within a ``meta_attributes`` element within the ``op`` element * in an ``nvpair`` element within a ``meta_attributes`` element within :ref:`operation defaults ` If not specified, the default from the table below is used. .. list-table:: **Operation Properties** :class: longtable :widths: 2 2 3 4 :header-rows: 1 * - Name - Type - Default - Description * - .. _op_id: .. index:: pair: op; id single: id; action property single: action; property, id id - :ref:`id ` - - A unique identifier for the XML element *(required)* * - .. _op_name: .. index:: pair: op; name single: name; action property single: action; property, name name - :ref:`text ` - - An action name supported by the resource agent *(required)* * - .. _op_interval: .. index:: pair: op; interval single: interval; action property single: action; property, interval interval - :ref:`duration ` - 0 - If this is a positive value, Pacemaker will schedule recurring instances of this operation at the given interval (which makes sense only with :ref:`name ` set to :ref:`monitor `). If this is 0, Pacemaker will apply other properties configured for this operation to instances that are scheduled as needed during normal cluster operation. *(required)* * - .. _op_role: .. index:: pair: op; role single: role; action property single: action; property, role role - :ref:`enumeration ` - - If this is set, the operation configuration applies only on nodes where the cluster expects the resource to be in the specified role. This makes sense only for recurring monitors. Allowed values: ``Started``, ``Stopped``, and in the case of :ref:`promotable clone resources `, ``Unpromoted`` and ``Promoted``. * - .. _op_timeout: .. index:: pair: op; timeout single: timeout; action property single: action; property, timeout timeout - :ref:`timeout ` - 20s - If resource agent execution does not complete within this amount of time, the action will be considered failed. **Note:** timeouts for fencing agents are handled specially (see the :ref:`fencing` chapter). * - .. _op_on_fail: .. index:: pair: op; on-fail single: on-fail; action property single: action; property, on-fail on-fail - :ref:`enumeration ` - * If ``name`` is ``stop``: ``fence`` if :ref:`stonith-enabled ` is true, otherwise ``block`` * If ``name`` is ``demote``: ``on-fail`` of the ``monitor`` action with ``role`` set to ``Promoted``, if present, enabled, and configured to a value other than ``demote``, or ``restart`` otherwise * Otherwise: ``restart`` - How the cluster should respond to a failure of this action. Allowed values: * ``ignore:`` Pretend the resource did not fail * ``block:`` Do not perform any further operations on the resource * ``stop:`` Stop the resource and leave it stopped * ``demote:`` Demote the resource, without a full restart. This is valid only for ``promote`` actions, and for ``monitor`` actions with both a nonzero ``interval`` and ``role`` set to ``Promoted``; for any other action, a configuration error will be logged, and the default behavior will be used. *(since 2.0.5)* * ``restart:`` Stop the resource, and start it again if allowed (possibly on a different node) * ``fence:`` Fence the node on which the resource failed * ``standby:`` Put the node on which the resource failed in standby mode (forcing *all* resources away) * - .. _op_enabled: .. index:: pair: op; enabled single: enabled; action property single: action; property, enabled enabled - :ref:`boolean ` - true - If ``false``, ignore this operation definition. This does not suppress all actions of this type, but is typically used to pause a recurring monitor. This can complement the resource being unmanaged (:ref:`is-managed ` set to ``false``), which does not stop recurring operations. Maintenance mode, which does stop configured monitors, overrides this setting. * - .. _op_record_pending: .. index:: pair: op; record-pending single: record-pending; action property single: action; property, record-pending record-pending - :ref:`boolean ` - true - Operation results are always recorded when the operation completes (successful or not). If this is ``true``, operations will also be recorded when initiated, so that status output can indicate that the - operation is in progress. + operation is in progress. *(deprecated since 3.0.0)* .. note:: Only one action can be configured for any given combination of ``name`` and ``interval``. .. note:: When ``on-fail`` is set to ``demote``, recovery from failure by a successful demote causes the cluster to recalculate whether and where a new instance should be promoted. The node with the failure is eligible, so if promotion scores have not changed, it will be promoted again. There is no direct equivalent of ``migration-threshold`` for the promoted role, but the same effect can be achieved with a location constraint using a :ref:`rule ` with a node attribute expression for the resource's fail count. For example, to immediately ban the promoted role from a node with any failed promote or promoted instance monitor: .. code-block:: xml This example assumes that there is a promotable clone of the ``my_primitive`` resource (note that the primitive name, not the clone name, is used in the rule), and that there is a recurring 10-second-interval monitor configured for the promoted role (fail count attributes specify the interval in milliseconds). .. _s-resource-monitoring: Monitoring Resources for Failure ################################ When Pacemaker first starts a resource, it runs one-time ``monitor`` operations (referred to as *probes*) to ensure the resource is running where it's supposed to be, and not running where it's not supposed to be. (This behavior can be affected by the ``resource-discovery`` location constraint property.) Other than those initial probes, Pacemaker will *not* (by default) check that the resource continues to stay healthy [#]_. You must configure ``monitor`` operations explicitly to perform these checks. .. topic:: An OCF resource with a recurring health check .. code-block:: xml By default, a ``monitor`` operation will ensure that the resource is running where it is supposed to. The ``target-role`` property can be used for further checking. For example, if a resource has one ``monitor`` operation with ``interval=10 role=Started`` and a second ``monitor`` operation with ``interval=11 role=Stopped``, the cluster will run the first monitor on any nodes it thinks *should* be running the resource, and the second monitor on any nodes that it thinks *should not* be running the resource (for the truly paranoid, who want to know when an administrator manually starts a service by mistake). .. note:: Currently, monitors with ``role=Stopped`` are not implemented for :ref:`clone ` resources. Custom Recurring Operations ########################### Typically, only ``monitor`` operations should be configured as recurring. However, it is possible to implement a custom action name in an OCF agent and then configure that as a recurring operation. This could be useful, for example, to run a report, rotate a log, or clean temporary files related to a particular service. Failures of custom recurring operations will be ignored by the cluster and will not be reported in cluster status *(since 3.0.0; previously, they would be treated like failed monitors)*. A fail count and last failure timestamp will be recorded as transient node attributes, and those node attributes will be erased by the ``crm_resource --cleanup`` command. .. _s-operation-defaults: Setting Global Defaults for Operations ###################################### You can change the global default values for operation properties in a given cluster. These are defined in an ``op_defaults`` section of the CIB's ``configuration`` section, and can be set with ``crm_attribute``. For example, .. code-block:: none # crm_attribute --type op_defaults --name timeout --update 20s would default each operation's ``timeout`` to 20 seconds. If an operation's definition also includes a value for ``timeout``, then that value would be used for that operation instead. When Implicit Operations Take a Long Time ######################################### The cluster will always perform a number of implicit operations: ``start``, ``stop`` and a non-recurring ``monitor`` operation used at startup to check whether the resource is already active. If one of these is taking too long, then you can create an entry for them and specify a longer timeout. .. topic:: An OCF resource with custom timeouts for its implicit actions .. code-block:: xml Multiple Monitor Operations ########################### Provided no two operations (for a single resource) have the same name and interval, you can have as many ``monitor`` operations as you like. In this way, you can do a superficial health check every minute and progressively more intense ones at higher intervals. To tell the resource agent what kind of check to perform, you need to provide each monitor with a different value for a common parameter. The OCF standard creates a special parameter called ``OCF_CHECK_LEVEL`` for this purpose and dictates that it is "made available to the resource agent without the normal ``OCF_RESKEY`` prefix". Whatever name you choose, you can specify it by adding an ``instance_attributes`` block to the ``op`` tag. It is up to each resource agent to look for the parameter and decide how to use it. .. topic:: An OCF resource with two recurring health checks, performing different levels of checks specified via ``OCF_CHECK_LEVEL``. .. code-block:: xml Disabling a Monitor Operation ############################# The easiest way to stop a recurring monitor is to just delete it. However, there can be times when you only want to disable it temporarily. In such cases, simply add ``enabled=false`` to the operation's definition. .. topic:: Example of an OCF resource with a disabled health check .. code-block:: xml This can be achieved from the command line by executing: .. code-block:: none # cibadmin --modify --xml-text '' Once you've done whatever you needed to do, you can then re-enable it with .. code-block:: none # cibadmin --modify --xml-text '' .. index:: single: start-delay; operation attribute single: interval-origin; operation attribute single: interval; interval-origin single: operation; interval-origin single: operation; start-delay Specifying When Recurring Actions are Performed ############################################### By default, recurring actions are scheduled relative to when the resource started. In some cases, you might prefer that a recurring action start relative to a specific date and time. For example, you might schedule an in-depth monitor to run once every 24 hours, and want it to run outside business hours. To do this, set the operation's ``interval-origin``. The cluster uses this point to calculate the correct ``start-delay`` such that the operation will occur at ``interval-origin`` plus a multiple of the operation interval. For example, if the recurring operation's interval is 24h, its ``interval-origin`` is set to 02:00, and it is currently 14:32, then the cluster would initiate the operation after 11 hours and 28 minutes. The value specified for ``interval`` and ``interval-origin`` can be any date/time conforming to the `ISO8601 standard `_. By way of example, to specify an operation that would run on the first Monday of 2021 and every Monday after that, you would add: .. topic:: Example recurring action that runs relative to base date/time .. code-block:: xml .. index:: single: resource; failure recovery single: operation; failure recovery .. _failure-handling: Handling Resource Failure ######################### By default, Pacemaker will attempt to recover failed resources by restarting them. However, failure recovery is highly configurable. .. index:: single: resource; failure count single: operation; failure count Failure Counts ______________ Pacemaker tracks resource failures for each combination of node, resource, and operation (start, stop, monitor, etc.). You can query the fail count for a particular node, resource, and/or operation using the ``crm_failcount`` command. For example, to see how many times the 10-second monitor for ``myrsc`` has failed on ``node1``, run: .. code-block:: none # crm_failcount --query -r myrsc -N node1 -n monitor -I 10s If you omit the node, ``crm_failcount`` will use the local node. If you omit the operation and interval, ``crm_failcount`` will display the sum of the fail counts for all operations on the resource. You can use ``crm_resource --cleanup`` or ``crm_failcount --delete`` to clear fail counts. For example, to clear the above monitor failures, run: .. code-block:: none # crm_resource --cleanup -r myrsc -N node1 -n monitor -I 10s If you omit the resource, ``crm_resource --cleanup`` will clear failures for all resources. If you omit the node, it will clear failures on all nodes. If you omit the operation and interval, it will clear the failures for all operations on the resource. .. note:: Even when cleaning up only a single operation, all failed operations will disappear from the status display. This allows us to trigger a re-check of the resource's current status. Higher-level tools may provide other commands for querying and clearing fail counts. The ``crm_mon`` tool shows the current cluster status, including any failed operations. To see the current fail counts for any failed resources, call ``crm_mon`` with the ``--failcounts`` option. This shows the fail counts per resource (that is, the sum of any operation fail counts for the resource). .. index:: single: migration-threshold; resource meta-attribute single: resource; migration-threshold Failure Response ________________ Normally, if a running resource fails, pacemaker will try to stop it and start it again. Pacemaker will choose the best location to start it each time, which may be the same node that it failed on. However, if a resource fails repeatedly, it is possible that there is an underlying problem on that node, and you might desire trying a different node in such a case. Pacemaker allows you to set your preference via the ``migration-threshold`` resource meta-attribute. [#]_ If you define ``migration-threshold`` to *N* for a resource, it will be banned from the original node after *N* failures there. .. note:: The ``migration-threshold`` is per *resource*, even though fail counts are tracked per *operation*. The operation fail counts are added together to compare against the ``migration-threshold``. By default, fail counts remain until manually cleared by an administrator using ``crm_resource --cleanup`` or ``crm_failcount --delete`` (hopefully after first fixing the failure's cause). It is possible to have fail counts expire automatically by setting the ``failure-timeout`` resource meta-attribute. .. important:: A successful operation does not clear past failures. If a recurring monitor operation fails once, succeeds many times, then fails again days later, its fail count is 2. Fail counts are cleared only by manual intervention or failure timeout. For example, setting ``migration-threshold`` to 2 and ``failure-timeout`` to ``60s`` would cause the resource to move to a new node after 2 failures, and allow it to move back (depending on stickiness and constraint scores) after one minute. .. note:: ``failure-timeout`` is measured since the most recent failure. That is, older failures do not individually time out and lower the fail count. Instead, all failures are timed out simultaneously (and the fail count is reset to 0) if there is no new failure for the timeout period. There are two exceptions to the migration threshold: when a resource either fails to start or fails to stop. If the cluster property ``start-failure-is-fatal`` is set to ``true`` (which is the default), start failures cause the fail count to be set to ``INFINITY`` and thus always cause the resource to move immediately. Stop failures are slightly different and crucial. If a resource fails to stop and fencing is enabled, then the cluster will fence the node in order to be able to start the resource elsewhere. If fencing is disabled, then the cluster has no way to continue and will not try to start the resource elsewhere, but will try to stop it again after any failure timeout or clearing. .. index:: single: reload single: reload-agent Reloading an Agent After a Definition Change ############################################ The cluster automatically detects changes to the configuration of active resources. The cluster's normal response is to stop the service (using the old definition) and start it again (with the new definition). This works, but some resource agents are smarter and can be told to use a new set of options without restarting. To take advantage of this capability, the resource agent must: * Implement the ``reload-agent`` action. What it should do depends completely on your application! .. note:: Resource agents may also implement a ``reload`` action to make the managed service reload its own *native* configuration. This is different from ``reload-agent``, which makes effective changes in the resource's *Pacemaker* configuration (specifically, the values of the agent's reloadable parameters). * Advertise the ``reload-agent`` operation in the ``actions`` section of its meta-data. * Set the ``reloadable`` attribute to 1 in the ``parameters`` section of its meta-data for any parameters eligible to be reloaded after a change. Once these requirements are satisfied, the cluster will automatically know to reload the resource (instead of restarting) when a reloadable parameter changes. .. note:: Metadata will not be re-read unless the resource needs to be started. If you edit the agent of an already active resource to set a parameter reloadable, the resource may restart the first time the parameter value changes. .. note:: If both a reloadable and non-reloadable parameter are changed simultaneously, the resource will be restarted. .. _live-migration: Migrating Resources ################### Normally, when the cluster needs to move a resource, it fully restarts the resource (that is, it stops the resource on the current node and starts it on the new node). However, some types of resources, such as many virtual machines, are able to move to another location without loss of state (often referred to as live migration or hot migration). In pacemaker, this is called live migration. Pacemaker can be configured to migrate a resource when moving it, rather than restarting it. Not all resources are able to migrate; see the :ref:`migration checklist ` below. Even those that can, won't do so in all situations. Conceptually, there are two requirements from which the other prerequisites follow: * The resource must be active and healthy at the old location; and * everything required for the resource to run must be available on both the old and new locations. The cluster is able to accommodate both *push* and *pull* migration models by requiring the resource agent to support two special actions: ``migrate_to`` (performed on the current location) and ``migrate_from`` (performed on the destination). In push migration, the process on the current location transfers the resource to the new location where is it later activated. In this scenario, most of the work would be done in the ``migrate_to`` action and, if anything, the activation would occur during ``migrate_from``. Conversely for pull, the ``migrate_to`` action is practically empty and ``migrate_from`` does most of the work, extracting the relevant resource state from the old location and activating it. There is no wrong or right way for a resource agent to implement migration, as long as it works. .. _migration_checklist: .. topic:: Migration Checklist * The resource may not be a clone. * The resource agent standard must be OCF. * The resource must not be in a failed or degraded state. * The resource agent must support ``migrate_to`` and ``migrate_from`` actions, and advertise them in its meta-data. * The resource must have the ``allow-migrate`` meta-attribute set to ``true`` (which is not the default). If an otherwise migratable resource depends on another resource via an ordering constraint, there are special situations in which it will be restarted rather than migrated. For example, if the resource depends on a clone, and at the time the resource needs to be moved, the clone has instances that are stopping and instances that are starting, then the resource will be restarted. The scheduler is not yet able to model this situation correctly and so takes the safer (if less optimal) path. Also, if a migratable resource depends on a non-migratable resource, and both need to be moved, the migratable resource will be restarted. .. rubric:: Footnotes .. [#] Currently, anyway. Automatic monitoring operations may be added in a future version of Pacemaker. .. [#] The naming of this option was perhaps unfortunate as it is easily confused with live migration, the process of moving a resource from one node to another without stopping it. Xen virtual guests are the most common example of resources that can be migrated in this manner. diff --git a/include/crm/common/logging_internal.h b/include/crm/common/logging_internal.h index 451770e790..ec63e7fe9f 100644 --- a/include/crm/common/logging_internal.h +++ b/include/crm/common/logging_internal.h @@ -1,239 +1,238 @@ /* * Copyright 2015-2024 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. */ #ifndef PCMK__CRM_COMMON_LOGGING_INTERNAL__H #define PCMK__CRM_COMMON_LOGGING_INTERNAL__H #include #include #include #ifdef __cplusplus extern "C" { #endif /* Some warnings are too noisy when logged every time a given function is called * (for example, using a deprecated feature). As an alternative, we allow * warnings to be logged once per invocation of the calling program. Each of * those warnings needs a flag defined here. */ enum pcmk__warnings { - pcmk__wo_blind = (1 << 0), - pcmk__wo_require_all = (1 << 4), - pcmk__wo_order_score = (1 << 5), - pcmk__wo_group_order = (1 << 11), - pcmk__wo_group_coloc = (1 << 12), - pcmk__wo_set_ordering = (1 << 15), - pcmk__wo_rdisc_enabled = (1 << 16), - pcmk__wo_op_attr_expr = (1 << 19), - pcmk__wo_instance_defaults = (1 << 20), - pcmk__wo_multiple_rules = (1 << 21), + pcmk__wo_blind = (1 << 0), + pcmk__wo_record_pending = (1 << 1), + pcmk__wo_require_all = (1 << 4), + pcmk__wo_order_score = (1 << 5), + pcmk__wo_group_order = (1 << 11), + pcmk__wo_group_coloc = (1 << 12), + pcmk__wo_set_ordering = (1 << 15), + pcmk__wo_rdisc_enabled = (1 << 16), + pcmk__wo_op_attr_expr = (1 << 19), pcmk__wo_clone_master_max = (1 << 23), pcmk__wo_clone_master_node_max = (1 << 24), - pcmk__wo_master_role = (1 << 26), - pcmk__wo_slave_role = (1 << 27), + pcmk__wo_master_role = (1 << 26), + pcmk__wo_slave_role = (1 << 27), }; /*! * \internal * \brief Log a warning once per invocation of calling program * * \param[in] wo_flag enum pcmk__warnings value for this warning * \param[in] fmt... printf(3)-style format and arguments */ #define pcmk__warn_once(wo_flag, fmt...) do { \ if (!pcmk_is_set(pcmk__warnings, wo_flag)) { \ if (wo_flag == pcmk__wo_blind) { \ crm_warn(fmt); \ } else { \ pcmk__config_warn(fmt); \ } \ pcmk__warnings = pcmk__set_flags_as(__func__, __LINE__, \ LOG_TRACE, \ "Warn-once", "logging", \ pcmk__warnings, \ (wo_flag), #wo_flag); \ } \ } while (0) typedef void (*pcmk__config_error_func) (void *ctx, const char *msg, ...) G_GNUC_PRINTF(2, 3); typedef void (*pcmk__config_warning_func) (void *ctx, const char *msg, ...) G_GNUC_PRINTF(2, 3); extern pcmk__config_error_func pcmk__config_error_handler; extern pcmk__config_warning_func pcmk__config_warning_handler; extern void *pcmk__config_error_context; extern void *pcmk__config_warning_context; void pcmk__set_config_error_handler(pcmk__config_error_func error_handler, void *error_context); void pcmk__set_config_warning_handler(pcmk__config_warning_func warning_handler, void *warning_context); /* Pacemaker library functions set this when a configuration error is found, * which turns on extra messages at the end of processing. */ extern bool pcmk__config_has_error; /* Pacemaker library functions set this when a configuration warning is found, * which turns on extra messages at the end of processing. */ extern bool pcmk__config_has_warning; /*! * \internal * \brief Log an error and make crm_verify return failure status * * \param[in] fmt... printf(3)-style format string and arguments */ #define pcmk__config_err(fmt...) do { \ pcmk__config_has_error = true; \ if (pcmk__config_error_handler == NULL) { \ crm_err(fmt); \ } else { \ pcmk__config_error_handler(pcmk__config_error_context, fmt); \ } \ } while (0) /*! * \internal * \brief Log a warning and make crm_verify return failure status * * \param[in] fmt... printf(3)-style format string and arguments */ #define pcmk__config_warn(fmt...) do { \ pcmk__config_has_warning = true; \ if (pcmk__config_warning_handler == NULL) { \ crm_warn(fmt); \ } else { \ pcmk__config_warning_handler(pcmk__config_warning_context, fmt);\ } \ } while (0) /*! * \internal * \brief Execute code depending on whether trace logging is enabled * * This is similar to \p do_crm_log_unlikely() except instead of logging, it * selects one of two code blocks to execute. * * \param[in] if_action Code block to execute if trace logging is enabled * \param[in] else_action Code block to execute if trace logging is not enabled * * \note Neither \p if_action nor \p else_action can contain a \p break or * \p continue statement. */ #define pcmk__if_tracing(if_action, else_action) do { \ static struct qb_log_callsite *trace_cs = NULL; \ \ if (trace_cs == NULL) { \ trace_cs = qb_log_callsite_get(__func__, __FILE__, \ "if_tracing", LOG_TRACE, \ __LINE__, crm_trace_nonlog); \ } \ if (crm_is_callsite_active(trace_cs, LOG_TRACE, \ crm_trace_nonlog)) { \ if_action; \ } else { \ else_action; \ } \ } while (0) /*! * \internal * \brief Log XML changes line-by-line in a formatted fashion * * \param[in] level Priority at which to log the messages * \param[in] xml XML to log * * \note This does nothing when \p level is \c LOG_STDOUT. */ #define pcmk__log_xml_changes(level, xml) do { \ uint8_t _level = pcmk__clip_log_level(level); \ static struct qb_log_callsite *xml_cs = NULL; \ \ switch (_level) { \ case LOG_STDOUT: \ case LOG_NEVER: \ break; \ default: \ if (xml_cs == NULL) { \ xml_cs = qb_log_callsite_get(__func__, __FILE__, \ "xml-changes", _level, \ __LINE__, 0); \ } \ if (crm_is_callsite_active(xml_cs, _level, 0)) { \ pcmk__log_xml_changes_as(__FILE__, __func__, __LINE__, \ 0, _level, xml); \ } \ break; \ } \ } while(0) /*! * \internal * \brief Log an XML patchset line-by-line in a formatted fashion * * \param[in] level Priority at which to log the messages * \param[in] patchset XML patchset to log * * \note This does nothing when \p level is \c LOG_STDOUT. */ #define pcmk__log_xml_patchset(level, patchset) do { \ uint8_t _level = pcmk__clip_log_level(level); \ static struct qb_log_callsite *xml_cs = NULL; \ \ switch (_level) { \ case LOG_STDOUT: \ case LOG_NEVER: \ break; \ default: \ if (xml_cs == NULL) { \ xml_cs = qb_log_callsite_get(__func__, __FILE__, \ "xml-patchset", _level, \ __LINE__, 0); \ } \ if (crm_is_callsite_active(xml_cs, _level, 0)) { \ pcmk__log_xml_patchset_as(__FILE__, __func__, __LINE__, \ 0, _level, patchset); \ } \ break; \ } \ } while(0) void pcmk__log_xml_changes_as(const char *file, const char *function, uint32_t line, uint32_t tags, uint8_t level, const xmlNode *xml); void pcmk__log_xml_patchset_as(const char *file, const char *function, uint32_t line, uint32_t tags, uint8_t level, const xmlNode *patchset); /*! * \internal * \brief Initialize logging for command line tools * * \param[in] name The name of the program * \param[in] verbosity How verbose to be in logging * * \note \p verbosity is not the same as the logging level (LOG_ERR, etc.). */ void pcmk__cli_init_logging(const char *name, unsigned int verbosity); int pcmk__add_logfile(const char *filename); void pcmk__add_logfiles(gchar **log_files, pcmk__output_t *out); void pcmk__free_common_logger(void); #ifdef __cplusplus } #endif #endif // PCMK__CRM_COMMON_LOGGING_INTERNAL__H