diff --git a/include/pacemaker.h b/include/pacemaker.h index 0ec0d5a802..1f3740d5dc 100644 --- a/include/pacemaker.h +++ b/include/pacemaker.h @@ -1,300 +1,307 @@ /* * Copyright 2019-2021 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 PACEMAKER__H # define PACEMAKER__H #ifdef __cplusplus extern "C" { #endif /** * \file * \brief High Level API * \ingroup pacemaker */ # include # include # include # include /*! * \brief Synthetic cluster events that can be injected into the cluster * for running simulations. */ typedef struct { /*! A list of node names (gchar *) to simulate bringing online */ GList *node_up; /*! A list of node names (gchar *) to simulate bringing offline */ GList *node_down; /*! A list of node names (gchar *) to simulate failing */ GList *node_fail; /*! A list of operations (gchar *) to inject. The format of these strings * is described in the "Operation Specification" section of crm_simulate * help output. */ GList *op_inject; /*! A list of operations (gchar *) that should return a given error code * if they fail. The format of these strings is described in the * "Operation Specification" section of crm_simulate help output. */ GList *op_fail; /*! A list of tickets (gchar *) to simulate granting */ GList *ticket_grant; /*! A list of tickets (gchar *) to simulate revoking */ GList *ticket_revoke; /*! A list of tickets (gchar *) to simulate putting on standby */ GList *ticket_standby; /*! A list of tickets (gchar *) to simulate activating */ GList *ticket_activate; /*! Does the cluster have an active watchdog device? */ char *watchdog; /*! Does the cluster have quorum? */ char *quorum; } pcmk_injections_t; /*! * \brief Get controller status * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] dest_node Destination node for request * \param[in] message_timeout_ms Message timeout * * \return Standard Pacemaker return code */ int pcmk_controller_status(xmlNodePtr *xml, char *dest_node, unsigned int message_timeout_ms); /*! * \brief Get designated controller * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] message_timeout_ms Message timeout * * \return Standard Pacemaker return code */ int pcmk_designated_controller(xmlNodePtr *xml, unsigned int message_timeout_ms); +/*! + * \brief Free a :pcmk_injections_t structure + * + * \param[in,out] injections The structure to be freed + */ +void pcmk_free_injections(pcmk_injections_t *injections); + /*! * \brief Get pacemakerd status * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] ipc_name IPC name for request * \param[in] message_timeout_ms Message timeout * * \return Standard Pacemaker return code */ int pcmk_pacemakerd_status(xmlNodePtr *xml, char *ipc_name, unsigned int message_timeout_ms); /*! * \brief Calculate and output resource operation digests * * \param[out] xml Where to store XML with result * \param[in] 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 * \param[in] data_set Cluster working set (with status) * * \return Standard Pacemaker return code */ int pcmk_resource_digests(xmlNodePtr *xml, pe_resource_t *rsc, pe_node_t *node, GHashTable *overrides, pe_working_set_t *data_set); /*! * \brief Get nodes list * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] node_types Node type(s) to return (default: all) * * \return Standard Pacemaker return code */ int pcmk_list_nodes(xmlNodePtr *xml, char *node_types); #ifdef BUILD_PUBLIC_LIBPACEMAKER /*! * \brief Perform a STONITH action. * * \param[in] st A connection to the STONITH API. * \param[in] target The node receiving the action. * \param[in] action The action to perform. * \param[in] name Who requested the fence action? * \param[in] timeout How long to wait for the operation to complete (in ms). * \param[in] tolerance If a successful action for \p target happened within * this many ms, return 0 without performing the action * again. * \param[in] delay Apply a fencing delay. Value -1 means disable also any * static/random fencing delays from pcmk_delay_base/max. * * \return Standard Pacemaker return code */ int pcmk_fence_action(stonith_t *st, const char *target, const char *action, const char *name, unsigned int timeout, unsigned int tolerance, int delay); /*! * \brief List the fencing operations that have occurred for a specific node. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] st A connection to the STONITH API. * \param[in] target The node to get history for. * \param[in] timeout How long to wait for the operation to complete (in ms). * \param[in] quiet Suppress most output. * \param[in] verbose Include additional output. * \param[in] broadcast Gather fencing history from all nodes. * \param[in] cleanup Clean up fencing history after listing. * * \return Standard Pacemaker return code */ int pcmk_fence_history(xmlNodePtr *xml, stonith_t *st, char *target, unsigned int timeout, bool quiet, int verbose, bool broadcast, bool cleanup); /*! * \brief List all installed STONITH agents. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] st A connection to the STONITH API. * \param[in] timeout How long to wait for the operation to complete (in ms). * * \return Standard Pacemaker return code */ int pcmk_fence_installed(xmlNodePtr *xml, stonith_t *st, unsigned int timeout); /*! * \brief When was a device last fenced? * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] target The node that was fenced. * \param[in] as_nodeid * * \return Standard Pacemaker return code */ int pcmk_fence_last(xmlNodePtr *xml, const char *target, bool as_nodeid); /*! * \brief List nodes that can be fenced. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree * \param[in] st A connection to the STONITH API * \param[in] device_id Resource ID of fence device to check * \param[in] timeout How long to wait for the operation to complete (in ms) * * \return Standard Pacemaker return code */ int pcmk_fence_list_targets(xmlNodePtr *xml, stonith_t *st, const char *device_id, unsigned int timeout); /*! * \brief Get metadata for a resource. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] st A connection to the STONITH API. * \param[in] agent The fence agent to get metadata for. * \param[in] timeout How long to wait for the operation to complete (in ms). * * \return Standard Pacemaker return code */ int pcmk_fence_metadata(xmlNodePtr *xml, stonith_t *st, char *agent, unsigned int timeout); /*! * \brief List registered fence devices. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] st A connection to the STONITH API. * \param[in] target If not NULL, only return devices that can fence * this node. * \param[in] timeout How long to wait for the operation to complete (in ms). * * \return Standard Pacemaker return code */ int pcmk_fence_registered(xmlNodePtr *xml, stonith_t *st, char *target, unsigned int timeout); /*! * \brief Register a fencing level for a specific node, node regex, or attribute. * * \p target can take three different forms: * - name=value, in which case \p target is an attribute. * - @pattern, in which case \p target is a node regex. * - Otherwise, \p target is a node name. * * \param[in] st A connection to the STONITH API. * \param[in] target The object to register a fencing level for. * \param[in] fence_level Index number of level to add. * \param[in] devices Devices to use in level. * * \return Standard Pacemaker return code */ int pcmk_fence_register_level(stonith_t *st, char *target, int fence_level, stonith_key_value_t *devices); /*! * \brief Unregister a fencing level for a specific node, node regex, or attribute. * * \p target can take three different forms: * - name=value, in which case \p target is an attribute. * - @pattern, in which case \p target is a node regex. * - Otherwise, \p target is a node name. * * \param[in] st A connection to the STONITH API. * \param[in] target The object to unregister a fencing level for. * \param[in] fence_level Index number of level to remove. * * \return Standard Pacemaker return code */ int pcmk_fence_unregister_level(stonith_t *st, char *target, int fence_level); /*! * \brief Validate a STONITH device configuration. * * \note If \p xml is not NULL, it will be freed first and the previous * contents lost. * * \param[in,out] xml The destination for the result, as an XML tree. * \param[in] st A connection to the STONITH API. * \param[in] agent The agent to validate (for example, "fence_xvm"). * \param[in] id STONITH device ID (may be NULL). * \param[in] params STONITH device configuration parameters. * \param[in] timeout How long to wait for the operation to complete (in ms). * * \return Standard Pacemaker return code */ int pcmk_fence_validate(xmlNodePtr *xml, stonith_t *st, const char *agent, const char *id, stonith_key_value_t *params, unsigned int timeout); #endif #ifdef __cplusplus } #endif #endif diff --git a/lib/pacemaker/pcmk_sched_utils.c b/lib/pacemaker/pcmk_sched_utils.c index cc5d388f72..9efaf9eb96 100644 --- a/lib/pacemaker/pcmk_sched_utils.c +++ b/lib/pacemaker/pcmk_sched_utils.c @@ -1,852 +1,875 @@ /* * Copyright 2004-2021 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 // lrmd_event_data_t #include #include +#include pe__location_t * rsc2node_new(const char *id, pe_resource_t *rsc, int node_weight, const char *discover_mode, pe_node_t *foo_node, pe_working_set_t *data_set) { pe__location_t *new_con = NULL; if (rsc == NULL || id == NULL) { pe_err("Invalid constraint %s for rsc=%p", crm_str(id), rsc); return NULL; } else if (foo_node == NULL) { CRM_CHECK(node_weight == 0, return NULL); } new_con = calloc(1, sizeof(pe__location_t)); if (new_con != NULL) { new_con->id = strdup(id); new_con->rsc_lh = rsc; new_con->node_list_rh = NULL; new_con->role_filter = RSC_ROLE_UNKNOWN; if (pcmk__str_eq(discover_mode, "always", pcmk__str_null_matches | pcmk__str_casei)) { new_con->discover_mode = pe_discover_always; } else if (pcmk__str_eq(discover_mode, "never", pcmk__str_casei)) { new_con->discover_mode = pe_discover_never; } else if (pcmk__str_eq(discover_mode, "exclusive", pcmk__str_casei)) { new_con->discover_mode = pe_discover_exclusive; rsc->exclusive_discover = TRUE; } else { pe_err("Invalid %s value %s in location constraint", XML_LOCATION_ATTR_DISCOVERY, discover_mode); } if (foo_node != NULL) { pe_node_t *copy = pe__copy_node(foo_node); copy->weight = node_weight; new_con->node_list_rh = g_list_prepend(NULL, copy); } data_set->placement_constraints = g_list_prepend(data_set->placement_constraints, new_con); rsc->rsc_location = g_list_prepend(rsc->rsc_location, new_con); } return new_con; } gboolean can_run_resources(const pe_node_t * node) { if (node == NULL) { return FALSE; } #if 0 if (node->weight < 0) { return FALSE; } #endif if (node->details->online == FALSE || node->details->shutdown || node->details->unclean || node->details->standby || node->details->maintenance) { crm_trace("%s: online=%d, unclean=%d, standby=%d, maintenance=%d", node->details->uname, node->details->online, node->details->unclean, node->details->standby, node->details->maintenance); return FALSE; } return TRUE; } /*! * \internal * \brief Copy a hash table of node objects * * \param[in] nodes Hash table to copy * * \return New copy of nodes (or NULL if nodes is NULL) */ GHashTable * pcmk__copy_node_table(GHashTable *nodes) { GHashTable *new_table = NULL; GHashTableIter iter; pe_node_t *node = NULL; if (nodes == NULL) { return NULL; } new_table = pcmk__strkey_table(NULL, free); g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) { pe_node_t *new_node = pe__copy_node(node); g_hash_table_insert(new_table, (gpointer) new_node->details->id, new_node); } return new_table; } /*! * \internal * \brief Copy a list of node objects * * \param[in] list List to copy * \param[in] reset Set copies' scores to 0 * * \return New list of shallow copies of nodes in original list */ GList * pcmk__copy_node_list(const GList *list, bool reset) { GList *result = NULL; for (const GList *gIter = list; gIter != NULL; gIter = gIter->next) { pe_node_t *new_node = NULL; pe_node_t *this_node = (pe_node_t *) gIter->data; new_node = pe__copy_node(this_node); if (reset) { new_node->weight = 0; } result = g_list_prepend(result, new_node); } return result; } struct node_weight_s { pe_node_t *active; pe_working_set_t *data_set; }; /* return -1 if 'a' is more preferred * return 1 if 'b' is more preferred */ static gint sort_node_weight(gconstpointer a, gconstpointer b, gpointer data) { const pe_node_t *node1 = (const pe_node_t *)a; const pe_node_t *node2 = (const pe_node_t *)b; struct node_weight_s *nw = data; int node1_weight = 0; int node2_weight = 0; int result = 0; if (a == NULL) { return 1; } if (b == NULL) { return -1; } node1_weight = node1->weight; node2_weight = node2->weight; if (can_run_resources(node1) == FALSE) { node1_weight = -INFINITY; } if (can_run_resources(node2) == FALSE) { node2_weight = -INFINITY; } if (node1_weight > node2_weight) { crm_trace("%s (%d) > %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return -1; } if (node1_weight < node2_weight) { crm_trace("%s (%d) < %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); return 1; } crm_trace("%s (%d) == %s (%d) : weight", node1->details->uname, node1_weight, node2->details->uname, node2_weight); if (pcmk__str_eq(nw->data_set->placement_strategy, "minimal", pcmk__str_casei)) { goto equal; } if (pcmk__str_eq(nw->data_set->placement_strategy, "balanced", pcmk__str_casei)) { result = compare_capacity(node1, node2); if (result < 0) { crm_trace("%s > %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return -1; } else if (result > 0) { crm_trace("%s < %s : capacity (%d)", node1->details->uname, node2->details->uname, result); return 1; } } /* now try to balance resources across the cluster */ if (node1->details->num_resources < node2->details->num_resources) { crm_trace("%s (%d) > %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (node1->details->num_resources > node2->details->num_resources) { crm_trace("%s (%d) < %s (%d) : resources", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } if (nw->active && nw->active->details == node1->details) { crm_trace("%s (%d) > %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return -1; } else if (nw->active && nw->active->details == node2->details) { crm_trace("%s (%d) < %s (%d) : active", node1->details->uname, node1->details->num_resources, node2->details->uname, node2->details->num_resources); return 1; } equal: crm_trace("%s = %s", node1->details->uname, node2->details->uname); return strcmp(node1->details->uname, node2->details->uname); } GList * sort_nodes_by_weight(GList *nodes, pe_node_t *active_node, pe_working_set_t *data_set) { struct node_weight_s nw = { active_node, data_set }; return g_list_sort_with_data(nodes, sort_node_weight, &nw); } void native_deallocate(pe_resource_t * rsc) { if (rsc->allocated_to) { pe_node_t *old = rsc->allocated_to; crm_info("Deallocating %s from %s", rsc->id, old->details->uname); pe__set_resource_flags(rsc, pe_rsc_provisional); rsc->allocated_to = NULL; old->details->allocated_rsc = g_list_remove(old->details->allocated_rsc, rsc); old->details->num_resources--; /* old->count--; */ calculate_utilization(old->details->utilization, rsc->utilization, TRUE); free(old); } } gboolean native_assign_node(pe_resource_t *rsc, pe_node_t *chosen, gboolean force) { pcmk__output_t *out = rsc->cluster->priv; CRM_ASSERT(rsc->variant == pe_native); if (force == FALSE && chosen != NULL) { bool unset = FALSE; if(chosen->weight < 0) { unset = TRUE; // Allow the graph to assume that the remote resource will come up } else if (!can_run_resources(chosen) && !pe__is_guest_node(chosen)) { unset = TRUE; } if(unset) { crm_debug("All nodes for resource %s are unavailable" ", unclean or shutting down (%s: %d, %d)", rsc->id, chosen->details->uname, can_run_resources(chosen), chosen->weight); pe__set_next_role(rsc, RSC_ROLE_STOPPED, "node availability"); chosen = NULL; } } /* todo: update the old node for each resource to reflect its * new resource count */ native_deallocate(rsc); pe__clear_resource_flags(rsc, pe_rsc_provisional); if (chosen == NULL) { GList *gIter = NULL; char *rc_inactive = pcmk__itoa(PCMK_OCF_NOT_RUNNING); crm_debug("Could not allocate a node for %s", rsc->id); pe__set_next_role(rsc, RSC_ROLE_STOPPED, "unable to allocate"); for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) { pe_action_t *op = (pe_action_t *) gIter->data; const char *interval_ms_s = g_hash_table_lookup(op->meta, XML_LRM_ATTR_INTERVAL_MS); crm_debug("Processing %s", op->uuid); if(pcmk__str_eq(RSC_STOP, op->task, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_optional); } else if(pcmk__str_eq(RSC_START, op->task, pcmk__str_casei)) { pe__clear_action_flags(op, pe_action_runnable); //pe__set_resource_flags(rsc, pe_rsc_block); } else if (interval_ms_s && !pcmk__str_eq(interval_ms_s, "0", pcmk__str_casei)) { if(pcmk__str_eq(rc_inactive, g_hash_table_lookup(op->meta, XML_ATTR_TE_TARGET_RC), pcmk__str_casei)) { /* This is a recurring monitor for the stopped state, leave it alone */ } else { /* Normal monitor operation, cancel it */ pe__clear_action_flags(op, pe_action_runnable); } } } free(rc_inactive); return FALSE; } crm_debug("Assigning %s to %s", chosen->details->uname, rsc->id); rsc->allocated_to = pe__copy_node(chosen); chosen->details->allocated_rsc = g_list_prepend(chosen->details->allocated_rsc, rsc); chosen->details->num_resources++; chosen->count++; calculate_utilization(chosen->details->utilization, rsc->utilization, FALSE); if (pcmk_is_set(rsc->cluster->flags, pe_flag_show_utilization)) { out->message(out, "resource-util", rsc, chosen, __func__); } return TRUE; } void log_action(unsigned int log_level, const char *pre_text, pe_action_t * action, gboolean details) { const char *node_uname = NULL; const char *node_uuid = NULL; const char *desc = NULL; if (action == NULL) { crm_trace("%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": "); return; } if (pcmk_is_set(action->flags, pe_action_pseudo)) { node_uname = NULL; node_uuid = NULL; } else if (action->node != NULL) { node_uname = action->node->details->uname; node_uuid = action->node->details->id; } else { node_uname = ""; node_uuid = NULL; } switch (text2task(action->task)) { case stonith_node: case shutdown_crm: if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (pcmk_is_set(action->flags, pe_action_optional)) { desc = "Optional "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } 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, pe_action_optional)) { desc = "Optional "; } else if (pcmk_is_set(action->flags, pe_action_pseudo)) { desc = "Pseudo "; } else if (!pcmk_is_set(action->flags, pe_action_runnable)) { desc = "!!Non-Startable!! "; } else if (pcmk_is_set(action->flags, pe_action_processed)) { desc = ""; } 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) { GList *gIter = NULL; crm_trace("\t\t====== Preceding Actions"); gIter = action->actions_before; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\t\t", other->action, FALSE); } crm_trace("\t\t====== Subsequent Actions"); gIter = action->actions_after; for (; gIter != NULL; gIter = gIter->next) { pe_action_wrapper_t *other = (pe_action_wrapper_t *) gIter->data; log_action(log_level + 1, "\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)); } } gboolean can_run_any(GHashTable * nodes) { GHashTableIter iter; pe_node_t *node = NULL; if (nodes == NULL) { return FALSE; } g_hash_table_iter_init(&iter, nodes); while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) { if (can_run_resources(node) && node->weight >= 0) { return TRUE; } } return FALSE; } pe_action_t * create_pseudo_resource_op(pe_resource_t * rsc, const char *task, bool optional, bool runnable, pe_working_set_t *data_set) { pe_action_t *action = custom_action(rsc, pcmk__op_key(rsc->id, task, 0), task, NULL, optional, TRUE, data_set); pe__set_action_flags(action, pe_action_pseudo); if(runnable) { pe__set_action_flags(action, pe_action_runnable); } return action; } /*! * \internal * \brief Create an executor cancel op * * \param[in] rsc Resource of action to cancel * \param[in] task Name of action to cancel * \param[in] interval_ms Interval of action to cancel * \param[in] node Node of action to cancel * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * pe_cancel_op(pe_resource_t *rsc, const char *task, guint interval_ms, pe_node_t *node, pe_working_set_t *data_set) { pe_action_t *cancel_op; char *interval_ms_s = crm_strdup_printf("%u", interval_ms); // @TODO dangerous if possible to schedule another action with this key char *key = pcmk__op_key(rsc->id, task, interval_ms); cancel_op = custom_action(rsc, key, RSC_CANCEL, node, FALSE, TRUE, data_set); free(cancel_op->task); cancel_op->task = strdup(RSC_CANCEL); free(cancel_op->cancel_task); cancel_op->cancel_task = strdup(task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_TASK, task); add_hash_param(cancel_op->meta, XML_LRM_ATTR_INTERVAL_MS, interval_ms_s); free(interval_ms_s); return cancel_op; } /*! * \internal * \brief Create a shutdown op for a scheduler transition * * \param[in] node Node being shut down * \param[in] data_set Working set of cluster * * \return Created op */ pe_action_t * sched_shutdown_op(pe_node_t *node, pe_working_set_t *data_set) { char *shutdown_id = crm_strdup_printf("%s-%s", CRM_OP_SHUTDOWN, node->details->uname); pe_action_t *shutdown_op = custom_action(NULL, shutdown_id, CRM_OP_SHUTDOWN, node, FALSE, TRUE, data_set); shutdown_constraints(node, shutdown_op, data_set); add_hash_param(shutdown_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE); return shutdown_op; } static char * generate_transition_magic(const char *transition_key, int op_status, int op_rc) { CRM_CHECK(transition_key != NULL, return NULL); return crm_strdup_printf("%d:%d;%s", op_status, op_rc, transition_key); } static void append_digest(lrmd_event_data_t *op, xmlNode *update, const char *version, const char *magic, int level) { /* this will enable us to later determine that the * resource's parameters have changed and we should force * a restart */ 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, version); #if 0 if (level < get_crm_log_level() && op->interval_ms == 0 && pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { char *digest_source = dump_xml_unformatted(args_xml); do_crm_log(level, "Calculated digest %s for %s (%s). Source: %s\n", digest, ID(update), magic, digest_source); free(digest_source); } #endif 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 * \param[in] level A log message will be logged at this level * * \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, int level) { 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); do_crm_log(level, "%s: Updating resource %s after %s op %s (interval=%u)", origin, op->rsc_id, op->op_type, pcmk_exec_status_str(op->op_status), op->interval_ms); crm_trace("DC version: %s", caller_version); 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, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { if (op->op_status == PCMK_EXEC_DONE) { task = CRMD_ACTION_START; } else { task = CRMD_ACTION_STATUS; } } key = pcmk__op_key(op->rsc_id, task, op->interval_ms); if (pcmk__str_eq(task, CRMD_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(). */ op->op_status = PCMK_EXEC_DONE; op->rc = 0; } } 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, XML_ATTR_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 = generate_transition_magic(op->user_data, op->op_status, op->rc); } crm_xml_add(xml_op, XML_ATTR_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, XML_ATTR_ORIGIN, origin); crm_xml_add(xml_op, XML_ATTR_CRM_VERSION, 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, exit_reason == NULL ? "" : 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, CRMD_ACTION_MIGRATE, CRMD_ACTION_MIGRATED, 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)); } append_digest(op, xml_op, caller_version, magic, LOG_DEBUG); 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; } pcmk__output_t * pcmk__new_logger(void) { int rc = pcmk_rc_ok; pcmk__output_t *out = NULL; const char* argv[] = { "", NULL }; pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_LOG, { NULL, NULL, NULL } }; pcmk__register_formats(NULL, formats); rc = pcmk__output_new(&out, "log", NULL, (char**)argv); if ((rc != pcmk_rc_ok) || (out == NULL)) { crm_err("Can't log resource details due to internal error: %s\n", pcmk_rc_str(rc)); return NULL; } pe__register_messages(out); pcmk__register_lib_messages(out); return out; } /*! * \internal * \brief Check whether a resource has reached its migration threshold on a node * * \param[in] rsc Resource to check * \param[in] node Node to check * \param[in] data_set Cluster working set * \param[out] failed If the threshold has been reached, this will be set to * the resource that failed (possibly a parent of \p rsc) * * \return true if the migration threshold has been reached, false otherwise */ bool pcmk__threshold_reached(pe_resource_t *rsc, pe_node_t *node, pe_working_set_t *data_set, pe_resource_t **failed) { int fail_count, remaining_tries; pe_resource_t *rsc_to_ban = rsc; // Migration threshold of 0 means never force away if (rsc->migration_threshold == 0) { return false; } // If we're ignoring failures, also ignore the migration threshold if (pcmk_is_set(rsc->flags, pe_rsc_failure_ignored)) { return false; } // If there are no failures, there's no need to force away fail_count = pe_get_failcount(node, rsc, NULL, pe_fc_effective|pe_fc_fillers, NULL, data_set); if (fail_count <= 0) { return false; } // If failed resource is anonymous clone instance, we'll force clone away if (!pcmk_is_set(rsc->flags, pe_rsc_unique)) { rsc_to_ban = uber_parent(rsc); } // How many more times recovery will be tried on this node remaining_tries = rsc->migration_threshold - fail_count; if (remaining_tries <= 0) { crm_warn("%s cannot run on %s due to reaching migration threshold " "(clean up resource to allow again)" CRM_XS " failures=%d migration-threshold=%d", rsc_to_ban->id, node->details->uname, fail_count, rsc->migration_threshold); if (failed != NULL) { *failed = rsc_to_ban; } return true; } crm_info("%s can fail %d more time%s on " "%s before reaching migration threshold (%d)", rsc_to_ban->id, remaining_tries, pcmk__plural_s(remaining_tries), node->details->uname, rsc->migration_threshold); return false; } + +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/tools/crm_simulate.c b/tools/crm_simulate.c index 66ccb91faa..2b96cedb08 100644 --- a/tools/crm_simulate.c +++ b/tools/crm_simulate.c @@ -1,809 +1,798 @@ /* * Copyright 2009-2021 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 #define SUMMARY "crm_simulate - simulate a Pacemaker cluster's response to events" struct { gboolean all_actions; char *dot_file; char *graph_file; gchar *input_file; guint modified; pcmk_injections_t *injections; gchar *output_file; gboolean print_pending; gboolean process; long long repeat; gboolean show_scores; gboolean show_utilization; gboolean simulate; gboolean store; gchar *test_dir; char *use_date; char *xml_file; } options = { .print_pending = TRUE, .repeat = 1 }; unsigned int section_opts = 0; cib_t *global_cib = NULL; char *temp_shadow = NULL; extern gboolean bringing_nodes_online; crm_exit_t exit_code = CRM_EX_OK; #define INDENT " " static pcmk__supported_format_t formats[] = { PCMK__SUPPORTED_FORMAT_NONE, PCMK__SUPPORTED_FORMAT_TEXT, PCMK__SUPPORTED_FORMAT_XML, { NULL, NULL, NULL } }; static gboolean attrs_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { section_opts |= pcmk_section_attributes; return TRUE; } static gboolean failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { section_opts |= pcmk_section_failcounts | pcmk_section_failures; return TRUE; } static gboolean in_place_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.store = TRUE; options.process = TRUE; options.simulate = TRUE; return TRUE; } static gboolean live_check_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.xml_file) { free(options.xml_file); } options.xml_file = NULL; return TRUE; } static gboolean node_down_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->node_down = g_list_append(options.injections->node_down, g_strdup(optarg)); return TRUE; } static gboolean node_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->node_fail = g_list_append(options.injections->node_fail, g_strdup(optarg)); return TRUE; } static gboolean node_up_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; bringing_nodes_online = TRUE; options.injections->node_up = g_list_append(options.injections->node_up, g_strdup(optarg)); return TRUE; } static gboolean op_fail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.process = TRUE; options.simulate = TRUE; options.injections->op_fail = g_list_append(options.injections->op_fail, g_strdup(optarg)); return TRUE; } static gboolean op_inject_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->op_inject = g_list_append(options.injections->op_inject, g_strdup(optarg)); return TRUE; } static gboolean quorum_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.injections->quorum) { free(options.injections->quorum); } options.modified++; options.injections->quorum = strdup(optarg); return TRUE; } static gboolean save_dotfile_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.dot_file) { free(options.dot_file); } options.process = TRUE; options.dot_file = strdup(optarg); return TRUE; } static gboolean save_graph_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.graph_file) { free(options.graph_file); } options.process = TRUE; options.graph_file = strdup(optarg); return TRUE; } static gboolean show_scores_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.process = TRUE; options.show_scores = TRUE; return TRUE; } static gboolean simulate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.process = TRUE; options.simulate = TRUE; return TRUE; } static gboolean ticket_activate_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->ticket_activate = g_list_append(options.injections->ticket_activate, g_strdup(optarg)); return TRUE; } static gboolean ticket_grant_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->ticket_grant = g_list_append(options.injections->ticket_grant, g_strdup(optarg)); return TRUE; } static gboolean ticket_revoke_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->ticket_revoke = g_list_append(options.injections->ticket_revoke, g_strdup(optarg)); return TRUE; } static gboolean ticket_standby_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.modified++; options.injections->ticket_standby = g_list_append(options.injections->ticket_standby, g_strdup(optarg)); return TRUE; } static gboolean utilization_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { options.process = TRUE; options.show_utilization = TRUE; return TRUE; } static gboolean watchdog_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.injections->watchdog) { free(options.injections->watchdog); } options.modified++; options.injections->watchdog = strdup(optarg); return TRUE; } static gboolean xml_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.xml_file) { free(options.xml_file); } options.xml_file = strdup(optarg); return TRUE; } static gboolean xml_pipe_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) { if (options.xml_file) { free(options.xml_file); } options.xml_file = strdup("-"); return TRUE; } static GOptionEntry operation_entries[] = { { "run", 'R', 0, G_OPTION_ARG_NONE, &options.process, "Process the supplied input and show what actions the cluster will take in response", NULL }, { "simulate", 'S', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, simulate_cb, "Like --run, but also simulate taking those actions and show the resulting new status", NULL }, { "in-place", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, in_place_cb, "Like --simulate, but also store the results back to the input file", NULL }, { "show-attrs", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, attrs_cb, "Show node attributes", NULL }, { "show-failcounts", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, failcounts_cb, "Show resource fail counts", NULL }, { "show-scores", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_scores_cb, "Show allocation scores", NULL }, { "show-utilization", 'U', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, utilization_cb, "Show utilization information", NULL }, { "profile", 'P', 0, G_OPTION_ARG_FILENAME, &options.test_dir, "Process all the XML files in the named directory to create profiling data", "DIR" }, { "repeat", 'N', 0, G_OPTION_ARG_INT, &options.repeat, "With --profile, repeat each test N times and print timings", "N" }, /* Deprecated */ { "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 synthetic_entries[] = { { "node-up", 'u', 0, G_OPTION_ARG_CALLBACK, node_up_cb, "Simulate bringing a node online", "NODE" }, { "node-down", 'd', 0, G_OPTION_ARG_CALLBACK, node_down_cb, "Simulate taking a node offline", "NODE" }, { "node-fail", 'f', 0, G_OPTION_ARG_CALLBACK, node_fail_cb, "Simulate a node failing", "NODE" }, { "op-inject", 'i', 0, G_OPTION_ARG_CALLBACK, op_inject_cb, "Generate a failure for the cluster to react to in the simulation.\n" INDENT "See `Operation Specification` help for more information.", "OPSPEC" }, { "op-fail", 'F', 0, G_OPTION_ARG_CALLBACK, op_fail_cb, "If the specified task occurs during the simulation, have it fail with return code ${rc}.\n" INDENT "The transition will normally stop at the failed action.\n" INDENT "Save the result with --save-output and re-run with --xml-file.\n" INDENT "See `Operation Specification` help for more information.", "OPSPEC" }, { "set-datetime", 't', 0, G_OPTION_ARG_STRING, &options.use_date, "Set date/time (ISO 8601 format, see https://en.wikipedia.org/wiki/ISO_8601)", "DATETIME" }, { "quorum", 'q', 0, G_OPTION_ARG_CALLBACK, quorum_cb, "Set to '1' (or 'true') to indicate cluster has quorum", "QUORUM" }, { "watchdog", 'w', 0, G_OPTION_ARG_CALLBACK, watchdog_cb, "Set to '1' (or 'true') to indicate cluster has an active watchdog device", "DEVICE" }, { "ticket-grant", 'g', 0, G_OPTION_ARG_CALLBACK, ticket_grant_cb, "Simulate granting a ticket", "TICKET" }, { "ticket-revoke", 'r', 0, G_OPTION_ARG_CALLBACK, ticket_revoke_cb, "Simulate revoking a ticket", "TICKET" }, { "ticket-standby", 'b', 0, G_OPTION_ARG_CALLBACK, ticket_standby_cb, "Simulate making a ticket standby", "TICKET" }, { "ticket-activate", 'e', 0, G_OPTION_ARG_CALLBACK, ticket_activate_cb, "Simulate activating a ticket", "TICKET" }, { NULL } }; static GOptionEntry artifact_entries[] = { { "save-input", 'I', 0, G_OPTION_ARG_FILENAME, &options.input_file, "Save the input configuration to the named file", "FILE" }, { "save-output", 'O', 0, G_OPTION_ARG_FILENAME, &options.output_file, "Save the output configuration to the named file", "FILE" }, { "save-graph", 'G', 0, G_OPTION_ARG_CALLBACK, save_graph_cb, "Save the transition graph (XML format) to the named file", "FILE" }, { "save-dotfile", 'D', 0, G_OPTION_ARG_CALLBACK, save_dotfile_cb, "Save the transition graph (DOT format) to the named file", "FILE" }, { "all-actions", 'a', 0, G_OPTION_ARG_NONE, &options.all_actions, "Display all possible actions in DOT graph (even if not part of transition)", NULL }, { NULL } }; static GOptionEntry source_entries[] = { { "live-check", 'L', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, live_check_cb, "Connect to CIB manager and use the current CIB contents as input", NULL }, { "xml-file", 'x', 0, G_OPTION_ARG_CALLBACK, xml_file_cb, "Retrieve XML from the named file", "FILE" }, { "xml-pipe", 'p', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, xml_pipe_cb, "Retrieve XML from stdin", NULL }, { NULL } }; static void print_cluster_status(pe_working_set_t * data_set, unsigned int show_opts, unsigned int section_opts, const char *title, bool print_spacer) { pcmk__output_t *out = data_set->priv; GList *all = NULL; section_opts |= pcmk_section_nodes | pcmk_section_resources; 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", data_set, 0, NULL, FALSE, section_opts, show_opts | pcmk_show_inactive_rscs, NULL, all, all); out->end_list(out); g_list_free(all); } static void print_transition_summary(pe_working_set_t *data_set, bool print_spacer) { pcmk__output_t *out = data_set->priv; PCMK__OUTPUT_SPACER_IF(out, print_spacer); out->begin_list(out, NULL, NULL, "Transition Summary"); LogNodeActions(data_set); g_list_foreach(data_set->resources, (GFunc) LogActions, data_set); out->end_list(out); } static int setup_input(const char *input, const char *output, GError **error) { int rc = pcmk_rc_ok; xmlNode *cib_object = NULL; char *local_output = NULL; if (input == NULL) { /* Use live CIB */ rc = cib__signon_query(NULL, &cib_object); if (rc != pcmk_rc_ok) { g_set_error(error, PCMK__RC_ERROR, rc, "CIB query failed: %s", pcmk_rc_str(rc)); return rc; } } else if (pcmk__str_eq(input, "-", pcmk__str_casei)) { cib_object = filename2xml(NULL); } else { cib_object = filename2xml(input); } if (get_object_root(XML_CIB_TAG_STATUS, cib_object) == NULL) { create_xml_node(cib_object, XML_CIB_TAG_STATUS); } if (cli_config_update(&cib_object, NULL, FALSE) == FALSE) { free_xml(cib_object); return pcmk_rc_transform_failed; } if (validate_xml(cib_object, NULL, FALSE) != TRUE) { free_xml(cib_object); return pcmk_rc_schema_validation; } if (output == NULL) { char *pid = pcmk__getpid_s(); local_output = get_shadow_file(pid); temp_shadow = strdup(local_output); output = local_output; free(pid); } rc = write_xml_file(cib_object, output, FALSE); free_xml(cib_object); cib_object = NULL; if (rc < 0) { rc = pcmk_legacy2rc(rc); g_set_error(error, PCMK__EXITC_ERROR, CRM_EX_CANTCREAT, "Could not create '%s': %s", output, pcmk_rc_str(rc)); return rc; } else { setenv("CIB_file", output, 1); free(local_output); return pcmk_rc_ok; } } 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), "Display only essential output", NULL }, { NULL } }; const char *description = "Operation Specification:\n\n" "The OPSPEC in any command line option is of the form\n" "${resource}_${task}_${interval_in_ms}@${node}=${rc}\n" "(memcached_monitor_20000@bart.example.com=7, for example).\n" "${rc} is an OCF return code. For more information on these\n" "return codes, refer to https://clusterlabs.org/pacemaker/doc/2.1/Pacemaker_Administration/html/agents.html#ocf-return-codes\n\n" "Examples:\n\n" "Pretend a recurring monitor action found memcached stopped on node\n" "fred.example.com and, during recovery, that the memcached stop\n" "action failed:\n\n" "\tcrm_simulate -LS --op-inject memcached:0_monitor_20000@bart.example.com=7 " "--op-fail memcached:0_stop_0@fred.example.com=1 --save-output /tmp/memcached-test.xml\n\n" "Now see what the reaction to the stop failed would be:\n\n" "\tcrm_simulate -S --xml-file /tmp/memcached-test.xml\n\n"; context = pcmk__build_arg_context(args, "text (default), xml", group, NULL); pcmk__add_main_args(context, extra_prog_entries); g_option_context_set_description(context, description); pcmk__add_arg_group(context, "operations", "Operations:", "Show operations options", operation_entries); pcmk__add_arg_group(context, "synthetic", "Synthetic Cluster Events:", "Show synthetic cluster event options", synthetic_entries); pcmk__add_arg_group(context, "artifact", "Artifact Options:", "Show artifact options", artifact_entries); pcmk__add_arg_group(context, "source", "Data Source:", "Show data source options", source_entries); return context; } static void reset(pe_working_set_t *data_set, xmlNodePtr input, pcmk__output_t *out) { data_set->input = input; data_set->priv = out; pcmk__set_effective_date(data_set, true, options.use_date); if(options.xml_file) { pe__set_working_set_flags(data_set, pe_flag_sanitized); } if (options.show_scores) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (options.show_utilization) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } } int main(int argc, char **argv) { int printed = pcmk_rc_no_output; int rc = pcmk_rc_ok; pe_working_set_t *data_set = NULL; pcmk__output_t *out = NULL; xmlNode *input = NULL; GError *error = NULL; GOptionGroup *output_group = NULL; pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY); gchar **processed_args = pcmk__cmdline_preproc(argv, "bdefgiqrtuwxDFGINOP"); GOptionContext *context = build_arg_context(args, &output_group); options.injections = calloc(1, sizeof(pcmk_injections_t)); if (options.injections == NULL) { rc = ENOMEM; goto done; } /* This must come before g_option_context_parse_strv. */ options.xml_file = strdup("-"); 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_simulate", args->verbosity); rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv); if (rc != pcmk_rc_ok) { fprintf(stderr, "Error creating output format %s: %s\n", args->output_ty, pcmk_rc_str(rc)); exit_code = CRM_EX_ERROR; goto done; } if (pcmk__str_eq(args->output_ty, "text", pcmk__str_null_matches) && !options.show_scores && !options.show_utilization) { pcmk__force_args(context, &error, "%s --text-fancy", g_get_prgname()); } else if (pcmk__str_eq(args->output_ty, "xml", pcmk__str_none)) { pcmk__force_args(context, &error, "%s --xml-simple-list --xml-substitute", g_get_prgname()); } pe__register_messages(out); pcmk__register_lib_messages(out); out->quiet = args->quiet; if (args->version) { out->version(out, false); goto done; } if (args->verbosity > 0) { #ifdef PCMK__COMPAT_2_0 /* Redirect stderr to stdout so we can grep the output */ close(STDERR_FILENO); dup2(STDOUT_FILENO, STDERR_FILENO); #endif } data_set = pe_new_working_set(); if (data_set == NULL) { rc = ENOMEM; g_set_error(&error, PCMK__RC_ERROR, rc, "Could not allocate working set"); goto done; } if (options.show_scores) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (options.show_utilization) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } pe__set_working_set_flags(data_set, pe_flag_no_compat); if (options.test_dir != NULL) { data_set->priv = out; pcmk__profile_dir(options.test_dir, options.repeat, data_set, options.use_date); rc = pcmk_rc_ok; goto done; } rc = setup_input(options.xml_file, options.store ? options.xml_file : options.output_file, &error); if (rc != pcmk_rc_ok) { goto done; } rc = cib__signon_query(&global_cib, &input); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "CIB query failed: %s", pcmk_rc_str(rc)); goto done; } reset(data_set, input, out); cluster_status(data_set); if (!out->is_quiet(out)) { if (pcmk_is_set(data_set->flags, pe_flag_maintenance_mode)) { printed = out->message(out, "maint-mode", data_set->flags); } if (data_set->disabled_resources || data_set->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", data_set->disabled_resources, data_set->ninstances, data_set->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(data_set, options.print_pending ? pcmk_show_pending : 0, section_opts, "Current cluster status", printed == pcmk_rc_ok); printed = pcmk_rc_ok; } if (options.modified) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); modify_configuration(data_set, global_cib, options.injections); printed = pcmk_rc_ok; rc = global_cib->cmds->query(global_cib, NULL, &input, cib_sync_call); if (rc != pcmk_rc_ok) { rc = pcmk_legacy2rc(rc); g_set_error(&error, PCMK__RC_ERROR, rc, "Could not get modified CIB: %s", pcmk_rc_str(rc)); goto done; } cleanup_calculations(data_set); reset(data_set, input, out); cluster_status(data_set); } if (options.input_file != NULL) { rc = write_xml_file(input, options.input_file, FALSE); if (rc < 0) { rc = pcmk_legacy2rc(rc); g_set_error(&error, PCMK__RC_ERROR, rc, "Could not create '%s': %s", options.input_file, pcmk_rc_str(rc)); goto done; } } if (options.process || options.simulate) { crm_time_t *local_date = NULL; pcmk__output_t *logger_out = NULL; if (pcmk_all_flags_set(data_set->flags, pe_flag_show_scores|pe_flag_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Allocation Scores and Utilization Information"); printed = pcmk_rc_ok; } else if (pcmk_is_set(data_set->flags, pe_flag_show_scores)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Allocation Scores"); printed = pcmk_rc_ok; } else if (pcmk_is_set(data_set->flags, pe_flag_show_utilization)) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); out->begin_list(out, NULL, NULL, "Utilization Information"); printed = pcmk_rc_ok; } else { logger_out = pcmk__new_logger(); if (logger_out == NULL) { goto done; } data_set->priv = logger_out; } pcmk__schedule_actions(data_set, input, local_date); if (logger_out == NULL) { out->end_list(out); } else { logger_out->finish(logger_out, CRM_EX_OK, true, NULL); pcmk__output_free(logger_out); data_set->priv = out; } input = NULL; /* Don't try and free it twice */ if (options.graph_file != NULL) { write_xml_file(data_set->graph, options.graph_file, FALSE); } if (options.dot_file != NULL) { rc = pcmk__write_sim_dotfile(data_set, options.dot_file, options.all_actions, args->verbosity > 0); if (rc != pcmk_rc_ok) { g_set_error(&error, PCMK__RC_ERROR, rc, "Could not open %s for writing: %s", options.dot_file, pcmk_rc_str(rc)); goto done; } } if (!out->is_quiet(out)) { print_transition_summary(data_set, printed == pcmk_rc_ok); } } rc = pcmk_rc_ok; if (options.simulate) { PCMK__OUTPUT_SPACER_IF(out, printed == pcmk_rc_ok); if (run_simulation(data_set, global_cib, options.injections->op_fail) != pcmk_rc_ok) { rc = pcmk_rc_error; } if (!out->is_quiet(out)) { pcmk__set_effective_date(data_set, true, options.use_date); if (options.show_scores) { pe__set_working_set_flags(data_set, pe_flag_show_scores); } if (options.show_utilization) { pe__set_working_set_flags(data_set, pe_flag_show_utilization); } cluster_status(data_set); print_cluster_status(data_set, 0, section_opts, "Revised Cluster Status", true); } } done: pcmk__output_and_clear_error(error, NULL); /* There sure is a lot to free in options. */ free(options.dot_file); free(options.graph_file); g_free(options.input_file); - g_list_free_full(options.injections->node_up, g_free); - g_list_free_full(options.injections->node_down, g_free); - g_list_free_full(options.injections->node_fail, g_free); - g_list_free_full(options.injections->op_fail, g_free); - g_list_free_full(options.injections->op_inject, g_free); g_free(options.output_file); g_free(options.test_dir); - g_list_free_full(options.injections->ticket_grant, g_free); - g_list_free_full(options.injections->ticket_revoke, g_free); - g_list_free_full(options.injections->ticket_standby, g_free); - g_list_free_full(options.injections->ticket_activate, g_free); - free(options.injections->quorum); - free(options.injections->watchdog); free(options.use_date); free(options.xml_file); - free(options.injections); + pcmk_free_injections(options.injections); pcmk__free_arg_context(context); g_strfreev(processed_args); if (data_set) { pe_free_working_set(data_set); } if (global_cib) { global_cib->cmds->signoff(global_cib); cib_delete(global_cib); } fflush(stderr); if (temp_shadow) { unlink(temp_shadow); free(temp_shadow); } if (rc != pcmk_rc_ok) { exit_code = pcmk_rc2exitc(rc); } if (out != NULL) { out->finish(out, exit_code, true, NULL); pcmk__output_free(out); } crm_exit(exit_code); }