diff --git a/cts/cli/regression.daemons.exp b/cts/cli/regression.daemons.exp
index 7b784b49e1..98b83b6267 100644
--- a/cts/cli/regression.daemons.exp
+++ b/cts/cli/regression.daemons.exp
@@ -1,451 +1,456 @@
=#=#=#= Begin test: Get CIB manager metadata =#=#=#=
1.1Cluster options used by Pacemaker's Cluster Information Base managerCluster Information Base manager optionsEnable Access Control Lists (ACLs) for the CIBEnable Access Control Lists (ACLs) for the CIBRaise this if log has "Evicting client" messages for cluster daemon PIDs (a good value is the number of resources in the cluster multiplied by the number of nodes).Maximum IPC message backlog before disconnecting a cluster daemon
=#=#=#= End test: Get CIB manager metadata - OK (0) =#=#=#=
* Passed: pacemaker-based - Get CIB manager metadata
=#=#=#= Begin test: Get controller metadata =#=#=#=
1.1Cluster options used by Pacemaker's controllerPacemaker controller optionsIncludes a hash which identifies the exact changeset the code was built from. Used for diagnostic purposes.Pacemaker version on cluster node elected Designated Controller (DC)Used for informational and diagnostic purposes.The messaging stack on which Pacemaker is currently runningThis optional value is mostly for users' convenience as desired in administration, but may also be used in Pacemaker configuration rules via the #cluster-name node attribute, and by higher-level tools and resource agents.An arbitrary name for the clusterThe optimal value will depend on the speed and load of your network and the type of switches used.How long to wait for a response from other nodes during start-upPacemaker is primarily event-driven, and looks ahead to know when to recheck cluster state for failure timeouts and most time-based rules. However, it will also recheck the cluster after this amount of inactivity, to evaluate rules with date specifications and serve as a fail-safe for certain types of scheduler bugs. Allowed values: Zero disables polling, while positive values are an interval in seconds(unless other units are specified, for example "5min")Polling interval to recheck cluster state and evaluate rules with date specificationsThe cluster will slow down its recovery process when the amount of system resources used (currently CPU) approaches this limitMaximum amount of system load that should be used by cluster nodesMaximum number of jobs that can be scheduled per node (defaults to 2x cores)Maximum number of jobs that can be scheduled per node (defaults to 2x cores)A cluster node may receive notification of its own fencing if fencing is misconfigured, or if fabric fencing is in use that doesn't cut cluster communication. Allowed values are "stop" to attempt to immediately stop Pacemaker and stay stopped, or "panic" to attempt to immediately reboot the local node, falling back to stop on failure.How a cluster node should react if notified of its own fencingDeclare an election failed if it is not decided within this much time. If you need to adjust this value, it probably indicates the presence of a bug.*** Advanced Use Only ***Exit immediately if shutdown does not complete within this much time. If you need to adjust this value, it probably indicates the presence of a bug.*** Advanced Use Only ***If you need to adjust this value, it probably indicates the presence of a bug.*** Advanced Use Only ***If you need to adjust this value, it probably indicates the presence of a bug.*** Advanced Use Only ***Delay cluster recovery for this much time to allow for additional events to occur. Useful if your configuration is sensitive to the order in which ping updates arrive.*** Advanced Use Only *** Enabling this option will slow down cluster recovery under all conditionsIf this is set to a positive value, lost nodes are assumed to self-fence using watchdog-based SBD within this much time. This does not require a fencing resource to be explicitly configured, though a fence_watchdog resource can be configured, to limit use to specific nodes. If this is set to 0 (the default), the cluster will never assume watchdog-based self-fencing. If this is set to a negative value, the cluster will use twice the local value of the `SBD_WATCHDOG_TIMEOUT` environment variable if that is positive, or otherwise treat this as 0. WARNING: When used, this timeout must be larger than `SBD_WATCHDOG_TIMEOUT` on all nodes that use watchdog-based SBD, and Pacemaker will refuse to start on any of those nodes where this is not true for the local value or SBD is not active. When this is set to a negative value, `SBD_WATCHDOG_TIMEOUT` must be set to the same value on all nodes that use SBD, otherwise data corruption or loss could occur.How long before nodes can be assumed to be safely down when watchdog-based self-fencing via SBD is in useHow many times fencing can fail before it will no longer be immediately re-attempted on a targetHow many times fencing can fail before it will no longer be immediately re-attempted on a targetWhat to do when the cluster does not have quorum Allowed values: stop, freeze, ignore, demote, suicideWhat to do when the cluster does not have quorumWhen true, resources active on a node when it is cleanly shut down are kept "locked" to that node (not allowed to run elsewhere) until they start again on that node after it rejoins (or for at most shutdown-lock-limit, if set). Stonith resources and Pacemaker Remote connections are never locked. Clone and bundle instances and the promoted role of promotable clones are currently never locked, though support could be added in a future release.Whether to lock resources to a cleanly shut down nodeIf shutdown-lock is true and this is set to a nonzero time duration, shutdown locks will expire after this much time has passed since the shutdown was initiated, even if the node has not rejoined.Do not lock resources to a cleanly shut down node longer than this
+
+ Fence nodes that do not join the controller process group within this much time after joining the cluster, to allow the cluster to continue managing resources. A value of 0 means never fence pending nodes.
+ How long to wait for a node that has joined the cluster to join the controller process group
+
+
=#=#=#= End test: Get controller metadata - OK (0) =#=#=#=
* Passed: pacemaker-controld - Get controller metadata
=#=#=#= Begin test: Get fencer metadata =#=#=#=
1.1Instance attributes available for all "stonith"-class resources and used by Pacemaker's fence daemon, formerly known as stonithdInstance attributes available for all "stonith"-class resourcessome devices do not support the standard 'port' parameter or may provide additional ones. Use this to specify an alternate, device-specific, parameter that should indicate the machine to be fenced. A value of none can be used to tell the cluster not to supply any additional parameters.Advanced use only: An alternate parameter to supply instead of 'port'Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and 3 for node2A mapping of host names to ports numbers for devices that do not support host names.A list of machines controlled by this device (Optional unless pcmk_host_list=static-list)Eg. node1,node2,node3Allowed values: dynamic-list (query the device via the 'list' command), static-list (check the pcmk_host_list attribute), status (query the device via the 'status' command), none (assume every device can fence every machine)How to determine which machines are controlled by the device.Enable a delay of no more than the time specified before executing fencing actions. Pacemaker derives the overall delay by taking the value of pcmk_delay_base and adding a random delay value such that the sum is kept below this maximum.Enable a base delay for fencing actions and specify base delay value.This enables a static delay for fencing actions, which can help avoid "death matches" where two nodes try to fence each other at the same time. If pcmk_delay_max is also used, a random delay will be added such that the total delay is kept below that value.This can be set to a single time value to apply to any node targeted by this device (useful if a separate device is configured for each target), or to a node map (for example, "node1:1s;node2:5") to set a different value per target.Enable a base delay for fencing actions and specify base delay value.Cluster property concurrent-fencing=true needs to be configured first.Then use this to specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.The maximum number of actions can be performed in parallel on this deviceSome devices do not support the standard commands or may provide additional ones.\nUse this to specify an alternate, device-specific, command that implements the 'reboot' action.Advanced use only: An alternate command to run instead of 'reboot'Some devices need much more/less time to complete than normal.Use this to specify an alternate, device-specific, timeout for 'reboot' actions.Advanced use only: Specify an alternate timeout to use for reboot actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'reboot' actions before giving up.Advanced use only: The maximum number of times to retry the 'reboot' command within the timeout periodSome devices do not support the standard commands or may provide additional ones.Use this to specify an alternate, device-specific, command that implements the 'off' action.Advanced use only: An alternate command to run instead of 'off'Some devices need much more/less time to complete than normal.Use this to specify an alternate, device-specific, timeout for 'off' actions.Advanced use only: Specify an alternate timeout to use for off actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'off' actions before giving up.Advanced use only: The maximum number of times to retry the 'off' command within the timeout periodSome devices do not support the standard commands or may provide additional ones.Use this to specify an alternate, device-specific, command that implements the 'on' action.Advanced use only: An alternate command to run instead of 'on'Some devices need much more/less time to complete than normal.Use this to specify an alternate, device-specific, timeout for 'on' actions.Advanced use only: Specify an alternate timeout to use for on actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'on' actions before giving up.Advanced use only: The maximum number of times to retry the 'on' command within the timeout periodSome devices do not support the standard commands or may provide additional ones.Use this to specify an alternate, device-specific, command that implements the 'list' action.Advanced use only: An alternate command to run instead of 'list'Some devices need much more/less time to complete than normal.Use this to specify an alternate, device-specific, timeout for 'list' actions.Advanced use only: Specify an alternate timeout to use for list actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'list' actions before giving up.Advanced use only: The maximum number of times to retry the 'list' command within the timeout periodSome devices do not support the standard commands or may provide additional ones.Use this to specify an alternate, device-specific, command that implements the 'monitor' action.Advanced use only: An alternate command to run instead of 'monitor'Some devices need much more/less time to complete than normal.\nUse this to specify an alternate, device-specific, timeout for 'monitor' actions.Advanced use only: Specify an alternate timeout to use for monitor actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'monitor' actions before giving up.Advanced use only: The maximum number of times to retry the 'monitor' command within the timeout periodSome devices do not support the standard commands or may provide additional ones.Use this to specify an alternate, device-specific, command that implements the 'status' action.Advanced use only: An alternate command to run instead of 'status'Some devices need much more/less time to complete than normal.Use this to specify an alternate, device-specific, timeout for 'status' actions.Advanced use only: Specify an alternate timeout to use for status actions instead of stonith-timeoutSome devices do not support multiple connections. Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining. Use this option to alter the number of times Pacemaker retries 'status' actions before giving up.Advanced use only: The maximum number of times to retry the 'status' command within the timeout period
=#=#=#= End test: Get fencer metadata - OK (0) =#=#=#=
* Passed: pacemaker-fenced - Get fencer metadata
=#=#=#= Begin test: Get scheduler metadata =#=#=#=
1.1Cluster options used by Pacemaker's schedulerPacemaker scheduler optionsWhat to do when the cluster does not have quorum Allowed values: stop, freeze, ignore, demote, suicideWhat to do when the cluster does not have quorumWhether resources can run on any node by defaultWhether resources can run on any node by defaultWhether the cluster should refrain from monitoring, starting, and stopping resourcesWhether the cluster should refrain from monitoring, starting, and stopping resourcesWhen true, the cluster will immediately ban a resource from a node if it fails to start there. When false, the cluster will instead check the resource's fail count against its migration-threshold.Whether a start failure should prevent a resource from being recovered on the same nodeWhether the cluster should check for active resources during start-upWhether the cluster should check for active resources during start-upWhen true, resources active on a node when it is cleanly shut down are kept "locked" to that node (not allowed to run elsewhere) until they start again on that node after it rejoins (or for at most shutdown-lock-limit, if set). Stonith resources and Pacemaker Remote connections are never locked. Clone and bundle instances and the promoted role of promotable clones are currently never locked, though support could be added in a future release.Whether to lock resources to a cleanly shut down nodeIf shutdown-lock is true and this is set to a nonzero time duration, shutdown locks will expire after this much time has passed since the shutdown was initiated, even if the node has not rejoined.Do not lock resources to a cleanly shut down node longer than thisIf false, unresponsive nodes are immediately assumed to be harmless, and resources that were active on them may be recovered elsewhere. This can result in a "split-brain" situation, potentially leading to data loss and/or service unavailability.*** Advanced Use Only *** Whether nodes may be fenced as part of recoveryAction to send to fence device when a node needs to be fenced ("poweroff" is a deprecated alias for "off") Allowed values: reboot, off, poweroffAction to send to fence device when a node needs to be fenced ("poweroff" is a deprecated alias for "off")This value is not used by Pacemaker, but is kept for backward compatibility, and certain legacy fence agents might use it.*** Advanced Use Only *** Unused by PacemakerThis is set automatically by the cluster according to whether SBD is detected to be in use. User-configured values are ignored. The value `true` is meaningful if diskless SBD is used and `stonith-watchdog-timeout` is nonzero. In that case, if fencing is required, watchdog-based self-fencing will be performed via SBD without requiring a fencing resource explicitly configured.Whether watchdog integration is enabledAllow performing fencing operations in parallelAllow performing fencing operations in parallelSetting this to false may lead to a "split-brain" situation,potentially leading to data loss and/or service unavailability.*** Advanced Use Only *** Whether to fence unseen nodes at start-upApply specified delay for the fencings that are targeting the lost nodes with the highest total resource priority in case we don't have the majority of the nodes in our cluster partition, so that the more significant nodes potentially win any fencing match, which is especially meaningful under split-brain of 2-node cluster. A promoted resource instance takes the base priority + 1 on calculation if the base priority is not 0. Any static/random delays that are introduced by `pcmk_delay_base/max` configured for the corresponding fencing resources will be added to this delay. This delay should be significantly greater than, safely twice, the maximum `pcmk_delay_base/max`. By default, priority fencing delay is disabled.Apply fencing delay targeting the lost nodes with the highest total resource priorityFence nodes that do not join the controller process group within this much time after joining the cluster, to allow the cluster to continue managing resources. A value of 0 means never fence pending nodes.How long to wait for a node that has joined the cluster to join the controller process groupThe node elected Designated Controller (DC) will consider an action failed if it does not get a response from the node executing the action within this time (after considering the action's own timeout). The "correct" value will depend on the speed and load of your network and cluster nodes.Maximum time for node-to-node communicationThe "correct" value will depend on the speed and load of your network and cluster nodes. If set to 0, the cluster will impose a dynamically calculated limit when any node has a high load.Maximum number of jobs that the cluster may execute in parallel across all nodesThe number of live migration actions that the cluster is allowed to execute in parallel on a node (-1 means no limit)The number of live migration actions that the cluster is allowed to execute in parallel on a node (-1 means no limit)Whether the cluster should stop all active resourcesWhether the cluster should stop all active resourcesWhether to stop resources that were removed from the configurationWhether to stop resources that were removed from the configurationWhether to cancel recurring actions removed from the configurationWhether to cancel recurring actions removed from the configurationValues other than default are poorly tested and potentially dangerous. This option will be removed in a future release.*** Deprecated *** Whether to remove stopped resources from the executorZero to disable, -1 to store unlimited.The number of scheduler inputs resulting in errors to saveZero to disable, -1 to store unlimited.The number of scheduler inputs resulting in warnings to saveZero to disable, -1 to store unlimited.The number of scheduler inputs without errors or warnings to saveRequires external entities to create node attributes (named with the prefix "#health") with values "red", "yellow", or "green". Allowed values: none, migrate-on-red, only-green, progressive, customHow cluster should react to node health attributesOnly used when "node-health-strategy" is set to "progressive".Base health score assigned to a nodeOnly used when "node-health-strategy" is set to "custom" or "progressive".The score to use for a node health attribute whose value is "green"Only used when "node-health-strategy" is set to "custom" or "progressive".The score to use for a node health attribute whose value is "yellow"Only used when "node-health-strategy" is set to "custom" or "progressive".The score to use for a node health attribute whose value is "red"How the cluster should allocate resources to nodes Allowed values: default, utilization, minimal, balancedHow the cluster should allocate resources to nodes
=#=#=#= End test: Get scheduler metadata - OK (0) =#=#=#=
* Passed: pacemaker-schedulerd - Get scheduler metadata
diff --git a/daemons/controld/controld_control.c b/daemons/controld/controld_control.c
index d5c97e2bff..540b3ff9f8 100644
--- a/daemons/controld/controld_control.c
+++ b/daemons/controld/controld_control.c
@@ -1,853 +1,863 @@
/*
* Copyright 2004-2023 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
static qb_ipcs_service_t *ipcs = NULL;
static crm_trigger_t *config_read_trigger = NULL;
#if SUPPORT_COROSYNC
extern gboolean crm_connect_corosync(crm_cluster_t * cluster);
#endif
void crm_shutdown(int nsig);
static gboolean crm_read_options(gpointer user_data);
/* A_HA_CONNECT */
void
do_ha_control(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
gboolean registered = FALSE;
static crm_cluster_t *cluster = NULL;
if (cluster == NULL) {
cluster = pcmk_cluster_new();
}
if (action & A_HA_DISCONNECT) {
crm_cluster_disconnect(cluster);
crm_info("Disconnected from the cluster");
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
}
if (action & A_HA_CONNECT) {
crm_set_status_callback(&peer_update_callback);
crm_set_autoreap(FALSE);
#if SUPPORT_COROSYNC
if (is_corosync_cluster()) {
registered = crm_connect_corosync(cluster);
}
#endif // SUPPORT_COROSYNC
if (registered) {
controld_election_init(cluster->uname);
controld_globals.our_nodename = cluster->uname;
controld_globals.our_uuid = cluster->uuid;
if(cluster->uuid == NULL) {
crm_err("Could not obtain local uuid");
registered = FALSE;
}
}
if (!registered) {
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
return;
}
populate_cib_nodes(node_update_none, __func__);
controld_clear_fsa_input_flags(R_HA_DISCONNECTED);
crm_info("Connected to the cluster");
}
if (action & ~(A_HA_CONNECT | A_HA_DISCONNECT)) {
crm_err("Unexpected action %s in %s", fsa_action2string(action),
__func__);
}
}
/* A_SHUTDOWN */
void
do_shutdown(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
/* just in case */
controld_set_fsa_input_flags(R_SHUTDOWN);
controld_disconnect_fencer(FALSE);
}
/* A_SHUTDOWN_REQ */
void
do_shutdown_req(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
xmlNode *msg = NULL;
controld_set_fsa_input_flags(R_SHUTDOWN);
//controld_set_fsa_input_flags(R_STAYDOWN);
crm_info("Sending shutdown request to all peers (DC is %s)",
pcmk__s(controld_globals.dc_name, "not set"));
msg = create_request(CRM_OP_SHUTDOWN_REQ, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
if (send_cluster_message(NULL, crm_msg_crmd, msg, TRUE) == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
free_xml(msg);
}
void
crmd_fast_exit(crm_exit_t exit_code)
{
if (pcmk_is_set(controld_globals.fsa_input_register, R_STAYDOWN)) {
crm_warn("Inhibiting respawn "CRM_XS" remapping exit code %d to %d",
exit_code, CRM_EX_FATAL);
exit_code = CRM_EX_FATAL;
} else if ((exit_code == CRM_EX_OK)
&& pcmk_is_set(controld_globals.fsa_input_register,
R_IN_RECOVERY)) {
crm_err("Could not recover from internal error");
exit_code = CRM_EX_ERROR;
}
if (controld_globals.logger_out != NULL) {
controld_globals.logger_out->finish(controld_globals.logger_out,
exit_code, true, NULL);
pcmk__output_free(controld_globals.logger_out);
controld_globals.logger_out = NULL;
}
crm_exit(exit_code);
}
crm_exit_t
crmd_exit(crm_exit_t exit_code)
{
GMainLoop *mloop = controld_globals.mainloop;
static bool in_progress = FALSE;
if (in_progress && (exit_code == CRM_EX_OK)) {
crm_debug("Exit is already in progress");
return exit_code;
} else if(in_progress) {
crm_notice("Error during shutdown process, exiting now with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
in_progress = TRUE;
crm_trace("Preparing to exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
/* Suppress secondary errors resulting from us disconnecting everything */
controld_set_fsa_input_flags(R_HA_DISCONNECTED);
/* Close all IPC servers and clients to ensure any and all shared memory files are cleaned up */
if(ipcs) {
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs);
ipcs = NULL;
}
controld_close_attrd_ipc();
controld_shutdown_schedulerd_ipc();
controld_disconnect_fencer(TRUE);
if ((exit_code == CRM_EX_OK) && (controld_globals.mainloop == NULL)) {
crm_debug("No mainloop detected");
exit_code = CRM_EX_ERROR;
}
/* On an error, just get out.
*
* Otherwise, make the effort to have mainloop exit gracefully so
* that it (mostly) cleans up after itself and valgrind has less
* to report on - allowing real errors stand out
*/
if (exit_code != CRM_EX_OK) {
crm_notice("Forcing immediate exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
crm_write_blackbox(SIGTRAP, NULL);
crmd_fast_exit(exit_code);
}
/* Clean up as much memory as possible for valgrind */
for (GList *iter = controld_globals.fsa_message_queue; iter != NULL;
iter = iter->next) {
fsa_data_t *fsa_data = (fsa_data_t *) iter->data;
crm_info("Dropping %s: [ state=%s cause=%s origin=%s ]",
fsa_input2string(fsa_data->fsa_input),
fsa_state2string(controld_globals.fsa_state),
fsa_cause2string(fsa_data->fsa_cause), fsa_data->origin);
delete_fsa_input(fsa_data);
}
controld_clear_fsa_input_flags(R_MEMBERSHIP);
g_list_free(controld_globals.fsa_message_queue);
controld_globals.fsa_message_queue = NULL;
controld_free_node_pending_timers();
controld_election_fini();
/* Tear down the CIB manager connection, but don't free it yet -- it could
* be used when we drain the mainloop later.
*/
controld_disconnect_cib_manager();
verify_stopped(controld_globals.fsa_state, LOG_WARNING);
controld_clear_fsa_input_flags(R_LRM_CONNECTED);
lrm_state_destroy_all();
mainloop_destroy_trigger(config_read_trigger);
config_read_trigger = NULL;
controld_destroy_fsa_trigger();
controld_destroy_transition_trigger();
pcmk__client_cleanup();
crm_peer_destroy();
controld_free_fsa_timers();
te_cleanup_stonith_history_sync(NULL, TRUE);
controld_free_sched_timer();
free(controld_globals.our_nodename);
controld_globals.our_nodename = NULL;
free(controld_globals.our_uuid);
controld_globals.our_uuid = NULL;
free(controld_globals.dc_name);
controld_globals.dc_name = NULL;
free(controld_globals.dc_version);
controld_globals.dc_version = NULL;
free(controld_globals.cluster_name);
controld_globals.cluster_name = NULL;
free(controld_globals.te_uuid);
controld_globals.te_uuid = NULL;
free_max_generation();
controld_destroy_failed_sync_table();
controld_destroy_outside_events_table();
mainloop_destroy_signal(SIGPIPE);
mainloop_destroy_signal(SIGUSR1);
mainloop_destroy_signal(SIGTERM);
mainloop_destroy_signal(SIGTRAP);
/* leave SIGCHLD engaged as we might still want to drain some service-actions */
if (mloop) {
GMainContext *ctx = g_main_loop_get_context(controld_globals.mainloop);
/* Don't re-enter this block */
controld_globals.mainloop = NULL;
/* no signals on final draining anymore */
mainloop_destroy_signal(SIGCHLD);
crm_trace("Draining mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
{
int lpc = 0;
while((g_main_context_pending(ctx) && lpc < 10)) {
lpc++;
crm_trace("Iteration %d", lpc);
g_main_context_dispatch(ctx);
}
}
crm_trace("Closing mainloop %d %d", g_main_loop_is_running(mloop), g_main_context_pending(ctx));
g_main_loop_quit(mloop);
/* Won't do anything yet, since we're inside it now */
g_main_loop_unref(mloop);
} else {
mainloop_destroy_signal(SIGCHLD);
}
cib_delete(controld_globals.cib_conn);
controld_globals.cib_conn = NULL;
throttle_fini();
/* Graceful */
crm_trace("Done preparing for exit with status %d (%s)",
exit_code, crm_exit_str(exit_code));
return exit_code;
}
/* A_EXIT_0, A_EXIT_1 */
void
do_exit(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_exit_t exit_code = CRM_EX_OK;
if (pcmk_is_set(action, A_EXIT_1)) {
exit_code = CRM_EX_ERROR;
crm_err("Exiting now due to errors");
}
verify_stopped(cur_state, LOG_ERR);
crmd_exit(exit_code);
}
static void sigpipe_ignore(int nsig) { return; }
/* A_STARTUP */
void
do_startup(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_debug("Registering Signal Handlers");
mainloop_add_signal(SIGTERM, crm_shutdown);
mainloop_add_signal(SIGPIPE, sigpipe_ignore);
config_read_trigger = mainloop_add_trigger(G_PRIORITY_HIGH,
crm_read_options, NULL);
controld_init_fsa_trigger();
controld_init_transition_trigger();
crm_debug("Creating CIB manager and executor objects");
controld_globals.cib_conn = cib_new();
lrm_state_init_local();
if (controld_init_fsa_timers() == FALSE) {
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
}
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
accept_controller_client(qb_ipcs_connection_t *c, uid_t uid, gid_t gid)
{
crm_trace("Accepting new IPC client connection");
if (pcmk__new_client(c, uid, gid) == NULL) {
return -EIO;
}
return 0;
}
// \return libqb error code (0 on success, -errno on error)
static int32_t
dispatch_controller_ipc(qb_ipcs_connection_t * c, void *data, size_t size)
{
uint32_t id = 0;
uint32_t flags = 0;
pcmk__client_t *client = pcmk__find_client(c);
xmlNode *msg = pcmk__client_data2xml(client, data, &id, &flags);
if (msg == NULL) {
pcmk__ipc_send_ack(client, id, flags, "ack", NULL, CRM_EX_PROTOCOL);
return 0;
}
pcmk__ipc_send_ack(client, id, flags, "ack", NULL, CRM_EX_INDETERMINATE);
CRM_ASSERT(client->user != NULL);
pcmk__update_acl_user(msg, F_CRM_USER, client->user);
crm_xml_add(msg, F_CRM_SYS_FROM, client->id);
if (controld_authorize_ipc_message(msg, client, NULL)) {
crm_trace("Processing IPC message from client %s",
pcmk__client_name(client));
route_message(C_IPC_MESSAGE, msg);
}
controld_trigger_fsa();
free_xml(msg);
return 0;
}
static int32_t
ipc_client_disconnected(qb_ipcs_connection_t *c)
{
pcmk__client_t *client = pcmk__find_client(c);
if (client) {
crm_trace("Disconnecting %sregistered client %s (%p/%p)",
(client->userdata? "" : "un"), pcmk__client_name(client),
c, client);
free(client->userdata);
pcmk__free_client(client);
controld_trigger_fsa();
}
return 0;
}
static void
ipc_connection_destroyed(qb_ipcs_connection_t *c)
{
crm_trace("Connection %p", c);
ipc_client_disconnected(c);
}
/* A_STOP */
void
do_stop(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
crm_trace("Closing IPC server");
mainloop_del_ipc_server(ipcs); ipcs = NULL;
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
/* A_STARTED */
void
do_started(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
static struct qb_ipcs_service_handlers crmd_callbacks = {
.connection_accept = accept_controller_client,
.connection_created = NULL,
.msg_process = dispatch_controller_ipc,
.connection_closed = ipc_client_disconnected,
.connection_destroyed = ipc_connection_destroyed
};
if (cur_state != S_STARTING) {
crm_err("Start cancelled... %s", fsa_state2string(cur_state));
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_MEMBERSHIP)) {
crm_info("Delaying start, no membership data (%.16llx)", R_MEMBERSHIP);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_LRM_CONNECTED)) {
crm_info("Delaying start, not connected to executor (%.16llx)", R_LRM_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_CIB_CONNECTED)) {
crm_info("Delaying start, CIB not connected (%.16llx)", R_CIB_CONNECTED);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register,
R_READ_CONFIG)) {
crm_info("Delaying start, Config not read (%.16llx)", R_READ_CONFIG);
crmd_fsa_stall(TRUE);
return;
} else if (!pcmk_is_set(controld_globals.fsa_input_register, R_PEER_DATA)) {
crm_info("Delaying start, No peer data (%.16llx)", R_PEER_DATA);
crmd_fsa_stall(TRUE);
return;
}
crm_debug("Init server comms");
ipcs = pcmk__serve_controld_ipc(&crmd_callbacks);
if (ipcs == NULL) {
crm_err("Failed to create IPC server: shutting down and inhibiting respawn");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
} else {
crm_notice("Pacemaker controller successfully started and accepting connections");
}
controld_set_fsa_input_flags(R_ST_REQUIRED);
controld_timer_fencer_connect(GINT_TO_POINTER(TRUE));
controld_clear_fsa_input_flags(R_STARTING);
register_fsa_input(msg_data->fsa_cause, I_PENDING, NULL);
}
/* A_RECOVER */
void
do_recover(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
controld_set_fsa_input_flags(R_IN_RECOVERY);
crm_warn("Fast-tracking shutdown in response to errors");
register_fsa_input(C_FSA_INTERNAL, I_TERMINATE, NULL);
}
static pcmk__cluster_option_t controller_options[] = {
/* name, old name, type, allowed values,
* default value, validator,
* short description,
* long description
*/
{
"dc-version", NULL, "string", NULL, PCMK__VALUE_NONE, NULL,
N_("Pacemaker version on cluster node elected Designated Controller (DC)"),
N_("Includes a hash which identifies the exact changeset the code was "
"built from. Used for diagnostic purposes.")
},
{
"cluster-infrastructure", NULL, "string", NULL, "corosync", NULL,
N_("The messaging stack on which Pacemaker is currently running"),
N_("Used for informational and diagnostic purposes.")
},
{
"cluster-name", NULL, "string", NULL, NULL, NULL,
N_("An arbitrary name for the cluster"),
N_("This optional value is mostly for users' convenience as desired "
"in administration, but may also be used in Pacemaker "
"configuration rules via the #cluster-name node attribute, and "
"by higher-level tools and resource agents.")
},
{
XML_CONFIG_ATTR_DC_DEADTIME, NULL, "time",
NULL, "20s", pcmk__valid_interval_spec,
N_("How long to wait for a response from other nodes during start-up"),
N_("The optimal value will depend on the speed and load of your network "
"and the type of switches used.")
},
{
XML_CONFIG_ATTR_RECHECK, NULL, "time",
N_("Zero disables polling, while positive values are an interval in seconds"
"(unless other units are specified, for example \"5min\")"),
"15min", pcmk__valid_interval_spec,
N_("Polling interval to recheck cluster state and evaluate rules "
"with date specifications"),
N_("Pacemaker is primarily event-driven, and looks ahead to know when to "
"recheck cluster state for failure timeouts and most time-based "
"rules. However, it will also recheck the cluster after this "
"amount of inactivity, to evaluate rules with date specifications "
"and serve as a fail-safe for certain types of scheduler bugs.")
},
{
"load-threshold", NULL, "percentage", NULL,
"80%", pcmk__valid_percentage,
N_("Maximum amount of system load that should be used by cluster nodes"),
N_("The cluster will slow down its recovery process when the amount of "
"system resources used (currently CPU) approaches this limit"),
},
{
"node-action-limit", NULL, "integer", NULL,
"0", pcmk__valid_number,
N_("Maximum number of jobs that can be scheduled per node "
"(defaults to 2x cores)")
},
{ XML_CONFIG_ATTR_FENCE_REACTION, NULL, "string", NULL, "stop", NULL,
N_("How a cluster node should react if notified of its own fencing"),
N_("A cluster node may receive notification of its own fencing if fencing "
"is misconfigured, or if fabric fencing is in use that doesn't cut "
"cluster communication. Allowed values are \"stop\" to attempt to "
"immediately stop Pacemaker and stay stopped, or \"panic\" to attempt "
"to immediately reboot the local node, falling back to stop on failure.")
},
{
XML_CONFIG_ATTR_ELECTION_FAIL, NULL, "time", NULL,
"2min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
N_("Declare an election failed if it is not decided within this much "
"time. If you need to adjust this value, it probably indicates "
"the presence of a bug.")
},
{
XML_CONFIG_ATTR_FORCE_QUIT, NULL, "time", NULL,
"20min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
N_("Exit immediately if shutdown does not complete within this much "
"time. If you need to adjust this value, it probably indicates "
"the presence of a bug.")
},
{
"join-integration-timeout", "crmd-integration-timeout", "time", NULL,
"3min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
N_("If you need to adjust this value, it probably indicates "
"the presence of a bug.")
},
{
"join-finalization-timeout", "crmd-finalization-timeout", "time", NULL,
"30min", pcmk__valid_interval_spec,
"*** Advanced Use Only ***",
N_("If you need to adjust this value, it probably indicates "
"the presence of a bug.")
},
{
"transition-delay", "crmd-transition-delay", "time", NULL,
"0s", pcmk__valid_interval_spec,
N_("*** Advanced Use Only *** Enabling this option will slow down "
"cluster recovery under all conditions"),
N_("Delay cluster recovery for this much time to allow for additional "
"events to occur. Useful if your configuration is sensitive to "
"the order in which ping updates arrive.")
},
{
"stonith-watchdog-timeout", NULL, "time", NULL,
"0", controld_verify_stonith_watchdog_timeout,
N_("How long before nodes can be assumed to be safely down when "
"watchdog-based self-fencing via SBD is in use"),
N_("If this is set to a positive value, lost nodes are assumed to "
"self-fence using watchdog-based SBD within this much time. This "
"does not require a fencing resource to be explicitly configured, "
"though a fence_watchdog resource can be configured, to limit use "
"to specific nodes. If this is set to 0 (the default), the cluster "
"will never assume watchdog-based self-fencing. If this is set to a "
"negative value, the cluster will use twice the local value of the "
"`SBD_WATCHDOG_TIMEOUT` environment variable if that is positive, "
"or otherwise treat this as 0. WARNING: When used, this timeout "
"must be larger than `SBD_WATCHDOG_TIMEOUT` on all nodes that use "
"watchdog-based SBD, and Pacemaker will refuse to start on any of "
"those nodes where this is not true for the local value or SBD is "
"not active. When this is set to a negative value, "
"`SBD_WATCHDOG_TIMEOUT` must be set to the same value on all nodes "
"that use SBD, otherwise data corruption or loss could occur.")
},
{
"stonith-max-attempts", NULL, "integer", NULL,
"10", pcmk__valid_positive_number,
N_("How many times fencing can fail before it will no longer be "
"immediately re-attempted on a target")
},
// Already documented in libpe_status (other values must be kept identical)
{
"no-quorum-policy", NULL, "select",
"stop, freeze, ignore, demote, suicide", "stop", pcmk__valid_quorum,
N_("What to do when the cluster does not have quorum"), NULL
},
{
XML_CONFIG_ATTR_SHUTDOWN_LOCK, NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
N_("Whether to lock resources to a cleanly shut down node"),
N_("When true, resources active on a node when it is cleanly shut down "
"are kept \"locked\" to that node (not allowed to run elsewhere) "
"until they start again on that node after it rejoins (or for at "
"most shutdown-lock-limit, if set). Stonith resources and "
"Pacemaker Remote connections are never locked. Clone and bundle "
"instances and the promoted role of promotable clones are "
"currently never locked, though support could be added in a future "
"release.")
},
{
XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT, NULL, "time", NULL,
"0", pcmk__valid_interval_spec,
N_("Do not lock resources to a cleanly shut down node longer than "
"this"),
N_("If shutdown-lock is true and this is set to a nonzero time "
"duration, shutdown locks will expire after this much time has "
"passed since the shutdown was initiated, even if the node has not "
"rejoined.")
},
+ {
+ XML_CONFIG_ATTR_NODE_PENDING_TIMEOUT, NULL, "time", NULL,
+ "2h", pcmk__valid_interval_spec,
+ N_("How long to wait for a node that has joined the cluster to join "
+ "the controller process group"),
+ N_("Fence nodes that do not join the controller process group within "
+ "this much time after joining the cluster, to allow the cluster "
+ "to continue managing resources. A value of 0 means never fence "
+ "pending nodes.")
+ },
};
void
crmd_metadata(void)
{
const char *desc_short = "Pacemaker controller options";
const char *desc_long = "Cluster options used by Pacemaker's controller";
gchar *s = pcmk__format_option_metadata("pacemaker-controld", desc_short,
desc_long, controller_options,
PCMK__NELEM(controller_options));
printf("%s", s);
g_free(s);
}
static void
config_query_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
const char *value = NULL;
GHashTable *config_hash = NULL;
crm_time_t *now = crm_time_new(NULL);
xmlNode *crmconfig = NULL;
xmlNode *alerts = NULL;
if (rc != pcmk_ok) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query resulted in an error: %s", pcmk_strerror(rc));
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
if (rc == -EACCES || rc == -pcmk_err_schema_validation) {
crm_err("The cluster is mis-configured - shutting down and staying down");
controld_set_fsa_input_flags(R_STAYDOWN);
}
goto bail;
}
crmconfig = output;
if ((crmconfig != NULL)
&& !pcmk__xe_is(crmconfig, XML_CIB_TAG_CRMCONFIG)) {
crmconfig = first_named_child(crmconfig, XML_CIB_TAG_CRMCONFIG);
}
if (!crmconfig) {
fsa_data_t *msg_data = NULL;
crm_err("Local CIB query for " XML_CIB_TAG_CRMCONFIG " section failed");
register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL);
goto bail;
}
crm_debug("Call %d : Parsing CIB options", call_id);
config_hash = pcmk__strkey_table(free, free);
pe_unpack_nvpairs(crmconfig, crmconfig, XML_CIB_TAG_PROPSET, NULL,
config_hash, CIB_OPTIONS_FIRST, FALSE, now, NULL);
// Validate all options, and use defaults if not already present in hash
pcmk__validate_cluster_options(config_hash, controller_options,
PCMK__NELEM(controller_options));
value = g_hash_table_lookup(config_hash, "no-quorum-policy");
if (pcmk__str_eq(value, "suicide", pcmk__str_casei) && pcmk__locate_sbd()) {
controld_set_global_flags(controld_no_quorum_suicide);
}
value = g_hash_table_lookup(config_hash, XML_CONFIG_ATTR_SHUTDOWN_LOCK);
if (crm_is_true(value)) {
controld_set_global_flags(controld_shutdown_lock_enabled);
} else {
controld_clear_global_flags(controld_shutdown_lock_enabled);
}
value = g_hash_table_lookup(config_hash,
XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT);
controld_globals.shutdown_lock_limit = crm_parse_interval_spec(value)
/ 1000;
value = g_hash_table_lookup(config_hash,
XML_CONFIG_ATTR_NODE_PENDING_TIMEOUT);
controld_globals.node_pending_timeout = crm_parse_interval_spec(value) / 1000;
value = g_hash_table_lookup(config_hash, "cluster-name");
pcmk__str_update(&(controld_globals.cluster_name), value);
// Let subcomponents initialize their own static variables
controld_configure_election(config_hash);
controld_configure_fencing(config_hash);
controld_configure_fsa_timers(config_hash);
controld_configure_throttle(config_hash);
alerts = first_named_child(output, XML_CIB_TAG_ALERTS);
crmd_unpack_alerts(alerts);
controld_set_fsa_input_flags(R_READ_CONFIG);
controld_trigger_fsa();
g_hash_table_destroy(config_hash);
bail:
crm_time_free(now);
}
/*!
* \internal
* \brief Trigger read and processing of the configuration
*
* \param[in] fn Calling function name
* \param[in] line Line number where call occurred
*/
void
controld_trigger_config_as(const char *fn, int line)
{
if (config_read_trigger != NULL) {
crm_trace("%s:%d - Triggered config processing", fn, line);
mainloop_set_trigger(config_read_trigger);
}
}
gboolean
crm_read_options(gpointer user_data)
{
cib_t *cib_conn = controld_globals.cib_conn;
int call_id = cib_conn->cmds->query(cib_conn,
"//" XML_CIB_TAG_CRMCONFIG
" | //" XML_CIB_TAG_ALERTS,
NULL, cib_xpath|cib_scope_local);
fsa_register_cib_callback(call_id, NULL, config_query_callback);
crm_trace("Querying the CIB... call %d", call_id);
return TRUE;
}
/* A_READCONFIG */
void
do_read_config(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
throttle_init();
controld_trigger_config();
}
void
crm_shutdown(int nsig)
{
const char *value = NULL;
guint default_period_ms = 0;
if ((controld_globals.mainloop == NULL)
|| !g_main_loop_is_running(controld_globals.mainloop)) {
crmd_exit(CRM_EX_OK);
return;
}
if (pcmk_is_set(controld_globals.fsa_input_register, R_SHUTDOWN)) {
crm_err("Escalating shutdown");
register_fsa_input_before(C_SHUTDOWN, I_ERROR, NULL);
return;
}
controld_set_fsa_input_flags(R_SHUTDOWN);
register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL);
/* If shutdown timer doesn't have a period set, use the default
*
* @TODO: Evaluate whether this is still necessary. As long as
* config_query_callback() has been run at least once, it doesn't look like
* anything could have changed the timer period since then.
*/
value = pcmk__cluster_option(NULL, controller_options,
PCMK__NELEM(controller_options),
XML_CONFIG_ATTR_FORCE_QUIT);
default_period_ms = crm_parse_interval_spec(value);
controld_shutdown_start_countdown(default_period_ms);
}