*/
#include "includes.h"
-#include "lib/events/events.h"
+#include "lib/tevent/tevent.h"
#include "system/filesys.h"
#include "system/wait.h"
#include "../include/ctdb_private.h"
return ret;
}
-static void ctdb_run_notification_script(struct ctdb_context *ctdb, const char *event)
+void ctdb_run_notification_script(struct ctdb_context *ctdb, const char *event)
{
pid_t child;
if (child == 0) {
int ret;
+ debug_extra = talloc_asprintf(NULL, "notification-%s:", event);
ret = ctdb_run_notification_script_child(ctdb, event);
if (ret != 0) {
DEBUG(DEBUG_ERR,(__location__ " Notification script failed\n"));
TDB_DATA data;
struct ctdb_node_flag_change c;
uint32_t next_interval;
+ int ret;
+ TDB_DATA rddata;
+ struct takeover_run_reply rd;
c.pnn = ctdb->pnn;
c.old_flags = node->flags;
+ rd.pnn = ctdb->pnn;
+ rd.srvid = CTDB_SRVID_TAKEOVER_RUN_RESPONSE;
+
+ rddata.dptr = (uint8_t *)&rd;
+ rddata.dsize = sizeof(rd);
+
+ if (status == -ETIME) {
+ ctdb->event_script_timeouts++;
+
+ if (ctdb->event_script_timeouts >= ctdb->tunable.script_timeout_count) {
+ DEBUG(DEBUG_ERR, ("Maximum timeout count %u reached for eventscript. Making node unhealthy\n", ctdb->tunable.script_timeout_count));
+ } else {
+ /* We pretend this is OK. */
+ goto after_change_status;
+ }
+ }
+
if (status != 0 && !(node->flags & NODE_FLAGS_UNHEALTHY)) {
DEBUG(DEBUG_NOTICE,("monitor event failed - disabling node\n"));
node->flags |= NODE_FLAGS_UNHEALTHY;
- ctdb->monitor->next_interval = 1;
- if (ctdb->tunable.disable_when_unhealthy != 0) {
- DEBUG(DEBUG_INFO, ("DISABLING node since it became unhealthy\n"));
- node->flags |= NODE_FLAGS_DISABLED;
- }
+ ctdb->monitor->next_interval = 5;
ctdb_run_notification_script(ctdb, "unhealthy");
+
+ /* ask the recmaster to reallocate all addresses */
+ DEBUG(DEBUG_ERR,("Node became UNHEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
+ ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
+ }
+
} else if (status == 0 && (node->flags & NODE_FLAGS_UNHEALTHY)) {
DEBUG(DEBUG_NOTICE,("monitor event OK - node re-enabled\n"));
node->flags &= ~NODE_FLAGS_UNHEALTHY;
- ctdb->monitor->next_interval = 1;
+ ctdb->monitor->next_interval = 5;
ctdb_run_notification_script(ctdb, "healthy");
+
+ /* ask the recmaster to reallocate all addresses */
+ DEBUG(DEBUG_ERR,("Node became HEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
+ ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
+ }
+
}
+after_change_status:
next_interval = ctdb->monitor->next_interval;
ctdb->monitor->next_interval *= 2;
} else if (status == 0) {
DEBUG(DEBUG_NOTICE,("startup event OK - enabling monitoring\n"));
ctdb->done_startup = true;
- ctdb->monitor->next_interval = 1;
+ ctdb->monitor->next_interval = 2;
ctdb_run_notification_script(ctdb, "startup");
}
}
+/*
+ wait until we have finished initial recoveries before we start the
+ monitoring events
+ */
+static void ctdb_wait_until_recovered(struct event_context *ev, struct timed_event *te,
+ struct timeval t, void *private_data)
+{
+ struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
+ int ret;
+
+ DEBUG(DEBUG_NOTICE,("CTDB_WAIT_UNTIL_RECOVERED\n"));
+
+ if (ctdb->vnn_map->generation == INVALID_GENERATION) {
+ ctdb->db_persistent_startup_generation = INVALID_GENERATION;
+
+ DEBUG(DEBUG_NOTICE,(__location__ " generation is INVALID. Wait one more second\n"));
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current_ofs(1, 0),
+ ctdb_wait_until_recovered, ctdb);
+ return;
+ }
+
+ if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
+ ctdb->db_persistent_startup_generation = INVALID_GENERATION;
+
+ DEBUG(DEBUG_NOTICE,(__location__ " in recovery. Wait one more second\n"));
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current_ofs(1, 0),
+ ctdb_wait_until_recovered, ctdb);
+ return;
+ }
+
+
+ if (!fast_start && timeval_elapsed(&ctdb->last_recovery_finished) < (ctdb->tunable.rerecovery_timeout + 3)) {
+ ctdb->db_persistent_startup_generation = INVALID_GENERATION;
+
+ DEBUG(DEBUG_NOTICE,(__location__ " wait for pending recoveries to end. Wait one more second.\n"));
+
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current_ofs(1, 0),
+ ctdb_wait_until_recovered, ctdb);
+ return;
+ }
+
+ if (ctdb->vnn_map->generation == ctdb->db_persistent_startup_generation) {
+ DEBUG(DEBUG_INFO,(__location__ " skip ctdb_recheck_persistent_health() "
+ "until the next recovery\n"));
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current_ofs(1, 0),
+ ctdb_wait_until_recovered, ctdb);
+ return;
+ }
+
+ ctdb->db_persistent_startup_generation = ctdb->vnn_map->generation;
+ ret = ctdb_recheck_persistent_health(ctdb);
+ if (ret != 0) {
+ ctdb->db_persistent_check_errors++;
+ if (ctdb->db_persistent_check_errors < ctdb->max_persistent_check_errors) {
+ DEBUG(ctdb->db_persistent_check_errors==1?DEBUG_ERR:DEBUG_WARNING,
+ (__location__ "ctdb_recheck_persistent_health() "
+ "failed (%llu of %llu times) - retry later\n",
+ (unsigned long long)ctdb->db_persistent_check_errors,
+ (unsigned long long)ctdb->max_persistent_check_errors));
+ event_add_timed(ctdb->ev,
+ ctdb->monitor->monitor_context,
+ timeval_current_ofs(1, 0),
+ ctdb_wait_until_recovered, ctdb);
+ return;
+ }
+ DEBUG(DEBUG_ALERT,(__location__
+ "ctdb_recheck_persistent_health() failed (%llu times) - prepare shutdown\n",
+ (unsigned long long)ctdb->db_persistent_check_errors));
+ ctdb_stop_recoverd(ctdb);
+ ctdb_stop_keepalive(ctdb);
+ ctdb_stop_monitoring(ctdb);
+ ctdb_release_all_ips(ctdb);
+ if (ctdb->methods != NULL) {
+ ctdb->methods->shutdown(ctdb);
+ }
+ ctdb_event_script(ctdb, CTDB_EVENT_SHUTDOWN);
+ DEBUG(DEBUG_ALERT,("ctdb_recheck_persistent_health() failed - Stopping CTDB daemon\n"));
+ exit(11);
+ }
+ ctdb->db_persistent_check_errors = 0;
+ DEBUG(DEBUG_NOTICE,(__location__
+ "ctdb_start_monitoring: ctdb_recheck_persistent_health() OK\n"));
+
+ DEBUG(DEBUG_NOTICE,(__location__ " Recoveries finished. Running the \"startup\" event.\n"));
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current(),
+ ctdb_check_health, ctdb);
+}
+
+
/*
see if the event scripts think we are healthy
*/
struct timeval t, void *private_data)
{
struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
- int ret;
+ int ret = 0;
if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
(ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup)) {
if (!ctdb->done_startup) {
ret = ctdb_event_script_callback(ctdb,
- timeval_current_ofs(ctdb->tunable.script_timeout, 0),
ctdb->monitor->monitor_context, ctdb_startup_callback,
- ctdb, "startup");
+ ctdb, false,
+ CTDB_EVENT_STARTUP, "%s", "");
} else {
- ret = ctdb_event_script_callback(ctdb,
- timeval_current_ofs(ctdb->tunable.script_timeout, 0),
- ctdb->monitor->monitor_context, ctdb_health_callback,
- ctdb, "monitor");
+ int i;
+ int skip_monitoring = 0;
+
+ if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
+ skip_monitoring = 1;
+ DEBUG(DEBUG_ERR,("Skip monitoring during recovery\n"));
+ }
+ for (i=1; i<=NUM_DB_PRIORITIES; i++) {
+ if (ctdb->freeze_handles[i] != NULL) {
+ DEBUG(DEBUG_ERR,("Skip monitoring since databases are frozen\n"));
+ skip_monitoring = 1;
+ break;
+ }
+ }
+ if (skip_monitoring != 0) {
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ timeval_current_ofs(ctdb->monitor->next_interval, 0),
+ ctdb_check_health, ctdb);
+ return;
+ } else {
+ ret = ctdb_event_script_callback(ctdb,
+ ctdb->monitor->monitor_context, ctdb_health_callback,
+ ctdb, false,
+ CTDB_EVENT_MONITOR, "%s", "");
+ }
}
if (ret != 0) {
DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
- ctdb->monitor->next_interval = 1;
+ ctdb->monitor->next_interval = 5;
event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
- timeval_current_ofs(1, 0),
- ctdb_check_health, ctdb);
- }
+ timeval_current_ofs(5, 0),
+ ctdb_check_health, ctdb);
+ }
}
/*
void ctdb_enable_monitoring(struct ctdb_context *ctdb)
{
ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
- ctdb->monitor->next_interval = 2;
+ ctdb->monitor->next_interval = 5;
DEBUG(DEBUG_INFO,("Monitoring has been enabled\n"));
}
ctdb->monitor->monitor_context = NULL;
ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
- ctdb->monitor->next_interval = 1;
+ ctdb->monitor->next_interval = 5;
DEBUG(DEBUG_NOTICE,("Monitoring has been stopped\n"));
}
*/
void ctdb_start_monitoring(struct ctdb_context *ctdb)
{
- struct timed_event *te;
-
if (ctdb->monitor != NULL) {
return;
}
ctdb->monitor = talloc(ctdb, struct ctdb_monitor_state);
CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor);
- ctdb->monitor->next_interval = 1;
+ ctdb->monitor->next_interval = 5;
ctdb->monitor->monitor_context = talloc_new(ctdb->monitor);
CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor->monitor_context);
- te = event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
+ event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
timeval_current_ofs(1, 0),
- ctdb_check_health, ctdb);
- CTDB_NO_MEMORY_FATAL(ctdb, te);
+ ctdb_wait_until_recovered, ctdb);
ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
DEBUG(DEBUG_NOTICE,("Monitoring has been started\n"));
node = ctdb->nodes[c->pnn];
old_flags = node->flags;
- c->old_flags = node->flags;
+ if (c->pnn != ctdb->pnn) {
+ c->old_flags = node->flags;
+ }
node->flags = c->new_flags & ~NODE_FLAGS_DISCONNECTED;
node->flags |= (c->old_flags & NODE_FLAGS_DISCONNECTED);
DEBUG(DEBUG_INFO, ("Control modflags on node %u - flags now 0x%x\n", c->pnn, node->flags));
+ if (node->flags == 0 && !ctdb->done_startup) {
+ DEBUG(DEBUG_ERR, (__location__ " Node %u became healthy - force recovery for startup\n",
+ c->pnn));
+ ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
+ }
/* tell the recovery daemon something has changed */
ctdb_daemon_send_message(ctdb, ctdb->pnn,