2 monitoring links to all other nodes to detect dead nodes
5 Copyright (C) Ronnie Sahlberg 2007
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, see <http://www.gnu.org/licenses/>.
22 #include "lib/events/events.h"
23 #include "system/filesys.h"
24 #include "system/wait.h"
25 #include "../include/ctdb_private.h"
27 struct ctdb_monitor_state {
28 uint32_t monitoring_mode;
29 TALLOC_CTX *monitor_context;
30 uint32_t next_interval;
33 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
34 struct timeval t, void *private_data);
37 setup the notification script
39 int ctdb_set_notification_script(struct ctdb_context *ctdb, const char *script)
41 ctdb->notification_script = talloc_strdup(ctdb, script);
42 CTDB_NO_MEMORY(ctdb, ctdb->notification_script);
46 static int ctdb_run_notification_script_child(struct ctdb_context *ctdb, const char *event)
52 if (stat(ctdb->notification_script, &st) != 0) {
53 DEBUG(DEBUG_ERR,("Could not stat notification script %s. Can not send notifications.\n", ctdb->notification_script));
56 if (!(st.st_mode & S_IXUSR)) {
57 DEBUG(DEBUG_ERR,("Notification script %s is not executable.\n", ctdb->notification_script));
61 cmd = talloc_asprintf(ctdb, "%s %s\n", ctdb->notification_script, event);
62 CTDB_NO_MEMORY(ctdb, cmd);
65 /* if the system() call was successful, translate ret into the
66 return code from the command
69 ret = WEXITSTATUS(ret);
72 DEBUG(DEBUG_ERR,("Notification script \"%s\" failed with error %d\n", cmd, ret));
78 static void ctdb_run_notification_script(struct ctdb_context *ctdb, const char *event)
82 if (ctdb->notification_script == NULL) {
87 if (child == (pid_t)-1) {
88 DEBUG(DEBUG_ERR,("Failed to fork() a notification child process\n"));
94 debug_extra = talloc_asprintf(NULL, "notification-%s:", event);
95 ret = ctdb_run_notification_script_child(ctdb, event);
97 DEBUG(DEBUG_ERR,(__location__ " Notification script failed\n"));
106 called when a health monitoring event script finishes
108 static void ctdb_health_callback(struct ctdb_context *ctdb, int status, void *p)
110 struct ctdb_node *node = ctdb->nodes[ctdb->pnn];
112 struct ctdb_node_flag_change c;
113 uint32_t next_interval;
116 struct takeover_run_reply rd;
117 const char *state_str = NULL;
120 c.old_flags = node->flags;
123 rd.srvid = CTDB_SRVID_TAKEOVER_RUN_RESPONSE;
125 rddata.dptr = (uint8_t *)&rd;
126 rddata.dsize = sizeof(rd);
128 if (status == -ETIME) {
129 ctdb->event_script_timeouts++;
131 if (ctdb->event_script_timeouts >= ctdb->tunable.script_timeout_count) {
132 DEBUG(DEBUG_ERR, ("Maximum timeout count %u reached for eventscript. Making node unhealthy\n", ctdb->tunable.script_timeout_count));
134 /* We pretend this is OK. */
135 goto after_change_status;
139 if (status != 0 && !(node->flags & NODE_FLAGS_UNHEALTHY)) {
140 DEBUG(DEBUG_NOTICE,("monitor event failed - disabling node\n"));
141 node->flags |= NODE_FLAGS_UNHEALTHY;
142 ctdb->monitor->next_interval = 5;
144 ctdb_run_notification_script(ctdb, "unhealthy");
145 } else if (status == 0 && (node->flags & NODE_FLAGS_UNHEALTHY)) {
146 DEBUG(DEBUG_NOTICE,("monitor event OK - node re-enabled\n"));
147 node->flags &= ~NODE_FLAGS_UNHEALTHY;
148 ctdb->monitor->next_interval = 5;
150 ctdb_run_notification_script(ctdb, "healthy");
154 next_interval = ctdb->monitor->next_interval;
156 ctdb->monitor->next_interval *= 2;
157 if (ctdb->monitor->next_interval > ctdb->tunable.monitor_interval) {
158 ctdb->monitor->next_interval = ctdb->tunable.monitor_interval;
161 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
162 timeval_current_ofs(next_interval, 0),
163 ctdb_check_health, ctdb);
165 if (c.old_flags == node->flags) {
169 c.new_flags = node->flags;
171 data.dptr = (uint8_t *)&c;
172 data.dsize = sizeof(c);
174 /* ask the recovery daemon to push these changes out to all nodes */
175 ctdb_daemon_send_message(ctdb, ctdb->pnn,
176 CTDB_SRVID_PUSH_NODE_FLAGS, data);
178 if (c.new_flags & NODE_FLAGS_UNHEALTHY) {
179 state_str = "UNHEALTHY";
181 state_str = "HEALTHY";
184 /* ask the recmaster to reallocate all addresses */
185 DEBUG(DEBUG_ERR,("Node became %s. Ask recovery master %u to perform ip reallocation\n",
186 state_str, ctdb->recovery_master));
187 ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
189 DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
195 called when the startup event script finishes
197 static void ctdb_startup_callback(struct ctdb_context *ctdb, int status, void *p)
200 DEBUG(DEBUG_ERR,("startup event failed\n"));
201 } else if (status == 0) {
202 DEBUG(DEBUG_NOTICE,("startup event OK - enabling monitoring\n"));
203 ctdb->done_startup = true;
204 ctdb->monitor->next_interval = 5;
205 ctdb_run_notification_script(ctdb, "startup");
208 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
209 timeval_current_ofs(ctdb->monitor->next_interval, 0),
210 ctdb_check_health, ctdb);
215 wait until we have finished initial recoveries before we start the
218 static void ctdb_wait_until_recovered(struct event_context *ev, struct timed_event *te,
219 struct timeval t, void *private_data)
221 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
224 DEBUG(DEBUG_NOTICE,("CTDB_WAIT_UNTIL_RECOVERED\n"));
226 if (ctdb->vnn_map->generation == INVALID_GENERATION) {
227 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
229 DEBUG(DEBUG_NOTICE,(__location__ " generation is INVALID. Wait one more second\n"));
230 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
231 timeval_current_ofs(1, 0),
232 ctdb_wait_until_recovered, ctdb);
236 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
237 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
239 DEBUG(DEBUG_NOTICE,(__location__ " in recovery. Wait one more second\n"));
240 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
241 timeval_current_ofs(1, 0),
242 ctdb_wait_until_recovered, ctdb);
247 if (timeval_elapsed(&ctdb->last_recovery_finished) < (ctdb->tunable.rerecovery_timeout + 3)) {
248 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
250 DEBUG(DEBUG_NOTICE,(__location__ " wait for pending recoveries to end. Wait one more second.\n"));
252 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
253 timeval_current_ofs(1, 0),
254 ctdb_wait_until_recovered, ctdb);
258 if (ctdb->vnn_map->generation == ctdb->db_persistent_startup_generation) {
259 DEBUG(DEBUG_INFO,(__location__ " skip ctdb_recheck_persistent_health() "
260 "until the next recovery\n"));
261 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
262 timeval_current_ofs(1, 0),
263 ctdb_wait_until_recovered, ctdb);
267 ctdb->db_persistent_startup_generation = ctdb->vnn_map->generation;
268 ret = ctdb_recheck_persistent_health(ctdb);
270 ctdb->db_persistent_check_errors++;
271 if (ctdb->db_persistent_check_errors < ctdb->max_persistent_check_errors) {
272 DEBUG(ctdb->db_persistent_check_errors==1?DEBUG_ERR:DEBUG_WARNING,
273 (__location__ "ctdb_recheck_persistent_health() "
274 "failed (%llu of %llu times) - retry later\n",
275 (unsigned long long)ctdb->db_persistent_check_errors,
276 (unsigned long long)ctdb->max_persistent_check_errors));
277 event_add_timed(ctdb->ev,
278 ctdb->monitor->monitor_context,
279 timeval_current_ofs(1, 0),
280 ctdb_wait_until_recovered, ctdb);
283 DEBUG(DEBUG_ALERT,(__location__
284 "ctdb_recheck_persistent_health() failed (%llu times) - prepare shutdown\n",
285 (unsigned long long)ctdb->db_persistent_check_errors));
286 ctdb_stop_recoverd(ctdb);
287 ctdb_stop_keepalive(ctdb);
288 ctdb_stop_monitoring(ctdb);
289 ctdb_release_all_ips(ctdb);
290 if (ctdb->methods != NULL) {
291 ctdb->methods->shutdown(ctdb);
293 ctdb_event_script(ctdb, CTDB_EVENT_SHUTDOWN);
294 DEBUG(DEBUG_ALERT,("ctdb_recheck_persistent_health() failed - Stopping CTDB daemon\n"));
297 ctdb->db_persistent_check_errors = 0;
298 DEBUG(DEBUG_NOTICE,(__location__
299 "ctdb_start_monitoring: ctdb_recheck_persistent_health() OK\n"));
301 DEBUG(DEBUG_NOTICE,(__location__ " Recoveries finished. Running the \"startup\" event.\n"));
302 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
303 timeval_current_ofs(1, 0),
304 ctdb_check_health, ctdb);
309 see if the event scripts think we are healthy
311 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
312 struct timeval t, void *private_data)
314 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
317 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
318 (ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup)) {
319 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
320 timeval_current_ofs(ctdb->monitor->next_interval, 0),
321 ctdb_check_health, ctdb);
325 if (!ctdb->done_startup) {
326 ret = ctdb_event_script_callback(ctdb,
327 ctdb->monitor->monitor_context, ctdb_startup_callback,
329 CTDB_EVENT_STARTUP, "%s", "");
332 int skip_monitoring = 0;
334 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
336 DEBUG(DEBUG_ERR,("Skip monitoring during recovery\n"));
338 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
339 if (ctdb->freeze_handles[i] != NULL) {
340 DEBUG(DEBUG_ERR,("Skip monitoring since databases are frozen\n"));
345 if (skip_monitoring != 0) {
346 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
347 timeval_current_ofs(ctdb->monitor->next_interval, 0),
348 ctdb_check_health, ctdb);
351 ret = ctdb_event_script_callback(ctdb,
352 ctdb->monitor->monitor_context, ctdb_health_callback,
354 CTDB_EVENT_MONITOR, "%s", "");
359 DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
360 ctdb->monitor->next_interval = 5;
361 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
362 timeval_current_ofs(5, 0),
363 ctdb_check_health, ctdb);
368 (Temporaily) Disabling monitoring will stop the monitor event scripts
369 from running but node health checks will still occur
371 void ctdb_disable_monitoring(struct ctdb_context *ctdb)
373 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
374 DEBUG(DEBUG_INFO,("Monitoring has been disabled\n"));
378 Re-enable running monitor events after they have been disabled
380 void ctdb_enable_monitoring(struct ctdb_context *ctdb)
382 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
383 ctdb->monitor->next_interval = 5;
384 DEBUG(DEBUG_INFO,("Monitoring has been enabled\n"));
387 /* stop any monitoring
388 this should only be done when shutting down the daemon
390 void ctdb_stop_monitoring(struct ctdb_context *ctdb)
392 talloc_free(ctdb->monitor->monitor_context);
393 ctdb->monitor->monitor_context = NULL;
395 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
396 ctdb->monitor->next_interval = 5;
397 DEBUG(DEBUG_NOTICE,("Monitoring has been stopped\n"));
401 start watching for nodes that might be dead
403 void ctdb_start_monitoring(struct ctdb_context *ctdb)
405 if (ctdb->monitor != NULL) {
409 ctdb->monitor = talloc(ctdb, struct ctdb_monitor_state);
410 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor);
412 ctdb->monitor->next_interval = 5;
414 ctdb->monitor->monitor_context = talloc_new(ctdb->monitor);
415 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor->monitor_context);
417 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
418 timeval_current_ofs(1, 0),
419 ctdb_wait_until_recovered, ctdb);
421 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
422 DEBUG(DEBUG_NOTICE,("Monitoring has been started\n"));
427 modify flags on a node
429 int32_t ctdb_control_modflags(struct ctdb_context *ctdb, TDB_DATA indata)
431 struct ctdb_node_flag_change *c = (struct ctdb_node_flag_change *)indata.dptr;
432 struct ctdb_node *node;
435 if (c->pnn >= ctdb->num_nodes) {
436 DEBUG(DEBUG_ERR,(__location__ " Node %d is invalid, num_nodes :%d\n", c->pnn, ctdb->num_nodes));
440 node = ctdb->nodes[c->pnn];
441 old_flags = node->flags;
442 if (c->pnn != ctdb->pnn) {
443 c->old_flags = node->flags;
445 node->flags = c->new_flags & ~NODE_FLAGS_DISCONNECTED;
446 node->flags |= (c->old_flags & NODE_FLAGS_DISCONNECTED);
448 /* we dont let other nodes modify our STOPPED status */
449 if (c->pnn == ctdb->pnn) {
450 node->flags &= ~NODE_FLAGS_STOPPED;
451 if (old_flags & NODE_FLAGS_STOPPED) {
452 node->flags |= NODE_FLAGS_STOPPED;
456 /* we dont let other nodes modify our BANNED status */
457 if (c->pnn == ctdb->pnn) {
458 node->flags &= ~NODE_FLAGS_BANNED;
459 if (old_flags & NODE_FLAGS_BANNED) {
460 node->flags |= NODE_FLAGS_BANNED;
464 if (node->flags == c->old_flags) {
465 DEBUG(DEBUG_INFO, ("Control modflags on node %u - Unchanged - flags 0x%x\n", c->pnn, node->flags));
469 DEBUG(DEBUG_INFO, ("Control modflags on node %u - flags now 0x%x\n", c->pnn, node->flags));
471 if (node->flags == 0 && !ctdb->done_startup) {
472 DEBUG(DEBUG_ERR, (__location__ " Node %u became healthy - force recovery for startup\n",
474 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
477 /* tell the recovery daemon something has changed */
478 ctdb_daemon_send_message(ctdb, ctdb->pnn,
479 CTDB_SRVID_SET_NODE_FLAGS, indata);
481 /* if we have become banned, we should go into recovery mode */
482 if ((node->flags & NODE_FLAGS_BANNED) && !(c->old_flags & NODE_FLAGS_BANNED) && (node->pnn == ctdb->pnn)) {
483 return ctdb_local_node_got_banned(ctdb);
490 return the monitoring mode
492 int32_t ctdb_monitoring_mode(struct ctdb_context *ctdb)
494 if (ctdb->monitor == NULL) {
495 return CTDB_MONITORING_DISABLED;
497 return ctdb->monitor->monitoring_mode;