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 ret = ctdb_run_notification_script_child(ctdb, event);
96 DEBUG(DEBUG_ERR,(__location__ " Notification script failed\n"));
105 called when a health monitoring event script finishes
107 static void ctdb_health_callback(struct ctdb_context *ctdb, int status, void *p)
109 struct ctdb_node *node = ctdb->nodes[ctdb->pnn];
111 struct ctdb_node_flag_change c;
112 uint32_t next_interval;
115 struct takeover_run_reply rd;
118 c.old_flags = node->flags;
121 rd.srvid = CTDB_SRVID_TAKEOVER_RUN_RESPONSE;
123 rddata.dptr = (uint8_t *)&rd;
124 rddata.dsize = sizeof(rd);
126 if (status == -ETIME) {
127 ctdb->event_script_timeouts++;
129 if (ctdb->event_script_timeouts >= ctdb->tunable.script_timeout_count) {
130 DEBUG(DEBUG_ERR, ("Maximum timeout count %u reached for eventscript. Making node unhealthy\n", ctdb->tunable.script_timeout_count));
132 /* We pretend this is OK. */
133 goto after_change_status;
137 if (status != 0 && !(node->flags & NODE_FLAGS_UNHEALTHY)) {
138 DEBUG(DEBUG_NOTICE,("monitor event failed - disabling node\n"));
139 node->flags |= NODE_FLAGS_UNHEALTHY;
140 ctdb->monitor->next_interval = 5;
142 ctdb_run_notification_script(ctdb, "unhealthy");
144 /* ask the recmaster to reallocate all addresses */
145 DEBUG(DEBUG_ERR,("Node became UNHEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
146 ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
148 DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
151 } else if (status == 0 && (node->flags & NODE_FLAGS_UNHEALTHY)) {
152 DEBUG(DEBUG_NOTICE,("monitor event OK - node re-enabled\n"));
153 node->flags &= ~NODE_FLAGS_UNHEALTHY;
154 ctdb->monitor->next_interval = 5;
156 ctdb_run_notification_script(ctdb, "healthy");
158 /* ask the recmaster to reallocate all addresses */
159 DEBUG(DEBUG_ERR,("Node became HEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
160 ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
162 DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
168 next_interval = ctdb->monitor->next_interval;
170 ctdb->monitor->next_interval *= 2;
171 if (ctdb->monitor->next_interval > ctdb->tunable.monitor_interval) {
172 ctdb->monitor->next_interval = ctdb->tunable.monitor_interval;
175 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
176 timeval_current_ofs(next_interval, 0),
177 ctdb_check_health, ctdb);
179 if (c.old_flags == node->flags) {
183 c.new_flags = node->flags;
185 data.dptr = (uint8_t *)&c;
186 data.dsize = sizeof(c);
188 /* ask the recovery daemon to push these changes out to all nodes */
189 ctdb_daemon_send_message(ctdb, ctdb->pnn,
190 CTDB_SRVID_PUSH_NODE_FLAGS, data);
196 called when the startup event script finishes
198 static void ctdb_startup_callback(struct ctdb_context *ctdb, int status, void *p)
201 DEBUG(DEBUG_ERR,("startup event failed\n"));
202 } else if (status == 0) {
203 DEBUG(DEBUG_NOTICE,("startup event OK - enabling monitoring\n"));
204 ctdb->done_startup = true;
205 ctdb->monitor->next_interval = 5;
206 ctdb_run_notification_script(ctdb, "startup");
209 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
210 timeval_current_ofs(ctdb->monitor->next_interval, 0),
211 ctdb_check_health, ctdb);
216 wait until we have finished initial recoveries before we start the
219 static void ctdb_wait_until_recovered(struct event_context *ev, struct timed_event *te,
220 struct timeval t, void *private_data)
222 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
225 DEBUG(DEBUG_NOTICE,("CTDB_WAIT_UNTIL_RECOVERED\n"));
227 if (ctdb->vnn_map->generation == INVALID_GENERATION) {
228 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
230 DEBUG(DEBUG_NOTICE,(__location__ " generation is INVALID. Wait one more second\n"));
231 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
232 timeval_current_ofs(1, 0),
233 ctdb_wait_until_recovered, ctdb);
237 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
238 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
240 DEBUG(DEBUG_NOTICE,(__location__ " in recovery. Wait one more second\n"));
241 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
242 timeval_current_ofs(1, 0),
243 ctdb_wait_until_recovered, ctdb);
248 if (timeval_elapsed(&ctdb->last_recovery_finished) < (ctdb->tunable.rerecovery_timeout + 3)) {
249 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
251 DEBUG(DEBUG_NOTICE,(__location__ " wait for pending recoveries to end. Wait one more second.\n"));
253 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
254 timeval_current_ofs(1, 0),
255 ctdb_wait_until_recovered, ctdb);
259 if (ctdb->vnn_map->generation == ctdb->db_persistent_startup_generation) {
260 DEBUG(DEBUG_INFO,(__location__ " skip ctdb_recheck_persistent_health() "
261 "until the next recovery\n"));
262 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
263 timeval_current_ofs(1, 0),
264 ctdb_wait_until_recovered, ctdb);
268 ctdb->db_persistent_startup_generation = ctdb->vnn_map->generation;
269 ret = ctdb_recheck_persistent_health(ctdb);
271 ctdb->db_persistent_check_errors++;
272 if (ctdb->db_persistent_check_errors < ctdb->max_persistent_check_errors) {
273 DEBUG(ctdb->db_persistent_check_errors==1?DEBUG_ERR:DEBUG_WARNING,
274 (__location__ "ctdb_recheck_persistent_health() "
275 "failed (%llu of %llu times) - retry later\n",
276 (unsigned long long)ctdb->db_persistent_check_errors,
277 (unsigned long long)ctdb->max_persistent_check_errors));
278 event_add_timed(ctdb->ev,
279 ctdb->monitor->monitor_context,
280 timeval_current_ofs(1, 0),
281 ctdb_wait_until_recovered, ctdb);
284 DEBUG(DEBUG_ALERT,(__location__
285 "ctdb_recheck_persistent_health() failed (%llu times) - prepare shutdown\n",
286 (unsigned long long)ctdb->db_persistent_check_errors));
287 ctdb_stop_recoverd(ctdb);
288 ctdb_stop_keepalive(ctdb);
289 ctdb_stop_monitoring(ctdb);
290 ctdb_release_all_ips(ctdb);
291 if (ctdb->methods != NULL) {
292 ctdb->methods->shutdown(ctdb);
294 ctdb_event_script(ctdb, CTDB_EVENT_SHUTDOWN);
295 DEBUG(DEBUG_ALERT,("ctdb_recheck_persistent_health() failed - Stopping CTDB daemon\n"));
298 ctdb->db_persistent_check_errors = 0;
299 DEBUG(DEBUG_NOTICE,(__location__
300 "ctdb_start_monitoring: ctdb_recheck_persistent_health() OK\n"));
302 DEBUG(DEBUG_NOTICE,(__location__ " Recoveries finished. Running the \"startup\" event.\n"));
303 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
304 timeval_current_ofs(1, 0),
305 ctdb_check_health, ctdb);
310 see if the event scripts think we are healthy
312 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
313 struct timeval t, void *private_data)
315 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
318 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
319 (ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup)) {
320 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
321 timeval_current_ofs(ctdb->monitor->next_interval, 0),
322 ctdb_check_health, ctdb);
326 if (!ctdb->done_startup) {
327 ret = ctdb_event_script_callback(ctdb,
328 ctdb->monitor->monitor_context, ctdb_startup_callback,
330 CTDB_EVENT_STARTUP, "%s", "");
333 int skip_monitoring = 0;
335 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
337 DEBUG(DEBUG_ERR,("Skip monitoring during recovery\n"));
339 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
340 if (ctdb->freeze_handles[i] != NULL) {
341 DEBUG(DEBUG_ERR,("Skip monitoring since databases are frozen\n"));
346 if (skip_monitoring != 0) {
347 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
348 timeval_current_ofs(ctdb->monitor->next_interval, 0),
349 ctdb_check_health, ctdb);
352 ret = ctdb_event_script_callback(ctdb,
353 ctdb->monitor->monitor_context, ctdb_health_callback,
355 CTDB_EVENT_MONITOR, "%s", "");
360 DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
361 ctdb->monitor->next_interval = 5;
362 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
363 timeval_current_ofs(5, 0),
364 ctdb_check_health, ctdb);
369 (Temporaily) Disabling monitoring will stop the monitor event scripts
370 from running but node health checks will still occur
372 void ctdb_disable_monitoring(struct ctdb_context *ctdb)
374 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
375 DEBUG(DEBUG_INFO,("Monitoring has been disabled\n"));
379 Re-enable running monitor events after they have been disabled
381 void ctdb_enable_monitoring(struct ctdb_context *ctdb)
383 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
384 ctdb->monitor->next_interval = 5;
385 DEBUG(DEBUG_INFO,("Monitoring has been enabled\n"));
388 /* stop any monitoring
389 this should only be done when shutting down the daemon
391 void ctdb_stop_monitoring(struct ctdb_context *ctdb)
393 talloc_free(ctdb->monitor->monitor_context);
394 ctdb->monitor->monitor_context = NULL;
396 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
397 ctdb->monitor->next_interval = 5;
398 DEBUG(DEBUG_NOTICE,("Monitoring has been stopped\n"));
402 start watching for nodes that might be dead
404 void ctdb_start_monitoring(struct ctdb_context *ctdb)
406 if (ctdb->monitor != NULL) {
410 ctdb->monitor = talloc(ctdb, struct ctdb_monitor_state);
411 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor);
413 ctdb->monitor->next_interval = 5;
415 ctdb->monitor->monitor_context = talloc_new(ctdb->monitor);
416 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor->monitor_context);
418 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
419 timeval_current_ofs(1, 0),
420 ctdb_wait_until_recovered, ctdb);
422 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
423 DEBUG(DEBUG_NOTICE,("Monitoring has been started\n"));
428 modify flags on a node
430 int32_t ctdb_control_modflags(struct ctdb_context *ctdb, TDB_DATA indata)
432 struct ctdb_node_flag_change *c = (struct ctdb_node_flag_change *)indata.dptr;
433 struct ctdb_node *node;
437 if (c->pnn >= ctdb->num_nodes) {
438 DEBUG(DEBUG_ERR,(__location__ " Node %d is invalid, num_nodes :%d\n", c->pnn, ctdb->num_nodes));
442 node = ctdb->nodes[c->pnn];
443 old_flags = node->flags;
444 if (c->pnn != ctdb->pnn) {
445 c->old_flags = node->flags;
447 node->flags = c->new_flags & ~NODE_FLAGS_DISCONNECTED;
448 node->flags |= (c->old_flags & NODE_FLAGS_DISCONNECTED);
450 /* we dont let other nodes modify our STOPPED status */
451 if (c->pnn == ctdb->pnn) {
452 node->flags &= ~NODE_FLAGS_STOPPED;
453 if (old_flags & NODE_FLAGS_STOPPED) {
454 node->flags |= NODE_FLAGS_STOPPED;
458 /* we dont let other nodes modify our BANNED status */
459 if (c->pnn == ctdb->pnn) {
460 node->flags &= ~NODE_FLAGS_BANNED;
461 if (old_flags & NODE_FLAGS_BANNED) {
462 node->flags |= NODE_FLAGS_BANNED;
466 if (node->flags == c->old_flags) {
467 DEBUG(DEBUG_INFO, ("Control modflags on node %u - Unchanged - flags 0x%x\n", c->pnn, node->flags));
471 DEBUG(DEBUG_INFO, ("Control modflags on node %u - flags now 0x%x\n", c->pnn, node->flags));
473 if (node->flags == 0 && !ctdb->done_startup) {
474 DEBUG(DEBUG_ERR, (__location__ " Node %u became healthy - force recovery for startup\n",
476 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
479 /* tell the recovery daemon something has changed */
480 ctdb_daemon_send_message(ctdb, ctdb->pnn,
481 CTDB_SRVID_SET_NODE_FLAGS, indata);
483 /* if we have become banned, we should go into recovery mode */
484 if ((node->flags & NODE_FLAGS_BANNED) && !(c->old_flags & NODE_FLAGS_BANNED) && (node->pnn == ctdb->pnn)) {
485 /* make sure we are frozen */
486 DEBUG(DEBUG_NOTICE,("This node has been banned - forcing freeze and recovery\n"));
487 /* Reset the generation id to 1 to make us ignore any
488 REQ/REPLY CALL/DMASTER someone sends to us.
489 We are now banned so we shouldnt service database calls
492 ctdb->vnn_map->generation = INVALID_GENERATION;
494 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
495 if (ctdb_start_freeze(ctdb, i) != 0) {
496 DEBUG(DEBUG_ERR,(__location__ " Failed to freeze db priority %u\n", i));
499 ctdb_release_all_ips(ctdb);
500 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
507 return the monitoring mode
509 int32_t ctdb_monitoring_mode(struct ctdb_context *ctdb)
511 if (ctdb->monitor == NULL) {
512 return CTDB_MONITORING_DISABLED;
514 return ctdb->monitor->monitoring_mode;