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/tevent/tevent.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 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;
119 c.old_flags = node->flags;
122 rd.srvid = CTDB_SRVID_TAKEOVER_RUN_RESPONSE;
124 rddata.dptr = (uint8_t *)&rd;
125 rddata.dsize = sizeof(rd);
127 if (status == -ETIME) {
128 ctdb->event_script_timeouts++;
130 if (ctdb->event_script_timeouts >= ctdb->tunable.script_timeout_count) {
131 DEBUG(DEBUG_ERR, ("Maximum timeout count %u reached for eventscript. Making node unhealthy\n", ctdb->tunable.script_timeout_count));
133 /* We pretend this is OK. */
134 goto after_change_status;
138 if (status != 0 && !(node->flags & NODE_FLAGS_UNHEALTHY)) {
139 DEBUG(DEBUG_NOTICE,("monitor event failed - disabling node\n"));
140 node->flags |= NODE_FLAGS_UNHEALTHY;
141 ctdb->monitor->next_interval = 5;
143 ctdb_run_notification_script(ctdb, "unhealthy");
145 /* ask the recmaster to reallocate all addresses */
146 DEBUG(DEBUG_ERR,("Node became UNHEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
147 ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
149 DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
152 } else if (status == 0 && (node->flags & NODE_FLAGS_UNHEALTHY)) {
153 DEBUG(DEBUG_NOTICE,("monitor event OK - node re-enabled\n"));
154 node->flags &= ~NODE_FLAGS_UNHEALTHY;
155 ctdb->monitor->next_interval = 5;
157 ctdb_run_notification_script(ctdb, "healthy");
159 /* ask the recmaster to reallocate all addresses */
160 DEBUG(DEBUG_ERR,("Node became HEALTHY. Ask recovery master %u to perform ip reallocation\n", ctdb->recovery_master));
161 ret = ctdb_daemon_send_message(ctdb, ctdb->recovery_master, CTDB_SRVID_TAKEOVER_RUN, rddata);
163 DEBUG(DEBUG_ERR,(__location__ " Failed to send ip takeover run request message to %u\n", ctdb->recovery_master));
169 next_interval = ctdb->monitor->next_interval;
171 ctdb->monitor->next_interval *= 2;
172 if (ctdb->monitor->next_interval > ctdb->tunable.monitor_interval) {
173 ctdb->monitor->next_interval = ctdb->tunable.monitor_interval;
176 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
177 timeval_current_ofs(next_interval, 0),
178 ctdb_check_health, ctdb);
180 if (c.old_flags == node->flags) {
184 c.new_flags = node->flags;
186 data.dptr = (uint8_t *)&c;
187 data.dsize = sizeof(c);
189 /* ask the recovery daemon to push these changes out to all nodes */
190 ctdb_daemon_send_message(ctdb, ctdb->pnn,
191 CTDB_SRVID_PUSH_NODE_FLAGS, data);
197 called when the startup event script finishes
199 static void ctdb_startup_callback(struct ctdb_context *ctdb, int status, void *p)
202 DEBUG(DEBUG_ERR,("startup event failed\n"));
203 } else if (status == 0) {
204 DEBUG(DEBUG_NOTICE,("startup event OK - enabling monitoring\n"));
205 ctdb->done_startup = true;
206 ctdb->monitor->next_interval = 2;
207 ctdb_run_notification_script(ctdb, "startup");
210 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
211 timeval_current_ofs(ctdb->monitor->next_interval, 0),
212 ctdb_check_health, ctdb);
217 wait until we have finished initial recoveries before we start the
220 static void ctdb_wait_until_recovered(struct event_context *ev, struct timed_event *te,
221 struct timeval t, void *private_data)
223 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
226 DEBUG(DEBUG_NOTICE,("CTDB_WAIT_UNTIL_RECOVERED\n"));
228 if (ctdb->vnn_map->generation == INVALID_GENERATION) {
229 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
231 DEBUG(DEBUG_NOTICE,(__location__ " generation is INVALID. Wait one more second\n"));
232 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
233 timeval_current_ofs(1, 0),
234 ctdb_wait_until_recovered, ctdb);
238 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
239 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
241 DEBUG(DEBUG_NOTICE,(__location__ " in recovery. Wait one more second\n"));
242 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
243 timeval_current_ofs(1, 0),
244 ctdb_wait_until_recovered, ctdb);
249 if (!fast_start && timeval_elapsed(&ctdb->last_recovery_finished) < (ctdb->tunable.rerecovery_timeout + 3)) {
250 ctdb->db_persistent_startup_generation = INVALID_GENERATION;
252 DEBUG(DEBUG_NOTICE,(__location__ " wait for pending recoveries to end. Wait one more second.\n"));
254 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
255 timeval_current_ofs(1, 0),
256 ctdb_wait_until_recovered, ctdb);
260 if (ctdb->vnn_map->generation == ctdb->db_persistent_startup_generation) {
261 DEBUG(DEBUG_INFO,(__location__ " skip ctdb_recheck_persistent_health() "
262 "until the next recovery\n"));
263 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
264 timeval_current_ofs(1, 0),
265 ctdb_wait_until_recovered, ctdb);
269 ctdb->db_persistent_startup_generation = ctdb->vnn_map->generation;
270 ret = ctdb_recheck_persistent_health(ctdb);
272 ctdb->db_persistent_check_errors++;
273 if (ctdb->db_persistent_check_errors < ctdb->max_persistent_check_errors) {
274 DEBUG(ctdb->db_persistent_check_errors==1?DEBUG_ERR:DEBUG_WARNING,
275 (__location__ "ctdb_recheck_persistent_health() "
276 "failed (%llu of %llu times) - retry later\n",
277 (unsigned long long)ctdb->db_persistent_check_errors,
278 (unsigned long long)ctdb->max_persistent_check_errors));
279 event_add_timed(ctdb->ev,
280 ctdb->monitor->monitor_context,
281 timeval_current_ofs(1, 0),
282 ctdb_wait_until_recovered, ctdb);
285 DEBUG(DEBUG_ALERT,(__location__
286 "ctdb_recheck_persistent_health() failed (%llu times) - prepare shutdown\n",
287 (unsigned long long)ctdb->db_persistent_check_errors));
288 ctdb_stop_recoverd(ctdb);
289 ctdb_stop_keepalive(ctdb);
290 ctdb_stop_monitoring(ctdb);
291 ctdb_release_all_ips(ctdb);
292 if (ctdb->methods != NULL) {
293 ctdb->methods->shutdown(ctdb);
295 ctdb_event_script(ctdb, CTDB_EVENT_SHUTDOWN);
296 DEBUG(DEBUG_ALERT,("ctdb_recheck_persistent_health() failed - Stopping CTDB daemon\n"));
299 ctdb->db_persistent_check_errors = 0;
300 DEBUG(DEBUG_NOTICE,(__location__
301 "ctdb_start_monitoring: ctdb_recheck_persistent_health() OK\n"));
303 DEBUG(DEBUG_NOTICE,(__location__ " Recoveries finished. Running the \"startup\" event.\n"));
304 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
306 ctdb_check_health, ctdb);
311 see if the event scripts think we are healthy
313 static void ctdb_check_health(struct event_context *ev, struct timed_event *te,
314 struct timeval t, void *private_data)
316 struct ctdb_context *ctdb = talloc_get_type(private_data, struct ctdb_context);
319 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL ||
320 (ctdb->monitor->monitoring_mode == CTDB_MONITORING_DISABLED && ctdb->done_startup)) {
321 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
322 timeval_current_ofs(ctdb->monitor->next_interval, 0),
323 ctdb_check_health, ctdb);
327 if (!ctdb->done_startup) {
328 ret = ctdb_event_script_callback(ctdb,
329 ctdb->monitor->monitor_context, ctdb_startup_callback,
331 CTDB_EVENT_STARTUP, "%s", "");
334 int skip_monitoring = 0;
336 if (ctdb->recovery_mode != CTDB_RECOVERY_NORMAL) {
338 DEBUG(DEBUG_ERR,("Skip monitoring during recovery\n"));
340 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
341 if (ctdb->freeze_handles[i] != NULL) {
342 DEBUG(DEBUG_ERR,("Skip monitoring since databases are frozen\n"));
347 if (skip_monitoring != 0) {
348 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
349 timeval_current_ofs(ctdb->monitor->next_interval, 0),
350 ctdb_check_health, ctdb);
353 ret = ctdb_event_script_callback(ctdb,
354 ctdb->monitor->monitor_context, ctdb_health_callback,
356 CTDB_EVENT_MONITOR, "%s", "");
361 DEBUG(DEBUG_ERR,("Unable to launch monitor event script\n"));
362 ctdb->monitor->next_interval = 5;
363 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
364 timeval_current_ofs(5, 0),
365 ctdb_check_health, ctdb);
370 (Temporaily) Disabling monitoring will stop the monitor event scripts
371 from running but node health checks will still occur
373 void ctdb_disable_monitoring(struct ctdb_context *ctdb)
375 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
376 DEBUG(DEBUG_INFO,("Monitoring has been disabled\n"));
380 Re-enable running monitor events after they have been disabled
382 void ctdb_enable_monitoring(struct ctdb_context *ctdb)
384 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
385 ctdb->monitor->next_interval = 5;
386 DEBUG(DEBUG_INFO,("Monitoring has been enabled\n"));
389 /* stop any monitoring
390 this should only be done when shutting down the daemon
392 void ctdb_stop_monitoring(struct ctdb_context *ctdb)
394 talloc_free(ctdb->monitor->monitor_context);
395 ctdb->monitor->monitor_context = NULL;
397 ctdb->monitor->monitoring_mode = CTDB_MONITORING_DISABLED;
398 ctdb->monitor->next_interval = 5;
399 DEBUG(DEBUG_NOTICE,("Monitoring has been stopped\n"));
403 start watching for nodes that might be dead
405 void ctdb_start_monitoring(struct ctdb_context *ctdb)
407 if (ctdb->monitor != NULL) {
411 ctdb->monitor = talloc(ctdb, struct ctdb_monitor_state);
412 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor);
414 ctdb->monitor->next_interval = 5;
416 ctdb->monitor->monitor_context = talloc_new(ctdb->monitor);
417 CTDB_NO_MEMORY_FATAL(ctdb, ctdb->monitor->monitor_context);
419 event_add_timed(ctdb->ev, ctdb->monitor->monitor_context,
420 timeval_current_ofs(1, 0),
421 ctdb_wait_until_recovered, ctdb);
423 ctdb->monitor->monitoring_mode = CTDB_MONITORING_ACTIVE;
424 DEBUG(DEBUG_NOTICE,("Monitoring has been started\n"));
429 modify flags on a node
431 int32_t ctdb_control_modflags(struct ctdb_context *ctdb, TDB_DATA indata)
433 struct ctdb_node_flag_change *c = (struct ctdb_node_flag_change *)indata.dptr;
434 struct ctdb_node *node;
438 if (c->pnn >= ctdb->num_nodes) {
439 DEBUG(DEBUG_ERR,(__location__ " Node %d is invalid, num_nodes :%d\n", c->pnn, ctdb->num_nodes));
443 node = ctdb->nodes[c->pnn];
444 old_flags = node->flags;
445 if (c->pnn != ctdb->pnn) {
446 c->old_flags = node->flags;
448 node->flags = c->new_flags & ~NODE_FLAGS_DISCONNECTED;
449 node->flags |= (c->old_flags & NODE_FLAGS_DISCONNECTED);
451 /* we dont let other nodes modify our STOPPED status */
452 if (c->pnn == ctdb->pnn) {
453 node->flags &= ~NODE_FLAGS_STOPPED;
454 if (old_flags & NODE_FLAGS_STOPPED) {
455 node->flags |= NODE_FLAGS_STOPPED;
459 /* we dont let other nodes modify our BANNED status */
460 if (c->pnn == ctdb->pnn) {
461 node->flags &= ~NODE_FLAGS_BANNED;
462 if (old_flags & NODE_FLAGS_BANNED) {
463 node->flags |= NODE_FLAGS_BANNED;
467 if (node->flags == c->old_flags) {
468 DEBUG(DEBUG_INFO, ("Control modflags on node %u - Unchanged - flags 0x%x\n", c->pnn, node->flags));
472 DEBUG(DEBUG_INFO, ("Control modflags on node %u - flags now 0x%x\n", c->pnn, node->flags));
474 if (node->flags == 0 && !ctdb->done_startup) {
475 DEBUG(DEBUG_ERR, (__location__ " Node %u became healthy - force recovery for startup\n",
477 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
480 /* tell the recovery daemon something has changed */
481 ctdb_daemon_send_message(ctdb, ctdb->pnn,
482 CTDB_SRVID_SET_NODE_FLAGS, indata);
484 /* if we have become banned, we should go into recovery mode */
485 if ((node->flags & NODE_FLAGS_BANNED) && !(c->old_flags & NODE_FLAGS_BANNED) && (node->pnn == ctdb->pnn)) {
486 /* make sure we are frozen */
487 DEBUG(DEBUG_NOTICE,("This node has been banned - forcing freeze and recovery\n"));
488 /* Reset the generation id to 1 to make us ignore any
489 REQ/REPLY CALL/DMASTER someone sends to us.
490 We are now banned so we shouldnt service database calls
493 ctdb->vnn_map->generation = INVALID_GENERATION;
495 for (i=1; i<=NUM_DB_PRIORITIES; i++) {
496 if (ctdb_start_freeze(ctdb, i) != 0) {
497 DEBUG(DEBUG_ERR,(__location__ " Failed to freeze db priority %u\n", i));
500 ctdb_release_all_ips(ctdb);
501 ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
508 return the monitoring mode
510 int32_t ctdb_monitoring_mode(struct ctdb_context *ctdb)
512 if (ctdb->monitor == NULL) {
513 return CTDB_MONITORING_DISABLED;
515 return ctdb->monitor->monitoring_mode;