3 * Bacula Director Job processing routines
5 * Kern Sibbald, October MM
10 Copyright (C) 2000-2003 Kern Sibbald and John Walker
12 This program is free software; you can redistribute it and/or
13 modify it under the terms of the GNU General Public License as
14 published by the Free Software Foundation; either version 2 of
15 the License, or (at your option) any later version.
17 This program is distributed in the hope that it will be useful,
18 but WITHOUT ANY WARRANTY; without even the implied warranty of
19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 General Public License for more details.
22 You should have received a copy of the GNU General Public
23 License along with this program; if not, write to the Free
24 Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
32 /* Forward referenced subroutines */
33 static void *job_thread(void *arg);
34 static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg);
35 static void release_resource_locks(JCR *jcr);
36 static int acquire_resource_locks(JCR *jcr);
38 static void backoff_resource_locks(JCR *jcr, int count);
41 /* Exported subroutines */
42 void run_job(JCR *jcr);
45 /* Imported subroutines */
46 extern void term_scheduler();
47 extern void term_ua_server();
48 extern int do_backup(JCR *jcr);
49 extern int do_admin(JCR *jcr);
50 extern int do_restore(JCR *jcr);
51 extern int do_verify(JCR *jcr);
54 static semlock_t job_lock;
55 static pthread_mutex_t mutex;
56 static pthread_cond_t resource_wait;
57 static int waiting = 0; /* count of waiting threads */
59 /* Queue of jobs to be run */
60 workq_t job_wq; /* our job work queue */
63 void init_job_server(int max_workers)
67 if ((stat = sem_init(&job_lock, max_workers)) != 0) {
68 Emsg1(M_ABORT, 0, _("Could not init job lock: ERR=%s\n"), strerror(stat));
70 if ((stat = pthread_mutex_init(&mutex, NULL)) != 0) {
71 Emsg1(M_ABORT, 0, _("Could not init resource mutex: ERR=%s\n"), strerror(stat));
73 if ((stat = pthread_cond_init(&resource_wait, NULL)) != 0) {
74 Emsg1(M_ABORT, 0, _("Could not init resource wait: ERR=%s\n"), strerror(stat));
78 if ((stat = workq_init(&job_wq, max_workers, job_thread)) != 0) {
79 Emsg1(M_ABORT, 0, _("Could not init job work queue: ERR=%s\n"), strerror(stat));
86 * Run a job -- typically called by the scheduler, but may also
87 * be called by the UA (Console program).
90 void run_job(JCR *jcr)
96 workq_ele_t *work_item;
99 sm_check(__FILE__, __LINE__, True);
100 init_msg(jcr, jcr->messages);
101 create_unique_job_name(jcr, jcr->job->hdr.name);
102 set_jcr_job_status(jcr, JS_Created);
103 jcr->jr.SchedTime = jcr->sched_time;
104 jcr->jr.StartTime = jcr->start_time;
105 jcr->jr.Type = jcr->JobType;
106 jcr->jr.Level = jcr->JobLevel;
107 jcr->jr.JobStatus = jcr->JobStatus;
108 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
109 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
111 /* Initialize termination condition variable */
112 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
113 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
114 set_jcr_job_status(jcr, JS_ErrorTerminated);
122 Dmsg0(50, "Open database\n");
123 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
124 jcr->catalog->db_password, jcr->catalog->db_address,
125 jcr->catalog->db_port, jcr->catalog->db_socket);
126 if (!db_open_database(jcr, jcr->db)) {
127 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
128 set_jcr_job_status(jcr, JS_ErrorTerminated);
132 Dmsg0(50, "DB opened\n");
137 jcr->jr.JobStatus = jcr->JobStatus;
138 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
139 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
140 set_jcr_job_status(jcr, JS_ErrorTerminated);
144 jcr->JobId = jcr->jr.JobId;
145 ASSERT(jcr->jr.JobId > 0);
147 Dmsg4(30, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
148 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
149 Dmsg0(200, "Add jrc to work queue\n");
152 if ((stat = pthread_create(&tid, NULL, job_thread, (void *)jcr)) != 0) {
153 Emsg1(M_ABORT, 0, _("Unable to create job thread: ERR=%s\n"), strerror(stat));
156 /* Queue the job to be run */
157 if ((stat = workq_add(&job_wq, (void *)jcr, &work_item, 0)) != 0) {
158 Emsg1(M_ABORT, 0, _("Could not add job to work queue: ERR=%s\n"), strerror(stat));
160 jcr->work_item = work_item;
162 Dmsg0(200, "Done run_job()\n");
166 * This is the engine called by workq_add() when we were pulled
167 * from the work queue.
168 * At this point, we are running in our own thread
170 static void *job_thread(void *arg)
172 JCR *jcr = (JCR *)arg;
174 pthread_detach(pthread_self());
175 sm_check(__FILE__, __LINE__, True);
178 if (!acquire_resource_locks(jcr)) {
179 set_jcr_job_status(jcr, JS_Canceled);
182 Dmsg0(200, "=====Start Job=========\n");
183 jcr->start_time = time(NULL); /* set the real start time */
184 set_jcr_job_status(jcr, JS_Running);
186 if (job_canceled(jcr)) {
187 update_job_end_record(jcr);
188 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
189 (utime_t)(jcr->start_time - jcr->sched_time)) {
190 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
191 set_jcr_job_status(jcr, JS_Canceled);
192 update_job_end_record(jcr);
196 if (jcr->job->RunBeforeJob) {
197 POOLMEM *before = get_pool_memory(PM_FNAME);
200 char line[MAXSTRING];
202 before = edit_run_codes(jcr, before, jcr->job->RunBeforeJob);
203 bpipe = open_bpipe(before, 0, "r");
204 free_pool_memory(before);
205 while (fgets(line, sizeof(line), bpipe->rfd)) {
206 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
208 status = close_bpipe(bpipe);
210 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
212 set_jcr_job_status(jcr, JS_FatalError);
213 update_job_end_record(jcr);
217 switch (jcr->JobType) {
220 if (jcr->JobStatus == JS_Terminated) {
226 if (jcr->JobStatus == JS_Terminated) {
232 if (jcr->JobStatus == JS_Terminated) {
238 if (jcr->JobStatus == JS_Terminated) {
243 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
246 if (jcr->job->RunAfterJob) {
247 POOLMEM *after = get_pool_memory(PM_FNAME);
250 char line[MAXSTRING];
252 after = edit_run_codes(jcr, after, jcr->job->RunAfterJob);
253 bpipe = open_bpipe(after, 0, "r");
254 free_pool_memory(after);
255 while (fgets(line, sizeof(line), bpipe->rfd)) {
256 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
258 status = close_bpipe(bpipe);
260 Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
262 set_jcr_job_status(jcr, JS_FatalError);
263 update_job_end_record(jcr);
268 release_resource_locks(jcr);
269 if (jcr->job->RescheduleOnError &&
270 jcr->JobStatus != JS_Terminated &&
271 jcr->JobStatus != JS_Canceled &&
272 jcr->job->RescheduleTimes > 0 &&
273 jcr->reschedule_count < jcr->job->RescheduleTimes) {
276 * Reschedule this job by cleaning it up, but
277 * reuse the same JobId if possible.
279 jcr->reschedule_count++;
280 jcr->sched_time = time(NULL) + jcr->job->RescheduleInterval;
281 Dmsg2(100, "Rescheduled Job %s to re-run in %d seconds.\n", jcr->Job,
282 (int)jcr->job->RescheduleInterval);
283 jcr->JobStatus = JS_Created; /* force new status */
284 dird_free_jcr(jcr); /* partial cleanup old stuff */
285 if (jcr->JobBytes == 0) {
286 continue; /* reschedule the job */
289 * Something was actually backed up, so we cannot reuse
290 * the old JobId or there will be database record
291 * conflicts. We now create a new job, copying the
292 * appropriate fields.
294 JCR *njcr = new_jcr(sizeof(JCR), dird_free_jcr);
295 set_jcr_defaults(njcr, jcr->job);
296 njcr->reschedule_count = jcr->reschedule_count;
297 njcr->JobLevel = jcr->JobLevel;
298 njcr->JobStatus = jcr->JobStatus;
299 njcr->pool = jcr->pool;
300 njcr->store = jcr->store;
301 njcr->messages = jcr->messages;
308 Dmsg0(200, "Close DB\n");
309 db_close_database(jcr, jcr->db);
313 Dmsg0(50, "======== End Job ==========\n");
314 sm_check(__FILE__, __LINE__, True);
319 * Acquire the resources needed. These locks limit the
320 * number of jobs by each resource. We have limits on
321 * Jobs, Clients, Storage, and total jobs.
323 static int acquire_resource_locks(JCR *jcr)
325 time_t now = time(NULL);
326 time_t wtime = jcr->sched_time - now;
328 /* Wait until scheduled time arrives */
329 if (wtime > 0 && verbose) {
330 Jmsg(jcr, M_INFO, 0, _("Job %s waiting %d seconds for scheduled start time.\n"),
332 set_jcr_job_status(jcr, JS_WaitStartTime);
334 /* Check every 30 seconds if canceled */
336 Dmsg2(100, "Waiting on sched time, jobid=%d secs=%d\n", jcr->JobId, wtime);
340 bmicrosleep(wtime, 0);
341 if (job_canceled(jcr)) {
344 wtime = jcr->sched_time - time(NULL);
351 /* Initialize semaphores */
352 if (jcr->store->sem.valid != SEMLOCK_VALID) {
353 if ((stat = sem_init(&jcr->store->sem, jcr->store->MaxConcurrentJobs)) != 0) {
354 Emsg1(M_ABORT, 0, _("Could not init Storage semaphore: ERR=%s\n"), strerror(stat));
357 if (jcr->client->sem.valid != SEMLOCK_VALID) {
358 if ((stat = sem_init(&jcr->client->sem, jcr->client->MaxConcurrentJobs)) != 0) {
359 Emsg1(M_ABORT, 0, _("Could not init Client semaphore: ERR=%s\n"), strerror(stat));
362 if (jcr->job->sem.valid != SEMLOCK_VALID) {
363 if ((stat = sem_init(&jcr->job->sem, jcr->job->MaxConcurrentJobs)) != 0) {
364 Emsg1(M_ABORT, 0, _("Could not init Job semaphore: ERR=%s\n"), strerror(stat));
369 /* Acquire semaphore */
370 set_jcr_job_status(jcr, JS_WaitJobRes);
371 if ((stat = sem_lock(&jcr->job->sem)) != 0) {
372 Emsg1(M_ABORT, 0, _("Could not acquire Job max jobs lock: ERR=%s\n"), strerror(stat));
374 set_jcr_job_status(jcr, JS_WaitClientRes);
375 if ((stat = sem_trylock(&jcr->client->sem)) != 0) {
377 backoff_resource_locks(jcr, 1);
380 Emsg1(M_ABORT, 0, _("Could not acquire Client max jobs lock: ERR=%s\n"), strerror(stat));
383 set_jcr_job_status(jcr, JS_WaitStoreRes);
384 if ((stat = sem_trylock(&jcr->store->sem)) != 0) {
386 backoff_resource_locks(jcr, 2);
389 Emsg1(M_ABORT, 0, _("Could not acquire Storage max jobs lock: ERR=%s\n"), strerror(stat));
392 set_jcr_job_status(jcr, JS_WaitMaxJobs);
393 if ((stat = sem_trylock(&job_lock)) != 0) {
395 backoff_resource_locks(jcr, 3);
398 Emsg1(M_ABORT, 0, _("Could not acquire max jobs lock: ERR=%s\n"), strerror(stat));
404 if (job_canceled(jcr)) {
409 * Wait for a resource to be released either by backoff or
410 * by a job terminating.
413 pthread_cond_wait(&resource_wait, &mutex);
418 jcr->acquired_resource_locks = 1;
425 * We could not get all the resource locks because
426 * too many jobs are running, so release any locks
427 * we did acquire, giving others a chance to use them
430 static void backoff_resource_locks(JCR *jcr, int count)
435 sem_unlock(&jcr->store->sem);
436 /* Fall through wanted */
438 sem_unlock(&jcr->client->sem);
439 /* Fall through wanted */
441 sem_unlock(&jcr->job->sem);
445 * Since we released a lock, if there are any threads
446 * waiting, wake them up so that they can try again.
449 pthread_cond_broadcast(&resource_wait);
456 * This is called at the end of the job to release
457 * any resource limits on the number of jobs. If
458 * there are any other jobs waiting, we wake them
459 * up so that they can try again.
461 static void release_resource_locks(JCR *jcr)
463 if (!jcr->acquired_resource_locks) {
464 return; /* Job canceled, no locks acquired */
468 sem_unlock(&jcr->store->sem);
469 sem_unlock(&jcr->client->sem);
470 sem_unlock(&jcr->job->sem);
471 sem_unlock(&job_lock);
473 pthread_cond_broadcast(&resource_wait);
475 jcr->acquired_resource_locks = 0;
481 * Get or create a Client record for this Job
483 int get_or_create_client_record(JCR *jcr)
487 memset(&cr, 0, sizeof(cr));
488 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
489 cr.AutoPrune = jcr->client->AutoPrune;
490 cr.FileRetention = jcr->client->FileRetention;
491 cr.JobRetention = jcr->client->JobRetention;
492 if (!jcr->client_name) {
493 jcr->client_name = get_pool_memory(PM_NAME);
495 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
496 if (!db_create_client_record(jcr, jcr->db, &cr)) {
497 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
498 db_strerror(jcr->db));
501 jcr->jr.ClientId = cr.ClientId;
503 if (!jcr->client_uname) {
504 jcr->client_uname = get_pool_memory(PM_NAME);
506 pm_strcpy(&jcr->client_uname, cr.Uname);
508 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
515 * Write status and such in DB
517 void update_job_end_record(JCR *jcr)
519 if (jcr->jr.EndTime == 0) {
520 jcr->jr.EndTime = time(NULL);
522 jcr->end_time = jcr->jr.EndTime;
523 jcr->jr.JobId = jcr->JobId;
524 jcr->jr.JobStatus = jcr->JobStatus;
525 jcr->jr.JobFiles = jcr->JobFiles;
526 jcr->jr.JobBytes = jcr->JobBytes;
527 jcr->jr.VolSessionId = jcr->VolSessionId;
528 jcr->jr.VolSessionTime = jcr->VolSessionTime;
529 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
530 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
531 db_strerror(jcr->db));
536 * Takes base_name and appends (unique) current
537 * date and time to form unique job name.
539 * Returns: unique job name in jcr->Job
540 * date/time in jcr->start_time
542 void create_unique_job_name(JCR *jcr, char *base_name)
544 /* Job start mutex */
545 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
546 static time_t last_start_time = 0;
549 char dt[MAX_TIME_LENGTH];
550 char name[MAX_NAME_LENGTH];
553 /* Guarantee unique start time -- maximum one per second, and
554 * thus unique Job Name
556 P(mutex); /* lock creation of jobs */
558 while (now == last_start_time) {
559 bmicrosleep(0, 500000);
562 last_start_time = now;
563 V(mutex); /* allow creation of jobs */
564 jcr->start_time = now;
565 /* Form Unique JobName */
566 localtime_r(&now, &tm);
567 /* Use only characters that are permitted in Windows filenames */
568 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
569 bstrncpy(name, base_name, sizeof(name));
570 name[sizeof(name)-22] = 0; /* truncate if too long */
571 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
572 /* Convert spaces into underscores */
573 for (p=jcr->Job; *p; p++) {
581 * Free the Job Control Record if no one is still using it.
582 * Called from main free_jcr() routine in src/lib/jcr.c so
583 * that we can do our Director specific cleanup of the jcr.
585 void dird_free_jcr(JCR *jcr)
587 Dmsg0(200, "Start dird free_jcr\n");
589 if (jcr->sd_auth_key) {
590 free(jcr->sd_auth_key);
591 jcr->sd_auth_key = NULL;
597 if (jcr->file_bsock) {
598 Dmsg0(200, "Close File bsock\n");
599 bnet_close(jcr->file_bsock);
600 jcr->file_bsock = NULL;
602 if (jcr->store_bsock) {
603 Dmsg0(200, "Close Store bsock\n");
604 bnet_close(jcr->store_bsock);
605 jcr->store_bsock = NULL;
608 Dmsg0(200, "Free JCR fname\n");
609 free_pool_memory(jcr->fname);
613 Dmsg0(200, "Free JCR stime\n");
614 free_pool_memory(jcr->stime);
617 if (jcr->RestoreBootstrap) {
618 free(jcr->RestoreBootstrap);
619 jcr->RestoreBootstrap = NULL;
621 if (jcr->client_uname) {
622 free_pool_memory(jcr->client_uname);
623 jcr->client_uname = NULL;
625 Dmsg0(200, "End dird free_jcr\n");
629 * Set some defaults in the JCR necessary to
630 * run. These items are pulled from the job
631 * definition as defaults, but can be overridden
632 * later either by the Run record in the Schedule resource,
633 * or by the Console program.
635 void set_jcr_defaults(JCR *jcr, JOB *job)
638 jcr->JobType = job->JobType;
639 jcr->JobLevel = job->level;
640 jcr->store = job->storage;
641 jcr->client = job->client;
642 if (!jcr->client_name) {
643 jcr->client_name = get_pool_memory(PM_NAME);
645 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
646 jcr->pool = job->pool;
647 jcr->catalog = job->client->catalog;
648 jcr->fileset = job->fileset;
649 jcr->messages = job->messages;
650 if (jcr->RestoreBootstrap) {
651 free(jcr->RestoreBootstrap);
653 /* This can be overridden by Console program */
654 if (job->RestoreBootstrap) {
655 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
657 /* If no default level given, set one */
658 if (jcr->JobLevel == 0) {
659 switch (jcr->JobType) {
661 jcr->JobLevel = L_VERIFY_CATALOG;
664 jcr->JobLevel = L_INCREMENTAL;
668 jcr->JobLevel = L_FULL;
677 * Edit codes into Run command
680 * %d = Director's name
688 * omsg = edited output message
689 * imsg = input string containing edit codes (%x)
692 static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg)
699 Dmsg1(200, "edit_run_codes: %s\n", imsg);
700 for (p=imsg; *p; p++) {
707 str = jcr->client_name;
716 str = job_status_to_str(jcr->JobStatus);
719 sprintf(add, "%d", jcr->JobId);
726 str = job_level_to_str(jcr->JobLevel);
729 str = jcr->job->hdr.name;
732 str = job_type_to_str(jcr->JobType);
746 Dmsg1(200, "add_str %s\n", str);
747 pm_strcat(&omsg, (char *)str);
748 Dmsg1(200, "omsg=%s\n", omsg);