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_restore(JCR *jcr);
50 extern int do_verify(JCR *jcr);
51 extern void backup_cleanup(void);
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 db_close_database(jcr, jcr->db);
129 set_jcr_job_status(jcr, JS_ErrorTerminated);
133 Dmsg0(50, "DB opened\n");
138 jcr->jr.JobStatus = jcr->JobStatus;
139 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
140 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
141 db_close_database(jcr, jcr->db);
142 set_jcr_job_status(jcr, JS_ErrorTerminated);
146 jcr->JobId = jcr->jr.JobId;
147 ASSERT(jcr->jr.JobId > 0);
149 Dmsg4(30, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
150 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
151 Dmsg0(200, "Add jrc to work queue\n");
154 if ((stat = pthread_create(&tid, NULL, job_thread, (void *)jcr)) != 0) {
155 Emsg1(M_ABORT, 0, _("Unable to create job thread: ERR=%s\n"), strerror(stat));
158 /* Queue the job to be run */
159 if ((stat = workq_add(&job_wq, (void *)jcr, &work_item, 0)) != 0) {
160 Emsg1(M_ABORT, 0, _("Could not add job to work queue: ERR=%s\n"), strerror(stat));
162 jcr->work_item = work_item;
164 Dmsg0(200, "Done run_job()\n");
168 * This is the engine called by workq_add() when we were pulled
169 * from the work queue.
170 * At this point, we are running in our own thread
172 static void *job_thread(void *arg)
174 JCR *jcr = (JCR *)arg;
176 pthread_detach(pthread_self());
177 sm_check(__FILE__, __LINE__, True);
179 if (!acquire_resource_locks(jcr)) {
180 set_jcr_job_status(jcr, JS_Canceled);
183 Dmsg0(200, "=====Start Job=========\n");
184 jcr->start_time = time(NULL); /* set the real start time */
185 set_jcr_job_status(jcr, JS_Running);
187 if (job_canceled(jcr)) {
188 update_job_end_record(jcr);
189 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
190 (utime_t)(jcr->start_time - jcr->sched_time)) {
191 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
192 set_jcr_job_status(jcr, JS_Canceled);
193 update_job_end_record(jcr);
197 if (jcr->job->RunBeforeJob) {
198 POOLMEM *before = get_pool_memory(PM_FNAME);
201 before = edit_run_codes(jcr, before, jcr->job->RunBeforeJob);
202 status = run_program(before, 0, NULL);
204 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
206 set_jcr_job_status(jcr, JS_FatalError);
207 update_job_end_record(jcr);
208 free_pool_memory(before);
211 free_pool_memory(before);
213 switch (jcr->JobType) {
216 if (jcr->JobStatus == JS_Terminated) {
222 if (jcr->JobStatus == JS_Terminated) {
228 if (jcr->JobStatus == JS_Terminated) {
235 set_jcr_job_status(jcr, JS_Terminated);
238 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
241 if (jcr->job->RunAfterJob) {
242 POOLMEM *after = get_pool_memory(PM_FNAME);
245 after = edit_run_codes(jcr, after, jcr->job->RunAfterJob);
246 status = run_program(after, 0, NULL);
248 Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
250 set_jcr_job_status(jcr, JS_FatalError);
251 update_job_end_record(jcr);
253 free_pool_memory(after);
257 release_resource_locks(jcr);
258 Dmsg0(50, "Before free jcr\n");
260 Dmsg0(50, "======== End Job ==========\n");
261 sm_check(__FILE__, __LINE__, True);
266 * Acquire the resources needed. These locks limit the
267 * number of jobs by each resource. We have limits on
268 * Jobs, Clients, Storage, and total jobs.
270 static int acquire_resource_locks(JCR *jcr)
272 time_t now = time(NULL);
274 /* Wait until scheduled time arrives */
275 if (jcr->sched_time > now && verbose) {
276 Jmsg(jcr, M_INFO, 0, _("Waiting %d seconds for sched time.\n"),
277 jcr->sched_time - now);
279 while (jcr->sched_time > now) {
280 Dmsg2(100, "Waiting on sched time, jobid=%d secs=%d\n", jcr->JobId,
281 jcr->sched_time - now);
282 bmicrosleep(jcr->sched_time - now, 0);
290 /* Initialize semaphores */
291 if (jcr->store->sem.valid != SEMLOCK_VALID) {
292 if ((stat = sem_init(&jcr->store->sem, jcr->store->MaxConcurrentJobs)) != 0) {
293 Emsg1(M_ABORT, 0, _("Could not init Storage semaphore: ERR=%s\n"), strerror(stat));
296 if (jcr->client->sem.valid != SEMLOCK_VALID) {
297 if ((stat = sem_init(&jcr->client->sem, jcr->client->MaxConcurrentJobs)) != 0) {
298 Emsg1(M_ABORT, 0, _("Could not init Client semaphore: ERR=%s\n"), strerror(stat));
301 if (jcr->job->sem.valid != SEMLOCK_VALID) {
302 if ((stat = sem_init(&jcr->job->sem, jcr->job->MaxConcurrentJobs)) != 0) {
303 Emsg1(M_ABORT, 0, _("Could not init Job semaphore: ERR=%s\n"), strerror(stat));
308 /* Acquire semaphore */
309 set_jcr_job_status(jcr, JS_WaitJobRes);
310 if ((stat = sem_lock(&jcr->job->sem)) != 0) {
311 Emsg1(M_ABORT, 0, _("Could not acquire Job max jobs lock: ERR=%s\n"), strerror(stat));
313 set_jcr_job_status(jcr, JS_WaitClientRes);
314 if ((stat = sem_trylock(&jcr->client->sem)) != 0) {
316 backoff_resource_locks(jcr, 1);
319 Emsg1(M_ABORT, 0, _("Could not acquire Client max jobs lock: ERR=%s\n"), strerror(stat));
322 set_jcr_job_status(jcr, JS_WaitStoreRes);
323 if ((stat = sem_trylock(&jcr->store->sem)) != 0) {
325 backoff_resource_locks(jcr, 2);
328 Emsg1(M_ABORT, 0, _("Could not acquire Storage max jobs lock: ERR=%s\n"), strerror(stat));
331 set_jcr_job_status(jcr, JS_WaitMaxJobs);
332 if ((stat = sem_trylock(&job_lock)) != 0) {
334 backoff_resource_locks(jcr, 3);
337 Emsg1(M_ABORT, 0, _("Could not acquire max jobs lock: ERR=%s\n"), strerror(stat));
345 * Wait for a resource to be released either by backoff or
346 * by a job terminating.
349 pthread_cond_wait(&resource_wait, &mutex);
360 * We could not get all the resource locks because
361 * too many jobs are running, so release any locks
362 * we did acquire, giving others a chance to use them
365 static void backoff_resource_locks(JCR *jcr, int count)
370 sem_unlock(&jcr->store->sem);
371 /* Fall through wanted */
373 sem_unlock(&jcr->client->sem);
374 /* Fall through wanted */
376 sem_unlock(&jcr->job->sem);
380 * Since we released a lock, if there are any threads
381 * waiting, wake them up so that they can try again.
384 pthread_cond_broadcast(&resource_wait);
391 * This is called at the end of the job to release
392 * any resource limits on the number of jobs. If
393 * there are any other jobs waiting, we wake them
394 * up so that they can try again.
396 static void release_resource_locks(JCR *jcr)
400 sem_unlock(&jcr->store->sem);
401 sem_unlock(&jcr->client->sem);
402 sem_unlock(&jcr->job->sem);
403 sem_unlock(&job_lock);
405 pthread_cond_broadcast(&resource_wait);
412 * Get or create a Client record for this Job
414 int get_or_create_client_record(JCR *jcr)
418 memset(&cr, 0, sizeof(cr));
419 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
420 cr.AutoPrune = jcr->client->AutoPrune;
421 cr.FileRetention = jcr->client->FileRetention;
422 cr.JobRetention = jcr->client->JobRetention;
423 if (jcr->client_name) {
424 free_pool_memory(jcr->client_name);
426 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
427 strcpy(jcr->client_name, jcr->client->hdr.name);
428 if (!db_create_client_record(jcr, jcr->db, &cr)) {
429 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
430 db_strerror(jcr->db));
433 jcr->jr.ClientId = cr.ClientId;
435 if (jcr->client_uname) {
436 free_pool_memory(jcr->client_uname);
438 jcr->client_uname = get_memory(strlen(cr.Uname) + 1);
439 strcpy(jcr->client_uname, cr.Uname);
441 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
448 * Write status and such in DB
450 void update_job_end_record(JCR *jcr)
452 if (jcr->jr.EndTime == 0) {
453 jcr->jr.EndTime = time(NULL);
455 jcr->end_time = jcr->jr.EndTime;
456 jcr->jr.JobId = jcr->JobId;
457 jcr->jr.JobStatus = jcr->JobStatus;
458 jcr->jr.JobFiles = jcr->JobFiles;
459 jcr->jr.JobBytes = jcr->JobBytes;
460 jcr->jr.VolSessionId = jcr->VolSessionId;
461 jcr->jr.VolSessionTime = jcr->VolSessionTime;
462 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
463 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
464 db_strerror(jcr->db));
469 * Takes base_name and appends (unique) current
470 * date and time to form unique job name.
472 * Returns: unique job name in jcr->Job
473 * date/time in jcr->start_time
475 void create_unique_job_name(JCR *jcr, char *base_name)
477 /* Job start mutex */
478 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
479 static time_t last_start_time = 0;
482 char dt[MAX_TIME_LENGTH];
483 char name[MAX_NAME_LENGTH];
486 /* Guarantee unique start time -- maximum one per second, and
487 * thus unique Job Name
489 P(mutex); /* lock creation of jobs */
491 while (now == last_start_time) {
492 bmicrosleep(0, 500000);
495 last_start_time = now;
496 V(mutex); /* allow creation of jobs */
497 jcr->start_time = now;
498 /* Form Unique JobName */
499 localtime_r(&now, &tm);
500 /* Use only characters that are permitted in Windows filenames */
501 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
502 bstrncpy(name, base_name, sizeof(name));
503 name[sizeof(name)-22] = 0; /* truncate if too long */
504 sprintf(jcr->Job, "%s.%s", name, dt); /* add date & time */
505 /* Convert spaces into underscores */
506 for (p=jcr->Job; *p; p++) {
514 * Free the Job Control Record if no one is still using it.
515 * Called from main free_jcr() routine in src/lib/jcr.c so
516 * that we can do our Director specific cleanup of the jcr.
518 void dird_free_jcr(JCR *jcr)
520 Dmsg0(200, "Start dird free_jcr\n");
522 if (jcr->file_bsock) {
523 Dmsg0(200, "Close File bsock\n");
524 bnet_close(jcr->file_bsock);
526 if (jcr->store_bsock) {
527 Dmsg0(200, "Close Store bsock\n");
528 bnet_close(jcr->store_bsock);
531 Dmsg0(200, "Free JCR fname\n");
532 free_pool_memory(jcr->fname);
535 Dmsg0(200, "Free JCR stime\n");
536 free_pool_memory(jcr->stime);
539 Dmsg0(200, "Close DB\n");
540 db_close_database(jcr, jcr->db);
542 if (jcr->RestoreWhere) {
543 free(jcr->RestoreWhere);
545 if (jcr->RestoreBootstrap) {
546 free(jcr->RestoreBootstrap);
548 if (jcr->client_uname) {
549 free_pool_memory(jcr->client_uname);
551 Dmsg0(200, "End dird free_jcr\n");
555 * Set some defaults in the JCR necessary to
556 * run. These items are pulled from the job
557 * definition as defaults, but can be overridden
558 * later either by the Run record in the Schedule resource,
559 * or by the Console program.
561 void set_jcr_defaults(JCR *jcr, JOB *job)
564 jcr->JobType = job->JobType;
565 jcr->JobLevel = job->level;
566 jcr->store = job->storage;
567 jcr->client = job->client;
568 if (jcr->client_name) {
569 free_pool_memory(jcr->client_name);
571 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
572 strcpy(jcr->client_name, jcr->client->hdr.name);
573 jcr->pool = job->pool;
574 jcr->catalog = job->client->catalog;
575 jcr->fileset = job->fileset;
576 jcr->messages = job->messages;
577 if (jcr->RestoreBootstrap) {
578 free(jcr->RestoreBootstrap);
580 /* This can be overridden by Console program */
581 if (job->RestoreBootstrap) {
582 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
584 /* If no default level given, set one */
585 if (jcr->JobLevel == 0) {
586 switch (jcr->JobType) {
588 jcr->JobLevel = L_VERIFY_CATALOG;
591 jcr->JobLevel = L_INCREMENTAL;
600 * Edit codes into Run command
603 * %d = Director's name
611 * omsg = edited output message
612 * imsg = input string containing edit codes (%x)
615 static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg)
622 Dmsg1(200, "edit_run_codes: %s\n", imsg);
623 for (p=imsg; *p; p++) {
630 str = jcr->client_name;
639 str = job_status_to_str(jcr->JobStatus);
642 sprintf(add, "%d", jcr->JobId);
649 str = job_level_to_str(jcr->JobLevel);
652 str = jcr->job->hdr.name;
655 str = job_type_to_str(jcr->JobType);
669 Dmsg1(200, "add_str %s\n", str);
670 pm_strcat(&omsg, (char *)str);
671 Dmsg1(200, "omsg=%s\n", omsg);