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 jcr->jr.SchedTime = jcr->sched_time;
103 jcr->jr.StartTime = jcr->start_time;
104 jcr->jr.Type = jcr->JobType;
105 jcr->jr.Level = jcr->JobLevel;
106 jcr->jr.JobStatus = jcr->JobStatus;
107 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
108 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
110 /* Initialize termination condition variable */
111 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
112 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
113 set_jcr_job_status(jcr, JS_ErrorTerminated);
121 Dmsg0(50, "Open database\n");
122 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
123 jcr->catalog->db_password, jcr->catalog->db_address,
124 jcr->catalog->db_port, jcr->catalog->db_socket);
125 if (!db_open_database(jcr, jcr->db)) {
126 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
127 db_close_database(jcr, 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 db_close_database(jcr, jcr->db);
141 set_jcr_job_status(jcr, JS_ErrorTerminated);
145 jcr->JobId = jcr->jr.JobId;
146 ASSERT(jcr->jr.JobId > 0);
148 Dmsg4(30, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
149 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
150 Dmsg0(200, "Add jrc to work queue\n");
153 if ((stat = pthread_create(&tid, NULL, job_thread, (void *)jcr)) != 0) {
154 Emsg1(M_ABORT, 0, _("Unable to create job thread: ERR=%s\n"), strerror(stat));
157 /* Queue the job to be run */
158 if ((stat = workq_add(&job_wq, (void *)jcr, &work_item, 0)) != 0) {
159 Emsg1(M_ABORT, 0, _("Could not add job to work queue: ERR=%s\n"), strerror(stat));
161 jcr->work_item = work_item;
163 Dmsg0(200, "Done run_job()\n");
167 * This is the engine called by workq_add() when we were pulled
168 * from the work queue.
169 * At this point, we are running in our own thread
171 static void *job_thread(void *arg)
173 JCR *jcr = (JCR *)arg;
175 pthread_detach(pthread_self());
176 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 before = edit_run_codes(jcr, before, jcr->job->RunBeforeJob);
201 status = run_program(before, 0, NULL);
203 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
205 set_jcr_job_status(jcr, JS_FatalError);
206 update_job_end_record(jcr);
207 free_pool_memory(before);
210 free_pool_memory(before);
212 switch (jcr->JobType) {
215 if (jcr->JobStatus == JS_Terminated) {
221 if (jcr->JobStatus == JS_Terminated) {
227 if (jcr->JobStatus == JS_Terminated) {
234 set_jcr_job_status(jcr, JS_Terminated);
237 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
240 if (jcr->job->RunAfterJob) {
241 POOLMEM *after = get_pool_memory(PM_FNAME);
244 after = edit_run_codes(jcr, after, jcr->job->RunAfterJob);
245 status = run_program(after, 0, NULL);
247 Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
249 set_jcr_job_status(jcr, JS_FatalError);
250 update_job_end_record(jcr);
252 free_pool_memory(after);
256 release_resource_locks(jcr);
257 Dmsg0(50, "Before free jcr\n");
259 Dmsg0(50, "======== End Job ==========\n");
260 sm_check(__FILE__, __LINE__, True);
265 * Acquire the resources needed. These locks limit the
266 * number of jobs by each resource. We have limits on
267 * Jobs, Clients, Storage, and total jobs.
269 static int acquire_resource_locks(JCR *jcr)
271 time_t now = time(NULL);
273 /* Wait until scheduled time arrives */
274 while (jcr->sched_time > now) {
275 Dmsg2(100, "Waiting on sched time, jobid=%d secs=%d\n", jcr->JobId,
276 jcr->sched_time - now);
277 bmicrosleep(jcr->sched_time - now, 0);
285 /* Initialize semaphores */
286 if (jcr->store->sem.valid != SEMLOCK_VALID) {
287 if ((stat = sem_init(&jcr->store->sem, jcr->store->MaxConcurrentJobs)) != 0) {
288 Emsg1(M_ABORT, 0, _("Could not init Storage semaphore: ERR=%s\n"), strerror(stat));
291 if (jcr->client->sem.valid != SEMLOCK_VALID) {
292 if ((stat = sem_init(&jcr->client->sem, jcr->client->MaxConcurrentJobs)) != 0) {
293 Emsg1(M_ABORT, 0, _("Could not init Client semaphore: ERR=%s\n"), strerror(stat));
296 if (jcr->job->sem.valid != SEMLOCK_VALID) {
297 if ((stat = sem_init(&jcr->job->sem, jcr->job->MaxConcurrentJobs)) != 0) {
298 Emsg1(M_ABORT, 0, _("Could not init Job semaphore: ERR=%s\n"), strerror(stat));
303 /* Acquire semaphore */
304 set_jcr_job_status(jcr, JS_WaitJobRes);
305 if ((stat = sem_lock(&jcr->job->sem)) != 0) {
306 Emsg1(M_ABORT, 0, _("Could not acquire Job max jobs lock: ERR=%s\n"), strerror(stat));
308 set_jcr_job_status(jcr, JS_WaitClientRes);
309 if ((stat = sem_trylock(&jcr->client->sem)) != 0) {
311 backoff_resource_locks(jcr, 1);
314 Emsg1(M_ABORT, 0, _("Could not acquire Client max jobs lock: ERR=%s\n"), strerror(stat));
317 set_jcr_job_status(jcr, JS_WaitStoreRes);
318 if ((stat = sem_trylock(&jcr->store->sem)) != 0) {
320 backoff_resource_locks(jcr, 2);
323 Emsg1(M_ABORT, 0, _("Could not acquire Storage max jobs lock: ERR=%s\n"), strerror(stat));
326 set_jcr_job_status(jcr, JS_WaitMaxJobs);
327 if ((stat = sem_trylock(&job_lock)) != 0) {
329 backoff_resource_locks(jcr, 3);
332 Emsg1(M_ABORT, 0, _("Could not acquire max jobs lock: ERR=%s\n"), strerror(stat));
340 * Wait for a resource to be released either by backoff or
341 * by a job terminating.
344 pthread_cond_wait(&resource_wait, &mutex);
355 * We could not get all the resource locks because
356 * too many jobs are running, so release any locks
357 * we did acquire, giving others a chance to use them
360 static void backoff_resource_locks(JCR *jcr, int count)
365 sem_unlock(&jcr->store->sem);
366 /* Fall through wanted */
368 sem_unlock(&jcr->client->sem);
369 /* Fall through wanted */
371 sem_unlock(&jcr->job->sem);
375 * Since we released a lock, if there are any threads
376 * waiting, wake them up so that they can try again.
379 pthread_cond_broadcast(&resource_wait);
386 * This is called at the end of the job to release
387 * any resource limits on the number of jobs. If
388 * there are any other jobs waiting, we wake them
389 * up so that they can try again.
391 static void release_resource_locks(JCR *jcr)
395 sem_unlock(&jcr->store->sem);
396 sem_unlock(&jcr->client->sem);
397 sem_unlock(&jcr->job->sem);
398 sem_unlock(&job_lock);
400 pthread_cond_broadcast(&resource_wait);
407 * Get or create a Client record for this Job
409 int get_or_create_client_record(JCR *jcr)
413 memset(&cr, 0, sizeof(cr));
414 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
415 cr.AutoPrune = jcr->client->AutoPrune;
416 cr.FileRetention = jcr->client->FileRetention;
417 cr.JobRetention = jcr->client->JobRetention;
418 if (jcr->client_name) {
419 free_pool_memory(jcr->client_name);
421 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
422 strcpy(jcr->client_name, jcr->client->hdr.name);
423 if (!db_create_client_record(jcr, jcr->db, &cr)) {
424 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
425 db_strerror(jcr->db));
428 jcr->jr.ClientId = cr.ClientId;
430 if (jcr->client_uname) {
431 free_pool_memory(jcr->client_uname);
433 jcr->client_uname = get_memory(strlen(cr.Uname) + 1);
434 strcpy(jcr->client_uname, cr.Uname);
436 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
443 * Write status and such in DB
445 void update_job_end_record(JCR *jcr)
447 if (jcr->jr.EndTime == 0) {
448 jcr->jr.EndTime = time(NULL);
450 jcr->end_time = jcr->jr.EndTime;
451 jcr->jr.JobId = jcr->JobId;
452 jcr->jr.JobStatus = jcr->JobStatus;
453 jcr->jr.JobFiles = jcr->JobFiles;
454 jcr->jr.JobBytes = jcr->JobBytes;
455 jcr->jr.VolSessionId = jcr->VolSessionId;
456 jcr->jr.VolSessionTime = jcr->VolSessionTime;
457 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
458 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
459 db_strerror(jcr->db));
464 * Takes base_name and appends (unique) current
465 * date and time to form unique job name.
467 * Returns: unique job name in jcr->Job
468 * date/time in jcr->start_time
470 void create_unique_job_name(JCR *jcr, char *base_name)
472 /* Job start mutex */
473 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
474 static time_t last_start_time = 0;
477 char dt[MAX_TIME_LENGTH];
478 char name[MAX_NAME_LENGTH];
481 /* Guarantee unique start time -- maximum one per second, and
482 * thus unique Job Name
484 P(mutex); /* lock creation of jobs */
486 while (now == last_start_time) {
487 bmicrosleep(0, 500000);
490 last_start_time = now;
491 V(mutex); /* allow creation of jobs */
492 jcr->start_time = now;
493 /* Form Unique JobName */
494 localtime_r(&now, &tm);
495 /* Use only characters that are permitted in Windows filenames */
496 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
497 bstrncpy(name, base_name, sizeof(name));
498 name[sizeof(name)-22] = 0; /* truncate if too long */
499 sprintf(jcr->Job, "%s.%s", name, dt); /* add date & time */
500 /* Convert spaces into underscores */
501 for (p=jcr->Job; *p; p++) {
509 * Free the Job Control Record if no one is still using it.
510 * Called from main free_jcr() routine in src/lib/jcr.c so
511 * that we can do our Director specific cleanup of the jcr.
513 void dird_free_jcr(JCR *jcr)
515 Dmsg0(200, "Start dird free_jcr\n");
517 if (jcr->file_bsock) {
518 Dmsg0(200, "Close File bsock\n");
519 bnet_close(jcr->file_bsock);
521 if (jcr->store_bsock) {
522 Dmsg0(200, "Close Store bsock\n");
523 bnet_close(jcr->store_bsock);
526 Dmsg0(200, "Free JCR fname\n");
527 free_pool_memory(jcr->fname);
530 Dmsg0(200, "Free JCR stime\n");
531 free_pool_memory(jcr->stime);
534 Dmsg0(200, "Close DB\n");
535 db_close_database(jcr, jcr->db);
537 if (jcr->RestoreWhere) {
538 free(jcr->RestoreWhere);
540 if (jcr->RestoreBootstrap) {
541 free(jcr->RestoreBootstrap);
543 if (jcr->client_uname) {
544 free_pool_memory(jcr->client_uname);
546 Dmsg0(200, "End dird free_jcr\n");
550 * Set some defaults in the JCR necessary to
551 * run. These items are pulled from the job
552 * definition as defaults, but can be overridden
553 * later either by the Run record in the Schedule resource,
554 * or by the Console program.
556 void set_jcr_defaults(JCR *jcr, JOB *job)
559 jcr->JobType = job->JobType;
560 jcr->JobLevel = job->level;
561 jcr->store = job->storage;
562 jcr->client = job->client;
563 if (jcr->client_name) {
564 free_pool_memory(jcr->client_name);
566 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
567 strcpy(jcr->client_name, jcr->client->hdr.name);
568 jcr->pool = job->pool;
569 jcr->catalog = job->client->catalog;
570 jcr->fileset = job->fileset;
571 jcr->messages = job->messages;
572 if (jcr->RestoreBootstrap) {
573 free(jcr->RestoreBootstrap);
575 /* This can be overridden by Console program */
576 if (job->RestoreBootstrap) {
577 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
579 /* If no default level given, set one */
580 if (jcr->JobLevel == 0) {
581 switch (jcr->JobType) {
583 jcr->JobLevel = L_VERIFY_CATALOG;
586 jcr->JobLevel = L_INCREMENTAL;
595 * Edit codes into Run command
598 * %d = Director's name
606 * omsg = edited output message
607 * imsg = input string containing edit codes (%x)
610 static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg)
617 Dmsg1(200, "edit_run_codes: %s\n", imsg);
618 for (p=imsg; *p; p++) {
625 str = jcr->client_name;
634 str = job_status_to_str(jcr->JobStatus);
637 sprintf(add, "%d", jcr->JobId);
644 str = job_level_to_str(jcr->JobLevel);
647 str = jcr->job->hdr.name;
650 str = job_type_to_str(jcr->JobType);
664 Dmsg1(200, "add_str %s\n", str);
665 pm_strcat(&omsg, (char *)str);
666 Dmsg1(200, "omsg=%s\n", omsg);