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 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);
177 if (!acquire_resource_locks(jcr)) {
178 set_jcr_job_status(jcr, JS_Canceled);
181 Dmsg0(200, "=====Start Job=========\n");
182 jcr->start_time = time(NULL); /* set the real start time */
183 set_jcr_job_status(jcr, JS_Running);
185 if (job_canceled(jcr)) {
186 update_job_end_record(jcr);
187 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
188 (utime_t)(jcr->start_time - jcr->sched_time)) {
189 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
190 set_jcr_job_status(jcr, JS_Canceled);
191 update_job_end_record(jcr);
195 if (jcr->job->RunBeforeJob) {
196 POOLMEM *before = get_pool_memory(PM_FNAME);
199 before = edit_run_codes(jcr, before, jcr->job->RunBeforeJob);
200 status = run_program(before, 0, NULL);
202 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
204 set_jcr_job_status(jcr, JS_FatalError);
205 update_job_end_record(jcr);
206 free_pool_memory(before);
209 free_pool_memory(before);
211 switch (jcr->JobType) {
214 if (jcr->JobStatus == JS_Terminated) {
220 if (jcr->JobStatus == JS_Terminated) {
226 if (jcr->JobStatus == JS_Terminated) {
233 set_jcr_job_status(jcr, JS_Terminated);
236 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
239 if (jcr->job->RunAfterJob) {
240 POOLMEM *after = get_pool_memory(PM_FNAME);
243 after = edit_run_codes(jcr, after, jcr->job->RunAfterJob);
244 status = run_program(after, 0, NULL);
246 Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
248 set_jcr_job_status(jcr, JS_FatalError);
249 update_job_end_record(jcr);
251 free_pool_memory(after);
255 release_resource_locks(jcr);
256 Dmsg0(50, "Before free jcr\n");
258 Dmsg0(50, "======== End Job ==========\n");
259 sm_check(__FILE__, __LINE__, True);
264 * Acquire the resources needed. These locks limit the
265 * number of jobs by each resource. We have limits on
266 * Jobs, Clients, Storage, and total jobs.
268 static int acquire_resource_locks(JCR *jcr)
270 time_t now = time(NULL);
271 time_t wtime = jcr->sched_time - now;
273 /* Wait until scheduled time arrives */
274 if (wtime > 0 && verbose) {
275 Jmsg(jcr, M_INFO, 0, _("Waiting %d seconds for start time.\n"), wtime);
276 set_jcr_job_status(jcr, JS_WaitStartTime);
279 Dmsg2(100, "Waiting on sched time, jobid=%d secs=%d\n", jcr->JobId, wtime);
283 bmicrosleep(wtime, 0);
284 if (job_canceled(jcr)) {
287 wtime = jcr->sched_time - time(NULL);
295 /* Initialize semaphores */
296 if (jcr->store->sem.valid != SEMLOCK_VALID) {
297 if ((stat = sem_init(&jcr->store->sem, jcr->store->MaxConcurrentJobs)) != 0) {
298 Emsg1(M_ABORT, 0, _("Could not init Storage semaphore: ERR=%s\n"), strerror(stat));
301 if (jcr->client->sem.valid != SEMLOCK_VALID) {
302 if ((stat = sem_init(&jcr->client->sem, jcr->client->MaxConcurrentJobs)) != 0) {
303 Emsg1(M_ABORT, 0, _("Could not init Client semaphore: ERR=%s\n"), strerror(stat));
306 if (jcr->job->sem.valid != SEMLOCK_VALID) {
307 if ((stat = sem_init(&jcr->job->sem, jcr->job->MaxConcurrentJobs)) != 0) {
308 Emsg1(M_ABORT, 0, _("Could not init Job semaphore: ERR=%s\n"), strerror(stat));
313 /* Acquire semaphore */
314 set_jcr_job_status(jcr, JS_WaitJobRes);
315 if ((stat = sem_lock(&jcr->job->sem)) != 0) {
316 Emsg1(M_ABORT, 0, _("Could not acquire Job max jobs lock: ERR=%s\n"), strerror(stat));
318 set_jcr_job_status(jcr, JS_WaitClientRes);
319 if ((stat = sem_trylock(&jcr->client->sem)) != 0) {
321 backoff_resource_locks(jcr, 1);
324 Emsg1(M_ABORT, 0, _("Could not acquire Client max jobs lock: ERR=%s\n"), strerror(stat));
327 set_jcr_job_status(jcr, JS_WaitStoreRes);
328 if ((stat = sem_trylock(&jcr->store->sem)) != 0) {
330 backoff_resource_locks(jcr, 2);
333 Emsg1(M_ABORT, 0, _("Could not acquire Storage max jobs lock: ERR=%s\n"), strerror(stat));
336 set_jcr_job_status(jcr, JS_WaitMaxJobs);
337 if ((stat = sem_trylock(&job_lock)) != 0) {
339 backoff_resource_locks(jcr, 3);
342 Emsg1(M_ABORT, 0, _("Could not acquire max jobs lock: ERR=%s\n"), strerror(stat));
348 if (job_canceled(jcr)) {
353 * Wait for a resource to be released either by backoff or
354 * by a job terminating.
357 pthread_cond_wait(&resource_wait, &mutex);
368 * We could not get all the resource locks because
369 * too many jobs are running, so release any locks
370 * we did acquire, giving others a chance to use them
373 static void backoff_resource_locks(JCR *jcr, int count)
378 sem_unlock(&jcr->store->sem);
379 /* Fall through wanted */
381 sem_unlock(&jcr->client->sem);
382 /* Fall through wanted */
384 sem_unlock(&jcr->job->sem);
388 * Since we released a lock, if there are any threads
389 * waiting, wake them up so that they can try again.
392 pthread_cond_broadcast(&resource_wait);
399 * This is called at the end of the job to release
400 * any resource limits on the number of jobs. If
401 * there are any other jobs waiting, we wake them
402 * up so that they can try again.
404 static void release_resource_locks(JCR *jcr)
408 sem_unlock(&jcr->store->sem);
409 sem_unlock(&jcr->client->sem);
410 sem_unlock(&jcr->job->sem);
411 sem_unlock(&job_lock);
413 pthread_cond_broadcast(&resource_wait);
420 * Get or create a Client record for this Job
422 int get_or_create_client_record(JCR *jcr)
426 memset(&cr, 0, sizeof(cr));
427 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
428 cr.AutoPrune = jcr->client->AutoPrune;
429 cr.FileRetention = jcr->client->FileRetention;
430 cr.JobRetention = jcr->client->JobRetention;
431 if (jcr->client_name) {
432 free_pool_memory(jcr->client_name);
434 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
435 strcpy(jcr->client_name, jcr->client->hdr.name);
436 if (!db_create_client_record(jcr, jcr->db, &cr)) {
437 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
438 db_strerror(jcr->db));
441 jcr->jr.ClientId = cr.ClientId;
443 if (jcr->client_uname) {
444 free_pool_memory(jcr->client_uname);
446 jcr->client_uname = get_memory(strlen(cr.Uname) + 1);
447 strcpy(jcr->client_uname, cr.Uname);
449 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
456 * Write status and such in DB
458 void update_job_end_record(JCR *jcr)
460 if (jcr->jr.EndTime == 0) {
461 jcr->jr.EndTime = time(NULL);
463 jcr->end_time = jcr->jr.EndTime;
464 jcr->jr.JobId = jcr->JobId;
465 jcr->jr.JobStatus = jcr->JobStatus;
466 jcr->jr.JobFiles = jcr->JobFiles;
467 jcr->jr.JobBytes = jcr->JobBytes;
468 jcr->jr.VolSessionId = jcr->VolSessionId;
469 jcr->jr.VolSessionTime = jcr->VolSessionTime;
470 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
471 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
472 db_strerror(jcr->db));
477 * Takes base_name and appends (unique) current
478 * date and time to form unique job name.
480 * Returns: unique job name in jcr->Job
481 * date/time in jcr->start_time
483 void create_unique_job_name(JCR *jcr, char *base_name)
485 /* Job start mutex */
486 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
487 static time_t last_start_time = 0;
490 char dt[MAX_TIME_LENGTH];
491 char name[MAX_NAME_LENGTH];
494 /* Guarantee unique start time -- maximum one per second, and
495 * thus unique Job Name
497 P(mutex); /* lock creation of jobs */
499 while (now == last_start_time) {
500 bmicrosleep(0, 500000);
503 last_start_time = now;
504 V(mutex); /* allow creation of jobs */
505 jcr->start_time = now;
506 /* Form Unique JobName */
507 localtime_r(&now, &tm);
508 /* Use only characters that are permitted in Windows filenames */
509 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
510 bstrncpy(name, base_name, sizeof(name));
511 name[sizeof(name)-22] = 0; /* truncate if too long */
512 sprintf(jcr->Job, "%s.%s", name, dt); /* add date & time */
513 /* Convert spaces into underscores */
514 for (p=jcr->Job; *p; p++) {
522 * Free the Job Control Record if no one is still using it.
523 * Called from main free_jcr() routine in src/lib/jcr.c so
524 * that we can do our Director specific cleanup of the jcr.
526 void dird_free_jcr(JCR *jcr)
528 Dmsg0(200, "Start dird free_jcr\n");
530 if (jcr->file_bsock) {
531 Dmsg0(200, "Close File bsock\n");
532 bnet_close(jcr->file_bsock);
534 if (jcr->store_bsock) {
535 Dmsg0(200, "Close Store bsock\n");
536 bnet_close(jcr->store_bsock);
539 Dmsg0(200, "Free JCR fname\n");
540 free_pool_memory(jcr->fname);
543 Dmsg0(200, "Free JCR stime\n");
544 free_pool_memory(jcr->stime);
547 Dmsg0(200, "Close DB\n");
548 db_close_database(jcr, jcr->db);
550 if (jcr->RestoreWhere) {
551 free(jcr->RestoreWhere);
553 if (jcr->RestoreBootstrap) {
554 free(jcr->RestoreBootstrap);
556 if (jcr->client_uname) {
557 free_pool_memory(jcr->client_uname);
559 Dmsg0(200, "End dird free_jcr\n");
563 * Set some defaults in the JCR necessary to
564 * run. These items are pulled from the job
565 * definition as defaults, but can be overridden
566 * later either by the Run record in the Schedule resource,
567 * or by the Console program.
569 void set_jcr_defaults(JCR *jcr, JOB *job)
572 jcr->JobType = job->JobType;
573 jcr->JobLevel = job->level;
574 jcr->store = job->storage;
575 jcr->client = job->client;
576 if (jcr->client_name) {
577 free_pool_memory(jcr->client_name);
579 jcr->client_name = get_memory(strlen(jcr->client->hdr.name) + 1);
580 strcpy(jcr->client_name, jcr->client->hdr.name);
581 jcr->pool = job->pool;
582 jcr->catalog = job->client->catalog;
583 jcr->fileset = job->fileset;
584 jcr->messages = job->messages;
585 if (jcr->RestoreBootstrap) {
586 free(jcr->RestoreBootstrap);
588 /* This can be overridden by Console program */
589 if (job->RestoreBootstrap) {
590 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
592 /* If no default level given, set one */
593 if (jcr->JobLevel == 0) {
594 switch (jcr->JobType) {
596 jcr->JobLevel = L_VERIFY_CATALOG;
599 jcr->JobLevel = L_INCREMENTAL;
603 jcr->JobLevel = L_FULL;
612 * Edit codes into Run command
615 * %d = Director's name
623 * omsg = edited output message
624 * imsg = input string containing edit codes (%x)
627 static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg)
634 Dmsg1(200, "edit_run_codes: %s\n", imsg);
635 for (p=imsg; *p; p++) {
642 str = jcr->client_name;
651 str = job_status_to_str(jcr->JobStatus);
654 sprintf(add, "%d", jcr->JobId);
661 str = job_level_to_str(jcr->JobLevel);
664 str = jcr->job->hdr.name;
667 str = job_type_to_str(jcr->JobType);
681 Dmsg1(200, "add_str %s\n", str);
682 pm_strcat(&omsg, (char *)str);
683 Dmsg1(200, "omsg=%s\n", omsg);