3 * Bacula Director Job processing routines
5 * Kern Sibbald, October MM
10 Copyright (C) 2000-2004 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 void job_monitor_watchdog(watchdog_t *self);
35 static void job_monitor_destructor(watchdog_t *self);
36 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr);
37 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr);
39 /* Exported subroutines */
41 /* Imported subroutines */
42 extern void term_scheduler();
43 extern void term_ua_server();
44 extern int do_backup(JCR *jcr);
45 extern int do_admin(JCR *jcr);
46 extern int do_restore(JCR *jcr);
47 extern int do_verify(JCR *jcr);
49 /* Imported variables */
50 extern time_t watchdog_time;
54 void init_job_server(int max_workers)
59 if ((stat = jobq_init(&job_queue, max_workers, job_thread)) != 0) {
60 Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), strerror(stat));
62 if ((wd = new_watchdog()) == NULL) {
63 Emsg0(M_ABORT, 0, _("Could not init job monitor watchdogs\n"));
65 wd->callback = job_monitor_watchdog;
66 wd->destructor = job_monitor_destructor;
69 wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
70 register_watchdog(wd);
74 * Run a job -- typically called by the scheduler, but may also
75 * be called by the UA (Console program).
77 * Returns: 0 on failure
81 JobId_t run_job(JCR *jcr)
87 sm_check(__FILE__, __LINE__, true);
88 init_msg(jcr, jcr->messages);
90 /* Initialize termination condition variable */
91 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
92 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
95 jcr->term_wait_inited = true;
100 Dmsg0(50, "Open database\n");
101 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
102 jcr->catalog->db_password, jcr->catalog->db_address,
103 jcr->catalog->db_port, jcr->catalog->db_socket);
104 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
105 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
106 jcr->catalog->db_name);
108 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
112 Dmsg0(50, "DB opened\n");
117 create_unique_job_name(jcr, jcr->job->hdr.name);
118 set_jcr_job_status(jcr, JS_Created);
119 init_jcr_job_record(jcr);
120 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
121 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
124 JobId = jcr->JobId = jcr->jr.JobId;
126 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
127 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
128 Dmsg0(200, "Add jrc to work queue\n");
130 /* Queue the job to be run */
131 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
132 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
136 Dmsg0(100, "Done run_job()\n");
142 set_jcr_job_status(jcr, JS_ErrorTerminated);
150 * This is the engine called by jobq.c:jobq_add() when we were pulled
151 * from the work queue.
152 * At this point, we are running in our own thread and all
153 * necessary resources are allocated -- see jobq.c
155 static void *job_thread(void *arg)
157 JCR *jcr = (JCR *)arg;
159 jcr->my_thread_id = pthread_self();
160 pthread_detach(jcr->my_thread_id);
161 sm_check(__FILE__, __LINE__, true);
164 Dmsg0(200, "=====Start Job=========\n");
165 jcr->start_time = time(NULL); /* set the real start time */
166 set_jcr_job_status(jcr, JS_Running);
168 if (job_canceled(jcr)) {
169 update_job_end_record(jcr);
170 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
171 (utime_t)(jcr->start_time - jcr->sched_time)) {
172 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
173 set_jcr_job_status(jcr, JS_Canceled);
174 update_job_end_record(jcr);
178 if (jcr->job->RunBeforeJob) {
179 POOLMEM *before = get_pool_memory(PM_FNAME);
182 char line[MAXSTRING];
184 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
185 bpipe = open_bpipe(before, 0, "r");
186 free_pool_memory(before);
187 while (fgets(line, sizeof(line), bpipe->rfd)) {
188 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
190 status = close_bpipe(bpipe);
192 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
194 set_jcr_job_status(jcr, JS_FatalError);
195 update_job_end_record(jcr);
199 switch (jcr->JobType) {
202 if (jcr->JobStatus == JS_Terminated) {
208 if (jcr->JobStatus == JS_Terminated) {
214 if (jcr->JobStatus == JS_Terminated) {
220 if (jcr->JobStatus == JS_Terminated) {
225 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
228 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
229 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
230 POOLMEM *after = get_pool_memory(PM_FNAME);
233 char line[MAXSTRING];
235 if (jcr->JobStatus == JS_Terminated) {
236 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
238 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
240 bpipe = open_bpipe(after, 0, "r");
241 free_pool_memory(after);
242 while (fgets(line, sizeof(line), bpipe->rfd)) {
243 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
245 status = close_bpipe(bpipe);
247 * Note, if we get an error here, do not mark the
248 * job in error, simply report the error condition.
251 if (jcr->JobStatus == JS_Terminated) {
252 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
255 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
260 /* Send off any queued messages */
261 if (jcr->msg_queue->size() > 0) {
262 dequeue_messages(jcr);
269 Dmsg0(50, "======== End Job ==========\n");
270 sm_check(__FILE__, __LINE__, true);
276 * Cancel a job -- typically called by the UA (Console program), but may also
277 * be called by the job watchdog.
279 * Returns: 1 if cancel appears to be successful
280 * 0 on failure. Message sent to ua->jcr.
282 int cancel_job(UAContext *ua, JCR *jcr)
286 switch (jcr->JobStatus) {
289 case JS_WaitClientRes:
290 case JS_WaitStoreRes:
291 case JS_WaitPriority:
293 case JS_WaitStartTime:
294 set_jcr_job_status(jcr, JS_Canceled);
295 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
296 jcr->JobId, jcr->Job);
297 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
301 set_jcr_job_status(jcr, JS_Canceled);
303 /* Cancel File daemon */
304 if (jcr->file_bsock) {
305 ua->jcr->client = jcr->client;
306 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
307 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
310 Dmsg0(200, "Connected to file daemon\n");
311 fd = ua->jcr->file_bsock;
312 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
313 while (bnet_recv(fd) >= 0) {
314 bsendmsg(ua, "%s", fd->msg);
316 bnet_sig(fd, BNET_TERMINATE);
318 ua->jcr->file_bsock = NULL;
321 /* Cancel Storage daemon */
322 if (jcr->store_bsock) {
323 ua->jcr->store = jcr->store;
324 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
325 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
328 Dmsg0(200, "Connected to storage daemon\n");
329 sd = ua->jcr->store_bsock;
330 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
331 while (bnet_recv(sd) >= 0) {
332 bsendmsg(ua, "%s", sd->msg);
334 bnet_sig(sd, BNET_TERMINATE);
336 ua->jcr->store_bsock = NULL;
344 static void job_monitor_destructor(watchdog_t *self)
346 JCR *control_jcr = (JCR *) self->data;
348 free_jcr(control_jcr);
351 static void job_monitor_watchdog(watchdog_t *self)
353 JCR *control_jcr, *jcr;
355 control_jcr = (JCR *)self->data;
357 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
364 if (jcr->JobId == 0) {
365 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
367 /* Keep reference counts correct */
368 free_locked_jcr(jcr);
372 /* check MaxWaitTime */
373 cancel = job_check_maxwaittime(control_jcr, jcr);
375 /* check MaxRunTime */
376 cancel |= job_check_maxruntime(control_jcr, jcr);
379 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
380 jcr, jcr->JobId, jcr->Job);
382 UAContext *ua = new_ua_context(jcr);
383 ua->jcr = control_jcr;
387 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
390 /* Keep reference counts correct */
391 free_locked_jcr(jcr);
397 * Check if the maxwaittime has expired and it is possible
400 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
404 if (jcr->job->MaxWaitTime == 0) {
407 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
408 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
409 jcr, jcr->Job, jcr->job->MaxWaitTime);
412 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
414 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
415 switch (jcr->JobStatus) {
420 case JS_WaitStoreRes:
421 case JS_WaitClientRes:
423 case JS_WaitPriority:
425 case JS_WaitStartTime:
427 Dmsg0(200, "JCR blocked in #1\n");
430 Dmsg0(200, "JCR running, checking SD status\n");
431 switch (jcr->SDJobStatus) {
436 Dmsg0(200, "JCR blocked in #2\n");
439 Dmsg0(200, "JCR not blocked in #2\n");
444 case JS_ErrorTerminated:
447 Dmsg0(200, "JCR already dead in #3\n");
450 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
453 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
454 cancel ? "" : "do not ", jcr, jcr->job);
460 * Check if maxruntime has expired and if the job can be
463 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
467 if (jcr->job->MaxRunTime == 0) {
470 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
471 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
472 jcr, jcr->Job, jcr->job->MaxRunTime);
476 switch (jcr->JobStatus) {
482 case JS_WaitStoreRes:
483 case JS_WaitClientRes:
485 case JS_WaitPriority:
487 case JS_WaitStartTime:
492 case JS_ErrorTerminated:
498 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
502 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
503 cancel ? "" : "do not ", jcr, jcr->job);
510 * Get or create a Client record for this Job
512 bool get_or_create_client_record(JCR *jcr)
516 memset(&cr, 0, sizeof(cr));
517 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
518 cr.AutoPrune = jcr->client->AutoPrune;
519 cr.FileRetention = jcr->client->FileRetention;
520 cr.JobRetention = jcr->client->JobRetention;
521 if (!jcr->client_name) {
522 jcr->client_name = get_pool_memory(PM_NAME);
524 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
525 if (!db_create_client_record(jcr, jcr->db, &cr)) {
526 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
527 db_strerror(jcr->db));
530 jcr->jr.ClientId = cr.ClientId;
532 if (!jcr->client_uname) {
533 jcr->client_uname = get_pool_memory(PM_NAME);
535 pm_strcpy(&jcr->client_uname, cr.Uname);
537 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
542 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
545 * Get or Create FileSet record
547 memset(fsr, 0, sizeof(FILESET_DBR));
548 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
549 if (jcr->fileset->have_MD5) {
550 struct MD5Context md5c;
551 unsigned char signature[16];
552 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
553 MD5Final(signature, &md5c);
554 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
555 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
557 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
559 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
560 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
561 fsr->FileSet, db_strerror(jcr->db));
564 jcr->jr.FileSetId = fsr->FileSetId;
566 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
567 fsr->FileSet, fsr->cCreateTime);
569 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
574 void init_jcr_job_record(JCR *jcr)
576 jcr->jr.SchedTime = jcr->sched_time;
577 jcr->jr.StartTime = jcr->start_time;
578 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
579 jcr->jr.JobType = jcr->JobType;
580 jcr->jr.JobLevel = jcr->JobLevel;
581 jcr->jr.JobStatus = jcr->JobStatus;
582 jcr->jr.JobId = jcr->JobId;
583 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
584 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
588 * Write status and such in DB
590 void update_job_end_record(JCR *jcr)
592 jcr->jr.EndTime = time(NULL);
593 jcr->end_time = jcr->jr.EndTime;
594 jcr->jr.JobId = jcr->JobId;
595 jcr->jr.JobStatus = jcr->JobStatus;
596 jcr->jr.JobFiles = jcr->JobFiles;
597 jcr->jr.JobBytes = jcr->JobBytes;
598 jcr->jr.VolSessionId = jcr->VolSessionId;
599 jcr->jr.VolSessionTime = jcr->VolSessionTime;
600 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
601 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
602 db_strerror(jcr->db));
607 * Takes base_name and appends (unique) current
608 * date and time to form unique job name.
610 * Returns: unique job name in jcr->Job
611 * date/time in jcr->start_time
613 void create_unique_job_name(JCR *jcr, const char *base_name)
615 /* Job start mutex */
616 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
617 static time_t last_start_time = 0;
620 char dt[MAX_TIME_LENGTH];
621 char name[MAX_NAME_LENGTH];
624 /* Guarantee unique start time -- maximum one per second, and
625 * thus unique Job Name
627 P(mutex); /* lock creation of jobs */
629 while (now == last_start_time) {
630 bmicrosleep(0, 500000);
633 last_start_time = now;
634 V(mutex); /* allow creation of jobs */
635 jcr->start_time = now;
636 /* Form Unique JobName */
637 localtime_r(&now, &tm);
638 /* Use only characters that are permitted in Windows filenames */
639 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
640 bstrncpy(name, base_name, sizeof(name));
641 name[sizeof(name)-22] = 0; /* truncate if too long */
642 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
643 /* Convert spaces into underscores */
644 for (p=jcr->Job; *p; p++) {
652 * Free the Job Control Record if no one is still using it.
653 * Called from main free_jcr() routine in src/lib/jcr.c so
654 * that we can do our Director specific cleanup of the jcr.
656 void dird_free_jcr(JCR *jcr)
658 Dmsg0(200, "Start dird free_jcr\n");
660 if (jcr->sd_auth_key) {
661 free(jcr->sd_auth_key);
662 jcr->sd_auth_key = NULL;
668 if (jcr->file_bsock) {
669 Dmsg0(200, "Close File bsock\n");
670 bnet_close(jcr->file_bsock);
671 jcr->file_bsock = NULL;
673 if (jcr->store_bsock) {
674 Dmsg0(200, "Close Store bsock\n");
675 bnet_close(jcr->store_bsock);
676 jcr->store_bsock = NULL;
679 Dmsg0(200, "Free JCR fname\n");
680 free_pool_memory(jcr->fname);
684 Dmsg0(200, "Free JCR stime\n");
685 free_pool_memory(jcr->stime);
688 if (jcr->RestoreBootstrap) {
689 free(jcr->RestoreBootstrap);
690 jcr->RestoreBootstrap = NULL;
692 if (jcr->client_uname) {
693 free_pool_memory(jcr->client_uname);
694 jcr->client_uname = NULL;
696 if (jcr->term_wait_inited) {
697 pthread_cond_destroy(&jcr->term_wait);
699 jcr->job_end_push.destroy();
700 Dmsg0(200, "End dird free_jcr\n");
704 * Set some defaults in the JCR necessary to
705 * run. These items are pulled from the job
706 * definition as defaults, but can be overridden
707 * later either by the Run record in the Schedule resource,
708 * or by the Console program.
710 void set_jcr_defaults(JCR *jcr, JOB *job)
713 jcr->JobType = job->JobType;
714 switch (jcr->JobType) {
717 jcr->JobLevel = L_NONE;
720 jcr->JobLevel = job->level;
723 jcr->JobPriority = job->Priority;
724 jcr->store = job->storage;
725 jcr->client = job->client;
726 if (!jcr->client_name) {
727 jcr->client_name = get_pool_memory(PM_NAME);
729 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
730 jcr->pool = job->pool;
731 jcr->full_pool = job->full_pool;
732 jcr->inc_pool = job->inc_pool;
733 jcr->dif_pool = job->dif_pool;
734 jcr->catalog = job->client->catalog;
735 jcr->fileset = job->fileset;
736 jcr->messages = job->messages;
737 jcr->spool_data = job->spool_data;
738 if (jcr->RestoreBootstrap) {
739 free(jcr->RestoreBootstrap);
740 jcr->RestoreBootstrap = NULL;
742 /* This can be overridden by Console program */
743 if (job->RestoreBootstrap) {
744 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
746 /* If no default level given, set one */
747 if (jcr->JobLevel == 0) {
748 switch (jcr->JobType) {
750 jcr->JobLevel = L_VERIFY_CATALOG;
753 jcr->JobLevel = L_INCREMENTAL;
757 jcr->JobLevel = L_NONE;