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 bool do_mac(JCR *jcr);
46 extern int do_admin(JCR *jcr);
47 extern int do_restore(JCR *jcr);
48 extern bool do_verify(JCR *jcr);
50 /* Imported variables */
51 extern time_t watchdog_time;
55 void init_job_server(int max_workers)
60 if ((stat = jobq_init(&job_queue, max_workers, job_thread)) != 0) {
62 Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror(stat));
64 if ((wd = new_watchdog()) == NULL) {
65 Emsg0(M_ABORT, 0, _("Could not init job monitor watchdogs\n"));
67 wd->callback = job_monitor_watchdog;
68 wd->destructor = job_monitor_destructor;
71 wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
72 register_watchdog(wd);
75 void term_job_server()
77 jobq_destroy(&job_queue); /* ignore any errors */
81 * Run a job -- typically called by the scheduler, but may also
82 * be called by the UA (Console program).
84 * Returns: 0 on failure
88 JobId_t run_job(JCR *jcr)
94 sm_check(__FILE__, __LINE__, true);
95 init_msg(jcr, jcr->messages);
97 /* Initialize termination condition variable */
98 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
100 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.strerror(errstat));
103 jcr->term_wait_inited = true;
108 Dmsg0(50, "Open database\n");
109 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
110 jcr->catalog->db_password, jcr->catalog->db_address,
111 jcr->catalog->db_port, jcr->catalog->db_socket,
112 jcr->catalog->mult_db_connections);
113 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
114 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
115 jcr->catalog->db_name);
117 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
121 Dmsg0(50, "DB opened\n");
126 create_unique_job_name(jcr, jcr->job->hdr.name);
127 set_jcr_job_status(jcr, JS_Created);
128 init_jcr_job_record(jcr);
129 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
130 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
133 JobId = jcr->JobId = jcr->jr.JobId;
135 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
136 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
137 Dmsg0(200, "Add jrc to work queue\n");
139 /* Queue the job to be run */
140 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
142 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror(stat));
146 Dmsg0(100, "Done run_job()\n");
152 set_jcr_job_status(jcr, JS_ErrorTerminated);
160 * This is the engine called by jobq.c:jobq_add() when we were pulled
161 * from the work queue.
162 * At this point, we are running in our own thread and all
163 * necessary resources are allocated -- see jobq.c
165 static void *job_thread(void *arg)
167 JCR *jcr = (JCR *)arg;
169 jcr->my_thread_id = pthread_self();
170 pthread_detach(jcr->my_thread_id);
171 sm_check(__FILE__, __LINE__, true);
174 Dmsg0(200, "=====Start Job=========\n");
175 jcr->start_time = time(NULL); /* set the real start time */
176 jcr->jr.StartTime = jcr->start_time;
177 set_jcr_job_status(jcr, JS_Running);
178 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
179 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
182 if (job_canceled(jcr)) {
183 update_job_end_record(jcr);
184 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
185 (utime_t)(jcr->start_time - jcr->sched_time)) {
186 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
187 set_jcr_job_status(jcr, JS_Canceled);
188 update_job_end_record(jcr);
192 if (jcr->job->RunBeforeJob) {
193 POOLMEM *before = get_pool_memory(PM_FNAME);
196 char line[MAXSTRING];
198 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
199 bpipe = open_bpipe(before, 0, "r");
200 free_pool_memory(before);
201 while (fgets(line, sizeof(line), bpipe->rfd)) {
202 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
204 status = close_bpipe(bpipe);
207 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror(status));
208 set_jcr_job_status(jcr, JS_FatalError);
209 update_job_end_record(jcr);
213 switch (jcr->JobType) {
216 if (jcr->JobStatus == JS_Terminated) {
222 if (jcr->JobStatus == JS_Terminated) {
228 if (jcr->JobStatus == JS_Terminated) {
234 if (jcr->JobStatus == JS_Terminated) {
241 do_mac(jcr); /* migration, archive, copy */
242 if (jcr->JobStatus == JS_Terminated) {
247 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
250 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
251 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
252 POOLMEM *after = get_pool_memory(PM_FNAME);
255 char line[MAXSTRING];
257 if (jcr->JobStatus == JS_Terminated) {
258 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
260 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
262 bpipe = open_bpipe(after, 0, "r");
263 free_pool_memory(after);
264 while (fgets(line, sizeof(line), bpipe->rfd)) {
265 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
267 status = close_bpipe(bpipe);
269 * Note, if we get an error here, do not mark the
270 * job in error, simply report the error condition.
274 if (jcr->JobStatus == JS_Terminated) {
275 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror(status));
277 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror(status));
281 /* Send off any queued messages */
282 if (jcr->msg_queue->size() > 0) {
283 dequeue_messages(jcr);
290 Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
291 sm_check(__FILE__, __LINE__, true);
297 * Cancel a job -- typically called by the UA (Console program), but may also
298 * be called by the job watchdog.
300 * Returns: 1 if cancel appears to be successful
301 * 0 on failure. Message sent to ua->jcr.
303 int cancel_job(UAContext *ua, JCR *jcr)
307 switch (jcr->JobStatus) {
310 case JS_WaitClientRes:
311 case JS_WaitStoreRes:
312 case JS_WaitPriority:
314 case JS_WaitStartTime:
315 set_jcr_job_status(jcr, JS_Canceled);
316 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
317 jcr->JobId, jcr->Job);
318 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
322 set_jcr_job_status(jcr, JS_Canceled);
324 /* Cancel File daemon */
325 if (jcr->file_bsock) {
326 ua->jcr->client = jcr->client;
327 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
328 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
331 Dmsg0(200, "Connected to file daemon\n");
332 fd = ua->jcr->file_bsock;
333 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
334 while (bnet_recv(fd) >= 0) {
335 bsendmsg(ua, "%s", fd->msg);
337 bnet_sig(fd, BNET_TERMINATE);
339 ua->jcr->file_bsock = NULL;
342 /* Cancel Storage daemon */
343 if (jcr->store_bsock) {
344 if (!ua->jcr->storage[0]) {
345 copy_storage(ua->jcr, jcr);
347 ua->jcr->store = jcr->store;
349 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
350 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
353 Dmsg0(200, "Connected to storage daemon\n");
354 sd = ua->jcr->store_bsock;
355 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
356 while (bnet_recv(sd) >= 0) {
357 bsendmsg(ua, "%s", sd->msg);
359 bnet_sig(sd, BNET_TERMINATE);
361 ua->jcr->store_bsock = NULL;
369 static void job_monitor_destructor(watchdog_t *self)
371 JCR *control_jcr = (JCR *) self->data;
373 free_jcr(control_jcr);
376 static void job_monitor_watchdog(watchdog_t *self)
378 JCR *control_jcr, *jcr;
380 control_jcr = (JCR *)self->data;
382 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
389 if (jcr->JobId == 0) {
390 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
392 /* Keep reference counts correct */
393 free_locked_jcr(jcr);
397 /* check MaxWaitTime */
398 cancel = job_check_maxwaittime(control_jcr, jcr);
400 /* check MaxRunTime */
401 cancel |= job_check_maxruntime(control_jcr, jcr);
404 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
405 jcr, jcr->JobId, jcr->Job);
407 UAContext *ua = new_ua_context(jcr);
408 ua->jcr = control_jcr;
412 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
415 /* Keep reference counts correct */
416 free_locked_jcr(jcr);
422 * Check if the maxwaittime has expired and it is possible
425 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
429 if (jcr->job->MaxWaitTime == 0) {
432 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
433 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
434 jcr, jcr->Job, jcr->job->MaxWaitTime);
437 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
439 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
440 switch (jcr->JobStatus) {
445 case JS_WaitStoreRes:
446 case JS_WaitClientRes:
448 case JS_WaitPriority:
450 case JS_WaitStartTime:
452 Dmsg0(200, "JCR blocked in #1\n");
455 Dmsg0(200, "JCR running, checking SD status\n");
456 switch (jcr->SDJobStatus) {
461 Dmsg0(200, "JCR blocked in #2\n");
464 Dmsg0(200, "JCR not blocked in #2\n");
469 case JS_ErrorTerminated:
472 Dmsg0(200, "JCR already dead in #3\n");
475 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
478 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
479 cancel ? "" : "do not ", jcr, jcr->job);
485 * Check if maxruntime has expired and if the job can be
488 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
492 if (jcr->job->MaxRunTime == 0) {
495 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
496 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
497 jcr, jcr->Job, jcr->job->MaxRunTime);
501 switch (jcr->JobStatus) {
507 case JS_WaitStoreRes:
508 case JS_WaitClientRes:
510 case JS_WaitPriority:
512 case JS_WaitStartTime:
517 case JS_ErrorTerminated:
523 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
527 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
528 cancel ? "" : "do not ", jcr, jcr->job);
535 * Get or create a Client record for this Job
537 bool get_or_create_client_record(JCR *jcr)
541 memset(&cr, 0, sizeof(cr));
542 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
543 cr.AutoPrune = jcr->client->AutoPrune;
544 cr.FileRetention = jcr->client->FileRetention;
545 cr.JobRetention = jcr->client->JobRetention;
546 if (!jcr->client_name) {
547 jcr->client_name = get_pool_memory(PM_NAME);
549 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
550 if (!db_create_client_record(jcr, jcr->db, &cr)) {
551 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
552 db_strerror(jcr->db));
555 jcr->jr.ClientId = cr.ClientId;
557 if (!jcr->client_uname) {
558 jcr->client_uname = get_pool_memory(PM_NAME);
560 pm_strcpy(jcr->client_uname, cr.Uname);
562 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
567 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
570 * Get or Create FileSet record
572 memset(fsr, 0, sizeof(FILESET_DBR));
573 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
574 if (jcr->fileset->have_MD5) {
575 struct MD5Context md5c;
576 unsigned char signature[16];
577 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
578 MD5Final(signature, &md5c);
579 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
580 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
582 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
584 if (!jcr->fileset->ignore_fs_changes ||
585 !db_get_fileset_record(jcr, jcr->db, fsr)) {
586 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
587 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
588 fsr->FileSet, db_strerror(jcr->db));
592 jcr->jr.FileSetId = fsr->FileSetId;
594 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
595 fsr->FileSet, fsr->cCreateTime);
597 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
602 void init_jcr_job_record(JCR *jcr)
604 jcr->jr.SchedTime = jcr->sched_time;
605 jcr->jr.StartTime = jcr->start_time;
606 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
607 jcr->jr.JobType = jcr->JobType;
608 jcr->jr.JobLevel = jcr->JobLevel;
609 jcr->jr.JobStatus = jcr->JobStatus;
610 jcr->jr.JobId = jcr->JobId;
611 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
612 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
616 * Write status and such in DB
618 void update_job_end_record(JCR *jcr)
620 jcr->jr.EndTime = time(NULL);
621 jcr->end_time = jcr->jr.EndTime;
622 jcr->jr.JobId = jcr->JobId;
623 jcr->jr.JobStatus = jcr->JobStatus;
624 jcr->jr.JobFiles = jcr->JobFiles;
625 jcr->jr.JobBytes = jcr->JobBytes;
626 jcr->jr.VolSessionId = jcr->VolSessionId;
627 jcr->jr.VolSessionTime = jcr->VolSessionTime;
628 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
629 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
630 db_strerror(jcr->db));
635 * Takes base_name and appends (unique) current
636 * date and time to form unique job name.
638 * Returns: unique job name in jcr->Job
639 * date/time in jcr->start_time
641 void create_unique_job_name(JCR *jcr, const char *base_name)
643 /* Job start mutex */
644 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
645 static time_t last_start_time = 0;
648 char dt[MAX_TIME_LENGTH];
649 char name[MAX_NAME_LENGTH];
652 /* Guarantee unique start time -- maximum one per second, and
653 * thus unique Job Name
655 P(mutex); /* lock creation of jobs */
657 while (now == last_start_time) {
658 bmicrosleep(0, 500000);
661 last_start_time = now;
662 V(mutex); /* allow creation of jobs */
663 jcr->start_time = now;
664 /* Form Unique JobName */
665 localtime_r(&now, &tm);
666 /* Use only characters that are permitted in Windows filenames */
667 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
668 bstrncpy(name, base_name, sizeof(name));
669 name[sizeof(name)-22] = 0; /* truncate if too long */
670 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
671 /* Convert spaces into underscores */
672 for (p=jcr->Job; *p; p++) {
680 * Free the Job Control Record if no one is still using it.
681 * Called from main free_jcr() routine in src/lib/jcr.c so
682 * that we can do our Director specific cleanup of the jcr.
684 void dird_free_jcr(JCR *jcr)
686 Dmsg0(200, "Start dird free_jcr\n");
688 if (jcr->sd_auth_key) {
689 free(jcr->sd_auth_key);
690 jcr->sd_auth_key = NULL;
696 if (jcr->file_bsock) {
697 Dmsg0(200, "Close File bsock\n");
698 bnet_close(jcr->file_bsock);
699 jcr->file_bsock = NULL;
701 if (jcr->store_bsock) {
702 Dmsg0(200, "Close Store bsock\n");
703 bnet_close(jcr->store_bsock);
704 jcr->store_bsock = NULL;
707 Dmsg0(200, "Free JCR fname\n");
708 free_pool_memory(jcr->fname);
712 Dmsg0(200, "Free JCR stime\n");
713 free_pool_memory(jcr->stime);
716 if (jcr->RestoreBootstrap) {
717 free(jcr->RestoreBootstrap);
718 jcr->RestoreBootstrap = NULL;
720 if (jcr->client_uname) {
721 free_pool_memory(jcr->client_uname);
722 jcr->client_uname = NULL;
724 if (jcr->term_wait_inited) {
725 pthread_cond_destroy(&jcr->term_wait);
727 jcr->job_end_push.destroy();
728 Dmsg0(200, "End dird free_jcr\n");
732 * Set some defaults in the JCR necessary to
733 * run. These items are pulled from the job
734 * definition as defaults, but can be overridden
735 * later either by the Run record in the Schedule resource,
736 * or by the Console program.
738 void set_jcr_defaults(JCR *jcr, JOB *job)
741 jcr->JobType = job->JobType;
742 switch (jcr->JobType) {
745 jcr->JobLevel = L_NONE;
748 jcr->JobLevel = job->JobLevel;
751 jcr->JobPriority = job->Priority;
752 for (int i=0; i<MAX_STORE; i++) {
753 jcr->storage[i] = job->storage[i];
755 if (!jcr->store && jcr->storage[0]) {
756 jcr->store = (STORE *)jcr->storage[0]->first();
758 jcr->client = job->client;
759 if (!jcr->client_name) {
760 jcr->client_name = get_pool_memory(PM_NAME);
762 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
763 jcr->pool = job->pool;
764 jcr->full_pool = job->full_pool;
765 jcr->inc_pool = job->inc_pool;
766 jcr->dif_pool = job->dif_pool;
767 jcr->catalog = job->client->catalog;
768 jcr->fileset = job->fileset;
769 jcr->messages = job->messages;
770 jcr->spool_data = job->spool_data;
771 if (jcr->RestoreBootstrap) {
772 free(jcr->RestoreBootstrap);
773 jcr->RestoreBootstrap = NULL;
775 /* This can be overridden by Console program */
776 if (job->RestoreBootstrap) {
777 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
779 /* This can be overridden by Console program */
780 jcr->verify_job = job->verify_job;
781 /* If no default level given, set one */
782 if (jcr->JobLevel == 0) {
783 switch (jcr->JobType) {
785 jcr->JobLevel = L_VERIFY_CATALOG;
788 jcr->JobLevel = L_INCREMENTAL;
792 jcr->JobLevel = L_NONE;
801 * copy the storage definitions from an old JCR to a new one
803 void copy_storage(JCR *new_jcr, JCR *old_jcr)
805 for (int i=0; i < MAX_STORE; i++) {
806 if (old_jcr->storage[i]) {
808 new_jcr->storage[i] = New(alist(10, not_owned_by_alist));
809 foreach_alist(st, old_jcr->storage[i]) {
810 new_jcr->storage[i]->append(st);
813 if (old_jcr->store) {
814 new_jcr->store = old_jcr->store;
815 } else if (new_jcr->storage[0]) {
816 new_jcr->store = (STORE *)new_jcr->storage[0]->first();