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) {
61 Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror(stat));
63 if ((wd = new_watchdog()) == NULL) {
64 Emsg0(M_ABORT, 0, _("Could not init job monitor watchdogs\n"));
66 wd->callback = job_monitor_watchdog;
67 wd->destructor = job_monitor_destructor;
70 wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
71 register_watchdog(wd);
74 void term_job_server()
76 jobq_destroy(&job_queue); /* ignore any errors */
80 * Run a job -- typically called by the scheduler, but may also
81 * be called by the UA (Console program).
83 * Returns: 0 on failure
87 JobId_t run_job(JCR *jcr)
93 sm_check(__FILE__, __LINE__, true);
94 init_msg(jcr, jcr->messages);
96 /* Initialize termination condition variable */
97 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
99 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.strerror(errstat));
102 jcr->term_wait_inited = true;
107 Dmsg0(50, "Open database\n");
108 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
109 jcr->catalog->db_password, jcr->catalog->db_address,
110 jcr->catalog->db_port, jcr->catalog->db_socket);
111 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
112 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
113 jcr->catalog->db_name);
115 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
119 Dmsg0(50, "DB opened\n");
124 create_unique_job_name(jcr, jcr->job->hdr.name);
125 set_jcr_job_status(jcr, JS_Created);
126 init_jcr_job_record(jcr);
127 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
128 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
131 JobId = jcr->JobId = jcr->jr.JobId;
133 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
134 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
135 Dmsg0(200, "Add jrc to work queue\n");
137 /* Queue the job to be run */
138 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
140 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror(stat));
144 Dmsg0(100, "Done run_job()\n");
150 set_jcr_job_status(jcr, JS_ErrorTerminated);
158 * This is the engine called by jobq.c:jobq_add() when we were pulled
159 * from the work queue.
160 * At this point, we are running in our own thread and all
161 * necessary resources are allocated -- see jobq.c
163 static void *job_thread(void *arg)
165 JCR *jcr = (JCR *)arg;
167 jcr->my_thread_id = pthread_self();
168 pthread_detach(jcr->my_thread_id);
169 sm_check(__FILE__, __LINE__, true);
172 Dmsg0(200, "=====Start Job=========\n");
173 jcr->start_time = time(NULL); /* set the real start time */
174 set_jcr_job_status(jcr, JS_Running);
176 if (job_canceled(jcr)) {
177 update_job_end_record(jcr);
178 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
179 (utime_t)(jcr->start_time - jcr->sched_time)) {
180 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
181 set_jcr_job_status(jcr, JS_Canceled);
182 update_job_end_record(jcr);
186 if (jcr->job->RunBeforeJob) {
187 POOLMEM *before = get_pool_memory(PM_FNAME);
190 char line[MAXSTRING];
192 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
193 bpipe = open_bpipe(before, 0, "r");
194 free_pool_memory(before);
195 while (fgets(line, sizeof(line), bpipe->rfd)) {
196 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
198 status = close_bpipe(bpipe);
201 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror(status));
202 set_jcr_job_status(jcr, JS_FatalError);
203 update_job_end_record(jcr);
207 switch (jcr->JobType) {
210 if (jcr->JobStatus == JS_Terminated) {
216 if (jcr->JobStatus == JS_Terminated) {
222 if (jcr->JobStatus == JS_Terminated) {
228 if (jcr->JobStatus == JS_Terminated) {
233 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
236 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
237 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
238 POOLMEM *after = get_pool_memory(PM_FNAME);
241 char line[MAXSTRING];
243 if (jcr->JobStatus == JS_Terminated) {
244 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
246 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
248 bpipe = open_bpipe(after, 0, "r");
249 free_pool_memory(after);
250 while (fgets(line, sizeof(line), bpipe->rfd)) {
251 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
253 status = close_bpipe(bpipe);
255 * Note, if we get an error here, do not mark the
256 * job in error, simply report the error condition.
260 if (jcr->JobStatus == JS_Terminated) {
261 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror(status));
263 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror(status));
267 /* Send off any queued messages */
268 if (jcr->msg_queue->size() > 0) {
269 dequeue_messages(jcr);
276 Dmsg0(50, "======== End Job ==========\n");
277 sm_check(__FILE__, __LINE__, true);
283 * Cancel a job -- typically called by the UA (Console program), but may also
284 * be called by the job watchdog.
286 * Returns: 1 if cancel appears to be successful
287 * 0 on failure. Message sent to ua->jcr.
289 int cancel_job(UAContext *ua, JCR *jcr)
293 switch (jcr->JobStatus) {
296 case JS_WaitClientRes:
297 case JS_WaitStoreRes:
298 case JS_WaitPriority:
300 case JS_WaitStartTime:
301 set_jcr_job_status(jcr, JS_Canceled);
302 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
303 jcr->JobId, jcr->Job);
304 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
308 set_jcr_job_status(jcr, JS_Canceled);
310 /* Cancel File daemon */
311 if (jcr->file_bsock) {
312 ua->jcr->client = jcr->client;
313 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
314 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
317 Dmsg0(200, "Connected to file daemon\n");
318 fd = ua->jcr->file_bsock;
319 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
320 while (bnet_recv(fd) >= 0) {
321 bsendmsg(ua, "%s", fd->msg);
323 bnet_sig(fd, BNET_TERMINATE);
325 ua->jcr->file_bsock = NULL;
328 /* Cancel Storage daemon */
329 if (jcr->store_bsock) {
330 ua->jcr->store = jcr->store;
331 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
332 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
335 Dmsg0(200, "Connected to storage daemon\n");
336 sd = ua->jcr->store_bsock;
337 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
338 while (bnet_recv(sd) >= 0) {
339 bsendmsg(ua, "%s", sd->msg);
341 bnet_sig(sd, BNET_TERMINATE);
343 ua->jcr->store_bsock = NULL;
351 static void job_monitor_destructor(watchdog_t *self)
353 JCR *control_jcr = (JCR *) self->data;
355 free_jcr(control_jcr);
358 static void job_monitor_watchdog(watchdog_t *self)
360 JCR *control_jcr, *jcr;
362 control_jcr = (JCR *)self->data;
364 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
371 if (jcr->JobId == 0) {
372 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
374 /* Keep reference counts correct */
375 free_locked_jcr(jcr);
379 /* check MaxWaitTime */
380 cancel = job_check_maxwaittime(control_jcr, jcr);
382 /* check MaxRunTime */
383 cancel |= job_check_maxruntime(control_jcr, jcr);
386 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
387 jcr, jcr->JobId, jcr->Job);
389 UAContext *ua = new_ua_context(jcr);
390 ua->jcr = control_jcr;
394 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
397 /* Keep reference counts correct */
398 free_locked_jcr(jcr);
404 * Check if the maxwaittime has expired and it is possible
407 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
411 if (jcr->job->MaxWaitTime == 0) {
414 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
415 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
416 jcr, jcr->Job, jcr->job->MaxWaitTime);
419 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
421 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
422 switch (jcr->JobStatus) {
427 case JS_WaitStoreRes:
428 case JS_WaitClientRes:
430 case JS_WaitPriority:
432 case JS_WaitStartTime:
434 Dmsg0(200, "JCR blocked in #1\n");
437 Dmsg0(200, "JCR running, checking SD status\n");
438 switch (jcr->SDJobStatus) {
443 Dmsg0(200, "JCR blocked in #2\n");
446 Dmsg0(200, "JCR not blocked in #2\n");
451 case JS_ErrorTerminated:
454 Dmsg0(200, "JCR already dead in #3\n");
457 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
460 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
461 cancel ? "" : "do not ", jcr, jcr->job);
467 * Check if maxruntime has expired and if the job can be
470 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
474 if (jcr->job->MaxRunTime == 0) {
477 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
478 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
479 jcr, jcr->Job, jcr->job->MaxRunTime);
483 switch (jcr->JobStatus) {
489 case JS_WaitStoreRes:
490 case JS_WaitClientRes:
492 case JS_WaitPriority:
494 case JS_WaitStartTime:
499 case JS_ErrorTerminated:
505 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
509 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
510 cancel ? "" : "do not ", jcr, jcr->job);
517 * Get or create a Client record for this Job
519 bool get_or_create_client_record(JCR *jcr)
523 memset(&cr, 0, sizeof(cr));
524 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
525 cr.AutoPrune = jcr->client->AutoPrune;
526 cr.FileRetention = jcr->client->FileRetention;
527 cr.JobRetention = jcr->client->JobRetention;
528 if (!jcr->client_name) {
529 jcr->client_name = get_pool_memory(PM_NAME);
531 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
532 if (!db_create_client_record(jcr, jcr->db, &cr)) {
533 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
534 db_strerror(jcr->db));
537 jcr->jr.ClientId = cr.ClientId;
539 if (!jcr->client_uname) {
540 jcr->client_uname = get_pool_memory(PM_NAME);
542 pm_strcpy(jcr->client_uname, cr.Uname);
544 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
549 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
552 * Get or Create FileSet record
554 memset(fsr, 0, sizeof(FILESET_DBR));
555 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
556 if (jcr->fileset->have_MD5) {
557 struct MD5Context md5c;
558 unsigned char signature[16];
559 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
560 MD5Final(signature, &md5c);
561 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
562 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
564 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
566 if (!jcr->fileset->ignore_fs_changes ||
567 !db_get_fileset_record(jcr, jcr->db, fsr)) {
568 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
569 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
570 fsr->FileSet, db_strerror(jcr->db));
574 jcr->jr.FileSetId = fsr->FileSetId;
576 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
577 fsr->FileSet, fsr->cCreateTime);
579 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
584 void init_jcr_job_record(JCR *jcr)
586 jcr->jr.SchedTime = jcr->sched_time;
587 jcr->jr.StartTime = jcr->start_time;
588 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
589 jcr->jr.JobType = jcr->JobType;
590 jcr->jr.JobLevel = jcr->JobLevel;
591 jcr->jr.JobStatus = jcr->JobStatus;
592 jcr->jr.JobId = jcr->JobId;
593 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
594 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
598 * Write status and such in DB
600 void update_job_end_record(JCR *jcr)
602 jcr->jr.EndTime = time(NULL);
603 jcr->end_time = jcr->jr.EndTime;
604 jcr->jr.JobId = jcr->JobId;
605 jcr->jr.JobStatus = jcr->JobStatus;
606 jcr->jr.JobFiles = jcr->JobFiles;
607 jcr->jr.JobBytes = jcr->JobBytes;
608 jcr->jr.VolSessionId = jcr->VolSessionId;
609 jcr->jr.VolSessionTime = jcr->VolSessionTime;
610 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
611 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
612 db_strerror(jcr->db));
617 * Takes base_name and appends (unique) current
618 * date and time to form unique job name.
620 * Returns: unique job name in jcr->Job
621 * date/time in jcr->start_time
623 void create_unique_job_name(JCR *jcr, const char *base_name)
625 /* Job start mutex */
626 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
627 static time_t last_start_time = 0;
630 char dt[MAX_TIME_LENGTH];
631 char name[MAX_NAME_LENGTH];
634 /* Guarantee unique start time -- maximum one per second, and
635 * thus unique Job Name
637 P(mutex); /* lock creation of jobs */
639 while (now == last_start_time) {
640 bmicrosleep(0, 500000);
643 last_start_time = now;
644 V(mutex); /* allow creation of jobs */
645 jcr->start_time = now;
646 /* Form Unique JobName */
647 localtime_r(&now, &tm);
648 /* Use only characters that are permitted in Windows filenames */
649 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
650 bstrncpy(name, base_name, sizeof(name));
651 name[sizeof(name)-22] = 0; /* truncate if too long */
652 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
653 /* Convert spaces into underscores */
654 for (p=jcr->Job; *p; p++) {
662 * Free the Job Control Record if no one is still using it.
663 * Called from main free_jcr() routine in src/lib/jcr.c so
664 * that we can do our Director specific cleanup of the jcr.
666 void dird_free_jcr(JCR *jcr)
668 Dmsg0(200, "Start dird free_jcr\n");
670 if (jcr->sd_auth_key) {
671 free(jcr->sd_auth_key);
672 jcr->sd_auth_key = NULL;
678 if (jcr->file_bsock) {
679 Dmsg0(200, "Close File bsock\n");
680 bnet_close(jcr->file_bsock);
681 jcr->file_bsock = NULL;
683 if (jcr->store_bsock) {
684 Dmsg0(200, "Close Store bsock\n");
685 bnet_close(jcr->store_bsock);
686 jcr->store_bsock = NULL;
689 Dmsg0(200, "Free JCR fname\n");
690 free_pool_memory(jcr->fname);
694 Dmsg0(200, "Free JCR stime\n");
695 free_pool_memory(jcr->stime);
698 if (jcr->RestoreBootstrap) {
699 free(jcr->RestoreBootstrap);
700 jcr->RestoreBootstrap = NULL;
702 if (jcr->client_uname) {
703 free_pool_memory(jcr->client_uname);
704 jcr->client_uname = NULL;
706 if (jcr->term_wait_inited) {
707 pthread_cond_destroy(&jcr->term_wait);
709 jcr->job_end_push.destroy();
710 Dmsg0(200, "End dird free_jcr\n");
714 * Set some defaults in the JCR necessary to
715 * run. These items are pulled from the job
716 * definition as defaults, but can be overridden
717 * later either by the Run record in the Schedule resource,
718 * or by the Console program.
720 void set_jcr_defaults(JCR *jcr, JOB *job)
723 jcr->JobType = job->JobType;
724 switch (jcr->JobType) {
727 jcr->JobLevel = L_NONE;
730 jcr->JobLevel = job->level;
733 jcr->JobPriority = job->Priority;
734 jcr->store = job->storage;
735 jcr->client = job->client;
736 if (!jcr->client_name) {
737 jcr->client_name = get_pool_memory(PM_NAME);
739 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
740 jcr->pool = job->pool;
741 jcr->full_pool = job->full_pool;
742 jcr->inc_pool = job->inc_pool;
743 jcr->dif_pool = job->dif_pool;
744 jcr->catalog = job->client->catalog;
745 jcr->fileset = job->fileset;
746 jcr->messages = job->messages;
747 jcr->spool_data = job->spool_data;
748 if (jcr->RestoreBootstrap) {
749 free(jcr->RestoreBootstrap);
750 jcr->RestoreBootstrap = NULL;
752 /* This can be overridden by Console program */
753 if (job->RestoreBootstrap) {
754 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
756 /* If no default level given, set one */
757 if (jcr->JobLevel == 0) {
758 switch (jcr->JobType) {
760 jcr->JobLevel = L_VERIFY_CATALOG;
763 jcr->JobLevel = L_INCREMENTAL;
767 jcr->JobLevel = L_NONE;