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) {
62 Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror());
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()
78 if ((stat=jobq_destroy(&job_queue)) != 0) {
81 Emsg1(M_INFO, 0, _("Could not term job queue: ERR=%s\n"), be.strerror());
86 * Run a job -- typically called by the scheduler, but may also
87 * be called by the UA (Console program).
89 * Returns: 0 on failure
93 JobId_t run_job(JCR *jcr)
99 sm_check(__FILE__, __LINE__, true);
100 init_msg(jcr, jcr->messages);
102 /* Initialize termination condition variable */
103 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
105 be.set_errno(errstat);
106 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.strerror());
109 jcr->term_wait_inited = true;
114 Dmsg0(50, "Open database\n");
115 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
116 jcr->catalog->db_password, jcr->catalog->db_address,
117 jcr->catalog->db_port, jcr->catalog->db_socket);
118 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
119 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
120 jcr->catalog->db_name);
122 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
126 Dmsg0(50, "DB opened\n");
131 create_unique_job_name(jcr, jcr->job->hdr.name);
132 set_jcr_job_status(jcr, JS_Created);
133 init_jcr_job_record(jcr);
134 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
135 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
138 JobId = jcr->JobId = jcr->jr.JobId;
140 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
141 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
142 Dmsg0(200, "Add jrc to work queue\n");
144 /* Queue the job to be run */
145 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
148 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror());
152 Dmsg0(100, "Done run_job()\n");
158 set_jcr_job_status(jcr, JS_ErrorTerminated);
166 * This is the engine called by jobq.c:jobq_add() when we were pulled
167 * from the work queue.
168 * At this point, we are running in our own thread and all
169 * necessary resources are allocated -- see jobq.c
171 static void *job_thread(void *arg)
173 JCR *jcr = (JCR *)arg;
175 jcr->my_thread_id = pthread_self();
176 pthread_detach(jcr->my_thread_id);
177 sm_check(__FILE__, __LINE__, true);
180 Dmsg0(200, "=====Start Job=========\n");
181 jcr->start_time = time(NULL); /* set the real start time */
182 set_jcr_job_status(jcr, JS_Running);
184 if (job_canceled(jcr)) {
185 update_job_end_record(jcr);
186 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
187 (utime_t)(jcr->start_time - jcr->sched_time)) {
188 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
189 set_jcr_job_status(jcr, JS_Canceled);
190 update_job_end_record(jcr);
194 if (jcr->job->RunBeforeJob) {
195 POOLMEM *before = get_pool_memory(PM_FNAME);
198 char line[MAXSTRING];
200 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
201 bpipe = open_bpipe(before, 0, "r");
202 free_pool_memory(before);
203 while (fgets(line, sizeof(line), bpipe->rfd)) {
204 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
206 status = close_bpipe(bpipe);
209 be.set_errno(status);
210 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror());
211 set_jcr_job_status(jcr, JS_FatalError);
212 update_job_end_record(jcr);
216 switch (jcr->JobType) {
219 if (jcr->JobStatus == JS_Terminated) {
225 if (jcr->JobStatus == JS_Terminated) {
231 if (jcr->JobStatus == JS_Terminated) {
237 if (jcr->JobStatus == JS_Terminated) {
242 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
245 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
246 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
247 POOLMEM *after = get_pool_memory(PM_FNAME);
250 char line[MAXSTRING];
252 if (jcr->JobStatus == JS_Terminated) {
253 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
255 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
257 bpipe = open_bpipe(after, 0, "r");
258 free_pool_memory(after);
259 while (fgets(line, sizeof(line), bpipe->rfd)) {
260 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
262 status = close_bpipe(bpipe);
264 * Note, if we get an error here, do not mark the
265 * job in error, simply report the error condition.
269 be.set_errno(status);
270 if (jcr->JobStatus == JS_Terminated) {
271 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror());
273 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror());
277 /* Send off any queued messages */
278 if (jcr->msg_queue->size() > 0) {
279 dequeue_messages(jcr);
286 Dmsg0(50, "======== End Job ==========\n");
287 sm_check(__FILE__, __LINE__, true);
293 * Cancel a job -- typically called by the UA (Console program), but may also
294 * be called by the job watchdog.
296 * Returns: 1 if cancel appears to be successful
297 * 0 on failure. Message sent to ua->jcr.
299 int cancel_job(UAContext *ua, JCR *jcr)
303 switch (jcr->JobStatus) {
306 case JS_WaitClientRes:
307 case JS_WaitStoreRes:
308 case JS_WaitPriority:
310 case JS_WaitStartTime:
311 set_jcr_job_status(jcr, JS_Canceled);
312 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
313 jcr->JobId, jcr->Job);
314 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
318 set_jcr_job_status(jcr, JS_Canceled);
320 /* Cancel File daemon */
321 if (jcr->file_bsock) {
322 ua->jcr->client = jcr->client;
323 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
324 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
327 Dmsg0(200, "Connected to file daemon\n");
328 fd = ua->jcr->file_bsock;
329 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
330 while (bnet_recv(fd) >= 0) {
331 bsendmsg(ua, "%s", fd->msg);
333 bnet_sig(fd, BNET_TERMINATE);
335 ua->jcr->file_bsock = NULL;
338 /* Cancel Storage daemon */
339 if (jcr->store_bsock) {
340 ua->jcr->store = jcr->store;
341 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
342 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
345 Dmsg0(200, "Connected to storage daemon\n");
346 sd = ua->jcr->store_bsock;
347 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
348 while (bnet_recv(sd) >= 0) {
349 bsendmsg(ua, "%s", sd->msg);
351 bnet_sig(sd, BNET_TERMINATE);
353 ua->jcr->store_bsock = NULL;
361 static void job_monitor_destructor(watchdog_t *self)
363 JCR *control_jcr = (JCR *) self->data;
365 free_jcr(control_jcr);
368 static void job_monitor_watchdog(watchdog_t *self)
370 JCR *control_jcr, *jcr;
372 control_jcr = (JCR *)self->data;
374 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
381 if (jcr->JobId == 0) {
382 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
384 /* Keep reference counts correct */
385 free_locked_jcr(jcr);
389 /* check MaxWaitTime */
390 cancel = job_check_maxwaittime(control_jcr, jcr);
392 /* check MaxRunTime */
393 cancel |= job_check_maxruntime(control_jcr, jcr);
396 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
397 jcr, jcr->JobId, jcr->Job);
399 UAContext *ua = new_ua_context(jcr);
400 ua->jcr = control_jcr;
404 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
407 /* Keep reference counts correct */
408 free_locked_jcr(jcr);
414 * Check if the maxwaittime has expired and it is possible
417 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
421 if (jcr->job->MaxWaitTime == 0) {
424 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
425 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
426 jcr, jcr->Job, jcr->job->MaxWaitTime);
429 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
431 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
432 switch (jcr->JobStatus) {
437 case JS_WaitStoreRes:
438 case JS_WaitClientRes:
440 case JS_WaitPriority:
442 case JS_WaitStartTime:
444 Dmsg0(200, "JCR blocked in #1\n");
447 Dmsg0(200, "JCR running, checking SD status\n");
448 switch (jcr->SDJobStatus) {
453 Dmsg0(200, "JCR blocked in #2\n");
456 Dmsg0(200, "JCR not blocked in #2\n");
461 case JS_ErrorTerminated:
464 Dmsg0(200, "JCR already dead in #3\n");
467 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
470 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
471 cancel ? "" : "do not ", jcr, jcr->job);
477 * Check if maxruntime has expired and if the job can be
480 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
484 if (jcr->job->MaxRunTime == 0) {
487 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
488 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
489 jcr, jcr->Job, jcr->job->MaxRunTime);
493 switch (jcr->JobStatus) {
499 case JS_WaitStoreRes:
500 case JS_WaitClientRes:
502 case JS_WaitPriority:
504 case JS_WaitStartTime:
509 case JS_ErrorTerminated:
515 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
519 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
520 cancel ? "" : "do not ", jcr, jcr->job);
527 * Get or create a Client record for this Job
529 bool get_or_create_client_record(JCR *jcr)
533 memset(&cr, 0, sizeof(cr));
534 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
535 cr.AutoPrune = jcr->client->AutoPrune;
536 cr.FileRetention = jcr->client->FileRetention;
537 cr.JobRetention = jcr->client->JobRetention;
538 if (!jcr->client_name) {
539 jcr->client_name = get_pool_memory(PM_NAME);
541 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
542 if (!db_create_client_record(jcr, jcr->db, &cr)) {
543 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
544 db_strerror(jcr->db));
547 jcr->jr.ClientId = cr.ClientId;
549 if (!jcr->client_uname) {
550 jcr->client_uname = get_pool_memory(PM_NAME);
552 pm_strcpy(jcr->client_uname, cr.Uname);
554 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
559 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
562 * Get or Create FileSet record
564 memset(fsr, 0, sizeof(FILESET_DBR));
565 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
566 if (jcr->fileset->have_MD5) {
567 struct MD5Context md5c;
568 unsigned char signature[16];
569 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
570 MD5Final(signature, &md5c);
571 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
572 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
574 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
576 if (!jcr->fileset->ignore_fs_changes ||
577 !db_get_fileset_record(jcr, jcr->db, fsr)) {
578 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
579 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
580 fsr->FileSet, db_strerror(jcr->db));
584 jcr->jr.FileSetId = fsr->FileSetId;
586 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
587 fsr->FileSet, fsr->cCreateTime);
589 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
594 void init_jcr_job_record(JCR *jcr)
596 jcr->jr.SchedTime = jcr->sched_time;
597 jcr->jr.StartTime = jcr->start_time;
598 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
599 jcr->jr.JobType = jcr->JobType;
600 jcr->jr.JobLevel = jcr->JobLevel;
601 jcr->jr.JobStatus = jcr->JobStatus;
602 jcr->jr.JobId = jcr->JobId;
603 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
604 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
608 * Write status and such in DB
610 void update_job_end_record(JCR *jcr)
612 jcr->jr.EndTime = time(NULL);
613 jcr->end_time = jcr->jr.EndTime;
614 jcr->jr.JobId = jcr->JobId;
615 jcr->jr.JobStatus = jcr->JobStatus;
616 jcr->jr.JobFiles = jcr->JobFiles;
617 jcr->jr.JobBytes = jcr->JobBytes;
618 jcr->jr.VolSessionId = jcr->VolSessionId;
619 jcr->jr.VolSessionTime = jcr->VolSessionTime;
620 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
621 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
622 db_strerror(jcr->db));
627 * Takes base_name and appends (unique) current
628 * date and time to form unique job name.
630 * Returns: unique job name in jcr->Job
631 * date/time in jcr->start_time
633 void create_unique_job_name(JCR *jcr, const char *base_name)
635 /* Job start mutex */
636 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
637 static time_t last_start_time = 0;
640 char dt[MAX_TIME_LENGTH];
641 char name[MAX_NAME_LENGTH];
644 /* Guarantee unique start time -- maximum one per second, and
645 * thus unique Job Name
647 P(mutex); /* lock creation of jobs */
649 while (now == last_start_time) {
650 bmicrosleep(0, 500000);
653 last_start_time = now;
654 V(mutex); /* allow creation of jobs */
655 jcr->start_time = now;
656 /* Form Unique JobName */
657 localtime_r(&now, &tm);
658 /* Use only characters that are permitted in Windows filenames */
659 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
660 bstrncpy(name, base_name, sizeof(name));
661 name[sizeof(name)-22] = 0; /* truncate if too long */
662 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
663 /* Convert spaces into underscores */
664 for (p=jcr->Job; *p; p++) {
672 * Free the Job Control Record if no one is still using it.
673 * Called from main free_jcr() routine in src/lib/jcr.c so
674 * that we can do our Director specific cleanup of the jcr.
676 void dird_free_jcr(JCR *jcr)
678 Dmsg0(200, "Start dird free_jcr\n");
680 if (jcr->sd_auth_key) {
681 free(jcr->sd_auth_key);
682 jcr->sd_auth_key = NULL;
688 if (jcr->file_bsock) {
689 Dmsg0(200, "Close File bsock\n");
690 bnet_close(jcr->file_bsock);
691 jcr->file_bsock = NULL;
693 if (jcr->store_bsock) {
694 Dmsg0(200, "Close Store bsock\n");
695 bnet_close(jcr->store_bsock);
696 jcr->store_bsock = NULL;
699 Dmsg0(200, "Free JCR fname\n");
700 free_pool_memory(jcr->fname);
704 Dmsg0(200, "Free JCR stime\n");
705 free_pool_memory(jcr->stime);
708 if (jcr->RestoreBootstrap) {
709 free(jcr->RestoreBootstrap);
710 jcr->RestoreBootstrap = NULL;
712 if (jcr->client_uname) {
713 free_pool_memory(jcr->client_uname);
714 jcr->client_uname = NULL;
716 if (jcr->term_wait_inited) {
717 pthread_cond_destroy(&jcr->term_wait);
719 jcr->job_end_push.destroy();
720 Dmsg0(200, "End dird free_jcr\n");
724 * Set some defaults in the JCR necessary to
725 * run. These items are pulled from the job
726 * definition as defaults, but can be overridden
727 * later either by the Run record in the Schedule resource,
728 * or by the Console program.
730 void set_jcr_defaults(JCR *jcr, JOB *job)
733 jcr->JobType = job->JobType;
734 switch (jcr->JobType) {
737 jcr->JobLevel = L_NONE;
740 jcr->JobLevel = job->level;
743 jcr->JobPriority = job->Priority;
744 jcr->store = job->storage;
745 jcr->client = job->client;
746 if (!jcr->client_name) {
747 jcr->client_name = get_pool_memory(PM_NAME);
749 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
750 jcr->pool = job->pool;
751 jcr->full_pool = job->full_pool;
752 jcr->inc_pool = job->inc_pool;
753 jcr->dif_pool = job->dif_pool;
754 jcr->catalog = job->client->catalog;
755 jcr->fileset = job->fileset;
756 jcr->messages = job->messages;
757 jcr->spool_data = job->spool_data;
758 if (jcr->RestoreBootstrap) {
759 free(jcr->RestoreBootstrap);
760 jcr->RestoreBootstrap = NULL;
762 /* This can be overridden by Console program */
763 if (job->RestoreBootstrap) {
764 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
766 /* If no default level given, set one */
767 if (jcr->JobLevel == 0) {
768 switch (jcr->JobType) {
770 jcr->JobLevel = L_VERIFY_CATALOG;
773 jcr->JobLevel = L_INCREMENTAL;
777 jcr->JobLevel = L_NONE;