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).
78 void run_job(JCR *jcr)
83 sm_check(__FILE__, __LINE__, true);
84 init_msg(jcr, jcr->messages);
86 /* Initialize termination condition variable */
87 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
88 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
91 jcr->term_wait_inited = true;
96 Dmsg0(50, "Open database\n");
97 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
98 jcr->catalog->db_password, jcr->catalog->db_address,
99 jcr->catalog->db_port, jcr->catalog->db_socket);
100 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
101 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
102 jcr->catalog->db_name);
104 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
108 Dmsg0(50, "DB opened\n");
113 create_unique_job_name(jcr, jcr->job->hdr.name);
114 set_jcr_job_status(jcr, JS_Created);
115 init_jcr_job_record(jcr);
116 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
117 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
120 jcr->JobId = jcr->jr.JobId;
122 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
123 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
124 Dmsg0(200, "Add jrc to work queue\n");
126 /* Queue the job to be run */
127 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
128 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
131 Dmsg0(100, "Done run_job()\n");
137 set_jcr_job_status(jcr, JS_ErrorTerminated);
145 * This is the engine called by jobq.c:jobq_add() when we were pulled
146 * from the work queue.
147 * At this point, we are running in our own thread and all
148 * necessary resources are allocated -- see jobq.c
150 static void *job_thread(void *arg)
152 JCR *jcr = (JCR *)arg;
154 jcr->my_thread_id = pthread_self();
155 pthread_detach(jcr->my_thread_id);
156 sm_check(__FILE__, __LINE__, true);
160 Dmsg0(200, "=====Start Job=========\n");
161 jcr->start_time = time(NULL); /* set the real start time */
162 set_jcr_job_status(jcr, JS_Running);
164 if (job_canceled(jcr)) {
165 update_job_end_record(jcr);
166 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
167 (utime_t)(jcr->start_time - jcr->sched_time)) {
168 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
169 set_jcr_job_status(jcr, JS_Canceled);
170 update_job_end_record(jcr);
174 if (jcr->job->RunBeforeJob) {
175 POOLMEM *before = get_pool_memory(PM_FNAME);
178 char line[MAXSTRING];
180 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
181 bpipe = open_bpipe(before, 0, "r");
182 free_pool_memory(before);
183 while (fgets(line, sizeof(line), bpipe->rfd)) {
184 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
186 status = close_bpipe(bpipe);
188 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
190 set_jcr_job_status(jcr, JS_FatalError);
191 update_job_end_record(jcr);
195 switch (jcr->JobType) {
198 if (jcr->JobStatus == JS_Terminated) {
204 if (jcr->JobStatus == JS_Terminated) {
210 if (jcr->JobStatus == JS_Terminated) {
216 if (jcr->JobStatus == JS_Terminated) {
221 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
224 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
225 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
226 POOLMEM *after = get_pool_memory(PM_FNAME);
229 char line[MAXSTRING];
231 if (jcr->JobStatus == JS_Terminated) {
232 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
234 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
236 bpipe = open_bpipe(after, 0, "r");
237 free_pool_memory(after);
238 while (fgets(line, sizeof(line), bpipe->rfd)) {
239 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
241 status = close_bpipe(bpipe);
243 * Note, if we get an error here, do not mark the
244 * job in error, simply report the error condition.
247 if (jcr->JobStatus == JS_Terminated) {
248 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
251 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
256 /* Send off any queued messages */
257 if (jcr->msg_queue->size() > 0) {
258 dequeue_messages(jcr);
265 Dmsg0(50, "======== End Job ==========\n");
266 sm_check(__FILE__, __LINE__, true);
272 * Cancel a job -- typically called by the UA (Console program), but may also
273 * be called by the job watchdog.
275 * Returns: 1 if cancel appears to be successful
276 * 0 on failure. Message sent to ua->jcr.
278 int cancel_job(UAContext *ua, JCR *jcr)
282 switch (jcr->JobStatus) {
285 case JS_WaitClientRes:
286 case JS_WaitStoreRes:
287 case JS_WaitPriority:
289 case JS_WaitStartTime:
290 set_jcr_job_status(jcr, JS_Canceled);
291 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
292 jcr->JobId, jcr->Job);
293 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
297 set_jcr_job_status(jcr, JS_Canceled);
299 /* Cancel File daemon */
300 if (jcr->file_bsock) {
301 ua->jcr->client = jcr->client;
302 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
303 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
306 Dmsg0(200, "Connected to file daemon\n");
307 fd = ua->jcr->file_bsock;
308 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
309 while (bnet_recv(fd) >= 0) {
310 bsendmsg(ua, "%s", fd->msg);
312 bnet_sig(fd, BNET_TERMINATE);
314 ua->jcr->file_bsock = NULL;
317 /* Cancel Storage daemon */
318 if (jcr->store_bsock) {
319 ua->jcr->store = jcr->store;
320 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
321 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
324 Dmsg0(200, "Connected to storage daemon\n");
325 sd = ua->jcr->store_bsock;
326 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
327 while (bnet_recv(sd) >= 0) {
328 bsendmsg(ua, "%s", sd->msg);
330 bnet_sig(sd, BNET_TERMINATE);
332 ua->jcr->store_bsock = NULL;
340 static void job_monitor_destructor(watchdog_t *self)
342 JCR *control_jcr = (JCR *) self->data;
344 free_jcr(control_jcr);
347 static void job_monitor_watchdog(watchdog_t *self)
349 JCR *control_jcr, *jcr;
351 control_jcr = (JCR *)self->data;
353 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
360 if (jcr->JobId == 0) {
361 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
363 /* Keep reference counts correct */
364 free_locked_jcr(jcr);
368 /* check MaxWaitTime */
369 cancel = job_check_maxwaittime(control_jcr, jcr);
371 /* check MaxRunTime */
372 cancel |= job_check_maxruntime(control_jcr, jcr);
375 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
376 jcr, jcr->JobId, jcr->Job);
378 UAContext *ua = new_ua_context(jcr);
379 ua->jcr = control_jcr;
383 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
386 /* Keep reference counts correct */
387 free_locked_jcr(jcr);
393 * Check if the maxwaittime has expired and it is possible
396 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
400 if (jcr->job->MaxWaitTime == 0) {
403 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
404 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
405 jcr, jcr->Job, jcr->job->MaxWaitTime);
408 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
410 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
411 switch (jcr->JobStatus) {
416 case JS_WaitStoreRes:
417 case JS_WaitClientRes:
419 case JS_WaitPriority:
421 case JS_WaitStartTime:
423 Dmsg0(200, "JCR blocked in #1\n");
426 Dmsg0(200, "JCR running, checking SD status\n");
427 switch (jcr->SDJobStatus) {
432 Dmsg0(200, "JCR blocked in #2\n");
435 Dmsg0(200, "JCR not blocked in #2\n");
440 case JS_ErrorTerminated:
443 Dmsg0(200, "JCR already dead in #3\n");
446 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
449 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
450 cancel ? "" : "do not ", jcr, jcr->job);
456 * Check if maxruntime has expired and if the job can be
459 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
463 if (jcr->job->MaxRunTime == 0) {
466 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
467 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
468 jcr, jcr->Job, jcr->job->MaxRunTime);
472 switch (jcr->JobStatus) {
478 case JS_WaitStoreRes:
479 case JS_WaitClientRes:
481 case JS_WaitPriority:
483 case JS_WaitStartTime:
488 case JS_ErrorTerminated:
494 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
498 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
499 cancel ? "" : "do not ", jcr, jcr->job);
506 * Get or create a Client record for this Job
508 bool get_or_create_client_record(JCR *jcr)
512 memset(&cr, 0, sizeof(cr));
513 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
514 cr.AutoPrune = jcr->client->AutoPrune;
515 cr.FileRetention = jcr->client->FileRetention;
516 cr.JobRetention = jcr->client->JobRetention;
517 if (!jcr->client_name) {
518 jcr->client_name = get_pool_memory(PM_NAME);
520 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
521 if (!db_create_client_record(jcr, jcr->db, &cr)) {
522 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
523 db_strerror(jcr->db));
526 jcr->jr.ClientId = cr.ClientId;
528 if (!jcr->client_uname) {
529 jcr->client_uname = get_pool_memory(PM_NAME);
531 pm_strcpy(&jcr->client_uname, cr.Uname);
533 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
538 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
541 * Get or Create FileSet record
543 memset(fsr, 0, sizeof(FILESET_DBR));
544 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
545 if (jcr->fileset->have_MD5) {
546 struct MD5Context md5c;
547 unsigned char signature[16];
548 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
549 MD5Final(signature, &md5c);
550 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
551 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
553 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
555 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
556 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
557 fsr->FileSet, db_strerror(jcr->db));
560 jcr->jr.FileSetId = fsr->FileSetId;
562 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
563 fsr->FileSet, fsr->cCreateTime);
565 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
570 void init_jcr_job_record(JCR *jcr)
572 jcr->jr.SchedTime = jcr->sched_time;
573 jcr->jr.StartTime = jcr->start_time;
574 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
575 jcr->jr.JobType = jcr->JobType;
576 jcr->jr.JobLevel = jcr->JobLevel;
577 jcr->jr.JobStatus = jcr->JobStatus;
578 jcr->jr.JobId = jcr->JobId;
579 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
580 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
584 * Write status and such in DB
586 void update_job_end_record(JCR *jcr)
588 if (jcr->jr.EndTime == 0) {
589 jcr->jr.EndTime = time(NULL);
591 jcr->end_time = jcr->jr.EndTime;
592 jcr->jr.JobId = jcr->JobId;
593 jcr->jr.JobStatus = jcr->JobStatus;
594 jcr->jr.JobFiles = jcr->JobFiles;
595 jcr->jr.JobBytes = jcr->JobBytes;
596 jcr->jr.VolSessionId = jcr->VolSessionId;
597 jcr->jr.VolSessionTime = jcr->VolSessionTime;
598 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
599 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
600 db_strerror(jcr->db));
605 * Takes base_name and appends (unique) current
606 * date and time to form unique job name.
608 * Returns: unique job name in jcr->Job
609 * date/time in jcr->start_time
611 void create_unique_job_name(JCR *jcr, const char *base_name)
613 /* Job start mutex */
614 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
615 static time_t last_start_time = 0;
618 char dt[MAX_TIME_LENGTH];
619 char name[MAX_NAME_LENGTH];
622 /* Guarantee unique start time -- maximum one per second, and
623 * thus unique Job Name
625 P(mutex); /* lock creation of jobs */
627 while (now == last_start_time) {
628 bmicrosleep(0, 500000);
631 last_start_time = now;
632 V(mutex); /* allow creation of jobs */
633 jcr->start_time = now;
634 /* Form Unique JobName */
635 localtime_r(&now, &tm);
636 /* Use only characters that are permitted in Windows filenames */
637 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
638 bstrncpy(name, base_name, sizeof(name));
639 name[sizeof(name)-22] = 0; /* truncate if too long */
640 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
641 /* Convert spaces into underscores */
642 for (p=jcr->Job; *p; p++) {
650 * Free the Job Control Record if no one is still using it.
651 * Called from main free_jcr() routine in src/lib/jcr.c so
652 * that we can do our Director specific cleanup of the jcr.
654 void dird_free_jcr(JCR *jcr)
656 Dmsg0(200, "Start dird free_jcr\n");
658 if (jcr->sd_auth_key) {
659 free(jcr->sd_auth_key);
660 jcr->sd_auth_key = NULL;
666 if (jcr->file_bsock) {
667 Dmsg0(200, "Close File bsock\n");
668 bnet_close(jcr->file_bsock);
669 jcr->file_bsock = NULL;
671 if (jcr->store_bsock) {
672 Dmsg0(200, "Close Store bsock\n");
673 bnet_close(jcr->store_bsock);
674 jcr->store_bsock = NULL;
677 Dmsg0(200, "Free JCR fname\n");
678 free_pool_memory(jcr->fname);
682 Dmsg0(200, "Free JCR stime\n");
683 free_pool_memory(jcr->stime);
686 if (jcr->RestoreBootstrap) {
687 free(jcr->RestoreBootstrap);
688 jcr->RestoreBootstrap = NULL;
690 if (jcr->client_uname) {
691 free_pool_memory(jcr->client_uname);
692 jcr->client_uname = NULL;
694 if (jcr->term_wait_inited) {
695 pthread_cond_destroy(&jcr->term_wait);
697 jcr->job_end_push.destroy();
698 Dmsg0(200, "End dird free_jcr\n");
702 * Set some defaults in the JCR necessary to
703 * run. These items are pulled from the job
704 * definition as defaults, but can be overridden
705 * later either by the Run record in the Schedule resource,
706 * or by the Console program.
708 void set_jcr_defaults(JCR *jcr, JOB *job)
711 jcr->JobType = job->JobType;
712 switch (jcr->JobType) {
715 jcr->JobLevel = L_NONE;
718 jcr->JobLevel = job->level;
721 jcr->JobPriority = job->Priority;
722 jcr->store = job->storage;
723 jcr->client = job->client;
724 if (!jcr->client_name) {
725 jcr->client_name = get_pool_memory(PM_NAME);
727 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
728 jcr->pool = job->pool;
729 jcr->full_pool = job->full_pool;
730 jcr->inc_pool = job->inc_pool;
731 jcr->dif_pool = job->dif_pool;
732 jcr->catalog = job->client->catalog;
733 jcr->fileset = job->fileset;
734 jcr->messages = job->messages;
735 jcr->spool_data = job->spool_data;
736 if (jcr->RestoreBootstrap) {
737 free(jcr->RestoreBootstrap);
738 jcr->RestoreBootstrap = NULL;
740 /* This can be overridden by Console program */
741 if (job->RestoreBootstrap) {
742 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
744 /* If no default level given, set one */
745 if (jcr->JobLevel == 0) {
746 switch (jcr->JobType) {
748 jcr->JobLevel = L_VERIFY_CATALOG;
751 jcr->JobLevel = L_INCREMENTAL;
755 jcr->JobLevel = L_NONE;