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);
159 Dmsg0(200, "=====Start Job=========\n");
160 jcr->start_time = time(NULL); /* set the real start time */
161 set_jcr_job_status(jcr, JS_Running);
163 if (job_canceled(jcr)) {
164 update_job_end_record(jcr);
165 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
166 (utime_t)(jcr->start_time - jcr->sched_time)) {
167 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
168 set_jcr_job_status(jcr, JS_Canceled);
169 update_job_end_record(jcr);
173 if (jcr->job->RunBeforeJob) {
174 POOLMEM *before = get_pool_memory(PM_FNAME);
177 char line[MAXSTRING];
179 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
180 bpipe = open_bpipe(before, 0, "r");
181 free_pool_memory(before);
182 while (fgets(line, sizeof(line), bpipe->rfd)) {
183 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
185 status = close_bpipe(bpipe);
187 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
189 set_jcr_job_status(jcr, JS_FatalError);
190 update_job_end_record(jcr);
194 switch (jcr->JobType) {
197 if (jcr->JobStatus == JS_Terminated) {
203 if (jcr->JobStatus == JS_Terminated) {
209 if (jcr->JobStatus == JS_Terminated) {
215 if (jcr->JobStatus == JS_Terminated) {
220 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
223 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
224 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
225 POOLMEM *after = get_pool_memory(PM_FNAME);
228 char line[MAXSTRING];
230 if (jcr->JobStatus == JS_Terminated) {
231 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
233 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
235 bpipe = open_bpipe(after, 0, "r");
236 free_pool_memory(after);
237 while (fgets(line, sizeof(line), bpipe->rfd)) {
238 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
240 status = close_bpipe(bpipe);
242 * Note, if we get an error here, do not mark the
243 * job in error, simply report the error condition.
246 if (jcr->JobStatus == JS_Terminated) {
247 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
250 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
255 /* Send off any queued messages */
256 if (jcr->msg_queue->size() > 0) {
257 dequeue_messages(jcr);
264 Dmsg0(50, "======== End Job ==========\n");
265 sm_check(__FILE__, __LINE__, true);
271 * Cancel a job -- typically called by the UA (Console program), but may also
272 * be called by the job watchdog.
274 * Returns: 1 if cancel appears to be successful
275 * 0 on failure. Message sent to ua->jcr.
277 int cancel_job(UAContext *ua, JCR *jcr)
281 switch (jcr->JobStatus) {
284 case JS_WaitClientRes:
285 case JS_WaitStoreRes:
286 case JS_WaitPriority:
288 case JS_WaitStartTime:
289 set_jcr_job_status(jcr, JS_Canceled);
290 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
291 jcr->JobId, jcr->Job);
292 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
296 set_jcr_job_status(jcr, JS_Canceled);
298 /* Cancel File daemon */
299 if (jcr->file_bsock) {
300 ua->jcr->client = jcr->client;
301 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
302 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
305 Dmsg0(200, "Connected to file daemon\n");
306 fd = ua->jcr->file_bsock;
307 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
308 while (bnet_recv(fd) >= 0) {
309 bsendmsg(ua, "%s", fd->msg);
311 bnet_sig(fd, BNET_TERMINATE);
313 ua->jcr->file_bsock = NULL;
316 /* Cancel Storage daemon */
317 if (jcr->store_bsock) {
318 ua->jcr->store = jcr->store;
319 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
320 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
323 Dmsg0(200, "Connected to storage daemon\n");
324 sd = ua->jcr->store_bsock;
325 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
326 while (bnet_recv(sd) >= 0) {
327 bsendmsg(ua, "%s", sd->msg);
329 bnet_sig(sd, BNET_TERMINATE);
331 ua->jcr->store_bsock = NULL;
339 static void job_monitor_destructor(watchdog_t *self)
341 JCR *control_jcr = (JCR *) self->data;
343 free_jcr(control_jcr);
346 static void job_monitor_watchdog(watchdog_t *self)
348 JCR *control_jcr, *jcr;
350 control_jcr = (JCR *)self->data;
352 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
359 if (jcr->JobId == 0) {
360 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
362 /* Keep reference counts correct */
363 free_locked_jcr(jcr);
367 /* check MaxWaitTime */
368 cancel = job_check_maxwaittime(control_jcr, jcr);
370 /* check MaxRunTime */
371 cancel |= job_check_maxruntime(control_jcr, jcr);
374 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
375 jcr, jcr->JobId, jcr->Job);
377 UAContext *ua = new_ua_context(jcr);
378 ua->jcr = control_jcr;
382 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
385 /* Keep reference counts correct */
386 free_locked_jcr(jcr);
392 * Check if the maxwaittime has expired and it is possible
395 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
399 if (jcr->job->MaxWaitTime == 0) {
402 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
403 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
404 jcr, jcr->Job, jcr->job->MaxWaitTime);
407 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
409 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
410 switch (jcr->JobStatus) {
415 case JS_WaitStoreRes:
416 case JS_WaitClientRes:
418 case JS_WaitPriority:
420 case JS_WaitStartTime:
422 Dmsg0(200, "JCR blocked in #1\n");
425 Dmsg0(200, "JCR running, checking SD status\n");
426 switch (jcr->SDJobStatus) {
431 Dmsg0(200, "JCR blocked in #2\n");
434 Dmsg0(200, "JCR not blocked in #2\n");
439 case JS_ErrorTerminated:
442 Dmsg0(200, "JCR already dead in #3\n");
445 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
448 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
449 cancel ? "" : "do not ", jcr, jcr->job);
455 * Check if maxruntime has expired and if the job can be
458 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
462 if (jcr->job->MaxRunTime == 0) {
465 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
466 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
467 jcr, jcr->Job, jcr->job->MaxRunTime);
471 switch (jcr->JobStatus) {
477 case JS_WaitStoreRes:
478 case JS_WaitClientRes:
480 case JS_WaitPriority:
482 case JS_WaitStartTime:
487 case JS_ErrorTerminated:
493 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
497 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
498 cancel ? "" : "do not ", jcr, jcr->job);
505 * Get or create a Client record for this Job
507 bool get_or_create_client_record(JCR *jcr)
511 memset(&cr, 0, sizeof(cr));
512 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
513 cr.AutoPrune = jcr->client->AutoPrune;
514 cr.FileRetention = jcr->client->FileRetention;
515 cr.JobRetention = jcr->client->JobRetention;
516 if (!jcr->client_name) {
517 jcr->client_name = get_pool_memory(PM_NAME);
519 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
520 if (!db_create_client_record(jcr, jcr->db, &cr)) {
521 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
522 db_strerror(jcr->db));
525 jcr->jr.ClientId = cr.ClientId;
527 if (!jcr->client_uname) {
528 jcr->client_uname = get_pool_memory(PM_NAME);
530 pm_strcpy(&jcr->client_uname, cr.Uname);
532 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
537 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
540 * Get or Create FileSet record
542 memset(fsr, 0, sizeof(FILESET_DBR));
543 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
544 if (jcr->fileset->have_MD5) {
545 struct MD5Context md5c;
546 unsigned char signature[16];
547 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
548 MD5Final(signature, &md5c);
549 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
550 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
552 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
554 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
555 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
556 fsr->FileSet, db_strerror(jcr->db));
559 jcr->jr.FileSetId = fsr->FileSetId;
561 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
562 fsr->FileSet, fsr->cCreateTime);
564 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
569 void init_jcr_job_record(JCR *jcr)
571 jcr->jr.SchedTime = jcr->sched_time;
572 jcr->jr.StartTime = jcr->start_time;
573 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
574 jcr->jr.JobType = jcr->JobType;
575 jcr->jr.JobLevel = jcr->JobLevel;
576 jcr->jr.JobStatus = jcr->JobStatus;
577 jcr->jr.JobId = jcr->JobId;
578 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
579 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
583 * Write status and such in DB
585 void update_job_end_record(JCR *jcr)
587 jcr->jr.EndTime = time(NULL);
588 jcr->end_time = jcr->jr.EndTime;
589 jcr->jr.JobId = jcr->JobId;
590 jcr->jr.JobStatus = jcr->JobStatus;
591 jcr->jr.JobFiles = jcr->JobFiles;
592 jcr->jr.JobBytes = jcr->JobBytes;
593 jcr->jr.VolSessionId = jcr->VolSessionId;
594 jcr->jr.VolSessionTime = jcr->VolSessionTime;
595 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
596 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
597 db_strerror(jcr->db));
602 * Takes base_name and appends (unique) current
603 * date and time to form unique job name.
605 * Returns: unique job name in jcr->Job
606 * date/time in jcr->start_time
608 void create_unique_job_name(JCR *jcr, const char *base_name)
610 /* Job start mutex */
611 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
612 static time_t last_start_time = 0;
615 char dt[MAX_TIME_LENGTH];
616 char name[MAX_NAME_LENGTH];
619 /* Guarantee unique start time -- maximum one per second, and
620 * thus unique Job Name
622 P(mutex); /* lock creation of jobs */
624 while (now == last_start_time) {
625 bmicrosleep(0, 500000);
628 last_start_time = now;
629 V(mutex); /* allow creation of jobs */
630 jcr->start_time = now;
631 /* Form Unique JobName */
632 localtime_r(&now, &tm);
633 /* Use only characters that are permitted in Windows filenames */
634 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
635 bstrncpy(name, base_name, sizeof(name));
636 name[sizeof(name)-22] = 0; /* truncate if too long */
637 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
638 /* Convert spaces into underscores */
639 for (p=jcr->Job; *p; p++) {
647 * Free the Job Control Record if no one is still using it.
648 * Called from main free_jcr() routine in src/lib/jcr.c so
649 * that we can do our Director specific cleanup of the jcr.
651 void dird_free_jcr(JCR *jcr)
653 Dmsg0(200, "Start dird free_jcr\n");
655 if (jcr->sd_auth_key) {
656 free(jcr->sd_auth_key);
657 jcr->sd_auth_key = NULL;
663 if (jcr->file_bsock) {
664 Dmsg0(200, "Close File bsock\n");
665 bnet_close(jcr->file_bsock);
666 jcr->file_bsock = NULL;
668 if (jcr->store_bsock) {
669 Dmsg0(200, "Close Store bsock\n");
670 bnet_close(jcr->store_bsock);
671 jcr->store_bsock = NULL;
674 Dmsg0(200, "Free JCR fname\n");
675 free_pool_memory(jcr->fname);
679 Dmsg0(200, "Free JCR stime\n");
680 free_pool_memory(jcr->stime);
683 if (jcr->RestoreBootstrap) {
684 free(jcr->RestoreBootstrap);
685 jcr->RestoreBootstrap = NULL;
687 if (jcr->client_uname) {
688 free_pool_memory(jcr->client_uname);
689 jcr->client_uname = NULL;
691 if (jcr->term_wait_inited) {
692 pthread_cond_destroy(&jcr->term_wait);
694 jcr->job_end_push.destroy();
695 Dmsg0(200, "End dird free_jcr\n");
699 * Set some defaults in the JCR necessary to
700 * run. These items are pulled from the job
701 * definition as defaults, but can be overridden
702 * later either by the Run record in the Schedule resource,
703 * or by the Console program.
705 void set_jcr_defaults(JCR *jcr, JOB *job)
708 jcr->JobType = job->JobType;
709 switch (jcr->JobType) {
712 jcr->JobLevel = L_NONE;
715 jcr->JobLevel = job->level;
718 jcr->JobPriority = job->Priority;
719 jcr->store = job->storage;
720 jcr->client = job->client;
721 if (!jcr->client_name) {
722 jcr->client_name = get_pool_memory(PM_NAME);
724 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
725 jcr->pool = job->pool;
726 jcr->full_pool = job->full_pool;
727 jcr->inc_pool = job->inc_pool;
728 jcr->dif_pool = job->dif_pool;
729 jcr->catalog = job->client->catalog;
730 jcr->fileset = job->fileset;
731 jcr->messages = job->messages;
732 jcr->spool_data = job->spool_data;
733 if (jcr->RestoreBootstrap) {
734 free(jcr->RestoreBootstrap);
735 jcr->RestoreBootstrap = NULL;
737 /* This can be overridden by Console program */
738 if (job->RestoreBootstrap) {
739 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
741 /* If no default level given, set one */
742 if (jcr->JobLevel == 0) {
743 switch (jcr->JobType) {
745 jcr->JobLevel = L_VERIFY_CATALOG;
748 jcr->JobLevel = L_INCREMENTAL;
752 jcr->JobLevel = L_NONE;