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 = create_control_jcr("*JobMonitor*", JT_SYSTEM);
70 register_watchdog(wd);
77 * Run a job -- typically called by the scheduler, but may also
78 * be called by the UA (Console program).
81 void run_job(JCR *jcr)
86 sm_check(__FILE__, __LINE__, True);
87 init_msg(jcr, jcr->messages);
88 create_unique_job_name(jcr, jcr->job->hdr.name);
89 set_jcr_job_status(jcr, JS_Created);
90 jcr->jr.SchedTime = jcr->sched_time;
91 jcr->jr.StartTime = jcr->start_time;
92 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
93 jcr->jr.Type = jcr->JobType;
94 jcr->jr.Level = jcr->JobLevel;
95 jcr->jr.JobStatus = jcr->JobStatus;
96 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
97 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
99 /* Initialize termination condition variable */
100 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
101 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
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 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
113 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
114 jcr->catalog->db_name);
116 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
120 Dmsg0(50, "DB opened\n");
125 jcr->jr.JobStatus = jcr->JobStatus;
126 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
127 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
130 jcr->JobId = jcr->jr.JobId;
132 Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
133 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
134 Dmsg0(200, "Add jrc to work queue\n");
136 /* Queue the job to be run */
137 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
138 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
141 Dmsg0(100, "Done run_job()\n");
147 set_jcr_job_status(jcr, JS_ErrorTerminated);
155 * This is the engine called by jobq.c:jobq_add() when we were pulled
156 * from the work queue.
157 * At this point, we are running in our own thread and all
158 * necessary resources are allocated -- see jobq.c
160 static void *job_thread(void *arg)
162 JCR *jcr = (JCR *)arg;
164 pthread_detach(pthread_self());
165 sm_check(__FILE__, __LINE__, True);
169 Dmsg0(200, "=====Start Job=========\n");
170 jcr->start_time = time(NULL); /* set the real start time */
171 set_jcr_job_status(jcr, JS_Running);
173 if (job_canceled(jcr)) {
174 update_job_end_record(jcr);
175 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
176 (utime_t)(jcr->start_time - jcr->sched_time)) {
177 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
178 set_jcr_job_status(jcr, JS_Canceled);
179 update_job_end_record(jcr);
183 if (jcr->job->RunBeforeJob) {
184 POOLMEM *before = get_pool_memory(PM_FNAME);
187 char line[MAXSTRING];
189 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
190 bpipe = open_bpipe(before, 0, "r");
191 free_pool_memory(before);
192 while (fgets(line, sizeof(line), bpipe->rfd)) {
193 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
195 status = close_bpipe(bpipe);
197 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
199 set_jcr_job_status(jcr, JS_FatalError);
200 update_job_end_record(jcr);
204 switch (jcr->JobType) {
207 if (jcr->JobStatus == JS_Terminated) {
213 if (jcr->JobStatus == JS_Terminated) {
219 if (jcr->JobStatus == JS_Terminated) {
225 if (jcr->JobStatus == JS_Terminated) {
230 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
233 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
234 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
235 POOLMEM *after = get_pool_memory(PM_FNAME);
238 char line[MAXSTRING];
240 if (jcr->JobStatus == JS_Terminated) {
241 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
243 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
245 bpipe = open_bpipe(after, 0, "r");
246 free_pool_memory(after);
247 while (fgets(line, sizeof(line), bpipe->rfd)) {
248 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
250 status = close_bpipe(bpipe);
252 * Note, if we get an error here, do not mark the
253 * job in error, simply report the error condition.
256 if (jcr->JobStatus == JS_Terminated) {
257 Jmsg(jcr, M_ERROR, 0, _("RunAfterJob returned non-zero status=%d\n"),
260 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
265 /* Send off any queued messages */
266 if (jcr->msg_queue->size() > 0) {
267 dequeue_messages(jcr);
274 Dmsg0(50, "======== End Job ==========\n");
275 sm_check(__FILE__, __LINE__, True);
281 * Cancel a job -- typically called by the UA (Console program), but may also
282 * be called by the job watchdog.
284 * Returns: 1 if cancel appears to be successful
285 * 0 on failure. Message sent to ua->jcr.
287 int cancel_job(UAContext *ua, JCR *jcr)
291 switch (jcr->JobStatus) {
294 case JS_WaitClientRes:
295 case JS_WaitStoreRes:
296 case JS_WaitPriority:
298 case JS_WaitStartTime:
299 set_jcr_job_status(jcr, JS_Canceled);
300 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
301 jcr->JobId, jcr->Job);
302 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
306 set_jcr_job_status(jcr, JS_Canceled);
308 /* Cancel File daemon */
309 if (jcr->file_bsock) {
310 ua->jcr->client = jcr->client;
311 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
312 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
315 Dmsg0(200, "Connected to file daemon\n");
316 fd = ua->jcr->file_bsock;
317 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
318 while (bnet_recv(fd) >= 0) {
319 bsendmsg(ua, "%s", fd->msg);
321 bnet_sig(fd, BNET_TERMINATE);
323 ua->jcr->file_bsock = NULL;
326 /* Cancel Storage daemon */
327 if (jcr->store_bsock) {
328 ua->jcr->store = jcr->store;
329 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
330 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
333 Dmsg0(200, "Connected to storage daemon\n");
334 sd = ua->jcr->store_bsock;
335 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
336 while (bnet_recv(sd) >= 0) {
337 bsendmsg(ua, "%s", sd->msg);
339 bnet_sig(sd, BNET_TERMINATE);
341 ua->jcr->store_bsock = NULL;
349 static void job_monitor_destructor(watchdog_t *self)
351 JCR *control_jcr = (JCR *) self->data;
353 free_jcr(control_jcr);
356 static void job_monitor_watchdog(watchdog_t *self)
358 JCR *control_jcr, *jcr;
360 control_jcr = (JCR *)self->data;
362 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
369 if (jcr->JobId == 0) {
370 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
372 /* Keep reference counts correct */
373 free_locked_jcr(jcr);
377 /* check MaxWaitTime */
378 cancel = job_check_maxwaittime(control_jcr, jcr);
380 /* check MaxRunTime */
381 cancel |= job_check_maxruntime(control_jcr, jcr);
384 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
385 jcr, jcr->JobId, jcr->Job);
387 UAContext *ua = new_ua_context(jcr);
388 ua->jcr = control_jcr;
392 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
395 /* Keep reference counts correct */
396 free_locked_jcr(jcr);
402 * Check if the maxwaittime has expired and it is possible
405 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
409 if (jcr->job->MaxWaitTime == 0) {
412 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
413 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
414 jcr, jcr->Job, jcr->job->MaxWaitTime);
417 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
419 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
420 switch (jcr->JobStatus) {
425 case JS_WaitStoreRes:
426 case JS_WaitClientRes:
428 case JS_WaitPriority:
430 case JS_WaitStartTime:
432 Dmsg0(200, "JCR blocked in #1\n");
435 Dmsg0(200, "JCR running, checking SD status\n");
436 switch (jcr->SDJobStatus) {
441 Dmsg0(200, "JCR blocked in #2\n");
444 Dmsg0(200, "JCR not blocked in #2\n");
449 case JS_ErrorTerminated:
452 Dmsg0(200, "JCR already dead in #3\n");
455 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
458 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
459 cancel ? "" : "do not ", jcr, jcr->job);
465 * Check if maxruntime has expired and if the job can be
468 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
472 if (jcr->job->MaxRunTime == 0) {
475 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
476 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
477 jcr, jcr->Job, jcr->job->MaxRunTime);
481 switch (jcr->JobStatus) {
487 case JS_WaitStoreRes:
488 case JS_WaitClientRes:
490 case JS_WaitPriority:
492 case JS_WaitStartTime:
497 case JS_ErrorTerminated:
503 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
507 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
508 cancel ? "" : "do not ", jcr, jcr->job);
515 * Get or create a Client record for this Job
517 int get_or_create_client_record(JCR *jcr)
521 memset(&cr, 0, sizeof(cr));
522 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
523 cr.AutoPrune = jcr->client->AutoPrune;
524 cr.FileRetention = jcr->client->FileRetention;
525 cr.JobRetention = jcr->client->JobRetention;
526 if (!jcr->client_name) {
527 jcr->client_name = get_pool_memory(PM_NAME);
529 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
530 if (!db_create_client_record(jcr, jcr->db, &cr)) {
531 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
532 db_strerror(jcr->db));
535 jcr->jr.ClientId = cr.ClientId;
537 if (!jcr->client_uname) {
538 jcr->client_uname = get_pool_memory(PM_NAME);
540 pm_strcpy(&jcr->client_uname, cr.Uname);
542 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
549 * Write status and such in DB
551 void update_job_end_record(JCR *jcr)
553 if (jcr->jr.EndTime == 0) {
554 jcr->jr.EndTime = time(NULL);
556 jcr->end_time = jcr->jr.EndTime;
557 jcr->jr.JobId = jcr->JobId;
558 jcr->jr.JobStatus = jcr->JobStatus;
559 jcr->jr.JobFiles = jcr->JobFiles;
560 jcr->jr.JobBytes = jcr->JobBytes;
561 jcr->jr.VolSessionId = jcr->VolSessionId;
562 jcr->jr.VolSessionTime = jcr->VolSessionTime;
563 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
564 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
565 db_strerror(jcr->db));
570 * Takes base_name and appends (unique) current
571 * date and time to form unique job name.
573 * Returns: unique job name in jcr->Job
574 * date/time in jcr->start_time
576 void create_unique_job_name(JCR *jcr, char *base_name)
578 /* Job start mutex */
579 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
580 static time_t last_start_time = 0;
583 char dt[MAX_TIME_LENGTH];
584 char name[MAX_NAME_LENGTH];
587 /* Guarantee unique start time -- maximum one per second, and
588 * thus unique Job Name
590 P(mutex); /* lock creation of jobs */
592 while (now == last_start_time) {
593 bmicrosleep(0, 500000);
596 last_start_time = now;
597 V(mutex); /* allow creation of jobs */
598 jcr->start_time = now;
599 /* Form Unique JobName */
600 localtime_r(&now, &tm);
601 /* Use only characters that are permitted in Windows filenames */
602 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
603 bstrncpy(name, base_name, sizeof(name));
604 name[sizeof(name)-22] = 0; /* truncate if too long */
605 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
606 /* Convert spaces into underscores */
607 for (p=jcr->Job; *p; p++) {
615 * Free the Job Control Record if no one is still using it.
616 * Called from main free_jcr() routine in src/lib/jcr.c so
617 * that we can do our Director specific cleanup of the jcr.
619 void dird_free_jcr(JCR *jcr)
621 Dmsg0(200, "Start dird free_jcr\n");
623 if (jcr->sd_auth_key) {
624 free(jcr->sd_auth_key);
625 jcr->sd_auth_key = NULL;
631 if (jcr->file_bsock) {
632 Dmsg0(200, "Close File bsock\n");
633 bnet_close(jcr->file_bsock);
634 jcr->file_bsock = NULL;
636 if (jcr->store_bsock) {
637 Dmsg0(200, "Close Store bsock\n");
638 bnet_close(jcr->store_bsock);
639 jcr->store_bsock = NULL;
642 Dmsg0(200, "Free JCR fname\n");
643 free_pool_memory(jcr->fname);
647 Dmsg0(200, "Free JCR stime\n");
648 free_pool_memory(jcr->stime);
651 if (jcr->RestoreBootstrap) {
652 free(jcr->RestoreBootstrap);
653 jcr->RestoreBootstrap = NULL;
655 if (jcr->client_uname) {
656 free_pool_memory(jcr->client_uname);
657 jcr->client_uname = NULL;
659 pthread_cond_destroy(&jcr->term_wait);
660 Dmsg0(200, "End dird free_jcr\n");
664 * Set some defaults in the JCR necessary to
665 * run. These items are pulled from the job
666 * definition as defaults, but can be overridden
667 * later either by the Run record in the Schedule resource,
668 * or by the Console program.
670 void set_jcr_defaults(JCR *jcr, JOB *job)
673 jcr->JobType = job->JobType;
674 switch (jcr->JobType) {
677 jcr->JobLevel = L_NONE;
680 jcr->JobLevel = job->level;
683 jcr->JobPriority = job->Priority;
684 jcr->store = job->storage;
685 jcr->client = job->client;
686 if (!jcr->client_name) {
687 jcr->client_name = get_pool_memory(PM_NAME);
689 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
690 jcr->pool = job->pool;
691 jcr->full_pool = job->full_pool;
692 jcr->inc_pool = job->inc_pool;
693 jcr->dif_pool = job->dif_pool;
694 jcr->catalog = job->client->catalog;
695 jcr->fileset = job->fileset;
696 jcr->messages = job->messages;
697 if (jcr->RestoreBootstrap) {
698 free(jcr->RestoreBootstrap);
699 jcr->RestoreBootstrap = NULL;
701 /* This can be overridden by Console program */
702 if (job->RestoreBootstrap) {
703 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
705 /* If no default level given, set one */
706 if (jcr->JobLevel == 0) {
707 switch (jcr->JobType) {
709 jcr->JobLevel = L_VERIFY_CATALOG;
712 jcr->JobLevel = L_INCREMENTAL;
716 jcr->JobLevel = L_NONE;