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);
85 create_unique_job_name(jcr, jcr->job->hdr.name);
86 set_jcr_job_status(jcr, JS_Created);
87 jcr->jr.SchedTime = jcr->sched_time;
88 jcr->jr.StartTime = jcr->start_time;
89 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
90 jcr->jr.Type = jcr->JobType;
91 jcr->jr.Level = jcr->JobLevel;
92 jcr->jr.JobStatus = jcr->JobStatus;
93 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
94 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
96 /* Initialize termination condition variable */
97 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
98 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
105 Dmsg0(50, "Open database\n");
106 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
107 jcr->catalog->db_password, jcr->catalog->db_address,
108 jcr->catalog->db_port, jcr->catalog->db_socket);
109 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
110 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
111 jcr->catalog->db_name);
113 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
117 Dmsg0(50, "DB opened\n");
122 jcr->jr.JobStatus = jcr->JobStatus;
123 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
124 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
127 jcr->JobId = jcr->jr.JobId;
129 Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
130 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
131 Dmsg0(200, "Add jrc to work queue\n");
133 /* Queue the job to be run */
134 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
135 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
138 Dmsg0(100, "Done run_job()\n");
144 set_jcr_job_status(jcr, JS_ErrorTerminated);
152 * This is the engine called by jobq.c:jobq_add() when we were pulled
153 * from the work queue.
154 * At this point, we are running in our own thread and all
155 * necessary resources are allocated -- see jobq.c
157 static void *job_thread(void *arg)
159 JCR *jcr = (JCR *)arg;
161 jcr->my_thread_id = pthread_self();
162 pthread_detach(jcr->my_thread_id);
163 sm_check(__FILE__, __LINE__, true);
167 Dmsg0(200, "=====Start Job=========\n");
168 jcr->start_time = time(NULL); /* set the real start time */
169 set_jcr_job_status(jcr, JS_Running);
171 if (job_canceled(jcr)) {
172 update_job_end_record(jcr);
173 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
174 (utime_t)(jcr->start_time - jcr->sched_time)) {
175 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
176 set_jcr_job_status(jcr, JS_Canceled);
177 update_job_end_record(jcr);
181 if (jcr->job->RunBeforeJob) {
182 POOLMEM *before = get_pool_memory(PM_FNAME);
185 char line[MAXSTRING];
187 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
188 bpipe = open_bpipe(before, 0, "r");
189 free_pool_memory(before);
190 while (fgets(line, sizeof(line), bpipe->rfd)) {
191 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
193 status = close_bpipe(bpipe);
195 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
197 set_jcr_job_status(jcr, JS_FatalError);
198 update_job_end_record(jcr);
202 switch (jcr->JobType) {
205 if (jcr->JobStatus == JS_Terminated) {
211 if (jcr->JobStatus == JS_Terminated) {
217 if (jcr->JobStatus == JS_Terminated) {
223 if (jcr->JobStatus == JS_Terminated) {
228 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
231 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
232 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
233 POOLMEM *after = get_pool_memory(PM_FNAME);
236 char line[MAXSTRING];
238 if (jcr->JobStatus == JS_Terminated) {
239 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
241 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
243 bpipe = open_bpipe(after, 0, "r");
244 free_pool_memory(after);
245 while (fgets(line, sizeof(line), bpipe->rfd)) {
246 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
248 status = close_bpipe(bpipe);
250 * Note, if we get an error here, do not mark the
251 * job in error, simply report the error condition.
254 if (jcr->JobStatus == JS_Terminated) {
255 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
258 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
263 /* Send off any queued messages */
264 if (jcr->msg_queue->size() > 0) {
265 dequeue_messages(jcr);
272 Dmsg0(50, "======== End Job ==========\n");
273 sm_check(__FILE__, __LINE__, true);
279 * Cancel a job -- typically called by the UA (Console program), but may also
280 * be called by the job watchdog.
282 * Returns: 1 if cancel appears to be successful
283 * 0 on failure. Message sent to ua->jcr.
285 int cancel_job(UAContext *ua, JCR *jcr)
289 switch (jcr->JobStatus) {
292 case JS_WaitClientRes:
293 case JS_WaitStoreRes:
294 case JS_WaitPriority:
296 case JS_WaitStartTime:
297 set_jcr_job_status(jcr, JS_Canceled);
298 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
299 jcr->JobId, jcr->Job);
300 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
304 set_jcr_job_status(jcr, JS_Canceled);
306 /* Cancel File daemon */
307 if (jcr->file_bsock) {
308 ua->jcr->client = jcr->client;
309 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
310 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
313 Dmsg0(200, "Connected to file daemon\n");
314 fd = ua->jcr->file_bsock;
315 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
316 while (bnet_recv(fd) >= 0) {
317 bsendmsg(ua, "%s", fd->msg);
319 bnet_sig(fd, BNET_TERMINATE);
321 ua->jcr->file_bsock = NULL;
324 /* Cancel Storage daemon */
325 if (jcr->store_bsock) {
326 ua->jcr->store = jcr->store;
327 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
328 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
331 Dmsg0(200, "Connected to storage daemon\n");
332 sd = ua->jcr->store_bsock;
333 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
334 while (bnet_recv(sd) >= 0) {
335 bsendmsg(ua, "%s", sd->msg);
337 bnet_sig(sd, BNET_TERMINATE);
339 ua->jcr->store_bsock = NULL;
347 static void job_monitor_destructor(watchdog_t *self)
349 JCR *control_jcr = (JCR *) self->data;
351 free_jcr(control_jcr);
354 static void job_monitor_watchdog(watchdog_t *self)
356 JCR *control_jcr, *jcr;
358 control_jcr = (JCR *)self->data;
360 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
367 if (jcr->JobId == 0) {
368 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
370 /* Keep reference counts correct */
371 free_locked_jcr(jcr);
375 /* check MaxWaitTime */
376 cancel = job_check_maxwaittime(control_jcr, jcr);
378 /* check MaxRunTime */
379 cancel |= job_check_maxruntime(control_jcr, jcr);
382 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
383 jcr, jcr->JobId, jcr->Job);
385 UAContext *ua = new_ua_context(jcr);
386 ua->jcr = control_jcr;
390 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
393 /* Keep reference counts correct */
394 free_locked_jcr(jcr);
400 * Check if the maxwaittime has expired and it is possible
403 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
407 if (jcr->job->MaxWaitTime == 0) {
410 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
411 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
412 jcr, jcr->Job, jcr->job->MaxWaitTime);
415 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
417 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
418 switch (jcr->JobStatus) {
423 case JS_WaitStoreRes:
424 case JS_WaitClientRes:
426 case JS_WaitPriority:
428 case JS_WaitStartTime:
430 Dmsg0(200, "JCR blocked in #1\n");
433 Dmsg0(200, "JCR running, checking SD status\n");
434 switch (jcr->SDJobStatus) {
439 Dmsg0(200, "JCR blocked in #2\n");
442 Dmsg0(200, "JCR not blocked in #2\n");
447 case JS_ErrorTerminated:
450 Dmsg0(200, "JCR already dead in #3\n");
453 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
456 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
457 cancel ? "" : "do not ", jcr, jcr->job);
463 * Check if maxruntime has expired and if the job can be
466 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
470 if (jcr->job->MaxRunTime == 0) {
473 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
474 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
475 jcr, jcr->Job, jcr->job->MaxRunTime);
479 switch (jcr->JobStatus) {
485 case JS_WaitStoreRes:
486 case JS_WaitClientRes:
488 case JS_WaitPriority:
490 case JS_WaitStartTime:
495 case JS_ErrorTerminated:
501 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
505 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
506 cancel ? "" : "do not ", jcr, jcr->job);
513 * Get or create a Client record for this Job
515 int get_or_create_client_record(JCR *jcr)
519 memset(&cr, 0, sizeof(cr));
520 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
521 cr.AutoPrune = jcr->client->AutoPrune;
522 cr.FileRetention = jcr->client->FileRetention;
523 cr.JobRetention = jcr->client->JobRetention;
524 if (!jcr->client_name) {
525 jcr->client_name = get_pool_memory(PM_NAME);
527 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
528 if (!db_create_client_record(jcr, jcr->db, &cr)) {
529 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
530 db_strerror(jcr->db));
533 jcr->jr.ClientId = cr.ClientId;
535 if (!jcr->client_uname) {
536 jcr->client_uname = get_pool_memory(PM_NAME);
538 pm_strcpy(&jcr->client_uname, cr.Uname);
540 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
547 * Write status and such in DB
549 void update_job_end_record(JCR *jcr)
551 if (jcr->jr.EndTime == 0) {
552 jcr->jr.EndTime = time(NULL);
554 jcr->end_time = jcr->jr.EndTime;
555 jcr->jr.JobId = jcr->JobId;
556 jcr->jr.JobStatus = jcr->JobStatus;
557 jcr->jr.JobFiles = jcr->JobFiles;
558 jcr->jr.JobBytes = jcr->JobBytes;
559 jcr->jr.VolSessionId = jcr->VolSessionId;
560 jcr->jr.VolSessionTime = jcr->VolSessionTime;
561 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
562 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
563 db_strerror(jcr->db));
568 * Takes base_name and appends (unique) current
569 * date and time to form unique job name.
571 * Returns: unique job name in jcr->Job
572 * date/time in jcr->start_time
574 void create_unique_job_name(JCR *jcr, char *base_name)
576 /* Job start mutex */
577 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
578 static time_t last_start_time = 0;
581 char dt[MAX_TIME_LENGTH];
582 char name[MAX_NAME_LENGTH];
585 /* Guarantee unique start time -- maximum one per second, and
586 * thus unique Job Name
588 P(mutex); /* lock creation of jobs */
590 while (now == last_start_time) {
591 bmicrosleep(0, 500000);
594 last_start_time = now;
595 V(mutex); /* allow creation of jobs */
596 jcr->start_time = now;
597 /* Form Unique JobName */
598 localtime_r(&now, &tm);
599 /* Use only characters that are permitted in Windows filenames */
600 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
601 bstrncpy(name, base_name, sizeof(name));
602 name[sizeof(name)-22] = 0; /* truncate if too long */
603 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
604 /* Convert spaces into underscores */
605 for (p=jcr->Job; *p; p++) {
613 * Free the Job Control Record if no one is still using it.
614 * Called from main free_jcr() routine in src/lib/jcr.c so
615 * that we can do our Director specific cleanup of the jcr.
617 void dird_free_jcr(JCR *jcr)
619 Dmsg0(200, "Start dird free_jcr\n");
621 if (jcr->sd_auth_key) {
622 free(jcr->sd_auth_key);
623 jcr->sd_auth_key = NULL;
629 if (jcr->file_bsock) {
630 Dmsg0(200, "Close File bsock\n");
631 bnet_close(jcr->file_bsock);
632 jcr->file_bsock = NULL;
634 if (jcr->store_bsock) {
635 Dmsg0(200, "Close Store bsock\n");
636 bnet_close(jcr->store_bsock);
637 jcr->store_bsock = NULL;
640 Dmsg0(200, "Free JCR fname\n");
641 free_pool_memory(jcr->fname);
645 Dmsg0(200, "Free JCR stime\n");
646 free_pool_memory(jcr->stime);
649 if (jcr->RestoreBootstrap) {
650 free(jcr->RestoreBootstrap);
651 jcr->RestoreBootstrap = NULL;
653 if (jcr->client_uname) {
654 free_pool_memory(jcr->client_uname);
655 jcr->client_uname = NULL;
657 pthread_cond_destroy(&jcr->term_wait);
658 Dmsg0(200, "End dird free_jcr\n");
662 * Set some defaults in the JCR necessary to
663 * run. These items are pulled from the job
664 * definition as defaults, but can be overridden
665 * later either by the Run record in the Schedule resource,
666 * or by the Console program.
668 void set_jcr_defaults(JCR *jcr, JOB *job)
671 jcr->JobType = job->JobType;
672 switch (jcr->JobType) {
675 jcr->JobLevel = L_NONE;
678 jcr->JobLevel = job->level;
681 jcr->JobPriority = job->Priority;
682 jcr->store = job->storage;
683 jcr->client = job->client;
684 if (!jcr->client_name) {
685 jcr->client_name = get_pool_memory(PM_NAME);
687 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
688 jcr->pool = job->pool;
689 jcr->full_pool = job->full_pool;
690 jcr->inc_pool = job->inc_pool;
691 jcr->dif_pool = job->dif_pool;
692 jcr->catalog = job->client->catalog;
693 jcr->fileset = job->fileset;
694 jcr->messages = job->messages;
695 jcr->spool_data = job->spool_data;
696 if (jcr->RestoreBootstrap) {
697 free(jcr->RestoreBootstrap);
698 jcr->RestoreBootstrap = NULL;
700 /* This can be overridden by Console program */
701 if (job->RestoreBootstrap) {
702 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
704 /* If no default level given, set one */
705 if (jcr->JobLevel == 0) {
706 switch (jcr->JobType) {
708 jcr->JobLevel = L_VERIFY_CATALOG;
711 jcr->JobLevel = L_INCREMENTAL;
715 jcr->JobLevel = L_NONE;