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);
76 * Run a job -- typically called by the scheduler, but may also
77 * be called by the UA (Console program).
80 void run_job(JCR *jcr)
85 sm_check(__FILE__, __LINE__, True);
86 init_msg(jcr, jcr->messages);
87 create_unique_job_name(jcr, jcr->job->hdr.name);
88 set_jcr_job_status(jcr, JS_Created);
89 jcr->jr.SchedTime = jcr->sched_time;
90 jcr->jr.StartTime = jcr->start_time;
91 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
92 jcr->jr.Type = jcr->JobType;
93 jcr->jr.Level = jcr->JobLevel;
94 jcr->jr.JobStatus = jcr->JobStatus;
95 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
96 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
98 /* Initialize termination condition variable */
99 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
100 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
107 Dmsg0(50, "Open database\n");
108 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
109 jcr->catalog->db_password, jcr->catalog->db_address,
110 jcr->catalog->db_port, jcr->catalog->db_socket);
111 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
112 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
113 jcr->catalog->db_name);
115 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
119 Dmsg0(50, "DB opened\n");
124 jcr->jr.JobStatus = jcr->JobStatus;
125 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
126 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
129 jcr->JobId = jcr->jr.JobId;
131 Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
132 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
133 Dmsg0(200, "Add jrc to work queue\n");
135 /* Queue the job to be run */
136 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
137 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
140 Dmsg0(100, "Done run_job()\n");
146 set_jcr_job_status(jcr, JS_ErrorTerminated);
154 * This is the engine called by jobq.c:jobq_add() when we were pulled
155 * from the work queue.
156 * At this point, we are running in our own thread and all
157 * necessary resources are allocated -- see jobq.c
159 static void *job_thread(void *arg)
161 JCR *jcr = (JCR *)arg;
163 pthread_detach(pthread_self());
164 sm_check(__FILE__, __LINE__, True);
168 Dmsg0(200, "=====Start Job=========\n");
169 jcr->start_time = time(NULL); /* set the real start time */
170 set_jcr_job_status(jcr, JS_Running);
172 if (job_canceled(jcr)) {
173 update_job_end_record(jcr);
174 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
175 (utime_t)(jcr->start_time - jcr->sched_time)) {
176 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
177 set_jcr_job_status(jcr, JS_Canceled);
178 update_job_end_record(jcr);
182 if (jcr->job->RunBeforeJob) {
183 POOLMEM *before = get_pool_memory(PM_FNAME);
186 char line[MAXSTRING];
188 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
189 bpipe = open_bpipe(before, 0, "r");
190 free_pool_memory(before);
191 while (fgets(line, sizeof(line), bpipe->rfd)) {
192 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
194 status = close_bpipe(bpipe);
196 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
198 set_jcr_job_status(jcr, JS_FatalError);
199 update_job_end_record(jcr);
203 switch (jcr->JobType) {
206 if (jcr->JobStatus == JS_Terminated) {
212 if (jcr->JobStatus == JS_Terminated) {
218 if (jcr->JobStatus == JS_Terminated) {
224 if (jcr->JobStatus == JS_Terminated) {
229 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
232 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
233 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
234 POOLMEM *after = get_pool_memory(PM_FNAME);
237 char line[MAXSTRING];
239 if (jcr->JobStatus == JS_Terminated) {
240 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
242 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
244 bpipe = open_bpipe(after, 0, "r");
245 free_pool_memory(after);
246 while (fgets(line, sizeof(line), bpipe->rfd)) {
247 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
249 status = close_bpipe(bpipe);
251 * Note, if we get an error here, do not mark the
252 * job in error, simply report the error condition.
255 if (jcr->JobStatus == JS_Terminated) {
256 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
259 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
264 /* Send off any queued messages */
265 if (jcr->msg_queue->size() > 0) {
266 dequeue_messages(jcr);
273 Dmsg0(50, "======== End Job ==========\n");
274 sm_check(__FILE__, __LINE__, True);
280 * Cancel a job -- typically called by the UA (Console program), but may also
281 * be called by the job watchdog.
283 * Returns: 1 if cancel appears to be successful
284 * 0 on failure. Message sent to ua->jcr.
286 int cancel_job(UAContext *ua, JCR *jcr)
290 switch (jcr->JobStatus) {
293 case JS_WaitClientRes:
294 case JS_WaitStoreRes:
295 case JS_WaitPriority:
297 case JS_WaitStartTime:
298 set_jcr_job_status(jcr, JS_Canceled);
299 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
300 jcr->JobId, jcr->Job);
301 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
305 set_jcr_job_status(jcr, JS_Canceled);
307 /* Cancel File daemon */
308 if (jcr->file_bsock) {
309 ua->jcr->client = jcr->client;
310 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
311 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
314 Dmsg0(200, "Connected to file daemon\n");
315 fd = ua->jcr->file_bsock;
316 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
317 while (bnet_recv(fd) >= 0) {
318 bsendmsg(ua, "%s", fd->msg);
320 bnet_sig(fd, BNET_TERMINATE);
322 ua->jcr->file_bsock = NULL;
325 /* Cancel Storage daemon */
326 if (jcr->store_bsock) {
327 ua->jcr->store = jcr->store;
328 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
329 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
332 Dmsg0(200, "Connected to storage daemon\n");
333 sd = ua->jcr->store_bsock;
334 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
335 while (bnet_recv(sd) >= 0) {
336 bsendmsg(ua, "%s", sd->msg);
338 bnet_sig(sd, BNET_TERMINATE);
340 ua->jcr->store_bsock = NULL;
348 static void job_monitor_destructor(watchdog_t *self)
350 JCR *control_jcr = (JCR *) self->data;
352 free_jcr(control_jcr);
355 static void job_monitor_watchdog(watchdog_t *self)
357 JCR *control_jcr, *jcr;
359 control_jcr = (JCR *)self->data;
361 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
368 if (jcr->JobId == 0) {
369 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
371 /* Keep reference counts correct */
372 free_locked_jcr(jcr);
376 /* check MaxWaitTime */
377 cancel = job_check_maxwaittime(control_jcr, jcr);
379 /* check MaxRunTime */
380 cancel |= job_check_maxruntime(control_jcr, jcr);
383 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
384 jcr, jcr->JobId, jcr->Job);
386 UAContext *ua = new_ua_context(jcr);
387 ua->jcr = control_jcr;
391 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
394 /* Keep reference counts correct */
395 free_locked_jcr(jcr);
401 * Check if the maxwaittime has expired and it is possible
404 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
408 if (jcr->job->MaxWaitTime == 0) {
411 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
412 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
413 jcr, jcr->Job, jcr->job->MaxWaitTime);
416 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
418 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
419 switch (jcr->JobStatus) {
424 case JS_WaitStoreRes:
425 case JS_WaitClientRes:
427 case JS_WaitPriority:
429 case JS_WaitStartTime:
431 Dmsg0(200, "JCR blocked in #1\n");
434 Dmsg0(200, "JCR running, checking SD status\n");
435 switch (jcr->SDJobStatus) {
440 Dmsg0(200, "JCR blocked in #2\n");
443 Dmsg0(200, "JCR not blocked in #2\n");
448 case JS_ErrorTerminated:
451 Dmsg0(200, "JCR already dead in #3\n");
454 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
457 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
458 cancel ? "" : "do not ", jcr, jcr->job);
464 * Check if maxruntime has expired and if the job can be
467 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
471 if (jcr->job->MaxRunTime == 0) {
474 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
475 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
476 jcr, jcr->Job, jcr->job->MaxRunTime);
480 switch (jcr->JobStatus) {
486 case JS_WaitStoreRes:
487 case JS_WaitClientRes:
489 case JS_WaitPriority:
491 case JS_WaitStartTime:
496 case JS_ErrorTerminated:
502 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
506 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
507 cancel ? "" : "do not ", jcr, jcr->job);
514 * Get or create a Client record for this Job
516 int get_or_create_client_record(JCR *jcr)
520 memset(&cr, 0, sizeof(cr));
521 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
522 cr.AutoPrune = jcr->client->AutoPrune;
523 cr.FileRetention = jcr->client->FileRetention;
524 cr.JobRetention = jcr->client->JobRetention;
525 if (!jcr->client_name) {
526 jcr->client_name = get_pool_memory(PM_NAME);
528 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
529 if (!db_create_client_record(jcr, jcr->db, &cr)) {
530 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
531 db_strerror(jcr->db));
534 jcr->jr.ClientId = cr.ClientId;
536 if (!jcr->client_uname) {
537 jcr->client_uname = get_pool_memory(PM_NAME);
539 pm_strcpy(&jcr->client_uname, cr.Uname);
541 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
548 * Write status and such in DB
550 void update_job_end_record(JCR *jcr)
552 if (jcr->jr.EndTime == 0) {
553 jcr->jr.EndTime = time(NULL);
555 jcr->end_time = jcr->jr.EndTime;
556 jcr->jr.JobId = jcr->JobId;
557 jcr->jr.JobStatus = jcr->JobStatus;
558 jcr->jr.JobFiles = jcr->JobFiles;
559 jcr->jr.JobBytes = jcr->JobBytes;
560 jcr->jr.VolSessionId = jcr->VolSessionId;
561 jcr->jr.VolSessionTime = jcr->VolSessionTime;
562 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
563 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
564 db_strerror(jcr->db));
569 * Takes base_name and appends (unique) current
570 * date and time to form unique job name.
572 * Returns: unique job name in jcr->Job
573 * date/time in jcr->start_time
575 void create_unique_job_name(JCR *jcr, char *base_name)
577 /* Job start mutex */
578 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
579 static time_t last_start_time = 0;
582 char dt[MAX_TIME_LENGTH];
583 char name[MAX_NAME_LENGTH];
586 /* Guarantee unique start time -- maximum one per second, and
587 * thus unique Job Name
589 P(mutex); /* lock creation of jobs */
591 while (now == last_start_time) {
592 bmicrosleep(0, 500000);
595 last_start_time = now;
596 V(mutex); /* allow creation of jobs */
597 jcr->start_time = now;
598 /* Form Unique JobName */
599 localtime_r(&now, &tm);
600 /* Use only characters that are permitted in Windows filenames */
601 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
602 bstrncpy(name, base_name, sizeof(name));
603 name[sizeof(name)-22] = 0; /* truncate if too long */
604 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
605 /* Convert spaces into underscores */
606 for (p=jcr->Job; *p; p++) {
614 * Free the Job Control Record if no one is still using it.
615 * Called from main free_jcr() routine in src/lib/jcr.c so
616 * that we can do our Director specific cleanup of the jcr.
618 void dird_free_jcr(JCR *jcr)
620 Dmsg0(200, "Start dird free_jcr\n");
622 if (jcr->sd_auth_key) {
623 free(jcr->sd_auth_key);
624 jcr->sd_auth_key = NULL;
630 if (jcr->file_bsock) {
631 Dmsg0(200, "Close File bsock\n");
632 bnet_close(jcr->file_bsock);
633 jcr->file_bsock = NULL;
635 if (jcr->store_bsock) {
636 Dmsg0(200, "Close Store bsock\n");
637 bnet_close(jcr->store_bsock);
638 jcr->store_bsock = NULL;
641 Dmsg0(200, "Free JCR fname\n");
642 free_pool_memory(jcr->fname);
646 Dmsg0(200, "Free JCR stime\n");
647 free_pool_memory(jcr->stime);
650 if (jcr->RestoreBootstrap) {
651 free(jcr->RestoreBootstrap);
652 jcr->RestoreBootstrap = NULL;
654 if (jcr->client_uname) {
655 free_pool_memory(jcr->client_uname);
656 jcr->client_uname = NULL;
658 pthread_cond_destroy(&jcr->term_wait);
659 Dmsg0(200, "End dird free_jcr\n");
663 * Set some defaults in the JCR necessary to
664 * run. These items are pulled from the job
665 * definition as defaults, but can be overridden
666 * later either by the Run record in the Schedule resource,
667 * or by the Console program.
669 void set_jcr_defaults(JCR *jcr, JOB *job)
672 jcr->JobType = job->JobType;
673 switch (jcr->JobType) {
676 jcr->JobLevel = L_NONE;
679 jcr->JobLevel = job->level;
682 jcr->JobPriority = job->Priority;
683 jcr->store = job->storage;
684 jcr->client = job->client;
685 if (!jcr->client_name) {
686 jcr->client_name = get_pool_memory(PM_NAME);
688 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
689 jcr->pool = job->pool;
690 jcr->full_pool = job->full_pool;
691 jcr->inc_pool = job->inc_pool;
692 jcr->dif_pool = job->dif_pool;
693 jcr->catalog = job->client->catalog;
694 jcr->fileset = job->fileset;
695 jcr->messages = job->messages;
696 jcr->spool_data = job->spool_data;
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;