3 * Bacula Director Job processing routines
5 * Kern Sibbald, October MM
10 Copyright (C) 2000-2003 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 = watchdog_new()) == 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);
75 static void job_monitor_destructor(watchdog_t *self)
77 JCR *control_jcr = (JCR *) self->data;
79 free_jcr(control_jcr);
82 static void job_monitor_watchdog(watchdog_t *self)
84 JCR *control_jcr, *jcr;
86 control_jcr = (JCR *) self->data;
88 Dmsg1(200, "job_monitor_watchdog %p called\n", self);
92 for (jcr = NULL; (jcr = get_next_jcr(jcr)); /* nothing */) {
95 if (jcr->JobId == 0) {
96 Dmsg2(200, "Skipping JCR %p (%s) with JobId 0\n",
98 /* Keep reference counts correct */
103 /* check MaxWaitTime */
104 cancel = job_check_maxwaittime(control_jcr, jcr);
106 /* check MaxRunTime */
107 cancel |= job_check_maxruntime(control_jcr, jcr);
110 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
111 jcr, jcr->JobId, jcr->Job);
113 UAContext *ua = new_ua_context(jcr);
114 ua->jcr = control_jcr;
118 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
121 /* Keep reference counts correct */
122 free_locked_jcr(jcr);
127 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
131 if (jcr->job->MaxWaitTime == 0) {
134 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
135 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
136 jcr, jcr->Job, jcr->job->MaxWaitTime);
139 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
141 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
142 switch (jcr->JobStatus) {
147 case JS_WaitStoreRes:
148 case JS_WaitClientRes:
150 case JS_WaitPriority:
152 case JS_WaitStartTime:
154 Dmsg0(200, "JCR blocked in #1\n");
157 Dmsg0(200, "JCR running, checking SD status\n");
158 switch (jcr->SDJobStatus) {
163 Dmsg0(200, "JCR blocked in #2\n");
166 Dmsg0(200, "JCR not blocked in #2\n");
171 case JS_ErrorTerminated:
173 Dmsg0(200, "JCR already dead in #3\n");
176 Emsg1(M_ABORT, 0, _("Unhandled job status code %d\n"),
179 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
180 cancel ? "" : "do not ", jcr, jcr->job);
185 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
189 if (jcr->job->MaxRunTime == 0) {
192 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
193 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
194 jcr, jcr->Job, jcr->job->MaxRunTime);
198 switch (jcr->JobStatus) {
203 case JS_WaitStoreRes:
204 case JS_WaitClientRes:
206 case JS_WaitPriority:
208 case JS_WaitStartTime:
213 case JS_ErrorTerminated:
218 Emsg1(M_ABORT, 0, _("Unhandled job status code %d\n"),
222 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
223 cancel ? "" : "do not ", jcr, jcr->job);
229 * Run a job -- typically called by the scheduler, but may also
230 * be called by the UA (Console program).
233 void run_job(JCR *jcr)
238 sm_check(__FILE__, __LINE__, True);
239 init_msg(jcr, jcr->messages);
240 create_unique_job_name(jcr, jcr->job->hdr.name);
241 set_jcr_job_status(jcr, JS_Created);
242 jcr->jr.SchedTime = jcr->sched_time;
243 jcr->jr.StartTime = jcr->start_time;
244 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
245 jcr->jr.Type = jcr->JobType;
246 jcr->jr.Level = jcr->JobLevel;
247 jcr->jr.JobStatus = jcr->JobStatus;
248 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
249 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
251 /* Initialize termination condition variable */
252 if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
253 Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
260 Dmsg0(50, "Open database\n");
261 jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
262 jcr->catalog->db_password, jcr->catalog->db_address,
263 jcr->catalog->db_port, jcr->catalog->db_socket);
264 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
265 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
266 jcr->catalog->db_name);
268 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
272 Dmsg0(50, "DB opened\n");
277 jcr->jr.JobStatus = jcr->JobStatus;
278 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
279 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
282 jcr->JobId = jcr->jr.JobId;
283 ASSERT(jcr->jr.JobId > 0);
285 Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
286 jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
287 Dmsg0(200, "Add jrc to work queue\n");
289 /* Queue the job to be run */
290 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
291 Emsg1(M_ABORT, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
293 Dmsg0(100, "Done run_job()\n");
299 set_jcr_job_status(jcr, JS_ErrorTerminated);
306 * Cancel a job -- typically called by the UA (Console program), but may also
307 * be called by the job watchdog.
309 * Returns: 1 if cancel appears to be successful
310 * 0 on failure. Message sent to ua->jcr.
312 int cancel_job(UAContext *ua, JCR *jcr)
316 switch (jcr->JobStatus) {
319 case JS_WaitClientRes:
320 case JS_WaitStoreRes:
321 case JS_WaitPriority:
323 case JS_WaitStartTime:
324 set_jcr_job_status(jcr, JS_Canceled);
325 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
326 jcr->JobId, jcr->Job);
327 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
331 set_jcr_job_status(jcr, JS_Canceled);
333 /* Cancel File daemon */
334 if (jcr->file_bsock) {
335 ua->jcr->client = jcr->client;
336 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
337 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
340 Dmsg0(200, "Connected to file daemon\n");
341 fd = ua->jcr->file_bsock;
342 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
343 while (bnet_recv(fd) >= 0) {
344 bsendmsg(ua, "%s", fd->msg);
346 bnet_sig(fd, BNET_TERMINATE);
348 ua->jcr->file_bsock = NULL;
351 /* Cancel Storage daemon */
352 if (jcr->store_bsock) {
353 ua->jcr->store = jcr->store;
354 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
355 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
358 Dmsg0(200, "Connected to storage daemon\n");
359 sd = ua->jcr->store_bsock;
360 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
361 while (bnet_recv(sd) >= 0) {
362 bsendmsg(ua, "%s", sd->msg);
364 bnet_sig(sd, BNET_TERMINATE);
366 ua->jcr->store_bsock = NULL;
374 * This is the engine called by jobq.c:jobq_add() when we were pulled
375 * from the work queue.
376 * At this point, we are running in our own thread and all
377 * necessary resources are allocated -- see jobq.c
379 static void *job_thread(void *arg)
381 JCR *jcr = (JCR *)arg;
383 pthread_detach(pthread_self());
384 sm_check(__FILE__, __LINE__, True);
388 Dmsg0(200, "=====Start Job=========\n");
389 jcr->start_time = time(NULL); /* set the real start time */
390 set_jcr_job_status(jcr, JS_Running);
392 if (job_canceled(jcr)) {
393 update_job_end_record(jcr);
394 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
395 (utime_t)(jcr->start_time - jcr->sched_time)) {
396 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
397 set_jcr_job_status(jcr, JS_Canceled);
398 update_job_end_record(jcr);
402 if (jcr->job->RunBeforeJob) {
403 POOLMEM *before = get_pool_memory(PM_FNAME);
406 char line[MAXSTRING];
408 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
409 bpipe = open_bpipe(before, 0, "r");
410 free_pool_memory(before);
411 while (fgets(line, sizeof(line), bpipe->rfd)) {
412 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
414 status = close_bpipe(bpipe);
416 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
418 set_jcr_job_status(jcr, JS_FatalError);
419 update_job_end_record(jcr);
423 switch (jcr->JobType) {
426 if (jcr->JobStatus == JS_Terminated) {
432 if (jcr->JobStatus == JS_Terminated) {
438 if (jcr->JobStatus == JS_Terminated) {
444 if (jcr->JobStatus == JS_Terminated) {
449 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
452 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
453 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
454 POOLMEM *after = get_pool_memory(PM_FNAME);
457 char line[MAXSTRING];
459 if (jcr->JobStatus == JS_Terminated) {
460 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
462 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
464 bpipe = open_bpipe(after, 0, "r");
465 free_pool_memory(after);
466 while (fgets(line, sizeof(line), bpipe->rfd)) {
467 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
469 status = close_bpipe(bpipe);
471 if (jcr->JobStatus == JS_Terminated) {
472 Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
475 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
478 set_jcr_job_status(jcr, JS_FatalError);
479 update_job_end_record(jcr);
487 Dmsg0(50, "======== End Job ==========\n");
488 sm_check(__FILE__, __LINE__, True);
494 * Get or create a Client record for this Job
496 int get_or_create_client_record(JCR *jcr)
500 memset(&cr, 0, sizeof(cr));
501 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
502 cr.AutoPrune = jcr->client->AutoPrune;
503 cr.FileRetention = jcr->client->FileRetention;
504 cr.JobRetention = jcr->client->JobRetention;
505 if (!jcr->client_name) {
506 jcr->client_name = get_pool_memory(PM_NAME);
508 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
509 if (!db_create_client_record(jcr, jcr->db, &cr)) {
510 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
511 db_strerror(jcr->db));
514 jcr->jr.ClientId = cr.ClientId;
516 if (!jcr->client_uname) {
517 jcr->client_uname = get_pool_memory(PM_NAME);
519 pm_strcpy(&jcr->client_uname, cr.Uname);
521 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
528 * Write status and such in DB
530 void update_job_end_record(JCR *jcr)
532 if (jcr->jr.EndTime == 0) {
533 jcr->jr.EndTime = time(NULL);
535 jcr->end_time = jcr->jr.EndTime;
536 jcr->jr.JobId = jcr->JobId;
537 jcr->jr.JobStatus = jcr->JobStatus;
538 jcr->jr.JobFiles = jcr->JobFiles;
539 jcr->jr.JobBytes = jcr->JobBytes;
540 jcr->jr.VolSessionId = jcr->VolSessionId;
541 jcr->jr.VolSessionTime = jcr->VolSessionTime;
542 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
543 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
544 db_strerror(jcr->db));
549 * Takes base_name and appends (unique) current
550 * date and time to form unique job name.
552 * Returns: unique job name in jcr->Job
553 * date/time in jcr->start_time
555 void create_unique_job_name(JCR *jcr, char *base_name)
557 /* Job start mutex */
558 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
559 static time_t last_start_time = 0;
562 char dt[MAX_TIME_LENGTH];
563 char name[MAX_NAME_LENGTH];
566 /* Guarantee unique start time -- maximum one per second, and
567 * thus unique Job Name
569 P(mutex); /* lock creation of jobs */
571 while (now == last_start_time) {
572 bmicrosleep(0, 500000);
575 last_start_time = now;
576 V(mutex); /* allow creation of jobs */
577 jcr->start_time = now;
578 /* Form Unique JobName */
579 localtime_r(&now, &tm);
580 /* Use only characters that are permitted in Windows filenames */
581 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
582 bstrncpy(name, base_name, sizeof(name));
583 name[sizeof(name)-22] = 0; /* truncate if too long */
584 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
585 /* Convert spaces into underscores */
586 for (p=jcr->Job; *p; p++) {
594 * Free the Job Control Record if no one is still using it.
595 * Called from main free_jcr() routine in src/lib/jcr.c so
596 * that we can do our Director specific cleanup of the jcr.
598 void dird_free_jcr(JCR *jcr)
600 Dmsg0(200, "Start dird free_jcr\n");
602 if (jcr->sd_auth_key) {
603 free(jcr->sd_auth_key);
604 jcr->sd_auth_key = NULL;
610 if (jcr->file_bsock) {
611 Dmsg0(200, "Close File bsock\n");
612 bnet_close(jcr->file_bsock);
613 jcr->file_bsock = NULL;
615 if (jcr->store_bsock) {
616 Dmsg0(200, "Close Store bsock\n");
617 bnet_close(jcr->store_bsock);
618 jcr->store_bsock = NULL;
621 Dmsg0(200, "Free JCR fname\n");
622 free_pool_memory(jcr->fname);
626 Dmsg0(200, "Free JCR stime\n");
627 free_pool_memory(jcr->stime);
630 if (jcr->RestoreBootstrap) {
631 free(jcr->RestoreBootstrap);
632 jcr->RestoreBootstrap = NULL;
634 if (jcr->client_uname) {
635 free_pool_memory(jcr->client_uname);
636 jcr->client_uname = NULL;
638 pthread_cond_destroy(&jcr->term_wait);
639 Dmsg0(200, "End dird free_jcr\n");
643 * Set some defaults in the JCR necessary to
644 * run. These items are pulled from the job
645 * definition as defaults, but can be overridden
646 * later either by the Run record in the Schedule resource,
647 * or by the Console program.
649 void set_jcr_defaults(JCR *jcr, JOB *job)
652 jcr->JobType = job->JobType;
653 switch (jcr->JobType) {
656 jcr->JobLevel = L_NONE;
659 jcr->JobLevel = job->level;
662 jcr->JobPriority = job->Priority;
663 jcr->store = job->storage;
664 jcr->client = job->client;
665 if (!jcr->client_name) {
666 jcr->client_name = get_pool_memory(PM_NAME);
668 pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
669 jcr->pool = job->pool;
670 jcr->catalog = job->client->catalog;
671 jcr->fileset = job->fileset;
672 jcr->messages = job->messages;
673 if (jcr->RestoreBootstrap) {
674 free(jcr->RestoreBootstrap);
675 jcr->RestoreBootstrap = NULL;
677 /* This can be overridden by Console program */
678 if (job->RestoreBootstrap) {
679 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
681 /* If no default level given, set one */
682 if (jcr->JobLevel == 0) {
683 switch (jcr->JobType) {
685 jcr->JobLevel = L_VERIFY_CATALOG;
688 jcr->JobLevel = L_INCREMENTAL;
692 jcr->JobLevel = L_NONE;