3 * Bacula Director Job processing routines
5 * Kern Sibbald, October MM
10 Copyright (C) 2000-2004 Kern Sibbald
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 bool do_mac(JCR *jcr);
46 extern int do_admin(JCR *jcr);
47 extern int do_restore(JCR *jcr);
48 extern bool do_verify(JCR *jcr);
50 /* Imported variables */
51 extern time_t watchdog_time;
55 void init_job_server(int max_workers)
60 if ((stat = jobq_init(&job_queue, max_workers, job_thread)) != 0) {
62 Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror(stat));
64 if ((wd = new_watchdog()) == NULL) {
65 Emsg0(M_ABORT, 0, _("Could not init job monitor watchdogs\n"));
67 wd->callback = job_monitor_watchdog;
68 wd->destructor = job_monitor_destructor;
71 wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
72 register_watchdog(wd);
75 void term_job_server()
77 jobq_destroy(&job_queue); /* ignore any errors */
81 * Run a job -- typically called by the scheduler, but may also
82 * be called by the UA (Console program).
84 * Returns: 0 on failure
88 JobId_t run_job(JCR *jcr)
94 sm_check(__FILE__, __LINE__, true);
95 init_msg(jcr, jcr->messages);
97 /* Initialize termination condition variable */
98 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"), be.strerror(errstat));
103 jcr->term_wait_inited = true;
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 jcr->catalog->mult_db_connections);
113 if (!jcr->db || !db_open_database(jcr, jcr->db)) {
114 Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
115 jcr->catalog->db_name);
117 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
121 Dmsg0(50, "DB opened\n");
126 create_unique_job_name(jcr, jcr->job->hdr.name);
127 set_jcr_job_status(jcr, JS_Created);
128 init_jcr_job_record(jcr);
129 if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
130 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
133 JobId = jcr->JobId = jcr->jr.JobId;
135 Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
136 jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
137 Dmsg0(200, "Add jrc to work queue\n");
139 /* Queue the job to be run */
140 if ((stat = jobq_add(&job_queue, jcr)) != 0) {
142 Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror(stat));
146 Dmsg0(100, "Done run_job()\n");
152 set_jcr_job_status(jcr, JS_ErrorTerminated);
160 * This is the engine called by jobq.c:jobq_add() when we were pulled
161 * from the work queue.
162 * At this point, we are running in our own thread and all
163 * necessary resources are allocated -- see jobq.c
165 static void *job_thread(void *arg)
167 JCR *jcr = (JCR *)arg;
169 jcr->my_thread_id = pthread_self();
170 pthread_detach(jcr->my_thread_id);
171 sm_check(__FILE__, __LINE__, true);
174 Dmsg0(200, "=====Start Job=========\n");
175 jcr->start_time = time(NULL); /* set the real start time */
176 jcr->jr.StartTime = jcr->start_time;
177 set_jcr_job_status(jcr, JS_Running);
178 if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
179 Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
182 if (job_canceled(jcr)) {
183 update_job_end_record(jcr);
184 } else if (jcr->job->MaxStartDelay != 0 && jcr->job->MaxStartDelay <
185 (utime_t)(jcr->start_time - jcr->sched_time)) {
186 Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
187 set_jcr_job_status(jcr, JS_Canceled);
188 update_job_end_record(jcr);
192 generate_event(jcr, "StartJob");
193 if (jcr->job->RunBeforeJob) {
194 POOLMEM *before = get_pool_memory(PM_FNAME);
197 char line[MAXSTRING];
199 before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
200 bpipe = open_bpipe(before, 0, "r");
201 free_pool_memory(before);
202 while (fgets(line, sizeof(line), bpipe->rfd)) {
203 Jmsg(jcr, M_INFO, 0, _("RunBefore: %s"), line);
205 status = close_bpipe(bpipe);
208 Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror(status));
209 set_jcr_job_status(jcr, JS_FatalError);
210 update_job_end_record(jcr);
214 switch (jcr->JobType) {
217 if (jcr->JobStatus == JS_Terminated) {
223 if (jcr->JobStatus == JS_Terminated) {
229 if (jcr->JobStatus == JS_Terminated) {
235 if (jcr->JobStatus == JS_Terminated) {
242 do_mac(jcr); /* migration, archive, copy */
243 if (jcr->JobStatus == JS_Terminated) {
248 Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
251 if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
252 (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
253 POOLMEM *after = get_pool_memory(PM_FNAME);
256 char line[MAXSTRING];
258 if (jcr->JobStatus == JS_Terminated) {
259 after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
261 after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
263 bpipe = open_bpipe(after, 0, "r");
264 free_pool_memory(after);
265 while (fgets(line, sizeof(line), bpipe->rfd)) {
266 Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
268 status = close_bpipe(bpipe);
270 * Note, if we get an error here, do not mark the
271 * job in error, simply report the error condition.
275 if (jcr->JobStatus == JS_Terminated) {
276 Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror(status));
278 Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror(status));
282 generate_event(jcr, "EndJob");
283 /* Send off any queued messages */
284 if (jcr->msg_queue->size() > 0) {
285 dequeue_messages(jcr);
292 Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
293 sm_check(__FILE__, __LINE__, true);
299 * Cancel a job -- typically called by the UA (Console program), but may also
300 * be called by the job watchdog.
302 * Returns: 1 if cancel appears to be successful
303 * 0 on failure. Message sent to ua->jcr.
305 int cancel_job(UAContext *ua, JCR *jcr)
309 switch (jcr->JobStatus) {
312 case JS_WaitClientRes:
313 case JS_WaitStoreRes:
314 case JS_WaitPriority:
316 case JS_WaitStartTime:
317 set_jcr_job_status(jcr, JS_Canceled);
318 bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
319 jcr->JobId, jcr->Job);
320 jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
324 set_jcr_job_status(jcr, JS_Canceled);
326 /* Cancel File daemon */
327 if (jcr->file_bsock) {
328 ua->jcr->client = jcr->client;
329 if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
330 bsendmsg(ua, _("Failed to connect to File daemon.\n"));
333 Dmsg0(200, "Connected to file daemon\n");
334 fd = ua->jcr->file_bsock;
335 bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
336 while (bnet_recv(fd) >= 0) {
337 bsendmsg(ua, "%s", fd->msg);
339 bnet_sig(fd, BNET_TERMINATE);
341 ua->jcr->file_bsock = NULL;
344 /* Cancel Storage daemon */
345 if (jcr->store_bsock) {
346 if (!ua->jcr->storage[0]) {
347 copy_storage(ua->jcr, jcr);
349 ua->jcr->store = jcr->store;
351 if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
352 bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
355 Dmsg0(200, "Connected to storage daemon\n");
356 sd = ua->jcr->store_bsock;
357 bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
358 while (bnet_recv(sd) >= 0) {
359 bsendmsg(ua, "%s", sd->msg);
361 bnet_sig(sd, BNET_TERMINATE);
363 ua->jcr->store_bsock = NULL;
371 static void job_monitor_destructor(watchdog_t *self)
373 JCR *control_jcr = (JCR *) self->data;
375 free_jcr(control_jcr);
378 static void job_monitor_watchdog(watchdog_t *self)
380 JCR *control_jcr, *jcr;
382 control_jcr = (JCR *)self->data;
384 Dmsg1(400, "job_monitor_watchdog %p called\n", self);
391 if (jcr->JobId == 0) {
392 Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
394 /* Keep reference counts correct */
395 free_locked_jcr(jcr);
399 /* check MaxWaitTime */
400 cancel = job_check_maxwaittime(control_jcr, jcr);
402 /* check MaxRunTime */
403 cancel |= job_check_maxruntime(control_jcr, jcr);
406 Dmsg3(200, "Cancelling JCR %p jobid %d (%s)\n",
407 jcr, jcr->JobId, jcr->Job);
409 UAContext *ua = new_ua_context(jcr);
410 ua->jcr = control_jcr;
414 Dmsg1(200, "Have cancelled JCR %p\n", jcr);
417 /* Keep reference counts correct */
418 free_locked_jcr(jcr);
424 * Check if the maxwaittime has expired and it is possible
427 static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
431 if (jcr->job->MaxWaitTime == 0) {
434 if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
435 Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
436 jcr, jcr->Job, jcr->job->MaxWaitTime);
439 Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
441 jcr->JobId, jcr->Job, jcr->job->MaxWaitTime);
442 switch (jcr->JobStatus) {
447 case JS_WaitStoreRes:
448 case JS_WaitClientRes:
450 case JS_WaitPriority:
452 case JS_WaitStartTime:
454 Dmsg0(200, "JCR blocked in #1\n");
457 Dmsg0(200, "JCR running, checking SD status\n");
458 switch (jcr->SDJobStatus) {
463 Dmsg0(200, "JCR blocked in #2\n");
466 Dmsg0(200, "JCR not blocked in #2\n");
471 case JS_ErrorTerminated:
474 Dmsg0(200, "JCR already dead in #3\n");
477 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
480 Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
481 cancel ? "" : "do not ", jcr, jcr->job);
487 * Check if maxruntime has expired and if the job can be
490 static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
494 if (jcr->job->MaxRunTime == 0) {
497 if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
498 Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
499 jcr, jcr->Job, jcr->job->MaxRunTime);
503 switch (jcr->JobStatus) {
509 case JS_WaitStoreRes:
510 case JS_WaitClientRes:
512 case JS_WaitPriority:
514 case JS_WaitStartTime:
519 case JS_ErrorTerminated:
525 Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
529 Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
530 cancel ? "" : "do not ", jcr, jcr->job);
537 * Get or create a Client record for this Job
539 bool get_or_create_client_record(JCR *jcr)
543 memset(&cr, 0, sizeof(cr));
544 bstrncpy(cr.Name, jcr->client->hdr.name, sizeof(cr.Name));
545 cr.AutoPrune = jcr->client->AutoPrune;
546 cr.FileRetention = jcr->client->FileRetention;
547 cr.JobRetention = jcr->client->JobRetention;
548 if (!jcr->client_name) {
549 jcr->client_name = get_pool_memory(PM_NAME);
551 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
552 if (!db_create_client_record(jcr, jcr->db, &cr)) {
553 Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
554 db_strerror(jcr->db));
557 jcr->jr.ClientId = cr.ClientId;
559 if (!jcr->client_uname) {
560 jcr->client_uname = get_pool_memory(PM_NAME);
562 pm_strcpy(jcr->client_uname, cr.Uname);
564 Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
569 bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
572 * Get or Create FileSet record
574 memset(fsr, 0, sizeof(FILESET_DBR));
575 bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
576 if (jcr->fileset->have_MD5) {
577 struct MD5Context md5c;
578 unsigned char signature[16];
579 memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
580 MD5Final(signature, &md5c);
581 bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
582 bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
584 Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
586 if (!jcr->fileset->ignore_fs_changes ||
587 !db_get_fileset_record(jcr, jcr->db, fsr)) {
588 if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
589 Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
590 fsr->FileSet, db_strerror(jcr->db));
594 jcr->jr.FileSetId = fsr->FileSetId;
596 Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
597 fsr->FileSet, fsr->cCreateTime);
599 Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
604 void init_jcr_job_record(JCR *jcr)
606 jcr->jr.SchedTime = jcr->sched_time;
607 jcr->jr.StartTime = jcr->start_time;
608 jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
609 jcr->jr.JobType = jcr->JobType;
610 jcr->jr.JobLevel = jcr->JobLevel;
611 jcr->jr.JobStatus = jcr->JobStatus;
612 jcr->jr.JobId = jcr->JobId;
613 bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
614 bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
618 * Write status and such in DB
620 void update_job_end_record(JCR *jcr)
622 jcr->jr.EndTime = time(NULL);
623 jcr->end_time = jcr->jr.EndTime;
624 jcr->jr.JobId = jcr->JobId;
625 jcr->jr.JobStatus = jcr->JobStatus;
626 jcr->jr.JobFiles = jcr->JobFiles;
627 jcr->jr.JobBytes = jcr->JobBytes;
628 jcr->jr.VolSessionId = jcr->VolSessionId;
629 jcr->jr.VolSessionTime = jcr->VolSessionTime;
630 if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
631 Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
632 db_strerror(jcr->db));
637 * Takes base_name and appends (unique) current
638 * date and time to form unique job name.
640 * Returns: unique job name in jcr->Job
641 * date/time in jcr->start_time
643 void create_unique_job_name(JCR *jcr, const char *base_name)
645 /* Job start mutex */
646 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
647 static time_t last_start_time = 0;
650 char dt[MAX_TIME_LENGTH];
651 char name[MAX_NAME_LENGTH];
654 /* Guarantee unique start time -- maximum one per second, and
655 * thus unique Job Name
657 P(mutex); /* lock creation of jobs */
659 while (now == last_start_time) {
660 bmicrosleep(0, 500000);
663 last_start_time = now;
664 V(mutex); /* allow creation of jobs */
665 jcr->start_time = now;
666 /* Form Unique JobName */
667 localtime_r(&now, &tm);
668 /* Use only characters that are permitted in Windows filenames */
669 strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
670 bstrncpy(name, base_name, sizeof(name));
671 name[sizeof(name)-22] = 0; /* truncate if too long */
672 bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
673 /* Convert spaces into underscores */
674 for (p=jcr->Job; *p; p++) {
682 * Free the Job Control Record if no one is still using it.
683 * Called from main free_jcr() routine in src/lib/jcr.c so
684 * that we can do our Director specific cleanup of the jcr.
686 void dird_free_jcr(JCR *jcr)
688 Dmsg0(200, "Start dird free_jcr\n");
690 if (jcr->sd_auth_key) {
691 free(jcr->sd_auth_key);
692 jcr->sd_auth_key = NULL;
698 if (jcr->file_bsock) {
699 Dmsg0(200, "Close File bsock\n");
700 bnet_close(jcr->file_bsock);
701 jcr->file_bsock = NULL;
703 if (jcr->store_bsock) {
704 Dmsg0(200, "Close Store bsock\n");
705 bnet_close(jcr->store_bsock);
706 jcr->store_bsock = NULL;
709 Dmsg0(200, "Free JCR fname\n");
710 free_pool_memory(jcr->fname);
714 Dmsg0(200, "Free JCR stime\n");
715 free_pool_memory(jcr->stime);
718 if (jcr->RestoreBootstrap) {
719 free(jcr->RestoreBootstrap);
720 jcr->RestoreBootstrap = NULL;
722 if (jcr->client_uname) {
723 free_pool_memory(jcr->client_uname);
724 jcr->client_uname = NULL;
726 if (jcr->term_wait_inited) {
727 pthread_cond_destroy(&jcr->term_wait);
729 jcr->job_end_push.destroy();
730 Dmsg0(200, "End dird free_jcr\n");
734 * Set some defaults in the JCR necessary to
735 * run. These items are pulled from the job
736 * definition as defaults, but can be overridden
737 * later either by the Run record in the Schedule resource,
738 * or by the Console program.
740 void set_jcr_defaults(JCR *jcr, JOB *job)
743 jcr->JobType = job->JobType;
744 switch (jcr->JobType) {
747 jcr->JobLevel = L_NONE;
750 jcr->JobLevel = job->JobLevel;
753 jcr->JobPriority = job->Priority;
754 for (int i=0; i<MAX_STORE; i++) {
755 jcr->storage[i] = job->storage[i];
757 if (!jcr->store && jcr->storage[0]) {
758 jcr->store = (STORE *)jcr->storage[0]->first();
760 jcr->client = job->client;
761 if (!jcr->client_name) {
762 jcr->client_name = get_pool_memory(PM_NAME);
764 pm_strcpy(jcr->client_name, jcr->client->hdr.name);
765 jcr->pool = job->pool;
766 jcr->full_pool = job->full_pool;
767 jcr->inc_pool = job->inc_pool;
768 jcr->dif_pool = job->dif_pool;
769 jcr->catalog = job->client->catalog;
770 jcr->fileset = job->fileset;
771 jcr->messages = job->messages;
772 jcr->spool_data = job->spool_data;
773 if (jcr->RestoreBootstrap) {
774 free(jcr->RestoreBootstrap);
775 jcr->RestoreBootstrap = NULL;
777 /* This can be overridden by Console program */
778 if (job->RestoreBootstrap) {
779 jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
781 /* This can be overridden by Console program */
782 jcr->verify_job = job->verify_job;
783 /* If no default level given, set one */
784 if (jcr->JobLevel == 0) {
785 switch (jcr->JobType) {
787 jcr->JobLevel = L_VERIFY_CATALOG;
790 jcr->JobLevel = L_INCREMENTAL;
794 jcr->JobLevel = L_NONE;
803 * copy the storage definitions from an old JCR to a new one
805 void copy_storage(JCR *new_jcr, JCR *old_jcr)
807 for (int i=0; i < MAX_STORE; i++) {
808 if (old_jcr->storage[i]) {
810 new_jcr->storage[i] = New(alist(10, not_owned_by_alist));
811 foreach_alist(st, old_jcr->storage[i]) {
812 new_jcr->storage[i]->append(st);
815 if (old_jcr->store) {
816 new_jcr->store = old_jcr->store;
817 } else if (new_jcr->storage[0]) {
818 new_jcr->store = (STORE *)new_jcr->storage[0]->first();