* Version $Id$
*/
/*
- Copyright (C) 2000-2003 Kern Sibbald and John Walker
+ Copyright (C) 2000-2005 Kern Sibbald
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
/* Forward referenced subroutines */
static void *job_thread(void *arg);
-static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg);
-static void release_resource_locks(JCR *jcr);
-static int acquire_resource_locks(JCR *jcr);
-#ifdef USE_SEMAPHORE
-static void backoff_resource_locks(JCR *jcr, int count);
-#endif
-
-/* Exported subroutines */
-void run_job(JCR *jcr);
-
+static void job_monitor_watchdog(watchdog_t *self);
+static void job_monitor_destructor(watchdog_t *self);
+static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr);
+static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr);
/* Imported subroutines */
extern void term_scheduler();
extern void term_ua_server();
-extern int do_backup(JCR *jcr);
-extern int do_admin(JCR *jcr);
-extern int do_restore(JCR *jcr);
-extern int do_verify(JCR *jcr);
-
-#ifdef USE_SEMAPHORE
-static semlock_t job_lock;
-static pthread_mutex_t mutex;
-static pthread_cond_t resource_wait;
-static int waiting = 0; /* count of waiting threads */
-#else
-#ifdef JOB_QUEUE
-jobq_t job_queue;
-#else
-/* Queue of jobs to be run */
-workq_t job_wq; /* our job work queue */
-#endif
-#endif
+
+/* Imported variables */
+extern time_t watchdog_time;
+
+jobq_t job_queue;
void init_job_server(int max_workers)
{
int stat;
-#ifdef USE_SEMAPHORE
- if ((stat = sem_init(&job_lock, max_workers)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init job lock: ERR=%s\n"), strerror(stat));
- }
- if ((stat = pthread_mutex_init(&mutex, NULL)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init resource mutex: ERR=%s\n"), strerror(stat));
- }
- if ((stat = pthread_cond_init(&resource_wait, NULL)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init resource wait: ERR=%s\n"), strerror(stat));
- }
+ watchdog_t *wd;
+
+ if ((stat = jobq_init(&job_queue, max_workers, job_thread)) != 0) {
+ berrno be;
+ Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror(stat));
+ }
+ wd = new_watchdog();
+ wd->callback = job_monitor_watchdog;
+ wd->destructor = job_monitor_destructor;
+ wd->one_shot = false;
+ wd->interval = 60;
+ wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
+ register_watchdog(wd);
+}
-#else
-#ifdef JOB_QUEUE
- if ((stat = job_init(&job_queue, max_workers, job_thread)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), strerror(stat));
- }
-#else
- /* This is the OLD work queue code to go away */
- if ((stat = workq_init(&job_wq, max_workers, job_thread)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init job work queue: ERR=%s\n"), strerror(stat));
- }
-#endif
-#endif
- return;
+void term_job_server()
+{
+ jobq_destroy(&job_queue); /* ignore any errors */
}
/*
* Run a job -- typically called by the scheduler, but may also
* be called by the UA (Console program).
*
+ * Returns: 0 on failure
+ * JobId on success
+ *
*/
-void run_job(JCR *jcr)
+JobId_t run_job(JCR *jcr)
{
int stat, errstat;
-#ifdef USE_SEMAPHORE
- pthread_t tid;
-#else
-#ifndef JOB_QUEUE
- workq_ele_t *work_item;
-#endif
-#endif
-
- sm_check(__FILE__, __LINE__, True);
+ JobId_t JobId = 0;
+
+ P(jcr->mutex);
+ sm_check(__FILE__, __LINE__, true);
init_msg(jcr, jcr->messages);
- create_unique_job_name(jcr, jcr->job->hdr.name);
- set_jcr_job_status(jcr, JS_Created);
- jcr->jr.SchedTime = jcr->sched_time;
- jcr->jr.StartTime = jcr->start_time;
- jcr->jr.Type = jcr->JobType;
- jcr->jr.Level = jcr->JobLevel;
- jcr->jr.JobStatus = jcr->JobStatus;
- bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
- bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
/* Initialize termination condition variable */
if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
- Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), strerror(errstat));
- set_jcr_job_status(jcr, JS_ErrorTerminated);
- free_jcr(jcr);
- return;
+ berrno be;
+ Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.strerror(errstat));
+ goto bail_out;
}
+ jcr->term_wait_inited = true;
/*
* Open database
Dmsg0(50, "Open database\n");
jcr->db=db_init_database(jcr, jcr->catalog->db_name, jcr->catalog->db_user,
jcr->catalog->db_password, jcr->catalog->db_address,
- jcr->catalog->db_port, jcr->catalog->db_socket);
- if (!db_open_database(jcr, jcr->db)) {
- Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
- set_jcr_job_status(jcr, JS_ErrorTerminated);
- free_jcr(jcr);
- return;
+ jcr->catalog->db_port, jcr->catalog->db_socket,
+ jcr->catalog->mult_db_connections);
+ if (!jcr->db || !db_open_database(jcr, jcr->db)) {
+ Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
+ jcr->catalog->db_name);
+ if (jcr->db) {
+ Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
+ }
+ goto bail_out;
}
Dmsg0(50, "DB opened\n");
/*
- * Create Job record
+ * Create Job record
*/
- jcr->jr.JobStatus = jcr->JobStatus;
+ create_unique_job_name(jcr, jcr->job->hdr.name);
+ set_jcr_job_status(jcr, JS_Created);
+ init_jcr_job_record(jcr);
if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
+ goto bail_out;
+ }
+ JobId = jcr->JobId = jcr->jr.JobId;
+
+ Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
+ jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
+
+ if (!get_or_create_client_record(jcr)) {
+ goto bail_out;
+ }
+
+ if (!jcr->fname) {
+ jcr->fname = get_pool_memory(PM_FNAME);
+ }
+
+ /* Now, do pre-run stuff, like setting job level (Inc/diff, ...) */
+ switch (jcr->JobType) {
+ case JT_BACKUP:
+ if (!do_backup_init(jcr)) {
+ backup_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ case JT_VERIFY:
+ if (!do_verify_init(jcr)) {
+ verify_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ case JT_RESTORE:
+ if (!do_restore_init(jcr)) {
+ restore_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ case JT_ADMIN:
+ if (!do_admin_init(jcr)) {
+ admin_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ case JT_MIGRATION:
+ case JT_COPY:
+ case JT_ARCHIVE:
+ if (!do_mac_init(jcr)) { /* migration, archive, copy */
+ mac_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ default:
+ Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
set_jcr_job_status(jcr, JS_ErrorTerminated);
- free_jcr(jcr);
- return;
+ break;
+ }
+ if (job_canceled(jcr)) {
+ goto bail_out;
}
- jcr->JobId = jcr->jr.JobId;
- ASSERT(jcr->jr.JobId > 0);
- Dmsg4(30, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
- jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
Dmsg0(200, "Add jrc to work queue\n");
-#ifdef USE_SEMAPHORE
- if ((stat = pthread_create(&tid, NULL, job_thread, (void *)jcr)) != 0) {
- Emsg1(M_ABORT, 0, _("Unable to create job thread: ERR=%s\n"), strerror(stat));
- }
-#else
-#ifdef JOB_QUEUE
/* Queue the job to be run */
if ((stat = jobq_add(&job_queue, jcr)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
+ berrno be;
+ Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror(stat));
+ JobId = 0;
+ goto bail_out;
}
-#else
- /* Queue the job to be run */
- if ((stat = workq_add(&job_wq, (void *)jcr, &work_item, 0)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not add job to work queue: ERR=%s\n"), strerror(stat));
+ Dmsg0(100, "Done run_job()\n");
+
+ V(jcr->mutex);
+ return JobId;
+
+bail_out:
+ if (jcr->fname) {
+ free_memory(jcr->fname);
+ jcr->fname = NULL;
}
- jcr->work_item = work_item;
-#endif
-#endif
- Dmsg0(200, "Done run_job()\n");
+ V(jcr->mutex);
+ return JobId;
+
}
-/*
- * This is the engine called by workq_add() when we were pulled
+
+/*
+ * This is the engine called by jobq.c:jobq_add() when we were pulled
* from the work queue.
- * At this point, we are running in our own thread
+ * At this point, we are running in our own thread and all
+ * necessary resources are allocated -- see jobq.c
*/
static void *job_thread(void *arg)
{
JCR *jcr = (JCR *)arg;
- pthread_detach(pthread_self());
- sm_check(__FILE__, __LINE__, True);
+ jcr->my_thread_id = pthread_self();
+ pthread_detach(jcr->my_thread_id);
+ sm_check(__FILE__, __LINE__, true);
for ( ;; ) {
- if (!acquire_resource_locks(jcr)) {
- set_jcr_job_status(jcr, JS_Canceled);
- }
-
Dmsg0(200, "=====Start Job=========\n");
jcr->start_time = time(NULL); /* set the real start time */
+ jcr->jr.StartTime = jcr->start_time;
set_jcr_job_status(jcr, JS_Running);
+ if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
+ Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
+ }
if (job_canceled(jcr)) {
update_job_end_record(jcr);
} else {
/* Run Job */
+ generate_event(jcr, "StartJob");
if (jcr->job->RunBeforeJob) {
POOLMEM *before = get_pool_memory(PM_FNAME);
int status;
BPIPE *bpipe;
char line[MAXSTRING];
-
- before = edit_run_codes(jcr, before, jcr->job->RunBeforeJob);
+
+ before = edit_job_codes(jcr, before, jcr->job->RunBeforeJob, "");
bpipe = open_bpipe(before, 0, "r");
free_pool_memory(before);
while (fgets(line, sizeof(line), bpipe->rfd)) {
}
status = close_bpipe(bpipe);
if (status != 0) {
- Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
- status);
+ berrno be;
+ Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror(status));
set_jcr_job_status(jcr, JS_FatalError);
update_job_end_record(jcr);
goto bail_out;
}
switch (jcr->JobType) {
case JT_BACKUP:
- do_backup(jcr);
- if (jcr->JobStatus == JS_Terminated) {
+ if (do_backup(jcr)) {
do_autoprune(jcr);
+ } else {
+ backup_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_VERIFY:
- do_verify(jcr);
- if (jcr->JobStatus == JS_Terminated) {
+ if (do_verify(jcr)) {
do_autoprune(jcr);
+ } else {
+ verify_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_RESTORE:
- do_restore(jcr);
- if (jcr->JobStatus == JS_Terminated) {
+ if (do_restore(jcr)) {
do_autoprune(jcr);
+ } else {
+ restore_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_ADMIN:
- do_admin(jcr);
- if (jcr->JobStatus == JS_Terminated) {
+ if (do_admin(jcr)) {
do_autoprune(jcr);
+ } else {
+ admin_cleanup(jcr, JS_ErrorTerminated);
+ }
+ break;
+ case JT_MIGRATION:
+ case JT_COPY:
+ case JT_ARCHIVE:
+ if (do_mac(jcr)) { /* migration, archive, copy */
+ do_autoprune(jcr);
+ } else {
+ mac_cleanup(jcr, JS_ErrorTerminated);
}
break;
default:
Pmsg1(0, "Unimplemented job type: %d\n", jcr->JobType);
break;
}
- if (jcr->job->RunAfterJob) {
+ if ((jcr->job->RunAfterJob && jcr->JobStatus == JS_Terminated) ||
+ (jcr->job->RunAfterFailedJob && jcr->JobStatus != JS_Terminated)) {
POOLMEM *after = get_pool_memory(PM_FNAME);
int status;
BPIPE *bpipe;
char line[MAXSTRING];
-
- after = edit_run_codes(jcr, after, jcr->job->RunAfterJob);
+
+ if (jcr->JobStatus == JS_Terminated) {
+ after = edit_job_codes(jcr, after, jcr->job->RunAfterJob, "");
+ } else {
+ after = edit_job_codes(jcr, after, jcr->job->RunAfterFailedJob, "");
+ }
bpipe = open_bpipe(after, 0, "r");
free_pool_memory(after);
while (fgets(line, sizeof(line), bpipe->rfd)) {
Jmsg(jcr, M_INFO, 0, _("RunAfter: %s"), line);
}
status = close_bpipe(bpipe);
+ /*
+ * Note, if we get an error here, do not mark the
+ * job in error, simply report the error condition.
+ */
if (status != 0) {
- Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
- status);
- set_jcr_job_status(jcr, JS_FatalError);
- update_job_end_record(jcr);
+ berrno be;
+ if (jcr->JobStatus == JS_Terminated) {
+ Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror(status));
+ } else {
+ Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror(status));
+ }
}
}
- }
-bail_out:
- release_resource_locks(jcr);
- if (jcr->job->RescheduleOnError &&
- jcr->JobStatus != JS_Terminated &&
- jcr->JobStatus != JS_Canceled &&
- jcr->job->RescheduleTimes > 0 &&
- jcr->reschedule_count < jcr->job->RescheduleTimes) {
-
- /*
- * Reschedule this job by cleaning it up, but
- * reuse the same JobId if possible.
- */
- jcr->reschedule_count++;
- jcr->sched_time = time(NULL) + jcr->job->RescheduleInterval;
- Dmsg2(100, "Rescheduled Job %s to re-run in %d seconds.\n", jcr->Job,
- (int)jcr->job->RescheduleInterval);
- jcr->JobStatus = JS_Created; /* force new status */
- dird_free_jcr(jcr); /* partial cleanup old stuff */
- if (jcr->JobBytes == 0) {
- continue; /* reschedule the job */
+ generate_event(jcr, "EndJob");
+ /* Send off any queued messages */
+ if (jcr->msg_queue->size() > 0) {
+ dequeue_messages(jcr);
}
- /*
- * Something was actually backed up, so we cannot reuse
- * the old JobId or there will be database record
- * conflicts. We now create a new job, copying the
- * appropriate fields.
- */
- JCR *njcr = new_jcr(sizeof(JCR), dird_free_jcr);
- set_jcr_defaults(njcr, jcr->job);
- njcr->reschedule_count = jcr->reschedule_count;
- njcr->JobLevel = jcr->JobLevel;
- njcr->JobStatus = jcr->JobStatus;
- njcr->pool = jcr->pool;
- njcr->store = jcr->store;
- njcr->messages = jcr->messages;
- run_job(njcr);
}
+bail_out:
break;
}
- if (jcr->db) {
- Dmsg0(200, "Close DB\n");
- db_close_database(jcr, jcr->db);
- jcr->db = NULL;
- }
- free_jcr(jcr);
- Dmsg0(50, "======== End Job ==========\n");
- sm_check(__FILE__, __LINE__, True);
+ Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
+ sm_check(__FILE__, __LINE__, true);
return NULL;
}
+
/*
- * Acquire the resources needed. These locks limit the
- * number of jobs by each resource. We have limits on
- * Jobs, Clients, Storage, and total jobs.
+ * Cancel a job -- typically called by the UA (Console program), but may also
+ * be called by the job watchdog.
+ *
+ * Returns: 1 if cancel appears to be successful
+ * 0 on failure. Message sent to ua->jcr.
*/
-static int acquire_resource_locks(JCR *jcr)
+int cancel_job(UAContext *ua, JCR *jcr)
{
- time_t now = time(NULL);
- time_t wtime = jcr->sched_time - now;
-
- /* Wait until scheduled time arrives */
- if (wtime > 0 && verbose) {
- Jmsg(jcr, M_INFO, 0, _("Job %s waiting %d seconds for scheduled start time.\n"),
- jcr->Job, wtime);
- set_jcr_job_status(jcr, JS_WaitStartTime);
- }
- /* Check every 30 seconds if canceled */
- while (wtime > 0) {
- Dmsg2(100, "Waiting on sched time, jobid=%d secs=%d\n", jcr->JobId, wtime);
- if (wtime > 30) {
- wtime = 30;
+ BSOCK *sd, *fd;
+
+ switch (jcr->JobStatus) {
+ case JS_Created:
+ case JS_WaitJobRes:
+ case JS_WaitClientRes:
+ case JS_WaitStoreRes:
+ case JS_WaitPriority:
+ case JS_WaitMaxJobs:
+ case JS_WaitStartTime:
+ set_jcr_job_status(jcr, JS_Canceled);
+ bsendmsg(ua, _("JobId %d, Job %s marked to be canceled.\n"),
+ jcr->JobId, jcr->Job);
+ jobq_remove(&job_queue, jcr); /* attempt to remove it from queue */
+ return 1;
+
+ default:
+ set_jcr_job_status(jcr, JS_Canceled);
+
+ /* Cancel File daemon */
+ if (jcr->file_bsock) {
+ ua->jcr->client = jcr->client;
+ if (!connect_to_file_daemon(ua->jcr, 10, FDConnectTimeout, 1)) {
+ bsendmsg(ua, _("Failed to connect to File daemon.\n"));
+ return 0;
+ }
+ Dmsg0(200, "Connected to file daemon\n");
+ fd = ua->jcr->file_bsock;
+ bnet_fsend(fd, "cancel Job=%s\n", jcr->Job);
+ while (bnet_recv(fd) >= 0) {
+ bsendmsg(ua, "%s", fd->msg);
+ }
+ bnet_sig(fd, BNET_TERMINATE);
+ bnet_close(fd);
+ ua->jcr->file_bsock = NULL;
}
- bmicrosleep(wtime, 0);
- if (job_canceled(jcr)) {
- return 0;
+
+ /* Cancel Storage daemon */
+ if (jcr->store_bsock) {
+ if (!ua->jcr->storage) {
+ copy_storage(ua->jcr, jcr);
+ } else {
+ set_storage(ua->jcr, jcr->store);
+ }
+ if (!connect_to_storage_daemon(ua->jcr, 10, SDConnectTimeout, 1)) {
+ bsendmsg(ua, _("Failed to connect to Storage daemon.\n"));
+ return 0;
+ }
+ Dmsg0(200, "Connected to storage daemon\n");
+ sd = ua->jcr->store_bsock;
+ bnet_fsend(sd, "cancel Job=%s\n", jcr->Job);
+ while (bnet_recv(sd) >= 0) {
+ bsendmsg(ua, "%s", sd->msg);
+ }
+ bnet_sig(sd, BNET_TERMINATE);
+ bnet_close(sd);
+ ua->jcr->store_bsock = NULL;
}
- wtime = jcr->sched_time - time(NULL);
}
+ return 1;
+}
-#ifdef USE_SEMAPHORE
- int stat;
- /* Initialize semaphores */
- if (jcr->store->sem.valid != SEMLOCK_VALID) {
- if ((stat = sem_init(&jcr->store->sem, jcr->store->MaxConcurrentJobs)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init Storage semaphore: ERR=%s\n"), strerror(stat));
- }
- }
- if (jcr->client->sem.valid != SEMLOCK_VALID) {
- if ((stat = sem_init(&jcr->client->sem, jcr->client->MaxConcurrentJobs)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init Client semaphore: ERR=%s\n"), strerror(stat));
- }
- }
- if (jcr->job->sem.valid != SEMLOCK_VALID) {
- if ((stat = sem_init(&jcr->job->sem, jcr->job->MaxConcurrentJobs)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not init Job semaphore: ERR=%s\n"), strerror(stat));
- }
- }
+static void job_monitor_destructor(watchdog_t *self)
+{
+ JCR *control_jcr = (JCR *) self->data;
- for ( ;; ) {
- /* Acquire semaphore */
- set_jcr_job_status(jcr, JS_WaitJobRes);
- if ((stat = sem_lock(&jcr->job->sem)) != 0) {
- Emsg1(M_ABORT, 0, _("Could not acquire Job max jobs lock: ERR=%s\n"), strerror(stat));
- }
- set_jcr_job_status(jcr, JS_WaitClientRes);
- if ((stat = sem_trylock(&jcr->client->sem)) != 0) {
- if (stat == EBUSY) {
- backoff_resource_locks(jcr, 1);
- goto wait;
- } else {
- Emsg1(M_ABORT, 0, _("Could not acquire Client max jobs lock: ERR=%s\n"), strerror(stat));
- }
- }
- set_jcr_job_status(jcr, JS_WaitStoreRes);
- if ((stat = sem_trylock(&jcr->store->sem)) != 0) {
- if (stat == EBUSY) {
- backoff_resource_locks(jcr, 2);
- goto wait;
- } else {
- Emsg1(M_ABORT, 0, _("Could not acquire Storage max jobs lock: ERR=%s\n"), strerror(stat));
- }
- }
- set_jcr_job_status(jcr, JS_WaitMaxJobs);
- if ((stat = sem_trylock(&job_lock)) != 0) {
- if (stat == EBUSY) {
- backoff_resource_locks(jcr, 3);
- goto wait;
- } else {
- Emsg1(M_ABORT, 0, _("Could not acquire max jobs lock: ERR=%s\n"), strerror(stat));
- }
+ free_jcr(control_jcr);
+}
+
+static void job_monitor_watchdog(watchdog_t *self)
+{
+ JCR *control_jcr, *jcr;
+
+ control_jcr = (JCR *)self->data;
+
+ Dmsg1(800, "job_monitor_watchdog %p called\n", self);
+
+ lock_jcr_chain();
+
+ foreach_jcr(jcr) {
+ bool cancel;
+
+ if (jcr->JobId == 0) {
+ Dmsg2(800, "Skipping JCR %p (%s) with JobId 0\n",
+ jcr, jcr->Job);
+ /* Keep reference counts correct */
+ free_locked_jcr(jcr);
+ continue;
}
- break;
-wait:
- if (job_canceled(jcr)) {
- return 0;
+ /* check MaxWaitTime */
+ cancel = job_check_maxwaittime(control_jcr, jcr);
+
+ /* check MaxRunTime */
+ cancel |= job_check_maxruntime(control_jcr, jcr);
+
+ if (cancel) {
+ Dmsg3(800, "Cancelling JCR %p jobid %d (%s)\n",
+ jcr, jcr->JobId, jcr->Job);
+
+ UAContext *ua = new_ua_context(jcr);
+ ua->jcr = control_jcr;
+ cancel_job(ua, jcr);
+ free_ua_context(ua);
+
+ Dmsg1(800, "Have cancelled JCR %p\n", jcr);
}
- P(mutex);
- /*
- * Wait for a resource to be released either by backoff or
- * by a job terminating.
- */
- waiting++;
- pthread_cond_wait(&resource_wait, &mutex);
- waiting--;
- V(mutex);
- /* Try again */
- }
- jcr->acquired_resource_locks = true;
-#endif
- return 1;
+
+ /* Keep reference counts correct */
+ free_locked_jcr(jcr);
+ }
+ unlock_jcr_chain();
}
-#ifdef USE_SEMAPHORE
/*
- * We could not get all the resource locks because
- * too many jobs are running, so release any locks
- * we did acquire, giving others a chance to use them
- * while we wait.
+ * Check if the maxwaittime has expired and it is possible
+ * to cancel the job.
*/
-static void backoff_resource_locks(JCR *jcr, int count)
+static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
{
- P(mutex);
- switch (count) {
- case 3:
- sem_unlock(&jcr->store->sem);
- /* Fall through wanted */
- case 2:
- sem_unlock(&jcr->client->sem);
- /* Fall through wanted */
- case 1:
- sem_unlock(&jcr->job->sem);
+ bool cancel = false;
+ bool ok_to_cancel = false;
+ JOB *job = jcr->job;
+
+ if (job->MaxWaitTime == 0 && job->FullMaxWaitTime == 0 &&
+ job->IncMaxWaitTime == 0 && job->DiffMaxWaitTime == 0) {
+ return false;
+ }
+ if (jcr->JobLevel == L_FULL && job->FullMaxWaitTime != 0 &&
+ (watchdog_time - jcr->start_time) >= job->FullMaxWaitTime) {
+ ok_to_cancel = true;
+ } else if (jcr->JobLevel == L_DIFFERENTIAL && job->DiffMaxWaitTime != 0 &&
+ (watchdog_time - jcr->start_time) >= job->DiffMaxWaitTime) {
+ ok_to_cancel = true;
+ } else if (jcr->JobLevel == L_INCREMENTAL && job->IncMaxWaitTime != 0 &&
+ (watchdog_time - jcr->start_time) >= job->IncMaxWaitTime) {
+ ok_to_cancel = true;
+ } else if (job->MaxWaitTime != 0 &&
+ (watchdog_time - jcr->start_time) >= job->MaxWaitTime) {
+ ok_to_cancel = true;
+ }
+ if (!ok_to_cancel) {
+ return false;
+ }
+ Dmsg3(800, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
+ "checking status\n",
+ jcr->JobId, jcr->Job, job->MaxWaitTime);
+ switch (jcr->JobStatus) {
+ case JS_Created:
+ case JS_Blocked:
+ case JS_WaitFD:
+ case JS_WaitSD:
+ case JS_WaitStoreRes:
+ case JS_WaitClientRes:
+ case JS_WaitJobRes:
+ case JS_WaitPriority:
+ case JS_WaitMaxJobs:
+ case JS_WaitStartTime:
+ cancel = true;
+ Dmsg0(200, "JCR blocked in #1\n");
break;
+ case JS_Running:
+ Dmsg0(800, "JCR running, checking SD status\n");
+ switch (jcr->SDJobStatus) {
+ case JS_WaitMount:
+ case JS_WaitMedia:
+ case JS_WaitFD:
+ cancel = true;
+ Dmsg0(800, "JCR blocked in #2\n");
+ break;
+ default:
+ Dmsg0(800, "JCR not blocked in #2\n");
+ break;
+ }
+ break;
+ case JS_Terminated:
+ case JS_ErrorTerminated:
+ case JS_Canceled:
+ case JS_FatalError:
+ Dmsg0(800, "JCR already dead in #3\n");
+ break;
+ default:
+ Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
+ jcr->JobStatus);
}
- /*
- * Since we released a lock, if there are any threads
- * waiting, wake them up so that they can try again.
- */
- if (waiting > 0) {
- pthread_cond_broadcast(&resource_wait);
- }
- V(mutex);
+ Dmsg3(800, "MaxWaitTime result: %scancel JCR %p (%s)\n",
+ cancel ? "" : "do not ", jcr, jcr->job);
+
+ return cancel;
}
-#endif
/*
- * This is called at the end of the job to release
- * any resource limits on the number of jobs. If
- * there are any other jobs waiting, we wake them
- * up so that they can try again.
+ * Check if maxruntime has expired and if the job can be
+ * canceled.
*/
-static void release_resource_locks(JCR *jcr)
+static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
{
- if (!jcr->acquired_resource_locks) {
- return; /* Job canceled, no locks acquired */
- }
-#ifdef USE_SEMAPHORE
- P(mutex);
- sem_unlock(&jcr->store->sem);
- sem_unlock(&jcr->client->sem);
- sem_unlock(&jcr->job->sem);
- sem_unlock(&job_lock);
- if (waiting > 0) {
- pthread_cond_broadcast(&resource_wait);
- }
- jcr->acquired_resource_locks = false;
- V(mutex);
-#endif
+ bool cancel = false;
+
+ if (jcr->job->MaxRunTime == 0) {
+ return false;
+ }
+ if ((watchdog_time - jcr->start_time) < jcr->job->MaxRunTime) {
+ Dmsg3(200, "Job %p (%s) with MaxRunTime %d not expired\n",
+ jcr, jcr->Job, jcr->job->MaxRunTime);
+ return false;
+ }
+
+ switch (jcr->JobStatus) {
+ case JS_Created:
+ case JS_Running:
+ case JS_Blocked:
+ case JS_WaitFD:
+ case JS_WaitSD:
+ case JS_WaitStoreRes:
+ case JS_WaitClientRes:
+ case JS_WaitJobRes:
+ case JS_WaitPriority:
+ case JS_WaitMaxJobs:
+ case JS_WaitStartTime:
+ case JS_Differences:
+ cancel = true;
+ break;
+ case JS_Terminated:
+ case JS_ErrorTerminated:
+ case JS_Canceled:
+ case JS_FatalError:
+ cancel = false;
+ break;
+ default:
+ Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
+ jcr->JobStatus);
+ }
+
+ Dmsg3(200, "MaxRunTime result: %scancel JCR %p (%s)\n",
+ cancel ? "" : "do not ", jcr, jcr->job);
+
+ return cancel;
}
+
/*
* Get or create a Client record for this Job
*/
-int get_or_create_client_record(JCR *jcr)
+bool get_or_create_client_record(JCR *jcr)
{
CLIENT_DBR cr;
if (!jcr->client_name) {
jcr->client_name = get_pool_memory(PM_NAME);
}
- pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
+ pm_strcpy(jcr->client_name, jcr->client->hdr.name);
if (!db_create_client_record(jcr, jcr->db, &cr)) {
- Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
+ Jmsg(jcr, M_FATAL, 0, _("Could not create Client record. ERR=%s\n"),
db_strerror(jcr->db));
- return 0;
+ return false;
}
jcr->jr.ClientId = cr.ClientId;
if (cr.Uname[0]) {
if (!jcr->client_uname) {
jcr->client_uname = get_pool_memory(PM_NAME);
}
- pm_strcpy(&jcr->client_uname, cr.Uname);
+ pm_strcpy(jcr->client_uname, cr.Uname);
}
- Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
+ Dmsg2(100, "Created Client %s record %d\n", jcr->client->hdr.name,
jcr->jr.ClientId);
- return 1;
+ return true;
+}
+
+bool get_or_create_fileset_record(JCR *jcr, FILESET_DBR *fsr)
+{
+ /*
+ * Get or Create FileSet record
+ */
+ memset(fsr, 0, sizeof(FILESET_DBR));
+ bstrncpy(fsr->FileSet, jcr->fileset->hdr.name, sizeof(fsr->FileSet));
+ if (jcr->fileset->have_MD5) {
+ struct MD5Context md5c;
+ unsigned char signature[16];
+ memcpy(&md5c, &jcr->fileset->md5c, sizeof(md5c));
+ MD5Final(signature, &md5c);
+ bin_to_base64(fsr->MD5, (char *)signature, 16); /* encode 16 bytes */
+ bstrncpy(jcr->fileset->MD5, fsr->MD5, sizeof(jcr->fileset->MD5));
+ } else {
+ Jmsg(jcr, M_WARNING, 0, _("FileSet MD5 signature not found.\n"));
+ }
+ if (!jcr->fileset->ignore_fs_changes ||
+ !db_get_fileset_record(jcr, jcr->db, fsr)) {
+ if (!db_create_fileset_record(jcr, jcr->db, fsr)) {
+ Jmsg(jcr, M_ERROR, 0, _("Could not create FileSet \"%s\" record. ERR=%s\n"),
+ fsr->FileSet, db_strerror(jcr->db));
+ return false;
+ }
+ }
+ jcr->jr.FileSetId = fsr->FileSetId;
+ if (fsr->created) {
+ Jmsg(jcr, M_INFO, 0, _("Created new FileSet record \"%s\" %s\n"),
+ fsr->FileSet, fsr->cCreateTime);
+ }
+ Dmsg2(119, "Created FileSet %s record %u\n", jcr->fileset->hdr.name,
+ jcr->jr.FileSetId);
+ return true;
}
+void init_jcr_job_record(JCR *jcr)
+{
+ jcr->jr.SchedTime = jcr->sched_time;
+ jcr->jr.StartTime = jcr->start_time;
+ jcr->jr.EndTime = 0; /* perhaps rescheduled, clear it */
+ jcr->jr.JobType = jcr->JobType;
+ jcr->jr.JobLevel = jcr->JobLevel;
+ jcr->jr.JobStatus = jcr->JobStatus;
+ jcr->jr.JobId = jcr->JobId;
+ bstrncpy(jcr->jr.Name, jcr->job->hdr.name, sizeof(jcr->jr.Name));
+ bstrncpy(jcr->jr.Job, jcr->Job, sizeof(jcr->jr.Job));
+}
/*
* Write status and such in DB
*/
void update_job_end_record(JCR *jcr)
{
- if (jcr->jr.EndTime == 0) {
- jcr->jr.EndTime = time(NULL);
- }
+ jcr->jr.EndTime = time(NULL);
jcr->end_time = jcr->jr.EndTime;
jcr->jr.JobId = jcr->JobId;
jcr->jr.JobStatus = jcr->JobStatus;
jcr->jr.VolSessionId = jcr->VolSessionId;
jcr->jr.VolSessionTime = jcr->VolSessionTime;
if (!db_update_job_end_record(jcr, jcr->db, &jcr->jr)) {
- Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
+ Jmsg(jcr, M_WARNING, 0, _("Error updating job record. %s"),
db_strerror(jcr->db));
}
}
* Returns: unique job name in jcr->Job
* date/time in jcr->start_time
*/
-void create_unique_job_name(JCR *jcr, char *base_name)
+void create_unique_job_name(JCR *jcr, const char *base_name)
{
/* Job start mutex */
static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
char *p;
/* Guarantee unique start time -- maximum one per second, and
- * thus unique Job Name
+ * thus unique Job Name
*/
P(mutex); /* lock creation of jobs */
now = time(NULL);
/* Form Unique JobName */
localtime_r(&now, &tm);
/* Use only characters that are permitted in Windows filenames */
- strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
+ strftime(dt, sizeof(dt), "%Y-%m-%d_%H.%M.%S", &tm);
bstrncpy(name, base_name, sizeof(name));
name[sizeof(name)-22] = 0; /* truncate if too long */
bsnprintf(jcr->Job, sizeof(jcr->Job), "%s.%s", name, dt); /* add date & time */
}
}
-/*
- * Free the Job Control Record if no one is still using it.
- * Called from main free_jcr() routine in src/lib/jcr.c so
- * that we can do our Director specific cleanup of the jcr.
- */
-void dird_free_jcr(JCR *jcr)
+/* Called directly from job rescheduling */
+void dird_free_jcr_pointers(JCR *jcr)
{
- Dmsg0(200, "Start dird free_jcr\n");
-
if (jcr->sd_auth_key) {
free(jcr->sd_auth_key);
jcr->sd_auth_key = NULL;
bnet_close(jcr->store_bsock);
jcr->store_bsock = NULL;
}
- if (jcr->fname) {
+ if (jcr->fname) {
Dmsg0(200, "Free JCR fname\n");
free_pool_memory(jcr->fname);
jcr->fname = NULL;
free_pool_memory(jcr->client_uname);
jcr->client_uname = NULL;
}
+ if (jcr->term_wait_inited) {
+ pthread_cond_destroy(&jcr->term_wait);
+ jcr->term_wait_inited = false;
+ }
+}
+
+/*
+ * Free the Job Control Record if no one is still using it.
+ * Called from main free_jcr() routine in src/lib/jcr.c so
+ * that we can do our Director specific cleanup of the jcr.
+ */
+void dird_free_jcr(JCR *jcr)
+{
+ Dmsg0(200, "Start dird free_jcr\n");
+
+ dird_free_jcr_pointers(jcr);
+
+ /* Delete lists setup to hold storage pointers */
+ if (jcr->storage) {
+ delete jcr->storage;
+ }
+ jcr->job_end_push.destroy();
Dmsg0(200, "End dird free_jcr\n");
}
*/
void set_jcr_defaults(JCR *jcr, JOB *job)
{
+ STORE *st;
jcr->job = job;
jcr->JobType = job->JobType;
- jcr->JobLevel = job->level;
- jcr->store = job->storage;
+ switch (jcr->JobType) {
+ case JT_ADMIN:
+ case JT_RESTORE:
+ jcr->JobLevel = L_NONE;
+ break;
+ default:
+ jcr->JobLevel = job->JobLevel;
+ break;
+ }
+ jcr->JobPriority = job->Priority;
+ /* Copy storage definitions -- deleted in dir_free_jcr above */
+ if (job->storage) {
+ if (jcr->storage) {
+ delete jcr->storage;
+ }
+ jcr->storage = New(alist(10, not_owned_by_alist));
+ foreach_alist(st, job->storage) {
+ jcr->storage->append(st);
+ }
+ }
+ if (jcr->storage) {
+ jcr->store = (STORE *)jcr->storage->first();
+ }
jcr->client = job->client;
if (!jcr->client_name) {
jcr->client_name = get_pool_memory(PM_NAME);
}
- pm_strcpy(&jcr->client_name, jcr->client->hdr.name);
+ pm_strcpy(jcr->client_name, jcr->client->hdr.name);
jcr->pool = job->pool;
+ jcr->full_pool = job->full_pool;
+ jcr->inc_pool = job->inc_pool;
+ jcr->dif_pool = job->dif_pool;
jcr->catalog = job->client->catalog;
jcr->fileset = job->fileset;
- jcr->messages = job->messages;
+ jcr->messages = job->messages;
+ jcr->spool_data = job->spool_data;
+ jcr->write_part_after_job = job->write_part_after_job;
if (jcr->RestoreBootstrap) {
free(jcr->RestoreBootstrap);
+ jcr->RestoreBootstrap = NULL;
}
/* This can be overridden by Console program */
if (job->RestoreBootstrap) {
jcr->RestoreBootstrap = bstrdup(job->RestoreBootstrap);
}
+ /* This can be overridden by Console program */
+ jcr->verify_job = job->verify_job;
/* If no default level given, set one */
if (jcr->JobLevel == 0) {
switch (jcr->JobType) {
break;
case JT_RESTORE:
case JT_ADMIN:
- jcr->JobLevel = L_FULL;
+ jcr->JobLevel = L_NONE;
break;
default:
break;
}
/*
- * Edit codes into Run command
- * %% = %
- * %c = Client's name
- * %d = Director's name
- * %i = JobId
- * %e = Job Exit
- * %j = Job
- * %l = Job Level
- * %n = Job name
- * %t = Job type
- *
- * omsg = edited output message
- * imsg = input string containing edit codes (%x)
- *
+ * copy the storage definitions from an old JCR to a new one
*/
-static char *edit_run_codes(JCR *jcr, char *omsg, char *imsg)
+void copy_storage(JCR *new_jcr, JCR *old_jcr)
{
- char *p;
- const char *str;
- char add[20];
-
- *omsg = 0;
- Dmsg1(200, "edit_run_codes: %s\n", imsg);
- for (p=imsg; *p; p++) {
- if (*p == '%') {
- switch (*++p) {
- case '%':
- str = "%";
- break;
- case 'c':
- str = jcr->client_name;
- if (!str) {
- str = "";
- }
- break;
- case 'd':
- str = my_name;
- break;
- case 'e':
- str = job_status_to_str(jcr->JobStatus);
- break;
- case 'i':
- sprintf(add, "%d", jcr->JobId);
- str = add;
- break;
- case 'j': /* Job */
- str = jcr->Job;
- break;
- case 'l':
- str = job_level_to_str(jcr->JobLevel);
- break;
- case 'n':
- str = jcr->job->hdr.name;
- break;
- case 't':
- str = job_type_to_str(jcr->JobType);
- break;
- default:
- add[0] = '%';
- add[1] = *p;
- add[2] = 0;
- str = add;
- break;
- }
- } else {
- add[0] = *p;
- add[1] = 0;
- str = add;
+ if (old_jcr->storage) {
+ STORE *st;
+ if (old_jcr->storage) {
+ delete old_jcr->storage;
+ }
+ new_jcr->storage = New(alist(10, not_owned_by_alist));
+ foreach_alist(st, old_jcr->storage) {
+ new_jcr->storage->append(st);
+ }
+ }
+ if (old_jcr->store) {
+ new_jcr->store = old_jcr->store;
+ } else if (new_jcr->storage) {
+ new_jcr->store = (STORE *)new_jcr->storage->first();
+ }
+}
+
+/* Set storage override */
+void set_storage(JCR *jcr, STORE *store)
+{
+ STORE *storage;
+
+ jcr->store = store;
+ foreach_alist(storage, jcr->storage) {
+ if (store == storage) {
+ return;
}
- Dmsg1(200, "add_str %s\n", str);
- pm_strcat(&omsg, (char *)str);
- Dmsg1(200, "omsg=%s\n", omsg);
}
- return omsg;
+ /* Store not in list, so add it */
+ jcr->storage->prepend(store);
}
projects / bacula / bacula / blobdiff