* Version $Id$
*/
/*
- Copyright (C) 2000-2003 Kern Sibbald and John Walker
+ Copyright (C) 2000-2004 Kern Sibbald and John Walker
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
if ((stat = jobq_init(&job_queue, max_workers, job_thread)) != 0) {
Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), strerror(stat));
}
- if ((wd = watchdog_new()) == NULL) {
+ if ((wd = new_watchdog()) == NULL) {
Emsg0(M_ABORT, 0, _("Could not init job monitor watchdogs\n"));
}
wd->callback = job_monitor_watchdog;
wd->destructor = job_monitor_destructor;
wd->one_shot = false;
wd->interval = 60;
- wd->data = create_control_jcr("*JobMonitor*", JT_SYSTEM);
+ wd->data = new_control_jcr("*JobMonitor*", JT_SYSTEM);
register_watchdog(wd);
return;
}
-static void job_monitor_destructor(watchdog_t *self)
-{
- JCR *control_jcr = (JCR *) self->data;
-
- free_jcr(control_jcr);
-}
-
-static void job_monitor_watchdog(watchdog_t *self)
-{
- JCR *control_jcr, *jcr;
-
- control_jcr = (JCR *) self->data;
-
- Dmsg1(200, "job_monitor_watchdog %p called\n", self);
-
- lock_jcr_chain();
-
- for (jcr = NULL; (jcr = get_next_jcr(jcr)); /* nothing */) {
- bool cancel;
-
- if (jcr->JobId == 0) {
- Dmsg2(200, "Skipping JCR %p (%s) with JobId 0\n",
- jcr, jcr->Job);
- /* Keep reference counts correct */
- free_locked_jcr(jcr);
- continue;
- }
-
- /* check MaxWaitTime */
- cancel = job_check_maxwaittime(control_jcr, jcr);
-
- /* check MaxRunTime */
- cancel |= job_check_maxruntime(control_jcr, jcr);
-
- if (cancel) {
- Dmsg3(200, "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(200, "Have cancelled JCR %p\n", jcr);
- }
-
- /* Keep reference counts correct */
- free_locked_jcr(jcr);
- }
- unlock_jcr_chain();
-}
-
-static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
-{
- bool cancel = false;
-
- if (jcr->job->MaxWaitTime == 0) {
- return false;
- }
- if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
- Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
- jcr, jcr->Job, jcr->job->MaxWaitTime);
- return false;
- }
- Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
- "checking status\n",
- jcr->JobId, jcr->Job, jcr->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(200, "JCR running, checking SD status\n");
- switch (jcr->SDJobStatus) {
- case JS_WaitMount:
- case JS_WaitMedia:
- case JS_WaitFD:
- cancel = true;
- Dmsg0(200, "JCR blocked in #2\n");
- break;
- default:
- Dmsg0(200, "JCR not blocked in #2\n");
- break;
- }
- break;
- case JS_Terminated:
- case JS_ErrorTerminated:
- case JS_Canceled:
- Dmsg0(200, "JCR already dead in #3\n");
- break;
- default:
- Emsg1(M_ABORT, 0, _("Unhandled job status code %d\n"),
- jcr->JobStatus);
- }
- Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
- cancel ? "" : "do not ", jcr, jcr->job);
-
- return cancel;
-}
-
-static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
-{
- 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_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_Running:
- cancel = true;
- break;
- case JS_Terminated:
- case JS_ErrorTerminated:
- case JS_Canceled:
- cancel = false;
- break;
- default:
- Emsg1(M_ABORT, 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;
-}
-
/*
* Run a job -- typically called by the scheduler, but may also
* be called by the UA (Console program).
goto bail_out;
}
jcr->JobId = jcr->jr.JobId;
- ASSERT(jcr->jr.JobId > 0);
Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
/* 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));
+ Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), strerror(stat));
+ goto bail_out;
}
Dmsg0(100, "Done run_job()\n");
}
-/*
- * 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.
- */
-int cancel_job(UAContext *ua, JCR *jcr)
-{
- 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;
- }
-
- /* Cancel Storage daemon */
- if (jcr->store_bsock) {
- ua->jcr->store = 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;
- }
- }
-
- return 1;
-}
/*
* This is the engine called by jobq.c:jobq_add() when we were pulled
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) {
if (jcr->JobStatus == JS_Terminated) {
- Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
+ Jmsg(jcr, M_WARNING, 0, _("RunAfterJob returned non-zero status=%d\n"),
status);
} else {
Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
status);
}
- set_jcr_job_status(jcr, JS_FatalError);
- update_job_end_record(jcr);
}
}
+ /* Send off any queued messages */
+ if (jcr->msg_queue->size() > 0) {
+ dequeue_messages(jcr);
+ }
}
bail_out:
break;
}
+/*
+ * 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.
+ */
+int cancel_job(UAContext *ua, JCR *jcr)
+{
+ 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;
+ }
+
+ /* Cancel Storage daemon */
+ if (jcr->store_bsock) {
+ ua->jcr->store = 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;
+ }
+ }
+
+ return 1;
+}
+
+
+static void job_monitor_destructor(watchdog_t *self)
+{
+ JCR *control_jcr = (JCR *) self->data;
+
+ free_jcr(control_jcr);
+}
+
+static void job_monitor_watchdog(watchdog_t *self)
+{
+ JCR *control_jcr, *jcr;
+
+ control_jcr = (JCR *)self->data;
+
+ Dmsg1(400, "job_monitor_watchdog %p called\n", self);
+
+ lock_jcr_chain();
+
+ foreach_jcr(jcr) {
+ bool cancel;
+
+ if (jcr->JobId == 0) {
+ Dmsg2(400, "Skipping JCR %p (%s) with JobId 0\n",
+ jcr, jcr->Job);
+ /* Keep reference counts correct */
+ free_locked_jcr(jcr);
+ continue;
+ }
+
+ /* check MaxWaitTime */
+ cancel = job_check_maxwaittime(control_jcr, jcr);
+
+ /* check MaxRunTime */
+ cancel |= job_check_maxruntime(control_jcr, jcr);
+
+ if (cancel) {
+ Dmsg3(200, "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(200, "Have cancelled JCR %p\n", jcr);
+ }
+
+ /* Keep reference counts correct */
+ free_locked_jcr(jcr);
+ }
+ unlock_jcr_chain();
+}
+
+/*
+ * Check if the maxwaittime has expired and it is possible
+ * to cancel the job.
+ */
+static bool job_check_maxwaittime(JCR *control_jcr, JCR *jcr)
+{
+ bool cancel = false;
+
+ if (jcr->job->MaxWaitTime == 0) {
+ return false;
+ }
+ if ((watchdog_time - jcr->start_time) < jcr->job->MaxWaitTime) {
+ Dmsg3(200, "Job %p (%s) with MaxWaitTime %d not expired\n",
+ jcr, jcr->Job, jcr->job->MaxWaitTime);
+ return false;
+ }
+ Dmsg3(200, "Job %d (%s): MaxWaitTime of %d seconds exceeded, "
+ "checking status\n",
+ jcr->JobId, jcr->Job, jcr->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(200, "JCR running, checking SD status\n");
+ switch (jcr->SDJobStatus) {
+ case JS_WaitMount:
+ case JS_WaitMedia:
+ case JS_WaitFD:
+ cancel = true;
+ Dmsg0(200, "JCR blocked in #2\n");
+ break;
+ default:
+ Dmsg0(200, "JCR not blocked in #2\n");
+ break;
+ }
+ break;
+ case JS_Terminated:
+ case JS_ErrorTerminated:
+ case JS_Canceled:
+ case JS_FatalError:
+ Dmsg0(200, "JCR already dead in #3\n");
+ break;
+ default:
+ Jmsg1(jcr, M_ERROR, 0, _("Unhandled job status code %d\n"),
+ jcr->JobStatus);
+ }
+ Dmsg3(200, "MaxWaitTime result: %scancel JCR %p (%s)\n",
+ cancel ? "" : "do not ", jcr, jcr->job);
+
+ return cancel;
+}
+
+/*
+ * Check if maxruntime has expired and if the job can be
+ * canceled.
+ */
+static bool job_check_maxruntime(JCR *control_jcr, JCR *jcr)
+{
+ 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
*/
}
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->spool_data = job->spool_data;
if (jcr->RestoreBootstrap) {
free(jcr->RestoreBootstrap);
jcr->RestoreBootstrap = NULL;
projects / bacula / bacula / blobdiff