* 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
/* Forward referenced subroutines */
static void *job_thread(void *arg);
+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);
/* Exported subroutines */
-
/* Imported subroutines */
extern void term_scheduler();
extern void term_ua_server();
extern int do_restore(JCR *jcr);
extern int do_verify(JCR *jcr);
+/* Imported variables */
+extern time_t watchdog_time;
+
jobq_t job_queue;
void init_job_server(int max_workers)
{
int stat;
+ watchdog_t *wd;
+
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));
+ berrno be;
+ be.set_errno(stat);
+ Emsg1(M_ABORT, 0, _("Could not init job queue: ERR=%s\n"), be.strerror());
}
- return;
+ 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 = new_control_jcr("*JobMonitor*", JT_SYSTEM);
+ register_watchdog(wd);
}
/*
* 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;
+ JobId_t JobId = 0;
- sm_check(__FILE__, __LINE__, True);
+ 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.EndTime = 0; /* perhaps rescheduled, clear it */
- 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;
+ be.set_errno(errstat);
+ Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.strerror());
+ goto bail_out;
}
+ jcr->term_wait_inited = true;
/*
* Open database
if (jcr->db) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
}
- set_jcr_job_status(jcr, JS_ErrorTerminated);
- free_jcr(jcr);
- return;
+ goto bail_out;
}
Dmsg0(50, "DB opened\n");
/*
* 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));
- set_jcr_job_status(jcr, JS_ErrorTerminated);
- free_jcr(jcr);
- return;
+ goto bail_out;
}
- jcr->JobId = jcr->jr.JobId;
- ASSERT(jcr->jr.JobId > 0);
+ JobId = jcr->JobId = jcr->jr.JobId;
- Dmsg4(50, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
- jcr->JobId, jcr->Job, jcr->jr.Type, jcr->jr.Level);
+ Dmsg4(100, "Created job record JobId=%d Name=%s Type=%c Level=%c\n",
+ jcr->JobId, jcr->Job, jcr->jr.JobType, jcr->jr.JobLevel);
Dmsg0(200, "Add jrc to work queue\n");
/* 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;
+ be.set_errno(stat);
+ Jmsg(jcr, M_FATAL, 0, _("Could not add job queue: ERR=%s\n"), be.strerror());
+ JobId = 0;
+ goto bail_out;
}
Dmsg0(100, "Done run_job()\n");
+
+ V(jcr->mutex);
+ return JobId;
+
+bail_out:
+ set_jcr_job_status(jcr, JS_ErrorTerminated);
+ V(jcr->mutex);
+ return JobId;
+
}
+
/*
- * This is the engine called by job_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 and all
* necessary resources are allocated -- see jobq.c
{
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 ( ;; ) {
-
Dmsg0(200, "=====Start Job=========\n");
jcr->start_time = time(NULL); /* set the real start time */
set_jcr_job_status(jcr, JS_Running);
}
status = close_bpipe(bpipe);
if (status != 0) {
- Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob returned non-zero status=%d\n"),
- status);
+ berrno be;
+ be.set_errno(status);
+ Jmsg(jcr, M_FATAL, 0, _("RunBeforeJob error: ERR=%s\n"), be.strerror());
set_jcr_job_status(jcr, JS_FatalError);
update_job_end_record(jcr);
goto bail_out;
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) {
+ berrno be;
+ be.set_errno(status);
if (jcr->JobStatus == JS_Terminated) {
- Jmsg(jcr, M_FATAL, 0, _("RunAfterJob returned non-zero status=%d\n"),
- status);
+ Jmsg(jcr, M_WARNING, 0, _("RunAfterJob error: ERR=%s\n"), be.strerror());
} else {
- Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob returned non-zero status=%d\n"),
- status);
+ Jmsg(jcr, M_FATAL, 0, _("RunAfterFailedJob error: ERR=%s\n"), be.strerror());
}
- 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;
}
Dmsg0(50, "======== End Job ==========\n");
- sm_check(__FILE__, __LINE__, True);
+ sm_check(__FILE__, __LINE__, true);
return NULL;
}
+/*
+ * 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
*/
-int get_or_create_client_record(JCR *jcr)
+bool get_or_create_client_record(JCR *jcr)
{
CLIENT_DBR cr;
if (!db_create_client_record(jcr, jcr->db, &cr)) {
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]) {
}
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 (!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;
* 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;
free_pool_memory(jcr->client_uname);
jcr->client_uname = NULL;
}
- pthread_cond_destroy(&jcr->term_wait);
+ if (jcr->term_wait_inited) {
+ pthread_cond_destroy(&jcr->term_wait);
+ }
+ jcr->job_end_push.destroy();
Dmsg0(200, "End dird free_jcr\n");
}
}
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