bool setup_job(JCR *jcr)
{
int errstat;
jcr->lock();
Dsm_check(100);
init_msg(jcr, jcr->res.messages, job_code_callback_director);
/*
* Initialize termination condition variable
*/
if ((errstat = pthread_cond_init(&jcr->term_wait, NULL)) != 0) {
berrno be;
Jmsg1(jcr, M_FATAL, 0, _("Unable to init job cond variable: ERR=%s\n"), be.bstrerror(errstat));
jcr->unlock();
goto bail_out;
}
jcr->term_wait_inited = true;
/*
* Initialize nextrun ready condition variable
*/
if ((errstat = pthread_cond_init(&jcr->nextrun_ready, NULL)) != 0) {
berrno be;
Jmsg1(jcr, M_FATAL, 0, _("Unable to init job nextrun cond variable: ERR=%s\n"), be.bstrerror(errstat));
jcr->unlock();
goto bail_out;
}
jcr->nextrun_ready_inited = true;
create_unique_job_name(jcr, jcr->res.job->name());
jcr->setJobStatus(JS_Created);
jcr->unlock();
/*
* Open database
*/
Dmsg0(100, "Open database\n");
jcr->db = db_sql_get_pooled_connection(jcr,
jcr->res.catalog->db_driver,
jcr->res.catalog->db_name,
jcr->res.catalog->db_user,
jcr->res.catalog->db_password,
jcr->res.catalog->db_address,
jcr->res.catalog->db_port,
jcr->res.catalog->db_socket,
jcr->res.catalog->mult_db_connections,
jcr->res.catalog->disable_batch_insert);
if (jcr->db == NULL) {
Jmsg(jcr, M_FATAL, 0, _("Could not open database \"%s\".\n"),
jcr->res.catalog->db_name);
goto bail_out;
}
Dmsg0(150, "DB opened\n");
if (!jcr->fname) {
jcr->fname = get_pool_memory(PM_FNAME);
}
if (!jcr->res.pool_source) {
jcr->res.pool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.pool_source, _("unknown source"));
}
if (!jcr->res.npool_source) {
jcr->res.npool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.npool_source, _("unknown source"));
}
if (jcr->JobReads()) {
if (!jcr->res.rpool_source) {
jcr->res.rpool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->res.rpool_source, _("unknown source"));
}
}
/*
* Create Job record
*/
init_jcr_job_record(jcr);
if (!get_or_create_client_record(jcr)) {
goto bail_out;
}
if (!db_create_job_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
goto bail_out;
}
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);
new_plugins(jcr); /* instantiate plugins for this jcr */
generate_plugin_event(jcr, bDirEventJobStart);
if (job_canceled(jcr)) {
goto bail_out;
}
if (jcr->JobReads() && !jcr->rstorage) {
if (jcr->res.job->storage) {
copy_rwstorage(jcr, jcr->res.job->storage, _("Job resource"));
} else {
copy_rwstorage(jcr, jcr->res.job->pool->storage, _("Pool resource"));
}
}
if (!jcr->JobReads()) {
free_rstorage(jcr);
}
/*
* Now, do pre-run stuff, like setting job level (Inc/diff, ...)
* this allows us to setup a proper job start record for restarting
* in case of later errors.
*/
switch (jcr->getJobType()) {
case JT_BACKUP:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!do_ndmp_backup_init(jcr)) {
ndmp_backup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
if (!do_native_vbackup_init(jcr)) {
native_vbackup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
} else {
if (!do_native_backup_init(jcr)) {
native_backup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
}
break;
}
break;
case JT_VERIFY:
if (!do_verify_init(jcr)) {
verify_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_RESTORE:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!do_ndmp_restore_init(jcr)) {
ndmp_restore_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
/*
* Any non NDMP restore is not interested at the items
* that were selected for restore so drop them now.
*/
if (jcr->restore_tree_root) {
free_tree(jcr->restore_tree_root);
jcr->restore_tree_root = NULL;
}
if (!do_native_restore_init(jcr)) {
native_restore_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
}
break;
case JT_ADMIN:
if (!do_admin_init(jcr)) {
admin_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_COPY:
case JT_MIGRATE:
if (!do_migration_init(jcr)) {
migration_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
default:
Pmsg1(0, _("Unimplemented job type: %d\n"), jcr->getJobType());
jcr->setJobStatus(JS_ErrorTerminated);
goto bail_out;
}
generate_plugin_event(jcr, bDirEventJobInit);
Dsm_check(100);
return true;
bail_out:
return false;
}
/*
* Do a virtual backup, which consolidates all previous backups into
* a sort of synthetic Full.
*
* Returns: false on failure
* true on success
*/
bool do_native_vbackup(JCR *jcr)
{
char ed1[100];
BSOCK *sd;
char *p;
db_list_ctx jobids;
if (!jcr->res.rstorage) {
Jmsg(jcr, M_FATAL, 0, _("No storage for reading given.\n"));
return false;
}
if (!jcr->res.wstorage) {
Jmsg(jcr, M_FATAL, 0, _("No storage for writing given.\n"));
return false;
}
Dmsg2(100, "rstorage=%p wstorage=%p\n", jcr->res.rstorage, jcr->res.wstorage);
Dmsg2(100, "Read store=%s, write store=%s\n",
((STORERES *)jcr->res.rstorage->first())->name(),
((STORERES *)jcr->res.wstorage->first())->name());
/*
* Print Job Start message
*/
Jmsg(jcr, M_INFO, 0, _("Start Virtual Backup JobId %s, Job=%s\n"),
edit_uint64(jcr->JobId, ed1), jcr->Job);
if (!jcr->accurate) {
Jmsg(jcr, M_WARNING, 0,
_("This Job is not an Accurate backup so is not equivalent to a Full backup.\n"));
}
db_accurate_get_jobids(jcr, jcr->db, &jcr->jr, &jobids);
Dmsg1(10, "Accurate jobids=%s\n", jobids.list);
if (jobids.count == 0) {
Jmsg(jcr, M_FATAL, 0, _("No previous Jobs found.\n"));
return false;
}
/*
* Now we find the last job that ran and store it's info in
* the previous_jr record. We will set our times to the
* values from that job so that anything changed after that
* time will be picked up on the next backup.
*/
p = strrchr(jobids.list, ','); /* find last jobid */
if (p != NULL) {
p++;
} else {
p = jobids.list;
}
memset(&jcr->previous_jr, 0, sizeof(jcr->previous_jr));
jcr->previous_jr.JobId = str_to_int64(p);
Dmsg1(10, "Previous JobId=%s\n", p);
if (!db_get_job_record(jcr, jcr->db, &jcr->previous_jr)) {
Jmsg(jcr, M_FATAL, 0, _("Error getting Job record for previous Job: ERR=%s"),
db_strerror(jcr->db));
return false;
}
if (!create_bootstrap_file(jcr, jobids.list)) {
Jmsg(jcr, M_FATAL, 0, _("Could not get or create the FileSet record.\n"));
return false;
}
/*
* Open a message channel connection with the Storage
* daemon. This is to let him know that our client
* will be contacting him for a backup session.
*
*/
Dmsg0(110, "Open connection with storage daemon\n");
jcr->setJobStatus(JS_WaitSD);
/*
* Start conversation with Storage daemon
*/
if (!connect_to_storage_daemon(jcr, 10, me->SDConnectTimeout, true)) {
return false;
}
sd = jcr->store_bsock;
/*
* Now start a job with the Storage daemon
*/
if (!start_storage_daemon_job(jcr, jcr->res.rstorage, jcr->res.wstorage, /* send_bsr */ true)) {
return false;
}
Dmsg0(100, "Storage daemon connection OK\n");
/*
* We re-update the job start record so that the start
* time is set after the run before job. This avoids
* that any files created by the run before job will
* be saved twice. They will be backed up in the current
* job, but not in the next one unless they are changed.
* Without this, they will be backed up in this job and
* in the next job run because in that case, their date
* is after the start of this run.
*/
jcr->start_time = time(NULL);
jcr->jr.StartTime = jcr->start_time;
jcr->jr.JobTDate = jcr->start_time;
jcr->setJobStatus(JS_Running);
/*
* Update job start record
*/
if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
return false;
}
/*
* Declare the job started to start the MaxRunTime check
*/
jcr->setJobStarted();
/*
* Start the job prior to starting the message thread below
* to avoid two threads from using the BSOCK structure at
* the same time.
*/
if (!sd->fsend("run")) {
return false;
}
/*
* Now start a Storage daemon message thread
*/
if (!start_storage_daemon_message_thread(jcr)) {
return false;
}
jcr->setJobStatus(JS_Running);
/*
* Pickup Job termination data
* Note, the SD stores in jcr->JobFiles/ReadBytes/JobBytes/JobErrors
*/
wait_for_storage_daemon_termination(jcr);
jcr->setJobStatus(jcr->SDJobStatus);
db_write_batch_file_records(jcr); /* used by bulk batch file insert */
if (!jcr->is_JobStatus(JS_Terminated)) {
return false;
}
native_vbackup_cleanup(jcr, jcr->JobStatus);
return true;
}
/*
* 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
*/
static void *job_thread(void *arg)
{
JCR *jcr = (JCR *)arg;
pthread_detach(pthread_self());
Dsm_check(100);
Dmsg0(200, "=====Start Job=========\n");
jcr->setJobStatus(JS_Running); /* this will be set only if no error */
jcr->start_time = time(NULL); /* set the real start time */
jcr->jr.StartTime = jcr->start_time;
if (jcr->res.job->MaxStartDelay != 0 && jcr->res.job->MaxStartDelay <
(utime_t)(jcr->start_time - jcr->sched_time)) {
jcr->setJobStatus(JS_Canceled);
Jmsg(jcr, M_FATAL, 0, _("Job canceled because max start delay time exceeded.\n"));
}
if (job_check_maxrunschedtime(jcr)) {
jcr->setJobStatus(JS_Canceled);
Jmsg(jcr, M_FATAL, 0, _("Job canceled because max run sched time exceeded.\n"));
}
/* TODO : check if it is used somewhere */
if (jcr->res.job->RunScripts == NULL) {
Dmsg0(200, "Warning, job->RunScripts is empty\n");
jcr->res.job->RunScripts = New(alist(10, not_owned_by_alist));
}
if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
}
/* Run any script BeforeJob on dird */
run_scripts(jcr, jcr->res.job->RunScripts, "BeforeJob");
/*
* We re-update the job start record so that the start
* time is set after the run before job. This avoids
* that any files created by the run before job will
* be saved twice. They will be backed up in the current
* job, but not in the next one unless they are changed.
* Without this, they will be backed up in this job and
* in the next job run because in that case, their date
* is after the start of this run.
*/
jcr->start_time = time(NULL);
jcr->jr.StartTime = jcr->start_time;
if (!db_update_job_start_record(jcr, jcr->db, &jcr->jr)) {
Jmsg(jcr, M_FATAL, 0, "%s", db_strerror(jcr->db));
}
generate_plugin_event(jcr, bDirEventJobRun);
switch (jcr->getJobType()) {
case JT_BACKUP:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!job_canceled(jcr)) {
if (do_ndmp_backup(jcr)) {
do_autoprune(jcr);
} else {
ndmp_backup_cleanup(jcr, JS_ErrorTerminated);
}
} else {
ndmp_backup_cleanup(jcr, JS_Canceled);
}
break;
default:
if (!job_canceled(jcr)) {
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
if (do_native_vbackup(jcr)) {
do_autoprune(jcr);
} else {
native_vbackup_cleanup(jcr, JS_ErrorTerminated);
}
} else {
if (do_native_backup(jcr)) {
do_autoprune(jcr);
} else {
native_backup_cleanup(jcr, JS_ErrorTerminated);
}
}
} else {
if (jcr->is_JobLevel(L_VIRTUAL_FULL)) {
native_vbackup_cleanup(jcr, JS_Canceled);
} else {
native_backup_cleanup(jcr, JS_Canceled);
}
}
break;
}
break;
case JT_VERIFY:
if (!job_canceled(jcr)) {
if (do_verify(jcr)) {
do_autoprune(jcr);
} else {
verify_cleanup(jcr, JS_ErrorTerminated);
}
} else {
verify_cleanup(jcr, JS_Canceled);
}
break;
case JT_RESTORE:
switch (jcr->getJobProtocol()) {
case PT_NDMP:
if (!job_canceled(jcr)) {
if (do_ndmp_restore(jcr)) {
do_autoprune(jcr);
} else {
ndmp_restore_cleanup(jcr, JS_ErrorTerminated);
}
} else {
ndmp_restore_cleanup(jcr, JS_Canceled);
}
break;
default:
if (!job_canceled(jcr)) {
if (do_native_restore(jcr)) {
do_autoprune(jcr);
} else {
native_restore_cleanup(jcr, JS_ErrorTerminated);
}
} else {
native_restore_cleanup(jcr, JS_Canceled);
}
break;
}
break;
case JT_ADMIN:
if (!job_canceled(jcr)) {
if (do_admin(jcr)) {
do_autoprune(jcr);
} else {
admin_cleanup(jcr, JS_ErrorTerminated);
}
} else {
admin_cleanup(jcr, JS_Canceled);
}
break;
case JT_COPY:
case JT_MIGRATE:
if (!job_canceled(jcr)) {
if (do_migration(jcr)) {
do_autoprune(jcr);
} else {
migration_cleanup(jcr, JS_ErrorTerminated);
}
} else {
migration_cleanup(jcr, JS_Canceled);
}
break;
default:
Pmsg1(0, _("Unimplemented job type: %d\n"), jcr->getJobType());
break;
}
run_scripts(jcr, jcr->res.job->RunScripts, "AfterJob");
/* Send off any queued messages */
if (jcr->msg_queue && jcr->msg_queue->size() > 0) {
dequeue_messages(jcr);
}
generate_plugin_event(jcr, bDirEventJobEnd);
Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
Dsm_check(100);
return NULL;
}
