/*
* Run a File daemon Job -- File daemon already authorized
* Director sends us this command.
*
* Basic task here is:
* - Read a command from the File daemon
* - Execute it
*
*/
void run_job(JCR *jcr)
{
BSOCK *dir = jcr->dir_bsock;
char ec1[30];
dir->set_jcr(jcr);
Dmsg1(120, "Start run Job=%s\n", jcr->Job);
dir->fsend(Job_start, jcr->Job);
jcr->start_time = time(NULL);
jcr->run_time = jcr->start_time;
set_jcr_job_status(jcr, JS_Running);
dir_send_job_status(jcr); /* update director */
do_fd_commands(jcr);
jcr->end_time = time(NULL);
dequeue_messages(jcr); /* send any queued messages */
set_jcr_job_status(jcr, JS_Terminated);
generate_daemon_event(jcr, "JobEnd");
dir->fsend(Job_end, jcr->Job, jcr->JobStatus, jcr->JobFiles,
edit_uint64(jcr->JobBytes, ec1), jcr->JobErrors);
dir->signal(BNET_EOD); /* send EOD to Director daemon */
return;
}
bool setup_job(JCR *jcr)
{
int errstat;
jcr->lock();
Dsm_check(100);
init_msg(jcr, jcr->messages);
/* 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;
create_unique_job_name(jcr, jcr->job->name());
jcr->setJobStatus(JS_Created);
jcr->unlock();
/*
* Open database
*/
Dmsg0(100, "Open database\n");
jcr->db = db_init_database(jcr, jcr->catalog->db_driver, jcr->catalog->db_name,
jcr->catalog->db_user, jcr->catalog->db_password,
jcr->catalog->db_address, jcr->catalog->db_port,
jcr->catalog->db_socket, jcr->catalog->mult_db_connections,
jcr->catalog->disable_batch_insert);
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));
db_close_database(jcr, jcr->db);
}
goto bail_out;
}
Dmsg0(150, "DB opened\n");
if (!jcr->fname) {
jcr->fname = get_pool_memory(PM_FNAME);
}
if (!jcr->pool_source) {
jcr->pool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->pool_source, _("unknown source"));
}
if (jcr->JobReads()) {
if (!jcr->rpool_source) {
jcr->rpool_source = get_pool_memory(PM_MESSAGE);
pm_strcpy(jcr->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);
generate_daemon_event(jcr, "JobStart");
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->job->storage) {
copy_rwstorage(jcr, jcr->job->storage, _("Job resource"));
} else {
copy_rwstorage(jcr, jcr->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:
if (!do_backup_init(jcr)) {
backup_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_VERIFY:
if (!do_verify_init(jcr)) {
verify_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
break;
case JT_RESTORE:
if (!do_restore_init(jcr)) {
restore_cleanup(jcr, JS_ErrorTerminated);
goto bail_out;
}
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_job_event(jcr, "JobInit");
generate_plugin_event(jcr, bDirEventJobInit);
Dsm_check(100);
return true;
bail_out:
return false;
}
/*
* 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->job->MaxStartDelay != 0 && jcr->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->job->RunScripts == NULL) {
Dmsg0(200, "Warning, job->RunScripts is empty\n");
jcr->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->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_job_event(jcr, "JobRun");
generate_plugin_event(jcr, bDirEventJobRun);
switch (jcr->getJobType()) {
case JT_BACKUP:
if (!job_canceled(jcr) && do_backup(jcr)) {
do_autoprune(jcr);
} else {
backup_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_VERIFY:
if (!job_canceled(jcr) && do_verify(jcr)) {
do_autoprune(jcr);
} else {
verify_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_RESTORE:
if (!job_canceled(jcr) && do_restore(jcr)) {
do_autoprune(jcr);
} else {
restore_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_ADMIN:
if (!job_canceled(jcr) && do_admin(jcr)) {
do_autoprune(jcr);
} else {
admin_cleanup(jcr, JS_ErrorTerminated);
}
break;
case JT_COPY:
case JT_MIGRATE:
if (!job_canceled(jcr) && do_migration(jcr)) {
do_autoprune(jcr);
} else {
migration_cleanup(jcr, JS_ErrorTerminated);
}
break;
default:
Pmsg1(0, _("Unimplemented job type: %d\n"), jcr->getJobType());
break;
}
run_scripts(jcr, jcr->job->RunScripts, "AfterJob");
/* Send off any queued messages */
if (jcr->msg_queue && jcr->msg_queue->size() > 0) {
dequeue_messages(jcr);
}
generate_daemon_event(jcr, "JobEnd");
generate_plugin_event(jcr, bDirEventJobEnd);
Dmsg1(50, "======== End Job stat=%c ==========\n", jcr->JobStatus);
Dsm_check(100);
return NULL;
}
/*
* Director requests us to start a job
* Basic tasks done here:
* - We pickup the JobId to be run from the Director.
* - We pickup the device, media, and pool from the Director
* - Wait for a connection from the File Daemon (FD)
* - Accept commands from the FD (i.e. run the job)
* - Return when the connection is terminated or
* there is an error.
*/
bool job_cmd(JCR *jcr)
{
int JobId;
char auth_key[100];
char spool_size[30];
char seed[100];
BSOCK *dir = jcr->dir_bsock;
POOL_MEM job_name, client_name, job, fileset_name, fileset_md5;
int JobType, level, spool_attributes, no_attributes, spool_data;
int write_part_after_job, PreferMountedVols;
int stat;
JCR *ojcr;
/*
* Get JobId and permissions from Director
*/
Dmsg1(100, "<dird: %s", dir->msg);
bstrncpy(spool_size, "0", sizeof(spool_size));
stat = sscanf(dir->msg, jobcmd, &JobId, job.c_str(), job_name.c_str(),
client_name.c_str(),
&JobType, &level, fileset_name.c_str(), &no_attributes,
&spool_attributes, fileset_md5.c_str(), &spool_data,
&write_part_after_job, &PreferMountedVols, spool_size,
&jcr->rerunning, &jcr->VolSessionId, &jcr->VolSessionTime);
if (stat != 17) {
pm_strcpy(jcr->errmsg, dir->msg);
dir->fsend(BAD_job, stat, jcr->errmsg);
Dmsg1(100, ">dird: %s", dir->msg);
jcr->setJobStatus(JS_ErrorTerminated);
return false;
}
Dmsg3(100, "==== rerunning=%d VolSesId=%d VolSesTime=%d\n", jcr->rerunning,
jcr->VolSessionId, jcr->VolSessionTime);
/*
* Since this job could be rescheduled, we
* check to see if we have it already. If so
* free the old jcr and use the new one.
*/
ojcr = get_jcr_by_full_name(job.c_str());
if (ojcr && !ojcr->authenticated) {
Dmsg2(100, "Found ojcr=0x%x Job %s\n", (unsigned)(intptr_t)ojcr, job.c_str());
free_jcr(ojcr);
}
jcr->JobId = JobId;
Dmsg2(800, "Start JobId=%d %p\n", JobId, jcr);
/*
* If job rescheduled because previous was incomplete,
* the Resched flag is set and VolSessionId and VolSessionTime
* are given to us (same as restarted job).
*/
if (!jcr->rerunning) {
jcr->VolSessionId = newVolSessionId();
jcr->VolSessionTime = VolSessionTime;
}
bstrncpy(jcr->Job, job, sizeof(jcr->Job));
unbash_spaces(job_name);
jcr->job_name = get_pool_memory(PM_NAME);
pm_strcpy(jcr->job_name, job_name);
unbash_spaces(client_name);
jcr->client_name = get_pool_memory(PM_NAME);
pm_strcpy(jcr->client_name, client_name);
unbash_spaces(fileset_name);
jcr->fileset_name = get_pool_memory(PM_NAME);
pm_strcpy(jcr->fileset_name, fileset_name);
jcr->setJobType(JobType);
jcr->setJobLevel(level);
jcr->no_attributes = no_attributes;
jcr->spool_attributes = spool_attributes;
jcr->spool_data = spool_data;
jcr->spool_size = str_to_int64(spool_size);
jcr->write_part_after_job = write_part_after_job;
jcr->fileset_md5 = get_pool_memory(PM_NAME);
pm_strcpy(jcr->fileset_md5, fileset_md5);
jcr->PreferMountedVols = PreferMountedVols;
jcr->authenticated = false;
/*
* Pass back an authorization key for the File daemon
*/
bsnprintf(seed, sizeof(seed), "%p%d", jcr, JobId);
make_session_key(auth_key, seed, 1);
dir->fsend(OKjob, jcr->VolSessionId, jcr->VolSessionTime, auth_key);
Dmsg2(50, ">dird jid=%u: %s", (uint32_t)jcr->JobId, dir->msg);
jcr->sd_auth_key = bstrdup(auth_key);
memset(auth_key, 0, sizeof(auth_key));
new_plugins(jcr); /* instantiate the plugins */
generate_daemon_event(jcr, "JobStart");
generate_plugin_event(jcr, bsdEventJobStart, (void *)"JobStart");
return true;
}
