static int forced_jobpurge(
job *pjob,
struct batch_request *preq)
{
long owner_purge = FALSE;
/* check about possibly purging the job */
if (preq->rq_extend != NULL)
{
if (!strncmp(preq->rq_extend, delpurgestr, strlen(delpurgestr)))
{
get_svr_attr_l(SRV_ATR_OwnerPurge, &owner_purge);
if (((preq->rq_perm & (ATR_DFLAG_OPRD | ATR_DFLAG_OPWR | ATR_DFLAG_MGRD | ATR_DFLAG_MGWR)) != 0) ||
((svr_chk_owner(preq, pjob) == 0) && (owner_purge)))
{
force_purge_work(pjob);
return(PURGE_SUCCESS);
}
else
{
/* FAILURE */
req_reject(PBSE_PERM, 0, preq, NULL, NULL);
unlock_ji_mutex(pjob, __func__, "1", LOGLEVEL);
return(-1);
}
}
}
return(PBSE_NONE);
} /* END forced_jobpurge() */
int retry_job_exit(
job_exiting_retry_info *jeri)
{
char log_buf[LOCAL_LOG_BUF_SIZE];
job *pjob;
jeri->attempts++;
if (jeri->attempts >= MAX_EXITING_RETRY_ATTEMPTS)
{
/* job has been attempted the maximum number of times. Destroy the job */
if ((pjob = svr_find_job(jeri->jobid, TRUE)) != NULL)
{
force_purge_work(pjob);
}
remove_entry_from_exiting_list(jeri);
}
else
{
snprintf(log_buf, sizeof(log_buf), "Retrying job exiting for job %s",
jeri->jobid);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, __func__, log_buf);
jeri->last_attempt = time(NULL);
on_job_exit(NULL, strdup(jeri->jobid));
}
return(PBSE_NONE);
} /* END retry_job_exit() */
END_TEST
START_TEST(test_force_purge_work)
{
job *pjob = (job *)calloc(1, sizeof(job));
pjob->ji_wattr[JOB_ATR_exec_host].at_val.at_str = strdup("bob");
force_purge_work(pjob);
/* normally pjob wouldn't be valid at this point, but I've made the functions
* that free set these values in the scaffolding so we can test what happened */
fail_unless(pjob->ji_wattr[JOB_ATR_exec_host].at_val.at_str == NULL);
fail_unless(pjob->ji_qs.ji_state == JOB_STATE_COMPLETE);
}
char *get_next_retryable_jobid(
exiting_jobs_info_iterator **iter)
{
job_exiting_retry_info *jeri;
job *pjob;
time_t time_now = time(NULL);
char log_buf[LOCAL_LOG_BUF_SIZE];
exiting_jobs_info.lock();
if(*iter == NULL)
{
*iter = exiting_jobs_info.get_iterator();
}
while ((jeri = (*iter)->get_next_item()) != NULL)
{
if (time_now - jeri->last_attempt > EXITING_RETRY_TIME)
{
if (jeri->attempts >= MAX_EXITING_RETRY_ATTEMPTS)
{
std::string jid(jeri->jobid);
exiting_jobs_info.remove(jeri->jobid);
free(jeri);
exiting_jobs_info.unlock();
if ((pjob = svr_find_job((char *)jid.c_str(), TRUE)) != NULL)
{
snprintf(log_buf, sizeof(log_buf), "Job %s has had its exiting re-tried %d times, purging.",
jeri->jobid, MAX_EXITING_RETRY_ATTEMPTS);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, __func__, log_buf);
force_purge_work(pjob);
}
exiting_jobs_info.lock();
}
else
{
jeri->attempts++;
jeri->last_attempt = time_now;
exiting_jobs_info.unlock();
char *jobid = strdup(jeri->jobid);
return(jobid);
}
}
}
exiting_jobs_info.unlock();
return(NULL);
} /* END get_next_retryable_jobid() */
char *get_next_retryable_jobid(
int *iter)
{
job_exiting_retry_info *jeri;
job *pjob;
time_t time_now = time(NULL);
char log_buf[LOCAL_LOG_BUF_SIZE];
pthread_mutex_lock(exiting_jobs_info->hm_mutex);
mutex_mgr exit_mgr(exiting_jobs_info->hm_mutex, true);
while ((jeri = (job_exiting_retry_info *)next_from_hash_map(exiting_jobs_info, iter, true)) != NULL)
{
if (time_now - jeri->last_attempt > EXITING_RETRY_TIME)
{
if (jeri->attempts >= MAX_EXITING_RETRY_ATTEMPTS)
{
std::string jid(jeri->jobid);
remove_from_hash_map(exiting_jobs_info, jeri->jobid, true);
free(jeri);
exit_mgr.unlock(); //Don't hold on to a mutex when trying to lock another.
if ((pjob = svr_find_job((char *)jid.c_str(), TRUE)) != NULL)
{
snprintf(log_buf, sizeof(log_buf), "Job %s has had its exiting re-tried %d times, purging.",
jeri->jobid, MAX_EXITING_RETRY_ATTEMPTS);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, __func__, log_buf);
force_purge_work(pjob);
}
exit_mgr.lock();
}
else
{
jeri->attempts++;
jeri->last_attempt = time_now;
char *jobid = strdup(jeri->jobid);
return(jobid);
}
}
}
return(NULL);
} /* END get_next_retryable_jobid() */
void ensure_deleted(
struct work_task *ptask) /* I */
{
job *pjob;
char *jobid;
jobid = ptask->wt_parm1;
if (jobid != NULL)
{
if ((pjob = svr_find_job(jobid, FALSE)) != NULL)
{
force_purge_work(pjob);
}
}
free(jobid);
free(ptask->wt_mutex);
free(ptask);
} /* END ensure_deleted() */
