From 2c99c42a4fc1a570627099bd987b21baefcf6306 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Fri, 14 May 2010 15:29:25 +0000
Subject: [PATCH] Insure that we don't purge any jobs before testing after_ok,
not_ok dependencies upon that job
---
src/slurmctld/job_mgr.c | 30 +++++++++++++++++++++++++-----
src/slurmctld/slurmctld.h | 9 +++++----
2 files changed, 30 insertions(+), 9 deletions(-)
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 8c2bdd88fbe..213a5949ce5 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -4860,19 +4860,39 @@ static void _pack_pending_job_details(struct job_details *detail_ptr,
/*
* purge_old_job - purge old job records.
- * the jobs must have completed at least MIN_JOB_AGE minutes ago
- * global: job_list - global job table
- * last_job_update - time of last job table update
- * NOTE: READ lock_slurmctld config before entry
+ * The jobs must have completed at least MIN_JOB_AGE minutes ago.
+ * Test job dependencies, handle after_ok, after_not_ok before
+ * purging any jobs.
+ * NOTE: READ lock slurmctld config and WRITE lock jobs before entry
*/
void purge_old_job(void)
{
+ ListIterator job_iterator;
+ struct job_record *job_ptr;
+ time_t now = time(NULL);
int i;
+ job_iterator = list_iterator_create(job_list);
+ while ((job_ptr = (struct job_record *) list_next(job_iterator))) {
+ if (!IS_JOB_PENDING(job_ptr))
+ continue;
+ if (test_job_dependency(job_ptr) == 2) {
+ info("Job dependency can't be satisfied, cancelling "
+ "job %u", job_ptr->job_id);
+ job_ptr->job_state = JOB_CANCELLED;
+ xfree(job_ptr->state_desc);
+ job_ptr->start_time = now;
+ job_ptr->end_time = now;
+ job_completion_logger(job_ptr);
+ last_job_update = now;
+ }
+ }
+ list_iterator_destroy(job_iterator);
+
i = list_delete_all(job_list, &_list_find_job_old, "");
if (i) {
debug2("purge_old_job: purged %d old job records", i);
-/* last_job_update = time(NULL); don't worry about state save */
+/* last_job_update = now; don't worry about state save */
}
}
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index 4d7c90a8368..38853f1e270 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -1347,11 +1347,12 @@ extern void part_fini (void);
/*
* purge_old_job - purge old job records.
- * the jobs must have completed at least MIN_JOB_AGE minutes ago
- * global: job_list - global job table
- * last_job_update - time of last job table update
+ * The jobs must have completed at least MIN_JOB_AGE minutes ago.
+ * Test job dependencies, handle after_ok, after_not_ok before
+ * purging any jobs.
+ * NOTE: READ lock slurmctld config and WRITE lock jobs before entry
*/
-extern void purge_old_job (void);
+void purge_old_job(void);
/*
* rehash_jobs - Create or rebuild the job hash table.
--
GitLab