From 919b1b9892711bcf726ee657134aee3764ec02cb Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Tue, 11 May 2004 17:41:43 +0000
Subject: [PATCH] Better deal with KillWait==0. Don't immediately kill a job
after a timeout RPC is send (possible race condition in background threads).
---
src/slurmctld/job_mgr.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index f198bd71776..8314bf459f4 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -2110,6 +2110,7 @@ void job_time_limit(void)
continue;
}
+ /* Give srun command warning message about pending timeout */
if (job_ptr->end_time <= (now + 60))
srun_timeout (job_ptr->job_id, job_ptr->end_time);
@@ -2300,7 +2301,7 @@ static int _list_find_job_id(void *job_entry, void *key)
static int _list_find_job_old(void *job_entry, void *key)
{
time_t now = time(NULL);
- time_t kill_age = now - (slurmctld_conf.kill_wait * 2);
+ time_t kill_age = now - (slurmctld_conf.kill_wait + 20);
time_t min_age = now - slurmctld_conf.min_job_age;
struct job_record *job_ptr = (struct job_record *)job_entry;
--
GitLab