From f2b1c35fb899bb9fad5187e95d01ed4463290afe Mon Sep 17 00:00:00 2001
From: Danny Auble <da@schedmd.com>
Date: Thu, 14 Jul 2016 16:55:50 -0700
Subject: [PATCH] Move commit 30f4f81c1f4a7b8 to be above code that could call
delete_step_records which would delete the steps without the killing flag
set.
---
src/slurmctld/node_scheduler.c | 31 ++++++++++++++++---------------
1 file changed, 16 insertions(+), 15 deletions(-)
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 91a50e6fb99..03fc0ae07cb 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -3726,6 +3726,22 @@ extern void re_kill_job(struct job_record *job_ptr)
job_ptr->spank_job_env);
kill_job->spank_job_env_size = job_ptr->spank_job_env_size;
+ /* On a Cray system this will start the NHC early so it is
+ * able to gather any information it can from the apparent
+ * unkillable processes.
+ * NOTE: do not do a list_for_each here, that will hold on the list
+ * lock while processing the entire list which could
+ * potentially be needed to lock again in
+ * select_g_step_finish which could potentially call
+ * post_job_step which calls delete_step_record which locks
+ * the list to create a list_iterator on the same list and
+ * could cause deadlock :).
+ */
+ step_iterator = list_iterator_create(job_ptr->step_list);
+ while ((step_ptr = list_next(step_iterator)))
+ select_g_step_finish(step_ptr, true);
+ list_iterator_destroy(step_iterator);
+
#ifdef HAVE_FRONT_END
if (job_ptr->batch_host &&
(front_end_ptr = find_front_end_record(job_ptr->batch_host))) {
@@ -3819,21 +3835,6 @@ extern void re_kill_job(struct job_record *job_ptr)
job_ptr->job_id, host_str);
}
#endif
- /* On a Cray system this will start the NHC early so it is
- * able to gather any information it can from the apparent
- * unkillable processes.
- * NOTE: do not do a list_for_each here, that will hold on the list
- * lock while processing the entire list which could
- * potentially be needed to lock again in
- * select_g_step_finish which could potentially call
- * post_job_step which calls delete_step_record which locks
- * the list to create a list_iterator on the same list and
- * could cause deadlock :).
- */
- step_iterator = list_iterator_create(job_ptr->step_list);
- while ((step_ptr = list_next(step_iterator)))
- select_g_step_finish(step_ptr, true);
- list_iterator_destroy(step_iterator);
xfree(host_str);
last_job_id = job_ptr->job_id;
--
GitLab