From f915ef34783777e3604ca6433c77394dba17a0cc Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Thu, 9 Nov 2006 20:01:24 +0000
Subject: [PATCH] When a job epilog completes on all non-DOWN nodes,
immediately purge it's job steps that lack switch windows. Needed for LSF
operation. Based upon slurm.hp.node_fail.patch.
---
NEWS | 3 +++
src/slurmctld/job_mgr.c | 5 +++--
src/slurmctld/node_scheduler.c | 1 +
src/slurmctld/slurmctld.h | 9 ++++++---
src/slurmctld/step_mgr.c | 14 ++++++++++----
5 files changed, 23 insertions(+), 9 deletions(-)
diff --git a/NEWS b/NEWS
index d411524a37e..1ce084fc928 100644
--- a/NEWS
+++ b/NEWS
@@ -14,6 +14,9 @@ documents those changes that are of interest to users and admins.
-- Race condition for forwarding logic fix from Hongjia Cao
-- Add support for Portable Linux Processor Affinity (PLPA, see
http://www.open-mpi.org/software/plpa).
+ -- When a job epilog completes on all non-DOWN nodes, immediately purge
+ it's job steps that lack switch windows. Needed for LSF operation.
+ Based upon slurm.hp.node_fail.patch.
-- Modify srun to ignore entries on --nodelist for job step creation
if their count exceeds the task count. Based on slurm.hp.srun.patch.
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 46049981b43..835e0853f41 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -999,7 +999,8 @@ extern int kill_running_job_by_node_name(char *node_name, bool step_test)
error("node_cnt underflow on JobId=%u",
job_ptr->job_id);
if (job_ptr->node_cnt == 0) {
- job_ptr->job_state &= (~JOB_COMPLETING);
+ job_ptr->job_state &= (~JOB_COMPLETING);\
+ delete_step_records(job_ptr, 1);
slurm_sched_schedule();
}
if (node_ptr->comp_job_cnt)
@@ -2659,7 +2660,7 @@ static void _list_delete_job(void *job_entry)
xfree(job_ptr->comment);
select_g_free_jobinfo(&job_ptr->select_jobinfo);
if (job_ptr->step_list) {
- delete_all_step_records(job_ptr);
+ delete_step_records(job_ptr, 0);
list_destroy(job_ptr->step_list);
}
job_count--;
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 045c0dcdcc7..28972b6de66 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -216,6 +216,7 @@ extern void deallocate_nodes(struct job_record *job_ptr, bool timeout,
if ((agent_args->node_count - down_node_cnt) == 0) {
job_ptr->job_state &= (~JOB_COMPLETING);
+ delete_step_records(job_ptr, 1);
slurm_sched_schedule();
}
if (agent_args->node_count == 0) {
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index bb03003b7de..eeac384e02e 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -496,10 +496,13 @@ extern struct part_record *create_part_record (void);
extern struct step_record * create_step_record (struct job_record *job_ptr);
/*
- * delete_all_step_records - delete all step record for specified job_ptr
- * IN job_ptr - pointer to job table entry to have step record added
+ * delete_step_records - delete step record for specified job_ptr
+ * IN job_ptr - pointer to job table entry to have step records removed
+ * IN filter - determine which job steps to delete
+ * 0: delete all job steps
+ * 1: delete only job steps without a switch allocation
*/
-extern void delete_all_step_records (struct job_record *job_ptr);
+extern void delete_step_records (struct job_record *job_ptr, int filter);
/*
* delete_job_details - delete a job's detail record and clear it's pointer
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index 1e428108101..154b6e14f18 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -104,11 +104,14 @@ create_step_record (struct job_record *job_ptr)
/*
- * delete_all_step_records - delete all step record for specified job_ptr
- * IN job_ptr - pointer to job table entry to have step record added
+ * delete_step_records - delete step record for specified job_ptr
+ * IN job_ptr - pointer to job table entry to have step records removed
+ * IN filter - determine which job steps to delete
+ * 0: delete all job steps
+ * 1: delete only job steps without a switch allocation
*/
-void
-delete_all_step_records (struct job_record *job_ptr)
+extern void
+delete_step_records (struct job_record *job_ptr, int filter)
{
ListIterator step_iterator;
struct step_record *step_ptr;
@@ -118,6 +121,9 @@ delete_all_step_records (struct job_record *job_ptr)
last_job_update = time(NULL);
while ((step_ptr = (struct step_record *) list_next (step_iterator))) {
+ if ((filter == 1) && (step_ptr->switch_job))
+ continue;
+
list_remove (step_iterator);
if (step_ptr->switch_job) {
switch_g_job_step_complete(
--
GitLab