From 9128476a45426dc0fa59449332a211dc3f81084c Mon Sep 17 00:00:00 2001
From: Tim Wickberg <tim@schedmd.com>
Date: Thu, 25 May 2017 10:49:00 -0600
Subject: [PATCH] Prevent a job tested on multiple partitions from being marked
WHOLE_NODE_USER.
If a job is considered on a partition with ExclusiveUser=YES
then it would be marked as if it was submitted with the
--exclusive flag, which would lead to delays launching it
on ExclusiveUser=NO partitions, and cause lower-than-expected
cluster usage.
As a side effect, the job_ptr->part_ptr->flags need to be
tested wherever WHOLE_NODE_USER is considered, instead of
just job_ptr->details->whole_node.
Bug 3771.
---
NEWS | 2 ++
src/slurmctld/node_mgr.c | 14 +++++++++-----
src/slurmctld/node_scheduler.c | 7 ++-----
src/slurmctld/read_config.c | 6 ++++--
4 files changed, 17 insertions(+), 12 deletions(-)
diff --git a/NEWS b/NEWS
index 081019150ce..9566a589e13 100644
--- a/NEWS
+++ b/NEWS
@@ -51,6 +51,8 @@ documents those changes that are of interest to users and administrators.
to 365-days (as is done elsewhere), rather than 49710 days.
-- Check if variable given to scontrol show job is a valid jobid.
-- Fix WithSubAccounts option to not include WithDeleted unless requested.
+ -- Prevent a job tested on multiple partitions from being marked
+ WHOLE_NODE_USER.
* Changes in Slurm 17.02.3
==========================
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index 34ba315b70b..b0f8025fdee 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -3583,8 +3583,10 @@ extern void make_node_alloc(struct node_record *node_ptr,
(node_ptr->no_share_job_cnt)++;
}
- if (job_ptr->details &&
- (job_ptr->details->whole_node == WHOLE_NODE_USER)) {
+ if ((job_ptr->details &&
+ (job_ptr->details->whole_node == WHOLE_NODE_USER)) ||
+ (job_ptr->part_ptr &&
+ (job_ptr->part_ptr->flags & PART_FLAG_EXCLUSIVE_USER))) {
node_ptr->owner_job_cnt++;
node_ptr->owner = job_ptr->user_id;
}
@@ -3787,9 +3789,11 @@ void make_node_idle(struct node_record *node_ptr,
xfree(node_ptr->mcs_label);
}
}
-
- if (job_ptr && job_ptr->details &&
- (job_ptr->details->whole_node == WHOLE_NODE_USER)) {
+ if (job_ptr &&
+ ((job_ptr->details &&
+ (job_ptr->details->whole_node == WHOLE_NODE_USER)) ||
+ (job_ptr->part_ptr &&
+ (job_ptr->part_ptr->flags & PART_FLAG_EXCLUSIVE_USER)))) {
if (--node_ptr->owner_job_cnt == 0) {
node_ptr->owner = NO_VAL;
xfree(node_ptr->mcs_label);
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 41f3fe18998..5c61cca0dac 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -900,11 +900,8 @@ extern void filter_by_node_owner(struct job_record *job_ptr,
struct node_record *node_ptr;
int i;
- if ((job_ptr->details->whole_node == 0) &&
- (job_ptr->part_ptr->flags & PART_FLAG_EXCLUSIVE_USER))
- job_ptr->details->whole_node = WHOLE_NODE_USER;
-
- if (job_ptr->details->whole_node == WHOLE_NODE_USER) {
+ if ((job_ptr->details->whole_node == WHOLE_NODE_USER) ||
+ (job_ptr->part_ptr->flags & PART_FLAG_EXCLUSIVE_USER)) {
/* Need to remove all nodes allocated to any active job from
* any other user */
job_iterator = list_iterator_create(job_list);
diff --git a/src/slurmctld/read_config.c b/src/slurmctld/read_config.c
index 4c831d695b0..a9461b3ef5e 100644
--- a/src/slurmctld/read_config.c
+++ b/src/slurmctld/read_config.c
@@ -2168,8 +2168,10 @@ static int _sync_nodes_to_active_job(struct job_record *job_ptr)
} else if (bit_test(job_ptr->node_bitmap, i) == 0)
continue;
- if (job_ptr->details &&
- (job_ptr->details->whole_node == WHOLE_NODE_USER)) {
+ if ((job_ptr->details &&
+ (job_ptr->details->whole_node == WHOLE_NODE_USER)) ||
+ (job_ptr->part_ptr &&
+ (job_ptr->part_ptr->flags & PART_FLAG_EXCLUSIVE_USER))) {
node_ptr->owner_job_cnt++;
node_ptr->owner = job_ptr->user_id;
}
--
GitLab