From f17764718e11f4fc18b30a84f2d5b4273fe6e617 Mon Sep 17 00:00:00 2001
From: Brian Christiansen <brian@schedmd.com>
Date: Wed, 31 Jan 2018 16:43:11 -0700
Subject: [PATCH] Revert "When submitting a --test-only job respect the -M
option."
This reverts commit 516b0d598fb22ce9f00265577dc14771cfba41ab.
With the fixing of the NEWS file.
We want to keep the idea of only checking one federation.
---
NEWS | 1 -
src/api/allocate.c | 12 +-----------
src/common/slurmdb_defs.c | 11 +++++++++++
3 files changed, 12 insertions(+), 12 deletions(-)
diff --git a/NEWS b/NEWS
index 24817d80622..f2fd9cf1e10 100644
--- a/NEWS
+++ b/NEWS
@@ -62,7 +62,6 @@ documents those changes that are of interest to users and administrators.
job container where if the step was canceled would also cancel the stepd
erroneously.
-- Make sure the slurmstepd blocks signals like SIGTERM correctly.
- -- When submitting a --test-only job respect the -M option.
-- SPANK - When slurm_spank_init_post_opt() fails return error correctly.
-- When revoking a sibling job in the federation we want to send a start
message before purging the job record to get the uid of the revoked job.
diff --git a/src/api/allocate.c b/src/api/allocate.c
index f28cdb84c2a..d3a45fc4f15 100644
--- a/src/api/allocate.c
+++ b/src/api/allocate.c
@@ -604,17 +604,7 @@ int slurm_job_will_run(job_desc_msg_t *req)
cluster_name = working_cluster_rec->name;
else
cluster_name = slurmctld_conf.cluster_name;
-
- /*
- * If clusters is defined then slurmdb_get_first_avail_cluster() has
- * already been called and figured out the fastest cluster and a
- * will_run to the cluster is all that is needed. However if specific
- * clusters have been requested and the local cluster is in a federated
- * then check all of the clusters in the federation for the fastest
- * cluster.
- */
- if (!req->clusters &&
- !slurm_load_federation(&ptr) &&
+ if (!slurm_load_federation(&ptr) &&
cluster_in_federation(ptr, cluster_name))
rc = _fed_job_will_run(req, &will_run_resp, ptr);
else
diff --git a/src/common/slurmdb_defs.c b/src/common/slurmdb_defs.c
index 54cf55f6fa3..fbf8dd6e0ec 100644
--- a/src/common/slurmdb_defs.c
+++ b/src/common/slurmdb_defs.c
@@ -3006,6 +3006,7 @@ extern int slurmdb_get_first_avail_cluster(job_desc_msg_t *req,
ListIterator itr;
List cluster_list = NULL;
List ret_list = NULL;
+ List tried_feds = list_create(NULL);
*cluster_rec = NULL;
cluster_list = slurmdb_get_info_cluster(cluster_names);
@@ -3030,14 +3031,24 @@ extern int slurmdb_get_first_avail_cluster(job_desc_msg_t *req,
ret_list = list_create(_destroy_local_cluster_rec);
itr = list_iterator_create(cluster_list);
while ((working_cluster_rec = list_next(itr))) {
+ /* only try one cluster from each federation */
+ if (working_cluster_rec->fed.id &&
+ list_find_first(tried_feds, slurm_find_char_in_list,
+ working_cluster_rec->fed.name))
+ continue;
+
if ((local_cluster = _job_will_run(req))) {
list_append(ret_list, local_cluster);
+ if (working_cluster_rec->fed.id)
+ list_append(tried_feds,
+ working_cluster_rec->fed.name);
} else {
error("Problem with submit to cluster %s: %m",
working_cluster_rec->name);
}
}
list_iterator_destroy(itr);
+ FREE_NULL_LIST(tried_feds);
/* restore working_cluster_rec in case it was already set */
if (*cluster_rec) {
--
GitLab