From 58c211406d5182c6f188adcb57758b09d7265d5f Mon Sep 17 00:00:00 2001
From: jette <jette@schedmd.com>
Date: Tue, 4 Jun 2013 09:58:43 -0700
Subject: [PATCH] launch/poe - Fix for hostlist file support with repeated host
names.
Without this change, it appears that POE ignores the -procs argument
resulting in a job step request with multiple host names, but only
one ntask required
---
NEWS | 1 +
src/plugins/switch/nrt/libpermapi/shr_64.c | 4 ++++
2 files changed, 5 insertions(+)
diff --git a/NEWS b/NEWS
index dc5fc0b0de5..f509c95c0c2 100644
--- a/NEWS
+++ b/NEWS
@@ -41,6 +41,7 @@ documents those changes that are of interest to users and admins.
-- Select/cons_res - Fix bug resulting in error of "cons_res: sync loop not
progressing, holding job #"
-- checkpoint/blcr - Reset max_nodes from zero to NO_VAL on job restart.
+ -- launch/poe - Fix for hostlist file support with repeated host names.
* Changes in Slurm 2.5.6
========================
diff --git a/src/plugins/switch/nrt/libpermapi/shr_64.c b/src/plugins/switch/nrt/libpermapi/shr_64.c
index f4cc258351f..9d25cc13b2b 100644
--- a/src/plugins/switch/nrt/libpermapi/shr_64.c
+++ b/src/plugins/switch/nrt/libpermapi/shr_64.c
@@ -1844,6 +1844,7 @@ int pe_rm_submit_job(rmhandle_t resource_mgr, job_command_t job_cmd,
&& pe_job_req->host_names && *pe_job_req->host_names) {
/* This means there was a hostfile used for this job.
* So we need to set up the arbitrary distribution of it. */
+ int hostfile_count = 0;
char **names = pe_job_req->host_names;
opt.distribution = SLURM_DIST_ARBITRARY;
while (names && *names) {
@@ -1853,7 +1854,10 @@ int pe_rm_submit_job(rmhandle_t resource_mgr, job_command_t job_cmd,
else
opt.nodelist = xstrdup(strtok(*names, "."));
names++;
+ hostfile_count++;
}
+ if (pe_job_req->total_tasks == -1)
+ pe_job_req->total_tasks = hostfile_count;
}
if (pe_job_req->num_nodes != -1)
--
GitLab