From db36b0f0adabe232c0e68649ac8c97ca161effd4 Mon Sep 17 00:00:00 2001
From: Morris Jette <jette@schedmd.com>
Date: Tue, 3 Dec 2013 08:30:20 -0800
Subject: [PATCH] Make sure batch complete RPC has node name
This is a correction in the logic of commit
3f4b2d51fcdc2178d67739fbfbcebc5ed1a75872 on launch failures.
---
src/slurmctld/proc_req.c | 2 +-
src/slurmd/slurmd/req.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index 0f0cbd732ca..c112c17b801 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -1703,7 +1703,7 @@ static void _slurm_rpc_complete_batch_script(slurm_msg_t * msg)
lock_slurmctld(job_write_lock);
job_ptr = find_job_record(comp_msg->job_id);
- if (job_ptr && job_ptr->batch_host &&
+ if (job_ptr && job_ptr->batch_host && comp_msg->node_name &&
strcmp(job_ptr->batch_host, comp_msg->node_name)) {
/* This can be the result of the slurmd on the batch_host
* failing, but the slurmstepd continuing to run. Then the
diff --git a/src/slurmd/slurmd/req.c b/src/slurmd/slurmd/req.c
index 5c95d9465ee..b5d4b1b84e5 100644
--- a/src/slurmd/slurmd/req.c
+++ b/src/slurmd/slurmd/req.c
@@ -1637,7 +1637,7 @@ _abort_job(uint32_t job_id, uint32_t slurm_rc)
resp.job_id = job_id;
resp.job_rc = 1;
resp.slurm_rc = slurm_rc;
- resp.node_name = NULL; /* unused */
+ resp.node_name = conf->node_name;
resp.jobacct = NULL; /* unused */
resp_msg.msg_type = REQUEST_COMPLETE_BATCH_SCRIPT;
resp_msg.data = &resp;
--
GitLab