From 1a5966bc2b07f4a8f33edf7147a37a52ea7d1d26 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Mon, 31 Jul 2006 18:56:27 +0000
Subject: [PATCH] Don't treat lack of srun response to SRUN_JOB_COMPLETE as an
error for the node.
---
src/slurmctld/agent.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/src/slurmctld/agent.c b/src/slurmctld/agent.c
index bd35556b74f..107504224ab 100644
--- a/src/slurmctld/agent.c
+++ b/src/slurmctld/agent.c
@@ -757,6 +757,7 @@ static void *_thread_per_group_rpc(void *args)
is_kill_msg = ( (msg_type == REQUEST_KILL_TIMELIMIT) ||
(msg_type == REQUEST_TERMINATE_JOB) );
srun_agent = ( (msg_type == SRUN_PING) ||
+ (msg_type == SRUN_JOB_COMPLETE) ||
(msg_type == SRUN_TIMEOUT) ||
(msg_type == RESPONSE_RESOURCE_ALLOCATION) ||
(msg_type == SRUN_NODE_FAIL) );
@@ -770,7 +771,8 @@ static void *_thread_per_group_rpc(void *args)
enum job_states state = JOB_END;
struct job_record *job_ptr = NULL;
- if (msg_type == SRUN_PING) {
+ if ((msg_type == SRUN_PING)
+ || (msg_type == SRUN_JOB_COMPLETE)) {
srun_ping_msg_t *msg = task_ptr->msg_args_ptr;
job_id = msg->job_id;
} else if (msg_type == SRUN_TIMEOUT) {
--
GitLab