From 08b5678380abee81a57f6b4e5ee6e83ec7861c39 Mon Sep 17 00:00:00 2001
From: alejluther <alucero@os3sl.com>
Date: Tue, 15 May 2012 16:20:39 +0300
Subject: [PATCH] Avoiding a slurmctld crash when scheduling problems due to
resources. Setting an ADMIN hold instead.
---
src/plugins/select/cons_res/dist_tasks.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/plugins/select/cons_res/dist_tasks.c b/src/plugins/select/cons_res/dist_tasks.c
index f8b23adc2c6..d56af783f0d 100644
--- a/src/plugins/select/cons_res/dist_tasks.c
+++ b/src/plugins/select/cons_res/dist_tasks.c
@@ -499,7 +499,11 @@ static void _cyclic_sync_core_bitmap(struct job_record *job_ptr,
}
if (prev_cpus == cpus) {
/* we're stuck! */
- fatal("cons_res: sync loop not progressing");
+ job_ptr->priority = 0;
+ job_ptr->state_reason = WAIT_HELD;
+ error("cons_res: sync loop not progressing, "
+ "job %u", job_ptr->job_id);
+ return -1;
}
}
/* clear the rest of the cores in each socket
--
GitLab