From 813ec4c57d68f5ef71b77db4276e68507995ef48 Mon Sep 17 00:00:00 2001
From: Morris Jette <jette@schedmd.com>
Date: Sat, 31 Dec 2016 13:13:52 -0700
Subject: [PATCH] Revise salloc abort logic
Revision to 06fbdd751c4d4a7343c8f78ea63d8a52869f5fb0
Previous logic was causing tests 1.87 and 15.19 to fail sometimes
due to a race condition by adding a "Job allocation revoked"
message to salloc output even if no allocation occured.
bug 3351
---
src/salloc/salloc.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/src/salloc/salloc.c b/src/salloc/salloc.c
index 45bea337e25..35d4e6663e5 100644
--- a/src/salloc/salloc.c
+++ b/src/salloc/salloc.c
@@ -109,6 +109,7 @@ pthread_mutex_t allocation_state_lock = PTHREAD_MUTEX_INITIALIZER;
static bool exit_flag = false;
static bool suspend_flag = false;
static bool allocation_interrupted = false;
+static bool allocation_revoked = false;
static uint32_t pending_job_id = 0;
static time_t last_timeout = 0;
static struct termios saved_tty_attributes;
@@ -891,7 +892,7 @@ static void _job_complete_handler(srun_job_complete_msg_t *comp)
} else {
info("Job allocation %u has been revoked.",
comp->job_id);
- allocation_interrupted = true;
+ allocation_revoked = true;
}
}
allocation_state = REVOKED;
@@ -1059,7 +1060,7 @@ static int _wait_bluegene_block_ready(resource_allocation_response_msg_t *alloc)
is_ready = 1;
break;
}
- if (allocation_interrupted)
+ if (allocation_interrupted || allocation_revoked)
break;
}
if (is_ready)
@@ -1164,7 +1165,7 @@ static int _wait_nodes_ready(resource_allocation_response_msg_t *alloc)
is_ready = 1;
break;
}
- if (allocation_interrupted)
+ if (allocation_interrupted || allocation_revoked)
break;
}
if (is_ready) {
--
GitLab