From 73c2e82c1cea857785aa8a7ab224b51011943e15 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Mon, 4 Dec 2006 17:09:51 +0000
Subject: [PATCH] Validate product of socket*core*thread count on node
registration rather than individual values. Correct values will need to
be specified in slurm.conf with FastSchedule=1 for correct multi-core
scheduling behavior.
---
NEWS | 3 +++
src/slurmctld/node_mgr.c | 21 +++++++++++++++++++++
2 files changed, 24 insertions(+)
diff --git a/NEWS b/NEWS
index 8a03efa9f25..fe91b8a7c88 100644
--- a/NEWS
+++ b/NEWS
@@ -5,6 +5,9 @@ documents those changes that are of interest to users and admins.
=============================
-- Fix for select/cons_res state preservation over slurmctld restart,
patch.1.2.0-pre7.061130.cr_state from Dan Palermo.
+ -- Validate product of socket*core*thread count on node registration rather
+ than individual values. Correct values will need to be specified in slurm.conf
+ with FastSchedule=1 for correct multi-core scheduling behavior.
* Changes in SLURM 1.2.0-pre8
=============================
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index d8c1cc8961a..9bda8135ce7 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -1170,6 +1170,10 @@ validate_node_specs (char *node_name, uint16_t cpus,
config_ptr = node_ptr->config_ptr;
error_code = 0;
+#if 0
+ /* Testing the socket, core, and thread count here can produce
+ * an error if the user did not specify the values on slurm.conf
+ * for a multi-core system */
if ((slurmctld_conf.fast_schedule != 2)
&& (sockets < config_ptr->sockets)) {
error("Node %s has low socket count %u", node_name, sockets);
@@ -1193,6 +1197,23 @@ validate_node_specs (char *node_name, uint16_t cpus,
reason_down = "Low thread count";
}
node_ptr->threads = threads;
+#else
+ if (slurmctld_conf.fast_schedule != 2) {
+ int tot1, tot2;
+ tot1 = sockets * cores * threads;
+ tot2 = config_ptr->sockets * config_ptr->cores *
+ config_ptr->threads;
+ if (tot1 < tot2) {
+ error("Node %s has low socket*core*thread count %u",
+ node_name, tot1);
+ error_code = EINVAL;
+ reason_down = "Low socket*core*thread count";
+ }
+ }
+ node_ptr->sockets = sockets;
+ node_ptr->cores = cores;
+ node_ptr->threads = threads;
+#endif
if ((slurmctld_conf.fast_schedule != 2)
&& (cpus < config_ptr->cpus)) {
--
GitLab