From 2a534a8c72e604cf20db32d37490a5bb4e9fbcca Mon Sep 17 00:00:00 2001
From: Jason Kincl <jkincl@redhat.com>
Date: Thu, 9 Nov 2023 16:36:13 -0500
Subject: [PATCH] Updating Slurm config

---
 slurm/manifests/kustomization.yaml       |  1 +
 slurm/manifests/priorityclass.yaml       |  7 +++++++
 slurm/manifests/statefulset-compute.yaml | 22 +++++++++++++++++++++-
 3 files changed, 29 insertions(+), 1 deletion(-)
 create mode 100644 slurm/manifests/priorityclass.yaml

diff --git a/slurm/manifests/kustomization.yaml b/slurm/manifests/kustomization.yaml
index 74be77a..aad6fed 100644
--- a/slurm/manifests/kustomization.yaml
+++ b/slurm/manifests/kustomization.yaml
@@ -11,6 +11,7 @@ resources:
   - serviceaccount.yaml
   - clusterrolebinding.yaml
   - claim.yaml
+  - priorityclass.yaml
 
   - servicemonitor.yaml
   - scaledobject.yaml
diff --git a/slurm/manifests/priorityclass.yaml b/slurm/manifests/priorityclass.yaml
new file mode 100644
index 0000000..613dd73
--- /dev/null
+++ b/slurm/manifests/priorityclass.yaml
@@ -0,0 +1,7 @@
+apiVersion: scheduling.k8s.io/v1
+kind: PriorityClass
+metadata:
+  name: high-priority-batch
+value: 1000000
+globalDefault: false
+description: "This priority class should be used for high priority batch pods only."
\ No newline at end of file
diff --git a/slurm/manifests/statefulset-compute.yaml b/slurm/manifests/statefulset-compute.yaml
index 360b385..c9e3e8c 100644
--- a/slurm/manifests/statefulset-compute.yaml
+++ b/slurm/manifests/statefulset-compute.yaml
@@ -35,6 +35,7 @@ spec:
         app: slurm
         component: compute
     spec:
+      priorityClassName: high-priority-batch
       automountServiceAccountToken: false
       shareProcessNamespace: true
       dnsConfig:
@@ -43,6 +44,20 @@ spec:
       enableServiceLinks: false
       terminationGracePeriodSeconds: 15
       serviceAccountName: slurm
+      affinity:
+        podAntiAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+          - labelSelector:
+              matchExpressions:
+              - key: app
+                operator: In
+                values:
+                - slurm
+              - key: component
+                operator: In
+                values:
+                - compute
+            topologyKey: "kubernetes.io/hostname"
       initContainers:
       - name: fix-munge-paths
         image: ghcr.io/naps-product-sa/openshift-batch/munge:latest
@@ -74,7 +89,12 @@ spec:
               - /bin/bash
               - -c
               - |
-                scontrol update nodename=$HOSTNAME state=down reason=shutdown
+                scontrol delete nodename=$HOSTNAME
+        resources:
+          requests:
+            cpu: 3
+          limits:
+            cpu: 3
         securityContext:
           runAsUser: 0
           runAsGroup: 0