Add GPU runner nodes (#4982)

pytorch · Mar 4, 2024 · 09a4723 · 09a4723
1 parent 8c37cbf
commit 09a4723
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 3 deletions.
diff --git a/.github/arc-node-config.yaml b/.github/arc-node-config.yaml
@@ -16,6 +16,23 @@ nodeConfig:
       - key: kubernetes.io/os
         operator: In
         values: ["linux"]
-      # - key: node.kubernetes.io/instance-type
-      #   operator: In
-      #   values: ["p3.8xlarge", "p3.16xlarge"]
+  - nodeType: compute-gpu
+    requirements:
+      - key: "karpenter.k8s.aws/instance-category"
+        operator: In
+        values: ["p"]
+      - key: "karpenter.k8s.aws/instance-cpu"
+        operator: In
+        values: ["16", "32", "64"]
+      - key: "kubernetes.io/arch"
+        operator: In
+        values: ["amd64"]
+      - key: "karpenter.sh/capacity-type"
+        operator: In
+        values: ["spot", "on-demand"]
+      - key: kubernetes.io/os
+        operator: In
+        values: ["linux"]
+      - key: node.kubernetes.io/instance-type
+        operator: In
+        values: ["p3.8xlarge", "p3.16xlarge"]
diff --git a/.github/arc-runner-config.yaml b/.github/arc-runner-config.yaml
@@ -59,3 +59,18 @@ runnerConfig:
     scaleUpFactor: 1.20
     scaleUpThreshold: 0.80
     containerMode: k8s
+  - cpu: 4.0
+    isEphemeral: true
+    maxRunners: 150
+    memory: 16Gi
+    minRunners: 1
+    nodeType: compute-gpu
+    partDocker: 20Gi
+    partTmp: 50Gi
+    partWork: 150Gi
+    runnerLabel: linux-gpu.large.arc
+    scaleDownFactor: 0.80
+    scaleDownThreshold: 0.60
+    scaleUpFactor: 1.20
+    scaleUpThreshold: 0.80
+    containerMode: dind-rootless