The NVIDIA GPU Operator requires GPU tolerations

The NVIDIA DaemonSets are unable to run their pods without specifying the tolerations given to the GPU nodes in the AcceleratorProfiles. Fixes nerc-project/operations#913
OCP-on-NERC · Jan 31, 2025 · 402b7c8 · 402b7c8
1 parent a6290fb
commit 402b7c8
Showing 1 changed file with 10 additions and 0 deletions.
diff --git a/nvidia-gpu-operator/overlays/nerc-ocp-test/clusterpolicy/clusterpolicy_patch.yaml b/nvidia-gpu-operator/overlays/nerc-ocp-test/clusterpolicy/clusterpolicy_patch.yaml
@@ -16,3 +16,13 @@ spec:
         value: 'false'
       - name: ACCEPT_NVIDIA_VISIBLE_DEVICES_AS_VOLUME_MOUNTS
         value: "true"
+  daemonsets:
+    tolerations:
+    - effect: NoSchedule
+      key: nvidia.com/gpu.product
+      operator: Equal
+      value: NVIDIA-A100-SXM4-40GB
+    - effect: NoSchedule
+      key: nvidia.com/gpu.product
+      operator: Equal
+      value: Tesla-V100-PCIE-32GB