Add condition to helm chart to enable computeDomains and GPUs separately

For now, enabling GPUs is not supported (even though it's the default). We error out if the GPU resource is enabled with instructions on how to disable it. Signed-off-by: Kevin Klues <[email protected]>
NVIDIA · Feb 6, 2025 · 7a33e9a · 7a33e9a
1 parent e9c72fc
commit 7a33e9a
Show file tree

Hide file tree

Showing 6 changed files with 22 additions and 0 deletions.
diff --git a/demo/clusters/gke/install-dra-driver-gpu.sh b/demo/clusters/gke/install-dra-driver-gpu.sh
@@ -36,6 +36,7 @@ helm upgrade -i --create-namespace --namespace nvidia nvidia-dra-driver-gpu ${PR
   --set controller.priorityClassName="" \
   --set kubeletPlugin.priorityClassName="" \
   --set nvidiaDriverRoot="/opt/nvidia" \
+  --set resources.gpus.enabled=false \
   --set kubeletPlugin.tolerations[0].key=nvidia.com/gpu \
   --set kubeletPlugin.tolerations[0].operator=Exists \
   --set kubeletPlugin.tolerations[0].effect=NoSchedule
diff --git a/demo/clusters/kind/install-dra-driver-gpu.sh b/demo/clusters/kind/install-dra-driver-gpu.sh
@@ -28,6 +28,7 @@ helm upgrade -i --create-namespace --namespace nvidia nvidia-dra-driver-gpu ${PR
     ${NVIDIA_CTK_PATH:+--set nvidiaCtkPath=${NVIDIA_CTK_PATH}} \
     ${NVIDIA_DRIVER_ROOT:+--set nvidiaDriverRoot=${NVIDIA_DRIVER_ROOT}} \
     ${MASK_NVIDIA_DRIVER_PARAMS:+--set maskNvidiaDriverParams=${MASK_NVIDIA_DRIVER_PARAMS}} \
+    --set resources.gpus.enabled=false \
     --wait
 
 set +x

diff --git a/deployments/helm/nvidia-dra-driver-gpu/templates/controller.yaml b/deployments/helm/nvidia-dra-driver-gpu/templates/controller.yaml
@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+{{- if .Values.resources.computeDomains.enabled }}
 ---
 apiVersion: apps/v1
 kind: Deployment
@@ -74,3 +75,4 @@ spec:
       tolerations:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+{{- end }}
diff --git a/deployments/helm/nvidia-dra-driver-gpu/templates/kubeletplugin.yaml b/deployments/helm/nvidia-dra-driver-gpu/templates/kubeletplugin.yaml
@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+{{- if or .Values.resources.computeDomains.enabled .Values.resources.gpus.enabled }}
 ---
 apiVersion: apps/v1
 kind: DaemonSet
@@ -48,6 +49,7 @@ spec:
       securityContext:
         {{- toYaml .Values.kubeletPlugin.podSecurityContext | nindent 8 }}
       containers:
+      {{- if .Values.resources.computeDomains.enabled }}
       - name: compute-domain
         securityContext:
           {{- toYaml .Values.kubeletPlugin.containers.computeDomain.securityContext | nindent 10 }}
@@ -80,6 +82,7 @@ spec:
         - name: driver-root
           mountPath: /driver-root
           readOnly: true
+      {{- end }}
       volumes:
       - name: plugins-registry
         hostPath:
@@ -105,3 +108,4 @@ spec:
       tolerations:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+{{- end }}
diff --git a/deployments/helm/nvidia-dra-driver-gpu/templates/validation.yaml b/deployments/helm/nvidia-dra-driver-gpu/templates/validation.yaml
@@ -29,3 +29,11 @@
 {{- $error = printf "%s\nSee: https://helm.sh/docs/helm/helm_install/#options" $error }}
 {{- fail $error }}
 {{- end }}
+
+{{- if .Values.resources.gpus.enabled }}
+{{- $error := "" }}
+{{- $error = printf "%s\nThe default value of 'resources.gpus.enabled=true' is not yet supported." $error }}
+{{- $error = printf "%s\nIt is set to true by default to future proof it as the default once support for it becomes available." $error }}
+{{- $error = printf "%s\nUntil then, please explicitly set 'resources.gpus.enabled=false' when installing this chart." $error }}
+{{- fail $error }}
+{{- end }}
diff --git a/deployments/helm/nvidia-dra-driver-gpu/values.yaml b/deployments/helm/nvidia-dra-driver-gpu/values.yaml
@@ -50,6 +50,12 @@ serviceAccount:
   # If not set and create is true, a name is generated using the fullname template
   name: ""
 
+resources:
+  gpus:
+    enabled: true
+  computeDomains:
+    enabled: true
+
 controller:
   priorityClassName: "system-node-critical"
   podAnnotations: {}