Add runtimeClassName support for GPU workloads

README.md

@@ -211,11 +211,13 @@ Key values in `chart/values.yaml`:
 | `image.tag` | `2.43.0` | Ray version |
 | `head.resources.requests.cpu` | `1` | Head node CPU request |
 | `head.resources.requests.memory` | `2Gi` | Head node memory request |
+| `head.runtimeClassName` | - | Runtime class for head pod (e.g., `nvidia` for GPU) |
 | `worker.replicas` | `1` | Number of worker nodes |
 | `worker.minReplicas` | `1` | Min workers (for autoscaling) |
 | `worker.maxReplicas` | `1` | Max workers (for autoscaling) |
 | `worker.resources.requests.cpu` | `1` | Worker CPU request |
 | `worker.resources.requests.memory` | `2Gi` | Worker memory request |
+| `worker.runtimeClassName` | - | Runtime class for worker pods (e.g., `nvidia` for GPU) |
 
 ### Serve Applications
 

chart/templates/rayservice.yaml

@@ -23,6 +23,9 @@ spec:
         dashboard-host: "0.0.0.0"
       template:
         spec:
+          {{- with .Values.head.runtimeClassName }}
+          runtimeClassName: {{ . }}
+          {{- end }}
           containers:
             - name: ray-head
               image: "{{ .Values.image.repository }}:{{ .Values.image.tag }}"
@@ -49,6 +52,9 @@ spec:
         rayStartParams: {}
         template:
           spec:
+            {{- with .Values.worker.runtimeClassName }}
+            runtimeClassName: {{ . }}
+            {{- end }}
             containers:
               - name: ray-worker
                 image: "{{ .Values.image.repository }}:{{ .Values.image.tag }}"

chart/values.yaml

@@ -75,6 +75,7 @@ image:
 # Ray Head
 # =============================================================================
 head:
+  # runtimeClassName: nvidia  # Set for GPU-enabled head nodes
   containerEnv: []
   resources:
     limits:
@@ -91,6 +92,7 @@ worker:
   replicas: 1
   minReplicas: 1
   maxReplicas: 1
+  # runtimeClassName: nvidia  # Set for GPU-enabled worker nodes
   containerEnv: []
   resources:
     limits: