Add noew selector to make sure the apps run on the gpu worker'

manifests/ollama/values.yaml

@@ -1,6 +1,13 @@
 service:
   type: ClusterIP
 
+# Enable GPU support
+ollama:
+  gpu:
+    enabled: true
+    type: nvidia
+    number: 1
+
 resources:
   requests:
     cpu: 2
@@ -9,6 +16,10 @@ resources:
     cpu: 4
     memory: 8Gi
 
+# Schedule on GPU worker node
+nodeSelector:
+  kubernetes.io/hostname: k3s-worker-5-gpu
+
 persistentVolume:
   enabled: true
   size: 30Gi

manifests/open-webui/values.yaml

@@ -19,3 +19,7 @@ resources:
   limits:
     cpu: 500m
     memory: 1Gi
+
+# Schedule on GPU worker node (same as Ollama for low latency)
+nodeSelector:
+  kubernetes.io/hostname: k3s-worker-5-gpu