Add toleration to make sure the apps run on the gpu worker

manifests/ollama/values.yaml

@@ -20,6 +20,13 @@ resources:
 nodeSelector:
   kubernetes.io/hostname: k3s-worker-5-gpu
 
+# Tolerate GPU node taint
+tolerations:
+  - key: gpu
+    operator: Equal
+    value: "true"
+    effect: NoSchedule
+
 persistentVolume:
   enabled: true
   size: 30Gi

manifests/open-webui/values.yaml

@@ -23,3 +23,10 @@ resources:
 # Schedule on GPU worker node (same as Ollama for low latency)
 nodeSelector:
   kubernetes.io/hostname: k3s-worker-5-gpu
+
+# Tolerate GPU node taint
+tolerations:
+  - key: gpu
+    operator: Equal
+    value: "true"
+    effect: NoSchedule