diff --git a/clusters/k3s-dgx/nim-service/qwen.yaml b/clusters/k3s-dgx/nim-service/qwen.yaml index f07e0f7..0e985ce 100644 --- a/clusters/k3s-dgx/nim-service/qwen.yaml +++ b/clusters/k3s-dgx/nim-service/qwen.yaml @@ -94,35 +94,41 @@ spec: limits: nvidia.com/gpu: 1 livenessProbe: - httpGet: - path: /health - port: 8000 - initialDelaySeconds: 120 - periodSeconds: 30 - timeoutSeconds: 10 - failureThreshold: 10 + enabled: true + probe: + httpGet: + path: /health + port: 8000 + initialDelaySeconds: 120 + periodSeconds: 30 + timeoutSeconds: 10 + failureThreshold: 10 readinessProbe: - httpGet: - path: /health - port: 8000 - initialDelaySeconds: 30 - periodSeconds: 15 - timeoutSeconds: 10 - failureThreshold: 20 + enabled: true + probe: + httpGet: + path: /health + port: 8000 + initialDelaySeconds: 30 + periodSeconds: 15 + timeoutSeconds: 10 + failureThreshold: 20 startupProbe: - httpGet: - path: /health - port: 8000 - initialDelaySeconds: 10 - periodSeconds: 20 - timeoutSeconds: 10 - failureThreshold: 60 - expose: - service: - type: ClusterIP - port: 8000 + enabled: true + probe: + httpGet: + path: /health + port: 8000 + initialDelaySeconds: 10 + periodSeconds: 20 + timeoutSeconds: 10 + failureThreshold: 60 + expose: + service: + type: ClusterIP + port: 8000 --- apiVersion: gateway.networking.k8s.io/v1 kind: HTTPRoute