Skip to content

Symptoms

  • Requests to http://user-service.prod.svc.cluster.local:8080 from other pods return connection refused.
  • The Service user-service exists in the prod namespace and has a ClusterIP assigned.
  • kubectl get endpoints user-service -n prod shows no endpoints (empty subsets).
  • There are 3 healthy pods running for the user-service deployment, all in Running state with 1/1 READY.
  • The service was recently recreated by a junior engineer who wrote the manifest from scratch.
  • The application worked before the service was recreated.