Question

A backend application runs in pods that listen on container port 8080. A ClusterIP Service named 'api-svc' is defined with port 80 and targetPort 80. The pods are Running, readiness probes pass, and 'kubectl get endpoints api-svc' shows three healthy pod IPs listed. However, other pods that curl 'http://api-svc:80' consistently receive 'connection refused'. What is the most likely cause?

Accepted Answer

The Service targetPort (80) does not match the port the application listens on (8080), so traffic is forwarded to a closed port on the pods.

Answer

CoreDNS is failing to resolve 'api-svc', causing the connection to be refused before any packet is sent.

Answer

The Service selector does not match the pod labels, so no endpoints are populated.

Answer

kube-proxy is not running on the client node, so the ClusterIP virtual IP cannot be reached.