The Kubernetes Signals I Check First

Debugging Kubernetes becomes easier when you ask the control plane questions in a fixed order. Start with desired state, move to events, and inspect logs only after you know which container matters.

The five-minute pass

kubectl get deploy,pod -n payments -o wide
kubectl get events -n payments --sort-by=.lastTimestamp
kubectl describe pod -n payments "$POD"
kubectl logs -n payments "$POD" --all-containers --prefix

Follow the ownership chain

flowchart TD
  Deployment --> ReplicaSet
  ReplicaSet --> Pod
  Pod --> Container
  Service --> Pod
  Ingress --> Service

Check whether selectors and labels connect the chain you intended. Many apparent network failures are actually empty service endpoints.

Preserve evidence

Before restarting a failing workload, collect the pod description, previous container logs, events, and the deployed manifest. Recovery without evidence guarantees a longer incident next time.