tools-only Kubernetes Troubleshooting
Debugging Workflow
# 1. Overview
kubectl get pods -o wide
kubectl get events -n <namespace> --sort-by='.lastTimestamp'
# 2. Details
kubectl describe pod <pod-name>
# 3. Logs
kubectl logs <pod-name>
kubectl logs <pod-name> --previous # Crashed instance
kubectl logs <pod-name> -c <container>
Common Pod States
| State | Cause | Solution |
|---|---|---|
| Pending | No node resources | Check node capacity |
| ContainerCreating | Image pulling | Check image URI |
| CrashLoopBackOff | Container exits | Check logs, health checks |
| ImagePullBackOff | Failed image pull | Verify credentials |
| OOMKilled (137) | Out of memory | Increase memory limit |
Service & Network
kubectl exec -it <pod-name> -- nslookup kubernetes.default
kubectl exec -it <pod-name> -- curl http://myservice:8080
kubectl get endpoints <service-name>
kubectl port-forward service/myservice 8080:8080
kubectl get networkpolicies -A
Quick Fixes
| Problem | Command |
|---|---|
| Pod stuck | kubectl delete pod <name> --grace-period=0 --force |
| High CPU | kubectl top pods -A --sort-by=cpu |
| High memory | kubectl top pods -A --sort-by=memory |
| Restart | kubectl rollout restart deployment/<name> |
| Rollback | kubectl rollout undo deployment/<name> |
See kubernetes-troubleshooting-advanced.md for node issues, HPA, anti-patterns.