50 Kubernetes Issues Every DevOps Engineer Should Know How to Fix Kubernetes is powerful, but when things break, knowing what to check matters more than memorizing commands. Here are 50 common Kubernetes issues and the first thing I’d check: 1. Pod stuck Pending → kubectl describe pod 2. ImagePullBackOff → Check image name, tag & registry credentials 3. ErrImagePull → Verify image exists and registry access 4. CrashLoopBackOff → Check container logs and startup command 5. OOMKilled → Increase memory limits or investigate memory usage 6. Container stuck Terminating → Check finalizers and node status 7. Pod stuck ContainerCreating → Check volumes, CNI & events 8. Readiness probe failing → Verify endpoint, port & application health 9. Liveness probe failing → Fix health endpoint or probe configuration 10. Startup probe failing → Give slow-starting apps more time 11. Service not reachable → Verify selectors and endpoints 12. Service has no endpoints → Check pod labels vs service selector 13. Ingress returns 404 → Check Ingress rules and backend service 14. Ingress returns 502 → Check backend service and target port 15. DNS resolution failing → Check CoreDNS pods and configuration 16. CoreDNS CrashLoopBackOff → Inspect CoreDNS logs/config 17. Pod cannot reach another pod → Check NetworkPolicies and CNI 18. Pod cannot reach the internet → Check CNI, NAT and node routing 19. NetworkPolicy blocking traffic → Review ingress/egress rules 20. PersistentVolume stuck Pending → Check StorageClass and provisioner 21. PVC stuck Pending → Check available PVs/storage provisioner 22. PVC stuck Terminating → Check finalizers and attached resources 23. Volume mount failure → Check CSI driver and permissions 24. Read-only filesystem → Check mount options and application behavior 25. Node NotReady → Inspect kubelet, CNI and node conditions 26. Node DiskPressure → Clean images, logs and unused volumes 27. Node MemoryPressure → Find memory-heavy workloads 28. Node PIDPressure → Check processes and pod density 29. Kubelet not running → Check systemd and kubelet logs 30. Pods not scheduling → Check taints, tolerations and resources 31. Insufficient CPU → Adjust requests or scale the cluster 32. Insufficient memory → Adjust requests or add capacity 33. Taints preventing scheduling → Add the correct tolerations 34. Affinity preventing scheduling → Review node/pod affinity rules 35. Deployment rollout stuck → Check ReplicaSets, events and pod status 36. Deployment unavailable → Check replicas, probes and scheduling 37. Rolling update too slow → Review strategy and readiness probes 38. Old pods not terminating → Check PDBs, finalizers and termination hooks 39. HPA not scaling → Check metrics-server and resource requests 40. Metrics unavailable → Verify metrics-server installation/config 41. Secret not available → Check secret name, namespace and RBAC 42. ConfigMap changes not reflected → Understand mounting/env injection behavior 43. RBAC Forbidden errors → Check Role/ClusterRole and bindings 44. ServiceAccount authentication failing → Verify token/RBAC configuration 45. Webhook failures → Check webhook service, TLS and endpoints 46. API server unavailable → Check control-plane health 47. etcd issues → Check etcd health, disk and resource usage 48. Cluster certificate problems → Check certificate expiry and configuration 49. Unexpected pod evictions → Investigate node pressure and QoS 50. Random application failures → Start with events, logs, metrics and dependencies The most important Kubernetes troubleshooting commands to master: - kubectl get - kubectl describe - kubectl logs - kubectl exec - kubectl events - kubectl top - kubectl rollout - kubectl get endpoints Don’t just learn Kubernetes commands, learn how Kubernetes works. When you understand scheduling, networking, DNS, storage, probes, controllers, RBAC and the control plane, troubleshooting becomes significantly easier. Save this. You’ll need it when Kubernetes decides to ruin your day.