The Kubernetes Errors You Will See

If you run K8s long enough, you'll meet all 50 of these. This series groups them into 3 layers so you can jump straight to the right part when things break.

The debug pyramid

                +------------------+
                |   Cluster-wide   |  Part 3
                |  (RBAC, HPA,     |
                |   ETCD, Autoscale|
                +---------+--------+
                          |
                +---------+--------+
                |    Networking    |  Part 2
                | (Services, DNS,  |
                |  Ingress, CNI)   |
                +---------+--------+
                          |
                +---------+--------+
                |    Pod-level     |  Part 1
                | (Crash, Pending, |
                |  ImagePull, OOM) |
                +------------------+

Golden rule: always start at the bottom of the pyramid. kubectl describe pod + kubectl logs fix 70% of issues without touching the cluster.

The 5 commands you'll type every day

kubectl get pods -A                    # who is unhappy?
kubectl describe pod <name>            # events at the bottom = truth
kubectl logs <pod> [-c container]      # what did the app say?
kubectl logs <pod> --previous          # what did it say *before* it crashed?
kubectl get events --sort-by='.lastTimestamp'  # global recent activity

What each part covers

Part	Topic	Errors covered
1	Pod-level	CrashLoopBackOff, Pending, ImagePullBackOff, OOM, Probe failures, Terminating stuck
2	Networking	Service unreachable, DNS, Ingress, NodePort, LoadBalancer, NetworkPolicy, CNI
3	Cluster/Storage	RBAC, HPA, PVC pending, ETCD, Cert expiry, Node drain, Scheduler, Quotas

Real-world example

3 AM on-call ping: "checkout is down". You SSH into the jumpbox, run kubectl get pods -A → 4 pods in ImagePullBackOff. You open Part 1 of this series, follow the 4-step checklist → turns out the CI job pushed a typo'd tag last night. 8 minutes, incident closed.