I've run Kubernetes in production for years. I assumed the CKA would be a formality. Week one of prep humbled me — not because the material was new, but because the exam tests things production rarely makes you do by hand.
Surprise 1 — Speed is the real exam
The CKA is performance-based: you fix real clusters under a clock. Knowing the answer isn't enough; you need it fast. In production I'd open the docs, think, type carefully. The exam gives you minutes per task. That changes everything about how you work.
Surprise 2 — kubectl imperative commands matter more than YAML
In production I write declarative manifests and apply them. The exam rewards imperative speed:
Generating a manifest skeleton with --dry-run=client -o yaml and editing it is far faster than writing YAML from scratch. I'd never bothered learning this because in real life I have my templates. Under the clock, it's essential.
Surprise 3 — Set up these aliases on day one
alias k=kubectl
export do='--dry-run=client -o yaml' # k create deploy x --image=nginx $do
export now='--grace-period=0 --force'The exam lets you configure your shell. These three save real seconds on every task.
Surprise 4 — etcd backup/restore and static pods
The topics I'd never touched in managed clouds (EKS/GKE handle them) are exactly what the CKA drills:
etcdctl snapshot save/restore- static pod manifests in
/etc/kubernetes/manifests - manual
kubeadmupgrades
Managed Kubernetes had abstracted these away for me. The exam assumes you run the control plane yourself.
The honest takeaway
Years of using Kubernetes is not the same as years of operating the control plane by hand. The CKA tests the second thing.
If you're coming from managed cloud, budget time for etcd, kubeadm, and raw imperative kubectl. The concepts are familiar; the muscle memory isn't.