Certified Kubernetes Administrator (CKA)¶
The CKA wasn't an accident. I wanted it on purpose and treated it like a project. Study plan, daily labs, kubectl until the commands felt automatic. I didn't want to recognize answers. I wanted to sit in front of a terminal and fix things fast.
Why this one¶
The format sold me. The CKA is performance-based: a live terminal, real tasks, time pressure. No multiple choice. You either know how to do the job or you don't.
Kubernetes shows up everywhere in infrastructure work, and without fundamentals it's painful. You can copy manifests all day and still not understand what's happening. I wanted real confidence, not just familiarity.
The exam domains¶
I read the curriculum first and used it as a map. Five areas:
Cluster Architecture, Installation and Configuration. kubeadm, certificates, RBAC, control plane components. If you don't understand what the API server, etcd, scheduler, and controller manager do, everything else feels like magic.
Workloads and Scheduling. Pods, Deployments, DaemonSets, StatefulSets, Jobs. Scheduling rules, affinity, taints, tolerations, resource requests and limits.
Services and Networking. Pod-to-pod communication, Services, DNS, NetworkPolicies, Ingress. This part took me the most time. Kubernetes networking is layered and easy to get lost in.
Storage. PVs, PVCs, StorageClasses, access modes, reclaim policies. Smaller domain, but details matter.
Troubleshooting. Broken pods, misconfigured services, failing nodes. You only get good at this by breaking things repeatedly.
What worked for me¶
The official docs as a primary resource. They're the only reference allowed during the exam, so navigating them fast is a skill on its own.
KodeKloud for structure. Not just videos. It pushes you into labs quickly, and the tasks are close to what you face in the exam.
Local clusters constantly. kind, k3s, kubeadm. Every concept had to become hands-on. I broke control planes, recovered them, tested NetworkPolicies, drained nodes, rescheduled workloads.
Timed simulations. Killer.sh is harder than the real exam, and that's the point. It forces you to move fast and stay calm.
kubectl muscle memory. Imperative commands to generate YAML quickly (kubectl run, kubectl create, --dry-run=client -o yaml). kubectl explain constantly. Speed matters.
Exam day¶
Remotely proctored, browser-based environment, multiple clusters.
Three things that mattered most: always switch context before touching anything (forgetting costs points), prioritize high-value tasks first, and know the docs structure so you're not searching blindly.
The CKA gave me a much better mental model of Kubernetes. Troubleshooting started feeling methodical instead of chaotic. There's usually a sequence of checks that leads to the answer, but you only build that instinct through repetition.
It's hard in a good way. It rewards practice, not memorization.