CKAD — the Kubernetes exam for people who ship on it

CKAD is the exam for people who deploy applications onto Kubernetes rather than operate the cluster underneath them. Weightings and the target Kubernetes version move each cycle, so confirm both on the Linux Foundation’s exam page before you book.

Bundles run to $625 with a THRIVE-ONE subscription and $645 with Kubernetes for Developers.

“Easier than CKA” is the wrong frame

Believing that framing is how experienced developers fail this exam.

CKA is weighted toward diagnosing a cluster; 30% of it is troubleshooting. CKAD assumes the cluster works and asks whether you can express an application correctly inside it. The difficulty isn’t lower, it’s differently placed. CKAD’s pressure comes from volume and pace: a lot of small tasks, each of which you have to get exactly right, in two hours.

If CKA rewards knowing where to look, CKAD rewards knowing what to type.

The 25% domain

Application Environment, Configuration and Security covers ConfigMaps, Secrets, security contexts, resource requests and limits, service accounts, and admission-adjacent behaviour.

Requests and limits deserve extra study time. Not because they’re conceptually hard, but because the consequences of getting them wrong are the most commonly misunderstood thing in Kubernetes. Know precisely what happens when a container exceeds its memory limit versus its CPU limit. The two behave completely differently, and the exam can tell whether you know that.

Design and build

Multi-container pod patterns — sidecar, ambassador, adapter — plus init containers, jobs and cronjobs, and building a container image that isn’t needlessly awful. Know when a sidecar is the right answer, and when you’re really looking at two things that should have been two pods.

Deployment

Rolling updates, rollbacks, deployment strategies, scaling, and Helm basics. Rehearse the rollback path: inspecting rollout history and returning to a previous revision should be something you do without thinking.

Observability and maintenance

Probes (liveness, readiness, startup), plus logs and debugging a running container. Probes are the highest-value part of this domain for real work, never mind the exam. A wrong liveness probe is one of the few Kubernetes misconfigurations that makes uptime worse than having no probe at all, by restarting a container that was merely slow.

Preparing

The single highest-return habit is the same as CKA: generate manifests, don’t type them.

kubectl run, kubectl create and kubectl expose with --dry-run=client -o yaml produce a valid skeleton instantly. Typing a PodSpec from memory under time pressure is how people run out of clock with three tasks left.

Learn kubectl explain too. It answers “what’s the field called again” faster than the documentation site, and after a few weeks it works offline in your head.

Both habits need repetition against a real cluster rather than a text editor. The CKAD track on CertLabs, our practice platform, runs in a browser terminal and grades each objective against live system state, which is the closest thing to the exam’s own feedback loop.

Who should take it

Developers and platform users who ship onto Kubernetes and want a structured way to learn it properly, rather than by pattern-matching whatever the last team wrote.

If you run the cluster rather than deploy onto it, CKA is the better fit. If you’re not sure, we compared all three.