Studying CKA and CKAD together: the overlap, the order and one shared plan

Same format, different job

Both are performance-based exams on a remote Linux desktop, scored on the state you leave behind. CKA is described by the Linux Foundation as the exam for Kubernetes administrators; CKAD is for people who design, build and deploy cloud native applications on Kubernetes. The format numbers are nearly identical, so the difference is entirely in what the tasks ask you to do.

The overlap, topic by topic

Read the two curricula side by side and a large shared core appears. Both expect you to run and update workloads, configure them with ConfigMaps and Secrets, expose them with Services, restrict them with NetworkPolicies, and install packages with Helm and Kustomize. Both mention CRDs and operators. These are the topics to practise first, because every hour spent on them counts twice.

What only CKA asks

CKA's largest domains are Troubleshooting at 30% and Cluster Architecture, Installation and Configuration at 25%. Most of that is administration a developer rarely touches: preparing infrastructure for a kubeadm cluster, managing its lifecycle, a highly available control plane, the extension interfaces, RBAC, and diagnosing broken nodes and components. Storage classes and dynamic provisioning, CoreDNS and the Gateway API are also CKA items.

• kubeadm: initialise, join and upgrade a cluster; know where the static Pod manifests and certificates live.
• Troubleshooting nodes and control plane components with systemctl and journalctl, not only kubectl.
• StorageClasses, access modes, reclaim policies, PVs and PVCs.
• Gateway API resources for ingress traffic, alongside the classic Ingress.

What CKAD emphasises

CKAD spreads its weight evenly; its largest domain is Application Environment, Configuration and Security at 25%. It goes deeper into the application itself: defining and building container images, choosing the right workload resource, multi-container Pod patterns such as sidecars and init containers, probes, container logs and debugging, SecurityContexts and capabilities, ServiceAccounts, and deployment strategies such as blue/green and canary built from Kubernetes primitives. It also lists understanding API deprecations.

• Build and modify a container image, then run it in a Pod.
• Liveness, readiness and startup probes, and what happens when each fails.
• Jobs and CronJobs, and when a Deployment, DaemonSet or StatefulSet fits better.
• A canary with two Deployments behind one Service, and a blue/green switch by changing a selector.

Which to take first

The official pages impose no order between these two. Choose by background. If you run clusters, take CKA first: most of CKAD is then the shared core plus a few application topics. If you write and ship applications, take CKAD first: the shared core is familiar, and CKA becomes the administration layer on top. If you are planning for Kubestronaut, remember that CKS requires a passed CKA, so CKA first keeps the path to CKS open sooner. Whichever you choose, keep the gap between the two exams short. The shared core decays quickly when you stop using it daily, and two weeks between the exams is enough to cover the topics only the second one asks about.

• Operations or platform background: CKA, then CKAD.
• Developer background: CKAD, then CKA.
• Aiming at CKS soon: CKA first, because it is the CKS prerequisite.

An eight-week plan for both

The plan assumes CKA first and a few hours of practice most days. It builds the shared core once, then adds each exam’s own topics, and books the two exams a few weeks apart. Swap weeks five and six with the CKAD-only material if you take CKAD first.

Practising both in one cluster

A single local cluster with two worker nodes covers almost everything both exams ask, as long as you can break and rebuild it. Keep it on the exam's minor version, v1.37, so flags and API versions match. Rebuild it from scratch at least twice during the plan, because kubeadm installation and upgrades are CKA tasks and the practice of reading your own errors is the best troubleshooting training there is. Rebuild the cluster at least once during the eight weeks. Installation and upgrades are CKA tasks, and a fresh cluster also clears the leftovers of earlier exercises that can hide mistakes in later ones.

Hands-on labs for CKA, CKAD, CKS, LFCS, ICA and CNPE are in preview: the GoldenKube CLI runs them on your own machine, but it has no public release yet. Until there is a public release, plan on your own cluster as well.

Habits that pay off in both exams

Because the two exams share one environment and one tool, the habits you build for the first carry straight into the second. Practise them from the first week, not in the final days, so they cost you no thought during either exam.

• Start each task by connecting to the host or switching to the context it names, then check the namespace.
• Create objects imperatively where kubectl allows it and generate YAML with --dry-run=client -o yaml where it does not.
• Use kubectl explain to look up a field before you search the documentation.
• Read every object back after creating it and compare it with the task text: names, labels, ports, images, namespaces.
• Skip a task that stalls, note its number and come back after the first pass.
• Copy and paste with Ctrl+Shift+C and Ctrl+Shift+V in the exam terminal, which the instructions describe.

Booking two exams back to back

Each exam is a separate purchase with its own 12-month window to schedule and take it, its own retake and its own simulator access. For the three Kubernetes exams, the FAQ gives two Killer.sh simulator attempts per registration, each open for 36 hours, except for single-attempt purchases. In this plan that means two CKA sessions in week six and two CKAD sessions in week eight.

1. 1Book CKA once the week-four milestone is met, for a slot at the end of week six.
2. 2Book CKAD after you pass CKA, for the end of week eight.
3. 3If you need to move a booking, do it at least 24 hours before the start; a no-show forfeits the attempt.
4. 4If an exam does not go well, use the included retake after a week or two of targeted review and move the second exam by the same amount.