CKADExam PrepCertificationkubectl

How to Prepare for the CKAD Exam (Application-Focused Practice)

CKAD rewards fast, correct workload debugging — probes, rollouts, Services, ConfigMaps, and Jobs. A practical study plan without drowning in cluster internals.

2026-06-073 min read

The CKAD exam is a two-hour, hands-on sprint: roughly 15–20 tasks, 66% to pass, live clusters, official Kubernetes docs in a second tab. No one cares if you can recite every API field — they care if you can deploy, configure, expose, and fix applications quickly.

This guide is a study plan aligned with that reality, not a syllabus dump.

What CKAD actually measures

The exam blueprint rotates, but these domains show up repeatedly:

Domain	What you must do under pressure
Workloads	Deployments, ReplicaSets, rollbacks, multi-container pods
Configuration	ConfigMaps, Secrets, env, volumes, securityContext basics
Services & networking	ClusterIP, NodePort, DNS, NetworkPolicy
Observability	Probes, logs, resource requests/limits
Batch	Jobs, CronJobs, parallelism/completions

Notice what is not the focus: etcd backup, kubeadm upgrades, static control plane manifests on disk. Save that energy for CKA.

Build muscle memory for the exam workflow

1. Generate YAML fast

kubectl create deployment web --image=nginx:1.25 --dry-run=client -o yaml > deploy.yaml
kubectl expose deployment web --port=80 --dry-run=client -o yaml >> svc.yaml

Edit, kubectl apply -f. The exam rewards imperative → YAML → apply, not typing from scratch.

2. Debug in a fixed order

For workload problems:

kubectl get pods -o wide
kubectl describe pod <name>
kubectl logs <name> --previous   # CrashLoopBackOff
kubectl get deploy,rs,svc

Train yourself to pick the first command before reaching for delete pod.

3. Probes are exam bait

Slow-start apps need startupProbe or realistic initialDelaySeconds. Wrong probe port is a classic CKAD trap — the container runs, but Ready never becomes true.

4. Services need Endpoints

kubectl get endpoints (or EndpointSlices) — if empty, trace selector labels back to pod labels.

5. Resources and HPA

HPA with CPU utilization requires CPU requests on pods. Missing requests = metrics show <unknown>.

A 4-week prep outline

Week 1 — Foundations

Run through free Pod Debugging scenarios until describe + logs --previous is automatic
Read our CrashLoopBackOff guide and Pending pods guide

Week 2 — Application patterns

Practice Deployments, rollouts, ConfigMaps/Secrets, Jobs in a local cluster (kind/minikube)
Decision Trainer topics: Deployment Rollouts, Config & Secrets, Service Connectivity

Week 3 — Timed pressure

CKAD Exam Prep pack: 100 scenario cases + exam sprints (5-minute timed sets)
Bookmark docs: Pod probes, Services, Deployments, NetworkPolicy

Week 4 — Simulation

Use the official Killer.sh exam simulator included with exam registration
One full timed run; note where you lose minutes (YAML typing vs diagnosis)

Exam-day habits

Read the whole task before typing — some tasks share a cluster context
SSH first when the infobox tells you — work on the designated host
Flag and move on — partial credit beats one rabbit hole
Verify — kubectl get, curl a Service, check Ready count
Use k and completion — they are pre-configured for a reason

Practice with scenario grading

Flashcards that ask "what is a liveness probe?" do not prepare you for "this Deployment rolled out but traffic still hits old pods — what do you check first?"

CKAD Exam Prep on k8s Flashcards uses graded first-step decisions — best, acceptable, bad, and trap answers — plus timed exam sprints. It complements hands-on labs; it does not replace them.