
CKA (Certified Kubernetes Administrator) is the Linux Foundation and CNCF's hands-on exam for running Kubernetes clusters. You get a live command line and must build, configure, fix and upgrade clusters within 2 hours. It is the certification Hyderabad employers ask for most often in DevOps, SRE and platform engineering job descriptions, and you need it before you can take CKS.
This guide follows the official CKA exam page. It covers the exam facts, the five domains and their weightings, the kubectl and kubeadm tasks to practise, a 6-week plan and 15 original practice tasks.
Quick facts (Linux Foundation, November 2026)
| Item | Detail |
|---|---|
| Format | Online, proctored, performance-based tasks on a command line |
| Duration | 2 hours |
| Kubernetes version | v1.37 (updated 4 to 8 weeks after each Kubernetes minor release) |
| Price | US$445 for the exam; bundles with the LFS258 course or a THRIVE-ONE subscription cost more |
| Included | One free retake and two Killer.sh simulator sessions (36 hours each, 17 questions) |
| Prerequisites | None |
| Validity | 2 years; you have 12 months from purchase to take the exam |
Read the official CKA exam page. The linked curriculum PDF may show an older Kubernetes version than the exam itself, so practise on the current release.
Domains and weightings
| Domain | Weight | Competencies |
|---|---|---|
| Troubleshooting | 30% | Clusters and nodes; cluster components; resource usage; container output streams; services and networking |
| Cluster Architecture, Installation and Configuration | 25% | RBAC; preparing infrastructure; kubeadm clusters; cluster lifecycle; highly available control plane; Helm and Kustomize; CNI, CSI and CRI; CRDs and operators |
| Services and Networking | 20% | Pod connectivity; network policies; ClusterIP, NodePort and LoadBalancer services; Gateway API; Ingress; CoreDNS |
| Workloads and Scheduling | 15% | Deployments, rolling updates and rollbacks; ConfigMaps and Secrets; autoscaling; self-healing primitives; admission and scheduling (limits, node affinity) |
| Storage | 10% | Storage classes and dynamic provisioning; volume types, access modes and reclaim policies; PVs and PVCs |
Troubleshooting carries almost a third of the marks. Practise fixing broken clusters more than building new ones.
Commands and manifests to practise
# speed setup
alias k=kubectl; export do="--dry-run=client -o yaml"
k config use-context <context-from-the-question> # always switch context first
# RBAC
k create role pod-reader --verb=get,list,watch --resource=pods -n dev
k create rolebinding read-pods --role=pod-reader --user=jane -n dev
k auth can-i list pods -n dev --as jane
# workloads
k create deploy web --image=nginx --replicas=3
k set image deploy/web nginx=nginx:1.27; k rollout status deploy/web; k rollout undo deploy/web
# troubleshooting
k get nodes; k describe node worker-1; k get events -A --sort-by=.lastTimestamp
ssh worker-1 'sudo systemctl status kubelet; sudo journalctl -u kubelet --no-pager | tail'
sudo crictl ps -a; sudo crictl logs <container-id>
k logs pod/api -c app --previous; k top pods -A
# kubeadm upgrade (control plane, then each worker)
sudo kubeadm upgrade plan
sudo kubeadm upgrade apply v1.37.x
k drain worker-1 --ignore-daemonsets --delete-emptydir-data
# upgrade the kubelet and kubectl packages, then:
sudo systemctl daemon-reload && sudo systemctl restart kubelet
k uncordon worker-1
Gateway API HTTPRoute:
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
name: web
spec:
parentRefs:
- name: main-gateway
hostnames: ["web.example.com"]
rules:
- matches:
- path: {type: PathPrefix, value: /}
backendRefs:
- name: web
port: 80
HorizontalPodAutoscaler:
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
name: web
spec:
scaleTargetRef: {apiVersion: apps/v1, kind: Deployment, name: web}
minReplicas: 2
maxReplicas: 6
metrics:
- type: Resource
resource:
name: cpu
target: {type: Utilization, averageUtilization: 70}
6-week study plan
| Week | Focus | Lab (kubeadm on 3 VMs, or kind/minikube for objects) |
|---|---|---|
| 1 | kubectl speed, contexts, pods, deployments, rollouts, ConfigMaps, Secrets | Generate every manifest with --dry-run=client -o yaml |
| 2 | Scheduling: requests and limits, node affinity, taints and tolerations, autoscaling | Pin a pod to one node, then evict it with a taint |
| 3 | kubeadm install, HA control plane concepts, upgrades, RBAC, Helm, Kustomize, CRDs and operators | Build a 1-control-plane, 2-worker cluster and upgrade it by one minor version |
| 4 | Services, CoreDNS, network policies, Ingress, Gateway API | Expose an app through an HTTPRoute; block other namespaces with a policy |
| 5 | Storage classes, PVs, PVCs, access modes, reclaim policies | Dynamic provisioning with a local-path or CSI driver |
| 6 | Troubleshooting: kubelet, static pods, certificates, DNS, services; both Killer.sh sessions | Break the cluster in 5 ways and fix it; book the exam |
15 practice tasks with solutions
These are original tasks written for this guide, not exam questions.
1. A node shows NotReady. Where do you look first?
SSH to it and check the kubelet with systemctl status kubelet and journalctl -u kubelet.
2. Where are static pod manifests for the control plane on a kubeadm cluster?/etc/kubernetes/manifests.
3. Let user jane list pods only in namespace dev.
Create a Role with get, list and watch on pods in dev, and a RoleBinding to jane; test with k auth can-i.
4. Roll a deployment back to its previous version.k rollout undo deploy/<name>.
5. Keep pods off node-2 unless they tolerate it.k taint nodes node-2 key=value:NoSchedule.
6. Make a pod run only on nodes labelled disk=ssd.
Use node affinity with requiredDuringSchedulingIgnoredDuringExecution (or a simple nodeSelector).
7. What order do you upgrade a kubeadm cluster in?
Control plane nodes first (kubeadm, then kubelet and kubectl), then each worker after draining it.
8. Which service type gives a stable internal IP only?
ClusterIP.
9. Pods can't resolve service names. What do you check?
The CoreDNS pods and service in kube-system, the pod's /etc/resolv.conf, and network policies that block DNS on port 53.
10. Deny all ingress traffic to pods in a namespace.
A NetworkPolicy with an empty podSelector: {}, policyTypes: [Ingress] and no ingress rules.
11. Which Gateway API resource routes HTTP paths to services?
HTTPRoute, attached to a Gateway with parentRefs.
12. A PVC stays Pending. Name two causes.
No matching PV or StorageClass, or the requested access mode or size isn't available.
13. What happens to a PV with reclaim policy Retain when its PVC is deleted?
The PV and its data are kept, in Released state, until an administrator cleans it up.
14. See logs from a container that crashed and restarted.k logs <pod> -c <container> --previous.
15. Install a component from a Helm chart into its own namespace.helm install <release> <chart> -n <ns> --create-namespace.
Before and after CKA
- Start with KCNA; developers can take CKAD alongside CKA.
- After CKA, move on to CKS, CKNE or OpenShift with EX280.
- Learn it hands-on in our Kubernetes course and DevOps course; see live Kubernetes and DevOps jobs.
Cloud Soft Solutions in Ameerpet, Hyderabad teaches Kubernetes, DevOps and multi-cloud with real-time projects. See the APEX program and all courses.
📞 Call +91 96660 19191 · WhatsApp · Flat No. 513, 5th Floor, Nilagiri Block, Aditya Enclave, Ameerpet, Hyderabad
Frequently Asked Questions
What is the CKA exam?
The Certified Kubernetes Administrator exam from the Linux Foundation and CNCF: 2 hours of hands-on tasks on a live Kubernetes command line.
Which Kubernetes version does CKA use?
v1.37 as of November 2026. The exam is updated 4 to 8 weeks after each minor release.
How much does CKA cost?
US$445 for the exam, including one retake and two Killer.sh simulator sessions.
What are the CKA domains?
Troubleshooting 30%, Cluster Architecture, Installation and Configuration 25%, Services and Networking 20%, Workloads and Scheduling 15%, Storage 10%.
Are there prerequisites for CKA?
No.
How long is CKA valid?
2 years.



