Your hands decide how well you do on the Certified Kubernetes Administrator exam, not your memory of YAML templates. Any cka study guide for the 2026 exam has to start with one number: the Linux Foundation gives you 2 hours to work on a live cluster. The Cloud Native Computing Foundation splits the score across five domains whose weights spread from 10 percent for Storage to 30 percent for Troubleshooting, so the heaviest topic counts three times the lightest one. By the end of this guide you will know the current domains, what to memorize, what to build by hand, and a 6 week plan you can start on Monday.
The Linux Foundation runs the CKA as a live, online, hands-on exam. That single fact changes how you should study. You cannot pass by reading, and you cannot pass by slow lab work on your own laptop. You need speed on a real cluster, and you need it under a clock.
Last verified October 2026.
What should a cka study guide cover in 2026?
Start with the official weights. The Cloud Native Computing Foundation lists five domains for the CKA, and the gap between the biggest and the smallest is three to one.
| Exam domain | Weight | What it tests |
|---|---|---|
| Cluster Architecture, Installation and Configuration | 25 percent | Building a cluster with kubeadm, role based access control, Helm, and custom resources |
| Workloads and Scheduling | 15 percent | Deployments, rollouts, probes, ConfigMaps, Secrets, limits, and autoscaling |
| Services and Networking | 20 percent | Service types, DNS, Ingress, and network policies |
| Storage | 10 percent | Volumes, claims, storage classes, access modes, and reclaim policies |
| Troubleshooting | 30 percent | Finding faults in nodes, components, services, and container output |
Here is what those names mean in plain words. A workload is anything that runs in your cluster, such as a Deployment. A control plane is the set of parts that decides what runs and where. Storage covers the files your workloads keep. Networking covers how traffic moves between Pods and out to the internet. Troubleshooting covers working out why a broken thing is broken.
Troubleshooting carries 30 percent of the score, so it deserves the biggest slice of your lab time. If a Pod will not start, you have to read events, check logs, inspect a node, and read a crash report. Those four habits show up again and again in troubleshooting tasks.
If you are still weighing this exam against other credentials, our guide to the best DevOps certifications for 2026 careers puts the CKA next to the alternatives. It is a hands-on role credential, so it fits well beside a general DevOps path.
How long do you have and how do people run out of time?
The Linux Foundation exam tips page says the exam holds 15 to 20 performance-based tasks. Divide the clock and each task gets somewhere between six and eight minutes. That sounds comfortable until you meet a task that needs a rollout restart plus a service check.
The exam runs on the PSI Bridge platform. You work on a remote Linux desktop inside a secure browser while a proctor watches your webcam and screen share. The Linux Foundation asks for one active monitor, a screen of 15 inches or larger, and a 1080p resolution. Only the exam window may be open, and you cannot take the test inside a virtual machine.
The most common way candidates run out of time is not a hard question. It is one slow first task that eats 20 minutes, followed by a rush through everything else. Set a fixed budget instead. Decide that you will not spend more than seven minutes on any single task, and move on when the budget is gone.
The second most common mistake is going without documentation. You may browse the official Kubernetes documentation during the exam, and you may copy and paste from it. Practise looking things up so the habit feels normal on the day.
What should you memorize and what should you learn?
Memorize the shape of commands, not every flag. kubectl is the command line tool that talks to your cluster. You should reach for the right subcommand without thinking, and flags like -n for namespace, -o yaml for output, --all-namespaces, and --selector should feel automatic.
Facts to memorize before exam day
- The four service types and what each one does
- What a readiness probe and a liveness probe each control
- The difference between a Deployment, a ReplicaSet, and a StatefulSet
- How a role and a role binding fit together
Skills to build with your own hands
- Run a deployment with a rolling update and then roll it back
- Create a persistent volume claim and bind it to a workload
- Expose the same app as ClusterIP, NodePort, and LoadBalancer
- Set a resource limit and a network policy on a namespace
You do not need to memorise a Deployment manifest. You need to know what each field does. That is the line between reading Kubernetes docs and running a cluster.
Managed services change some answers. If your job uses Amazon EKS, Azure AKS, or Google GKE, read EKS vs AKS vs GKE managed Kubernetes compared. Some tasks will ask you to read cloud resources instead of raw nodes.
How do you practice in the same environment as the real exam?
The Linux Foundation bundles a simulator with the CKA. It runs through Killer.sh and it is the closest thing to exam day you can get at home. The Linux Foundation CKA page says you get two simulation attempts, that each session holds 17 questions, and that you get a different set of questions each time. You also get 36 hours of access for each attempt, and the simulator returns a graded score.
Two sessions is not many, so save them. Take the first one at the end of week four. Write down every task where you went over seven minutes and every task you could not finish. Any cka study guide that skips the simulator is guessing at the clock.
For everything else, build a local cluster. The official Kubernetes documentation has a setup page for each of the common local tools. Make sure you can destroy and rebuild the cluster in under five minutes, because a good share of troubleshooting tasks begin from a broken state.
Do not memorise answers to old practice questions. The exam draws fresh tasks every time, so learning the shape of a task is worth more than recalling one fix.
What does a week by week cka study plan look like?
Six weeks works if you touch the cluster most days. If you only have four, cut weeks one and four and keep the rest. Treat the table below as the backbone of any cka study guide you buy.
| Week | Focus | Lab target |
|---|---|---|
| Week 1 | Cluster setup and cluster architecture | Build a cluster with kubeadm, add a worker node, and read the control plane pod logs |
| Week 2 | Workloads and scheduling | Deploy, scale, roll out a new version, and roll back using only working commands |
| Week 3 | Services, networking, and storage | Expose three service types, write one Ingress rule, and bind one claim to one volume |
| Week 4 | Troubleshooting drill and first simulator | Break a cluster on purpose ten times and fix it ten times, then sit the simulator |
| Week 5 | Weak domains and speed work | Every task under six minutes, with one full simulator session held in reserve |
| Week 6 | Final mock and exam day logistics | One full mock on the real PSI system check, plus a run through of the exam day checklist |
Keep a lab notebook and write down every error you hit with the command that fixed it. Troubleshooting rewards pattern matching, and a notebook is how you build patterns fast.
What do you need to pass and what happens if you fail?
The Linux Foundation candidate FAQ says a CKA score of 66 percent or above is a pass. The same FAQ says the exam is scored automatically and your report arrives by email within 24 hours of finishing.
You do not have to pay for a second try. The Linux Foundation CKA page says your purchase gives you 12 months to schedule and take the exam, plus two exam attempts. Fees, bundles, and discounts move around, so read our CKA exam cost 2026 price breakdown before you buy anything.
Two attempts inside 12 months is plenty if you take the first one honestly. Do not spend both in your first week. Run the simulator, then book.
One last piece of career context. Teams that ship fast often blur the line between building and running software. The difference between Agile and DevOps in 2026 shows where the CKA sits in that picture and helps you explain the credential in an interview. If you already run clusters for a living, our Kubernetes cluster cost optimization 2026 guide covers the money side of the job that the exam never tests.
Photo: Pexels, Pixabay Content License, via Pixabay (https://pixabay.com/photos/port-pier-cargo-containers-crate-1845350/)






