What the CKA is and who it is for
The Certified Kubernetes Administrator is a performance based certification from the Cloud Native Computing Foundation and The Linux Foundation. It certifies that you can install, configure and manage production grade Kubernetes clusters. There is no multiple choice. You sit in a live terminal connected to real clusters and complete tasks, and you either get them working or you do not.
It is aimed at people who run Kubernetes for a living or want to: platform engineers, DevOps engineers, site reliability engineers and systems administrators moving into container platforms. It also suits developers who deploy to Kubernetes and want to understand the cluster underneath their workloads.
The credential is valid for two years, and each registration includes one free retake, which takes some of the pressure off a first attempt.
Linux Foundation CKA at a glance
| Item | Detail |
|---|---|
| Exam code | CKA |
| Questions | 15 to 20 performance based tasks |
| Time allowed | 2 hours |
| Passing score | 66% |
| Exam fee | US$445 per registration, including one free retake |
| Where you sit it | Online proctored via PSI |
| Certification valid for | 2 years |
What is on the exam
Troubleshooting (30%). The largest domain. You are given a cluster or an application that is broken and asked to fix it: nodes that are not ready, control plane components that are failing, pods stuck in a bad state, networking that does not connect and logs that need reading. These tasks are worth the most and take the longest, so your diagnostic routine needs to be fast and repeatable.
Cluster architecture, installation and configuration (25%). Building and maintaining the cluster itself: RBAC, managing the control plane, adding and upgrading nodes, backing up and restoring the cluster state, and installing the components a production cluster depends on. Expect to be asked to make configuration changes on the nodes, not just through kubectl.
Services and networking (20%). How traffic gets into and around the cluster: Service types, DNS inside the cluster, network policies that allow or deny traffic between pods, and exposing applications to the outside. Network policy tasks are a common place to lose marks because the YAML is fiddly and hard to verify at a glance.
Workloads and scheduling (15%). Deployments, rolling updates and rollbacks, scaling, configuring applications with ConfigMaps and Secrets, and controlling where pods run with labels, selectors, taints, tolerations and affinity.
Storage (10%). Persistent volumes, persistent volume claims, storage classes, access modes and attaching storage to pods. Small in weight but usually quick to score if you know the objects.
Why people fail it
There is no multiple choice, no definitions and no process of elimination. You get a live Kubernetes cluster, a terminal and a set of tasks, and you either complete them or you do not. Candidates who have studied Kubernetes from books and videos, but not from a terminal, find that the gap between knowing and doing is the whole exam.
Time is the real enemy. Fifteen to twenty tasks in two hours, with 66% needed to pass, leaves only a few minutes per task once you allow for reading and switching contexts. Most candidates who fail ran out of time, not knowledge. Knowing the commands is not enough: you need to type them fast, verify the result and move on without hesitation. Hand writing YAML from memory, hunting through documentation for every field or debugging a typo for five minutes is how a passing candidate becomes a failing one.
The other common cause is not verifying. A task looks done, but the pod never actually became ready, or the change went into the wrong namespace or the wrong cluster context. Unchecked work scores nothing, so verification is part of every task.
A study plan that fits the exam
Six weeks, with the time weighted to troubleshooting and cluster architecture, and every session done in a terminal. Start by trying the free CKA practice tasks against a cluster of your own so you know what a task looks like and how long it takes you.
- Week 1: cluster architecture, part one. Build a cluster yourself, learn the control plane components and where their configuration lives, and practise backing up and restoring cluster state.
- Week 2: cluster architecture, part two, plus workloads. RBAC, node maintenance and upgrades, then Deployments, rollouts, scaling, ConfigMaps, Secrets and scheduling controls.
- Week 3: services and networking, plus storage. Service types, cluster DNS, exposing applications and network policies until you can write one without looking. Then persistent volumes and claims.
- Week 4: troubleshooting, part one. Break your own cluster on purpose: stop a component, misconfigure a node, corrupt a manifest. Build a fixed diagnostic routine and time yourself repairing each fault.
- Week 5: troubleshooting, part two, plus speed. Application and networking faults. Work through the 23 tasks in the CKA practice task pack with the step by step solutions, noting the fastest route each one takes and the verification step it ends with.
- Week 6: timed runs. Use the tasks only copy against the clock. Aim to finish every task with a verification command, skip anything that stalls you and come back to it, and repeat until you are comfortably inside two hours.
On exam day
The CKA is delivered online with a proctor through PSI, so your own room and machine are the exam environment. Check the system requirements, clear the desk, and test your webcam and connection the day before. Expect the check in and environment inspection to take time before your two hours begin.
Once the clock starts, read each task for the namespace and the cluster context it names, and switch to them before you touch anything. Do the tasks you find easiest first to bank marks, and leave the long troubleshooting tasks until you have a buffer. Generate YAML with kubectl rather than writing it from scratch, verify every task with a quick command that proves the outcome, and if something is not working after a few minutes, move on. The pass mark is 66%, so a couple of skipped tasks are survivable, but a task that ate twenty minutes is not.
Frequently asked questions
What happens if I fail the CKA?
Each registration includes one free retake, so a first failure costs you time rather than another fee. Use the score breakdown to see which domains fell short, rebuild your timed practice around them, and book the retake once you are finishing full runs inside two hours. If you fail the retake as well, you need a new registration at US$445.
Is the practice pack enough on its own?
No. The pack is 23 practice tasks with full step by step solutions, covering the formats the exam uses: cluster troubleshooting, RBAC, network policies, storage and scheduling. It is a rehearsal for the exam, not a Kubernetes course. You need a cluster to practise on and enough hands on time that the commands are automatic. The pack then shows you the fastest route through each task type and the verification that proves it is done.
Can I use documentation during the exam?
Check The Linux Foundation’s current rules on which resources are allowed during the exam before your sitting. Whatever is allowed, do not plan to rely on it: looking things up costs time, and time is what most failing candidates run out of.
How long is the CKA valid?
Two years from the date you pass. After that you sit the exam again to renew.
When you are ready to practise against the clock, the CKA practice task pack is US$39, refunded if you fail.