CKA in 2026 — troubleshooting is now the biggest domain

The Certified Kubernetes Administrator exam is the closest thing the ecosystem has to a baseline credential. This is its current shape, checked against the Linux Foundation’s published exam page — which is also where you should confirm it before you book, since versions and weightings move with the release cycle.

What the price includes

The $445 covers two exam attempts within a 12-month eligibility window, plus two Killer.sh simulator attempts. Bundles run to $625 with a THRIVE-ONE annual subscription and $645 with Kubernetes Fundamentals (LFS258).

One wrinkle on validity: certifications earned before 1 April 2024 remain valid for three years. Everything since is two.

Why the weightings matter

Troubleshooting is the single largest domain, and that one line is the most useful thing on this page.

It means the exam isn’t testing whether you can recite the fields of a PodSpec. It’s testing whether you can take a cluster that’s wrong and work out why. Most study material builds the first skill and leaves the second alone.

Preparing for a troubleshooting exam

Reading the documentation until you know the API is necessary but nowhere near sufficient. What moves the needle is deliberately breaking things:

• Stop the kubelet on a node. Find out from the control plane that you did.
• Give a pod a node selector nothing satisfies. Read the scheduler’s account of why it’s pending.
• Break CoreDNS. Watch what fails, and in what order.
• Point a Service at the wrong label. Notice how far you get before anything says so.
• Fill a node’s disk until eviction kicks in.

Each of those teaches you a signal: a place to look, and what a healthy answer there looks like. Thirty percent of the exam is made of those signals.

Building a cluster you are willing to break is the part most people skip, usually because a broken lab is annoying to rebuild. Our own practice platform, CertLabs, exists for that: sandboxed clusters that reset between exercises, with each objective graded against live system state rather than the command you typed.

Two hours goes quickly

Nearly everyone who fails knew the material and ran out of clock. Three habits buy time, and they’re worth practising until they’re automatic.

Set the context first. Every task tells you which cluster and namespace. Getting this wrong means doing a task perfectly against the wrong target, which scores zero and costs you the time twice.

Generate, don’t author. kubectl run, kubectl create and kubectl expose with --dry-run=client -o yaml produce a manifest skeleton in two seconds that takes two minutes to type. Edit the skeleton.

Learn the documentation, not just Kubernetes. You’re allowed the official docs during the exam, which only helps if you know where things are. Practise finding the pod-security page or the CSI section from a cold start, under time pressure. Bookmarks are permitted, so use them.

The limits of the credential

CKA proves you can operate a cluster on the day. It says nothing about whether you can design one, or run one under load with a business depending on it.

We say this about every certification on this site, including the ones we teach: use it to structure your learning and to clear a procurement checkbox. Both are legitimate. Just don’t confuse the credential with the capability. That gap gets filled by operating something real, and there’s no exam for it.

Where to go next

• CKAD if you build on Kubernetes rather than run it
• CKS if you run it and security is your problem — CKA is a hard prerequisite
• EX280 if your organisation runs OpenShift specifically

If you’re unsure which of the three to take first, we wrote a comparison.