The CKA Is a Lab Exam, Not a Quiz, and That Changes Everything

Most certification exams ask whether you recognize the right answer. The Certified Kubernetes Administrator asks whether you can make a broken cluster work while a clock runs. There is no partial credit for knowing what you should have typed. A task either passes the grader or it does not.

That single design decision drives everything about how the exam should be studied, and it is the thing most candidates coming from a multiple-choice background get wrong first.

The format, from the official record

The Linux Foundation’s CKA page sets out the shape of it:

• $445 for the exam, which includes a second attempt if the first does not go your way.
• Two hours, performance based, in a live terminal.
• Kubernetes v1.35 is the version under test.
• Certification is valid for two years.
• Two Killer.sh simulator sessions come with enrollment, 36 hours of access each, seventeen questions per session.

The passing score is 66%, which the Linux Foundation documents in its CKA and CKAD FAQ. That number deserves a second look. It means you can get a full third of the exam wrong and still walk away certified. It also means that a single task you cannot finish is survivable, and that time spent stuck is more expensive than time spent wrong.

The domain weights are the study plan

The exam publishes how it is weighted, which is more generosity than most certifying bodies extend. The current split:

Domain	Weight
Troubleshooting	30%
Cluster Architecture, Installation and Configuration	25%
Services and Networking	20%
Workloads and Scheduling	15%
Storage	10%

Look at the top and bottom of that table together. Troubleshooting is worth three times what Storage is worth. It is worth more than Storage and Workloads combined.

Now consider how most Kubernetes courses are ordered. They start with Pods, move to Deployments, cover Services, reach Storage somewhere in the back half, and treat troubleshooting as a closing chapter if they treat it at all. That is almost exactly inverted from the exam’s own priorities.

If you allocate study time in proportion to the weights, roughly a third of your preparation is spent deliberately breaking clusters and fixing them. Not reading about failure modes. Producing them, on purpose, and working out what the symptoms look like from the inside.

Open book, and why that is not a gift

The Kubernetes documentation is permitted during the exam. Candidates hear this and relax, which is the wrong response.

The documentation is enormous. Two hours is not enough time to read it. What the open-book policy actually rewards is knowing the shape of the docs well enough to land on the right page in one move. If you find yourself using the site search during the exam, you have already lost more time than the task is worth.

The practical consequence is that the memorization target shifts. You do not need to hold every YAML field in your head, because you can look those up. You do need to know, without thinking:

• Which resource solves which problem.
• The imperative kubectl command that generates the manifest you want, so you can start from something rather than typing from scratch.
• Where in the documentation each common example lives.

That is a different kind of preparation from flashcards, and it is worth a post of its own.

Two hours is the real adversary

Roughly fifteen to twenty tasks in one hundred and twenty minutes gives you something on the order of six or seven minutes each. Some tasks take ninety seconds. Some are twenty-minute rabbit holes wearing a friendly disguise.

The single most useful exam-day habit is the least technical one: read every task first, and do the ones you can finish quickly before you touch anything hard. The exam lets you flag and return. At a 66% pass mark, sweeping up the cheap points is a better strategy than achieving completeness on the expensive ones.

The second most useful habit is setting your context correctly at the start of every task. Multiple clusters are in play. A perfect answer applied to the wrong cluster scores zero, and it is a mistake that feels impossible until the moment you make it.

Who this exam is actually for

The CKA is not an introduction to Kubernetes. It assumes you can already operate a cluster and asks how well. If you have never run kubectl in anger, the KCNA is the gentler entry point and covers the conceptual ground first.

If you are already running workloads and want the credential to say so, the CKA is unusually honest about what it measures. There is no way to pass it by recognizing vocabulary. That is inconvenient to study for and is exactly what makes it worth holding.

Our CKA Study Guide is built around the same weighting the exam publishes, which is why roughly a third of it is troubleshooting.

Sources: Linux Foundation CKA certification page, Linux Foundation CKA/CKAD/CKS FAQ.