CertGrid CertGrid
Concepts·Kubernetes and Cloud Native Associate

KCNA Exam Format and Domain Weights

KCNA is broad and shallow on purpose: four domains, one of them nearly half the exam, and no live cluster to fix. This cluster answers for 161 API resource types, 71 of them added by installed software - which is the whole argument for a separate cloud native exam.

Orientation Guide 1 of 46 Beginner

Written against the versions above. Domain weights come from the published KCNA curriculum. Check the current one before you book - the Linux Foundation revises it.

One control plane and three workers. Nothing in this path needs a second cluster.
Server NameIP AddressOSRolesCPURAMHDD
CKA1001192.168.0.175Ubuntu 26.04 LTSControl Plane Node2 Core4 GB50 GB
CKA1001-NODE01192.168.0.176Ubuntu 26.04 LTSWorker Node2 Core4 GB50 GB
CKA1001-NODE02192.168.0.177Ubuntu 26.04 LTSWorker Node2 Core4 GB50 GB
CKA1001-NODE03192.168.0.178Ubuntu 26.04 LTSWorker Node2 Core4 GB50 GB

Before you start

  1. Four domains, unevenly weighted

    KCNA is a multiple-choice exam, 90 minutes, and unlike CKA there is no live cluster to operate. That changes what preparation means: you are not drilling speed, you are building a vocabulary broad enough that nothing in the ecosystem is a surprise.

    The published weights, which changed on 24 November 2025 - if you are revising from anything older you will see five domains and a different split:

    | Domain | Weight | Competencies | |---|---|---| | Kubernetes Fundamentals | 44% | Core concepts, administration, scheduling, containerization | | Container Orchestration | 28% | Networking, security, troubleshooting, storage | | Cloud Native Application Delivery | 16% | Application delivery, debugging | | Cloud Native Architecture | 12% | Observability, ecosystem and principles, community and collaboration |

    Nearly half the exam is Kubernetes fundamentals. That is the practical headline: if Pods, Deployments, Services, ConfigMaps and the API model are solid, you are close to halfway there. Container orchestration is the next 28%, and it is where the exam gets specific - networking, storage, security and troubleshooting, each with its own vocabulary.

    Two things about that table are worth noticing. Observability used to be a domain of its own at 8%; it is now one competency inside a 12% domain, so it is worth less than it looks in older study guides. And cluster installation appears nowhere at any weight - that is CKA's territory. You need a cluster to run commands against, not the ability to build one.

    The 44% is not abstract, either. It is very nearly these two API groups: the core group and apps. Every object in the largest domain is in that list.

    bash Example session
    kubectl --context cka1001 api-resources --api-group='' --no-headers | awk '{print $1}' | tr '\n' ' 'bindings componentstatuses configmaps endpoints events limitranges namespaces nodes persistentvolumeclaims persistentvolumes pods podtemplates replicationcontrollers resourcequotas secrets serviceaccounts serviceskubectl --context cka1001 api-resources --api-group=apps --no-headers | awk '{print $1}' | tr '\n' ' 'controllerrevisions daemonsets deployments replicasets statefulsets

    Expected resultThe core group - Pods, Services, ConfigMaps, Secrets, Namespaces, Nodes - and the five workload controllers in apps.

    Success conditionYou can read the largest exam domain as a list of API resources.

  2. How much surface area there is

    Two numbers make the scope concrete. This cluster answers for 161 API resource types, of which 71 are CustomResourceDefinitions - added by Calico, the Prometheus Operator, Argo CD and the CSI drivers.

    So more than 40% of what this cluster knows about is not Kubernetes at all. It is the ecosystem, installed on top. That ratio is the best argument for why "Cloud Native" is a separate exam domain from "Kubernetes": in any real cluster, most of the API is other people's software.

    It is also a warning about study material. A tutorial written against a bare cluster describes maybe a third of what you will actually meet.

    bash Example session
    kubectl --context cka1001 api-resources --no-headers | wc -l161kubectl --context cka1001 get crd --no-headers | wc -l71kubectl --context cka1001 api-versions | head -8admissionregistration.k8s.io/v1apiextensions.k8s.io/v1apiregistration.k8s.io/v1apps/v1argoproj.io/v1alpha1authentication.k8s.io/v1authorization.k8s.io/v1autoscaling/v1

    Expected resultA resource-type count around 160, a CRD count around 70, and API groups including third-party ones like argoproj.io.

    Success conditionYou can see how much of the API surface came from installed software.

  3. Where KCNA sits among the five

    There are five Kubernetes certifications and they are not a ladder:

    • KCNA - associate, multiple choice. Breadth across the ecosystem.
    • KCSA - associate, multiple choice. Security concepts and threat models.
    • CKA - performance based. Operate and repair a cluster.
    • CKAD - performance based. Build and ship workloads onto one.
    • CKS - performance based, requires CKA. Harden one under attack.

    KCNA before CKA is a reasonable order but not a requirement, and it is not a prerequisite for anything. Its real value is different: it forces you to learn the *names* - what Argo does, what a service mesh is for, what OpenTelemetry collects - which makes every later exam's material land faster because you already know where each piece fits.

    What KCNA will not do is make you able to fix a cluster. That is CKA, and no amount of multiple choice substitutes for having broken and repaired one yourself.

    One last look at where the ecosystem actually comes from: group the 71 CRDs by the top level of their domain and you get 33 under .io, 28 under .org, 10 under .com. Every one is a vendor or project that agreed to extend the same API rather than build a parallel one. That agreement is what "cloud native" names.

    bash Example session
    kubectl --context cka1001 get crd --no-headers | awk '{print $1}' | sed 's/.*\.//' | sort | uniq -c | sort -rn | head -6     33 io     28 org     10 com

    Expected resultCRD counts grouped by domain suffix.

    Success conditionYou can see how many distinct third parties extended this cluster.

Official sources