Here are 20 free KCNA sample questions with answers, taken from the GoldenKube question bank and grouped by exam domain. Every question has the correct answer marked and a written explanation, so you can check what you know and see why the wrong options fail. Kubernetes and Cloud Native Associate is a multiple-choice exam; this set is independent practice and not affiliated with the Linux Foundation.
How to use these KCNA practice questions
Work through the 20 KCNA sample questions in one sitting without looking anything up. For each question, decide on an answer first, then open the answer and read the explanation, including when you were right, because a correct guess and a correct line of reasoning look identical until the exam rewords the question. Mark every question as right, wrong or guessed, and count a guess as wrong when you score yourself. Give the set about 30 minutes, which is the average time per question on the real exam multiplied by twenty. The aim is to find out which exam domains are already solid and which ones need study before you buy the exam.
The official exam format is 60 questions in 90 minutes, with a passing score of 75%, according to the Linux Foundation documentation checked on 2026-10-23.
With JavaScript enabled, each answer starts collapsed behind a Show answer button so you can test yourself. Without JavaScript, and for any crawler or screen reader that reads the raw page, every answer and explanation is already visible below its question.
These questions were written for GoldenKube practice. They are not copied from the real exam, and nothing on this page tells you what the real exam will ask.
Read the question twice and underline the constraint words: "most", "least", "first", "only", "not".
Eliminate options you can prove wrong before you pick between the rest.
Where a question says to select all that apply, every correct option must be chosen and none of the wrong ones.
Read the explanation even for questions you got right; it often names the nearby concept the exam likes to swap in.
Kubernetes Fundamentals (9 questions)
Question 1
What happens during preemption in Kubernetes scheduling?
A.High priority pods can cause lower priority pods to be evicted to make room
B.Nodes preempt pods when they are cordoned
C.System pods preempt user pods regardless of priority
D.Pods preempt network bandwidth from other pods
Correct answer: A
If a high-priority pod cannot be scheduled due to lack of resources, the scheduler can preempt (evict) lower-priority pods to free up space for the high-priority pod.
Question 2
What defines a standard format for declaring Kubernetes objects?
A.XML
B.CSV
C.YAML
D.TOML
Correct answer: C
YAML (and JSON) is the standard language used to define Kubernetes manifests, outlining the desired state of resources like Pods, Services, and Deployments.
Question 3
What happens when a pod's liveness probe fails?
A.The pod is marked as unhealthy but continues running
B.The container is restarted
C.The pod is deleted
D.Nothing, it's just a warning
Correct answer: B
When a liveness probe fails repeatedly, Kubernetes restarts the container. Readiness probes stop traffic but don't restart. Startup probes protect slow-starting containers.
Question 4
Which statement about container images and security is true?
A.Containers are inherently secure and do not require scanning
B.Only the base OS layer needs to be scanned for vulnerabilities
C.Image security scanning should be integrated into the CI/CD pipeline to detect known vulnerabilities
It is a best practice to regularly scan container images for known Common Vulnerabilities and Exposures (CVEs) as part of the CI/CD process before deploying them to Kubernetes.
Question 5
What happens when a startup probe fails?
A.The container is marked unhealthy but continues running
B.The container is killed and restarted according to restart policy
C.The pod is deleted immediately
D.Nothing, it's just logged
Correct answer: B
Startup probes protect slow-starting containers. If startup probe fails after all retries, the container is killed and restarted. Liveness and readiness probes are disabled until startup succeeds.
Question 6
Which Kubernetes component manages core control loops, such as the node controller and replicaset controller?
A.kube-apiserver
B.kube-scheduler
C.kube-controller-manager
D.kube-proxy
Correct answer: C
The kube-controller-manager runs controller processes. These are background threads that handle routine tasks in the cluster, like responding when nodes go down or maintaining correct replica counts.
Question 7
What is the default Service type in Kubernetes when no type is specified?
A.NodePort
B.LoadBalancer
C.ClusterIP
D.ExternalName
Correct answer: C
ClusterIP is the default Kubernetes Service type. It exposes the service on a cluster-internal IP, making the service only reachable from within the cluster.
How do you safely back up a Kubernetes cluster's state?
A.By backing up the node filesystems
B.By backing up the etcd database
C.By exporting all pods using kubectl get pods -o yaml
D.By copying the kube-apiserver logs
Correct answer: B
Since etcd holds all cluster data, state, and secrets, backing up the etcd data directory or taking an etcd snapshot is the correct way to back up a Kubernetes cluster.
Question 9
Which Kubernetes object automatically manages the lifecycle of pods in a deployment?
A.Service
B.ReplicaSet
C.ConfigMap
D.Namespace
Correct answer: B
Deployments create and manage ReplicaSets, which in turn manage pods. The ReplicaSet ensures the desired number of pod replicas are running. Deployments handle rolling updates by creating new ReplicaSets.
Container Orchestration (6 questions)
Question 10
What is the purpose of the Horizontal Pod Autoscaler (HPA) in Kubernetes?
A.To add more nodes to the cluster
B.To automatically scale the number of Pod replicas based on observed metrics
C.To vertically increase CPU and memory of existing Pods
D.To distribute Pods evenly across namespaces
Correct answer: B
The Horizontal Pod Autoscaler automatically scales the number of Pod replicas in a Deployment, ReplicaSet, or StatefulSet based on observed CPU utilization, memory utilization, or custom metrics. It adjusts the replica count to match the target metric value.
What scheduler task is NOT a primary responsibility of the Kubernetes scheduler?
A.Selecting which node a pod should run on
B.Monitoring pod health
C.Considering resource requirements when placing pods
D.Respecting node taints and pod tolerations
Correct answer: B
The scheduler assigns pods to nodes. It doesn't monitor pod health (kubelet does that via probes). Scheduler responsibilities:
- Select nodes for pods
- Consider resource requirements
- Respect taints, tolerations, and affinity rules
Question 12
What is the primary difference between a Deployment and a StatefulSet?
A.Deployments are for stateless apps; StatefulSets provide stable network identity and persistent storage
What happens if a pod doesn't have a toleration for a node's taint?
A.The pod runs anyway
B.The pod is scheduled but runs slowly
C.The pod cannot be scheduled on that node
D.The taint is removed
Correct answer: C
Taints repel pods. Without matching tolerations, the scheduler skips tainted nodes. This is how you dedicate nodes to specific workloads or prevent certain pods from running.
Question 14
Which of the following describes a 'local' persistent volume?
A.A volume provisioned on the developer's laptop
B.A volume that stores data in the Pod's memory (tmpfs)
C.A volume representing a mounted local disk directly attached to a single Kubernetes node
D.A cloud-based block storage volume attached to the cluster
Correct answer: C
A 'local' PV represents a mounted local storage device such as a disk, partition, or directory on a specific node. The scheduler ensures pods using the PVC are scheduled to that specific node.
Question 15
What is required for a Pod to use a Secret as a volume?
A.The Secret must be encrypted with a KMS provider
B.The Pod and the Secret must reside in the same Namespace
C.The Pod must be running with cluster-admin privileges
D.The Secret must be explicitly defined in a StorageClass
Correct answer: B
Secrets are namespaced resources. A Pod can only reference and mount a Secret if both the Pod and the Secret exist within the same namespace.
Cloud Native Application Delivery (3 questions)
Question 16
What is the main advantage of GitOps?
A.Faster container startup times
B.Using Git as the single source of truth for infrastructure and applications
C.Reduced storage costs
D.Improved network performance
Correct answer: B
GitOps core principles:
- Git is the source of truth
- Declarative configuration
- Automated synchronization
- Version control for infrastructure
- Audit trail via Git history
Question 17
What deployment strategy gradually shifts traffic from old to new versions?
A.Blue-green deployment
B.Canary deployment
C.Rolling update
D.Recreate
Correct answer: B
Deployment strategies:
- **Rolling update**: Gradual replacement (default)
- **Blue-green**: Complete switch between environments
- **Canary**: Gradual traffic shift (e.g., 10% → 50% → 100%)
- **Recreate**: Delete all, then create new
Canary reduces risk by exposing changes to small percentage first.
Question 18
What is an advantage of Gateway API over Ingress?
A.It only works with HTTP
B.It supports multiple protocols (HTTP, TCP, UDP, gRPC) and has role-oriented design
C.It requires less configuration
D.It uses less memory
Correct answer: B
Gateway API advantages:
- Multi-protocol: Not just HTTP/HTTPS
- Role-oriented: Infrastructure vs. application teams
- Expressive: Built-in advanced features
- Portable: Less vendor lock-in
- Extensible: Standardized extension mechanism
Cloud Native Architecture (2 questions)
Question 19
Which cloud-native principle emphasizes designing for failure?
A.Monolithic architecture
B.Manual intervention
C.Resilience and fault tolerance
D.Single point of failure
Correct answer: C
Cloud-native resilience principles:
- Design for failure (expect components to fail)
- Circuit breakers
- Retry and timeout strategies
- Graceful degradation
- Health checks and auto-healing
Avoid single points of failure.
Question 20
Which of the following are commonly referred to as the 'three pillars of observability' in cloud native systems?
A.Monitoring, Alerting, and Dashboards
B.Metrics, Logs, and Traces
C.Availability, Reliability, and Scalability
D.Events, Alerts, and Incidents
Correct answer: B
Metrics, logs, and traces are widely recognized as the three fundamental pillars of observability. Together, they provide a comprehensive view of system health, allowing engineers to understand both what went wrong and why.
Score yourself
The Linux Foundation lists a passing score of 75% for KCNA. On a set of 20 questions that is 15 correct. Reaching 15 here is a useful signal and not a prediction: 20 questions is a small sample, and the real exam covers every domain in its published weights. Use the table below to decide what to do next. Count only the questions you answered correctly without help, and be strict about it: a question you got right for the wrong reason belongs in the missed pile. Then repeat the exercise a few days later with a different set from the full bank so that you are measuring understanding and not memory of these specific questions.
Count your misses per domain, using the headings above. One missed question in a domain is noise; three misses in one domain is a study plan. The practice mode on GoldenKube does this sorting for you by sending you back to the domains where you are weakest.
Correct out of 20
What it suggests
What to do next
18 to 20
The basics are solid in the domains sampled.
Move to harder mixed practice and check the domains this set did not cover.
15 to 17
Close to the benchmark, with some gaps.
Review every missed explanation, then practise the domains where you missed questions.
11 to 14
Several domains need real study.
Read the official curriculum for those domains, then return to practice.
0 to 10
You are early in your preparation.
Start from the curriculum and the project documentation, then retry this set.
Kubernetes Fundamentals: what it tests
Kubernetes Fundamentals is the largest KCNA domain and covers what a cluster is made of and how you describe work to it. Expect the control plane components (API server, etcd, scheduler, controller manager), the node components (kubelet, kube-proxy and the container runtime), and the core API objects: Pods, ReplicaSets, Deployments, StatefulSets, DaemonSets, Jobs, Services, ConfigMaps, Secrets and Namespaces. Questions usually describe a situation and ask which object or component is responsible, so knowing the single job of each piece matters more than memorising field names. Typical traps are swapping similar objects, such as a Deployment and a StatefulSet, or crediting the kubelet with a decision that the scheduler makes.
Study by tracing the request path: a manifest reaches the API server, is stored in etcd, is noticed by a controller, is placed by the scheduler and is started by the kubelet. If you can narrate that path from memory, most component questions reduce to recalling one step of it.
Container Orchestration: what it tests
Container Orchestration covers how containers are run, connected and kept alive once Kubernetes owns them. Expect container runtimes and the Container Runtime Interface, images and registries, container networking and the CNI plugin model, service discovery and DNS, storage through volumes and the CSI, scheduling, and the basics of securing workloads with RBAC and Pod security settings. The domain rewards clear boundaries between similar interfaces: CRI is for runtimes, CNI for networking and CSI for storage. A frequent trap is mixing up the different autoscalers, since the Horizontal Pod Autoscaler changes replica counts, while other autoscalers change resource requests or the number of nodes. The wording is often a scenario, so identify the layer involved before you read the answer options.
A useful drill is to take each of the three plugin interfaces and name one thing it standardises, one thing it does not, and one example implementation. Doing the same for the autoscalers takes five minutes and closes a common source of lost points.
Cloud Native Application Delivery: what it tests
Cloud Native Application Delivery is about how code gets from a repository to a running cluster and how releases are made safe. Expect continuous integration and continuous delivery, the idea of GitOps with a pull-based agent, packaging with tools such as Helm and Kustomize, and rollout strategies including rolling updates, blue-green and canary releases. Questions tend to test definitions and trade-offs, for example when a canary release is preferable to a blue-green switch, or what separates continuous delivery from continuous deployment. A common trap is describing GitOps as any pipeline that stores files in Git, when the defining feature is that an agent continuously reconciles the cluster toward what Git declares.
Write one sentence for each strategy that starts with the risk it reduces and ends with the cost it adds. The exam usually asks for the better fit for a stated risk, and that sentence is the answer.
Cloud Native Architecture: what it tests
Cloud Native Architecture is the broadest and most conceptual KCNA domain. Expect the role of the CNCF and its project maturity levels (sandbox, incubating and graduated), the characteristics of cloud native systems such as resilience, elasticity and loose coupling, autoscaling and serverless ideas, the three observability signals of metrics, logs and traces, and community topics such as roles and open standards. Because the domain is small, a handful of questions decide it. A frequent trap is attributing the wrong job to a well-known project, for example confusing a metrics system with a tracing system, so learn one sentence per major CNCF project and which signal it handles.
Skim the CNCF landscape once with a single purpose: for each category, remember one graduated project and what it does. That is usually enough for the architecture questions at this level.
What a 20-question KCNA sample can and cannot tell you
A set of 20 questions is a screening test, not a measurement. With only a few questions per domain, one lucky guess or one unlucky wording moves your result by five percentage points, and a domain with only two questions behind it says almost nothing about that domain's true strength. The set is also fixed, so after one read-through it measures your memory of these items and not your command of KCNA. Use it to choose where to start studying, not to decide whether you are ready, and then rely on the larger bank, where questions repeat only after a gap and the app tracks which domains stay weak.
A second limit is style. Written practice questions share one author voice, while a real exam draws on many writers and reads differently. Treat any practice score, here or elsewhere, as one input next to the official curriculum, hands-on experience with the technology, and the documentation of the project itself.
A one-week plan after the KCNA sample set
After the sample set, spend a week turning the result into a plan. Start with the domains where you missed questions and read the matching part of the official curriculum before you try more questions, because practice is far more useful once you can explain the topic. Return to practice in short daily sessions, review every missed question the next day, and leave the last day for a fresh mixed set. Keep a simple note of which concepts caused misses; the same three or four ideas usually explain most of them, and naming them is what turns practice into progress.
The largest KCNA domain is Kubernetes Fundamentals, at 44% of the published weights. If two domains tie for weakest in your tally, start with the heavier one, and keep the other 3 domains ticking over with short reviews.
Adjust the pace to your calendar. If you have more time, spread the same steps over two or three weeks and add a second mixed session at the end of each week. If you have less, keep days 1 to 3 and day 6 and accept a rougher picture. What matters is that every study session starts from a specific weak domain and ends with a check, not that the plan is followed to the letter.
Day
Task
1
Tally misses per domain and rank the domains from weakest to strongest.
2
Read the official curriculum and documentation for the weakest domain.
3
Practise that domain in the app and read every explanation.
4
Repeat the exercise for the second weakest domain.
5
Review the questions you missed on days 3 and 4 without looking at notes.
6
Do a mixed session across all domains and compare the result with this page.
7
Rest, then retake the 20 questions on this page and note which misses persist.
Practice the full KCNA bank
These 20 questions are a small slice of the GoldenKube KCNA bank, which holds more than 250 questions. In practice mode the app serves the questions you are due to review first, then the domain where you are weakest, and it brings missed questions back after a gap until you answer them correctly. You can answer 10 questions as a guest without an account; after that a free account keeps your progress, and no credit card is asked for at any point. Each answer comes with the same kind of written explanation you saw on this page, so a wrong answer is always followed by the reasoning that fixes it.
Practice mode shows one question at a time and gives the verdict and explanation as soon as you answer. A focus option lets you stay inside a single domain until it feels solid, which is the natural next step after finding your weakest domain on this page. There is no timer and no exam simulation in practice mode; the goal is learning, not performance under pressure.
No. They are practice questions written for GoldenKube to cover the published KCNA domains. GoldenKube does not have the real exam, does not claim any overlap with it, and does not sell or share exam content. Use them to find gaps in your understanding, not to predict what you will be asked.
How many KCNA questions does GoldenKube have?
The KCNA bank holds more than 250 questions as of 2026-10-23. This page shows 20 of them. The full bank is available in practice mode, where the app chooses what to show you next based on what you have already answered and what you got wrong.
Do the answers come with explanations?
Yes. Every question on this page has the correct answer marked and a written explanation of why it is correct and, where useful, why the other options are not. Only questions that have an explanation are published on public pages, and where a question carries references they are listed under the explanation.
Is this free?
Yes. Reading this page needs no account. In the practice app you can answer 10 questions as a guest, and then you create a free account to continue. No credit card is requested at any point.
How were these 20 questions chosen?
A script picks them with a fixed seed, so the set does not change between visits. It considers only questions that have a written explanation and tries to follow the published domain weights. The domain labels in the bank do not always match the official wording, so a domain can be thinly represented in a single set of 20; the full bank is the better measure of coverage.
How often is this page updated?
The exam facts and the domain list were last checked against the Linux Foundation pages on 2026-10-23, and the page shows its update date at the top. When the curriculum changes or questions are added, corrected or removed in the bank, the page is reviewed again.
Independent community project, not affiliated with or endorsed by CNCF or The Linux Foundation.