SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
KCNA · Domain 3

Cloud Native Application Delivery practice questions

Cloud Native Application Delivery is worth 16% of the KCNA exam — the 3rd-heaviest of the 4 domains. Delivering and debugging applications on cloud-native platforms. Official weighting 16%. 6 fully worked examples are further down this page, answers included.

Exam weight
16%
the 3rd-heaviest of the 4 domains
Questions
62
across 2 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

5 free questions left today.

Mode

Ready when you are

10 fresh questions drawn across 1 of 4 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Cloud Native Application Delivery questions, fully explained

Questions from the KCNA bank mapped to domain 3, with the answer key and the reasoning behind every option. None of them repeat the examples on the main KCNA practice page.

Question 1Cloud Native Application Delivery

A team wants two complete, independently running environments, old and new, and plans to switch all production traffic from one to the other in a single cut-over. Which deployment pattern are they describing?

Choose one.

  • a
    RollingUpdate

    RollingUpdate replaces Pods gradually within a single Deployment; it does not run two full separate environments.

  • b
    Blue-green Correct

    Blue-green runs the old (blue) and new (green) versions as two full environments and switches traffic from one to the other in a single cut-over, typically by updating a Service selector or load balancer target.

  • c
    Canary

    Canary exposes a small percentage of traffic to the new version first rather than switching all traffic at once.

  • d
    Recreate

    Recreate is a Deployment strategy field value that tears down old Pods before creating new ones inside one Deployment, not a two-environment pattern.

The concept

Blue-green deployment is a release pattern, not a built-in Kubernetes Deployment strategy value.

Why that’s the answer

In blue-green, both versions are deployed and fully scaled at the same time as separate resources, for example two Deployments, and traffic is redirected all at once, often by changing which Pods a Service's label selector matches. This gives an instant, easily reversible cutover, at the cost of running double the capacity briefly.

How to reason it out
  1. Deploy the new (green) version alongside the running (blue) version, fully scaled.
  2. Validate the green environment separately from production traffic.
  3. Switch the Service or load balancer to point at green all at once, and keep blue available for a fast rollback.

Exam tip: Blue-green is an application-level pattern built from ordinary Kubernetes objects, not a Deployment strategy field value.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

Question 2Cloud Native Application Delivery

A team wants each release to reach about 5% of users first, then grow in steps while error rates stay low, and stop automatically if they rise. Which approach provides this on Kubernetes?

Choose one.

  • a
    Set the Deployment strategy type to Canary with a 5% step

    The Deployment strategy field accepts only RollingUpdate and Recreate. There is no Canary value, so this manifest would be rejected.

  • b
    Use RollingUpdate with maxSurge set to 5% of the replicas

    maxSurge only limits how many extra Pods exist during a rolling update. The rollout still proceeds as soon as new Pods are Ready, with no traffic percentage held and no metric check to stop it.

  • c
    Use Argo Rollouts to shift weighted traffic in analysed steps Correct

    Correct. Argo Rollouts is a progressive delivery controller: it sends a set weight of traffic to the new version (often through a service mesh or ingress), runs analysis against metrics at each step, and aborts if they fail.

  • d
    Use two Deployments and repoint the Service selector at once

    Repointing a Service selector moves all traffic in one switch. That is a blue-green release, not a small share that grows in steps.

The concept

A canary release sends a small share of real traffic to the new version and increases it only while health metrics stay good.

Why that’s the answer

Plain Deployments cannot hold a traffic percentage or watch error rates, and the strategy field has no Canary value. Weighted, stepwise traffic shifting with automatic abort comes from progressive delivery tooling such as Argo Rollouts, usually working with a service mesh (Istio, Linkerd) or ingress. maxSurge is a capacity limit, not a traffic split, and a selector switch is blue-green.

How to reason it out
  1. Notice the three requirements: a small first share, stepwise growth and automatic abort on bad metrics.
  2. Rule out the strategy field: only RollingUpdate and Recreate exist.
  3. Recognise that maxSurge controls Pod counts, not traffic weights or metrics.
  4. Choose the progressive delivery controller that combines weighted traffic with analysis.

Exam tip: Canary is a pattern; Kubernetes alone cannot hold a traffic percentage, so use a mesh or Argo Rollouts.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

Question 3Cloud Native Application Delivery

A CI script must wait until a Deployment rollout finishes and exit with a non-zero code if the rollout exceeds its progress deadline. Which kubectl command does this?

Choose one.

  • a
    kubectl get deployments

    kubectl get deployments prints a one-off snapshot of replica counts and exits successfully either way; it neither waits for the rollout nor fails when the progress deadline is exceeded.

  • b
    kubectl rollout history

    This lists past revisions of a Deployment; it does not report on the currently in-progress rollout.

  • c
    kubectl rollout status Correct

    Correct. kubectl rollout status deployment/NAME watches the rollout, returns 0 once it completes, and exits non-zero when the Deployment reports that it exceeded progressDeadlineSeconds.

  • d
    kubectl describe pod

    This shows detail about a single Pod, not the overall progress of a Deployment rollout.

The concept

kubectl rollout status is the standard way to watch a Deployment update reach completion.

Why that’s the answer

kubectl rollout status deployment/NAME attaches to the rollout and blocks until every replica is updated and available. If the Deployment's Progressing condition reports ProgressDeadlineExceeded, the command exits with a non-zero code, which is what lets a CI step fail the pipeline. get deployments only shows a snapshot, rollout history lists past revisions, and describe pod covers one Pod.

How to reason it out
  1. Identify the need: block until the rollout ends and signal failure through the exit code.
  2. Rule out snapshot and history commands, which return immediately.
  3. Choose kubectl rollout status, which waits and fails when the progress deadline is exceeded.

Exam tip: To gate a pipeline on a rollout, use kubectl rollout status: it waits and fails on a stalled rollout.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

Question 4Cloud Native Application Delivery

Which kubectl command reverts a Deployment to its previous revision?

Choose one.

  • a
    kubectl rollout restart

    This restarts the Pods of the current revision; it does not switch to a prior revision.

  • b
    kubectl apply -k

    This applies manifests built with Kustomize; it is unrelated to rolling back a Deployment revision.

  • c
    kubectl rollout undo Correct

    kubectl rollout undo deployment/NAME rolls the Deployment back to its previous revision by re-applying that revision's Pod template.

  • d
    kubectl scale

    This changes the replica count of a Deployment; it does not change which revision is running.

The concept

kubectl rollout undo is the command for reverting a Deployment to an earlier recorded revision.

Why that’s the answer

Because Deployments keep a history of ReplicaSets, bounded by revisionHistoryLimit, kubectl rollout undo can restore the Pod template of a previous revision, causing a new rollout back to that known-good state.

How to reason it out
  1. Identify that the current rollout is unhealthy.
  2. Run kubectl rollout undo deployment/NAME to revert to the previous revision.
  3. Kubernetes performs a new rolling update back to the prior Pod template.

Exam tip: kubectl rollout undo rolls a Deployment back by rolling forward to a previous revision's template.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

Question 5Cloud Native Application Delivery

In a GitOps setup, what is treated as the single source of truth for the desired state of a cluster?

Choose one.

  • a
    Snapshots of live cluster objects that a job exports to Git

    That copies actual state into Git, the reverse of GitOps. Live state is compared against the declared state in Git; it is never the source of truth.

  • b
    Declarative manifests versioned in a Git repository Correct

    Correct. GitOps keeps the whole desired state as declarative files in Git; every change is a commit, and agents reconcile the cluster to match.

  • c
    The CI pipeline definitions stored in the Git repository

    Pipeline files describe how to build and test artifacts. They do not declare which objects should run in the cluster.

  • d
    The release history Helm stores inside the cluster

    Helm's release records are live data kept in the cluster, not versioned desired state in Git, so a GitOps agent does not reconcile against them.

The concept

GitOps makes a Git repository of declarative configuration the single source of truth for desired state.

Why that’s the answer

The GitOps principles require desired state to be declarative and versioned, with Git the usual store, and pulled and reconciled by agents. Exported snapshots of live objects are actual state copied backwards, CI pipeline files describe builds rather than cluster state, and Helm release records live inside the cluster. Only declarative manifests versioned in Git are the desired state.

How to reason it out
  1. Separate desired state from actual state and from build tooling.
  2. Recall that GitOps stores desired state as declarative, versioned files in Git.
  3. Eliminate exported live state, CI pipeline files and in-cluster release records.

Exam tip: In GitOps, Git holds the desired state; the cluster is made to match it.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

Question 6Cloud Native Application Delivery

In a GitOps setup, which component fetches the desired state from the Git repository and applies it to the cluster?

Choose one.

  • a
    A CI job outside the cluster that holds a kubeconfig and runs kubectl apply

    That is the classic push model: an external pipeline stores cluster credentials and applies changes into the cluster. GitOps moves this work inside the cluster.

  • b
    A controller inside the cluster that watches Git and syncs each new commit Correct

    Correct. GitOps agents such as Argo CD and Flux run as controllers in the cluster, poll (or receive a webhook from) the repository, and apply what changed.

  • c
    An admission webhook in the API server that loads repository files per request

    Admission webhooks validate or mutate objects that clients submit. They do not fetch configuration from Git or create objects on their own.

  • d
    A server-side hook in the Git host that sends each merge to the API server

    A Git server hook that calls the API server is still a push from outside the cluster, with cluster credentials held by the Git host.

The concept

In the GitOps pull model, an agent running inside the cluster pulls desired state from Git and reconciles the cluster to it.

Why that’s the answer

The defining trait of GitOps delivery is that the cluster pulls: a controller such as Argo CD or Flux runs in the cluster, watches the repository and applies changes, so no outside system needs write credentials to the cluster. A CI job or a Git server hook that calls the API server is a push from outside, and admission webhooks only check or modify objects that someone else submits.

How to reason it out
  1. Ask where the component runs: inside or outside the cluster.
  2. Ask which direction the change flows: pulled by the cluster or pushed into it.
  3. Rule out admission webhooks, which never originate objects.
  4. Pick the in-cluster controller that watches Git.

Exam tip: GitOps agents run inside the cluster and pull from Git; anything that pushes from outside is the CI model.

Cloud Native Application Delivery: GitOps, Helm, and Deployment Strategies — the lesson that teaches this.

What KCNA domain 3 tests, topic by topic

The official exam guide breaks Cloud Native Application Delivery into 2 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published KCNA practice questions per topic in Cloud Native Application Delivery
TopicWhat it coversQuestions
Application DeliveryOfficial KCNA competency (Cloud Native Application Delivery). Delivering applications on Kubernetes: deployment strategies and rollouts, GitOps and CI/CD fundamentals, package management with Helm, and the application definition and image-build landscape.31
DebuggingOfficial KCNA competency. Debugging application delivery: inspecting Deployments and rollouts, diagnosing failed releases, and using observability signals during delivery.31
Total62

Revise Cloud Native Application Delivery before you drill it

Other KCNA domains

Cloud Native Application Delivery: your questions

Cloud Native Application Delivery is domain 3 of the KCNA exam guide and carries 16% of the scored content — the 3rd-heaviest of the 4 domains. On a 60-question paper that works out to roughly 10 questions, though CNCF does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official KCNA exam guide.