75+ Harness CD Interview Questions and Answers (2027)

80 Harness CD interview questions with answers: pipelines, services, environments, delegates, triggers, approvals, canary and blue green, GitOps, OPA.

Sep 25, 2025 - 12:49
Nov 7, 2026 - 00:05
 0  342
75+ Harness CD Interview Questions and Answers (2027)

Harness interview questions for continuous delivery roles test how you model a release in Harness CD: pipelines made of stages and steps, services deployed to environments and infrastructure definitions, delegates and connectors, triggers and approvals, deployment strategies such as rolling, canary and blue green, GitOps with Argo CD, and reusable templates. This page has 80 questions with direct answers, grouped by experience level and followed by scenarios.

The answers were checked against the Harness Developer Hub in November 2026. Harness is rolling out a new experience, documented as Harness 3.0, with a flatter v1 pipeline YAML and its modules regrouped into four product "Agents". The answers here use the NextGen pipeline model that existing accounts and most interview panels still use; Harness states that the new experience stays backward compatible with it. If the role also covers Harness CI, security scanning, feature flags, cost management or chaos testing, read the companion guide, Harness engineer interview questions, after this one.

Key takeaways

  • Freshers are asked to explain the CD building blocks: pipeline, stage, step, service, environment, infrastructure definition, connector, delegate and secret.
  • Candidates with 2 to 5 years of experience should be able to read pipeline YAML, use runtime inputs and expressions such as <+pipeline.variables.version>, and explain each deployment strategy and its rollback.
  • Senior candidates are tested on governance: RBAC, OPA policies, freeze windows, Git Experience, templates with versioning, and running many services across many environments.
  • The delegate makes only outbound HTTPS connections (port 443) to Harness, so no inbound firewall rule is needed. Know how delegate selectors and the upgrader work.
  • Harness GitOps is built on Argo CD. Flux is not supported as the reconciler.
  • Scenario questions decide most experienced-hire interviews. Practise describing symptom, cause, fix and prevention.

How to use this guide

Each answer starts with a short reply you can say in an interview, then adds the detail an interviewer usually asks next. For general rounds on Linux, Git, Docker and Kubernetes, pair this page with our 200+ DevOps interview questions guide.

Your experienceRead firstThen practise
Fresher or career switcherQuestions 1 to 25Create a free Harness account, install a Kubernetes delegate and deploy a sample app with a rolling deployment
2 to 5 yearsQuestions 26 to 62A pipeline with QA and prod stages, an approval, a canary in prod and a Git trigger
Senior or leadQuestions 63 to 70An OPA policy that blocks prod deployments without approval, plus a freeze window
Everyone with work experienceQuestions 71 to 80 (scenarios)Explain each scenario aloud in under two minutes

Harness CD interview questions for freshers

1. What is Harness CD?

Harness Continuous Delivery (CD) is the Harness module that deploys software to targets such as Kubernetes, VMs, serverless platforms and cloud services through pipelines. It models what you deploy (services), where you deploy (environments and infrastructure definitions) and how you deploy (execution steps and strategies), then adds approvals, verification, rollback and audit around each release.

2. What is a pipeline in Harness?

A pipeline is the end-to-end workflow for a release. Harness describes a CD pipeline as a series of stages, each deploying a service to an environment. Pipelines can also contain approval stages, custom stages and notifications, and they can be stored inline in Harness or as YAML in Git.

3. What is the difference between a stage and a step?

A stage is a major segment of the pipeline, usually one release milestone such as "deploy to QA", "approval" or "deploy to prod". A step is a single operation inside a stage, for example a Kubernetes rolling deployment, a shell script, an HTTP check or a Jira approval. Steps can be grouped into step groups and run in sequence or in parallel.

4. What is a Harness service?

A service is the logical representation of a microservice or workload you deploy. It holds the service definition: the deployment type (Kubernetes, Helm, ECS, Lambda and so on), the manifests, the artifact source and service variables. You create it once and reuse it in many pipelines and environments.

5. What is an environment?

An environment is a logical deployment target such as dev, QA or production. Environments have a type (Production or Pre-Production), which governance features like RBAC and freeze windows can use. An environment can also hold variables, config files and overrides that apply to every service deployed into it.

6. What is an infrastructure definition?

An infrastructure definition describes the physical target inside an environment: for Kubernetes, a cluster connector plus namespace and release name; for VMs, the hosts; for ECS, the cluster and region. One environment can have several infrastructure definitions, for example two clusters in different regions.

7. What is a Harness delegate?

The delegate is a service you install in your own network (in a Kubernetes cluster, in Docker or on a host) that performs tasks on behalf of Harness Manager: it pulls artifacts, applies manifests, runs scripts and calls your tools. Credentials and traffic to your infrastructure stay inside your network, and the delegate connects out to Harness to fetch work.

8. How does the delegate communicate with Harness?

It opens outbound HTTPS/WSS connections to Harness Manager on port 443. No inbound port needs to be opened in your firewall. The delegate also makes outbound calls to whatever providers your connectors point at, such as a cloud API, Git host or artifact registry.

9. What is a connector?

A connector stores the settings and credential references Harness needs to talk to a third-party system: a Git provider, Docker registry, Kubernetes cluster, cloud account, Jira, ServiceNow or a secret manager. Steps and services reference the connector instead of embedding credentials, so you can rotate a credential in one place.

10. How are secrets handled in Harness?

Secrets are stored in a secret manager and referenced by ID. Harness has a built-in secret manager and also integrates with external ones such as HashiCorp Vault, AWS Secrets Manager, Azure Key Vault and GCP Secret Manager. In YAML you reference a secret with an expression such as <+secrets.getValue("db_password")>, and Harness masks secret values in logs.

11. What are account, organization and project scopes?

Harness has three scope levels. The account is the top level, organizations group related projects (often a business unit), and projects hold pipelines and day-to-day resources for one team. Connectors, secrets, templates, delegates and environments can be created at any level and are visible to the scopes below, so shared resources usually live at account or org level.

12. What is a trigger?

A trigger starts a pipeline automatically. Harness documents four trigger types: Webhook (Git events and generic events), Artifact (a new artifact version in a registry), Manifest (a new Helm chart version) and Scheduled (cron). Custom webhook triggers let any system start a pipeline with a URL or cURL command.

13. What approval options does Harness provide?

Harness supports Harness Approval (named user groups approve in the UI), Jira Approval, ServiceNow Approval and Custom Approval (a script whose output decides approve or reject). Approvals can be steps inside a stage or a dedicated Approval stage. Jira and ServiceNow approvals poll the ticket until the approval criteria match.

14. Which deployment strategies does Harness support?

Rolling, canary, blue green and basic, plus multi-service deployments. Which strategies are available depends on the deployment type. Basic replaces everything at once and Harness advises against it for production.

15. What is a rolling deployment?

A rolling deployment replaces instances gradually, one by one or in batches, inside the same environment. For Kubernetes, Harness applies the manifests and lets the Deployment's rolling update replace pods. It needs no extra capacity, but old and new versions serve traffic at the same time during the rollout.

16. What is a canary deployment in Harness?

A canary deployment releases the new version to a small portion of the workload first, verifies it, then proceeds. For Kubernetes, Harness does this in two phases: it deploys a canary workload, verifies it and deletes it, then rolls the new version out to the primary workload with a rolling update.

17. What is a blue green deployment in Harness?

Two versions run side by side: the current one serves production traffic and the new one is deployed to the stage side. After verification, Harness swaps traffic to the new version. For Kubernetes, the swap changes the selectors on a primary service and a stage service, so it is fast and the rollback is another swap.

18. What is rollback in Harness?

Rollback returns the target to the last successful deployment. Each stage has rollback steps that run when the stage fails and the failure strategy says to roll back. For Kubernetes rolling deployments the rollback step restores the previous release; for blue green it swaps traffic back.

19. What is a failure strategy?

A failure strategy decides what Harness does when a step or stage fails. Options include manual intervention, retry, mark as success, ignore failure, abort, stage rollback and pipeline rollback. You can set strategies per error type. A step-level strategy overrides the stage-level one.

20. What is Continuous Verification?

Continuous Verification (CV) checks a deployment against metrics and logs from health sources such as Prometheus, Datadog, Dynatrace, New Relic, Splunk or Elasticsearch. The Verify step uses machine learning to compare the new version with a baseline and can fail the stage, which then triggers rollback, when it finds anomalies.

21. What is a runtime input?

A runtime input is a field marked <+input> whose value is supplied when the pipeline runs, either by the person running it, by a trigger or from an input set. It lets one pipeline deploy different artifact tags, services or environments without editing the YAML.

22. What are input sets and overlays?

An input set is a saved collection of runtime input values for a pipeline, for example "QA values" or "prod values". An overlay is a group of input sets applied together in order, so you can combine a base set with an environment-specific set.

23. What are Harness expressions?

Expressions reference values at runtime using the <+...> syntax. Common examples are <+pipeline.name>, <+pipeline.variables.version>, <+service.name>, <+env.name> and <+artifacts.primary.tag>. You can use them in manifests, values files, scripts and conditions.

24. What is a template?

A template is a reusable definition of a step, step group, stage or pipeline. Teams publish templates (often at account or org level) so every pipeline uses the same approved deployment logic. Templates are versioned, and a pipeline can pin a version or follow the version marked stable.

25. What is Harness GitOps?

Harness GitOps is the GitOps capability inside Harness CD, built on Argo CD. A GitOps Agent based on Argo CD runs in your cluster and syncs applications from Git to clusters, while Harness adds a central UI, RBAC, audit, and pipeline steps that update Git and trigger syncs. For background, see our post on what GitOps is.

Pipelines, services and environments (2 to 5 years experience)

26. Walk through the YAML of a simple Kubernetes deployment stage.

A Deployment stage references a service, an environment and an infrastructure definition, defines execution steps and rollback steps, and sets a failure strategy. The example below deploys with a rolling strategy and rolls back the stage on any error.

pipeline:
  name: deploy-web
  identifier: deploy_web
  orgIdentifier: default
  projectIdentifier: payments
  stages:
    - stage:
        name: Deploy to QA
        identifier: deploy_qa
        type: Deployment
        spec:
          deploymentType: Kubernetes
          service:
            serviceRef: web
          environment:
            environmentRef: qa
            deployToAll: false
            infrastructureDefinitions:
              - identifier: qa_cluster
          execution:
            steps:
              - step:
                  type: K8sRollingDeploy
                  name: Rolling Deployment
                  identifier: rolling_deploy
                  spec:
                    skipDryRun: false
                  timeout: 10m
            rollbackSteps:
              - step:
                  type: K8sRollingRollback
                  name: Rolling Rollback
                  identifier: rolling_rollback
                  spec: {}
                  timeout: 10m
        failureStrategies:
          - onFailure:
              errors:
                - AllErrors
              action:
                type: StageRollback

27. Where do the Kubernetes manifests and the artifact come from?

Both are defined in the service. Manifests (plain YAML, Helm charts, Kustomize or OpenShift templates) usually come from a Git connector or a Helm repository. The artifact source points at a registry such as Docker Hub, ECR, GCR, ACR, Nexus or Artifactory. Harness resolves the tag at runtime and makes it available to manifests through expressions.

28. How do you pass the image tag into a Kubernetes manifest?

Reference the artifact in the values file with an expression, for example image: <+artifacts.primary.image>, and make the tag a runtime input in the service so the trigger or the user supplies it. Harness renders the values file before applying the manifests, so the Deployment receives the exact image chosen for that run.

29. How do you deploy the same service to QA and prod with different settings?

Keep one service and use environment-level configuration: environment variables, environment-specific values files and service overrides. The pipeline has one stage per environment, and each stage resolves the overrides for its environment, so prod can get more replicas or a different config map without copying the service.

30. What are service overrides and what is the priority order?

Overrides replace service values such as variables, manifests (values files, Kustomize patches) and config files for a given environment or infrastructure. With Overrides V2, which Harness lists as a Beta feature that must be enabled, the priority from highest to lowest is: service and infrastructure specific, infrastructure specific, service specific, then global environment. Across scopes, project beats organization, which beats account.

31. What are pipeline, stage and service variables?

Variables hold values you reuse. Pipeline variables are referenced as <+pipeline.variables.NAME>, stage variables as <+stage.variables.NAME>, and service variables are defined in the service and can be overridden per environment. Account and org level variables exist for values shared by many projects.

32. How do you choose which delegate runs a task?

Use delegate selectors. You add tags to delegates and then reference those tags in connectors or in a step's Delegate Selector field. Harness only sends the task to delegates that match. Without selectors, Harness picks from eligible delegates in scope that can reach the target.

33. What is the difference between installing a delegate with Helm, Kubernetes YAML and Docker?

All three run the same delegate. Helm and Kubernetes manifests install it as a Deployment in a cluster, which suits most teams because it can scale replicas and auto-upgrade. The Docker install runs a single container on a VM. Harness does not provide auto-upgrade for Docker delegates, so you upgrade them yourself.

34. How do delegate upgrades work?

For Kubernetes and Helm installs, an upgrader cron job runs every hour by default. It asks Harness Manager which delegate version is published for the account and, if the running image differs, updates the Deployment image, which triggers a rolling update so running tasks can finish. Delegate images use a yy.mm.verno tag. You can suspend the cron job to pin a version.

35. How do you run a stage only for some branches or conditions?

Use conditional execution on the stage or step. You can run it only when the previous stage succeeded, always, or when a JEXL condition is true, for example when <+trigger.targetBranch> equals main. The failure strategy takes precedence over conditional execution when a step fails.

36. What looping strategies are available?

Matrix, repeat and parallelism. Matrix runs a stage or step once per combination of values, repeat runs it a number of times or once per item in a list, and parallelism runs N copies. The example deploys to two regions with at most two running at once.

strategy:
  matrix:
    region:
      - us-east-1
      - eu-west-1
  maxConcurrency: 2

Inside the stage, refer to the current value with <+matrix.region>.

37. How do you deploy several services in one stage?

Use a multi-service deployment, where the stage lists several services and Harness deploys each one, in parallel or in sequence. You can also deploy one service to several environments or infrastructures in the same stage. For large estates this keeps the pipeline count low.

38. What is pipeline chaining?

Pipeline chaining runs one pipeline as a stage of another. A parent pipeline can call a build pipeline and then a deploy pipeline, passing inputs and reading the child's outputs after its stage completes. Nested chaining (a child that chains another pipeline) is not supported.

39. What are barriers, resource constraints and Queue steps?

Barriers make parallel stages or step groups wait for each other at a point. Resource constraints stop two deployments to the same service and infrastructure from running at the same time; Harness adds them to deployment stages by default. Queue steps order executions that share a resource key, for example a shared test database.

40. How do you share data between steps?

Steps publish output variables that later steps read with expressions. For example, a Shell Script step can export a variable and a later step reads <+execution.steps.STEP_ID.output.outputVariables.NAME>. Built-in steps also expose outputs, such as the resolved artifact or the deployed release name.

Deployment strategies and verification (2 to 5 years experience)

41. Which steps make up a Kubernetes canary deployment in Harness?

A canary phase with a Canary Deployment step (which deploys a set number or percentage of pods), usually a Verify step, then a Canary Delete step that removes the canary workload, followed by a primary phase with a Rolling Deployment step. Rollback steps include Canary Delete and Rolling Rollback.

42. Does a Harness Kubernetes canary split traffic by percentage?

Not by itself. The canary pods carry the same labels as the primary pods, so the Kubernetes Service sends them a share of traffic roughly in proportion to pod count. For exact percentage routing you add traffic routing through a service mesh or ingress controller, which Harness supports for some providers.

43. Which steps make up a Kubernetes blue green deployment?

Stage Deployment deploys the new version to the stage side, an optional verification step checks it, Swap Primary with Stage switches the selectors so production traffic goes to the new version, and an optional scale-down step removes the old pods later. The swap step does not run its own health check, so add a Verify or HTTP step after it if you want one.

44. How does Harness identify the primary and stage services in blue green?

With two Kubernetes Services in the manifests, annotate one with harness.io/primary-service: "true" and the other with harness.io/stage-service: "true". With a single Service and no annotations, Harness creates the stage copy for you with a -stage suffix.

45. When would you choose blue green over canary?

Choose blue green when you need an instant, all-at-once switch and an instant rollback, and you can afford double capacity during the release. Choose canary when you want to expose a small share of users first and watch metrics before committing. Our post on blue green versus canary releases covers the trade-offs in more depth.

46. How do you configure the Verify step?

Pick a monitored service with one or more health sources, choose the verification type (for example canary, rolling, blue green or load test), set sensitivity (High, Medium or Low) and set duration. Harness suggests around 10 minutes for log providers and 15 minutes for APM and infrastructure providers. High sensitivity fails on even small anomalies.

47. Your Verify step fails although the release looks healthy. What do you check?

Check the baseline (was the previous version healthy?), the metric queries in the health source (do they filter to the right service and environment?), sensitivity (High may be too strict for a noisy metric) and duration (too short means too few data points). Per-metric sensitivity overrides help when one noisy metric drives failures.

48. What happens to the pipeline when verification fails?

The Verify step fails and the stage's failure strategy decides the outcome. Commonly the strategy is stage rollback, so Harness runs the rollback steps. You can also choose manual intervention so an engineer decides, which is useful while you are still tuning health sources.

49. How does Harness deploy Helm charts?

You can use the Kubernetes deployment type with a Helm chart as the manifest, where Harness renders the chart and applies the output with its own steps (canary and blue green are available), or the Native Helm deployment type, where Harness runs Helm commands and the release is tracked by Helm. The choice affects which strategies and rollback behaviour you get.

50. Which non-Kubernetes deployment types do you know?

Harness CD lists, among others, Native Helm, Amazon ECS, Amazon AMI/ASG, AWS Lambda, AWS SAM, Azure Web Apps, Azure Functions, Google Cloud Functions, Google Cloud Run, Google Managed Instance Groups, Tanzu Application Service, the Serverless.com Framework, SSH and WinRM deployments to VMs, and custom deployments. Strategy options vary by type.

51. How do you run Terraform or other infrastructure steps in a CD pipeline?

Harness CD has infrastructure provisioning steps for Terraform (plan, apply, destroy, rollback), Terraform Cloud, CloudFormation, Azure ARM and Blueprint, Terragrunt and shell script provisioning. A common pattern is to provision a namespace or environment in the infrastructure section of the stage, then deploy the service into it.

52. What is the difference between a deployment pipeline and a GitOps pipeline?

In a standard deployment pipeline the delegate pushes changes to the target by running steps. In a GitOps flow the pipeline changes Git (for example updates an image tag in a values file or opens a pull request), and the GitOps Agent pulls that change into the cluster. GitOps keeps Git as the source of truth and corrects drift automatically.

GitOps, templates, triggers and approvals (2 to 5 years experience)

53. What are the main Harness GitOps entities?

GitOps Agent (the Argo CD based worker in your cluster), repositories (Git or Helm sources), clusters (deployment targets), applications (source, target cluster and namespace, sync policy) and ApplicationSets. Harness also links GitOps applications to Harness services and environments so releases show up in deployment dashboards.

54. Can you bring an existing Argo CD into Harness?

Yes. Harness documents a bring-your-own Argo CD option, where the agent uses the existing repo server, Redis cache, application controller and ApplicationSet controller in your cluster. Be careful during cleanup: removing the Argo CD CRDs deletes Argo CD objects such as applications and ApplicationSets. Our Argo CD interview questions cover the Argo side.

55. How do ApplicationSets work in Harness GitOps?

You select the GitOps agent that manages the ApplicationSet, then define a generator and a template. Harness supports the Argo CD generator types, including List, Cluster, Git, Matrix and Pull Request. Template labels such as harness.io/serviceRef and harness.io/envRef can map generated apps to Harness services and environments.

56. How do you promote a release across environments with GitOps?

Store one values file or Kustomize overlay per environment in Git. A pipeline updates the dev file, syncs and verifies, then waits for approval before updating the prod file and syncing the prod application. Harness has GitOps steps for updating release repos, merging pull requests and syncing applications, so the promotion is audited in one pipeline.

57. What is the difference between template versions and the stable version?

Each template can have several versions. A pipeline either links to a specific version or always uses the stable one. When you mark a new version stable, pipelines linked to stable pick it up. Changes that alter the template's inputs need a new version; other edits can go into the same version.

58. How do you roll out a template change safely?

Create a new version, test it in a non-critical pipeline that pins that version, then mark it stable. For Git-stored templates, keep changes on a branch and merge when tested. After changing inputs, reconcile the pipelines that use the template so their YAML matches the new input set.

59. How do you set up a Git trigger that deploys on merge to main?

Create a Webhook trigger for your Git provider, choose the push event (or pull request closed and merged), set a condition on the target branch, and map trigger payload values to the pipeline's runtime inputs, for example the commit SHA as the image tag. Harness can register the webhook automatically when the connector has the right permissions.

60. How do artifact triggers work?

An Artifact trigger watches a registry path and starts the pipeline when a new tag appears. Harness polls artifacts at one-minute intervals, and the docs note a delay of five to ten minutes after creating or updating a trigger before polling starts. The new tag is available in the pipeline as a trigger expression.

61. How do you require a Jira approval before production?

Add a Jira Approval step (or a Jira Approval stage) before the prod stage. Configure the Jira connector, the issue key (often created earlier with a Jira Create step and passed by expression) and approval criteria, for example status equals "Approved". Optionally set rejection criteria so a "Rejected" status fails the pipeline immediately.

62. What is a ServiceNow change window in an approval?

The ServiceNow Approval step can enforce the change window from the ticket, so the pipeline proceeds only when the current time is inside the approved window. The start and end times use the time zone of the ServiceNow account behind the connector. For when gates help and when they slow teams down, see our post on approval gates in CI/CD.

Governance and scale (senior)

63. How does Harness RBAC work?

Access is granted through role bindings that combine three parts: a principal (user, user group or service account), a role (a set of permissions) and a resource group (all resources of a type, or named resources). Bindings apply at account, org or project scope. Harness RBAC is additive with no explicit deny, so effective access is the union of all bindings.

64. How would you restrict who can deploy to production?

Create a resource group containing only production environments, a custom role with pipeline execute and environment access permissions, and bind it to a small prod-deployers user group. Everyone else gets a role scoped to pre-production environments. Add a Harness Approval for prod and an OPA policy as a second control.

65. What is policy as code in Harness and when is it evaluated?

Harness uses Open Policy Agent (OPA) with policies written in Rego. Policies go into policy sets, and each set is tied to an entity type and an event. For pipelines the events are On Save, On Run (evaluated after preflight checks) and On Step Start. A set can warn or block when a policy fails.

66. Write a policy that blocks prod deployments without an approval stage.

The policy reads the pipeline JSON and denies it if a Deployment stage targets the prod environment and no Approval stage exists. Attach it to a policy set for pipelines on On Run (and On Save, if pipelines are edited in the UI).

package pipeline

deny[msg] {
  stage := input.pipeline.stages[_].stage
  stage.type == "Deployment"
  stage.spec.environment.environmentRef == "prod"
  not has_approval
  msg := sprintf("Stage '%s' deploys to prod without an approval stage", [stage.name])
}

has_approval {
  input.pipeline.stages[_].stage.type == "Approval"
}

If pipelines are edited in Git, On Save does not block the Git sync, so Harness recommends On Run policies as well. More examples are in our post on OPA policy as code.

67. How do deployment freeze windows work?

A freeze window blocks CD stages for selected services, environments or pipelines during a time range, with optional recurrence (daily, weekly, monthly or yearly). Freezes can be set at account, org or project level and apply only to CD stages. Users with the Override permission under Deployment Freeze, and account admins, can still deploy. Notifications can go out when a freeze starts or blocks a run.

68. What is Git Experience and why use remote pipelines?

Git Experience lets you store pipelines, templates, input sets, services, environments and infrastructure definitions as YAML in your Git repository (remote) instead of inside Harness (inline). Changes go through pull requests and code review, and branches let you test pipeline changes. Admins can enforce Git Experience so new pipelines and templates must be remote.

69. How do you structure Harness for 200 microservices and 30 teams?

Use orgs per business unit and projects per team. Put shared connectors, secret managers, delegates and golden templates at account or org level. Each team builds pipelines from stage templates, stores YAML in Git, and inherits OPA policies and RBAC. Use multi-service deployments or matrix strategies to avoid a pipeline per service per environment.

70. What are the delegate operating practices you would enforce?

Run at least two delegate replicas per cluster or network zone, keep auto-upgrade on (or pin and upgrade on a schedule), tag delegates by network and purpose and use selectors in connectors, give the delegate service account only the permissions its deployments need, and alert on delegates that disconnect or expire. Harness publishes an expiration policy for old delegate versions, so track versions.

Scenario-based Harness CD interview questions

71. A pipeline fails with "no eligible delegates". How do you troubleshoot?

Check that a delegate is connected in the right scope, that its tags match the selectors in the step or connector, and that it can reach the target (network, DNS, proxy). Then check delegate logs for capability check failures. A common cause is a connector that inherits credentials from a delegate whose tag was changed.

72. A Kubernetes rolling deployment times out waiting for steady state. What do you do?

Read the step log for the pod events Harness prints, then check with kubectl describe pod and kubectl logs. Typical causes are image pull errors (wrong tag or registry secret), failing readiness probes, resource quotas or insufficient node capacity. Fix the cause, then let the failure strategy roll back or rerun from the failed stage.

73. Production broke after a canary passed verification. What went wrong and how do you prevent it?

The canary may have been too small to receive enough traffic, the health source queries may not have covered the failing endpoint, or the issue appeared only under full load. Increase canary size or add mesh-based traffic routing, add metrics for error rate and latency on key transactions, add a Verify step after the primary rollout, and keep rollback on failure.

74. Someone deployed to prod during a release freeze. How do you investigate?

Check the Audit Trail and the execution details to see who ran it and whether a freeze override was used. Users with the Override permission and account admins can bypass freezes, and a trigger whose API key has override permission can too. Remove override from roles that do not need it and require approval for prod.

75. A Git trigger fires twice for each merge. What is the likely cause?

Usually two webhooks exist for the same repository (one registered automatically, one manually), or the trigger listens to both the push event and pull request merged. Check the webhook list in the Git provider and the trigger's event and condition settings, and keep one webhook and one event.

76. A secret value appears in a third-party tool's output. How do you respond?

Rotate the secret in the secret manager first, since the value is exposed. Then find how it leaked: a script that echoed a transformed value (masking matches the exact value), a file written to an artifact, or a step that passed it to a tool that logs inputs. Fix the script and restrict who can view logs for that pipeline.

77. Blue green rollback did not restore the old version. Why?

If the old pods had already been scaled down after the swap, there was nothing to swap back to. Keep the old version running until post-swap verification passes, and only then add the scale-down step. Also confirm that the primary and stage services are annotated correctly so the swap targets the right objects.

78. Teams keep copying pipelines and they drift apart. How do you fix it?

Create stage and pipeline templates for the common flows, publish them at org level with a stable version, and migrate teams to them. Store templates and pipelines in Git, add an OPA policy that requires approved templates for prod stages, and use template versioning for changes.

79. A GitOps application stays OutOfSync after the pipeline updates Git. What do you check?

Confirm the pipeline committed to the branch and path the application tracks, that the agent can reach the repo (credentials, network), whether auto-sync is enabled, and whether a field mutated by a controller (for example replicas set by an autoscaler) needs an ignore-differences rule. Check the agent logs and the application's sync status details.

80. Deployments to three regions must run one at a time and stop on the first failure. How do you build it?

Use a matrix strategy over the regions with maxConcurrency: 1, or three sequential stages created from one stage template. Set the failure strategy to roll back the failed stage and abort the pipeline so later regions do not start. Add a Verify step in each region before moving on.

How to prepare for a Harness CD interview

Build one pipeline end to end: a service with a Helm chart, QA and prod environments, a Kubernetes delegate, a Git trigger, a Harness Approval, a canary in prod with a Verify step, and a stage template. Then break it on purpose (wrong tag, failing probe, missing selector) and practise explaining the fix. Our list of CI/CD pipeline examples helps with system design questions about release flow.

If you want guided practice, DevOps Training Institute runs classroom batches in Kharadi, Pune and live online classes across India. The DevOps training covers CI/CD and deployment strategies, and the CKA training builds the Kubernetes troubleshooting skills that most Harness CD scenarios depend on.

Official references

FAQs

Is Harness CD hard to learn for a Jenkins user?

No. The main shift is from scripting every step to modelling services, environments and infrastructure, then choosing built-in deployment steps. If you know Kubernetes and Helm, you can build a working canary pipeline in a few days of practice.

What is the difference between Harness CD and Harness GitOps?

Harness CD pipelines push changes to targets through the delegate. Harness GitOps uses an Argo CD based agent that pulls desired state from Git into clusters. Many teams use both: pipelines handle approvals and promotion, and GitOps handles the sync.

Do I need to learn Harness 3.0 for interviews in 2027?

Know that it exists: a flatter v1 pipeline YAML, typed inputs, steps that run as containers and a single lighter delegate. Harness states the new experience is backward compatible, and most current accounts and interview questions still use the NextGen model covered here.

Which deployment strategy is asked about most?

Canary and blue green. Be ready to name the Kubernetes steps for each, explain how traffic moves, and describe what rollback does in each case.

Does the Harness delegate need inbound firewall access?

No. The delegate makes outbound HTTPS connections to Harness on port 443. It also needs outbound access to the systems your connectors use, such as cloud APIs, Git and registries.

Is there a Harness certification?

Yes. Harness University runs a Continuous Delivery & GitOps track with Developer, Administrator and Architect levels. Certifications help show structured knowledge, but interviewers usually weigh hands-on pipeline work you can explain in detail.

How many Harness questions should I practise before an interview?

Be able to answer the freshers' section without notes, write the stage YAML in question 26 from memory, and talk through at least six of the scenario questions in detail.

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Mridul I am a passionate technology enthusiast with a strong focus on DevOps, Cloud Computing, and Cybersecurity. Through my blogs at DevOps Training Institute, I aim to simplify complex concepts and share practical insights for learners and professionals. My goal is to empower readers with knowledge, hands-on tips, and industry best practices to stay ahead in the ever-evolving world of DevOps.