Harness Deployment Strategy Interview Questions (2027)

36 Harness deployment interview questions on rolling, canary, blue green and basic strategies, verification, approvals and rollback in Harness CD.

Sep 25, 2025 - 14:52
Nov 9, 2026 - 07:54
 0  59
Harness Deployment Strategy Interview Questions (2027)

Harness deployment interview questions test whether you can pick and configure the right release strategy in Harness CD: rolling, canary, blue green, basic or multi-service. Interviewers expect you to name the generated steps, explain how traffic moves between versions, add verification and approvals, and describe how rollback works for each strategy. This page has 36 questions about deployment strategies only, grouped by experience level, plus a scenario section.

It is narrower than our real-time Harness interview questions, which cover pipelines, delegates, connectors and troubleshooting across the whole platform. Read this one when the job description says "progressive delivery", "canary releases" or "zero downtime deployments". Facts were checked against the Harness Developer Hub in November 2026.

Key takeaways

  • Harness CD offers rolling, blue green, canary, basic, multi-service and build deployments as stage strategies.
  • In Kubernetes, the Canary and Blue Green strategies support only Deployment workloads, and only one managed workload per service. The Rolling Deployment step can handle several managed workloads.
  • A Harness Kubernetes canary has two groups: a Canary Deployment step plus Canary Delete, then a normal rolling update for the primary group.
  • Blue green uses the Stage Deployment and Swap Primary with Stage steps, and Harness tracks pod sets with the harness.io/color annotation.
  • Verification sits between the "test" step and the "promote" step. Approvals (Harness, Jira, ServiceNow) and failure strategies decide what happens when checks fail.
  • Expect at least one scenario where you must justify a strategy choice in terms of cost, risk and rollback speed.

Strategy comparison at a glance

StrategyHow it releasesExtra infrastructureRollbackTypical fit
RollingReplaces instances one by one or in batchesNoneRolling Rollback step redeploys previous versionStateless services, non-production stages
CanarySmall canary first, verify, then full rolloutA few extra pods during the canary phaseDelete canary or roll back the primaryProduction releases that need real-traffic checks
Blue greenDeploys to idle set, then swaps service selectorsA second full set of pods while both runSwap back to previous pod setZero-downtime cutover, fast rollback
BasicUpdates all nodes at onceNoneSlow, redeploy old versionLearning, non-critical apps
Multi-serviceDeploys several services togetherDependsPer serviceApps with version dependencies

Harness deployment questions for freshers

1. What deployment strategies does Harness CD support?

Harness documents rolling, blue green, canary, basic, multi-service and build deployments. You choose the strategy when you add execution steps to a deployment stage, and Harness generates the default steps for it. You can then add, remove or reorder steps, so a strategy is a starting template rather than a locked workflow.

2. What is a rolling deployment?

All nodes in one environment are replaced with the new version serially or in batches defined by a window size. It needs no extra infrastructure and carries less risk than a basic deployment. The trade-off is that the application and database must tolerate old and new versions running side by side during the rollout.

3. What is a canary deployment?

A canary releases the new version to a small part of the environment first. You check metrics and logs, and only if the canary looks healthy do you continue to the full rollout. Harness describes it as the lowest-risk strategy because business exposure is limited to the canary phase.

4. What is a blue green deployment?

Two identical environments run at the same time with different versions. The new version is deployed to the idle (stage) side, tested, and then traffic is switched to it. Rollback means switching traffic back. Some vendors call this red/black deployment.

5. What is a basic deployment and when is it acceptable?

Every node is updated at the same time with the new version. Harness says it is not intended for production because it is less safe than canary or blue green. It is acceptable for learning, experiments, off-hours deployments of apps nobody uses at that time, or services that are not business critical.

6. What are the service, environment and infrastructure definition in a Harness deployment?

The service describes what you deploy: manifests, artifacts and variables. The environment is where you deploy, such as dev or prod. The infrastructure definition is the concrete target inside that environment, for example a Kubernetes cluster and namespace reached through a connector. A deployment stage combines all three with an execution strategy.

7. What is a deployment stage in Harness?

A stage is a section of a pipeline that performs one major job. A Deploy stage takes a service, an environment and an infrastructure definition, and runs execution steps such as Rolling Deployment or Canary Deployment. A pipeline can chain several Deploy stages, for example QA then production.

8. What is the difference between gated and no-gate CD?

Gated CD inserts approvals before production changes. No-gate CD deploys without manual approval and depends on automated testing and verification. Harness supports both: you can add approval stages or steps, or leave them out.

9. Which approval options can you add to a deployment pipeline?

Harness supports manual Harness approval stages and steps, Jira approval stages and steps, and ServiceNow approval steps and stages. Teams often put an approval between a staging deploy and a production deploy, or link a production deploy to an approved change ticket.

10. What does Continuous Verification do during a deployment?

It analyses metrics and logs from your monitoring tools after a deployment step and decides whether the new version behaves normally. The Verify step can fail the stage when it detects anomalies, which triggers the failure strategy, often a rollback. Harness offers AI Verify (v1) and a newer AI Verify (v2) step.

11. Name some health sources Harness verification can use.

For AI Verify (v1), Harness lists APM tools such as AppDynamics, Amazon CloudWatch, Datadog, Dynatrace, New Relic, Prometheus and Splunk Observability, and log tools such as Elasticsearch, Grafana Loki, Splunk and Sumo Logic. A custom health source covers tools without a native connector. AI Verify (v2) supports a narrower set today.

12. Why do canary and blue green strategies cost different amounts?

Blue green keeps two full sets of pods running during the swap, so it can double compute for a period. A canary adds only a few extra pods during its first phase. Harness notes that canary is cheaper because it does not need two production environments.

Questions for engineers with 2 to 5 years of experience

13. How does a Harness Kubernetes canary differ from a textbook canary?

A textbook canary moves through many verified phases (2%, 10%, 25% and so on). Harness uses two groups for Kubernetes. Group 1 creates a canary copy of the Deployment, you verify it, and the Canary Delete step removes it. Group 2 runs a normal Kubernetes rolling update of the primary Deployment. Kubernetes rolling update already handles gradual pod replacement, so Harness does not repeat it.

14. How do you size the canary in the Canary Deployment step?

Choose Instance Count for a fixed number of pods, or Percentage of the replicas in your manifest. With replicas: 4 and 50%, two pods are deployed. With replicas: 3 and 50%, Harness rounds up and deploys two pods.

15. What does the Canary Delete step do, and why keep it?

It removes the canary workload created by the Canary Deployment step once that step has done its job. Without it, the extra canary pods keep running beside the primary Deployment, consume resources and can receive traffic through the same Service selector.

16. Which Kubernetes workloads can Harness Canary and Blue Green steps deploy?

Only Deployment workloads, and only one managed workload per service. Multiple workloads in the service manifests cause the deployment to fail. The Rolling Deployment step supports other workload types and multiple managed workloads, and the Apply step can deploy any object, commonly Jobs.

17. When would you use the Apply step instead of Rolling Deployment?

Use Apply for objects you want to push once, such as a database migration Job before the main rollout, or for manifests you keep separate from the managed workload. Harness documents that Apply can deploy Jobs but does not offer rollback, while Rolling Deployment offers rollback but is not meant for Jobs.

18. Walk through a Harness Kubernetes blue green deployment.

Stage Deployment creates a new pod set and points the stage Service at it, then waits for steady state. You add verification or tests between steps. Swap Primary with Stage switches the primary Service selector to the new pod set and the stage Service to the old one. On the first deployment the pod set is labelled harness.io/color: blue, the next one green, and later deployments alternate.

19. How does Harness know which Service is primary and which is stage?

With a single Service in the manifests, Harness treats it as primary and creates a copy with a -stage suffix automatically. With two Services, annotate them with harness.io/primary-service: "true" and harness.io/stage-service: "true".

apiVersion: v1
kind: Service
metadata:
  name: shop-api
  annotations:
    harness.io/primary-service: "true"
spec:
  selector:
    app: shop-api
  ports:
    - port: 80
      targetPort: 8080

20. What are the Blue Green Stage Scale Down and Scale Up steps for?

After the swap, Scale Down shrinks the last stage environment to save resources and removes its HPA and PodDisruptionBudget. Scale Up can restore a scaled-down environment. The cost is slower rollback: the old version is no longer warm, and HPA and PDB objects must be redeployed.

21. What does the Skip Dry Run option change?

By default Harness runs kubectl apply with --dry-run during the Initialize phase to print what would be sent to the cluster. Selecting Skip Dry Run removes that check. Keep the dry run on unless it causes a specific problem, because it catches invalid manifests before any change happens.

22. How does rollback work for a rolling deployment?

Add a Rolling Rollback step to the rollback section of the stage. It rolls back workloads deployed by the Rolling Deployment step to the previous release. Failure strategies decide when the rollback runs, for example on any step failure or on a failed Verify step.

23. What is a failure strategy and where do you set it?

A failure strategy defines the action when a step or stage fails: retry, manual intervention, mark as success, abort, or roll back the stage or pipeline. You can set it on a step, step group or stage. For deployment steps, "rollback stage on failure" is the usual production choice.

24. Where should the Verify step go in each strategy?

In a canary, after Canary Deployment and before the primary rollout. In blue green, after Stage Deployment and before Swap Primary with Stage, so you check the new version before users see it. In a rolling deployment, after the rollout, which means users are already on the new version when verification runs.

25. How do multi-service deployments work?

One stage deploys several services to the same environment, which helps when services depend on matching versions. Harness lists the downside: it is hard to test every dependency, and rollback is slower. Many teams prefer separate stages with explicit ordering.

26. How can you deploy to many clusters in one pipeline?

Harness supports multiple services and environments in a deployment stage, so one stage can fan out across several infrastructure definitions. You can run them in parallel or one after another. One after another lets you stop before the change reaches every cluster.

Senior and lead-level questions

27. How do you choose between canary and blue green for a payments API?

Ask three things: can you afford two full environments during cutover, do you need real traffic on the new version before full exposure, and how fast must rollback be. Blue green gives instant switch-back but tests on stage traffic. Canary exposes a small slice of real users and costs less. Harness notes many customers move from blue green to canary to test in production with less business impact.

28. What database risks apply to every deployment strategy?

During any overlap, old and new code hit the same schema. Use backward-compatible migrations (expand, then contract), run them as a separate step such as a Kubernetes Job, and never make a destructive change in the same release that first depends on it. Blue green does not remove this risk because the database is usually shared.

29. How would you build a progressive rollout with more than one canary phase?

Harness documents a progressive canary option for Kubernetes. You can also add phases with the Scale step between the canary and rolling groups, or use Argo Rollouts through Harness GitOps. Each phase should have a verification step and a clear stop condition.

30. How do deployment freezes fit into release governance?

Harness has a Freeze Deployments feature that blocks deployments in a time window or for chosen services and environments, for example during a festival sale. Pair it with approvals and policies so hotfixes follow a documented exception path.

31. How do you limit the blast radius of a bad release across regions?

Deploy to one region or cluster first with a canary, verify, then fan out. Use barriers or separate stages so later regions wait. Keep rollback automatic per region and make sure the failure strategy stops the remaining stages.

32. How do you stop verification from giving false results?

Make sure the canary gets enough traffic to produce data, pick metrics that reflect user impact (error rate, latency), and tune duration and sensitivity. Watch tool limits too. Harness warns that Datadog's default API rate limit can make verification fail when there are many pods.

Scenario-based questions

33. A canary deploy succeeded but users saw errors from "extra" pods for an hour. What happened?

The likely cause is that Canary Delete was removed or skipped, so canary pods kept matching the Service selector. Check the execution graph, restore Canary Delete in the canary group, and add a check that no canary workload remains after the stage.

34. Your blue green pipeline fails because the service has a Deployment and a StatefulSet.

Blue green supports one Deployment workload per service. Move the StatefulSet to its own service with a rolling strategy, or deploy it with the Apply step, and keep blue green for the stateless Deployment.

35. After a swap, the team scaled down the old environment and then found a bug. Rollback is slow.

Scale Down removed the warm previous version and its HPA and PDB. Use Scale Up or redeploy the previous artifact. Going forward, delay Scale Down until the new version has passed a soak period.

36. Verification failed but the stage did not roll back.

Check the failure strategy on the Verify step and stage. It may be set to manual intervention or ignore. Set "rollback stage" for verification failures, and confirm the rollback section contains the right rollback step for the strategy.

How to prepare

Build a small Kubernetes app and deploy it three times with different strategies, watching pods and Service selectors with kubectl get pods,svc -o wide. For broader Harness coverage, see the 75+ Harness CD interview questions, the scenario-based Harness questions and the advanced Harness questions. Our CI/CD pipeline examples show how large companies structure releases, and the 200 DevOps interview questions hub covers the rest of the stack.

If you want guided labs on CI/CD and Kubernetes releases, the DevOps training runs in our Kharadi, Pune classroom and live online.

Frequently asked questions

Which Harness strategy is best for production?

Harness says most customers use blue green or canary for mission-critical apps. Canary costs less and tests with real users. Blue green gives the fastest rollback.

Can Harness canary deploy a StatefulSet?

No. Canary and Blue Green steps support only Kubernetes Deployment workloads. Use the Rolling Deployment step for other workloads.

What steps does Harness generate for blue green?

Stage Deployment and Swap Primary with Stage. You can add verification, Blue Green Stage Scale Down and Scale Up steps.

Does a Harness canary use many traffic percentages?

The default Kubernetes canary uses two groups: a canary sized by count or percentage, then a rolling update of the primary. Harness also documents a progressive canary option.

Where does verification go in a canary?

Between the Canary Deployment step and the primary rollout, so a bad version fails before full exposure.

Is a basic deployment ever acceptable?

Harness does not recommend it for production. It suits learning, experiments and non-critical services.

References

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.