CI/CD Interview Questions on Checks, Artifacts and Release Gates
CI/CD interviews ask how a change moves from source to a release that meets its checks and approval policy. These original questions cover integration, delivery, deployment, pipelines and operational feedback. The worked case has a green pull-request result even though a required integration test never executed.
Topic: DevOps & Platform Engineering in IT Infrastructure · Interview preparation
Quick answer: CI/CD interviews ask how a change moves from source to a release that meets its checks and approval policy. These original questions cover integration, delivery, deployment, pipelines and operational feedback. The worked case has a green pull-request result even though a required integration test never executed. It distinguishes a platform’s reported status from an organization’s explicit acceptance rule, then asks what evidence a reviewer needs before promotion. No workflow, repository permission or deployment is changed by this paper exercise.
Key takeaways
- Define whether a release policy requires a check to execute.
- A skipped job and a skipped workflow can behave differently.
- Associate check results with the intended commit and run.
- Keep build artifacts separate from reusable dependency caches.
- Protect pipeline identities and review release exceptions explicitly.
Table of Contents
- How do CI, continuous delivery and continuous deployment differ?
- Can a green result hide a test that never ran?
- What evidence should a release gate evaluate?
- How do artifacts, caches and test layers fit together?
- How should you discuss pipeline security and secrets?
- What should answers about deployment and rollback include?
- How do you explain CI/CD feedback and improve the supplied case?
- References
How do CI, continuous delivery and continuous deployment differ?
1. What is continuous integration? Developers integrate changes frequently and receive automated feedback on the integrated result. It is a working practice supported by tests and tools, not merely the presence of a build server.
2. How do delivery and deployment differ? Continuous delivery keeps validated changes ready for release under the chosen approval process. Continuous deployment automatically releases qualifying changes. Neither term means every commit must bypass review.
3. What is a pipeline? It connects defined activities such as build, tests, packaging and promotion. GitHub’s workflow overview explains events, jobs and steps; Jenkins’ Pipeline guide describes a pipeline as code, including stages and controlled pauses.
Can a green result hide a test that never ran?
The invented team’s release policy requires its integration test to execute successfully on the relevant candidate. Unit tests passed, but a job condition skipped integration tests. The pull request displays a successful status.
4. Is the team’s acceptance rule satisfied? No. GitHub’s job-condition guide explicitly says a skipped job reports success and does not block merging even when required. The organization’s stronger rule must therefore check execution and outcome, not just the green label.
5. Is skipping a workflow identical? No. GitHub’s required-check troubleshooting guide distinguishes skipped-workflow and check behaviour. Read the actual event, filters, job conditions and branch-protection configuration before diagnosing a blocked or permitted merge.
| Check in supplied case | Recorded state | Team rule |
|---|---|---|
| Unit test | Executed and passed | Satisfied for that test |
| Integration test | Skipped by job condition | Execution requirement not satisfied |
| Summary status | Green | Insufficient evidence by itself |
What evidence should a release gate evaluate?
6. What identifies the tested candidate? Keep the commit identity, workflow or pipeline version, run identifier and relevant artifact association. A passing result for an older commit does not establish that the current candidate passed.
7. How should an exception be handled? A documented policy can allow a justified exception with an approver, scope and expiry. It should remain visible as an exception, rather than being relabelled as a test that ran.
8. What should the integration check produce? Record whether it started, its final outcome and useful test evidence. Distinguish failure, cancellation, timeout and “not applicable” according to the release rule.
9. How would you verify the fix? Use a controlled candidate that should require integration testing and confirm the gate refuses promotion when that test is skipped. A successful run should retain the evidence needed to reproduce the decision.
| Evidence field | Why it matters | Review question |
|---|---|---|
| Candidate identity | Avoids using an unrelated result | Which commit was evaluated? |
| Execution and outcome | Separates skipped from passed | Did the required test actually run? |
| Exception approval | Makes a policy deviation visible | Who approved this exact scope? |
How do artifacts, caches and test layers fit together?
10. What is a build artifact? GitHub’s artifact guide describes files retained from a workflow, such as test results or built output. A dependency cache serves a different reuse purpose and is not a substitute for a controlled release artifact.
11. How do unit and integration tests differ? Unit tests isolate a small behaviour; integration tests exercise interactions under a stated scope. Their names alone do not establish which dependencies or failure paths were covered.
12. Does a staging pass prove production success? Configuration, traffic, data and dependencies can differ. State what staging validates and what remains for controlled rollout and service acceptance. Keep the relevant environment assumptions with the test result.
How should you discuss pipeline security and secrets?
13. What is the pipeline’s trust boundary? Review who can change workflow definitions, execute untrusted contributions, access secrets and approve protected environments. Build automation can hold powerful deployment authority.
14. Are redacted logs a complete secret-protection mechanism? No. GitHub’s secure-use reference describes scoped permissions and secret-handling precautions. Prevent unnecessary exposure and review generated output instead of assuming every sensitive transformation will be masked.
15. What does secure development add to a pipeline? NIST’s SSDF 1.1 connects practices across the development lifecycle. Scanning is one input; dependency management, review and response ownership also matter. A clean scan is not proof that software contains no vulnerability.
What should answers about deployment and rollback include?
16. How do blue-green and canary approaches differ? Blue-green switches between prepared environments; a canary exposes a controlled subset to a candidate. Both need routing, acceptance signals and a plan for stateful effects.
17. Is rollback always a safe undo? Reversing code or traffic may not reverse writes, messages or schema effects. Define recovery for the actual change and preserve evidence of the affected interval.
18. What are feature flags for? They can separate selected behaviour from deployment, but flags need ownership, tested combinations and eventual removal. Turning a flag off does not automatically undo data changes already made.
Keep the interview answer tied to a concrete failure signal. Name who can stop promotion and what service-level evidence determines whether the rollout is accepted.
How do you explain CI/CD feedback and improve the supplied case?
19. Which measures are useful? Track feedback delays, recurring failures, release outcomes and restoration under definitions the team understands. A faster pipeline that silently skips required coverage may improve a time metric while weakening the release decision.
20. What change would you recommend in this case? Make the policy explicit: integration testing must run and succeed for this class of change, unless an authorized exception is recorded. Give the final gate access to execution state and candidate identity. Review conditional paths and protect the gate itself from unreviewed changes.
Rehearse the answer with a passing candidate, a failed test, a skipped test and a cancelled run. Explain which outcomes block promotion and why. That is more useful than naming a preferred CI product without showing how its status becomes a trustworthy release decision.
Related reading: DevOps interview questions; Cloud deployment strategy interview questions.
References
Frequently Asked Questions
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Wow
0
Sad
0
Angry
0