CI/CD · Ship safely, repeatedly

What checks should run after deployment?

Mediumpost-deployverificationsloUpdated Oct 2026
Quiz this track

Short answer

Run smoke tests on critical paths, synthetic journeys, and SLO checks for errors and latency, then compare against the baseline. Verify in a defined window and roll back automatically if the criteria fail.

A successful deployment only proves the process completed. Post-deployment verification asks whether the system is doing its job for real traffic, which is a different question from whether the pods started.

  • Smoke tests: unauthenticated health and a few authenticated critical requests
  • Synthetic journeys: scripted user flows run from outside the cluster
  • SLO checks: error ratio, latency and saturation compared with the baseline
  • Business checks: order completion, signup success or payment authorization
  • Dependency checks: database, queue and third-party integrations
  • A defined verification window with automatic rollback or rollback recommendation
sh
curl --fail --silent --show-error https://api.example.com/health
curl --fail --silent --show-error --oauth2-bearer $SMOKE_TOKEN https://api.example.com/v1/checkout/health
Verification that runs for thirty seconds after deploy can miss a slow leak or a problem that only appears under sustained traffic. Match the window to the failure modes you care about.

Follow-ups you should expect

Interviewers use these to check whether the first answer was understanding or memorisation.

  1. 1What is the difference between a smoke test and a synthetic journey?
  2. 2How long should a post-deployment verification window be?
  3. 3What should trigger an automatic rollback?
  4. 4How do you verify a deployment with low traffic?

Say it in the interview

After deployment I run smoke tests on health and critical authenticated paths, then synthetic journeys and SLO comparisons against the baseline. I include business signals and dependency checks where they matter. The verification window has to match the failure mode. A slow memory leak will not show up in thirty seconds, so if the criteria fail, rollback should be automatic or at least strongly recommended.

Use this as a model, not a script. Saying it in your own words is what makes it survive follow-up pressure.

Related questions

Reading is step one

Now say it out loud.

Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live CI/CD mock interview and get scored feedback on depth, clarity and confidence.