CI/CD · Ship safely, repeatedly

How do you monitor a deployment in progress?

MediummonitoringdeploymentobservabilityUpdated Oct 2026
Quiz this track

Short answer

Record a deployment marker, watch golden signals and business metrics for the new version, and compare it with the baseline. Alert on user-facing symptoms and be ready to abort or roll back automatically.

A deployment changes the system, so monitoring it means separating the effect of the new version from normal variation. Deploy markers on dashboards make that correlation visible and reduce confusion during an incident.

  • Emit an event or annotation when a deployment starts and completes
  • Watch latency, traffic, errors and saturation, ideally split by version
  • Include a small set of business indicators such as checkout success or signup completion
  • Compare canary and baseline over the same window rather than absolute thresholds alone
  • Define abort criteria before the rollout so the decision is not made under pressure
  • Link the deploy marker to the commit, artifact digest and change ticket
text
request_rate by version
error_ratio by version
p95_latency by version
saturation by version
checkout_success_rate by version
Splitting by version shows whether the new release changed behaviour.
Do not rely on a single aggregate error rate. A small canary can look healthy in aggregate while failing for one important route, region or customer segment.

Follow-ups you should expect

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

  1. 1What are the golden signals?
  2. 2How do deploy markers help during an incident?
  3. 3Which business metrics would you watch for a checkout service?
  4. 4How do you avoid alert fatigue during a rollout?

Say it in the interview

I put a deployment marker on dashboards so the rollout lines up with changes in behaviour. I watch latency, errors, saturation and traffic split by version, plus a few business metrics that show whether users can complete the core task. Abort criteria are defined before the rollout. I compare against the baseline rather than trusting absolute thresholds, because normal traffic variation can look like a regression.

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.