Roll back first and investigate after, unless the deploy included a database migration that makes rollback unsafe. Confirm the timeline lines up with the deploy, check whether the errors are concentrated in the new version's instances, then revert to the previous artifact or shift traffic back. Once the error rate recovers, keep the failing build available in a staging environment so you can diagnose without users paying for it.
Why interviewers ask this
The interviewer is checking your default instinct under pressure. Strong candidates mitigate immediately and treat curiosity as a luxury for later. They also want the important caveat about migrations, since a naive rollback can be worse than the outage. Follow ups usually probe how you would have caught this earlier, which is where progressive delivery and automated rollback come in.
How to structure your answer
- State the default: roll back first, diagnose after.
- Give the one caveat that changes it, an irreversible migration.
- Describe how you confirm the deploy is really the cause.
- Close with what would have caught it sooner.
Example answer
Default is roll back. An hour of climbing errors is not a mystery to solve live, it is a bleed to stop, and the previous artifact is known good. Before I pull the trigger I do two quick checks. One, does the error curve actually start at the deploy time, or was it climbing before, because if it started earlier the deploy is a coincidence and rolling back wastes ten minutes. Two, did this release include a schema migration, because if it did I need to know whether the old code still works against the new schema. If we did expand and contract properly it does, and rollback is safe. If somebody dropped a column, rollback is now the dangerous option and I go for a forward fix or a feature flag instead. Once errors recover I keep the bad build deployed in staging so we can reproduce without customers involved. Afterwards the question I would push on is why a canary did not catch it, because an hour of impact means the rollout was not watching the error rate against a baseline.
Walking into this interview soon? GhostPilot listens to your live call, spots the question the moment it is asked, and puts a structured answer on your screen in real time. Try it on your next mock, or grab a $29 Session Pass, no subscription, for the real thing.
See how it worksFollow-up questions to expect
- How would you handle it if the release included an irreversible migration?
- What automated signal should have caught this in the first ten minutes?
- How do you handle a rollback when several services deployed together?
Related devops engineer questions
Your interviewer will ask their own version of this. Paste your actual job description into the free Question Predictor and get the 20 questions that role is most likely to ask, with what each one is really probing.
Predict my questions