Use the strangler pattern: put a facade in front of the capability, build the new service behind it, and move traffic incrementally with a feature flag while the monolith remains the fallback. Split the data last, usually by dual writing and backfilling until the new store is authoritative, and keep a verification phase where both paths run and results are compared before the old code is deleted.
Why interviewers ask this
This tests migration judgment, which is most of senior backend work. The interviewer wants incremental steps with a rollback at every stage, awareness that the data split is harder than the code split, and a definition of done that includes deleting the old path. Candidates who describe a rewrite and a cutover weekend are telling you how the incident happened.
How to structure your answer
- Start with a seam: route calls through a facade or interface.
- Move traffic incrementally behind a flag, keeping rollback cheap.
- Handle the data split explicitly with dual writes and backfill.
- Define completion as the old path removed, with verification before that.
Example answer
I would first create a seam. Even inside the monolith, all callers of that capability go through one interface, which is often a refactor worth doing on its own because it exposes how tangled the boundary really is. Then the new service implements the same contract, and the facade routes a small percentage of traffic to it behind a flag, starting with read only calls where a wrong answer is cheap. I run both in shadow mode for a while, sending the same request to each and comparing responses, because that catches the edge cases no one documented. Data is the hard part: I dual write, backfill history, reconcile, and only then make the new store authoritative, with the monolith reading from the service rather than the shared tables. Every step has a rollback of flipping a flag. And the migration is not finished until the old code and its tables are deleted; I have seen strangler migrations stall at ninety percent and leave two systems to maintain for years, which is worse than either option alone.
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 a transaction that spanned the old boundary?
- How do you compare results in shadow mode without doubling side effects?
- What would make you stop and roll the whole migration back?
Related backend developer 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