Tell me about a time you had to refactor or redesign a critical backend service while it was in active use. How did you plan the work, mitigate risk, communicate with stakeholders, and ensure the system stayed reliable in production?

Question

VirtualInterview.AI · Accepted Answer

At my last role, I led a refactor of our monolithic billing service that handled about 15k requests per minute and over $1M in daily transactions. It had grown into a 20k-line class that only two people fully understood, and deployments were painful. I started by mapping the main flows and error hotspots using logs and application traces, then proposed splitting it into three smaller services: invoicing, payments, and notifications.

To reduce risk, we kept the existing API contract and introduced a strangler pattern behind a feature flag. We shipped the new services dark first, mirroring 10–20% of traffic and comparing results in logs and metrics. I set clear rollback criteria and documented them for on-call. I held short weekly check-ins with product, support, and finance so they knew what was changing and when. Over six weeks, we migrated fully with zero major incidents, cut deploy time from 45 to 10 minutes, and reduced billing-related on-call alerts by about 60%.

Sample Answer

Keywords