The Failed Deploy That Caused 45 Minutes of Downtime
#production-outage — 45 minutes down
user-api v2.3.1 deployed in-place. NOT NULL migration broke rollback — v2.3.0 cannot run against new schema. Manual DB rollback + redeploy took 45 minutes (~$27k impact).
The trap: migrate shared DB, deploy code, no smoke tests, no instant rollback.
Your task
Build `BlueGreenDeployer` in `blue_green_deploy.js`: deploy to inactive environment, smoke-test before switch, flip router for zero-downtime cutover, instant rollback.
Book: blue-green environments · Prod: smoke tests catch bad migrations before users do.
Skills you'll prove
Interview Tip
When asked about blue-green deployment or zero-downtime deploys, say: "Blue-green means two identical environments — one live, one idle. Deploy to idle, run smoke tests, then flip the router. If anything fails, the old environment is still running and rollback is instant — just flip the router back. I lived through a 45-minute outage where a NOT NULL migration broke rollback. With blue-green, we'd have caught it in smoke tests on the idle environment before a single user was affected. The key trade-off is cost — you're running double infrastructure — but for critical services, that's cheaper than 45 minutes of downtime."