When traffic grew, our single-instance setup started to creak. Here is what we changed.
- Containerized every service
- Added health checks and graceful shutdown
- Automated rollouts with a canary stage
The result: deploys that no longer wake anyone up at night.