Description of the issue:

I’m running into a cold start issue with my setup. The very first API request after some idle time takes anywhere from 15–22 seconds, but every request after that drops down to literally <1 second.

Error messages:

No explicit error messages, just extremely slow response times on the first request, but subsequent ones are super fast.

Logs:

Server logs show execution time for the first request being several seconds, while subsequent calls to the same endpoint are processed in milliseconds. I feel like this suggests some kind of warm-up or initialization overhead.

Setup:

• Backend: NestJS on Railway

• Database: Supabase (Postgres)

• Additional service: Separate AI server also hosted on Railway

• Both backend and AI servers are being pinged every ~2.5 minutes for health checks

Question:

Is there a recommended way to prevent this DB cold start issue / is anyone running into the same thing?

Any best practices for avoiding this long “first call” penalty would be greatly appreciated.