Hi Luciano, quick clarification - when you say "transaction time," are you referring to:

- HTTP request response times (what apdex is likely measuring)
- Database transaction duration specifically

Any insight on specific transactions spiking more than others? What does your APM breakdown show - is the time being spent in DB queries, application code, or network calls?

hey chandrika, I mean Database transaction duration specifically.
All transactions are taking way longer. The longest time is spent on DB queries.

Could you try to enable metal edge for your metal services? https://docs.railway.com/railway-metal#metal-edge-network-beta

Our metal mesh routes traffic through own network and being on it should reduce latency because it essentially provides better routes for your metal traffic. Let me know if this helps, and if not, I can escalate to the team for us to dig further into

Chandrika, the slow service is the DB, which doesn't seem to have an option for metal edge.

Got it! I was looking at your services to dig deeper but it seems like you've switched back to our legacy regions
1) Has the performance recovered after the switch back?
2) Would it be possible to create a small reproducible environment that mimics the way you were set up on metal and runs into the same latency issue you're experiencing? Having this would make it easier to investigate the setup etc as the current one is no longer on metal