2 months ago
Seeing noticeably higher latency than usual on queries between my Axum backend and Postgres 16 — both in the same Railway project. This is a regression; performance was previously fine with the same setup. Postgres (1 replica) is in US Southeast, Axum service has 4 replicas. We've checked query statistics on the Postgres side and execution times look acceptable, so it seems like network round-trip overhead rather than query execution time.
- Response times are slower than what we've seen historically with the same workload
- Latency is consistent regardless of query complexity
- Query stats on Postgres look normal — the slowness appears to be between service and DB
Has there been a recent infrastructure change or incident that could be affecting internal networking?
Stack: Rust/Axum + SQLx with a connection pool. Project ID ee793c47-c901-48c9-a217-684e910393e3
4 Replies
2 months ago
Hey, Railway provides network logs that allow you to identify if network round-trips are causing those issues (they're raw TCP packet logs). You can access these logs by navigating to your service and clicking on Network Flow Logs.
Do you see anything unusual there that could be causing this?
Attachments
2 months ago
You can also filter by packets going into your database by using the @peer filter.
Thanks! Most requests to/from Pgbouncer have ~200ms latency. So, doesn't look like that's the issue. I'll keep troubleshooting.
2 months ago
im running into the same issue and wondering why there isnt any visibility or transparency into this