3 months ago
Service: vocab-academy-production (Singapore region, Hobby plan)
Since the July 2 Edge Network incident, requests randomly stall 2-15 seconds.
We have isolated it to Railway infrastructure:
-
TCP connect is always ~0.04s
-
Our origin is instrumented: since yesterday's boot it recorded ZERO
DB operations over 150ms and ZERO event-loop stalls over 500ms
(in-process monitoring endpoint, ~20h window)
-
Yet external TTFB randomly hits 2-15s, equally on static assets,
prerendered pages, SSR and API routes
-
Worked normally until July 1. Persists across many fresh deploys.
Origin-clean + connect-fast + random TTFB stalls on all content types
points to the edge-to-origin path or host network layer.
Please investigate our service's routing/host.
1 Replies
3 months ago
This thread has been opened as a bounty so the community can help solve it.
Status changed to Open Railway • 3 months ago
2 months ago
Multisecond CPU descheduling would appear as event loop lag so throttling is unlikely.
Measure handler entry to response start with Server Timing and compare against external TTFB.
Large TTFB with tiny application time isolates the Railway ingress or host accept path.