18 days ago
One of my services deployed but never actually served traffic, and the logs give no indication of why. I'd like help understanding what happened at the platform level.
Service details:
Deployment: fdf0aacf (from GitHub repo Armen101/DLT)
Runtime: Dockerfile build, Java 17 / Spring Boot app, with a persistent volume mounted at /app/data
Timeline (Jul 7 - 8, 2026, GMT+4):
19:44:25 — Deploy Logs: "Mounting volume on: /var/lib/containers/railwayapp/bind-mounts/fd82f6cd-4d3a-413b-bc20-f3aac981f672/vol_k16t..."
19:44:26 — Deploy Logs: "Starting Container"
19:44:26 → 20:35:51 — complete silence. Not a single line of application output for 51 minutes. Our Spring Boot app normally prints its startup banner within 1–2 seconds of the JVM launching, so it appears the process either never executed or was killed before it could write anything to stdout.
During this entire window the deployment showed status "Active", while every HTTP request to any path returned 502 "Application failed to respond" in 1–24ms (visible in HTTP Logs).
20:35:51 and 20:35:57 — "Stopping Container" (twice), after which the deployment shows "Removed".
What I've already ruled out:
The build completed successfully (image was produced and the container start was attempted).
The exact same commit, built and run locally with equivalent configuration, starts cleanly in ~2 seconds and serves requests.
The app was not crash-looping — there are no restart entries and no exceptions in Deploy Logs.
Questions:
Was the container OOM-killed or terminated by the platform before the entrypoint process could produce output? If so, is that visible on your side (it isn't in ours)?
Could the volume mount (vol_k16t...) have hung or attached in a degraded state, blocking the process?
Why did the deployment report "Active" for 1 day while the proxy was returning 502 for 100% of requests? Is there a health-check setting we should configure so this state fails fast instead?
Happy to provide any additional IDs or reproduce on request.
Thanks,
Armen
1 Replies
18 days ago
This thread has been opened as a bounty so the community can help solve it.
Status changed to Open Railway • 18 days ago
18 days ago
A couple of things I'd check here:
Does your app bind to 0.0.0.0:$PORT? A container can be running but never become reachable if it's listening on a different interface or port.
Is your entrypoint writing logs before Spring Boot starts? Since there isn't a single line of output, it would help to confirm whether the JVM is actually being invoked or if it's hanging before your application starts.
Since you have a persistent volume mounted, try deploying once without the volume (if possible) or with a fresh volume attached. If the app starts normally, that would narrow the issue down to something happening during initialization around the mounted data.
The 502s suggest the proxy never saw a healthy application, so I'd focus on what could prevent the process from reaching the point where it starts listening rather than on Spring Boot itself.
A 502 Bad Gateway error means the website's server received an invalid response from another upstream server.