3 months ago
All builds for my service scanner (project 40e31277-ca72-4caf-9b76-746b3005f3bc, service 06e32daa-3698-440d-b01b-1acc5e921c04) have failed since 2026-07-14 ~08:39 UTC. Production is still Online, serving from the last successful image.
This looks like the same class of issue as the recent "All builds FAIL on Metal builder builder-ryvidd" thread — a Metal builder wedged, not a build-config problem.
Every failed build logs only:
scheduling build on Metal builder "builder-nhmelg"
then fails in under 30 seconds with NO build output and zero instances created (it never reaches "fetching snapshot").
Already tried:
- Multiple
railway updirect uploads — same result, always routed to builder-nhmelg - Region pin on the service — did not change the builder assignment (reverted)
- Redeploying the last successful image (usePreviousImageTag) — works fine; only NEW builds fail
- Confirmed the
webservice in the SAME project builds fine on a different builder (builder-sskgan), and the Dockerfile builds locally
Please either:
- Remove/repair builder-nhmelg from the pool, or
- Re-route my scanner service to a different Metal builder, or
- Disable Metal builder for this service so it uses the legacy non-Metal builder.
Happy to trigger a fresh build on request. Thanks!
2 Replies
Status changed to Awaiting Railway Response Railway • 3 months ago
3 months ago
Could you try triggering a fresh build now and let us know if it goes through? Your service is using the Railpack builder in the sfo region, and we want to confirm whether builds are currently succeeding for you. If it still fails at the scheduling step, reply back and we will dig further.
Status changed to Awaiting User Response Railway • 3 months ago
sam-a
Could you try triggering a fresh build now and let us know if it goes through? Your service is using the Railpack builder in the sfo region, and we want to confirm whether builds are currently succeeding for you. If it still fails at the scheduling step, reply back and we will dig further.
3 months ago
Just triggered a fresh build (deployment 25d6bdf2-e8ff-421b-8925-e51a4de23d03, 2026-07-15 23:53 UTC) — it still fails at the scheduling step.
Specifics from this run:
- Build log shows only: scheduling build on Metal builder "builder-nhmelg" — nothing after it.
- Event steps: SNAPSHOT_CODE (3s) then BUILD_IMAGE "completed" in ~4 seconds (23:53:15 → 23:53:19 UTC) with zero build output, then the deployment fails in under 15 seconds with zero instances created. It never reaches an actual build.
- It's still landing on builder-nhmelg specifically. (You mentioned Railpack in sfo — the wedged node it keeps getting scheduled onto is builder-nhmelg.)
- For contrast: my
webservice in the same project builds fine on builder-sskgan, and the Dockerfile builds locally.
So no, still not going through. Please dig further — it looks like builder-nhmelg is still wedged for this service.
Status changed to Awaiting Railway Response Railway • 3 months ago
Status changed to Solved Anonymous • 3 months ago