2 months ago
Subject: "web" service fails every fresh build — instant failure with no build logs (builder-qpcbvd)
Project: O Mercado Aventureiro (6d2ba6ca-9c0e-4767-982b-c369b1fb763d)
Environment: production (236cfdb9-d87c-4953-a142-ed83cad12c06)
Service: web (4dbf5621-0c89-4e9f-b38d-1b13fb962581)
Builder: RAILPACK, reported as "Metal builder builder-qpcbvd"
Source: GitHub repo NeiMac/mercado-aventureiro, root directory apps/web, branch main
Issue:
Since ~2026-08-26 21:00 UTC, every attempt to trigger a fresh build for the "web" service fails immediately, with no usable build log:
scheduling build on Metal builder "builder-qpcbvd"
scheduling build on Metal builder "builder-qpcbvd"
Deploy failed
No further output — fails before any dependency install or source copy step. Happens regardless of trigger:
GitHub push to main (webhook-triggered deploy)
Manual railway up via CLI (4 attempts, different commits, same result)
Manual redeploy via the Railway API
The only thing that succeeds is a "redeploy" that reuses an old image without rebuilding — keeps the service up, but on stale code.
Failed deployment IDs (same signature): 1866f203, 14565d89, f1d1ba2f, 5bd14124, 937d440e, bc2af56e, fe42968c, 46626dc5 (all between 2026-08-26 21:07 UTC and 2026-08-27 01:12 UTC).
For comparison: other services in the same project (api, scheduler) built and deployed successfully from the same push around 00:46–00:47 UTC — so this looks specific to the web service's build path, not account/repo-wide.
What we need: help finding why builder-qpcbvd fails to start a real build for this service, and getting a fresh build of current main (commit 74f1fa9) live.
2 Replies
2 months ago
Your railway up is running from inside apps/web. The root directory is already set to apps/web, so the builder looks for apps/web inside what you upload and doesn't find it. Run it from the repo root, and it will build.
Status changed to Awaiting User Response Railway • about 2 months ago
2 months ago
Thanks, that was it! Turns out I was passing an explicit path argument to railway up (either . from the repo root, or the apps/web directory with --path-as-root). Both failed, but in different ways:
railway up . from the repo root → failed instantly client-side with prefix not found during "Indexing", before even reaching the server.
railway up apps/web --path-as-root → uploaded fine, but the build failed immediately ("scheduling build" logged twice, then "Deploy failed", no real build output) — looks like the service's Root Directory (apps/web) was being re-applied on top of the already-scoped upload.
Calling railway up with no path argument at all (just railway up -s web -e production --ci --yes -m "...", run from the repo root) fixed it — got a real Railpack build (npm install → next build → route list → image push → "Deploy complete").
Appreciate the quick pointer — all good now.
Status changed to Awaiting Railway Response Railway • about 2 months ago
Status changed to Solved Railway • about 2 months ago