a month ago
Project: AppZero (19d230c3-e219-4f99-8d54-d6e877726638)
Repo: https://github.com/gruppetto-app/AppZero
PR Environments was working as recently as yesterday (Oct 5) — PRs opened and merged that day got auto-provisioned preview environments as expected (and those environments were correctly auto-deleted on merge, which is why none of them still exist now). Starting today (Oct 6), no new PR gets an environment at all. The only environment that still exists outside staging/production is AppZero-pr-134, left over only because that PR is deliberately on hold (a PoC), not because it's the last one that ever worked.
What we've already ruled out today, each with a dedicated test PR:
- Stale staged config patch (from an old manual environment repoint) — discarded it, opened PR #434 as a clean test → still no environment.
- An extra object-storage bucket ("photos-staging") added around the same window things broke — deleted it entirely, opened PR #435 as a clean test → still no environment.
- "Enable Bot PR Environments" setting — already enabled, and moot anyway since our test PRs are authored by a real user account (nlangebe), not a bot.
- GitHub App installation — verified directly in gruppetto-app's org settings: Railway App installed, scoped to exactly this one repo, with full pull_requests/deployments/checks/workflows permissions. Nothing misconfigured there.
- Production/staging/AppZero-pr-134 branch configuration — read directly via the Railway API each time, consistently correct.
We also have an open AI-assistant thread on this (station.railway.com/support/pr-environments-not-creating-for-new-prs-71c3aeac) where the automated diagnosis has twice stated things directly contradicted by what we verify live via the API. We've exhausted everything checkable from the project/GitHub side and need a human engineer to check Railway's internal webhook delivery and environment-provisioning logs for this project for the Oct 5 → Oct 6 window specifically — something changed on your end in that window.
1 Reply
Status changed to Awaiting Railway Response Railway • about 1 month ago
a month ago
This thread has been opened as a bounty so the community can help solve it.
Status changed to Open Railway • about 1 month ago
a month ago
Given that PR Environments were working on Oct 5 and stopped on Oct 6 without a project-side configuration change, this looks worth checking at the GitHub App webhook/provisioning boundary.
Railway’s documentation says that with PR Environments enabled, opening a PR against a linked repository should automatically provision the PR environment. GitHub also keeps recent GitHub App webhook delivery records including the event payload, timestamp, and response.
Since the Railway GitHub App installation and project configuration have already been verified, could Railway check the pull_request webhook deliveries for PRs #434/#435 and compare them with a known-good Oct 5 PR? That should distinguish between the event not reaching Railway and the event being received but failing during PR-environment provisioning.
In particular, checking the delivery GUID/status and Railway’s corresponding provisioning logs should identify which side of that boundary is failing.
Status changed to Open dev • about 1 month ago
Status changed to Solved dev • about 1 month ago
