Egress to googleapis.com failing with "Premature close" — US West, all deployments, since Jul 2
ivarking11
PROOP

5 days ago

Our Node.js service cannot reach Google APIs from its Railway container. Every call to Google's OAuth token endpoint fails with:

Invalid response body while trying to fetch https://www.googleapis.com/oauth2/v4/token: Premature close

Project: desirable-analysis · Service: we-care-agent · Environment: production · Region: US West

Affected deployments (all three, same failure):

  • 6c3dbd1f-a500-4989-ab7c-1741cfd1a65c — intermittent occurrences starting ~06:30Z Jul 2
  • 09906a30-f23c-451d-9da2-bf0e8cdf479d — chronic from its boot at 15:52Z Jul 2; a service restart at 17:03Z did not help
  • b2b237ae-1f13-4afe-b988-67e55a490477 — fresh redeploy (new image + container) at 17:14:54Z Jul 2; failure persists from the first minute (40 errors in the first ~15 minutes)

Frequency: our Gmail sync polls every 60 s across 7 mailboxes; 100% of cycles fail (≈40 error lines per 15 min, continuously).

Evidence this is not on our side: the exact same call (google-auth-library JWT domain-wide-delegation token grant + Gmail threads.get) from an external network with the same credentials succeeds 3/3 in under 1.4 s. Same code, same credentials, only the network path differs.

This is production-impacting: our email ingestion for a logistics operation has been down since ~15:50Z and is currently running through a temporary local workaround.

Request: please check the egress/network path from our service's host/region toward oauth2.googleapis.com / www.googleapis.com (connection resets / premature close at the HTTP layer), and advise if there is a known issue or if you can migrate us to a healthy host.

Note: this appears to be the same issue reported in your help thread

"Egress truncating large gzipped HTTPS responses in US West (California)

Metal" (June 18-19, ~12 days ago), where your engineer confirmed the

path-MTU blackhole signature on US West Metal egress. That thread was

auto-closed without a confirmed network fix. Our case (googleapis.com,

same region, same signature, onset Jul 2) suggests the egress issue is

still active. Please treat this as a recurrence with fresh evidence.

Awaiting Conductor Response$20 Bounty

1 Replies

Status changed to Awaiting Railway Response Railway 5 days ago


Railway
BOT

4 days ago

This thread has been opened as a bounty so the community can help solve it.

Status changed to Open Railway 4 days ago


This is known nodejs issue and has been patched, upgrade your node version to 24.18.0 or newer


Welcome!

Sign in to your Railway account to join the conversation.

Loading...