11 days ago
Subject: Public domain unreachable (connection refused) despite healthy container — service-specific, not regional
Project: olimpiadas-fiestas-patrias (ID: 86d578ba-19f0-416e-8ada-83bfc7832c21)
Service: olimpiadas-fiestas-patrias (ID: 6bb9831e-d551-4a87-83fc-04fddba94eb6)
Environment: production
Region: sfo
Symptom: The public domain (olimpiadas-fiestas-patrias-production.up.railway.app) has been unreachable from the outside (connection refused / timeout) for a few days, even though the container itself is healthy.
What I've verified:
Deployment status: RUNNING, no crash/restart loop.
App logs show gunicorn starting cleanly and listening on 0.0.0.0:8080.
Domain "Target port" is correctly set to 8080 (matches).
Domain sync status: ACTIVE.
I deleted and regenerated the service domain (new domain ID) — same result, and it resolves to the same edge IP (146.112.61.110) which still refuses/times out connections from multiple independent networks.
railway ssh into this service also hangs/times out — same failure mode as HTTP.
Another service in this same workspace, same region (sfo), works perfectly and resolves to a different edge IP (69.46.46.x) — so this doesn't look like a regional outage, just this specific service/domain routing.
Ask: Could you check the health/routing of edge node 146.112.61.110 for this service, or reassign it to a healthy edge? App-side config looks correct; this looks like a stuck routing entry on Railway's edge for this service.
Happy to provide deployment ID or further logs if useful.
2 Replies
11 days ago
This thread has been opened as a bounty so the community can help solve it.
Status changed to Open Railway • 11 days ago
11 days ago
Try using a VPN. I’m able to access the site just fine.
Attachments
3 hours ago
The reported IP is the key discrepancy: 146.112.61.110 is a Cisco Umbrella/OpenDNS block-page address, not the Railway edge to investigate. Cisco's official block-page table lists that exact address for several filtering categories. This does not establish that your site contains malware or identify which category/policy applied.
At 2026-09-08 06:32 UTC, I checked public DNS for olimpiadas-fiestas-patrias-production.up.railway.app through both Google and Cloudflare. Both returned 69.46.46.122, with a 60-second TTL—not 146.112.61.110. These are current, outside-network measurements, not a reconstruction of your original failure; DNS answers can change, so don't hard-code that IP.
On an affected device/network, check the answer from its configured resolver:
nslookup olimpiadas-fiestas-patrias-production.up.railway.appIf it still returns 146.112.61.110, ask the network/DNS administrator to inspect the Umbrella/OpenDNS event and policy/category for this exact hostname and request a classification review if appropriate. No service recreation or filter bypass is needed to identify that discrepancy. The moderator being able to load the site is consistent with different DNS/filter paths.
This explains why the quoted IP should not be treated as a failed Railway edge. It does not establish the cause of the separate SSH timeout, and I have not tested your affected network or changed any deployment.