a month ago
Hello Railway team,
I am experiencing a path-specific public Edge routing issue on my Railway service.
Service: fireal-sales-os
Environment: production
Active deployment: c352537e
Public domain: fi-real-sales-os-production-f18d.up.railway.app
Target port: 8080
The application itself is healthy and other routes on the same public domain work correctly.
Working routes:
- /dashboard
- /quotes
- /quotes/1
- /api/attachments/1
However, the following routes consistently fail:
- /api/drawings/mask/1
- /api/drawings/mask/1/original
- /api/drawings/test
The failing requests return Railway's Edge fallback response:
HTTP/1.1 404 Not Found
x-railway-fallback: true
Content-Type: application/json
Server: railway-hikari
Response:
{
"status": "error",
"code": 404,
"message": "Application not found",
"request_id": "aEP2FtVuSOeI7r3o0B_USg"
}
With X-Railway-Debug: 1, the response includes:
x-railway-edge: hnd1
x-railway-upstream-zone: railway/asia-southeast1-eqsg3a
I also tested with Cache-Control: no-cache and a unique query string to rule out caching, but the same Edge fallback response was returned.
Importantly, the same Mask API works correctly through Railway private networking. A request to:
/api/drawings/mask/1/original
through the private Railway network returns:
HTTP 200
Content-Type: image/png
This indicates that the application and Next.js route handler are functioning correctly internally.
Also, the failing public Mask API requests do not appear in Railway Network Logs, while /api/attachments/1 does reach the service successfully on the same public domain.
Current Railway settings:
- Edge Rules: none configured
- CDN Caching: disabled
- Attack Mode: disabled
There is also no issue with the public domain itself because /dashboard, /quotes, and /api/attachments/1 are accessible through the same domain.
Could you please investigate whether there is a stale or incorrect path-specific public Edge routing state for this Railway-generated domain/service?
In particular, could you check whether the public Edge is incorrectly routing or rejecting requests under /api/drawings/* before they reach the application?
Thank you.
7 Replies
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
milo
could you check if this also happens on a railway generated public domain?
a month ago
Yes. This also happens on a Railway-generated public domain.
The affected request is:
https://fi-real-sales-os-production-f18d.up.railway.app/api/drawings/mask/1/original
This is a Railway-generated *.up.railway.app domain.
On the same public domain, these routes work normally:
/dashboard
/quotes
/quotes/1
/api/attachments/1
However, these routes return HTTP 404:
/api/drawings/mask/1
/api/drawings/mask/1/original
/api/drawings/test
The response includes:
x-railway-fallback: true
and:
{
"status": "error",
"code": 404,
"message": "Application not found"
}
The same Mask API works correctly through Railway private networking and returns HTTP 200 with image/png.
So yes, the issue is reproducible on a Railway-generated public domain.
Thank you for helping us investigate this.
a month ago
i've just accessed that link, and everything seems to 404- did you take the domain/deployment down?
milo
i've just accessed that link, and everything seems to 404- did you take the domain/deployment down?
a month ago
No, I did not take the domain or deployment down.
It was working before, and the issue was originally limited to the /api/drawings/* routes. Other routes such as /dashboard and /api/attachments/1 were working normally on the same Railway-generated domain.
The active production deployment was c352537e.
I will re-check the current deployment and domain status now.
a month ago
No, I did not take the domain or deployment down.
I have just checked the Railway dashboard. The production deployment is still ACTIVE and shows "Deployment successful" on the fireal-sales-os service.
The active deployment is c352537e.
The Railway dashboard currently shows this public domain:
https://fi-real-sales-os-production.up.railway.app
The URL I previously tested and included in my report was:
https://fi-real-sales-os-production-f18d.up.railway.app
Both are Railway-generated *.up.railway.app domains.
I have not stopped or removed the deployment or domain.
Could you also check whether the f18d.up.railway.app domain is currently correctly attached to the active production deployment?
a month ago
i can't check that, but any route i go to on that domain doesn't connect, even thes ones u stated that work: /dashboard, /quotes, /quotes/1
milo
i can't check that, but any route i go to on that domain doesn't connect, even thes ones u stated that work: /dashboard, /quotes, /quotes/1
a month ago
Thanks, that is helpful.
No, I did not take the domain or deployment down.
I just checked the Railway dashboard again, and the production deployment is still ACTIVE, shows "Deployment successful", and the service is Online.
The active deployment is c352537e.
What you are seeing is different from what I was seeing earlier. At the time of my tests, /dashboard, /quotes, /quotes/1, and /api/attachments/1 were working on the same f18d.up.railway.app domain, while only the /api/drawings/* routes were failing.
So if the f18d domain is now unable to connect to any route at all, something may have changed on the Railway public routing side.
I have not stopped, removed, or changed the deployment or domain.
Could you check why the Railway-generated f18d.up.railway.app hostname is no longer connecting to the active production service?