#Improper routing of client -> edge -> service

11 messages · Page 1 of 1 (latest)

analog pebble
#

I am based in the Netherlands
My service's location is also in the Netherlands
My service is proxied via Cloudflare

When I made a request, my request is sent to an edge server in Asia (issue is not always happening)

modest walrusBOT
#

Project ID: N/A

analog pebble
#

Project 146b3390-088d-483b-9c1b-9a7e3b2be785

#

Request id YO2Z_cNfS1WEhn_HacI7Nw
Commit fdf8f62
Branch/env prod

fossil edge
#

does the routing issue happen if you directly access the railway provided domain?

#

I'm sorry but that doesn't quite answer my question

analog pebble
#

No, I am using a custom domain with Cloudflare proxy enabled
And not a domain generated by Railway

fossil edge
#

please try the domain generated by railway

warped trout
#

I noticed something similar for a railway service with region Amsterdam.

When requesting the public domain generated by railway directly from my browser (from EU west), i get railway/europe-west4, so all good.

But when I make an API request from certain cloud providers, like a Vercel serverless function or edge function hosted specifically in fra1 (Germany), I get railway/asia-southeast1-eqsg3a and the request time for the same route increases from ~200ms to ~1500-2000ms. I double-checked in Vercel's observability tab that it's actually requesting from fra1

I noticed that this occasionally lasts for multiple hours/days or something, then it eventually routes to railway/europe-west4 again as it should. 🤔

fossil edge
warped trout