a month ago
Hi Railway team,
My service is deployed only in Southeast Asia (Singapore, asia-southeast1) on the public domain volcano-dash-production.up.railway.app. Players in Brisbane, Australia reach it through your San José edge instead of Singapore, so every request crosses the Pacific twice.
Measured on 20 Oct 2026 from Brisbane (ISP Spintel, upstream Superloop AS38195):
- Every response has x-railway-edge: sjc1 and server: railway-hikari (example request ids 6CqWpVDdS1mEd8vxacI7Nw, yjR1MzfZR32hz7Yjmrpb1w, N9jL_C8ETb23LFSiVOLIQQ).
- The domain resolves to 69.46.46.30 (RLWY-HIKARI-01). Ping is 171 ms average.
- Traceroute: Brisbane -> Sydney (AS38195, ~20 ms) -> next hop ~170 ms, so the trans-Pacific hop comes before your network.
- A warm HTTPS request (HTTP/2, keep-alive) takes ~345 ms. A direct Brisbane -> Singapore path is normally ~100-130 ms.
Could you check the anycast or peering for Australian networks (Superloop AS38195, and Australian ISPs generally) so that they land on the Singapore edge? I understand edge selection follows BGP. In a similar case you sent preferred routes to a regional transit provider (the 2025 thread "Bad anycast routing with Metal edge network in sea region"). Is there anything I can change on my side, for example a setting for the Metal edge?
Happy to run more traceroutes or tests. Thank you!
1 Reply
a month ago
There is no setting on your service or project that chooses which edge location your users enter through. All edge locations apply to every service automatically, and your deployment region only controls where the app runs. Which edge a request lands on is decided by internet routing (BGP) between your users' ISP and our anycast addresses, and is based on network paths rather than distance. Requests that enter at San José are then carried over our internal network to your Singapore deployment.
Your measurements fit that picture. The jump from about 20 ms to about 170 ms happens on the hop right after Sydney, before traffic reaches our network, and every response shows sjc1. That means the Superloop path is handing traffic to a US entry point instead of an Asia-Pacific one.
If you collect more data, the most useful pieces are a traceroute to 69.46.46.46 from the affected network and a response captured with the X-Railway-Debug: 1 request header, which adds X-Railway-Upstream-Zone and shows which zone served the request (Edge Networking).
Status changed to Awaiting User Response Railway • about 1 month ago
24 days ago
This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!
Status changed to Solved Railway • 24 days ago