10 months ago
I have a basic golang REST API that takes 2-3ms to respond in localhost but between 30-40ms on railway (using metal for hosting with metal edge enabled for the TCP handler on the Amsterdam region).
My internet is not the problem: I have a multi-gigabit fiber with 1ms ping to get a response from google.com or other fast websites.
Is there something planed to address this ?
11 Replies
I don´t think that the problem is related to my particular project since I saw several tickets mentioning this and beside the extra latency everything is working fine.
And this is confirmed by the traceroute between something like google.com and my service hosted on Railway: there's a significant latency diff
there is no complexity in my config, it's only a simple REST API, everything is setup the way it should according to the Railway docs (+ the latency problem is also present on my Postgres service that was setup using a template with no config change)
10 months ago
Response times when running locally and running in the cloud will always be radically different. When running services locally, there is no network delay. When running services in the cloud, there is network delay
10 months ago
30ms seems fine. Is there a reason why that amount of latency doesn't work for you besides "not as fast as local"?
10 months ago
i would also like to note that the TCP Proxy has to round trip to GCP -> Metal -> GCP, so any traceroute to a TCP Proxy is not a good test
9 months ago
!s
Nothing, I was just wondering since when I disable the Metal Edge routing it lowers to 20ms and when I run a similar stack on something like AWS I can get sub 10ms
Status changed to Solved adam • 9 months ago
9 months ago
Your original traceroute was to a Google Cloud proxy. If you try to traceroute a service using metal edge now, you should hit our network instead. (66.33.22.xx)