Unstable http latency
kyvrixon
FREEOP

8 months ago

Simple express server just sending a json payload and doing nothing else has very unstable response times. The 499's is mostly /favicon.ico requests which i return a 204 immediately but some requests get 499'd as well occasionally for whatever reason. Expected response time should be around 50-100ms (geographical latency) rather than ~200+ average

Unable to see this on local. Other hosts, vercel for example, have stable response times of around 50ms despite the geographical difference

Project ID: 6f683200-7f99-4ad6-9431-ffe6ca564062
Service ID: 04b8d5c3-6743-4723-8e56-bad353cdf731

Solved

18 Replies

kyvrixon
FREEOP

8 months ago

the first one percy 😔


8 months ago

Hey,

Can you give me the output of this -

And then provide me with your geographical location, accurate to what you are okay sharing publicly.


kyvrixon
FREEOP

8 months ago

edge=railway/asia-southeast1-eqsg3a
zone=asia-southeast1-eqsg3a
ip=xxx
forwarded=xxx
hs=5c30016b09ac
req=_1L4gEb7RB-7QqjRDcO5xA

if you need the ip and forwarded lmk. Im over in australia. Made a few changes to the code assuming it could be a caching problem but made no difference. Internal reply speed you can see in the deployment logs (kept them there just in case, times from when the server first knows about the request to right before responding) but the http logs seem whacky if that makes sense


8 months ago

Can you check that 2.87s 499 response? Did it also hit the Asia edge proxy?


kyvrixon
FREEOP

8 months ago

yes


kyvrixon
FREEOP

8 months ago

incase you needed more info

1397069930092036000


8 months ago

Well, I would start by deploying the service into the Asia region.


8 months ago

Currently, it is in US-East, I am in US-East, and I just opened the root of the domain, and I am seeing a request for / that was 3 ms.


8 months ago

In fact, ALL services within that project are deployed in US-East.


kyvrixon
FREEOP

8 months ago

so it is geographical and not my crappy coding yes? If so thats manageable, anything over on my end is just development anyway but even then the response time should be stable or within a few hundred ms rather than going up to seconds correct? If you want the main server.ts file i can provide it if you'd like (no sensitive info there)


8 months ago

Yeah this is 100% caused by you routing your traffic from Asia to US-East, who knows what the sharks do with your packets at that point.


kyvrixon
FREEOP

8 months ago

haha, good to know its all normal. Cheers brody ❤️


8 months ago

Let me know what kind of improvements you see after you have moved all your services to Asia.


kyvrixon
FREEOP

8 months ago

will do <:salute:1137099685417451530>


kyvrixon
FREEOP

8 months ago

yup response time has significantly decreased and is much more stable, just looking here with multi region replicas, how does this work? Would the request get routed to the closest replica (for web services like api's, websites etc) as well as multiple replicas for the same region

whats the use case?


8 months ago

We have docs all on that 🙂


kyvrixon
FREEOP

8 months ago

forgot about docs lol, cheers bud


7 months ago

!s


Status changed to Solved brody 8 months ago


Loading...