2 months ago
I do not have a way to reproducing this. But noting this occasionally
In my nextjs server components, deployed on vercel, requests to my railway server timeout and there is nothing on the railway logs of any such request coming in/
We just migrated to a new backend architecture, a separate project on railway. When we linked our frontend nextjs app, client side requests took like 16secs for a simple GET request, in railway logs it shows 1sec to fulfill the request.
This results in really bad UX and causes frustration for users.
Why is it happening?
2 Replies
2 months ago
Do you perhaps have app-sleeping/serverless enabled on your Railway service? App sleeping can lead to cold boot which can make requests take longer if not time out
2 months ago
Hmmm, what framework are you using for the backend
2 months ago
Interesting. You had mentioned originally that the service would randomly do this. Does it still randomly occur and is it at all reproducable?
If not, are there any times/situations that it happens more then others?
2 months ago
>In my nextjs server components, deployed on vercel, requests to my railway server timeout and there is nothing on the railway logs of any such request coming in/
Could this indicate that the requests do not reach the Railway server, possibly due to some connectivity issue between the two? Are there too many hops between Vercel and Railway, or some other reason why the requests could get dropped before reaching Railway? ( Sorry if i misread this, I am new to Railway )
Recently it started happening alot on my office network. I used 1.1.1.1 and that fixed it. Prolly an issue with my internet provider.
But it still does not explain why it happens on other user devices, when the frontend tries connecting with the backend. Or when a server component deployed on vercel tries to access an api endpoint to render server html. Its totally random and unfrequent, neither can I reporduce it nor can I give a confident answer on when times/situations it happens more often.
I was hoping that I might be missing something that I could fix easily, but we will start monitoring this extensively and I will reach back with my findings.