Hello,

You can view the HTTP logs right besides the Deploy logs when you have a deployment open.

@Brody , when I do see a log, how can I figure out why it did have 502? Even today, I saw 2 requests getting this.

Typically you would be able to click them and see the reason, but that's currently unavailable

Look, for example this

But deployment region is this

Yep, response details are currently unavailable

I meant, the edgeRegion

edgeRegion is the region that accepted the request, and then forwarded it to your service in Asia

Aha, okay

So is there a way to figure out why this happens? Some sort of log or similar?

I look at this, and see it happened 1 every hour since 4am

^ Not at the moment, no.

But 502's 99% of the time are caused by application level issues.

Is this an NGINX under the hood?

No, custom proxy.

But 502's 99% of the time are caused by application level issues.
So I would agree with you but you see the upstream address is not logged. Meaning it could route traffic

Where as if I look at another call, I see routing to IPv6 ip

Provide the request IDs for the 502's and I can check them Tuesday.

I3f4oXmkQ8m22IwfAQeqjw
mt6FsGOJRdSE3VaGm3z_FQ
boMQYaeVRiaN9CwQn6XIxQ

Do you have a plan on Railway that gives us priority support?

We have the enterprise plan that comes with guaranteed response times within 1 hour

Okay, I book a call with your Sales

Thanks Brody

Awesome! and I'll let you know what I find Tuesday.

Hi @Brody , could we have an update on this?

I3f4oXmkQ8m22IwfAQeqjw - connection closed unexpectedly
mt6FsGOJRdSE3VaGm3z_FQ - connection closed unexpectedly
boMQYaeVRiaN9CwQn6XIxQ - connection closed unexpectedly

This means that your application is closing the connection that our proxy makes to your application.

Okay, but the logs didn't have upstream specified. Would this not mean the routing failed from edge proxy to the service?

Our infra engineer mentioned that it would be normal during application-level failures like this. The upstream IP would be part of the new response details that have yet to be made available in the UI.

Tried 2 endpoints, exit after 1ms

Don't have insights on this that much, as far as I can see the app is available to process, see no reason why it would exit immediately, not even with timeout.

I'm sorry, but we do not have any observability into what your application does or does not do.

@Brody , is there a way to bypass your edge proxy and route traffic directly to the service?

There is not.