Hey there! We've found the following might help you get unblocked faster:

• 🧵 502 Bad Gateway error despite successful container startup

• 🧵 502 Errors, think it might be Redis

• 🧵 Intermittent 502 Errors in Vendure Admin Panel

If you find the answer from one of these, please let us know by solving the thread!

i have been having these issues over the past 24 hours too and have not received a single word from support

I'm seeing similar - reported here: https://station.railway.com/questions/intermittent-502s-on-post-requests-fas-0a593b6b

There's also a few other reports:
https://station.railway.com/questions/frequent-timeouts-in-the-last-24-hours-b-c79e4932
https://station.railway.com/questions/recent-500s-and-timeouts-in-the-past-24-427e87b6

Seems something wider going on.

Help!

Hello,

We've added enhanced logging on our side and are actively working to make that information available to you in the HTTP logs.

We're not seeing elevated 502 rates across the platform, which suggests this may be related to your specific application configuration.

In the meantime, please let us know when you encounter a 502 status code so we can check the logs and help you identify what's causing the issue.

Best,

Brody

Thanks Brody, looking forward to having access to those logs.

An example from yesterday that you might be able to check. I got a 502 @ Jul 16, 2025 - 12:56:37.249 AM (UTC) which did not appear in my server logs but was recorded by my FE

We would need a request ID to be able to lookup the error on our side, that would be returned in a X-Railway-Request-Id header.

Note sure if relevant at all, so apologies if not, but worth mentioning...

We recently had elevated 502s as well (as soon as we migrated to Railway from AWS).

We discovered that the Apache KeepAliveTimeout on our end was 5 seconds, but at Railway, it's currently 15 seconds.

As a result, say then a subsequent request may come in after 7 seconds, Railway thinks the connection is still alive (within 15 seconds), Apache at our has end dropped it. This can result in a 502.

We adjusted our apache config to set the KeepAliveTimeout to 15 seconds (or 16, for a small buffer) and it seems to have helped (still monitoring).

Railway mentioned they might look at changing their 15 seconds to 5seconds to provide more better compatibility with most HTTP servers

Yep, exactly it. We're evaluating this change to see if that will resolve a lot of 502s that people were facing.

✅ The ticket Temporary Service Disruptions has been marked as completed.

Hi,

Apologies for the delay, I needed to update some tooling in order to extract the request ids.

We are still seeing 502s with the requests not reaching our server. We been investigating further from our side but we have not been able to identify anything. As far as we can tell the server is healthy and accepting requests. It's a pretty simple setup, just a node server in a docker container.

Here are some request IDs where a 502: Application failed to respond error was returned.

Aug 14, 2025 - 04:51:26.153 AM (UTC) - fnsvqGLPReOnEEJ-xtoGcA

Aug 12, 2025 - 09:40:28.209 PM (UTC) - P_Wx6MlQSHGS5tIunjUZ_Q

Aug 10, 2025 - 03:21:35.060 AM (UTC) - Dq9-R_BrSMGiTcCnz8YAag

Thanks!

Hello,

We have now made the errors our proxy experienced while communicating with your application available in your HTTP logs.

You can filter for 502 status codes with @httpStatus:502, and then click the log to see the error details in the upstreamErrors attribute.

From the last 502: connection closed unexpectedly

Some users have solved this by increasing their keepalive timeout within their application to 15 minutes, or adding another replica as we will retry with another replica if available.

Best,
Brody

Thanks for the update Brody.

Do I need to do anything to enable searching with @httpStatus? Like add structured logs?

It looks like right now I'm not able to see/use any of the http filters. It looks like the only one available to me is '@level'

Please make sure you open the HTTP Logs tab within a specific deployment. The global Logs tab will not have what you are looking for.

Wow! I've been using railway for just over 2 years and didn't even know this logs view existed! I've just been using the 'Observability' page.

Confirming I can see the 502s on those logs and an upstream error of connection closed unexpectedly so I will look into adjusting our keepalive timeout.

Thanks for the help Brody, enjoy the rest of your week!

You as well, let me us know if you see any improvements by increasing the keepalive or by adding replicas.