Random 502 from Routelicious backend

[d2e]

Hello There
We have observed that the Fluid client occasionally receives random 502 errors from the Routelicious backend. The occurrence of these errors is highly unpredictable, sometimes happening once a day and other times only once every 2 or 3 days.

We are running Routelicious on AWS, fronted by an Application Load Balancer (ALB). After capturing the access log, we found that the proxy is receiving TCP RST signals from the target (Alfred), causing the proxy to send 502 errors to the Routelicious client. This issue also causes the client to crash with an unhandled error exception.

Within Routelicious, there is a configuration for setting the timeout, which is currently set to 0. By default, the keep-alive timeout on Node.js is 5 seconds, while the ALB idle timeout is set to 60 seconds. It is generally recommended that the target client keep-alive timeout should be greater than the proxy timeout.

I noticed that a similar issue occurred with Tinilicious in the past. As a fix, the keepAliveAgent package was configured in Tinilicious.
#11631

Any suggestion on why keepAliveAgent is not used in routelicious? And is there any way to configure keepalive timeout?