Keep Alive Timeout on Hydra Server

[sead-asurion]

We run an AWS ALB load balancer in front of Hydra pods running in a Kubernetes cluster. We are using Hydra 2.0.3. The ALB throws a small number of 502s every day.

I suspect that these are happening because the ALB idle timeout > the Hydra server's keep alive timeout. So sometimes a request gets to the ALB, the ALB has an existing connection to the Hydra server, it tries to use it, but mid-request the Hydra server's keep alive timeout expires for that connection, the Hydra server closes the connection, and the ALB returns a 502 to the client.

We've had a similar issue with our NodeJS and Scala services. This is a common problem when using an ALB. We solved it for NodeJS and Scala. This article describes the general issue in more detail.

Is there a way to increase the Hydra server's keep alive timeout to be > 60 seconds (which is the ALB default idle timeout)?