load balancing, nodejs and kubernetes

[Didier Colens]

Let's say we have pod A scaled to 4 replicas and pod B scaled to 2. pod B is a nodejs grpc client and pod A is a go grpc server, we can't use kubernetes' builtin service and kube-proxy load balancing because the client maintains a permanent connection to the server, so how can a client distribute its queries to all 4 pods and not only 2 (in the best case).

What is the proper way of doing this ?

[legen...@gmail.com]

I have a similar question. I setup a k8s load balancer for gRPC service, my client could connect to the service (both service and client use gRPC-java) using load balancer external IP, and my service received the request and could complete process, however gRPC client just can't receive the response. if i don't use k8s load balancing, and deploy client code to same cluster, or simply setup a local port forwarding proxy on client side, then it works without any issue. anyone know why k8s load balancing is not working for gRPC?