load balancer behavior after changing service to headless

[Michael Fong]

Hi all,

We are looking for help/advise/pointer on the particular problem below.

We want to use client-side load balancing from envoy, and we change our server application service to headless.

We now observe some behavior change that client grpc requests of the same connection do not go to the same pod; in other words, the requests could be routed to any of the application pod (depending on which LB strategy we pick). We want to know if there is a way make envoy always stay a particular pod for the same client connection once it's established?

Given that the server application might be stateful and perform based on client state kept in its memory.

Thanks in advance!

[Yan Avlasov]

It sounds like what you want is to run Envoy as a TCP proxy not HTTP proxy.

--
You received this message because you are subscribed to the Google Groups "envoy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to envoy-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/envoy-users/8114d88d-d701-4c15-9255-cc4bde385debn%40googlegroups.com.

[Tony Allen]

Alternatively, you may want gRPC streaming instead of unary requests.

To view this discussion on the web visit https://groups.google.com/d/msgid/envoy-users/CADZdZWEGp%3Df%3DshjhkCi3_Gae9fBD79n1tFS8_88-QQMe%3DDQD8A%40mail.gmail.com.