Setup: We are running three istio/envoy pods (each in a separate AZ) which are making "ExternalProcessor" GRPC calls to worker pods to add an http header to the original request.
The issue is: even when running six worker pods (two per AZ), only one in each AZ receives traffic. I assume this is due to the nature of GRPC/http2's long-lived connections.
If my assumption is correct, is it possible to open (at least one) connection to each worker pod (multiple connections from the envoy) and distribute the load evenly?