Right now the setup that we have is the following:

1. pgBouncer (connection pooler) points to Primary node of the cluster

   1. all reads and writes go to one node

2. we also have replicas Kubernetes Service that points to Replica nodes of the cluster, but these connections are not pooled

The problem here, is that users cannot leverage scaling capabilities of PostgreSQL cluster on Kubernetes in full, cause connections to replicas are not pooled. 

We need to come up with the design and implement it, that is going to provide easy failover and scalability for PostgreSQL on Kubernetes.

 is researching the approach that we are using on non-Kubernetes environments.