Three kinds of configuration could be possible for different kinds of cases: - Default strong consistency - Two replica one of which is raft learner - No RSM, only single replica for each shard/group without any write and read overheads