Intra-cluster communication


I have a question about the intra-cluster communication.
It is unicast , multicast or broadcast ?


DCP is all unicast and that’s covering replication between nodes. Can you clarify the type of communication you’re looking at?

DCP is used for replication ?
I thought that DCP is only use for XDCR … In the documentation, it is use to reduce the latency to update some views but i don’t see where it is use to replicate data to another node in the same cluster (or i may read it too fast :P)

In the typical case that i have some application clients and a node fail.
How application clients will be notify by the new cluster topology ?
How the other nodes will be notify by a node unreacheable ?