In case a slave node access the DBMS directly, then each node has its own connection pool?
Think about each domain/host instance as:
- one controller host or host acting as domain
- several server instances
Each in its own JVM. The domain/host is only an administrative process, there is no application request going to it.
Each server instance works similar to an standalone server (except administration control).
So each instance handle app-requests with its own pools/threads ... and might be part of a cluster or not.