The control connection is a dedicated connection used for administrative tasks:
checking schema agreement;
reacting to server events, which are used to notify the driver of external topology or schema changes.
When the driver starts, the control connection is established to the first contacted node. If that
node goes down, a reconnection is started to find another node; it is governed
by the same policy as regular connections (
advanced.reconnection-policy options in the
configuration), and tries the nodes according to a query plan from the
load balancing policy.
The control connection is managed independently from regular pooled connections, and
used exclusively for administrative requests. It shows up in Node.getOpenConnections, as well as
pool.open-connections metric; for example, if you’ve configured a pool size of
2, the control node will show 3 connections.
There are a few options to fine tune the control connection behavior in the
advanced.metadata sections; see the metadata
pages and the reference configuration for all the details.