Apache Druid

I'm trying to remember how historicals and coordinators communicate with each other. Do they communicate directly or just via zookeeper? Or is that something where it used to be via ZK but there's a different approach more recently?

OK this is `druid.serverview.type`, and we still have it as `batch` in our cluster.

In this case — do historicals ever do any outgoing connections to other Druid processes, with `druid.serverview.type=batch`? Or do they just talk to Zookeeper and to deep storage, and receive incoming queries from broker?