Set <default-queue-routing-type>ANYCAST</default-queue-routing-type> in address settings for all addresses ('#')
 
Start all three brokers - cluster is formed.
Start CLI consumer on node a - put --message-count to 2000 - so it waits for 2000 messages
Start CLI producer on node b - consumer may consume the messages. Or not.
Start CLI producer on node c - consumer may consume the messages. Or not.
 
You can switch the nodes. Consumer from node c, produce on others etc. It sometimes does not work.
 
Then remove <default-queue-routing-type>ANYCAST</default-queue-routing-type> from all broker.xml files.
 
Cluster works as expected.