broker uses anycast for amqp destination which is configured as multicast

using:

• broker configured with default-address-routing-type=MULTICAST (and default-queue-routing-type=MULTICAST)
• clients built with python-qpid-proton, but we have seen the same effect with other amqp clients

When a client is publishing to a new address, the address is created with routing type MULTICAST. this is ok.
When a client is subscribing to a new address, the address is created with routing type ANYCAST, which is unexpected.

To repeat:

• create plain broker using bin/artemis create --allow-anonymous --user admin --password admin broker1
• (optional) adjust broker1/etc/bootstrap.xml and/or adjust broker1/etc/jolokia-access.xml when the broker console can otherwise not be used/reached.
• update broker.xml: add these 2 lines to the address-settings for '#'
  <default-queue-routing-type>MULTICAST</default-queue-routing-type>
  <default-address-routing-type>MULTICAST</default-address-routing-type>
• start broker with broker1/bin/artemis run
• adjust the broker address in the 2 python files
• use supplied docker files to build and run a sender and/or a receiver

scenario 1:
run only the broker and the sender
see address 'demoaddress' appear as multicast address
any messages will disappear as expected because there is no consumer

scenario 2:
run only the broker and the receiver
see address 'demoaddress' appear as anycast (unexpected) address
a queue with the same name is created under it (consistent)

remove the address 'demoaddress' between runs because the effect only appears for new addresses

see the attached files

jbertram: as discussed in the mailing list