On 01/04/2010 04:45 PM, Robert Godfrey wrote:
As alluded to in my earlier mail, I spent a bit of time over the last couple
of weeks implementing federation in the Java Broker...
Trying to derive how it worked through playing with the C++ broker I did the
following
Set up three brokers A, B and C
Created an exchange on each of them called "myexchange", and set up two
dynamic links (one in each direction) between every pair of brokers
I then had a client create a queue on B and bind it to myexchange with the
binding key "mykey",
Now if I published a message to myexchange with routing-key "mykey" it
arrived at the client at B twice (one going direct from A to B, and once
going to from A to B via C).
Ted is probably the best person to comment on the above, but its quite
possibly 'expected' at present. We may of course want to address that.
Further when I then added a second client connected to C, creating a queue
on that broker and also binding it to myexchange with binding-key mykey...
And then delete consumer on B ... the federated bindings necessary to get
the messages to C are also deleted.
This is certainly a bug. It may simply be that the design relies on a
federated binding for each local binding but where the key is identical
the second binding doesn't actually result in a duplicate federated binding.
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
