[openstack-dev] [OSLO][RPC] AMQP / ZeroMQ control_exchange vs port numbers
doug.hellmann at dreamhost.com
Mon Apr 29 15:00:59 UTC 2013
On Mon, Apr 29, 2013 at 10:26 AM, Mark McLoughlin <markmc at redhat.com> wrote:
> On Mon, 2013-04-29 at 09:43 -0400, Eric Windisch wrote:
> > > My I'm thinking of an exchange is simply that it's a namespace
> > > under which topics live:
> > >
> > >
> > The discussion we were having was that we should NOT expose the
> > control_exchange in the API unless we also expose the queue connection
> > mechanism. The reason is that the control_exchange is presently a
> > point of clean sharding between projects.
> Yes, I'm proposing the concept of "exchange" purely as a namespace for
> projects to keep their topics separated ... and allow multiple instances
> of the project share the same space by allowing the exchange name to be
> user configurable.
> > Each unique value of control_exchange can be safely sharded to a
> > separate broker or, in the case of ZeroMQ, to a separate rpc_zmq_port.
> > This latter point is important because it feeds back into the ipc_dir
> > discussion we had; Projects can have their own IPC directory if they
> > run on their own rpc_zmq_port. Of course, sharding for RabbitMQ/Qpid
> > might also be useful...
> I don't think what I'm saying changes that. You can absolutely put each
> exchange on a separate broker.
It is difficult to do that using the current configuration schema, because
there are so many different options that have to be replicated for each
broker. So you end up with foo_exchange, foo_host, foo_port, etc. for every
service that you want to connect to. For ceilometer we just said that all
of the exchanges have to be on the same broker. For cells, I think they
reproduce the settings in the database.
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the OpenStack-dev