[rabbitmq-discuss] Questiong about plugins and clustering

[Alvaro Videla]

On Mon, Sep 23, 2013 at 3:41 AM, Simon MacMullen <simon at rabbitmq.com> wrote:

> What are you trying to do?
>

I want to have a plugin that works with a concept of topics and partitioned
queues. What does this mean?

I want to create a topic called "my_topic" which implies having a
consistent hash exchange (CHE) called "my_topic" + one queue called
"node_name.my_topic" bound to the my_topic exchange. The queue message
contents should live in the "node_name" node.

Whenever a node joins the cluster, I want to automatically declare a queue
called "new_node.my_topic" and bind it to the "my_topic" CHE. That queue
will be declared on new_node, so the message contents would end up in that
node.

With this I want to achieve some sort of "partitioned queues".

I plan this plugin to be an extension to the management plugin, so one
would create one of these topics via an HTTP call.

Also when connecting to a broker, one could ask this plugin via HTTP to
send the proper queue name for a particular topic, in order to consume from
it.

Does this makes sense?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20130923/011dcd5f/attachment.htm>