[rabbitmq-discuss] How to configure RabbitMQ cluster to get out of the scale.
simon at rabbitmq.com
Fri Mar 8 14:25:03 GMT 2013
On 07/03/13 23:31, t.otoda.freemail at gmail.com wrote:
> 1) What kind of point should I be careful about?
> the number of nodes of the cluster,
> the number of nodes mirrored queues,
> the number of queues which nodes of the cluster have, etc.
What do you mean by "be careful about"?
> 2) What kind of procedure should I let RabbitMQ get out of the scale in?
> I think RabbitMQ does not seem to get out of the scale only by
> having added a node.
> Because it doesn't do re-balance of queues between nodes automatically.
Yes, currently the expectation is you have a load balancer out front, or
that you have some other way to make your clients select nodes evenly.
Then new queues are declared on the node that the client declaring them
was connected to.
> If anyone know how to do re-balance of queues, please tell me.
This is something that comes up occasionally - in 3.1 it will be
possible to manually migrate queues around by using the HA "nodes"
policy, but doing so will involve deliberately making the queue fail
over, with attendant consumer cancellation. It's possible something
could be scripted on top of that for automatic queue balancing - is this
something for which there would be demand?
More information about the rabbitmq-discuss