[rabbitmq-discuss] Using Clustering with Amazon EC2

[Marek Majkowski]

David,

On Mon, Aug 9, 2010 at 18:30, David Allison <david at brightscope.com> wrote:
> I was hoping someone could discuss experiences they've had with using RabbitMQ clustering (as described in http://www.rabbitmq.com/clustering.html ) on Amazon EC2.  We currently only run one queue broker, and were considering using clustering to increase reliability and availability of the queues.

Clustering is not a reliability/availability mechanism. Sorry.
It should be only used to achieve a greater scalability, and can bring
benefit only for some group of use cases.

If you want availability, please consult our Pacemaker guide:
http://www.rabbitmq.com/pacemaker.html

Leaving that aside, did you have any particular issues with clustering on EC2?
Have you configured the firewall between hosts properly?

> Also, because the queues are currently not replicated, if a node goes down that contains a queue, is that queue immediately inaccessible, or are its contents (messages)  just lost, but clients could potentially still write to the same queue name on another broker in the cluster?

No. If a node that holds a queue goes down bad things will happen.

Marek