[rabbitmq-discuss] Cluster, shovel, how to accomplish this scenario?

[Matthew Sackman]

On Thu, Apr 28, 2011 at 05:19:26AM -0700, Simone wrote:
> I need to have two instances of the server located in two different and 
> distant places connected by a network link with limited bandwidth. Clients 
> need to be able to connect to their nearest instance only and to communicate 
> with clients connected to the other server instance using the least 
> bandwidth possible. Is this accomplished using clustering, shovel or do they 
> have a different purpose? In any case, how would I set-up the infrastructure 
> described above?

Clustering is unlikely to work well over a WAN because Erlang's
clustering (specifically mnesia) doesn't cope well with recovery from
partitions.

Shovel is an alternative but can only do static configuration. If you
have, or can arrange at the boundaries between your two systems, a fixed
number of queues then the shovel can work well. However, note that with
clustering, you have one logical broker. With shovel, you do not: you
have two distinct brokers.

Matthew