[rabbitmq-discuss] ConnectionFactory and load balancing

[Simon MacMullen]

The cluster will connect consumers to the "master" node anyway. In 
theory it's likely to be more performant if you connect directly to the 
master node, but in practice I suspect the difference would be fairly minor.

Getting load balancing on a mirrored queue doesn't really make much 
sense as a concept since *all* the mirrors need to do *all* the work 
anyway - there's no real load to balance.

Cheers, Simon

On 10/01/12 19:57, Adam Rabung wrote:
> Let me restate :)
>
> Connection.newConnection(Address[]) seems to be the best way, short of
> using HAProxy, to safely attain a connection to a cluster in the
> presence of nodes going down.  When we are mirroring queues across a
> cluster, is it important to connect to the "master" node for a queue, or
> should we randomize the Address[] to achieve some load balancing?
>
> Any feedback appreciated,
> Adam
>
> On Mon, Jan 9, 2012 at 5:03 PM, Adam Rabung <adamrabung at gmail.com
> <mailto:adamrabung at gmail.com>> wrote:
>  > Hi,
>  > I am running an active/active cluster.  The newConnection(Address[])
>  > method of ConnectionFactory takes an array of addresses to try to
>  > connect to, in order.  If I'd like to avoid HAProxy for now, it seems
>  > like it would be wise for me to randomize this array of Addresses
>  > before calling newConnection, to achieve some degree of load
>  > balancing.  Is that a worthwhile approach?
>  > Thank you,
>  > Adam
>
>
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss


-- 
Simon MacMullen
RabbitMQ, VMware