[rabbitmq-discuss] RabbitMQ warrens

Tue Aug 9 22:54:17 BST 2011

Hm -- that seems like it would work -- but the "RabbitMQ In Action"
seems to indicate that haproxy will only send traffic to the standby
if the master isn't up.

Do you have any details/links about the active/active HA approach?

On Tue, Aug 9, 2011 at 2:52 PM, Alexis Richardson <alexis at rabbitmq.com> wrote:
> Joe,
>
> The idea is to send two (or more) copies of the whole data stream to
> the brokers, such that each copy shares the fate of a broker.
> Therefore the death of a broker does not imply the death of a copy of
> the stream.
>
> But hang in there because "active/active" HA is looming and this will
> go further than the 'warren' approach..
>
> alexis
>
>
>
> On Tue, Aug 9, 2011 at 5:17 PM, Joe Van Dyk <joevandyk at gmail.com> wrote:
>> The Manning RabbitMQ In Action book states:
>>
>> "Now if what we really need is absolute "high availability" with no
>> possibilities of message loss we can build a "warren" using two
>> standalone RabbitMQ servers with a load balancer making them appear as
>> one to our applications."
>>
>> A "warren" is two unconnected rabbitmq instances with a load balancer
>> in front -- one rabbitmq is the master and the other is the standby.
>> If the load balancer notices the master is down, it will send the
>> traffic to the standby one, using a config like:
>>
>> listen rabbitmq 0.0.0.0:5680
>>  mode tcp
>>  balance roundrobin
>>  server master :5675 check inter 5000 rise 2 fall 3
>>  server standby :5675 backup check inter 5000 rise 2 fall 3
>>
>>
>> I don't get how this prevents message loss.  If there's messages in
>> the master's queue and it goes down (hard drive explodes), those
>> messages are lost, right?
>> _______________________________________________
>> rabbitmq-discuss mailing list
>> rabbitmq-discuss at lists.rabbitmq.com
>> https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>>
>