[rabbitmq-discuss] Broker failover

Jason J. W. Williams jasonjwwilliams at gmail.com
Thu Aug 20 16:26:38 BST 2009


Hey Niko,

Can you set your SLB to persistent mapping based on client IP. That  
should keep each client on the server they are intially mapped to  
until that server fails.

That being said, I still believe HA should be done in Rabbit. SLB is  
not the right hammer in my opinion.

-J

Sent via iPhone

On Aug 20, 2009, at 4:09, Niko Felger <niko.felger at googlemail.com>  
wrote:

> Matthew,
>
> Thanks a lot for all this info!
>
> Is there a way to achieve some of this in a clustered setup? I guess  
> our requirements are not so much HA of the whole messaging  
> subsystem, but rather that an as-large-as-possible proportion of  
> messages gets processed _eventually_. The scenario I am mainly  
> worried about is when producers suddenly cannot publish anymore  
> because the server has gone away and thus any messages are lost at  
> that point.
>
> We tried using a dumb load balancer (in front of both producers and  
> consumers) to achieve this, but so far this has caused us more  
> trouble than it saved, see here: http://www.nabble.com/RabbitMQ-load-balancing-failover-with-LVS-td24683230.html#a24683230
>
> Thanks!
> niko
>
> On Wed, Aug 19, 2009 at 15:45, Matthew Sackman <matthew at lshift.net>  
> wrote:
> Niko,
>
> On Wed, Aug 19, 2009 at 03:41:59PM +0100, Matthew Sackman wrote:
> > Therefore, if HA and failover is important to you, we'd recommend  
> the
> > following:
>
> ...
>
> One further issue with this is that it means really all the nodes need
> to manually be configured the same, in terms of queues, exchanges and
> bindings. As producers don't know which node they're connected to,  
> this
> really demands that:
> a) Every producer can attempt configuration whenever it connects; or
> b) As consumers may need to be connected to every node, they could do
>   the configuration, as they're not in front of the load balancer; or
> c) You have some other process that does configuration.
>
> This is definitely one area where the clustered setup saves you effort
> as all nodes implicitly get configured in the same way.
>
> Matthew
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> http://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20090820/599f0b93/attachment.htm 


More information about the rabbitmq-discuss mailing list