[rabbitmq-discuss] Guidance on Using F5 Load Balancer with RabbitMQ

[Richard Raseley]

Tim,

With regard to your comment:

"Ok, but you do realise that a single queue will reside on only one node,
and therefore messages sent to other nodes in the cluster will still have
to be routed/forwarded to the node on which the queue resides?"

We are using highly available queues, so this wouldn't be the case
(right?). I understand that there is intra-node communication required to
support this.

Regards,

Richard


On Sat, Mar 16, 2013 at 8:27 AM, Tim Watson <tim at rabbitmq.com> wrote:

> Hi Richard,
>
> On 15 Mar 2013, at 20:04, Richard Raseley wrote:
> > The intent is to abstract all production and consumption operations via
> the VIP (which itself is mapped to a pool which contains all of our
> RabbitMQ nodes) so that there is a single configured endpoint. With regard
> to load balancing, it will load balance across the nodes so that if (for
> example) 10 producer instances opened up 10 channels to the VIP, 5 would
> have one node as an endpoint and 5 would have the other (assuming a 2 node
> cluster and round-robin configuration).
> >
>
> Ok, but you do realise that a single queue will reside on only one node,
> and therefore messages sent to other nodes in the cluster will still have
> to be routed/forwarded to the node on which the queue resides?
>
> > I understand that any timeout settings would have to be tuned for the
> application, but I was looking for some high level general guidelines (e.g.
> "in situation x, consider y and z"). We've ensured that our timeout is
> greater than the heartbeat interval.
> >
>
> I think heartbeats are the main (general) thing to be considered. Perhaps
> other users who're more familiar with production configurations such as
> these will have more to add though.
>
> Cheers,
> Tim
>
> > Regards,
> >
> > Richard
> >
> >
> > On Fri, Mar 15, 2013 at 3:46 AM, Tim Watson <tim at rabbitmq.com> wrote:
> > Hi Richard,
> >
> > On 11 Mar 2013, at 20:55, Richard Raseley wrote:
> > > Is anyone able to share some "best practices" as it relates to using a
> F5 load balancer to distribute load between nodes of a RabbitMQ cluster?
> > >
> >
> > Can you explain a bit more clearly what you're trying to achieve please?
> Using a load balancer to maintain connectivity to a cluster is recommended
> and quite common, but that isn't going to 'distribute load between nodes'
> per se. Are you trying to distribute the load of publishing over multiple
> nodes, or consuming, or what? There are various things to consider
> depending on your goals.
> >
> > > I am most interested in the persistence and timeout settings which we
> should be using in conjunction with RabbitMQ.
> > >
> >
> > There cannot be any specific settings that will work for all deployments
> - this will be application specific I would've thought. The only advice
> that springs to mind is if you are setting AMQP heartbeats on your
> connections, then make sure the proxy timeout is aligned with (i.e.,
> greater than) these to avoid disturbing connections unnecessarily - but
> that's pretty obvious.
> >
> > Cheers,
> > Tim
> > _______________________________________________
> > rabbitmq-discuss mailing list
> > rabbitmq-discuss at lists.rabbitmq.com
> > https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
> >
> > _______________________________________________
> > rabbitmq-discuss mailing list
> > rabbitmq-discuss at lists.rabbitmq.com
> > https://lists.rabbitmq.com/cgi-bin/mailman/listinfo/rabbitmq-discuss
>
> _______________________________________________
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
> https://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/20130318/35fa26a7/attachment.htm>