[rabbitmq-discuss] Query on clustering

[Francesco Mazzoli]

Hi Rawat,

At Sat, 18 Aug 2012 08:41:57 -0700,
Rawat wrote:
> From what i  understood is that RabbitMQ Cluster is not transparent to
> Publishers and Consumers. e.g publisher has to switch to a different node of
> RabbitMQ Cluster if one node fails. Same way for consumers.

That's correct.

> Is there a way publisher & consumer can just connect with RabbitMQ cluster via
> a single nodename ? Without knowing which cluster node is serving them ?

Are you suggesting some king of "cluster name" that identifies the whole cluster
instead of a single node?  How would that work?

Since you can't rely on any cluster node to always be online, you have to change
if the one you're connected to goes down.

> I want to  install RabbitMQ cluster in 1x1 configuration ( both RAM nodes
> ).

Having a RAM only cluster is, in most imaginable scenarios, a very bad idea.
The performance improvements are minimum and you risk loss of data.  Why did you
want to do that?

> And my intention is only that if one node fails, then all publishing and
> consuming tranparently gets switched to other cluster node.

I'm afraid that the easiest way will be by catching disconnections and
reconnecting to another cluster node.

--
Francesco * Often in error, never in doubt