[rabbitmq-discuss] Federation in a cluster

[Matthias Radestock]

Yican,

On 29/01/13 17:05, Cao, Yican - 0551 - MITLL wrote:
> You are saying I only need to declare one setting in E?

As Simon said, parameters and policies are cluster wide, so declaring 
them on one of the cluster nodes is sufficient. You *can* declare them 
on all nodes, but there is no point; you'd just be (re-)setting the same 
cluster-wide setting over and over.

> ie.
> in E:
> rabbitmqctl set_parameter federation-upstream my-upstream
> '{"uri":"amqp://A", "expires":3600}'
> rabbitmqctl set_policy...
> ....

Yes.

> Now, it comes that if A which is my upstream fails, what would
> happen? [A, B, C] is running with HAQ (active/active).
> Do I just lost federation?

Yes, the federation link will be broken. As Simon says, the federation 
plug-in currently does not have any built-in failover. One way around 
that is to stick a load-balancer in front of [A,B,C] and connect to that 
instead of an individual node.

> [E, F] also runs HAQ cluster, what if E dies, how to config F continue
> to perform as down-stream nodes as HAQ suppose to do ? Can we achieve this?

 From http://www.rabbitmq.com/federation.html#getting-started
<quote>
If a node fails, links to upstream exchanges will be recreated on a 
surviving node.
</quote>


Matthias.