[rabbitmq-discuss] Clustering Questions

[Michael Klishin]

On 16 Mar 2014, at 12:12, ratheesh kannoth <ratheesh.ksz at gmail.com> wrote:

>> http://www.rabbitmq.com/configure.html#configuration-file
> I saw a paramter  "cluster_nodes" . it mentions about automatic
> clustering.
> 
> 1 ) So here  clustering  refers to clustering of  brokers ...right ?

Yes, RabbitMQ nodes.

> 
> 2) Suppose a producer is sending  messages to consumer thru Broker A.
> Suppose Broker A goes down or
> too busy. Clustering would help to transfer the flow of messages thru
> Broker B, which is part of cluster ?

Cluster nodes transfer published messages to queue masters, which then
replicate them to queue mirrors (if any). If a node goes down, publishers
should reconnect to a different one. Queues that had masters on the unavailable
node will elect a new master automatically, consumers from those queues will
receive notifications via basic.cancel.

> 
> 3)  In question (2),  is producer and consumer has to take any action
> , which is informed by server ? or it is explicit ?

Clients that were connected to the unreachable node will see network connectivity
errors of some kind. Whether clients have a way to recover and how exactly it
works depends on your client.


MK

Software Engineer, Pivotal/RabbitMQ