[rabbitmq-discuss] Failover in Clustering mode

[Matthias Radestock]

On 29/08/12 12:42, CharSyam wrote:
> I made a Rabbitmq cluster with 3 nodes.
>
> Originally, there are 1 disk node and 2 ram nodes.
>
> It worked well.
>
> but when I kill the process of node1.
>
> the queue_declare command failed( queue name is "task_queue" )
>
> and I couldn't make it again unless I recovred it.

That is the expected behaviour when the queue is declared as 'durable'.

While a queue's "home" node is down, clients connected to other nodes 
and attempting to re-declare the queue will get a 404-NOT_FOUND.

If other nodes were able to re-declare such queues they would then 
conflict with the original queue when that gets recovered.

Matthias.