[rabbitmq-discuss] Cluster breakage

[Emile Joubert]

Hi,

On 19/02/13 10:07, PSL 88506 wrote:
> machine-1 shows machine - 1 and machine - 3 are running and machine-2 is
> Node not running.
> 
> machine-2 shows machine - 2 is running and machine-1 and  machine-3 are
> Node not running.
> 
> machine-3 shows machine - 1 and machine - 3 are running and machine-2 is
> Node not running.

This is entirely consistent with a network interruption.

> We checked with network team, if there are any network fluctuations
> during that period.

The most likely explanation for what you saw is a network interruption.
If clustered nodes are not able to communicate for some period (about a
minute by default) then the cluster will break. Rabbit clusters do not
cope well in such environments. You should use federation or the shovel
instead if you have poor network reliability:

http://www.rabbitmq.com/partitions.html




-Emile