[rabbitmq-discuss] rabbitmqctl stop_app start_app loses all messages

Mon Jul 23 10:26:44 BST 2012

Hi,

On 21/07/12 02:26, Farshad Meshkinpour wrote:
> Could someone let me know what might be wrong with my configuration or
> what I am doing wrong?

This is expected behaviour. When slaves join they are cleared. Queue
contents is not eagerly synchronised, so slaves will only become
synchronised when the contents of the master queue is completely
replaced with new messages. If you restart each node in the cluster too
quickly for slaves to synchronise then messages will be lost.

If you expect multiple failures then you should ensure that you have
even greater redundancy. If you need to restart nodes for administrative
purposes then you should wait for the last slave to become synchronised
before restarting the next one.

-Emile