[rabbitmq-discuss] Cluster reverted to old config!

Emile Joubert emile at rabbitmq.com
Thu Apr 4 15:07:05 BST 2013



Hi Pete,

On 04/04/13 13:06, Pete Kelly wrote:
> boxes lost network connectivity for a short period (or eth1 became eth0
> and vice versa) which caused rabbitmq on both boxes to crash.

Brokers will survive network interruptions, but probably not network
hardware reconfiguration.

> Is it possible it lost the definitions (because both cluster nodes
> crashed at the same time) so it just loaded in the last one which was
> imported via the management GUI?

If the management plugin was configured to load an out-of-date
definition at startup (available on brokers since v2.8.0) then it could
appear like a reversion. The same could happen on a pre-v2.8.0 broker if
you scripted the load of a configuration file at startup.

Disc nodes will retain configuration for durable resources during a
restart, assuming the Mnesia database directory survives intact.
Transient resources will not survive a broker restart, so that may
explain the discrepancy you see for exchanges and queues, but not for
vhosts. E.g. see

http://www.rabbitmq.com/amqp-0-9-1-reference.html#exchange.declare.durable

http://www.rabbitmq.com/amqp-0-9-1-reference.html#queue.declare.durable



-Emile






More information about the rabbitmq-discuss mailing list