[rabbitmq-discuss] Shovel plugin behavior with durable queues...

[Sean Heaton]

Hello,

I am new to the list and RabbitMQ in general, so I hope this is not a stupid question, but I have seen some behavior that is strange to me when using the RabbitMQ shovel plugin.  I am using RabbitMQ v 2.7.1 running on Centos 6.0.  I have three machines in my setup.  Machines A and B are both clustered together.  Machine C is not part of the cluster.  It has the shovel plugin installed.  The plugin is set up to shovel messages from a durable queue on Machine C to a durable queue on Machine A.  In case it helps, my /etc/rabbitmq/rabbitmq.conf file looks like this:

[{rabbitmq_shovel,
  [{shovels,
    [{myShovel,
      [{sources,      [{broker,"amqp://MachineC"}]},
       {destinations, [{broker, "amqp://MachineA"}]},
       {queue, <<"myQueue">>},
       {ack_mode, on_confirm},
       {publish_properties, [{delivery_mode, 2}]},
       {reconnect_delay, 5}
      ]
    }]
 }]
}].

With this configuration, everything seems to work fine.  When I send messages to the queue on Machine C, they are shoveled automatically out to A.

So here is the weird behavior I mentioned earlier...  If I shut down the RabbitMQ service on Machine A, messages stack up in the durable queue on Machine C.  This is expected since Machine A is down.  However, if I then restart the RabbitMQ service on Machine C, all of the messages waiting in the durable queue get nuked.

Has anyone else seen this behavior?  Do I have a setting wrong or is there something else I can change so that this doesn't happen?

Thanks in advance for your help!

Regards,
Sean

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20120402/7ae6a368/attachment.htm>