We have a RabbitMQ cluster with three nodes configured. Each node has two virtual hosts (Vhost-A and Vhost-B). We need a possibility to move messages from Vhost-A to Vhost-B. To accomplish this, we setup a shovel directing messages from Exchange-1 on Vhost-A to Exchange-2 on Vhost-B.
rabbitmqctl -p Vhost-A set_parameter shovel shovel-exchange-1-to-vhost-b /
'{"src-uri": "amqp://user@/Vhost-A", "src-exchange": "Exchange-1", /
"src-exchange-key": "#", "dest-uri": "amqp://user@/Vhost-B", /
"dest-exchange": "Exchange-2", "add-forward-headers": false, /
"ack-mode": "on-confirm", "delete-after": "never"}'
This has the side-effect of replicating the messages on the destination Exchange-2. Meaning, the test-queue we bound to Exchange-2 on Vhost-B receives the same message three times (once from each cluster node).
How can we prevent this? Does it require a change in the shovel configuration or in the cluster configuration?
RabbitMQ version: 3.6.15
UPDATE 1:
We have two exclusive queues that cannot be deleted, because they are locked. Those queues disappear, once we disable shovel plugin on all cluster nodes. As soon as we reactivate the plugin on one node, the queues are created again.