[rabbitmq-discuss] highly available shovel in a cluster

Žilvinas Šaltys zilvinas.saltys at gmail.com
Mon Aug 13 17:23:50 BST 2012


We have a cluster made of two nodes with 1 HA queue that has a shovel
on the master node that shovels messages to itself from another
RabbitMQ broker over WAN. The problem is if the master node goes down
the shovel plugin needs to become active on the slave and to keep
shoveling. How to solve this problem?

One way I see is to have the shovel active on both nodes. But due to
how rabbitmq cluster works the shovel on the slave node would just
forward everything to the master node which would replicate back to
the slave. This seems inefficient because of unnecessary connections
made by the slave when only the master node needs to be doing the

Another way is to monitor both slave/master nodes externally if one
goes down then switch the config with the right shovel config and
restart the broker service.

Has anyone dealt with this before or knows of an easy way to solve it?

More information about the rabbitmq-discuss mailing list