When trying to setup a failover safe master/slave setup for Search using JMS, it is important that index updates for a given index are only processed by a single consumer (message driven bean). In case there are multiple beans listening to the same JMS queue as part of the failover strategy, without so called message grouping, index updates could be applied out of order, since the queue consumer work independently. Message grouping ensures that all messages for the same group get consumed by the same consumer (while it is alive).

The key for message grouping is to set the standard JMS header JMSXGroupID. This is a per-condition for ActiveMQ as well as HornetQ.