On startup of the Broker, or restart of a Virtualhost, the Broker restores all durable queues and their messages from disk. In the Broker's default mode the Virtualhosts do not become active until this recovery process completes. If queues have a large number of entries, this may take considerable time. During this time no messaging can be performed.

The Broker has a background recovery feature allows the system to return to operation sooner. If enabled the recovery process takes place in the background allow producers and consumers to begin work earlier.

The feature respects the message delivery order requirements of standard queues, that is any messages arriving whilst the background recovery is in flight won't overtake older messages still to be recovered from disk. There is an exception for the out of order queue types whilst background recovery is in flight. For instance, with priority queues older lower priority messages may be delivered before newer, higher priority.

To activate the feature, set a context variable use_async_message_store_recovery at the desired Virtualhost, or at Broker or higher to enable the feature broker-wide.