In ActiveMQ, failover is commonly used to refer to the Failover Transport
. Based on your question, I suspect that you are referring to ActiveMQ Master/Slave
and, more specifically, Shared File System Master Slave
. This is the only thing that would refer to disk storage in ActiveMQ besides message persistence
and you haven't mentioned that.
The ActiveMQ shared filesystem master/slave functionality utilizes a shared filesystem to provide a high availability configuration. The only requirement is that the filesystem support locking. Using the shared filesystem to store broker data, many instances of ActiveMQ can be pointed at the same data store location on the filesystem. The first broker instance that gets a lock on the data store location is considered the master and any other brokers pointed at this same location will remain in a warm standby state. If the master broker fails, then the lock on the data store location is released and the next broker to grab the lock will become the new master. There are some diagrams demonstrating this functionality in the Shared File System Master Slave
doc. And there's some more information in ActiveMQ In Action