0
votes

We are thinking about using the JDBC datastore to set up a Master/Slave network of 3 brokers. According to the documentation this will cause one broker to acquire an exclusive table lock on the JMS table and thus become the master. The other brokers will not fully start and wait until they can acquire the lock and become the new master.

How are the messages stored that are waiting to be redelivered by the ActiveMQ Broker Redelivery Plugin? Are they also persisted in the JDBC datastore or are they local to the broker and wont failover to the next machine?

1

1 Answers

1
votes

The redelivery plugin sends messages to the Job Scheduler for later redelivery. These messages are placed in the JobSchedulerStore which is separate from the standard KahaDB or JDBC stores. You can configure the store location to be some shared location that supports file system locks like an NFSv4 mount so that this store folder is shared. There is no JDBC implementation of the scheduler store at this time.