Automatic failover works, but does not fail-back-over...

[aaron.whiteside]

Hey Guys,

I seem to be having a problem with getting a JMS Client to fail-back-over. I am using HornetQ 2.0.0CR3.

I have manually created the ConnectionFactory and are specifying the following parameters. (the ConnectionFactory has been created passing two TransportConfiguration's)

        connectionFactory.setRetryInterval( 2000 );

        connectionFactory.setRetryIntervalMultiplier( 1.1 );

        connectionFactory.setMaxRetryInterval( 60000 );

        connectionFactory.setReconnectAttempts( -1 );

        connectionFactory.setConfirmationWindowSize( 1048576 );

        connectionFactory.setFailoverOnServerShutdown( true );

When I shutdown the primary broker as expected the client transparently fails-over to using the backup.

However when I proceed to bring up the primary again and shutdown the backup, expecting that the client would retry the original primary but it does not. As you would expect my application freezes waiting for the reconnection to the backup.

If I now startup the backup again, everything starts working again.

Is this a feature or a bug?

Cheers,

Aaron

P.S. The two brokers are configuration as a cluster.

[jmesnil]

This is a current limitation of HornetQ http://hornetq.sourceforge.net/docs/hornetq-2.0.0.CR3/user-manual/en/html/ha.html#d0e9456

When a backup takes over, it is not currently possible to restart the live server without stopping the backup server first.

This is planned for HornetQ 2.1

[timfox]

This is actually pretty common behaviour. ActiveMQ for instance does the same.

However, as Jeff says we're going to add functionality in 2.1 to allow a new backup to be associated to a live while running.