Master/slave (http://activemq.apache.org/masterslave.html) is the standard way to do it, so that a second broker takes over if the first goes down. You have to make all your messages persistent if you want them available after a failover, because the persistence store is how the slave figures out what to do when it takes over.
There are three persistence store technologies: JDBC, KahaDB, and LevelDB. JDBC is considerably slower than the other two and is used much less frequently, and the primary reason I've heard for using it is when someone wants the ability to look at what messages are in the store. KahaDB is the default technology at the moment so it's widely used, but it relies on a shared filesystem between the master and the slave hosts. If you have a way to make your shared filesystem redundant (including both the storage device and the network path to get to it), this could give you full fault tolerance, or you could choose to live with having a single point of failure in either of those. LevelDB is meant to eliminate the shared filesystem while maintaining fault tolerance by requiring writes to go to a quorum of the brokers before considering it successful. But it's newer so there's a higher risk of bugs (though it has seemed to become more stable in 5.12.x; I wouldn't even consider running it on a previous version) and it'll require more hosts (the recommended cluster size is three brokers instead of two, plus you'll need to run three Zookeeper instances and I'm not sure if those can be run on the same hosts or if they need their own). Also, recently there have been a number of LevelDB questions going unanswered because I don't know LevelDB well enough to answer them and no one who does has done so, so it might be harder to get help with LevelDB right at the moment (but that's always subject to change based on who is willing to donate their time to answering questions on the list). But if you can live with the added complexity, its purpose is to do exactly what you want. Tim On Oct 20, 2015 12:14 AM, "shackman" <[email protected]> wrote: > Thanks for all the references. I'll start my reading schedule there. > > > Tim Bain wrote > > Also, your current setup doesn't provide fault tolerance for the > > standalone broker itself. Do you need it to? > > In short, yes. Do you have a particular reference I can read about this > topic? Or do you happen to have an xml file that is the "best practices" > for > fault tolerance? > > thanks again for all your help. > > > > > > ----- > Regards, Shackman > -- > View this message in context: > http://activemq.2283324.n4.nabble.com/Persistent-messages-not-moving-from-embedded-broker-to-remote-broker-tp4702884p4703142.html > Sent from the ActiveMQ - User mailing list archive at Nabble.com. >
