Clustering and Failover are different things on HornetQ as you realized by running it and reading the docs.
How you phisically set your nodes is totally up to your requirements.
You have two boxes.
Box 1 will run the serverA and the backupForServerB
Box 2 will run the serverB and the backupForServerA
Note that this will only work if you have enough load available to handle both servers.
You could do the above also with shared storage. Which on that case the backup server will be completely idle and not having to do any work.
And that's just one example... don't take my example here as your solution without understanding it. I don't know your architecture and limitations.
You can configure this in many ways.
I'd suggest reading the HA chapter in the user manual. It goes into detail on this subject.