I am not an RHCS expert, but JBoss clustering is shared-nothing architecture, based on in-memory replication. So it makes no sense to use GFS with it.
Actually it shouldn't work at all as two JBoss instances can't share the same configuration directories (deploy, work, ...). But RHCS should be nice to balance / failover front-end apache_md_jk servers for your HTTP clients.
Why do you think you need a shared global filesystem for your jboss cluster?