We're in the middle of moving our JBossMQ queuing to HornetQ as a part of an effort to enhance the system, and offer High Availability.
What we want:
- no multicasting (cloud won't let us)
- individual message failover (via master-slave or otherwise)
- fail-back without any downtime
- no database
Currently, I can't get failover-failback functionality without multicasting, and the documentation indicates 'failover is not supported for a static set of nodes'. This is okay for now as we willing to put up with messages being trapped on down servers during failover.
My question is: what HA functionality will 2.2 have that I should consider? Should we wait? Will the new HA features allow us to failover without multicasting? (master-master? master-slave?)
All this will be do able in the next release and is currently being worked on in trunk. We will support multiple backups, failback to master node all with needing multicast. However we will only support shared journal for this version, the version after will also have this but using replication.