More on replication...
Right now, the live will send its nodeID to a replicating backup, so that the backup can use the same nodeID. The backup receives the nodeID as soon as possible because it (currently) uses it to monitor whether the live goes down or not.
Actually announcing the backup with that nodeID only happens after the backup is brought up-to-date with the live's journals, pages and largeMessages. So far so good.
What I need some help with are the consequences of initializing some instances at the backup with the wrong nodeID.
Specifically, at HornetQServerImpl.initializePart1() a ClusterManager is instantiated. A comment there explicitly mentions the need of only doing that after settling on a nodeID.
The ClusterManager initialized with the wrong nodeID is started at SharedNothingBackupActivation. But in any case, already at HornetQServerImpl.initializePart1() it is used to deploy a clusterConnection and that already uses the (old & non-final) nodeID.
Can anyone familiar with the cluster code (Clebert, I am thinking of you) comment on which would be the best way to deal with this?
