> You are starting with the master and then the slave hook up
> to this master (more ore less you create an administration
> server).
Not really, the only administration that is done by a master node is
starting/stopping child nodes. That is it.
> When the master is going down the slaves run on their own
> and become zombies because all the configurations are on
> the master.
No, the configuration is on the configserver, not the master node.
If the master node goes down then its children will vote to see who will
spawn a new master server, then attach to it.
> My idea is that every server contains the configurations and
> the changes are propagated on all the other servers in the
> group. This would allow an administrator to pick up any
> server and able to manage the whole group. When any server
> goes down all the other server knows each others and can
> still propagate the changes and when a new server joins the
> group it can take the configuration from any other server
> and deploy the services and applications.
You could cluster the configserver and use jini as a fail over mechanism.
Having each node keep updated configuration information is an uneeded
burden, especially when configuration comes in the form of large files and
such.
--jason
_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development
