Regarding Guillaume's comment about equal nodes, I can describe a requirement we currently have in one of our projects.
The requirement is to partition functionality across a set of karaf nodes. I.e. some nodes (in a load balancing scenario) provides certain services, while others provide other services. This is to, for instance, have a layered ESB architecture or have some services with completely different SLA's isolated from the rest. We have for now implemented this with the Puppet provision system where we have configured our nodes and put them in "named sets" (classes in Puppet). These sets of containers is then provisioned onto each host by a puppet agent. So, returning to the subject, it would be nice to be able to "name" clusters and put nodes in a named cluster, and it would provision itself. Magic. Or would you rather create different hazelcast configurations? I don't know Hazelcast that well. Regards Bjorn On Wed, Apr 13, 2011 at 3:02 PM, Ioannis Canellos <ioca...@gmail.com> wrote: > I am starting this discussion in order to gather the maximum possible > feedback about the features we want our clustering engine to have. > > Here is my view on the subject: > > a) Support auto discovery of nodes (via multicast and / or unicast). > b) Configuration replication (replicate configuration changes across the > nodes). > c) Features/Repositories replication (replicate feature repositories and > states). > d) Bundle state replication (replicate bundle states). > e) Event Admin (broadcast events across the nodes of the cluster). > f) Distributed Service Registry. > > -- > *Ioannis Canellos* > * > http://iocanel.blogspot.com > > Apache Karaf <http://karaf.apache.org/> Committer & PMC > Apache ServiceMix <http://servicemix.apache.org/> Committer > * >
