Understanding Buddy Replication

[hollowm]

I'm just trying to get a clearer picture of the Buddy Replication process and how it works. I've read the articles on Buddy Replication for JBoss Cache 1.4.0 "Jalapeno" (http://wiki.jboss.org/wiki/Wiki.jsp?page=JBossCacheBuddyReplicationDesign), and on BuddyReplicationAndHttpSessions (http://wiki.jboss.org/wiki/Wiki.jsp?page=BuddyReplicationAndHttpSessions). Initially, my perception was that the configuration was manual, and that two nodes in a cluster would need matching configuration in the jboss-service.xml in order to participate in buddy replication... after reading the "Jalepeno" article (I read it second), I think it is more automated in buddy selection.

My question then, is this:

Is the Buddy Replication configurable so that you can (in a 3 node cluster) have app1 replicate on node 1 and node 2, and app2 replicate on node 2 and node 3, and app 3 replicate on node 3 and node 1? Is it configurable at the app level, and configurable enough to allow for this manual configuration, or is it automated, and everything just replicates to 'x' number of nodes based on the 'numbuddies' property?
[/url]

[brian.stansberry]

It's configurable per cache. If different apps share the same cache (e.g. web session replication), then, no, it's not configurable per app.

As for the process of picking buddies, I'd say its more extensible than configurable. That is, you specify your buddy locator class, so you could write your own implementation of the BuddyLocator interface that gives you custom behavior you want. (And, of course, if it's generically useful, please contribute it back.)