1 Reply Latest reply on Sep 8, 2006 10:42 AM by Scott Stark

    Breaking clustering code out of the AS

    Brian Stansberry Master

      Looking for input on repository naming conventions.

      To help support Remoting and Messaging, I'm looking to extract some of the common usage clustering code out of AS cluster module into a separate project so it can produce a binary independent of the AS (see http://jira.jboss.com/jira/browse/JBCLUSTER-135).

      I'm thinking this will probably produce 2 binaries -- a "ha-client.jar" with the client-side classes (load balance policies, FamilyClusterInfo, etc) and server-side jar with whatever server-side elements I need for Messaging. The client-side jar is intended to be as lightweight as possible for use by Remoting. Wouldn't be surprised if the client-side stuff ends up in Remoting itself someday.

      I'm planning to put this in SVN and use Maven2 for the build. I've got that working for the ha-client.jar; trivial to do by copying from the jboss common project.

      I want to get this checked in so Remoting guys can use a snapshot of the client-side stuff.

      Here's what I'm thinking for the client-side stuff:

      SVN:

      https://svn.jboss.org/repos/cluster/ha-client/trunk

      An alternative is https://svn.jboss.org/repos/jbossas/projects/cluster/ha-client/trunk but I'm not clear on what the "jbossas/projects" namespace is meant to mean.

      When any server-side stuff is ready it would go in another subproject(s) at the same level as ha-client, e.g.
      https://svn.jboss.org/repos/cluster/ha-server/trunk

      Thirdparty repository (initially):

      repository.jboss.com/jboss/cluster/ha-client/snapshot/lib/jboss-ha-client.jar

      Maven repository (when needed):

      repository.jboss.com/maven2/jboss/cluster/ha-client/0.1/jboss-ha-client-0.1.jar

      An alternative to the use of "cluster" in the above name is "ha", but

      1) in labs and JIRA there is already a JBoss Cluster project, so using the same name is clearer.
      2) Clustering deals with both scalability and HA.

      Perhaps I should change ha-client.jar etc. to cluster-client.jar; it just seemed too long.

      Thoughts?