"Farm" Deployments in JBoss AS 5.x

Version 2

    Beginning with release AS 5.1.0.CR1, the "farm deployment" functionality previously available in AS 4.x has been restored. It was not available JBoss AS 5.0.0 or 5.0.1.

     

    In the future, this page will serve as documentation for the new AS 5.1.0 farm deployment functionality. At this point it is just a stub article.

     

    The AS 5.1.0 version of farming includes a few significant enhancements over what was available in the AS 4.x releases:

     

    • Handling of trees of content in the farm folder, including exploded deployments.  In AS 4, only leaf nodes were allowed in the farm folder, which precluded organizing content into subfolders, and, more significantly, precluded using exploded archives. In AS 5.1.0 this limitation is removed.
    • Unnecessary copies of existing content during startup eliminated. In AS 4, when a new node joined the cluster, the entire contents of the existing nodes' farm folder would be copied to the newly joining node, even if that node already had the same content on disk. This could be expensive if the farm folder includes a lot of content.
    • Improper "reincarnation" of removed content eliminated. In AS 4, if a deployment was removed from a running cluster, but was not removed from the farm folder on a node that was shut down, if that node was started again, the removed deployment would be pushed back out to the cluster and deployed. With AS 5.1.0 this no longer occurs.

     

    While the internals of how farming works are quite different from AS 4.x, the user experience is largely the same, excepting the above mentioned enhancements.  Basically, content copied to the farm folder will be distributed to the cluster and deployed on each node.

     

    One difference from AS 4 is the farm folder no longer has its own hot deployment scanner thread. Instead farm scanning runs off the main HDScanner service configured in deploy/hdscanner-jboss-beans.xml.