Hi all, Were continuing to tune our fault tolerant JackRabbit environment.
Thus far we have: · 2 machines · Each running Tomcat 5.0.30 with JackRabbit 1.3 as a model 2 shared resource · JackRabbit is configured with MySqlPersistenceManager, LocalFileSystem and Database Journal · Each Tomcat node hosts an web application allowing users to create/delete/edit JCR nodes. The web application also allows users to create new workspaces. The problem weve hit is that new workspaces are not clustered. Creating a new workspace (Workspace2) on node A does not result in a new workspace being created on node B e.g. Node A file system: · Repository o Repository o Version o Workspaces § Workspace1 § Workspace2 Node B file system: · Repository o Repository o Version o Workspaces § Workspace1 This is problem for us as new workspaces are being created all the time, hence they must be automatically clustered. Im guessing this problem occurs as the create workspace mechanism is JackRabbit specific and is outside of the JCR spec and event model? Potential solutions: · Move to a DatabaseFileSystem there are a couple of disadvantages to this for us. We like to be able to view and amend the files and in particular we regularly updated the custom_nodetypes.xml with new types as part of an upgrade process. Presumably the nodes types would be less accessible within the database. · Use a Network File Share for local files were trying to achieve the simplest solution for fault tolerance therefore wed rather avoid the complication of filesystem replication schemes (Linux DRBD etc) or the expense of SAN solutions. · Manually create copy new workspaces also not really an option as new workspaces are created all the time All suggestions welcome. Regards, Shaun.
