Problem with DefaultServicesManagerImpl in cluster (caches too much)

[Joram Barrez]

Hi List,

Currently we have CAS (3.3.0) running on a JBoss cluster. Services are
persisted to a database using the JPA implementation.

The problem that we now face is that changes done through the service
management on server A are not propagated on server B.

When inspecting the source code the cause is clear: the class
DefaultServicesManagerImpl has a local member 'services' which is used to
cache the services. When doing deletes, saves, ... this local member is
changed to reflect the change.
However, this change is *local* and is never propagated to the other server,
nor is the reload() method called.


Currently, we think of fixing it by calling the reload() every hour or so
the synchronize our CAS instances.

Am I correct in assuming there is problem with this class when running in a
cluster or am I missing something?

Regards

Joram
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://tp.its.yale.edu/pipermail/cas/attachments/20081118/a5c3aacc/attachment.html