I am assuming that you are using Toplink Essentials (TLE), the default
JPA provider of GlassFish V2. It uses a L2 cache that is not cluster
aware (Please see
http://weblogs.java.net/blog/guruwons/archive/2006/09/understanding_t_1.html
to know more about TLE caching). You can use explicit refresh and
Optimistic Locking in your data access layer to make sure that you get
the desired behavior while running in a clustered environment. You can
also switch to use EclipseLink, the default JPA provider for V3.
EclipseLink's L2 cache can be made cluster aware.
Rebecca Searls wrote:
>
>> Would a TopLink expert help us resolve a design issues
>> that MEP (Mobile Enterprise Platform) has with persistence.
>>
>> MEP uses Oracel TopLink. MEP was initially designed to run in a
>> (single) domain. We are moving it to run in a cluster. We are
>> having a problem with the TopLink data caching. We are using
>> the
>> default settings, so each instance is retaining a private
>> copy of
>> the DB data. When one instance updates a field, the other
>> instances
>> are not re-reading the data to get the changes. Configuring
>> TopLink
>> for our configuration is quite daunting, as there is no way
>> to determine
>> how close we are to having the correct combination of
>> settings. We
>> need someone who can give us guidance on this.
>>