Hi guys,

I'm struggling to understand a key part of the Hibernate documentation. Specifically 1.2.4...

http://docs.jboss.org/hibernate/stable/ ... ropagation

...says "In a transaction-scoped container managed entity manager (common case in a Java EE environment), the JTA transaction propagation is the same as the persistence context resource propagation".

During my HTTP request/response cycle, my app seems to be giving a new EntityManager to each @PersistenceContext-injected EJB. I assume therefore I am not correctly establishing a 'transaction-scoped container managed entity manager'? But the doc isn't clear how you actually establish the transaction scope?

I tried putting @TransactionAttribute at the class level, and at the method level, but my EJBs still receive different EntityManagers. If I were using a non-container-managed EntityManager I guess I could do something in a ServletFilter, but what is the correct way for container managed?

Thanks in advance,

Richard.

An answer for those that come this way after me :)

The documentation is a bit confusing. If you look here:

http://docs.jboss.org/hibernate/stable/ ... ml#d0e2823

There are some great examples of how to do it outside of a Java EE container. Documentation about 'em.createEntityManager' and 'em.getTransaction' and the JTA variant 'utx.commit()' and so forth. There is also discussion of how to look after your own EntityManager through the use of Filters or ThreadLocals or whatnot. Then there is a bit that says:

"all you have to do in a managed environment is to inject the EntityManager, do your data access work, and leave the rest to the container. Transaction boundaries are set declaratively in the annotations or deployment descriptors of your session beans. The lifecycle of the entity manager and persistence context is completely managed by the container."

And there are no further examples of JTA management. But this is misleading if you are trying to implement the (recommended) EntityManager-per-request pattern in a container-managed environment. When they say 'Transaction boundaries are set declaratively in the annotations... of your session beans' that's actually pretty useless, because it means you will start and end a transaction with each bean call during your HTTP request.

What you really need is to start and end the transaction around the entire HTTP request, and to do that you still need to use 'utx.commit()' and Filters and whatnot. Even if you are using container-managed persistence contexts, you'll still need a ServletFilter around your HTTP Request something like:


Alternatively, you can use Seam or CDI or something like that.

Hope that helps!