Bean Managed Transactions and rollback

Question

Hi Everybody,
I am using Bean Managed Transactions in a Message Bean which is called every some time by an EJB3 timer. This Message Bean subsequently calls a Session Bean which uses Container Managed Transactions and uses the default transaction attribute which is SUPPORTS. The Session Bean methods might sometime throw a System Exception(inheriting from RuntimeException) which will rollback the Bean Managed Transaction which was started from the Message Bean.
When this happens and I try to invoke userTransaction.rollback() I get invalid transaction state exception and I suppose this means that is already rolled back.
Is there a way to get another transaction or set the transaction back to a valid state so I can carry on with some persistence tasks or the only way to do that is by suppressing the RuntimeException and throwing an Application Exception having the *@ApplicationException(rollback=false)* annotation? Can I suppress a System Exception though?
Thank you in advance!

Answer 1

Saroj wrote:
Hi All,
I would like to know whether we can use JDBC Connection Object's commit and rollback
methods to control Transaction in Bean Managed Transactions or not.You may use the JDBC connection's transaction support from an EJB. That being said, you
need to understand that it won't be the transaction that started declaratively by the
EJB container nor would it be a bean-managed transaction started through
UserTransaction.
FWIW, I question why you'd want to do this though. I'd use container-managed
transactions and let the container handle this for you. The transaction manager
includes a 1PC optimization so it's not going to do an XA/2PC tx if you only have a
single resource in the tx. Also, the EJB container includes all the logic to properly
handle rollbacks when exceptions are thrown etc.
Finally, your code will more maintainable and reusable if you use the container-managed
tx + JTA resources. If I later wanted to call another EJB or another JTA resource (eg
JMS perhaps) I could do it without having to rewrite all of your code.
-- Rob
>
>
Why is it required that we should use Java Transaction API to control the Transaction
in Our Beans?
I understand that if we are using Multiple Resources and need to use Transaction
then going for JTA makes sense. If I am using only Resource,for example, Only One
Connection then we should be able to use Connection's Transaction control.
I understand that other way to do the transaction is to use Container's transaction
services.
Please respond at the earliest.
Thanks in Advance,
Saroj