JTA Transaction and CUrrent Session

Question

We have the following problem. We configured hibernate with
jta and when we execute a simple code that performs some hibernate
queries we get the following error:
"org.hibernate.LazyInitializationException: could not
initialize proxy - the owning Session was closed"
We discovered that if we leave the session opened the error
disapears, but on the long run this crashes the server throwing a
JDBC connection error.
We tried to use getCurrentSession() instead of opening the
session manualy but in this case we get the following error:
"org.hibernate.HibernateException : Unable to locate current
UserTransaction"
This its quite strange because the current session should be
bound to the current user transaction and its seems to us that our
hibernate cfg file its ok.
I´am posting part of my hibernate cfg file and the code
that I execute.
hibernate.cfg.xml
<hibernate-configuration>
<session-factory name="java:/hibernate/SessionFactory">
<property
name="hibernate.connection.datasource">java:fiap</property>
<property
name="hibernate.dialect">org.hibernate.dialect.PostgreSQLDialect</property>
<!-- Enable Hibernate's automatic session context
management -->
<property
name="hibernate.current_session_context_class">jta</property>
<property
name="hibernate.transaction.factory_class">org.hibernate.transaction.JTATransactionFactor y</property>
<property
name="hibernate.transaction.manager_lookup_class">org.hibernate.transaction.JBossTransact ionManagerLookup</property>
<property
name="jta.UserTransaction">java:comp/UserTransaction</property>
<!-- <property
name="connection.release_mode">after_statement</property>
-->
<!-- <property
name="hibernate.transaction.flush_before_completion">true</property>-->
<property
name="hibernate.transaction.auto_close_session">true</property>
and my java Code:
try {
tx = HUtil.getTx(); //JNDI lookup
tx.begin();
session = HUtil.getSessionFactory().getCurrentSession();
//JNDI Lookup
//session = HUtil.getSessionFactory().openSession();
Query query = session.createQuery("from Persona where
username = ?");
query.setString(0, userName);
persona = (Persona) query.uniqueResult();
query = session.createQuery("from Festival where id = ?");
query.setInteger(0, idFestival);
festival = (Festival) query.uniqueResult();
if (persona != null && festival != null) {
query = session.createQuery("from Inscripcion where
inscribeA = :persona " +
"AND inscriptosEn = :festival");
query.setParameter("persona", persona);
query.setParameter("festival", festival);
inscripcion = (Inscripcion) query.uniqueResult();
return inscripcion;
} catch (Exception e) {
e.printStackTrace();
throw e;
} finally {
tx.commit();
//session.close();
thanks

Answer 1

The scenario you described is typical ... "If a client calls m1" ... there are 2 cases:
(1) the client is already in a trans
(2) not (1)
Let (1). This is the case of a Client Controlled Transaction ... etc ... when s1 is called it begins a transaction in the client transactional context ... hence s1 can't call a kind of roolback but can force the rollback of the global transaction marking EJBContext.setRollbackOnly() ... s2 uses the same transactional context ... anayway the client is the only one that can commit. This happens because of "transaction required" as the transaction attribute, that in this case forces a flat transaction model. Different is the case if you uses "requiredNew" --> nested trans model ...
(2) is the same of (1) except that the trans starts at application tier.
"I still have one doubt. Wouldn't the stateless session bean loose the transaction context once I exit the second stateles session bean's method? Isn't that what stateless means? "
No:
stateless doesn't stay for the transaction context but for the conversational state with the client ... that is serializable ... it makes the beans more poolable ... that the stateful session bean has
bye