Atomikos Forum

Hi

I am using Atomikos transaction manager with JBosscache and Hibernate. 

I have injected java TransactionManager into the JBosscache as stated in their documentation.  Hibernate looks up the transactionManager using the properties configuration.


The following is what I have for test purposes:

1. I programatically start a Java Transaction.
2. I do two things:
  a. I save an object into JBosscache programmatically using "put" operation. 
  b. Then I call my PersonDaoImpl to save a "Person" object.
3. Finally I commit the Java Transaction.

Now this is what has happened according to the log file:

1.  The object is saved in the table. I verified it in the table.
2.  After the transaction state has entered in the "TERMINATED" state, then the JBossCache tried to commit the changes.  However, there is no transaction at this point.  JBossCache gives the following error:
- Received JTA STATUS_UNKNOWN in afterCompletion()!  XA resources may not be in sync


Can any one shed light on this?


thanks

What versions are you using?

I am using Atomikos 3.6.5 with Hibernate 3.3.2.GA and JBoss Cache 3.2.1.

When using a Oracle database, only the JBoss Cache is rolling
back after encountering a JTA STATUS_UKNOWN. 
However, when I tested with a MySQL database the JBoss Cache is committing fine. 

I even gave on the Oracle DB the "grant execute on sys.dbms_system to <username>;" and still does not work with OracleDB.

thanks

I noticed that on documentation from JBoss <a href="http://docs.jboss.org/jbosscache/3.2.1.GA/faq_en/html/TreeCache.html#d0e453">http://docs.jboss.org/jbosscache/3.2.1.GA/faq_en/html/TreeCache.html#d0e453</a> that JBoss Cache currently does not participate in XA Transactions.  Could this have anything to do with the XA transaction not working?

thanks

Did you see/try this?

http://bit.ly/hNjMXf

HTH

I am not using JBoss Cache as a 2nd Level cache for Hibernate.  I am using JBoss Cache as a separate entity for storing my own objects/data.

Then I can't help, sorry. Would you have access to a budget for developer access? Then we could scan through the logs and provide some more one-on-one support...

Best