Transactionhandling broken since JCR-2523

["Zschorn, Andreas" <AZ...@zeb.de>]

Hi,

we upgraded from jackrabbit 1.6.0 to 2.2.4 and we have massive problems with the transaction handling afterwards.
For us the complete transaction handling is broken.

I assume it is a problem of our side so perhaps somebody can help.

Situation is:
Jackrabbit as connected via an jca adapter on jboss 4.2.3 via spring modules.

Our error is that that the internal sessions are not connected to the transaction anymore after the connection was going back to the pool.

After some debugging, I  found problem comes through the following changes in the JCAManagedConnection with ticket JCR-2523 at the JCAManagedConnection.
In the constructor the current session is connected to the xaResource via
<code>
        if (this.mcf.getBindSessionToTransaction().booleanValue()) {
            this.xaResource =  new TransactionBoundXAResource(this, (XAResource) session);
</code>

And when the connection is going back to the pool, a new session is created in the cleanup code via openSession and the xaResource is not anymore connected to the current session. The old code was just to call session.refresh(false) during cleanup.

The problem is that only the xaResource is updated from the current transaction with the appropriate transaction context. Not the session object anymore. The result is, that session.save immediately saves to the database, because it does not know it is in a transaction context. In our case the complete transaction safety does not work anymore. (XAItemStateManager tx is null)

I am not deep enough in the code of jackrabbit, but this is, as far as I can see it, a wrong behavior.
I will see that I will write a testcase for this, but I have the feeling that I missed something fundamental, that's the reason I did not write an ticket yet.

Has anyone the same problems, or know for sure that the transaction handlings works? The current testcases are not sufficient enough, as far as I can see.
Thanks and I hope somebody can help me to understand how it is supposed to work.

Andreas Zschorn
________________________________________
zeb/information.technology gmbh & co. kg
Hammer Straße 165
48153 Münster
Phone

+4925197128-593

Fax

+4925197128-101

E-Mail

AZschorn@zeb.de<ma...@zeb.de>

WWW

http://www.zeb.de



Amtsgericht Münster | HRA 5127
Geschäftsführer: Martin Danne | Jürgen Hofner | Andreas Schick

Diese E-Mail und alle angefügten Dateien sind vertraulich und ausschließlich für den Adressaten bestimmt. Sollten Sie nicht der bezeichnete Adressat sein, informieren Sie bitte umgehend den Absender. Die Inhalte dieser E-Mail dürfen in diesem Fall nicht an Dritte weitergegeben, für keine Zwecke genutzt und in keiner Form gespeichert oder kopiert werden. Im Fall technischer Probleme mit dieser E-Mail wenden Sie sich bitte an den Absender.


This e-mail and any attachments are confidential and may also be privileged. If you are not the named recipient, please notify the sender immediately and do not disclose the contents to another person, use it for any purpose, or store or copy the information in any medium. In the event of any technical difficulty with this e-mail, please contact the sender.

AW: Transactionhandling broken since JCR-2523

["Zschorn, Andreas" <AZ...@zeb.de>]

Hi,

I created and sample project and created an JIRA ticket.
JCR-2901

I would appreciate if someone could have a look at it.
For us, it makes the current version unusable. 
Thanks


-----Ursprüngliche Nachricht-----
Von: KÖLL Claus [mailto:C.KOELL@TIROL.GV.AT] 
Gesendet: Montag, 21. Februar 2011 10:55
An: users@jackrabbit.apache.org
Betreff: AW: Transactionhandling broken since JCR-2523 

As i know, the JCR-2523 fix closes the session and creates a new one to refresh the internal Caches.

This behaviour should not break the whole transaction handling...

Prior this fix a JCA Container can close the Connection depending on pool configuration and this was a problem.

If you can create a testcase to illustrate your problem we can look at it.

greets
claus

AW: Transactionhandling broken since JCR-2523

[KÖLL Claus <C....@TIROL.GV.AT>]

As i know, the JCR-2523 fix closes the session and creates a new one to refresh the internal Caches.

This behaviour should not break the whole transaction handling...

Prior this fix a JCA Container can close the Connection depending on pool configuration and this was a problem.

If you can create a testcase to illustrate your problem we can look at it.

greets
claus