Skip to main content

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

  • From: michael keith <michael.keith@...>
  • To: jsr338-experts@...
  • Cc: Christian Romberg <cromberg@...>, eireland@...
  • Subject: [jsr338-experts] Re: question regarding JPA-JTA interaction (again)
  • Date: Mon, 18 Jun 2012 16:11:39 -0400
  • Organization: Oracle Corporation

Hi Christian,

A few comments.

The process described in the JTA spec on page 28 is illustrative, not prescriptive, so there is no requirement for the app server to delist the resource. That was just an example walkthrough of what *could* happen when a client closed a connection. I think you probably already knew that, though, because I think your point was just that if that *could* happen then it could break a JPA EM that wanted to write to the connection later on? Although, container data sources all pool the connections so in practice the connections never get closed, anyway (so the above case does not ever really occur).

Perhaps the strongest case is that the TransactionSynchronizationRegistry.registerInterprosedSynchronization() method in the JTA spec actually does state that beforeCompletion callback methods can access the resources, but that no transactional work can be performed on resources in afterCompletion. I would say that since the app server must provide this semantic we might be able to say that we are on reasonably firm ground. Would you agree?

-Mike

On 15/06/2012 9:56 AM, Christian Romberg wrote:
Finished my research of the ejb-core (3.0), jta (1.1) and jpa (2.0) specs and found nothing, that guarantees these two behaviors.

So for me it looks like, that we rely on undefined behavior.

I might have overseen something, and if anyone knows more, I would appreciate it, if you let me know.

Christian

On Fri, Jun 15, 2012 at 2:46 PM, Christian Romberg <cromberg@...> wrote:
Evan,

You are right, the issue would not occur then.

However so far I have been unsuccessful to find out, whether it is portable behavior, for the container to call Transaction.commit() while
there are any XAResource enlisted with this Transaction, where start() had been invoked, but end() has not.

Also I have not found out yet, whether there are scenarios, where the container must not delist the XAResource. (Although not
directly applicable, the control flow diagram p.28/29 in the JTA spec suggests otherwise.)

If both is defined, i.e. container does not delist XAResource it keeps in it's invocation context before calling Transaction.commit() and
Transaction.commit() handles still enlisted resources fine (by calling beforeCompletion() first, calling end() after that, and then prepare+commit)
then we (i.e. the JPA spec) would not have a problem.

If anyone knows, that that is defined somewhere, please let me know, I will also continue researching the related specs.

If that is not explicitly defined, than we (i.e. the JPA spec) do have a problem, because we rely on undefined behavior.

Regards,

Christian


On Thu, Jun 14, 2012 at 10:31 PM, Evan Ireland <eireland@...> wrote:
Christian,

In your step 6:

 "when the method invocation (of e.g. a session bean) ends, the container
will call Transaction.delistResource()"

I wonder if that is the root of the difficulty.

If the TM keeps JDBC connections (and/or XA resources) enlisted (in case
they are reused later in the same TX) until just before step 7b, then the
issue you are concerned about will not occur.

> -----Original Message-----
> From: Christian Romberg [mailto:cromberg@...]
> Sent: Friday, 15 June 2012 3:38 a.m.
> To: jsr338-experts@...
> Subject: [jsr338-experts] question regarding JPA-JTA
> interaction (again)
>
> Dear group,
>
> While investigating and implementing our JTA integration, I
> came across the following problem for the standard
> implementation approach:
>
> 1. container starts a JTA transaction
> 2. createEntityManager is called, and registers a
> Synchronization instance
> 3. as soon as the EntityManager acquires a JDBC connection,
> the container gets the XAConnection from the connection pool,
> passes the result of XAConnection.getConnection() to the
> EntityManager and registers the result of XAConnection.getXAResource()
> with the current JTA transaction
> 4. the TM calls XAResource.start(xid, TMNOFLAGS);
> 5. all work done by the EntityManager with that connection
> happens in the context of the transaction (branch) denoted by xid now
> 6. when the method invocation (of e.g. a session bean) ends,
> the container will call Transaction.delistResource() which in
> turn will
> call XAResource.end(xid, TMSUCCESS) [mandated by JTA spec,
> JTA 1.1 sequence diagram on page 29]
> 7. Transaction.commit() will drive the
> 2-phase-commit-protocol, before invoking prepare and commit,
> all "beforeCompletion()" calls
> of registered Synchronization instances are invoked
> 7a. the beforeCompletion() implementation of the
> Synchronization instance registered in step 2 is supposed to
> flush all changes to the database
> 7b. XAResource.prepare(xid) is invoked
> 7c. XAResource.commit(xid) is invoked
>
> The problem ist step 7a: The connection is not associated
> with xid, so this simply does not work.
> Even if the connection is lazily acquired and released, there
> is no portable way to trigger the necessary
> start() and end() invocations by the JPA implementations.
>
> So how is this supposed to work?
>
> (Btw. we have direct control over our XAResource
> implementation, so we can trigger flushing by having a
> callback from our XAResource to
> our EntityManager during XAResource.end(), but this naturally
> is not an option, if a 3rd-party XAResource implementation is used.)
>
> Regards,
>
> Christian
>
> --
> Christian Romberg
> Chief Engineer | Versant GmbH
> (T) +49 40 60990-0
> (F) +49 40 60990-113
> (E) cromberg@...
> www.versant.com
> <http://www.google.com/url?q=http%3A%2F%2Fwww.versant.com%2F&s
> a=D&sntz=1&usg=AFrqEzeeEBc_gN_8mxtt8xDB0tjXDXQVlw>  |
> www.db4o.com
> <http://www.google.com/url?q=http%3A%2F%2Fwww.db4o.com%2F&sa=D
> &sntz=1&usg=AFrqEzdo3Q40RwKQPBtnPIuBYQd1diFxJQ>
>
> --
> Versant
> GmbH is incorporated in Germany. Company registration number: HRB
> 54723, Amtsgericht Hamburg. Registered Office: Halenreie 42, 22359
> Hamburg, Germany. Geschäftsführer: Bernhard Wöbker, Volker John
>
> CONFIDENTIALITY
> NOTICE: This e-mail message, including any attachments, is
> for the sole
> use of the intended recipient(s) and may contain confidential or
> proprietary information. Any unauthorized review, use, disclosure or
> distribution is prohibited. If you are not the intended recipient,
> immediately contact the sender by reply e-mail and destroy
> all copies of
> the original message.
>
>
>
>
>




--
Christian Romberg
Chief Engineer | Versant GmbH
(T) +49 40 60990-0
(F) +49 40 60990-113
(E) cromberg@...
www.versant.com | www.db4o.com


--
Versant
GmbH is incorporated in Germany. Company registration number: HRB
54723, Amtsgericht Hamburg. Registered Office: Halenreie 42, 22359
Hamburg, Germany. Geschäftsführer: Bernhard Wöbker, Volker John


CONFIDENTIALITY
NOTICE: This e-mail message, including any attachments, is for the sole
use of the intended recipient(s) and may contain confidential or
proprietary information. Any unauthorized review, use, disclosure or
distribution is prohibited. If you are not the intended recipient,
immediately contact the sender by reply e-mail and destroy all copies of
the original message.







--
Christian Romberg
Chief Engineer | Versant GmbH
(T) +49 40 60990-0
(F) +49 40 60990-113
(E) cromberg@...
www.versant.com | www.db4o.com

--
Versant
GmbH is incorporated in Germany. Company registration number: HRB
54723, Amtsgericht Hamburg. Registered Office: Halenreie 42, 22359
Hamburg, Germany. Geschäftsführer: Bernhard Wöbker, Volker John


CONFIDENTIALITY
NOTICE: This e-mail message, including any attachments, is for the sole
use of the intended recipient(s) and may contain confidential or
proprietary information. Any unauthorized review, use, disclosure or
distribution is prohibited. If you are not the intended recipient,
immediately contact the sender by reply e-mail and destroy all copies of
the original message.






[jsr338-experts] question regarding JPA-JTA interaction (again)

Christian Romberg 06/14/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Evan Ireland 06/14/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Christian Romberg 06/15/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Christian Romberg 06/15/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Werner Keil 06/15/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

michael keith 06/18/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Christian Romberg 06/19/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Linda DeMichiel 06/18/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Werner Keil 06/19/2012

[jsr338-experts] Re: question regarding JPA-JTA interaction (again)

Christian Romberg 06/19/2012
 
 
Close
loading
Please Confirm
Close