In which case it is starting to sound as though there might be a
problem with the db driver. Have you tried it with derby or similar?
Tim
------------------------------------------------------------------------
Date: Sun, 21 Oct 2012 00:49:14 +1030
From: [email protected]
To: [email protected]
Subject: Re: Aries eclipselink.adapter
Yes.
On 21/10/2012 12:48 AM, Timothy Ward wrote:
This was in addition to changing your datasource blueprint to:
<service id="xaDataSource" ref="stagingXADataSource"
interface="javax.sql.XADataSource">
<service-properties>
<entry key="osgi.jndi.service.name" value="jpa/jtaStagingDb" />
</service-properties>
</service>
wasn't it?
Tim
------------------------------------------------------------------------
Date: Sun, 21 Oct 2012 00:02:21 +1030
From: [email protected] <mailto:[email protected]>
To: [email protected] <mailto:[email protected]>
Subject: Re: Aries eclipselink.adapter
... my progress so far :
With this line in the persistence.xml
<jta-data-source>osgi:service/javax.sql.DataSource/(&(osgi.jndi.service.name=jpa/jtaStagingDb)(xa=true))</jta-data-source>
Yet, the result is still as before: all the participating
transactions apart from the offending one got commited.
My only choice is to MsSql, and the problem might be with the driver.
Of course, I might try to build the test case with other database
to verify this.
As for the latest suggestions to tweak the config.ini ... This one
is generated (Eclipse default option: generate with default
content) and
doesn't bother to specify the org.osgi.framework.system.packages
property.
I definitely not going to create that listing manually - it's crazy.
At this very moment I tend to point finger to MsSql ...
Regards,
Anatoly
On 20/10/2012 10:43 PM, Timothy Ward wrote:
Hi,
Have you set up your system packages to prevent people getting
wired to the split package in the JDK?
If you look at
https://svn.apache.org/repos/asf/aries/trunk/samples/blog/blog-assembly/src/main/filtered-resources/configuration/config.ini
You can see that the org.osgi.framework.system.packages
property sets a mandatory partial=true parameter on
javax.transaction and javax.transaction.xa
Regards,
Tim
------------------------------------------------------------------------
Date: Sat, 20 Oct 2012 21:37:05 +1030
From: [email protected] <mailto:[email protected]>
To: [email protected] <mailto:[email protected]>
Subject: Re: Aries eclipselink.adapter
I tried the option 2) - I recollect tried it some time before.
And now it also doesn't work - problem with MsSql driver
Caused by: java.lang.LinkageError: loader constraint violation
in interface itable initialization: when resolving method
"net.sourceforge.jtds.jdbcx.JtdsXAConnection.getXAResource()Ljavax/transaction/xa/XAResource;"
the class loader (instance of
org/eclipse/osgi/internal/baseadaptor/DefaultClassLoader) of
the current class,
net/sourceforge/jtds/jdbcx/JtdsXAConnection, and the class
loader (instance of <bootloader>) for interface
javax/sql/XAConnection have different Class objects for the
type javax/transaction/xa/XAResource used in the signature
at
net.sourceforge.jtds.jdbcx.JtdsDataSource.getXAConnection(JtdsDataSource.java:116)
at
org.apache.aries.transaction.jdbc.XADatasourceEnlistingWrapper.getConnection(XADatasourceEnlistingWrapper.java:63)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
So, the ball on the MsSql driver side? The bad news is that I
am not aware of any other driver, and that one worked
perfectly well with Spring framework JTA declarative support
in non-OSGi environment.
What do you think?
On 20/10/2012 9:25 PM, Timothy Ward wrote:
I'm glad I could help!
I'll check with Manning about the discount code, I wasn't
aware that it had an expiry date.
Regards,
Tim
------------------------------------------------------------------------
Date: Sat, 20 Oct 2012 21:08:05 +1030
From: [email protected] <mailto:[email protected]>
To: [email protected] <mailto:[email protected]>
Subject: Re: Aries eclipselink.adapter
Hi, Timothy.
Thank you, your message arrived very handy just when I am
wasting my Saturday night (UTC +9.30) in another attempt
to make JTA work.
For a more than a couple of week I put it on the back
burner, while doing other tasks.
I will look at your suggestions and also will purchase the
book - thanks for the generous offer. I had only green
paper and the source code so far.
So, I may want to ask for an autograph from the author :).
My OSGi crash course is lasting couple of month by now, so
I had only chance to digest (albeit still suffering
heartburn :) ) "OSGi in Action".
Though I realized at rather early stage that I have to
deal with OSGi enterprise as far as JPA/JTA container
support concerned.
Kind regards,
Anatoly
PS "The coupon code you have entered has expired" I
received this message when applied the code.
On 20/10/2012 8:25 PM, Timothy Ward wrote:
Hi,
I'm afraid I haven't had time to do a full review, but
from the log I see that your datasource services are
both being registered by blueprint using the
DataSource interface.
Unless you're enlisting the JTA datasource with
transactions yourself then this looks like the source
of the problem. If you want the Aries runtime to do
the enlistment then you need to register the
datasource as an XADataSource.
There are then two options:
1. You can let the transaction wrappers bundle do the
enlistment and add (xa=true) to your JTA-data-source
jndi name
2. You can let the JPA container do the enlistment by
changing the JTA-data-source jndi name to use
XADataSource as the interface. This will only work for
Aries JPA 1.0 and higher.
If you are after more information about setting up
OSGi applications with JPA then there's a whole
chapter about it in Enterprise OSGi in Action, along
with chapters about tools, testing, web applications
and remoting. You can get it at
http://www.manning.com/cummins and get 37% off using
the code eosgi37.
I hope this helps you get set up ok.
Regards,
Tim
------------------------------------------------------------------------
Date: Mon, 15 Oct 2012 01:55:59 +1030
From: [email protected]
<mailto:[email protected]>
To: [email protected] <mailto:[email protected]>
Subject: Re: Aries eclipselink.adapter
Thank you, Timothy.
I use the current release 1.0.0 of aries
http://aries.apache.org/downloads/currentrelease.html
I have attached the blueprint context files -
blueprint seems to be the only way at the moment to
use declarative, AOP style, transactions support.
And the test class invoked in blueprint-test.xml
blueprint-datasource.xml -- blueprint context of the
database bundle
blueprint-employee.xml - blueprint of employee bundle
blueprint-test - blueprint of the integration test
routines bundle
persistence-jta.xml - the JPA/JTA persistence unit
configuration
I must confess, that I don't know what is
auto-enlisting datasource.
Regards,
Anatoly
PS I also attached the log of the test with the
deliberately induced sql error. The rollback is
announced, but all the insert statements, except for
the offending one, still get committed, including the
cascade insertions.
So, no actual rollback is performed.
On 12/10/2012 9:40 PM, Timothy Ward wrote:
Hi Anatoly,
I'd be interested in seeing the configuration for
the transactions that failed to roll back, and in
knowing what version of Aries JPA you were using.
If you don't give the JPA container an
auto-enlisting datasource then you can end up with
non-transactional behaviour.
This is why we have the transaction-wrappers bundle.
Tim
> Date: Fri, 12 Oct 2012 09:46:39 +1030
> From: [email protected]
<mailto:[email protected]>
> To: [email protected]
<mailto:[email protected]>
> Subject: Re: Aries eclipselink.adapter
>
> Yeah, that's what I did a month ago:
>
>
~/projects/aries/jpa/jpa-container-eclipselink-adapter
> --- pom.xml (revision 1388340)
> +++ pom.xml (working copy)
> @@ -81,7 +81,10 @@
> <dependency>
> <groupId>org.apache.aries</groupId>
> <artifactId>org.apache.aries.util</artifactId>
> + <version>1.0.0</version>
> +<!--
> <version>0.4</version>
> +-->
> <scope>provided</scope>
> </dependency>
> </dependencies>
>
> Still, the transactions don't work as expected
neither with eclipselink
> nor with openjpa.
> For instance, if two methods participate in the
transaction (the same of
> tx id testified that that was the case),
> and the second fails, then the first one still
got committed.
> The message was that transaction is nominated to
rollback, but then
> Rollback exception followed.
>
> I send the message some time ago to the user
list asking if anyone knows
> why the eclipse link adapter has never been
included into the release.
> And what the actual status of it.
> Anyway, as far as my experience go the the aries
container failed for me
> on transactional support.
>
>
> On 11/10/2012 8:51 PM, Christian Eugster wrote:
> > Hi,
> >
> > I managed this by changing the version range
of aries.util in the pom.
> >
> > But now I have another problem. After
packaging I tried to run an
> > example in the osgi-container. I get a
ComponentDefinitionException
> > saying Unable to validate xml: Caused by
SAXParseException saying:
> > cvc-complex-type.2.3: Element 'blueprint'
cannot have character
> > [children], because the type's content is
element-only.
> >
> > My blueprint looks like following:
> >
> > <?xml version="1.0" encoding="utf-8"?>
> > <blueprint
xmlns="http://www.osgi.org/xmlns/blueprint/v1.0.0";
<http://www.osgi.org/xmlns/blueprint/v1.0.0>
> >
xmlns:tx="http://aries.apache.org/xmlns/transactions/v1.0.0";
<http://aries.apache.org/xmlns/transactions/v1.0.0>
> >
xmlns:jpa="http://aries.apache.org/xmlns/jpa/v1.1.0";
<http://aries.apache.org/xmlns/jpa/v1.1.0>>
> > >
> > <bean
> > id="testDAOBean"
> > class="ch.persistence.TestDAOImpl"
> > >
> > <tx:transaction method="*" value="Required"/>
> > <jpa:context property="em" unitname="herakles"/>
> > </bean>
> > </blueprint>
> >
> > as I see, there are no character children. But
what am I doing wrong?
> >
> > Thank you for help!
> >
> >
> >
>
>
> --
> Anatoly Osiko
> Software Engineer, Integration
> SolveIT Software Pty Ltd
>
> Adelaide | Brisbane | Chisinau | Melbourne | Perth
>
> D: +61 8 7071 4918
> T: +61 8 8221 5533
> M: +61 4 1980 0386
> F: +61 8 8221 5677
>
> SolveIT Software Building
> Level 1, 99 Frome Street,
> Adelaide, SA 5000
>
> www.SolveITSoftware.com
<http://www.SolveITSoftware.com>
>
>
--
Anatoly Osiko
Software Engineer, Integration
SolveIT Software Pty Ltd
Adelaide | Brisbane | Chisinau | Melbourne | Perth
D: +61 8 7071 4918
T: +61 8 8221 5533
M: +61 4 1980 0386
F: +61 8 8221 5677
SolveIT Software Building
Level 1, 99 Frome Street,
Adelaide, SA 5000
www.SolveITSoftware.com <http://www.SolveITSoftware.com>
--
Anatoly Osiko
Software Engineer, Integration
SolveIT Software Pty Ltd
Adelaide | Brisbane | Chisinau | Melbourne | Perth
D: +61 8 7071 4918
T: +61 8 8221 5533
M: +61 4 1980 0386
F: +61 8 8221 5677
SolveIT Software Building
Level 1, 99 Frome Street,
Adelaide, SA 5000
www.SolveITSoftware.com <http://www.SolveITSoftware.com>
--
Anatoly Osiko
Software Engineer, Integration
SolveIT Software Pty Ltd
Adelaide | Brisbane | Chisinau | Melbourne | Perth
D: +61 8 7071 4918
T: +61 8 8221 5533
M: +61 4 1980 0386
F: +61 8 8221 5677
SolveIT Software Building
Level 1, 99 Frome Street,
Adelaide, SA 5000
www.SolveITSoftware.com <http://www.SolveITSoftware.com>
--
Anatoly Osiko
Software Engineer, Integration
SolveIT Software Pty Ltd
Adelaide | Brisbane | Chisinau | Melbourne | Perth
D: +61 8 7071 4918
T: +61 8 8221 5533
M: +61 4 1980 0386
F: +61 8 8221 5677
SolveIT Software Building
Level 1, 99 Frome Street,
Adelaide, SA 5000
www.SolveITSoftware.com <http://www.SolveITSoftware.com>
--
Anatoly Osiko
Software Engineer, Integration
SolveIT Software Pty Ltd
Adelaide | Brisbane | Chisinau | Melbourne | Perth
D: +61 8 7071 4918
T: +61 8 8221 5533
M: +61 4 1980 0386
F: +61 8 8221 5677
SolveIT Software Building
Level 1, 99 Frome Street,
Adelaide, SA 5000
www.SolveITSoftware.com <http://www.SolveITSoftware.com>
