Re: EntityManager Propagation Problem under 3.1
Hi Jim, I finally got to the bottom of the issue. There definitely was a bug, but it was in OpenEJB 3.0. We must have inadvertently fixed it when we did the stateful container rework between 3.0 and 3.1. It took me a while to notice this because of all the parent/child names in the sample, but it turns out your stateful session beans are syblings as they are created like so: ParentLocal parent = context.lookup(ParentBeanLocal); ChildLocal child = context.lookup(ChildBeanLocal); In order for the ChildBean stateful session bean to really be a child it must be created by the ParentBean. Meaning ParentBean must be the one to lookup the ChildBean or have it injected via @EJB. Here's a relevant JPA spec quote: 5.6.2.1 Inheritance of Extended Persistence Context If a stateful session bean instantiates a stateful session bean which also has such an extended persistence context, the extended persistence context of the first stateful session bean is inherited by the second stateful session bean and bound to it, and this rule recursively applies -- independently of whether transactions are active or not at the point of the creation of the stateful session beans. If the persistence context has been inherited by any stateful session beans, the container does not close the persistence context until all such stateful session beans have been removed or otherwise destroyed. If you add an @EJB ChildLocal child ref inside ParentBean and update your test code to not lookup the ChildBean and instead get it from the ParentBean, than all the tests pass as expected. Not sure what kind of leak we had in the 3.0 release that allowed non- child beans to mess with the persistence context of another stateful bean -- sort of a double bug as it also means two unrelated stateful beans wouldn't be allowed to have their own private extended persistence contexts -- but I'm certainly glad it's gone. Note that if you rely critically on this bug we could add it in as a feature. I use bug in quotes because tracking the persistence context by the thread as if it were in a transaction and regardless if stateful inheritance could be a useful technique. One might think of it as a new kind of persistence context management that has yet to make the spec. -David On Jan 27, 2009, at 8:01 PM, David Blevins wrote: Hi Jim, Wanted to give you an update. I've been looking into this one and so far haven't been able to find the source of the issue. In addition to your sample I've been adding more tests in this area -- the java ee tck has plenty but it's nice to have some of our own. Seems the basic functionality of extended persistence contexts is working: http://svn.apache.org/repos/asf/openejb/trunk/openejb3/container/openejb-core/src/test/java/org/apache/openejb/core/stateful/EntityManagerPropogationTest.java I'm going to keep working with your sample, but if you wanted to help expand that test case that could be very useful. You did some good work in the sample app. -David On Dec 10, 2008, at 9:33 AM, David Blevins wrote: On Nov 21, 2008, at 3:04 PM, JimOR wrote: David, Looks like OpenEJB3.1 is the culprit. EJB3.0 with JPA 1.0.1, 1.1, 1.2 and 1.3-SNAPSHOT work as expected, but 3.1 with any of the JPA jars fails. This sample http://www.nabble.com/file/p20631026/sample.zip sample.zip should give you a head start recreating the scenario. I had to remove the OpenJPA jars to fit within the file upload size limit, drop them into sample/lib and you should be good to go. The sample/test folder contains the junit xml results of the test case under the various classpaths. FYI, filled a JIRA for this: https://issues.apache.org/jira/browse/OPENEJB-970 (notifications: https://issues.apache.org/jira/secure/ViewIssue.jspa?id=12410384watch=true) Thanks for putting that sample together! -David
Re: EntityManager Propagation Problem under 3.1
Hi Jim, Wanted to give you an update. I've been looking into this one and so far haven't been able to find the source of the issue. In addition to your sample I've been adding more tests in this area -- the java ee tck has plenty but it's nice to have some of our own. Seems the basic functionality of extended persistence contexts is working: http://svn.apache.org/repos/asf/openejb/trunk/openejb3/container/openejb-core/src/test/java/org/apache/openejb/core/stateful/EntityManagerPropogationTest.java I'm going to keep working with your sample, but if you wanted to help expand that test case that could be very useful. You did some good work in the sample app. -David On Dec 10, 2008, at 9:33 AM, David Blevins wrote: On Nov 21, 2008, at 3:04 PM, JimOR wrote: David, Looks like OpenEJB3.1 is the culprit. EJB3.0 with JPA 1.0.1, 1.1, 1.2 and 1.3-SNAPSHOT work as expected, but 3.1 with any of the JPA jars fails. This sample http://www.nabble.com/file/p20631026/sample.zip sample.zip should give you a head start recreating the scenario. I had to remove the OpenJPA jars to fit within the file upload size limit, drop them into sample/lib and you should be good to go. The sample/test folder contains the junit xml results of the test case under the various classpaths. FYI, filled a JIRA for this: https://issues.apache.org/jira/browse/OPENEJB-970 (notifications: https://issues.apache.org/jira/secure/ViewIssue.jspa?id=12410384watch=true) Thanks for putting that sample together! -David
Re: EntityManager Propagation Problem under 3.1
On Nov 21, 2008, at 3:04 PM, JimOR wrote: David, Looks like OpenEJB3.1 is the culprit. EJB3.0 with JPA 1.0.1, 1.1, 1.2 and 1.3-SNAPSHOT work as expected, but 3.1 with any of the JPA jars fails. This sample http://www.nabble.com/file/p20631026/sample.zip sample.zip should give you a head start recreating the scenario. I had to remove the OpenJPA jars to fit within the file upload size limit, drop them into sample/lib and you should be good to go. The sample/test folder contains the junit xml results of the test case under the various classpaths. FYI, filled a JIRA for this: https://issues.apache.org/jira/browse/OPENEJB-970 (notifications: https://issues.apache.org/jira/secure/ViewIssue.jspa?id=12410384watch=true) Thanks for putting that sample together! -David
Re: EntityManager Propagation Problem under 3.1
On Nov 18, 2008, at 4:29 PM, JimOR wrote: Apologies in advance for cross-posting, but my post to OpenJPA ( http://n2.nabble.com/EntityManager-Propagation-with-Extended-PersistenceContext-td1495977.html http://n2.nabble.com/EntityManager-Propagation-with-Extended-PersistenceContext-td1495977.html ) hasn't gotten any hits. To repeat from that post, I have: A) Parent entity with LAZY OneToMany ListChild, cascade: PERSIST,MERGE,REFRESH B) Child entity with EAGER ManyToOne Parent, cascade: PERSIST,MERGE,REFRESH C)Abstract CRUD class with injected Extended PersistenceContext, and requisite CRUD methods D) Parent Stateful bean that extends CRUD E) Child Stateful bean that extends CRUD Classes are enhanced in place via ant prior to junit test run. Under OpenEJB3.0 (OpenJPA 1.0.1), Any/All managed entities show as managed by either bean, but under OpenEJB 3.1 (with deployed OpenJPA 1.1.0 AND 'latest binary' 1.2.0), only the bean responsible for the load/ persist/merge of an entity shows the entity as managed. The following snippet works fine under 3.0, but throws an exception under 3.1: Parent p = parentBean.insert(new Parent(Foo)); Child c = childBean.insert(new Child(Bar, p)); WARN org .apache .geronimo.transaction.manager.TransactionImpl.beforeCompletion(Ln 516) - Unexpected exception from beforeCompletion; transaction will roll back openjpa-1.1.0-r422266:659716 nonfatal user error org.apache.openjpa.persistence.ArgumentException: The given instance entity.Parent-6264 is not managed by this context. FailedObject: entity.Parent-6264 Hi Jim, I'm not sure what would cause this issue. We should try isolating things so we can figure out where to look. First thing would be either: a. swap out the OpenJPA version in your OpenEJB 3.1 install with OpenJPA 1.0.1, or b. take your working OpenEJB 3.0 install and upgrade your OpenJPA version from OpenJPA 1.0.1 to OpenJPA 1.1.0 or OpenJPA 1.2.0 I'd go with b. If that recreates the issue than you can rule out OpenEJB and update your post on the OpenJPA list saying you know it isn't an OpenEJB issue. If it doesn't recreate the issue than we know it's something on our end and we can dig deeper. -David
