Is OSIV with a stateful EJB reliable? What happens, when under load (or otherwise), the container decides to passivate the stateful bean. Would that event be completely transparent to a conversation that is using the bean, the PC, and managed entities therein?
The Seam project appears to recognize that there is a problem with a PC surviving passivation. see http://docs.redhat.com/docs/en-US/JBoss_Enterprise_Web_Platform/5/html/Seam_Reference_Guide/ch33s02s02.html Seams solution from what i can gather, is to stash a reference to the PC at the end of an invocation on the bean, and restore the reference on the next invocation. The fact that the Seam project found it necessary to do this rather than leave it to the container raises a big red flag. IBM discusses the problem here http://publib.boulder.ibm.com/infocenter/wasinfo/v6r1/index.jsp?topic=/com.ibm.websphere.ejbfep.multiplatform.doc/info/ae/ae/cejb_main.html but their solution appears to be focused preventing the passivation from failing as opposed to preserving the PC. It would seem to me that the only reliable thing to do would be to ensure that passivation is veto'd for the bean, as passivating the bean in such a way that it could be seamlessly activated and the PC restored, on a lazy init event sounds like a very hard problem. Adam Bien, in his post http://www.adam-bien.com/roller/abien/entry/why_stateful_and_local_anti answers the issue with (the somewhat glib) "just disable passivation. It isn't needed and possible on most application servers I know...". (We are on WAS 7). I'm not sure if passivation can be disabled and at which level, eg all beans, specific bean type or specific instance. This portion of the EJB spec (JSR 220): A container must not passivate a stateful session bean with an extended persistence context unless the following conditions are met: • All the entities in the persistence context are serializable. • The EntityManager is serializable. suggests to me another possibility: just arrange that the PC always contains some entity that is not serializable. A bit of a hack though ... I'm also curious how Spring deals with this, since as pointed out by Tobias in this thread, it offers an implementation of OSIV that might tend to validate the concept. -- View this message in context: http://openjpa.208410.n2.nabble.com/Avoiding-detachment-aka-Open-Session-in-View-tp6618821p6633375.html Sent from the OpenJPA Users mailing list archive at Nabble.com.
