I may be jumping the gun since 3.0 is still just Alpha, but I've been anxious to get started with 3.0, AS 7 and BDBJE so I've been starting with some tests. Here is one that did not work. Configuration:
- AS 7.1 and 3.0 Alpha 5 kit, using modified standalone-modeshape.xml (configured for BDBJE cache store instead of file store, with purge set to false)
- modeshape-schematic jar installed directly in AS 7 infinispan module to avoid classloading issues (see JIRA), and added as a resource in module.xml
- BDBJE cache loader jars installed directly in AS 7 infinispan module to avoid classloading issues (see JIRA), and added as a resource in module.xml
- SLSB with BMT, creating a hierarchy of 1000 nodes, depth of 3, in batches of 100, each batch in a separate transaction. (I've started with 1000, but ultimately want to see 1M working).
The node hierarchy being created is:
Each top level branch is created in a separate BMT transaction calling: UserTransaction.begin(), create the nodes, Session.save(), and finally UserTransaction.commit(). (Typical approach when there is too much work to do in a single CMT transaction.)
Problem: After running the test, and with AS7 still running, only the node1 branch (and all descendents) can be seen (e.g. by browsing using the REST client).
However, to my surprise, after restarting AS7, the rest of the branches (node2, ..., node10) can then be seen. Some kind of caching/flushing problem? I don't think I'm doing anything wrong. I've done this same kind of thing with JPA many times......
I've attached the Eclipse project for the EJB, the ejb-jar, and the config files.
I'm going to try the same thing w/o transactions, and then in a single CMT transaction, to see if that makes any difference.....
If you'd like me to file a JIRA, please let me know.