So we ran into a production issue earlier this week. The solution to the problem was to change a previously transactional database pool to be nontransactional (check the box next to Returns NXA connections).
This makes no sense though.
The error would come when a Singleton EJB would do a lot of work - essentially a batch process that would go through and touch 30k database records, do some arithmetic, and then do the update. This was done using JPA for persistence, inside of a Singleton. It looked like it would go through some of the records, but then suddenly say that it needs more NXA resources.
Any ideas? Why would this be happening?
Thanks!
[Message sent by forum member 'johnament']
http://forums.java.net/jive/thread.jspa?messageID=393895