Scenario:

1. enlist jms XA resource
2. enlist test xa resource
3. prepare jms XA resource
4. prepare test xa resource
5. commit jms XA resource
6. crash the connection and the TM does not know that jms XA resource was committed correctly
7. by spec XAER_RMFAIL is expected to be returned from RA (jdbc driver)
   • AMQ returns XA_RETRY which should lead to the same behavior as XAER_RMFAIL (even by spec it should be so)
8. commit test xa resource
9. recovery when log about db/jms XA resource is still "active"
   • recovery tries to commit but error could be returned as no such XID is in DB - this issue is fixed by JBTM-860

XA recovery on onMessage of MDB (two queues in use - one inbound to receive in MDB and one outbound to send out of bean). This scenario runs separate EAP container where messaging server is started. The MDB of the first container sends data there (there is no inVM message sending).

This issue is valid with JTS only (not JTA)

This issue is a regression, so the priority of this jira is a blocker.

Although I see this issues always locally on my notebook, this issue seems to be intermittent in lab, but it fails in ~75% of cases.

This issue has been caused by 9c962e9f99c3849970edaf5ea02da90d4dee6520 commit.