[ https://issues.jboss.org/browse/JBTM-2122?page=com.atlassian.jira.plugin.... ] Tom Jenkinson commented on JBTM-2122: ------------------------------------- Observation: when a transaction that is using JTS with JCA in-flow crashes after writing the prepare message, a subsequent XATerminator::commit() message is not sufficient to flush the xa_commit to enlisted XAR. This is different from JTA because we can't know if the TX is making slow progress or it's a genuine recovery scenario. For these types of application the follow is expected: XATerminator::prepare(xid1) CRASH xid1 = XATerminator::recover() XATerminator::commit(xid1) Then, in the JVM where the resource lives: periodicRecoveryScan() // Wait for orphan detection interval periodicRecoveryScan() The second periodicRecoveryScan will effect bottom-up recovery on the XAR and it will try to contact the potentially remote coordinator to ask the state of the transaction (which will be of state committing) -- This message was sent by Atlassian JIRA (v6.4.11#64026)