jta-crash-rec quick start [1] sometimes fails on not properly recovering XA transaction.
The issue happens once pod with EAP is scaled up again. There should be record Rover updated via JMS in H2 database.
By checking transaction logs there is issue that original EAP server, MDB commits subtransaction for H2 database but subtransaction for AMQ is in prepared state. XA transaction in such state should be committed by recovery manager.
However migration pod (with recovery manager) rolled back AMQ subtransaction. What is weird that commit for H2 database is not visible after scaling up the pod with EAP.
It's probable that all issues are caused by H2 database which is not reliable for this use case. See linked issue CLOUD-2795
[1] https://github.com/jboss-openshift/openshift-quickstarts/tree/master/jta-crash-rec-eap
- is related to
-
CLOUD-3252 [Quickstart] jta-crash-rec-eap emit `Unable to acquire JDBC Connection` error
- Closed
- relates to
-
CLOUD-2795 [EAP][XA] jta-crash-rec-eap quickstarts needs to use proper messaging and database datasource
- Closed