[ https://issues.redhat.com/browse/JBTM-3351?page=com.atlassian.jira.plugin... ] Ondrej Chaloupka commented on JBTM-3351: ---------------------------------------- After the discussion with [~mmusgrov] it's clear that the `:recover` call from WildFly jboss-cli.sh which invokes the `clearHeuristic` should make changes for the participant record marked with `_heuristic` It could be used the `AbstractRecord#forgetHeuristic`. The `XAResource#forgetHeuristic`does some more things as it calls `XAResource.forget(Xid)` and it removes the connection to the XAResource. But neither of two is an issue as the `XAResource` ignores the `forget(Xid)` call if the transaction is not heuristically finished (by XA spec) and the connection is no more needed as the recovery will re-create the connection to the resource if neeeded. If the XAResource transaction is not heuristically finished then the forget(Xid) call is ignored by RM. In such case the recovery may retry the attempt and it's ok. If the XAREsource was heuristically finished then there is no way to succesfully retry the transaction. For `:recover` works the XAResourceRecord needs to be reactivated and replayed by recovery manager. The operation of `forgetHeuristic` clears the problematic flag and let the next recovery cycle to retry the attempt. -- This message was sent by Atlassian Jira (v7.13.8#713008)