NOTE FOR PEOPLE HAVING THIS ISSUE:
You can 'workaround' this issue by restarting the database, which causes the locks by the XA transaction that is already prepared to be gone and replication will just continue.
(no need to restore slaves from backups, skip binlog events...)
NOTE FOR PEOPLE HAVING THIS ISSUE:
You can 'workaround' this issue by restarting the database, which causes the locks by the XA transaction that is already prepared to be gone and replication will just continue.
(no need to restore slaves from backups, skip binlog events...)