Details
Description
In a cluster/replication topology, where galera cluster operates as a slave for MariaDB master, and async replication transactions are conflicting with transactions executed in the galera cluster, it may happen that the async replication slave thread cannot handle gracefully the situation where high priority galera replication thread aborts the async replication transaction.
There are several different symptoms, which may result of such BF abort situation, depending on the state and phase of the async replication slave transaction processing in the time of BF aborting. The most critical problem happens, when the conflict with galera replication is false positive, and and async replication transactions should rollback and replay.
Attachments
Issue Links
- relates to
-
MDEV-21473 Galera node that is also an async slave from other cluster may crash unexpectedly on rollback or commit
- Closed