Details
-
Bug
-
Status: Closed (View Workflow)
-
Critical
-
Resolution: Incomplete
-
10.6.19
-
Red Hat Enterprise Linux release 8.10 & mariadb Ver 15.1 Distrib 10.6.19-MariaDB, for Linux (x86_64) using readline 5.1
Description
While resolving a replication error related to a foreign key constraint, we observed that issuing a START SLAVE after dropping the foreign key unexpectedly reset the relay log sequence from relay-log.00004404 to relay-log.00000001. As a result, some transactions were not applied on the replica, and replication broke again after appearing temporarily healthy.
This behavior is unexpected because STOP SLAVE; START SLAVE; should not reset the relay log sequence under normal circumstances, unless a RESET SLAVE or CHANGE MASTER was explicitly issued — which we did not execute in this case.
This appears to be a destructive relay log reset triggered by START SLAVE, which should not occur and may indicate a bug in MariaDB 10.6.19.