Details
-
Bug
-
Status: Open (View Workflow)
-
Major
-
Resolution: Unresolved
-
10.0.14
-
CentOS release 6.4 (Final)
kernel-2.6.32-358.el6.x86_64
Description
I had bidirectional replication setup on two instances of MariaDB (DB1 and DB2). DB1 was the only one used for reading or writing and DB2 was a hot spare.
On 2016-09-10 DB2 went out of sync and had to manually reloaded, after doing so the replication has only been one way - from DB1 to DB2.
On 2016-10-08 'start slave;' was run on DB1 and turned off a few minutes later. I would have presumed this to have no effect as the binary logs expire after 3 days and the replication was off for 2 months. But at that point the following happened:
1) Records between 2016-09-10 and 2016-11-08 disappeared on DB1 for these same tables.
2) Replication to DB2 failed with PRIMARY key already exists errors on these tables as the primary key began to be re-used
3) The DB1 binlogs of 2016-11-08 include the records which disappeared, these records were replicated to DB2
4) The DB2 binlogs which were still relayed to DB1 from 2016-09-10 were still present on DB1 from the 10th of September and do not involve the tables in question.
The tables in question are InnoDB, rowformat Compact and Collation latin1_swedish_ci.