[MDEV-31328] GTID replication error Created: 2023-05-22 Updated: 2023-05-25 |
|
| Status: | Open |
| Project: | MariaDB Server |
| Component/s: | Replication |
| Affects Version/s: | 10.6.10 |
| Fix Version/s: | 10.6 |
| Type: | Bug | Priority: | Major |
| Reporter: | Jeremy Knapp | Assignee: | Unassigned |
| Resolution: | Unresolved | Votes: | 1 |
| Labels: | None | ||
| Environment: |
RDS MariaDB 10.6.10 Primary to Aurora V2 8.0.mysql_aurora.3.03.0 Secondary |
||
| Description |
|
When running binlog replication from the primary to the secondary we see replication work for a while. We then get an error on the secondary.
The binlog in question looks like the following at the start.
It looks like a similar issue was listed here However it was fixed ages ago. |
| Comments |
| Comment by Daniel Black [ 2023-05-23 ] |
|
Are you able to extract the full binary log event and upload via sftp? Given its an Aurora secondary generating this error perhaps its their bug. At least they might be able to point to where in the MariaDB binary log message the incompatibility is. I'm assuming by the binary log filenames/positions this has been running successfully for quite a while. Has the MariaDB version or its replication settings, or the Aurora version/settings changed recently that triggered this? I'm slightly surprised it works at all. |
| Comment by Jeremy Knapp [ 2023-05-23 ] |
|
For sure I can upload it. I have ben looking at the MariaDB due to the from Master part. I can try to dig into the other side more. We probably can upgrade MariaDB versions on our primary. I have been searching for some sort of matrix that shows working. Or does binlog replication from MariaDB to Mysql just not work anymore? Actually I rolled the target log off. I can reproduce this pretty easily though. |
| Comment by Daniel Black [ 2023-05-25 ] |
|
Testing can occur on the MySQL side as at least there's source there and assume Aurora follows their implementation close enough. It takes two to ensure compatibility. I'm not sure MariaDB or MySQL are going to publicly document compatibility we can't control. I'm curious as to how well the compatibility has occurred with little testing effort around it. SQL steps on reproducibility welcome. |