[BUG] Switch over with mariadb-repmgr generates an inconsistent state for GTID about mariadb-tools HOT 13 CLOSED

Hi Cédric,

Which repmgr version are you running? I have made some bug fixes and would like to know if this behavior was before or after those changes. Please let me know. Thanks (also retry with the latest version from source, if you can)

from mariadb-tools.

mariadb-repmgr --version
MariaDB Tools version 0.0.1

from mariadb-tools.

Don't believe the version string, it has not been updated in a while :) Did you download it from the releases or compile it from source?

from mariadb-tools.

Héhé, ok ;-)
I downloaded the binaries from the releases section.

from mariadb-tools.

Please try this one

https://github.com/tanji/mariadb-tools/releases/download/0.2.0/mariadb-repmgr

from mariadb-tools.

Thanks, same issue.

from mariadb-tools.

It seems that you are using different domain IDs on that topology:
180533626 on srv-01
180533630 on the others

So I think that's a matter of supporting different domains in the same replication topology.

from mariadb-tools.

Hum, this is not domain ID here but server ID.

from mariadb-tools.

You're correct, I was mistaken. Thanks for the information, I think I have a possible solution and will let you know.

from mariadb-tools.

Cédric,

is there a chance that you're using -gtidcheck=true?

from mariadb-tools.

Yes

from mariadb-tools.

OK. It does not work in its current form because it's checking slave current GTID pos vs. master binlog GTID. If nothing has been written to the master, master binlog GTID is not updated and reflects an older state. I will patch so both slave and master use current GTID pos.

from mariadb-tools.

Understood. Thanks

from mariadb-tools.