on multithreaded slaves when gtid_mode is off the slave status could be inconsistence

[liuguoshun]

Please refer doc about "gap-free low-watermark position".
https://dev.mysql.com/doc/refman/5.7/en/replication-features-transaction-inconsistencies.html

on multithreaded slaves when gtid_mode is off ,need print an error info or abort the backup.

[davidducos]

Can you explain a bit more about this scenario? as from what I understood, you are trying to take a backup from a multithreaded database and the position saved in metadata file might be no correct. Am i right?

[liuguoshun]

Yes, mydumper use "FLUSH TABLES WITH READ LOCK" and "SHOW SLAVE STATUS" to get the Exec_master_log_pos.
Pos of "SLAVE STATUS" maybe not correct in metadata file, when backup multithreaded replica database.

[davidducos]

Ok, we are going to show a warning so users can disregard that info and start the replication from the replica where they are taking the backup instead of the source of the replication.

[davidducos]

@liuguoshun please, let me know if #417 is ok, so we can merge to master.

Thank you!!

[liuguoshun]

Great! :-)