Re: Read & Exec Master Log Position on Master-Master setup

[Devananda van der Veen]

Bubba,

That is correct. For master<->master replication, log-slave-updates is not necessary, and enabling it does create this "problem" that you've encountered. It's quite confusing, and I would like to see it fixed for this reason, even though nothing is really "broken" about it.

The IO thread has read all the bin log, and the SQL thread has not executed up to the very end (because each master ignores statements that it wrote to the binlog (unless --replicate-same-server-id is enabled)). So, the server isn't wrong in saying that the IO and SQL thread positions don't match, but it would be helpful if the SQL thread position advanced when statements were skipped.

Like Tim, I also have to use the Seconds_Behind_Master field to know if my servers are lagging, even though I would much prefer to compare the actual binlog positions (as read from SHOW MASTER STATUS and SHOW SLAVE STATUS).

For cascading or circular replication, log-slave-updates is definitely required.

Regards,

Devananda van der Veen
http://devananda-vdv.blogspot.com/

Open-source tools available at
http://mycat.sourceforge.net/