← Back to team overview

maria-developers team mailing list archive

MWL#47: On the question of annotations persisting from one binary log to another


Hi Alex,

My opinion is as follows:

When log-slave-updates is not set, the slave is not binlogging the updates
it is applying, so the comments do not go to the binary log either.

When log-slave-updates is set, the action should depend on the
binlog-annotate-row-events setting, like it's done on the master. That is,
if binlog-annotate-row-events is on, annotate events should be passed to 
slave's binary log, otherwise they should be discarded.

There is one hole left: if one does 
 mysqlbinlog /path/to/binlog | mysql --host=slave-which-has-log-slave-updates 

then the annotations will not make it to slave's binary log. I think that
should be ok, but we better get a confirmation of this to be sure.

One should also take care that the slave do not write annotations for SQL
statements in form of

   BINLOG 'xaqw324435324hnjsfp'

as they are not useful (or at least don't copy the entire 'xaqe23....' part).

Sergey Petrunia, Software Developer
Monty Program AB, http://askmonty.org
Blog: http://s.petrunia.net/blog