I have run across something that has me stumped. I have some systems that
have very large error logs because we haven't moved from statement-based to
mixed-based replication yet so they get a lot of warnings logged. I need to
rotate the error logs and have started looking at it doing so.
The problem is that on one system a normal course of action works perfectly,
but on anther it does not. And these systems were installed from the same
RPM packages (5.1.50 -- downloaded from the MySQL website).
Here is what I do:
log in with mysql client and 'flush logs' OR mysqladmin --flush-log
It should rename the old log file to mysqld.log-old and start a new
On one system it works perfectly
On the other...nothing.
I tried moving the error log (mv /var/log/mysqld/mysqld.log
/var/log/mysqld.log.old) and then issuing the flush logs command...it stays
writing to the "old" file and never makes a new one.
If I were to restart mysqld it would solve the problem but this is a
production system and that isn't very practical.
These systems are very similar. my.cnfs have been checked for differences. I
searched the interwebs and specifically bugs.mysql.com for something
similar. Not finding anything.
I would appreciate any ideas!