I have two mysql servers setup with default my.cnf config apart from server-id's and enabling binlogs on masters. Replication is setup and working and i have a simple bash scrip on master feeding in random data.
I am testing a script which stops and starts the replication periodically on the slave. The script records the Exec_Master_Log_Pos value and master binlog which will be used with the next CHANGE MASTER TO..
The problem i have is, if i poll the SHOW SLAVE STATUS every second i have noticed that during binlog rotation on the master i see the Read_Master_Log_Pos jumping very high, far to high to be actual written data to master which in turn seems to cause a knock on affect to the Exec_Master_Log_Pos so that is also wrong. So when i log this value and use it in the next CHANGE MASTER i get errors on slave about impossible position.
Interestingly at the same time of polling SHOW SLAVE i also log the output of SHOW MASTER STATUS and i don't see the same wrong binlog or position reported so it seems to be a slave thing only.
So my question is, does anyone know of recent bugs in mysql that could cause this? Is there a my.cnf setting i can magically enable to prevent this from happening or has anyone seem similar behavior as this?
This is running on Debian Wheezy 7.0 with MySQL 5.5.31.