In a MySQL master-slave replication architecture, how do you detect if there have been (erroneously, since this should never happen!) direct writes on the slave?
For now, I'm running this command on the slave:
tcpdump tcp port 3306 and not host my_master_IP -A | grep -i -e INSERT -e UPDATE -e DELETE
whose output, if any, means writes on the slave, which is bad.
Is there a better method?