A lot of guides explain that you can use MySQL master-slave replication to scale your database performance. But nobody explains how to keep the database instances in synch. MySQL replications are very shaky and tend to break and mess up stuff very easily. From my standpoint the whole replication scheme looks like a smart hack into the MySQL system.
Till today I used some self written scripts to check our systems. Nobody told me that there was a perfect toolchain for this problem already existing: Maatkit. Maatkit offers a few commandline tools that help you to keep the slaves in synch with your master.
Install it using brew:
brew install maatkit
Check tables for differences :
mk-table-checksum h=HOSTMASTER,u=USERMASTER,p=PASSWORDMASTER h=HOSTSLAVE,u=USERSLAVE,p=PASSWORDSLAVE -d DATABASE | mk-checksum-filter
And after you found differences you can even use Maatkit to repair those tables:
mk-table-sync --execute --sync-to-master h=SLAVEIP,u=USER,p=PASSWORD,D=DATABASE,t=TABLE
Besides repairing damaged tables Maatkit helps you to check if the slave replication is still working. Mk-heartbeat updates a heartbeat table in the master database and checks if the slave database has been updated. I have written a small script that wraps the heartbeat check and sends a mail if the slave replication has stopped.
And yes, using couchdb would make this toolchain obsolete :smile: .