Yeah Daevid! I know very well the issue! first set the slave to READ ONLY
[mysqld] read-only then there is a configuration option to tell the server to skip some type of errors automatically slave-skip-errors= http://dev.mysql.com/doc/refman/5.1/en/replication-options-slave.html#option_mysqld_slave-skip-errors But, But, BUT! What I did is to remove the constraint on the table of the slave so that you can control better the thing. Because if you systematically skip the 'foreign key forcing' error, you will skip them with any table, if you remove just that constraint on that table you have the situation more under control. I think one of these two are enough, the cron is very not recomended! Ciao Claudio 2009/5/20 Daevid Vincent <dae...@daevid.com> > We have a master / slave setup and as you know, one bad query can ruin your > whole day. Or if you accidentally write to the slave when you meant to > write > to the master, or any number of other things that break the fragility of a > replication setup. > > The magic incantation to get them synched again seems to be to login to the > slave and do this (over and over again until the Slave_IO_Running and > Slave_SQL_Running both say "Yes"): > > mysql> stop slave; SET GLOBAL SQL_SLAVE_SKIP_COUNTER = 1; start slave; show > slave status\G > > Is there a way to automate this a little bit. Maybe some bash script that > uses "mysql -e" and parses for those two strings? > Is this dangerous to do? > Is there a setting to have the slave do this already? > > In every case I've ever seen, it's always some SQL that got out of whack > like this: > > Last_Error: Error 'Duplicate key name 'id_operator'' on query. Default > database: 'core'. Query: 'ALTER TABLE `user_has_notification` ADD INDEX > `id_operator` (`id_operator`)' >
