Re: Point in time recovery with MySQL Cluster

[Matthew Montgomery]

Hello Mitchell,

The process is as follows:

1) backup database using START BACKUP
2) at some later point, before the restore make a backup of mysql.ndb_binlog_index table and binlogs. (perhaps hourly)
3) (failure/error occurs)
4) clear the ndbd nodes (--initial)
5) remove or truncate mysql.ndb_binlog_index
6) execute ndb_restore of all data (this populates the ndb_apply_status table)
7) restore the ndb_binlog_index table from mysql dump and binlog files from backup if necessary
8) Find current applied epoch : SELECT @LASTEPOCH:=MAX(epoch) FROM mysql.ndb_apply_status;
9) Find corresponding binlog position : SELECT Position, @FIRSTFILE:=File FROM mysql.ndb_binlog_index WHERE epoch > @LASTEPOCH ORDER BY epoch ASC LIMIT 1;
10) replay the binlog events using mysqlbinlog from the given file and position up to the point of the failure.