On Friday 08 October 2004 19.01, Gary Richardson wrote: > Hey guys, > > I'm running a master/slave setup with v4.0.20. There are a hand full > of databases being replicated. For backups, we stop replication on the > slave and pull a text dump using mysqldump. I also record a 'SHOW > SLAVE STATUS' from the time of the backup. > > My replica server crashed last night. It looks like it had something > to do with the disk cache as the replica was trying to replay already > committed transactions (lots of duplicate record errors).
> With MySQL's two phase replication, will the IO thread automatically > figure out what file to start downloading and where to resume? Nothing is automagic but if you did 1. stop slave; 2. show slave status; and recorded the info before you did your backup you would restore the backup data and start replication after jumping to the correct position on the slave with the following: CHANGE MASTER TO MASTER_LOG_FILE='george-bin.5528', -- Relay_Master_Log_File MASTER_LOG_POS=290303997; -- Exec_master_log_pos this command will purge any relay logs and reset the slave thread to the requested position. It can be a good idea to always have skip-slave-start set in your .cnf file on the slaves. consider just copying/tar'ing the mysql datadir - raw file backup and restore are more efficient than mysqldumps. -- ___ |K | Ongame e-Solutions AB | /\| Mikael Fridh / Technical Operations |_\/| tel: +46 18 606 538 / fax: +46 18 694 411 -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
