Yes thats very true, as the corrupt slave would pick up from where it left and its quite posssible that the record was already updated on the healthy slave from the master, causing the replication to fail, but having said that may be you could do this 1) stop the healthy slave 2) put a global read lock on master or stop it for a few mins (if you can) 3) clear the *.info and the *-bin.* files on corrupt slave 4) run rsync between the corrupt slave and master so now they are in sync.... 5) now start the master, and both the slaves and you are good to go
---All this could be easily scripted to make the task easier. ---Also my soultion may not be any better than what you are doing currently, except for the time saved in the rsync process instead of copying the whole data folder to the slave ...... Hope this helps , just my 2 cents Kishore Jalleda On 3/7/06, Michael Jeung <[EMAIL PROTECTED]> wrote: > > Hi Kishore, > > That's an interesting idea. However, given that the healthy slave > and the corrupt slave now have different values for > Exec_Master_Log_Pos, would restoring the tables from the healthy > slave necessarily be a good move? > > I would be worried that the corrupt slave's counter position would be > mismatched with the table. > > Regards, > Michael Jeung > > > On Mar 7, 2006, at 3:20 PM, Kishore Jalleda wrote: > > > an "rsync" with the right options from the unaffected slave to the > > corrupt > > one might prove to be an elegant technique > > > > Kishore Jalleda > > > > On 3/7/06, Michael Jeung <[EMAIL PROTECTED]> wrote: > >> > >> Hi Folks, > >> > >> We've got a Single Master/Multiple Slave environment. > >> Recently, we had some corruption on one of the slaves and I had to > >> repair the affected tables. > >> > >> After the repair completed, some of the rows on the slave had been > >> deleted - so the Master and the Slave weren't exactly in synch. The > >> slave was missing some records. I verified this myself with a > >> select count(*) on the table. > >> > >> Now, I know this is the normal way for MySQL to repair it's database. > >> > >> However, what I'd like to do is restore these missing rows to the > >> Slave without doing a dump and reload from my backups. Is this > >> possible? > >> > >> The way that we normally approach this problem is to take an > >> unaffected slave and copy the mysql directory from an unaffected > >> slave over to the corrupt slave. Then we restart MySQL on the > >> corrupt slave and things work out without too much difficulty. > >> However, this has always seemed like a terrible way to restore a > >> corrupt slave to me. Is there a more elegant way to do it? > >> > >> Regards, > >> Michael Jeung > >> > >> -- > >> MySQL General Mailing List > >> For list archives: http://lists.mysql.com/mysql > >> To unsubscribe: http://lists.mysql.com/mysql? > >> [EMAIL PROTECTED] > >> > >> > >
