6 messages in com.mysql.lists.bugsRe: Replication bugs ..
FromSent OnAttachments
Weslee Bilodeau11 Dec 2000 10:20 
Sasha Pachev11 Dec 2000 11:47 
Weslee Bilodeau11 Dec 2000 12:41 
Michael Widenius11 Dec 2000 14:25 
Sasha Pachev12 Dec 2000 06:37 
Michael Widenius14 Dec 2000 00:01 
Subject:Re: Replication bugs ..
From:Michael Widenius (mon@mysql.com)
Date:12/14/2000 12:01:48 AM
List:com.mysql.lists.bugs

Hi!

"Sasha" == Sasha Pachev <sas@mysql.com> writes:

I have now fixed the warning in MySQL 3.23.29 (by adding a T_VERY_SILENT flag to the repair process).

I think that we should in the future also copy the fist 2048 bytes of the index file from the master to the slave; This will avoid errors like this and also help us add more checks to ensure that the slave-table is really ok.

Sasha> Monty - I think we should have a special flag T_INDEX_REBUILD. LOAD
TABLE Sasha> FROM MASTER and RESTORE TABLE do not really repair a broken table - they
just Sasha> restore the missing index. So it would be good for the repair functions
to Sasha> know that this is what is happening, so certain things are to be
expected. It Sasha> may also be possible to introduce some optimizations as well as sanity
checks Sasha> to the repair process based on that knowledge.

Note that there is information in the header of the .MYI file that we MUST have to be able to do the above smoothly; For example, if the table is compressed, this is not recorded in the .frm file, but you must have the .MYI file to be able to repair the missing index.

It doesn't help to introduce a T_INDEX_REBUILD flag as we are not doing a rebuild; Without the .MYI header we must in many cases do a repair instead of a rebuild (this is always the case if we have deleted blocks in the index file)

Note that if we just copy the header on LOAD TABLE and RESTORE TABLE, you will get the behaveour you requested....

Is there any reason for not copying the header?

Regards, Monty