On 18 Apr 2008, at 10:27, Ed W wrote:
>> I´m going to set up a innodb database replication, and I have some
>> doubts
>> about the backing up the master database step.
>>
>> Normally, I do a "FLUSH TABLES WITH READ LOCK", backup the data
>> files (or
>> use mysqldump), get the current log coordinates with "SHOW MASTER
>> STATUS"
>> and unlock the tables. Using exactly the same steps work for innodb
>> databases?
>>
>> According to some texts, it seems not, but I haven´t found a clear
>> cookbook.
>>
>>
>> Just doing a "mysqldump --master-data --single-transaction" is
>> enough to
>> replace the above steps?
>>
>>
>
> How would you put these two things together in the case of a server
> using mixed database table types for a clean backup that can be used
> to restart synchronisation?
You can't. Only InnoDB supports single-transaction, so though it would
still generate a dump file for other table types, you'd have no
assurance that they are clean because writes could be happening as you
do it. This is one of the advantages of using innodb.
You can of course do separate dumps for other DBs, and because that
wouldn't need to include the innodb tables, it would at least reduce
the time that your tables are locked.
You could also look at maatkit's parallel dump/load scripts.
Marcus
--
Marcus Bointon
Synchromedia Limited: Creators of http://www.smartmessages.net/
UK resellers of info@hand CRM solutions
marcus@stripped | http://www.synchromedia.co.uk/
