Hi!
>>>>> "Steve" == Steve Ruby <[EMAIL PROTECTED]> writes:
Steve> Due to the fact that piping a table from mysqldump to an INNODB table
Steve> can be as much as 30X slower if the insert is not done with
Steve> transactions,
Steve> would it make sense to add some option to mysqldump to make every X
Steve> inserts a transaction.
Steve> for example
Steve> mysqldump --transactionrows=100000 mydatabase | mysql -hsomeserver
Steve> database
Steve> where the above would wrap every 100,000 inserts with begin; commit;
Steve> right now it is non-trivial to insert the transaction
Steve> lines every so often, or even around all of the data.
I don't think we actually need the above option.
Try using the --quick option to mysqldump.
This will MUCH faster for InnoDB, because it inserts up to 16M of data
at a time (instead of just single rows).
Could you test the above on your tables? If it doesn't give adequate
performance, then we have to consider adding begin-ends to mysqldump.
The problem with having very long transactions is that they could actually
make things go slower because you get much bigger rollback segments.
(I assume that if I am wrong, Heikki will correct me ...)
Regards,
Monty
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
