Cheers,
>
> Mark
>
> -Original Message-
> From: Dan Buettner [mailto:[EMAIL PROTECTED]
> Sent: Monday, July 10, 2006 3:21 PM
> To: Mark Steele
> Cc: mysql@lists.mysql.com
> Subject: Re: mysqldump with single-transaction with high-concurrency
DB
>
> Sorry Mark
5.0.22
-Original Message-
From: Dan Buettner [mailto:[EMAIL PROTECTED]
Sent: Tuesday, July 11, 2006 9:55 AM
To: Mark Steele
Cc: mysql@lists.mysql.com
Subject: Re: mysqldump with single-transaction with high-concurrency DB
Sorry Mark, appears you're right. --opt is the defaul
Monday, July 10, 2006 3:21 PM
To: Mark Steele
Cc: mysql@lists.mysql.com
Subject: Re: mysqldump with single-transaction with high-concurrency DB
Sorry Mark - I thought your question was more of a "does this seem
right" and "how do I" than a "something's wrong here&quo
ver, however MySQL
replication can lead to slave drift as it is statement based (as
opposed
to row-based replication). The only safe way to guarantee a real
backup
in a MySQL replication setup is via snapshots on the master.
-Original Message-----
From: Dan Buettner [mailto:[EM
napshot?
Cheers,
Mark
-Original Message-
From: Dan Buettner [mailto:[EMAIL PROTECTED]
Sent: Monday, July 10, 2006 3:21 PM
To: Mark Steele
Cc: mysql@lists.mysql.com
Subject: Re: mysqldump with single-transaction with high-concurrency DB
Sorry Mark - I thought your question was more of
ysql.com
Subject: Re: mysqldump with single-transaction with high-concurrency DB
Sorry Mark - I thought your question was more of a "does this seem
right" and "how do I" than a "something's wrong here" post.
I think your problem is coming in with the use of --opt
kup
in a MySQL replication setup is via snapshots on the master.
-Original Message-
From: Dan Buettner [mailto:[EMAIL PROTECTED]
Sent: Monday, July 10, 2006 2:42 PM
To: Mark Steele
Cc: mysql@lists.mysql.com
Subject: Re: mysqldump with single-transaction with high-concurrency DB
Mark,
Mark, that's the expected behavior of mysqldump with --opt and
--single-transaction; it locks all databases and all tables for the
duration of the dump, ensuring a consistent snapshot.
With a database this size (100 GB), it's an area where throwing
hardware at the problem may be your best bet. I