Comments on: An introduction to InnoDB error handling http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/ Stay curious! Mon, 13 Oct 2008 13:10:12 +0000 http://wordpress.org/?v=2.6.2 By: Xaprb http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/#comment-9802 Xaprb Wed, 30 May 2007 14:53:08 +0000 http://www.xaprb.com/blog/?p=219#comment-9802 Something not very well documented that affects this behavior: the innodb_rollback_on_timeout InnoDB startup variable. Something not very well documented that affects this behavior: the innodb_rollback_on_timeout InnoDB startup variable.

]]> By: Xaprb http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/#comment-2020 Xaprb Fri, 29 Sep 2006 13:11:02 +0000 http://www.xaprb.com/blog/?p=219#comment-2020 <p>I agree with you, Peter, different errors would have been a good idea. I thought about the system errors, but I couldn't think of when they would be really common -- like timeouts on a busy system. (Maybe that shouldn't be frequent either). My thinking was if the disk is full, lots of other things are probably going wrong, too.</p> I agree with you, Peter, different errors would have been a good idea. I thought about the system errors, but I couldn’t think of when they would be really common — like timeouts on a busy system. (Maybe that shouldn’t be frequent either). My thinking was if the disk is full, lots of other things are probably going wrong, too.

]]> By: Peter Zaitsev http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/#comment-2019 Peter Zaitsev Fri, 29 Sep 2006 09:01:04 +0000 http://www.xaprb.com/blog/?p=219#comment-2019 <p>Honestly I hate this change in 5.0.13 - not because it is good or bad but because it is very not transparent. </p> <p>I would really prefer MySQL to return different error codes if Full transaction was rolled back or only last statement. Especially keeping in mind other storage engines which may want to follow different rules. </p> <p>Also you could have mentioned system error cause for getting errors. It could be full tablespace and bunch of other errors.</p> Honestly I hate this change in 5.0.13 - not because it is good or bad but because it is very not transparent.

I would really prefer MySQL to return different error codes if Full transaction was rolled back or only last statement. Especially keeping in mind other storage engines which may want to follow different rules.

Also you could have mentioned system error cause for getting errors. It could be full tablespace and bunch of other errors.

]]> By: Xaprb http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/#comment-2008 Xaprb Thu, 28 Sep 2006 02:02:18 +0000 http://www.xaprb.com/blog/?p=219#comment-2008 <p>I meant multiple row queries. I've never seen part of a row inserted unless something crashed. Thanks for pointing out my vague-ness!</p> I meant multiple row queries. I’ve never seen part of a row inserted unless something crashed. Thanks for pointing out my vague-ness!

]]> By: Ronald Bradford http://www.xaprb.com/blog/2006/09/26/an-introduction-to-innodb-error-handling/#comment-2005 Ronald Bradford Wed, 27 Sep 2006 23:32:33 +0000 http://www.xaprb.com/blog/?p=219#comment-2005 <p>"it rolls back only the statement that times out , unless your version of MySQL is older than 5.0.13. In these cases, it rolls back the entire transaction. ". I didn't know that. That's a very important thing to know. Any application developed in 4 or early versions of 5 that use Innodb for ACID compliant transactions may fail ACID. I for example last year developed a CRM application that includes large financial batch processing and one of the reasons I could use MySQL was ACID compliance. (It runs MySQL 4.1)</p> <p>Regarding, "For example, if your INSERT against a MyISAM table causes a duplicate index violation when it’s halfway done, it stops, leaving half the data inserted." In this instance, you are refering to multi-row insert (either by multi value rows or INSERT SELECT) Correct? Or does this insert part of a row in a single row insert? (Ouch)</p> “it rolls back only the statement that times out , unless your version of MySQL is older than 5.0.13. In these cases, it rolls back the entire transaction. “. I didn’t know that. That’s a very important thing to know. Any application developed in 4 or early versions of 5 that use Innodb for ACID compliant transactions may fail ACID. I for example last year developed a CRM application that includes large financial batch processing and one of the reasons I could use MySQL was ACID compliance. (It runs MySQL 4.1)

Regarding, “For example, if your INSERT against a MyISAM table causes a duplicate index violation when it’s halfway done, it stops, leaving half the data inserted.” In this instance, you are refering to multi-row insert (either by multi value rows or INSERT SELECT) Correct? Or does this insert part of a row in a single row insert? (Ouch)

]]>