Re: random updates crashes nodes - jcwang - com.mysql.lists.cluster

7 messages in com.mysql.lists.clusterRe: random updates crashes nodes

From	Sent On	Attachments
jcwang	22 Jan 2008 18:55
jcwang	25 Jan 2008 09:19
jcwang	25 Jan 2008 15:14
Stewart Smith	28 Jan 2008 21:51
Stewart Smith	28 Jan 2008 21:52
jcwang	29 Jan 2008 17:23
Norbert Tretkowski	29 Jan 2008 23:20

Subject:	Re: random updates crashes nodes
From:	jcwang (jeff...@apple.com)
Date:	01/29/2008 05:23:39 PM
List:	com.mysql.lists.cluster

Hi Stewart,

Thanks for the reply. Can you point me to instructions on how I can check out from your repo? I didn't see any instructions on dev.mysql.com. Even better, any time estimates on when pre-compiled binaries will be available for the next 5.1.23 minor release?

thanks, Jeff

Stewart Smith wrote:

On Fri, 2008-01-25 at 15:14 -0800, jcwang wrote:

Although setting TransactionDeadlockTimeout higher avoids the locking issues, I've still been getting crashed data nodes. I've been able to consistently reproduce this bug and have filed it:

http://bugs.mysql.com/bug.php?id=34072

The crash seems to occur when you have Auto commit off and are manually committing. It seems like something is not being rolled back correctly in one data node but is rolled back in the other data node. This leads to inconsistencies and node crashes.

There's also been a lot of fixes since 5.1.22 - try latest 5.1 tree, or even latest -telco tree - these should be a lot better.