MySQL Bugs: #34045: Intermittent InnoDB crashes on shutdown

Bug #34045	Intermittent InnoDB crashes on shutdown - Page 28 3 still fixed or dirty
Submitted:	24 Jan 2008 22:07	Modified:	7 May 2008 8:58
Reporter:	Mark Callaghan	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Server: InnoDB storage engine	Severity:	S3 (Non-critical)
Version:	5.0.37	OS:	Any
Assigned to:	Assigned Account	CPU Architecture:	Any
Tags:	crash, dirty, fixed, innodb, shutdown

Description:
This occurs on servers that use 4 or 8 core servers.
The error message is: 
Page 28 3 still fixed or dirty
080124 13:48:42InnoDB: Assertion failure in thread 1197361504 in file buf0buf.c line 2453

How to repeat:
Run many servers with InnoDB. Shut all of them down. I shutdown 60 today to upgrade an RPM. 3 of them had this problem. Normally, the frequency is much lower.

This may be the same bug as http://bugs.mysql.com/bug.php?id=26081

Inaam, please put more diagnostic code to 5.0 so that we get detailed information what is the buf fix count and if the buffer page is dirty.

Mark, I assume these are AMD computers?

Regards,

Heikki

Mark,

Trying to tie it with http://bugs.mysql.com/bug.php?id=26081 do you use READ-COMMITTED isolation level?

regards,
inaam

I think most use repeatable read and repeatable read is the default. For now, if you give me code to improve debug output for this, I can get that into production.

We use AMD and Intel, but all of the problems have been on AMD with 4+ cores.

Mark,

I have a debug patch for this. It is based on some debug code written by Sunny a while ago to track another bug. I realize that since you are using 5.0.37 you may not have Sunny's code in your source. I am therefore going to attach a diff of storage/innobase/buf0buf.c of 5.0.38 (there seems to be no tag for 5.0.37 in bk) and the current buf0buf.c. You can apply that to your source.
Also, if possible, can you run your server under --core-file?

regards,
inaam

diff for debug patch

Attachment: block_debug.patch (text/x-diff), 5.31 KiB.

Need feedback from the customer.

Maybe we could also contact AMD?

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

Mark,

Heikki und Inaam are waiting for feeback from you.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".