Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 16 Dec 2005 14:24:21 +0800
From:      huang leo <leo.huang.gd@gmail.com>
To:        Greg 'groggy' Lehey <grog@freebsd.org>
Cc:        freebsd-performance@freebsd.org
Subject:   Re: Benchmark MySQL Performance On FreeBSD And Linux
Message-ID:  <a0cd7c070512152224m6304f398q@mail.gmail.com>
In-Reply-To: <20051216040722.GA503@wantadilla.lemis.com>
References:  <a0cd7c070512151920s67ca6cd6h@mail.gmail.com> <20051216040722.GA503@wantadilla.lemis.com>

next in thread | previous in thread | raw e-mail | index | archive | help
hi,
I experience the crashes myself. I used super-smack remotely to connect
MySQL with 200 clients first. After the test was done and the connections
were closed, we connected MySQL with 300 clients immediately. Then the MySQ=
L
restarted. The error log of MySQL is:
mysqld in free(): error: modified (chunk-) pointer
051130 10:12:04  mysqld restarted

I think that it's time out to cause the deadlocks, because our MySQL's
innodb_lock_wait_timeout is 50.

As you can see from our paper, we used super smack to test. The test script=
s
were given in the paper.

Best regards,
Leo Huang

2005/12/16, Greg 'groggy' Lehey <grog@freebsd.org>:
On Friday, 16 December 2005 at 11:20:12 +0800, huang leo wrote:
>
> We had evaluated MySQL performance on FreeBSD and Linux. The result is
> attached.

Thank you!

This is some of the most plausible information I've seen in a while.
I'm forwarding it to a MySQL internal list, and I expect we'll discuss
it in much detail.

A few comments:

- You refer to bug http://bugs.mysql.com/bug.php?12251.  Did you
experience these crashes yourself?  If you look through that bug
report, you'll find that we (MySQL) have had great difficulty
reproducing the problem.  Many people have reported that they have
no trouble under the same conditions, and we were half thinking it
might be related to (but not the fault of) the hardware in
question.  Your hardware is very different, so it would help us a
lot if you're also able to reproduce it.

- There's been a lot of discussion on the FreeBSD lists recently
suggesting that the performance problems are related to the choice
of timer.  I've never been convinced, and your results suggest to me
that this isn't the real problem.  Still, it would be nice to
compare the results using various timer options.

If you haven't been following the discussion, it's in the archives
for the FreeBSD-current mailing list.  Let me know (in private mail)
if you'd like me to forward them to you.

- I'd be very interested in the deadlocks you've been experiencing.
If you have any additional information, please let me know.

Do you have scripts for these tests?  I appreciate that the
information is mainly there, but it would be easier for us if the
scripts are available.

Thanks again for the report
Greg
--
See complete headers for address and phone numbers



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?a0cd7c070512152224m6304f398q>