xfs
[Top] [All Lists]

Re: performance degraded 2.4.8 -> 2.4.9?

To: Dan Yocum <yocum@xxxxxxxx>
Subject: Re: performance degraded 2.4.8 -> 2.4.9?
From: Stephen Lord <lord@xxxxxxx>
Date: Thu, 15 Nov 2001 09:00:49 -0600
Cc: tommy@xxxxxxxxxxxxxx, xfs-list <linux-xfs@xxxxxxxxxxx>
References: <3BF1A6F2.273B49C4@fnal.gov> <20011115095941.DB13.NEWSLETTER@teatime.com.tw> <3BF3D370.B1A520A8@fnal.gov>
Sender: owner-linux-xfs@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.2.1) Gecko/20010901
Dan Yocum wrote:

Tommy Wu wrote:

Dan Yocum <yocum@xxxxxxxx> wrote:

What happened between 2.4.8 and 2.4.9 that the disk I/O performance would
*dramatically* drop?  With 2.4.8 I'm getting ~34MB/s writes and ~160MB/s
reads.  As of 2.4.9 this has dropped to ~10MB/s writes and 80-100MB/s
reads.  That's pretty bad.  The problem was not fixed with 2.4.10 and is
still the same as of 2.4.13pre6 - I haven't checked the latest in a couple
weeks, maybe it's better?

I've decided to drop back to 2.4.5 since 2.4.8 was regularly crashing the
machines, the "kernel: __alloc_pages: 3-order allocation failed." errors
seem to have gone away, and the performance is right up there.

  I've got the same problem in 2.4.8, 2.4.9, 2.4.10...  and plus a
  xfs_force_shutdown problem in 2.4.9 in last month....

  I change my kernel to 2.4.8mdk from mandrank (base on -ac)... last month.
  It work great for me now! Running on a P3 1G SMP system with 1G ram
  for Oracle 9i.



I don't suppose you could take the system offline long enough to run a
bonnie++ test on it, could you?

Dan


I have been fighting Irix code for the last few weeks to meet some deadlines, and
I am about to go on vacation until after thanksgiving. Hopefully I will get some time
to look at this after I come back.


Steve




<Prev in Thread] Current Thread [Next in Thread>