[conspire] (forw) Re: [PenLUG] Wear and tear on disk drives

Rick Moen rick at linuxmafia.com
Mon Nov 12 12:18:44 PST 2007

Because some people have made overly broad claims about Ubuntu (and I'm
looking at you, Bruce  ;->  ).

----- Forwarded message from Rick Moen <rick at linuxmafia.com> -----

Date: Mon, 12 Nov 2007 12:15:20 -0800
From: Rick Moen <rick at linuxmafia.com>
To: penlug-members at penlug.org
Subject: Re: [PenLUG] Wear and tear on disk drives

I wrote:
> Quoting Jeffrey Siegal (jbs at quiotix.com):
> > Here's a better way:  Figure out what's spinning up the disk  
> > constantly and fix it.
> I don't have only a shallow understanding of this issue, but my sense of
> it is that this is _not_ about disks merely spinning up.  
> It's about load/unload cycles, which are something different, and are an
> inherent consequence of some combinations of motherboards and hard
> drives doing dumb things when given "hdparm -B 1" ACPI directives.

As it turns out, the term "Load_Cycle_Count" (from smartctl output) 
refers to the process of parking the head assembly at an "off ramp"
position (accompanied by audible clicking).  If the firmware and hard
drives are badly designed (implement ACPI particularly badly), then _if_
the OS implements laptop-mode or equivalent and thus sets ACPI setting
"hdparm -B 1", the drive will park/lock the heads too readily.

(Be aware that smartctl output aims to be predictive, but may be
incorrect.  See, e.g., this thread:
http://osdir.com/ml/utilities.smartmontools/2004-07/msg00072.html )

It appears that Ubuntu 7.04 and 7.10 set "hdparm -B 1" on laptops even
if the user hasn't enabled "laptop mode", the only respect in which
they're different from any other distro.  Ordinarily, the machine's
owner must enable "laptop-mode" manually.

(However, it _does_ appear to be, also, the default behaviour in
MS-Windows Vista, which provides no tools for changing that default:
http://blog.lynxworks.eu/?p=36 )

It also is claimed that "SPINDOWN_TIME=12" in /etc/default/acpi-support   
(spindown time in units of 5 seconds) setting (in recent Ubuntu
releases) is set to an unusually short interval -- and the
believed-to-be-excessive wear on drives probably results from
interaction between the two settings.

Best overview I've found:  https://wiki.ubuntu.com/DanielHahler/Bug59695

Desktop systems may also be affected if the OS cannot determine whether
the machine is on AC or battery power by parsing ACPI data -- and, as I
mentioned, ACPI is such a horrific mess that it's not surprising that it
sometimes is parsed wrong.

PenLUG-Members mailing list
PenLUG-Members at penlug.org

----- End forwarded message -----

More information about the conspire mailing list