I apologize for the very slow reply -- I started writing back to you and
then got interrupted by some travel.
> I've got a new Seagate Barracuda 7200.7 40Gb hard drive, a ST340014A.
> Its making a soft but annoying intermittent buzz, which I think might
> be related to Seagate's implemenation of SMART.
> I've found that any hard-drive activity (touch foo; sync) will stop the
> buzz for exactly 40 seconds (unless there is more disk activity in the
> meantime). Then it starts again.
Interesting.... that buzz may be related to self-testing or offline
testing, which then gets interrupted by drive activity.
> The only thing I've found that can make the buzz start on cue is to run
> a self test:
> smartctl -t long /dev/hda # -t long starts buzzing; -t offline doesn't
> In this case, disk activity doesn't stop the buzzing, but stopping the
> test does:
> touch foo; sync; # keeps on buzzing
> smartctl -X /dev/hda # stops the buzzing
> So, obviously, my hypothesis is that my drive is running an offline test
> on itself whenever the drive is idle for 40 seconds.
> So I try this:
> smartctl -o off -s off /dev/hda
> Unfortunately, it doens't work. The buzz persists.
Indeed, this would be the right thing to shut off the buzz, unless there
is a firmware bug that prevents the read scanning from stopping when smart
is disabled. For fun, you might try resetting the drive (eg, use hdparm
-w). Note that this *may* be dangerous, in the sense that the drive/ide
controller chipset may not cleanly reset. But I've done it lots it lots
of times without trouble
> Then the -a option reports that "Auto Offline Data Collection:
> Disabled". That's what I asked it to do. Now I use -s off; I'm not
> exactly sure what that will do, but it seems worth a try. Buzzing still
I think you must have meant to say that you try:
smartctl -o off -S off /dev/hda
and then when that fails to do anything, you try -s off and that still
fails to stop the buzzing.
> Now, I use hdparm -I to query the disk using a different utility. I see
> the following:
> Enabled Supported:
> SMART feature set
> * SMART self-test
> * SMART error logging
> This looks to me like "-s off" worked, but that maybe "-o off" didn't
> work, since hdparm is reporting that SMART self-test is still enabled.
No, in spite of what hdparm is printing, once you have disabled the SMART
feature set, smart self-test and error logging are disabled. This is just
printing a particular word of the DEVICE IDENTIFY device structure, that
happens to be badly defined in the ATA spec.
> Meanwhile, my buzzing has started up again. "touch foo; sync" stops it
> for 40 seconds.
> I haven't been able to discern a pattern for how long the buzzing lasts,
> but since this is the only drive on my system I can't unmount it and
> observe it in the total absence of disc activity. As far as I know, the
> buzzing keeps going until something accesses the drive.
Um, you can do these experiments. Use a bootable CD (knoppix, cool linux)
which will give you a complete operating OS, without using the disk at
all. Mount the disk and copy smartctl to /tmp (which is a ramdisk).
Then unmount the disk and run smartctl on the unmounted disk.
> Has anyone seen anything like this? Anyone familiar with the SMART
> capabilities of these new Seagate drives? Does the 40 second interval
> sound familiar?
I don't have any direct experience, but perhaps someone else will answer
> One last bit of info, relvance unknown: I've got my BIOS configured to
> spin down the disk after 3 minutes of inactivity. It hasn't been
> spinning down at all on that setting. When I had it set for 2 minutes,
> it spun down quite frequently (which would stop the buzzing, of course).
> Thanks for any help! I'm not a list subscriber, so if you're feeling
> charitable, please copy your response to <david at davidflanagan dot com>
Is this a laptop disk? If not, I'd advise against all the spindown/spinup
cycles. It's very hard on the disk's motor and bearings.