[PLUG] Using a hard disk with single bad sector.

Steve Bonds 1s7k8uhcd001 at sneakemail.com
Mon Sep 8 15:34:25 UTC 2003


On 8 Sep 2003, Michael C. Robinson michael-at-goose.robinson-west.com |PDX Linux| wrote:

> What is the most common cause of hard drive failure?

Try a quick read through this PDF from the fine folks at Seagate:

http://www.seagate.com/content/docs/pdf/whitepaper/D2c_More_than_Interface_ATA_vs_SCSI_042003.pdf

My take on the most common causes of hard drive failure are:
  + heat
  + failed spindle motor bearings
  + power surge/lag damage
  + vibration (often temporary "failure")

> When it failed it just disappeared.

Sounds like the electronics/controller died if you couldn't even get the
drive to identify itself.

One note on MTBFs-- in order to compare them you'll need to know the
assumptions made by each manufacturer with regard to the test
environment.  Were the drives on continually?  Were they only on 8
hours/day?  Was there a single power on per day or multiple?  What was the
ambient temperature kept at?  Etc. etc.

> I opened it up to look at it but there is nothing obvious inside.  

My guess is that the insides were probably completely functional when you
opened it.  Neat stuff in there, eh?

> What do you have to do to manually erase a hard disk beyond the point
> where anyone can recover data off of it short of sending it to a
> trusted recycling shop?

Writing once to each sector on the filesystem is enough to deter anyone
without access to a data recovery shop.  This is obviously hard to do if
the drive is inaccessible.

After that you'll simply need to weigh your paranoia vs. the time and
effort you want to spend.  For the truly paranoid, converting the entire
drive to plasma and dispersing it slowly into the atmosphere at high speed
and high altitude is probably sufficient.

> I've heard that hard disks have aluminum platters with something
> sprayed on them that's magnetic.

Glass was a more common substrate as of 2001.  It's smoother than
aluminum.  Manufacturers may well be on to something new now.

You're right about the thin layer of depositted magnetic material though.

> As far as bad sectors, are they mostly caused by spikes in power to
> the drive head or does the magnetic material deform beyond
> manipulation by the head after a while?

Hard drives use complex algorithms that extract a tiny signal from the
background noise.  The denser you pack the bits, the closer the signal
gets to the noise.  (And there is a strong financial motivation to pack
the bits as close as possible.)

If the magnetic properties of a microscopic magnetic domain change
slightly over time due to heating/cooling, oxidation, etc. the noise may
eventually engulf the signal.

Severe shock to the drive can also result in head/platter contact which
vaporizes some of the material from the surface and exposes nearby areas
to extremely high heat.  (Think of what your Dremel at 15,000 RPMs does to
what it touches.)  This alters the magnetic properties of a large area of
the surface and can easily create sudden nonrecoverable failures.

In modern drives, the slow creep of heat + oxidation happens so often that
when the algorithm can't cope with the change any more, and the logic in
the sector headers can't cope with the change any more that sector gets
remapped transparently to the host controller.  Note the key is that this
happens slowly-- the drive itself is aware that it can't read this area as
well as it used to before the area becomes completely unreadable.  (This
is where SMART comes in-- providing this information back to the host
controller.)

The magnetic material itself does not "wear out" from being magnetized.

> Is exposing the platters to the air enough to permanently prevent
> anyone from getting remnants of the data that was on it?

Anyone?  Nope.  Most folks?  Yep.

I find exposing the platters to a 3 pound sledge helps prevent even more
folks from getting remnants of the original data.  Microwaving is also
fun, though I usually reserve that fate for my CD-R media.

  -- Steve





More information about the PLUG mailing list