SMART

[Jeremy Chadwick]

On Thu, Nov 12, 2009 at 01:25:12PM +0100, Ivan Voras wrote:
> Jeremy Chadwick wrote:
> 
> >I can teach you how to decode/read SMART statistics correctly.
> >
> 
> Actually, it would be good if you taught more than him :)
> 
> I've always wondered how important are each of the dozen or so
> statistics and what indicates what...

I started a write-up but after writing about 300 lines realised that if
I continued the details would eventually be lost in the Sea of
Information Chaos that is a mailing list.  :-)  I've gone over how to
read SMART data 3 separate times in the past 2 months (at work, on a
public forum, and in private mail), so this would be the 4th...
  
I'll work on writing an actual HTML document to put up on my web site
and will respond with the URL once I finish it.

Sorry for the "yeah sure I can help you read this data" response
followed by what will probably be labelled as an excuse by some.
Admittedly reading the output is pretty simple, but "getting familiar"
with what the output looks like (on a per-vendor basis) takes exposure
to all sorts of drives, ditto with F/W bugs and so on.

In general though, don't let anyone tell you SMART is worthless.  The
"overall health assessment" status is generally worthless, but the
per-attribute data is of great use.  Don't let anyone tell you the
weighted/adjusted values (VALUE/WORST/THRESH) are useless either; in
some cases they're all you can safely rely on.  Don't damn SMART when
it's actually the manufacturers which need to be spanked for setting
such unreasonable health failure thresholds.

-- 
| Jeremy Chadwick                                   jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |