[sdiy] OT: how to save a dead Raided Drive?

[Andre Majorel]

On 2009-06-27 16:27 +0300, Samppa Tolvanen wrote:
> On Sat, Jun 27, 2009 at 4:04 PM, Andre Majorel<aym-htnys at teaser.fr> wrote:
> >
> >> it is very unlikely that both drives would mechanically fail at
> >> once right?
> >
> > Yes, though I've seen drives from the same batch fail a couple
> > days apart, while installed in different computers.
>
> How come MTBF has become so predictable these says?

Consider this model : the time of failure is a bell curve centred
around the MTBF so that, say, 50% of disks die between 0.8 MTBF
and 1.2 MTBF. When the MTBF is 5 years, the bulk of failures will
occur over a two-year span.

Now take a bad batch whose actual MTBF is 5 weeks. The bulk of
failures will occur over a two-week span. The possibility of two
disks dying the same day doesn't look so remote now, does it ?

This is of course a simplistic model but you get the idea.

And that is why setting up a RAID from N identical disks bought at
the same time is asking for it.

-- 
André Majorel <URL:http://www.teaser.fr/~amajorel/>