Minutes from the Technical Board, 2008-07-15
soren at ubuntu.com
Wed Aug 20 10:20:23 UTC 2008
On Wed, Aug 20, 2008 at 05:05:25PM +0800, Onno Benschop wrote:
>>> If you recall the google research about hard drive failures you will
>>> have remembered that SMART is no indication of impending failure.
>> And if you recall the very same research, you will have remembered
>> that SMART *does* indicate of impending failure. The conclusion of
>> the study was that if SMART says that he drive is about to go bust,
>> it's very likely to be true. However, if SMART says everything is
>> fine, that's not necessarily true.
> And your last sentence is precisely why I raised my point in the first
> The abstract from the report reads:
> [..] we conclude that models based on SMART parameters alone are
> unlikely to be useful for predicting individual drive failures.
The key word is "alone". SMART *does* provide an indication of impending
failure. A 99.9% full filesystem is an indication of impending system
failure. However, just because your filesystems are all only 25% full,
that doesn't mean that the building housing your servers is not on fire,
which is another quite sure indication of impending system failure.
Hence, a model based on filesystem fullness alone is not likely to be
useful for predicting individual system failure.
> And that is my point, "individual drive failures", that is, those on a
> single machine, which is what we were discussing if I'm not mistaken.
Right. Although, the idea of reporting this data for all of the worlds
Ubuntu systems to give the owners of identical devices an early warning
if others people's are starting to fail is interesting, but an entirely
Soren Hansen |
Virtualisation specialist | Ubuntu Server Team
Canonical Ltd. | http://www.ubuntu.com/
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 315 bytes
Desc: Digital signature
More information about the Ubuntu-devel-discuss