[mdlug] New hard drive

Paul set at pobox.com
Fri Jan 9 02:43:50 EST 2009


Drew <drew4096 at gmail.com>, on Fri Jan 09, 2009 [12:26:42 AM] said:
> 

[ ... long ugly formated smartctl dump ... ]

> It looks good as far as I could see. But I'd like expert comment on 
> it. Also I've been told that Google has
> found zero correlation between smartctl results and disk life. Should 
> I be doing something else to get
> the hard drive to tell the truth about its bad sector count?
> 

	Hi;

	First, be aware that *all* hardisks have bad sectors
(or at least the probable existance of them right from the begining).
In the old days, they came with the lists of the bad sectors
printed on the disk, and later when disks got bigger on sheets of
paper... So some poor bastard could enter them into a bad sector
list that the filesystem could avoid. There are also tools like
'badblocks' that linger on from this antique era that try to find
them.
	But that was when the OS could actually address sectors
via coordinate system of cylinder/head/sector. For over a decade
now, probably, disks are more like black boxes that completely
disassociate you from their inner realities. Sure, they can accept
CHS addressing, but what really goes on inside the disk has no
relation to this backwards fantasy. And 'modern' OS's and disks
dispense with this fantasy altogether, and are simply addressed as
a linear sequence of sectors. ie. sector 0 to MAXSECT.
	So
	a) your disk had bad sectors on it from the factory,
but they are mapped out.
	b) your disk will develope bad sectors over time, but the
disk will try to dectect them failing and map them out.
	c) Your smartctl log shows no *new* bad sectors so far.
	d) If you monitor your smartctl data, you can see if bad
sectors (or other problems) are developing, and being remapped,
and if they exceed the lifetime level deemed failure. (basicly,
you have 3 numbers, the current VALUE, the WORST value, and the
THRESHOLD-- if VALUE is < THRESHOLD, its considered a type of
failure)
	e) You disk may spontaneously fail with no warning at all
from the smartctl. Some people conclude that makes smartctl testing
and data worthless, which is like concluding that regular doctors
checkups are worthless, because the patient may get a good rating
and croak from a brain aneurysm the next day.

Paul
set at pobox.com

ps. Im sure if you care about the google hd survey, the limpest google-fu
will be sufficient to locate it.



More information about the mdlug mailing list