On 2/28/07, Volker Kuhlmann <[EMAIL PROTECTED]> wrote:
On Wed 28 Feb 2007 12:53:19 NZDT +1300, Kent Fredric wrote:
Btw smartmontools will tell you what the disk thinks the damage is.
I find smartmontools uncircumstantial at best, most of the time.The
best use I've had for it was on my box where I recently encouterd a
highly unusual problem with my Hitachi which has been resulting in the
disk becoming IO-Locked due to a disk timeout, which could _only_ be
fixed by a hard-power-off :/.
The problem did not occur to any of the "boot-off-cd" hard-drive
testing utilities, they all did a full disk read without a problem,
but for some particular reason, reading an actual file off that disk
resulted in the total denail of disk access. I managed to remedy this
tho, by an unusual technique. I extracted the exact block addresses
from the ext3 debug utility, and wrote a ruby script to extract them
directly from the disk using the in-disk-order, which was much
different to the actual file order, as for some reason, the locks
-this- technique resulted in were recoverable, ie: no reboot needed.
And after doing that to the file twice ( and seeing the
"block-relocated-count" skyrocket ), the file started working
normally! ( I don't know a lot about internals of hard drives, but
I've thus concluded about 30% of the disk is reserved for "oops"
copies! .... )
Oh, and running smartmon tools on a disk that doesn't spin up or
communicate with the IDE bus ... kinda hard. ;)
I've had 2 Samsungs die within 8 months (the one I bought, and the
replacement), although this particular model was reported by people with
larger sample sizes as being very reliable. I bought a Samsung again
because I like them (quiet, not hot-running).
I've been recently told Samsung make good drives, I do recall they
used to line the drive insides with rubber to get rid of the noise.
I've become a Hitachi advocate tho ;).
> me. Toms Hardware recently did a review on a bunch of -new- psu's and
> found their long-term stress tollerance below tollerable on many
> models,
Entirely plausible!
And those tests were on _HIGH_end PSUs, they didn't test any of the
multitudes of brandless abominations that often come with the case (
ie: I've seen case & PSU combos for sale for less than a price of a
case -without- a psu,... that -cant- be healthy )
--
Kent
ruby -e '[1, 2, 4, 7, 0, 9, 5, 8, 3, 10, 11, 6, 12, 13].each{|x|
print "enNOSPicAMreil [EMAIL PROTECTED]"[(2*x)..(2*x+1)]}'
