[gentoo-user] Checking an HD for problems

[gentoo-user] Checking an HD for problems

[Grant]

I just switched to a new WD Caviar Black hard drive (really fast and
quiet!) and I noticed some errors when I was cp -ax'ing everything
from my old drive to the new drive which were accompanied by loud
clicks.  Is there a way to do a comprehensive test/check of the old
drive to see if it has any problems?

- Grant

Re: [gentoo-user] Checking an HD for problems

[Mark Knecht]

On Wed, Sep 22, 2010 at 9:46 AM, Grant <emailgrant@gmail.com> wrote:
> I just switched to a new WD Caviar Black hard drive (really fast and
> quiet!) and I noticed some errors when I was cp -ax'ing everything
> from my old drive to the new drive which were accompanied by loud
> clicks.  Is there a way to do a comprehensive test/check of the old
> drive to see if it has any problems?
>
> - Grant

I know it's not a popular solution here in Linux-land, and it's pretty
slow for large drives, but I still use SpinRite for that sort of
thing.

As a quick test, if the old drive has S.M.A.R.T. is to read the data
held in the drive to tell you if the on-board controller is seeing
problems.

Hope this helps,
Mark

Re: [gentoo-user] Checking an HD for problems

[Stroller]

On 22 Sep 2010, at 17:46, Grant wrote:
> ... I noticed some errors when I was cp -ax'ing everything
> from my old drive to the new drive which were accompanied by loud
> clicks.  Is there a way to do a comprehensive test/check of the old
> drive to see if it has any problems?

You don't need to do a test. The disk that is making the noises is f**ked.

Assuming that it's the old drive that is knackered, and if you're not certain that all important data has been copied correctly, then use GNU ddrescue (there is more than one dd_rescue, and GNU's is the best one) to do a bitwise clone of the data. Follow the examples in the manual to do multiple passes - the first pass will get most of the data from good sectors, subsequent passes will make repeated attempts at the bad sectors.

Stroller.

[gentoo-user] Re: Checking an HD for problems

[walt]

On 09/22/2010 01:26 PM, Stroller wrote:
>
> On 22 Sep 2010, at 17:46, Grant wrote:
>> ... I noticed some errors when I was cp -ax'ing everything
>> from my old drive to the new drive which were accompanied by loud
>> clicks.  Is there a way to do a comprehensive test/check of the old
>> drive to see if it has any problems?
>
> You don't need to do a test. The disk that is making the noises is f**ked.
>
> Assuming that it's the old drive that is knackered...

I was thinking the same.  In the past three or four years I've had more
brand new drives go bad than older ones.  Funny, though, the replacement
drives I've received under warranty work spectacularly well.  Just luck?

Re: [gentoo-user] Re: Checking an HD for problems

[Alan McKinnon]

Apparently, though unproven, at 23:00 on Wednesday 22 September 2010, walt did 
opine thusly:

> On 09/22/2010 01:26 PM, Stroller wrote:
> > On 22 Sep 2010, at 17:46, Grant wrote:
> >> ... I noticed some errors when I was cp -ax'ing everything
> >> from my old drive to the new drive which were accompanied by loud
> >> clicks.  Is there a way to do a comprehensive test/check of the old
> >> drive to see if it has any problems?
> > 
> > You don't need to do a test. The disk that is making the noises is
> > f**ked.
> > 
> > Assuming that it's the old drive that is knackered...
> 
> I was thinking the same.  In the past three or four years I've had more
> brand new drives go bad than older ones.  Funny, though, the replacement
> drives I've received under warranty work spectacularly well.  Just luck?


No, not luck. It's a numbers game and that how the dice roll.

Modern drives are complex. As such they are more likely to fail than ancient 
drives simply because of the complexity. They are also better engineered than 
old ones but the loss from complexity is greater than the game from better 
engineering. Plus, they are incredibly cheap compared to ancient times.

Engineered products all have characteristic failure rates common across the 
model, the infamous bathtub curve. The factory can't do the full range of 
nurn-in tests they'd like to (bean counters rule), so you get a drive at the 
later end of the bathtub. Hence, you see elevated failure rates. The factory 
is willing to take a financial knock here as the loss from a few replacements 
is much lower than the gigantic loss from fully and properly testing every 
drive for hours and hours.

You get a replacement. Simple odds are that it is not one of the few that will 
fail early, so it doesn't and you think "Wow! The gods like me." Nope, 
statistics like me.

If the factory was real smart, they would keep a small stock of fully tested 
drives on the replacement shelf, only to be released as under-warranty 
replacements. You'd be certain these drives would NOT fail and it's trivially 
easy to get this past the bean counters because you'd be winning back customer 
loyalty. And the cost of testing those few drives fully is not that much. The 
average bean counter has a ballistic orgasm at the thought of this, and yes 
they can even tell you the price they attach to winning back that loyalty.

So now you know. Accountants do not think like techies.

-- 
alan dot mckinnon at gmail dot com