Re: Bad Clusters vs. Bad Sectors

[Arno]

Hello!

Am I understanding correctly that these terms are basically the same?

They are not. Sector are the storage unit of the
underlying storage device, while clusters are the Microsoft
filesystem block sizes. Clusters are typically
a multiple of the sector size (512 Bytes for HDDs and 2048
Bytes for some optical media).

If
so, then why did my client's updated Windows 2000 SP4's chkdsk (/r /f
parameters and rebooted to run it) on a HDD (NTFS) in an old Dell
Optiplex system say there was a bad cluster and was able to move a file
to a better place, but I rerun chkdsk in Windows 2000 and ran a chkdsk
(no parameters) and it found 0 KB of bad sector?

Well, a bad cluster is usually a cluster marked bad in the
filesystem This is a leftover artefact from the time when
HDDs diod expose their bad areas to the OS. Typically a
cluster gets marked bad if one or more sectors in it
experience a read error.

A HDD only marks a sector as bad when it cannot by extended
effoprt read that sector. However it will still try to read
it on a new request from the OS and it will not export that
marking to the OS, except in the SMART selftest log and there
only for the first found one. What the marking by the HDD is
for is to allow it to reallocated (replace with a good one)
the sector on a write.

So sectors is on the hardware and the low-level drivesr,
while clusters is in the (abstract) filesystem layer.

Arno

 

[Rod Speed]

Ant wrote

Ant wrote

Is bad block same as bad cluster (OS side)?

Yes. Is mostly MS OSs that call them clusters.

 

[Rod Speed]

Ant wrote

Rod Speed wrote

Ah OK. I guess there was no bad sector then on that old HDD then.

Or there was a bad sector and the drive fixed it.

 

[Rod Speed]

Ant wrote

Rod Speed wrote

If there was a bad sector, wouldn't I see a bad sector shown in chkdsk?

Not necessarily. They arent necessarily always bad for every read.

Hmm, maybe I should run SpinRite v5 on that HDD later.

Its completely useless now.

The only thing that makes any sense anymore is the SMART data.

 

[Arno]

On 10/5/2009 12:02 AM PT, Ant typed:

Is bad block same as bad cluster (OS side)?

Not quite. Low level there are no clusters. And high-level (filesystem)
a bad cluster is a cluster (e.g. 4kB) with a bad sector in it (512B)

Arno

 

[Rod Speed]

Ant wrote

Rod Speed wrote

Really? How so?

What it claims to be doing isnt even possible with modern hard drives.

I remember using it in the past and finding bad sectors.

A SMART scan does that much better.

 

[Arno]

On 10/12/2009 12:00 PM PT, Rod Speed typed:

Really? How so? I remember using it in the past and finding bad sectors.

SpinRite does find bad sectors, but not more or others than
a long SMART selftest. Before disk selftests SpinRite had its
justification. Today it has none.

Arno

 

[Rod Speed]

Ant wrote

Rod Speed wrote

How old were the the modern HDDs started?

The last time what it claims to be able to do was possible was
with stepper motor head actuator drives, mostly MFM and RLL
and a very few of the first IDEs.

I can't remember when SMART started.

That varied with the manufacturer.

I think my former Quantum Fireball 6.4 GB EIDE HDD had a simple SMART.

Yeah, most of that era did.

 

[Arno]

On 10/13/2009 6:41 AM PT, Arno typed:

OK, I have used smartctl's long tests before. I assume that was enough
to find them.

I believe so. I have done this repeatedly on several disks with problems
and the results were basically the same. One critial difference between
now and the time when SpinRiste made sense is that modern disks read
the signal in analog form while old disks did immerdiately convert
the signal to 1/0. Analog reads allow a much better signal quality
assessment, and hence assesment of the margin between the signal read
and one that would result in a non-recoverable sector. This way
marginal sectors can be flagged as bad before there are actual read
errors. Hostorically, this required the tricks SpinRite did, but
not anymore.

Arno

 

[Rod Speed]

Ant wrote

Ant wrote

I was reading SpinRite's FAQ (http://www.grc.com/sr/faq.htm) and Wikipedia (http://en.wikipedia.org/wiki/SpinRite):

"Does SpinRite support hard drive SMART capability? Yes. SpinRite supports SMART more completely and usefully than
anything else ever has.

Another bare faced lie.

Upon first use, SpinRite activates and enables a drive's disabled SMART subsystem. When SpinRite is not actively
running on the drive, the "static" state of the drive's SMART data is displayed.

When SpinRite is running on a drive, the drive's SMART data — both
the standard SMART parameters and the more detailed SMART event
counters — are continuously polled, monitored, and displayed.

Much more useful information about the true health and robustness of a drive can be gained by monitoring the SMART
system's feedback while the drive is under an actual workload. No other software has ever done this.

Another bare faced lie.

And it doesnt change enough to matter anyway.

Note: Some default SATA configurations can limit SpinRite's ability to obtain SMART information from SATA drives even
though all other data recovery operations will work without limitation. See the SATA
knowledgebase article for specific information about SpinRite v6.0's
operation with SATA drives and controllers."

"... Version 6 is rather different from previous versions. It offers full access to the entire disk surface regardless
of partitioning,

So does any SMART tool, fool.

Self-Monitoring, Analysis, and Reporting Technology (SMART) parameters
and control of partial scanning within a specified percentage range..."

Meaningless waffle.

It sounds SpinRite v6 is still useful.
Nope.

Am I missing something?

Yep, that its not doing anything that any SMART tool isnt doing.

 

[Arno]

On 10/13/2009 10:29 PM PT, Ant typed:

I was reading SpinRite's FAQ (http://www.grc.com/sr/faq.htm) and
Wikipedia (http://en.wikipedia.org/wiki/SpinRite):

"Does SpinRite support hard drive SMART capability? Yes. SpinRite
supports SMART more completely and usefully than anything else ever has.

A direct lie.

Upon first use, SpinRite activates and enables a drive's disabled SMART
subsystem.

Yes, I do that manually. HDDsentinel does it. Linux smartd does it.
Speedfan does it. Other tools do it too.

When SpinRite is not actively running on the drive, the
"static" state of the drive's SMART data is displayed. When SpinRite is
running on a drive, the drive's SMART data ? both the standard SMART
parameters and the more detailed SMART event counters ? are continuously
polled, monitored, and displayed.

So what? Nobody neds that. It is far better to have change-monitoring
for the critical attributes than to manually have to look at a
continuously updated display. If you really want it, many tools
allow you to do this.

Much more useful information about the
true health and robustness of a drive can be gained by monitoring the
SMART system's feedback while the drive is under an actual workload. No
other software has ever done this.

Another direct lie.

Note: Some default SATA configurations can limit SpinRite's ability to
obtain SMART information from SATA drives even though all other data
recovery operations will work without limitation. See the SATA
knowledgebase article for specific information about SpinRite v6.0's
operation with SATA drives and controllers."

"... Version 6 is rather different from previous versions. It offers
full access to the entire disk surface regardless of partitioning,
Self-Monitoring, Analysis, and Reporting Technology (SMART) parameters
and control of partial scanning within a specified percentage range..."

It sounds SpinRite v6 is still useful. Am I missing something?

Yes. These people have abandoned all atempts at honesty. I do not
say their tool is useless, but it is not better than the
alternatives.

Incidentially several other tools will run regular surface
scans, and they are critical for early problem detection.

Arno

 

[Yousuf Khan]

It sounds SpinRite v6 is still useful. Am I missing something?

A lot of SpinRite's claims are simply marketing. That's all you're missing.

Yousuf Khan

 

[Arno]

A lot of SpinRite's claims are simply marketing. That's all you're missing.

Yousuf Khan

Good summary.

Arno

 

[Arno]

On 10/18/2009 4:04 PM PT, Arno typed:

Thanks. Does Smartctl's long tests do surface scans?

Yes. I meant you can shedule them to run automatically at set
intervalls. Smartctl cannot do that, but the smartd in the
same package can and at least n Debian does as default.

Arno

 

[bbbl67]

Thanks all. Someone should update Wikipedia's datas too.

Ouch! It looks like there is a war going on within that Wiki article's
talk section. Almost as bad as any war that goes on in here.

Yousuf Khan

 

[Arno]

Ouch! It looks like there is a war going on within that Wiki article's
talk section. Almost as bad as any war that goes on in here.

Yousuf Khan

I am staying out of that. Some people just want to believe
in the snake-oil. Typically not those that actually understand
what a product does.

Arno

 

[Arno]

Yes. I meant you can shedule them to run automatically at set
intervalls. Smartctl cannot do that, but the smartd in the
same package can and at least n Debian does as default.

[/QUOTE]

Interesting. Can I run it as not a daemon and do it all at once when I
am not using the computer? Sort of like running chkdsk /f /r type of thing.

If you want to run the surface scan manually, just use
smartctl -t long <device>

Arno