read errors

[Squeeze]

Franc Zabkar wrote in news:[email protected]

On 30 May 2008 10:51:13 GMT, Arno Wagner <[email protected]> put finger to keyboard and composed:
My config is
c = 80 gig
d = 200 gig problem
both running ntfs

looked at smart from, everes,t first thing and saw no problems
recognizable by me.

[snip]

Franc
So should I return to seagate or what?
tia

I would try to obtain a warranty replacement. I recently took a Seagate
drive out of service after living with bad sectors for several years.
Towards the end it started to grow new defects on a weekly basis.
It still had only 130 bad sectors, which is well short of Seagate's SMART
threshold. Some people will retire a drive with a single bad sector
because they worry that a new defect may appear in a critical area.
It's your choice, but I definitely wouldn't continue to use your drive.

Right, good advice again, retire the drive and get a new one.
While time progresses, it should be fun to watch and see it
gather new defects again, just like that Samsung does already.
Who cares what data you'll loose on the new drive.
It will be a new drive and that's what counts, obviously.
You can always retire the new drive too, right, as long
as the warranty on your current one doesn't run out.

Here are a few SMART reports for various drives:
http://www.users.on.net/~fzabkar/SmartUDM

This 120MB Seagate drive is perfectly good even though several
attributes look very bad:
http://www.users.on.net/~fzabkar/SmartUDM/120GB.RPT

This is the drive I retired:
http://www.users.on.net/~fzabkar/SmartUDM/13GB.RPT

This is what Seagate's FAQ has to say about SMART:
http://preview.tinyurl.com/3k34qc

===================================================================

How do I interpret SMART diagnostic utilities results?
------------------------------------------------------
As a matter of policy, Seagate does not publish attributes and thresholds.

The SMART values that might be read out by third-party SMART software
are not based on how the values may be used within the Seagate hard
drives. Seagate does not provide support for software programs that
claim to read individual SMART attributes and thresholds. There may
be some historical correctness on older drives, but new drives, no
doubt, will have incorporated newer solutions, attributes and thresholds.

Seagate uses the general SMART Status, pass or fail. The individual
attributes and threshold values are proprietary and we do not offer a
utility that will read out the values. If the values that you are
seeing with a third party SMART utility are not displaying properly or
seem to be false, please contact your software vendor for further
explanation of the values.
===================================================================

Well, so much for your own Seagate drive SMART observations and conclusions then.

 

[DonLogan]

Rod Speed wrote in news:[email protected]

DonLogan <[email protected]> wrote

My config is
c = 80 gig
d = 200 gig problem
both running ntfs

looked at smart from, everes,t first thing and saw no problems
recognizable by me.

There are real problems tho, particularly with the second drive.

----------------------------------------------------------------------------

--------[ EVEREST Home Edition (c) 2003-2005 Lavalys, Inc. ]------------------------------------------------------------ [snip more unreadable stuff ]

<snipped stuff we didnt need>

sorry if formating an issue

Yeah, thats one real downside with Everest, along with the very
misleading OKs on those obscenely bad results with the bad sectors.

So the drive reports smart #s and the app, Everest, decides if it's OK & passes,

No, Everest reports the smart #s that the drive provides,

and sticks its own OK on the end of each line,

Sort of. And it's not just 'OK', it is 'OK: description' where description
depends on the Pre-fail/Advisory bit. It's a characterization of how the
drive regards these values when deciding on the drive's SMART status.

Each attribute has it's own internal SMART status, the significance of
which depends on the attribute's Prefail/Advisory bit and on whether
the normalized value has dipped below the threshold value.
If a value has dipped below threshold it will cause a SMART RETURN STATUS
of "threshold exceeded" which -depending on the Pre-fail/Advisory bit- is
'OK' or 'NotOK'. Some drives will only look at the values with pre-fail set.

So each individual attribute's SMART status is characterized as something
like
OK: always passing (advisory) or
OK: Value is normal (pre-fail) or
NotOK: Value exceeded threshold (pre-fail)>

does this mean that
ok - always passing is best
ok - value is normal is oh oh
and NotOk - warranty
and do the values display as billions or roll over at 999 million?

SMART says OK, but it isn't. I'm getting read errors!

You said 'S.M.A.R.T attributes have been abandoned several specs ago
already.' Oh oh. I'll investigate when i can get onto a highspeed.
I'm on dial-up. Do we have a successful replacement?

I'll try FindBad

thanks

Other apps may use different wording, like old_age and pre-fail.

and that OK should be ignored.

Not ignored but judged in the proper context.

Nope.

Nope, Rodbots?

And it isnt relevant to a drive thats dying anyway.

IF it's dying.

A few drives do have write verify enabled for the first few power cycles, but
thats mainly some Maxtors what are conservative about marginal sectors.

So no. And IDE drives have no Write Verify commands.

No, that what all drives do all the time.

All the time, Rodbots?

Its not relevant to a dying drive.

So what if it's *not* dying, Rodbots?

--------------------------

[snip]

What ?

I am going to replace or retire based on seagate accepting the smart as proof of failure.

Run Seagate's diagnostic on that drive

to get a warranty replacement.

And when completed you can do the same for the Samsung as it must be dying too.
That is, if you *must* believe the Rod- and Babblebots.

It can be interesting to play with a dying drive, but you'll
have to return the corpse to get a warranty replacement.

You could try replacing the power supply first
and see if that fixes the problem,

It won't fix it if you don't fix the drive too,
fixing the pending bad sectors first.

but I bet it wont.

Which is short for
'please don't do it so I won't run the chance of looking a fool'.

 

[Christian Franke]

Yes, the standard is that there is no real standard.
S.M.A.R.T attributes have been abandoned several specs ago already.

AFIAK, the attributes itself were never specified in any ATA standard.
The last spec of the attribute data format was in the last ATA-3 draft
from 1997.

See http://smartmontools.sourceforge.net/links.html#references for
further links.

Christian

 

[Arno Wagner]

Rod Speed wrote in news:[email protected]

My config is
c = 80 gig
d = 200 gig problem
both running ntfs

looked at smart from, everes,t first thing and saw no problems
recognizable by me.

There are real problems tho, particularly with the second drive.

----------------------------------------------------------------------------

--------[ EVEREST Home Edition (c) 2003-2005 Lavalys, Inc. ]------------------------------------------------------------ [snip more unreadable stuff ]

<snipped stuff we didnt need>

sorry if formating an issue

Yeah, thats one real downside with Everest, along with the very
misleading OKs on those obscenely bad results with the bad sectors.

So the drive reports smart #s and the app, Everest, decides if it's OK & passes,

No, Everest reports the smart #s that the drive provides,

and sticks its own OK on the end of each line,

Sort of. And it's not just 'OK', it is 'OK: description' where description
depends on the Pre-fail/Advisory bit. It's a characterization of how the
drive regards these values when deciding on the drive's SMART status.

Each attribute has it's own internal SMART status, the significance of
which depends on the attribute's Prefail/Advisory bit and on whether
the normalized value has dipped below the threshold value.
If a value has dipped below threshold it will cause a SMART RETURN STATUS
of "threshold exceeded" which -depending on the Pre-fail/Advisory bit- is
'OK' or 'NotOK'. Some drives will only look at the values with pre-fail set.

So each individual attribute's SMART status is characterized as something
like
OK: always passing (advisory) or
OK: Value is normal (pre-fail) or
NotOK: Value exceeded threshold (pre-fail)>

does this mean that
ok - always passing is best
ok - value is normal is oh oh
and NotOk - warranty
and do the values display as billions or roll over at 999 million?

SMART says OK, but it isn't. I'm getting read errors!

You said 'S.M.A.R.T attributes have been abandoned several specs ago
already.' Oh oh. I'll investigate when i can get onto a highspeed.
I'm on dial-up. Do we have a successful replacement?

I'll try FindBad

I would advise you to disregard what ''Squeeze'' says. He is
here for the posturing and dissing of others. His competence
is rather low. SMART attributes are perfectly fine, when
maintained in sync with the disks that are out there.

SMART thresholds are set by the vendores and they are typically
over-optimistic. Looking at the raw and cooked values still tells
a knowledgeable person a lot.

Arno

 

[Squeeze]

DonLogan wrote in news:[email protected]

Rod Speed wrote in news:[email protected]

My config is
c = 80 gig
d = 200 gig problem
both running ntfs

looked at smart from, everes,t first thing and saw no problems
recognizable by me.

There are real problems tho, particularly with the second drive.


--------[ EVEREST Home Edition (c) 2003-2005 Lavalys, Inc. ]------ [snip more unreadable stuff ]

<snipped stuff we didnt need>

sorry if formating an issue

Yeah, thats one real downside with Everest, along with the very
misleading OKs on those obscenely bad results with the bad sectors.

So the drive reports smart #s and the app, Everest, decides if it's OK & passes,

No, Everest reports the smart #s that the drive provides,

and sticks its own OK on the end of each line,

Sort of. And it's not just 'OK', it is 'OK: description' where description
depends on the Pre-fail/Advisory bit. It's a characterization of how the
drive regards these values when deciding on the drive's SMART status.

Each attribute has it's own internal SMART status, the significance of
which depends on the attribute's Prefail/Advisory bit and on whether
the normalized value has dipped below the threshold value.
If a value has dipped below threshold it will cause a SMART RETURN STATUS
of "threshold exceeded" which -depending on the Pre-fail/Advisory bit- is
'OK' or 'NotOK'. Some drives will only look at the values with pre-fail set.

So each individual attribute's SMART status is characterized as something
like
OK: always passing (advisory) or
OK: Value is normal (pre-fail) or
NotOK: Value exceeded threshold (pre-fail)

does this mean that
ok - always passing is best

Not at all, only that this value won't ever give you a failure is imminent
status. When such a value drops below theshold it says that the drive is
nearing it's useful lifes end, getting old of age.

ok - value is normal is oh oh

Depends on the value, of how far it deviated from the default value
that_it_started_out_life_with and towards the threshold value. All
'value is normal' means is that the normalized value is still above thres-
hold and that this value is not responsible for 'threshold exceeded'.

and NotOk - warranty
Probably.

and do the values display as billions or roll over at 999 million?

I have no idea where that comes from.

SMART says OK, but it isn't.

Yes it is.

I'm getting read errors!

No, really? Haven't you been explained why that is?
Oh wait, you were confused by the babblebots who said that it could
be caused by bad powersupply but then forgot all about that and said
that you *definetely* should replace the one drive though both your
drives have suffered from bad sectors and therefor the powersupply
should be the likely culprit. That didn't make sense, now did it.

You said 'S.M.A.R.T attributes have been abandoned several specs
ago already.' Oh oh.

It adds just some background to what Franc and Christian said.
I just posted an article on that subject in response to Christian.

I'll investigate when i can get onto a highspeed. I'm on dial-up.

Do we have a successful replacement?
What?


I'll try FindBad

Yeah. And to stop new bads being made while you write to the drive,
have a good look at the powersupply and supply of power to the drives.

thanks

[...]

 

[Squeeze]

Christian Franke wrote in news:[email protected]

AFIAK, the attributes itself were never specified in any ATA standard.
The last spec of the attribute data format was in the last ATA-3 draft
from 1997.

There is a proposal (e05148r0) by Jim Hatfield to add a SMART attributes list
as an informative annex to ATA8-ACS. It's about 3 years old already though.

Allow me my Franc Zabkar moment:

Quote

"2 Background

The documentation of how to access SMART attributes was removed before the final draft of ATA/ATAPI-3.
As I understand it, it was mainly a political issue. Drive vendors implemented attribute ‘X’ in different
ways, with different measurement scales and units. Customers (not understanding that) were trying to
compare vendor A with vendor B using the raw value of the attribute, and were making better/worse
judgements that were completely baseless.
Since then, the industry has stabilized many of the attributes through common customer requirements
being made of multiple drive vendors. Customers are more aware of the differences.\
Occasionally, there is a need for new attributes. Customers may dictate to drive vendors to implement
attribute ‘X’, defined in such and such a way.
Sometimes, these attributes are intentionally kept undocumented to the public, in order to provide market
differentiation between major OEMs.
There is another class of device users, however, in the open source community. These people still do not
understand the differences, and they publish assertions and software claiming to tell you information that
you ‘need to know’ about your ‘own property’ that ‘the others’ don’t want you to know. This proposal
would at least make a clear statement about attributes (in the absence of any standard) about how to
access and use the SMART attributes.

3.1.1 Overview
The information in this section was obsoleted in ATA/ATAPI-3. It is re-documented here for convenience,
as it continues to be used by some devices. Clarification has been added, since the original text was too
vague in places, and did not represent actual usage.
In late 1995, parts of SFF-8035i revision 2 (now obsolete) were merged with ATA/ATAPI-3.
Starting with ATA/ATAPI-4, there was no longer a requirement that a device maintain an attribute table.
Devices from then on were only required to return (via SMART RETURN STATUS) an OK or NotOK to
queries about their health. A ‘NotOK’ response indicates that the device considers itself ‘likely to fail’
(whatever that means). This left the interpretation of the values and threshholds completely up to the
device itself, eliminating a major source of confusion for host software.
ATA/ATAPI-5 added SMART error logs and self-tests to enhance the ability of a drive to report on its
health.

\Quote
from:
http://www.t13.org/Documents/UploadedDocuments/docs2005/e05148r0-ACS-SMARTAttributesAnnex.pdf

 

[Squeeze]

Arno Wagner wrote in news:[email protected]

Rod Speed wrote in
My config is
c = 80 gig
d = 200 gig problem
both running ntfs

looked at smart from, everes,t first thing and saw no problems
recognizable by me.

There are real problems tho, particularly with the second drive.

----------------------------------------------------------------------------

--------[ EVEREST Home Edition (c) 2003-2005 Lavalys, Inc.
]------------------------------------------------------------ [snip more unreadable stuff ]

<snipped stuff we didnt need>

sorry if formating an issue

Yeah, thats one real downside with Everest, along with the very
misleading OKs on those obscenely bad results with the bad sectors.

So the drive reports smart #s and the app, Everest, decides if it's OK & passes,

No, Everest reports the smart #s that the drive provides,

and sticks its own OK on the end of each line,

Sort of. And it's not just 'OK', it is 'OK: description' where description
depends on the Pre-fail/Advisory bit. It's a characterization of how the
drive regards these values when deciding on the drive's SMART status.

Each attribute has it's own internal SMART status, the significance of
which depends on the attribute's Prefail/Advisory bit and on whether
the normalized value has dipped below the threshold value.
If a value has dipped below threshold it will cause a SMART RETURN STATUS
of "threshold exceeded" which -depending on the Pre-fail/Advisory bit- is
'OK' or 'NotOK'. Some drives will only look at the values with pre-fail set.

So each individual attribute's SMART status is characterized as something
like
OK: always passing (advisory) or
OK: Value is normal (pre-fail) or
NotOK: Value exceeded threshold (pre-fail)>

does this mean that
ok - always passing is best
ok - value is normal is oh oh
and NotOk - warranty
and do the values display as billions or roll over at 999 million?

SMART says OK, but it isn't. I'm getting read errors!

You said 'S.M.A.R.T attributes have been abandoned several specs ago
already.' Oh oh. I'll investigate when i can get onto a highspeed.
I'm on dial-up. Do we have a successful replacement?

I'll try FindBad

I would advise you

That's what our babblebot does best and we all love him for it.

to disregard what ''Squeeze'' says.

He is here for the posturing and dissing of others.
His competence is rather low.

Sounds like that role fits you fine too, Babblebot.

SMART attributes are perfectly fine, when
maintained in sync with the disks that are out there.

Whatever that is supposed to mean.
Sounding academic without saying anything substantial.
A little bit of Babblebot posturing, perhaps?

SMART thresholds are set by the vendores and they are typically
over-optimistic.

Looking at the raw and cooked values still tells a knowledgeable person a lot.

So you just ruled yourself out then, Babblebot, since Franc
showed you rather clueless on the high Seagate numbers.

 

[DonLogan]

snip

I have no idea where that comes from.

Rae Read Error Rate & Reallocated Sector Count was
160,374,848
now is
49,678,669
so looks like it rolls over at 999,999,999

 

[Squeeze]

DonLogan wrote in news:[email protected]

snip

Rae Read Error Rate & Reallocated Sector Count was 160,374,848

Of course it was.

now is 49,678,669 , so looks like it rolls over at 999,999,999

Or that it is not a counter in the sense that it can only go up.
And it's a binary number, so no, a nice decimal number is not very likely.
Now think this over for a moment, would a number roll over without the threshold getting exceeded? What would the point of such
theshold be?

What is the number now? And what is it 5 minutes later.
What is it after powerup from cold.

 

[DonLogan]

DonLogan wrote in news:[email protected]


Of course it was. redudndantant

Or that it is not a counter in the sense that it can only go up.

was that a question?
i can think of several answers, but if you know a little help please

And it's a binary number, so no, a nice decimal number is not very likely. as long as its a number
Now think this over for a moment, would a number roll over without the threshold getting exceeded? no - bad idea
What would the point of such theshold be?

what's the threshold?
i see 4 numbers - w x y z
I'm quoting z

What is the number now? And what is it 5 minutes later.
What is it after powerup from cold.

yes, i can do that but isn't this old ground?
happy to record if it really means something and nobody knows.

(now 71,095,553 - so gone up this time - haven't rebooted)

 

[Squeeze]

Franc Zabkar wrote in news:[email protected]

I don't think you meant "Reallocated Sector Count". That figure grows
until the drive fails.

As for the "Raw Read Error Rate", it's difficult to understand how
Seagate uses this attribute. Seagate's FAQ states that ...

"The SMART values that might be read out by third-party SMART software
are not based on how the values may be used within the Seagate hard
drives. Seagate does not provide support for software programs that
claim to read individual SMART attributes and thresholds.

The individual attributes and threshold values are proprietary and we
do not offer a utility that will read out the values."

The "Seek Error Rate" value is also a strange one.

What "Seek Error Rate".
What exactly did you not understand in the lines from Seagate that you just quoted?

My testing suggests that it reflects a seek count, not an error, and not a rate:

http://groups.google.com/group/microsoft.public.windowsxp.hardware/msg/2ac63d875bfaf0d4
http://groups.google.com/group/microsoft.public.windowsxp.hardware/msg/b4eb3d937a08e2e3

However, if I look at the raw *totals* I get ...

http://www.users.on.net/~fzabkar/SmartUDM/13GB.RPT

Seek Error Rate 052E0E3000ECh
Power On Hours Count 0000000026C7h

If I divide "power on hours" by the "seek error rate" (aka seek count ???), ...

http://www.google.com/search?hl=en&q=0x26c7+/+0x52E0E3000EC+hours+in+microseconds&btnG=Search

... then I get 6.3 microseconds per seek, which seems absurd.

Let me give you some reflection on that, here is what Seagate writes:

"The SMART values that might be read out by third-party SMART
software are not based on how the values may be used within the Seagate
hard drives. Seagate does not provide support for software programs
that claim to read individual SMART attributes and thresholds."

Maybe you should read it sometime in between when you
are not occupied by your Franc Zabkar promotion.

Here is a spreadsheet of the last 3 month's of SMART data for my
failing 13GB Seagate hard drive:

http://www.users.on.net/~fzabkar/SmartUDM/SMART_13GB.XLS

Dude, that was so interesting.

 

[Squeeze]

Franc Zabkar wrote in news:[email protected]

I notice that there is an abrupt increase in the raw Seek Error Rate
value at row 25, from 0x052E0E39C161 to 0x052F0E39FD21. This
suggests that the uppermost 4 nibbles are not part of the seek count,
but instead reflect some other event or attribute. If we take only the
lower 32 bits, then we get an average of about 600msec between seeks
over the last 3 months of the drive's life, which makes much more sense.

Maybe the uppermost 16 bits represent the seek error count and the
lowermost 32 bits the total seek count ??? However, this still doesn't
explain why the "worst" normalised value for the Seek Error Rate
attribute is less than the current normalised value. Furthermore, my
13GB drive belongs to an earlier model family, so the SMART attributes
may be encoded differently to those of later models.

Perhaps you should re-read those lines from the Seagate FAQ again, no?
To get a clue?

Are you one of those people as mentioned by Jim Hatfield in that SMART annex proposal, Franc?

"There is another class of device users, however, in the open source community. These people still do not
understand the differences, and they publish assertions and software claiming to tell you information that
you ‘need to know’ about your ‘own property’ that ‘the others’ don’t want you to know"

 

[Christian Franke]

Christian Franke wrote in news:[email protected]

There is a proposal (e05148r0) by Jim Hatfield to add a SMART attributes list
as an informative annex to ATA8-ACS. It's about 3 years old already though.

I know ...

Allow me my Franc Zabkar moment:

Quote
[...]
\Quote
from:
http://www.t13.org/Documents/UploadedDocuments/docs2005/e05148r0-ACS-SMARTAttributesAnnex.pdf

.... because the above link and links to a more recent (3 part) version
are already on the mentioned link list:


The annex is still not included in the latest ATA-8 ACS draft 4c from
Dec 2007.

Christian

 

[Squeeze]

Christian Franke wrote in news:[email protected]

Christian Franke wrote in news:[email protected]

There is a proposal (e05148r0) by Jim Hatfield to add a SMART attributes list
as an informative annex to ATA8-ACS. It's about 3 years old already though.

I know ...

Allow me my Franc Zabkar moment:

Quote
[...]
\Quote
from:
http://www.t13.org/Documents/UploadedDocuments/docs2005/e05148r0-ACS-SMARTAttributesAnnex.pdf

... because the above link and links to a more recent (3 part) version
are already on the mentioned link list:

So it appears, thanks.