Seagate SMART, Raw Read Error Rate, Seek Error Rate

Franc Zabkar · Sep 1, 2007

I'm trying to make sense of the SMART reports for my 13GB and 120GB
Seagate hard drives. Both have very high numbers for the Raw Read
Error Rate and Seek Error Rate. At the moment the Raw Read Error Rate
for the 13GB seems to be unchanging, but the Seek Error Rate increases
every time I look at it. Also, if I compare today's Raw Read Error
Rate with the result from two years ago, the number is actually much
lower today. Does anyone know how these figures are calculated, or
even if they mean what they appear to mean?

These are recent reports produced by SmartUDM:
http://www.users.on.net/~fzabkar/SmartUDM/13GB.RPT
http://www.users.on.net/~fzabkar/SmartUDM/120GB.RPT

These reports were produced by Everest:
http://www.users.on.net/~fzabkar/SmartUDM/SMART_05.txt (2005)
http://www.users.on.net/~fzabkar/SmartUDM/SMART_07.txt (2007)
http://www.users.on.net/~fzabkar/SmartUDM/SMART_scandisk.txt

The first report was done in Sept 2005, the second in the last couple
of days. The last report is the result after running Scandisk.

BTW, the Current Pending Sector Count of 1 reflects a sector that has
been marked as bad by the OS. I suspect that the drive's controller is
aware that it is bad, but it cannot relocate it until such time as the
OS writes to it, thereby signalling that the data in that sector is no
longer of any consequence.

FWIW, SeaTools Desktop v3.00 says the 13GB drive is OK, apart from one
bad sector.

- Franc Zabkar

Arno Wagner · Sep 1, 2007

Previously Franc Zabkar said:
I'm trying to make sense of the SMART reports for my 13GB and 120GB
Seagate hard drives. Both have very high numbers for the Raw Read
Error Rate and Seek Error Rate.

Raw read error is very hard to interpret and usually not
important anyways. Seek errors are usually a poer problem
or a vibration problem. They may also indicate a problem
with the disk.

At the moment the Raw Read Error Rate
for the 13GB seems to be unchanging, but the Seek Error Rate increases
every time I look at it. Also, if I compare today's Raw Read Error
Rate with the result from two years ago, the number is actually much
lower today. Does anyone know how these figures are calculated, or
even if they mean what they appear to mean?

These are recent reports produced by SmartUDM:
http://www.users.on.net/~fzabkar/SmartUDM/13GB.RPT
http://www.users.on.net/~fzabkar/SmartUDM/120GB.RPT

These reports were produced by Everest:
http://www.users.on.net/~fzabkar/SmartUDM/SMART_05.txt (2005)
http://www.users.on.net/~fzabkar/SmartUDM/SMART_07.txt (2007)
http://www.users.on.net/~fzabkar/SmartUDM/SMART_scandisk.txt

The first report was done in Sept 2005, the second in the last couple
of days. The last report is the result after running Scandisk.

BTW, the Current Pending Sector Count of 1 reflects a sector that has
been marked as bad by the OS.

Not quite. It represents a sector that the drive has given up on, but
not yet been able to replace, because it was not written to it.
The OS does not factor into this.

A bad sector marked by the disk (and invisible to the OS) can
be counter as "reallocation event" or "reallocated sector
count". If these numbers start growing, something is seriously
wrong.

I suspect that the drive's controller is
aware that it is bad, but it cannot relocate it until such time as the
OS writes to it, thereby signalling that the data in that sector is no
longer of any consequence.
Yes.

FWIW, SeaTools Desktop v3.00 says the 13GB drive is OK, apart from one
bad sector.

One bad sector is no reason for concern. If they start to get more,
that would be.

Arno

Franc Zabkar · Sep 2, 2007

Raw read error is very hard to interpret and usually not
important anyways. Seek errors are usually a poer problem
or a vibration problem. They may also indicate a problem
with the disk.

Not quite. It represents a sector that the drive has given up on, but
not yet been able to replace, because it was not written to it.
The OS does not factor into this.

Sorry, my statement was ambiguous. Maybe I should have written that
"the Current Pending Sector Count of 1 coincides with a sector that
has been marked as bad by the OS".

A bad sector marked by the disk (and invisible to the OS) can
be counter as "reallocation event" or "reallocated sector
count". If these numbers start growing, something is seriously
wrong.

The numbers *are* growing. In fact they've grown from 34 to 119 in two
years. I've been preparing to replace the drive for quite some time
now. However, it's only in the last month or so that the drive has
been making occasional noises, ie a very soft clink, probably from the
voice coil positioner.

One bad sector is no reason for concern. If they start to get more,
that would be.

Arno

I now have a batch file that runs just prior to shutdown. Among other
things, it captures SMART data and appends it to a log file. It'll be
interesting to monitor the drive as it progresses toward total
failure. :-)

BTW, these are the SMART data for my Fujitsu 6GB drive:
http://www.users.on.net/~fzabkar/SmartUDM/6GB.RPT

Notice the raw value for "Power On Hours Count".

0000008EF98Ah = 9369994 dec
= 1069 years

In fact the figure appears to represent Power On Seconds.

- Franc Zabkar

Arno Wagner · Sep 2, 2007

Previously Franc Zabkar said:
On 1 Sep 2007 07:29:59 GMT, Arno Wagner <[email protected]> put finger to
keyboard and composed:

Sorry, my statement was ambiguous. Maybe I should have written that
"the Current Pending Sector Count of 1 coincides with a sector that
has been marked as bad by the OS".
Ok.

The numbers *are* growing. In fact they've grown from 34 to 119 in two
years. I've been preparing to replace the drive for quite some time
now. However, it's only in the last month or so that the drive has
been making occasional noises, ie a very soft clink, probably from the
voice coil positioner.

Well. Personally I stop trustinf a disk around 10 or so, unless
they all happened in one burst. I have had one Maxtor disk with
something like 200 reallocated sectors in one event, which
did run fine without any additional ones for three years afterwards.

So, it could be a problem with power (spikes, I would suspect),
mechanical shock/vibration or the like. Or the disk could have
a problem. I would replace that one. Also, at some time the disk will
run out of spare sectors.

I now have a batch file that runs just prior to shutdown. Among other
things, it captures SMART data and appends it to a log file. It'll be
interesting to monitor the drive as it progresses toward total
failure.

BTW, these are the SMART data for my Fujitsu 6GB drive:
http://www.users.on.net/~fzabkar/SmartUDM/6GB.RPT

Notice the raw value for "Power On Hours Count".

0000008EF98Ah = 9369994 dec
= 1069 years

In fact the figure appears to represent Power On Seconds.

This is a non-standardized field, AFAIK. Bogus readings are
no surprise here.

Arno

Franc Zabkar · Sep 2, 2007

This is a non-standardized field, AFAIK. Bogus readings are
no surprise here.

Arno

I suspect that the figures aren't necessarily bogus, they may just
need to be interpreted differently between manufacturers. That said, I
haven't been able to find any detailed SMART documentation at any of
the manufacturers' web sites.

- Franc Zabkar

Franc Zabkar · Sep 2, 2007

Raw read error is very hard to interpret and usually not
important anyways. Seek errors are usually a poer problem
or a vibration problem. They may also indicate a problem
with the disk.

I'm running Win98SE. While monitoring file accesses with Filemon (from
SysInternals), I used Everest Home Edition (ver 2.20.405) to monitor
the drive's SMART data. Every time I refreshed the SMART report (using
F5), the Seek Error Rate figure increased by 10 points. However the
Filemon capture window remained empty. How can a drive incur seek
errors if there are no file accesses? I would think that the SMART
data would be retrieved from the drive's RAM or flash EEPROM, so no
actual seeks would be required.

- Franc Zabkar

Folkert Rienstra · Sep 2, 2007

Franc Zabkar wrote in message news:[email protected]

I'm running Win98SE. While monitoring file accesses with Filemon (from
SysInternals), I used Everest Home Edition (ver 2.20.405) to monitor
the drive's SMART data. Every time I refreshed the SMART report (using
F5), the Seek Error Rate figure increased by 10 points. However the
Filemon capture window remained empty. How can a drive incur seek
errors if there are no file accesses?

What file access.

I would think that the SMART data would be retrieved from the
drive's RAM or flash EEPROM, so no actual seeks would be required.

Well, guess what.

Folkert Rienstra · Sep 2, 2007

Franc Zabkar wrote in message news:[email protected]
They all are, you babblebot moron.

Bogus readings are no surprise here.

As is your babbling, babblebot.

I suspect that the figures aren't necessarily bogus,

It's the babblebot, it just babbles.

they may just need to be interpreted differently between manufacturers.

Which is what 'Vendor specific' means.

Folkert Rienstra · Sep 2, 2007

Franc Zabkar wrote in message news

[email protected]

Sorry, my statement was ambiguous.

No it wasn't. Everybody else but the babblebot got it.

Maybe I should have written that "the Current Pending Sector Count
of 1 coincides with a sector that has been marked as bad by the OS".

Which would have gotten the same response from the babblebot.

The numbers *are* growing. In fact they've grown from 34 to 119 in
two years. I've been preparing to replace the drive for quite some
time now. However, it's only in the last month or so that the drive has
been making occasional noises, ie a very soft clink, probably from the
voice coil positioner.

Utter nonsense as always from the babblebot.

Folkert Rienstra · Sep 2, 2007

Arno Wagner wrote in message news:[email protected]

Well. Personally I stop trustinf a disk around
10 or so, unless they all happened in one burst.

Which makes perfect sense if you are babblebot.

I have had one Maxtor disk with something like 200 reallocated sectors in
one event, which did run fine without any additional ones for three years
afterwards.

Which obviously you only know by waiting for 3 years.

So, it could be a problem with power (spikes, I would suspect), mecha-
nical shock/vibration or the like. Or the disk could have a problem.

I would replace that one.

And subject the new disk to the same problem causes. Very good, babblebot.

Also, at some time the disk will run out of spare sectors.

At this rate somewhere in the next century which is very bad, eh babblebot.

This is a non-standardized field, AFAIK.

Little you know. They all are, babblebot.

Bogus readings are no surprise here.

As is your response.

Eric Gisin · Sep 2, 2007

Franc Zabkar said:
I'm running Win98SE. While monitoring file accesses with Filemon (from
SysInternals), I used Everest Home Edition (ver 2.20.405) to monitor
the drive's SMART data. Every time I refreshed the SMART report (using
F5), the Seek Error Rate figure increased by 10 points. However the
Filemon capture window remained empty. How can a drive incur seek
errors if there are no file accesses? I would think that the SMART
data would be retrieved from the drive's RAM or flash EEPROM, so no
actual seeks would be required.

SMART diagnostic I/O does not show up as Windows I/O.

Arno Wagner · Sep 2, 2007

Previously Franc Zabkar said:
On 2 Sep 2007 05:29:04 GMT, Arno Wagner <[email protected]> put finger to
keyboard and composed:

I suspect that the figures aren't necessarily bogus, they may just
need to be interpreted differently between manufacturers.

That is what I meant. The raw values ace accurate, but the interpreted
figures are ofteh wrong.

That said, I
haven't been able to find any detailed SMART documentation at any of
the manufacturers' web sites.

SMART is part of the ATA spec. You can find specs on the t13 comitte
website here: http://www.t13.org/

Arno

Folkert Rienstra · Sep 2, 2007

Arno Wagner wrote in news:[email protected]

That is what I meant. The raw values ace accurate, but the interpreted
figures are ofteh wrong.

SMART is part of the ATA spec. You can find specs on the t13 comitte
website here: http://www.t13.org/

Which obviously you didn't bother to consult.

Christian Franke · Sep 2, 2007

Yes, the FUJITSU MPE3064AT Attribute 9 raw value counts seconds.
Other drives use hours, minutes or half minutes.
Some SMART tools handle these differences, most don't.
See info about Attribute 9 in http://smartmontools.sourceforge.net/#FAQ

That is what I meant. The raw values ace accurate, but the interpreted
figures are ofteh wrong.

SMART is part of the ATA spec. You can find specs on the t13 comitte
website here: http://www.t13.org/

Unlike SMART status, self-tests and logs, SMART attributes are *not*
standardized in ATA-3...8. Even the general data format isn't standardized.

Specific Attributes are only listed in a proposed informal annex for
ATA-8. But it is still not included in the draft.

See "ATA References" at http://smartmontools.sourceforge.net/#references
for links & comments.

Christian

Franc Zabkar · Sep 2, 2007

SMART diagnostic I/O does not show up as Windows I/O.

Understood, and that's essentially what I wrote. I merely used Filemon
to confirm that Everest wasn't doing something else as well. So the
question remains, why does SMART diagnostic I/O cause the Seek Error
Rate figure to change?

- Franc Zabkar

Folkert Rienstra · Sep 2, 2007

Franc Zabkar wrote in news:[email protected]

Understood,

No, not really.

and that's essentially what I wrote.

No, you didn't.

I merely used Filemon to confirm that Everest wasn't doing something
else as well.

Like File IO maybe?

So the question remains, why does SMART diagnostic I/O cause the

Seek Error Rate figure to change?

Maybe because it does "SMART diagnostic I/O" ?
Just a calculated guess. What do you think.

Franc Zabkar · Sep 3, 2007

Franc Zabkar wrote in news:[email protected]

No, not really.

No, you didn't.

Like File IO maybe?

Maybe because it does "SMART diagnostic I/O" ?
Just a calculated guess. What do you think.

I think you are a putrefying dog turd that should go into my kill
file.

- Franc Zabkar

Franc Zabkar · Sep 3, 2007

Unlike SMART status, self-tests and logs, SMART attributes are *not*
standardized in ATA-3...8. Even the general data format isn't standardized.

Specific Attributes are only listed in a proposed informal annex for
ATA-8. But it is still not included in the draft.

See "ATA References" at http://smartmontools.sourceforge.net/#references
for links & comments.

Christian

Thanks for that. I'd been looking for SFF-8035 without success.

BTW, Seagate's docs weren't very useful at all. :-(

- Franc Zabkar

Arno Wagner · Sep 3, 2007

Yes, the FUJITSU MPE3064AT Attribute 9 raw value counts seconds.
Other drives use hours, minutes or half minutes.
Some SMART tools handle these differences, most don't.
See info about Attribute 9 in http://smartmontools.sourceforge.net/#FAQ

Unlike SMART status, self-tests and logs, SMART attributes are *not*
standardized in ATA-3...8. Even the general data format isn't standardized.

Specific Attributes are only listed in a proposed informal annex for
ATA-8. But it is still not included in the draft.

See "ATA References" at http://smartmontools.sourceforge.net/#references
for links & comments.

Aha. Interesting! That explains the mess some attributes are.

Arno

Christian Franke · Sep 3, 2007

Franc said:
...

Understood, and that's essentially what I wrote. I merely used Filemon
to confirm that Everest wasn't doing something else as well. So the
question remains, why does SMART diagnostic I/O cause the Seek Error
Rate figure to change?

Any attribute may be updated during a "Auto Offline Data Collection".
This is intended to update the SMART attributes marked "Offline".

There is more detailed info in the smartctl man page.
See the description of the "--offlineauto" option at:
http://smartmontools.sourceforge.net/man/smartctl.8.html
(BTW: this tool works also on Win9x/ME if driver provides smartvsd.vxd)

Christian

Seagate SMART, Raw Read Error Rate, Seek Error Rate

Franc Zabkar

Arno Wagner

Franc Zabkar

Arno Wagner

Franc Zabkar

Franc Zabkar

Folkert Rienstra

Folkert Rienstra

Folkert Rienstra

Folkert Rienstra

Eric Gisin

Arno Wagner

Folkert Rienstra

Christian Franke

Franc Zabkar

Folkert Rienstra

Franc Zabkar

Franc Zabkar

Arno Wagner

Christian Franke