Seagate SMART, Raw Read Error Rate, Seek Error Rate

[Franc Zabkar]

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

OK, but why does the Seek Error Rate increase by 10 points when this
happens, if indeed that is what is happening? SmartUDM indicates that
SMART error logging is not supported, nor is SMART self-test, so I
can't see why the drive would be accessing the platters. Wouldn't the
SMART data be written to RAM or flash? In any case it seems to me that
the Seek Error Rate parameter is more like a count than a rate.

I have also verified that the Seek Error Rate figure increases by 10
in real DOS mode when using SmartUDM. Smartdrv is caching the disc
when I do this.

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

I can't get Smartctl to run on my Win98SE box. Everest Home Edition
(Windows) has no problem, nor does SmartUDM (DOS). SMART is enabled in
the BIOS setup.

=====================================================================
smartctl -s on -a hda
smartctl version 5.37 [i686-mingw32-98] Copyright (C) 2002-6 Bruce
Allen
Home page is http://smartmontools.sourceforge.net/

=== START OF INFORMATION SECTION ===
Model Family: Seagate U8 family
Device Model: ST313021A
Serial Number: 6CT0C4JE
Firmware Version: 3.03
User Capacity: 13,022,324,736 bytes
Device is: In smartctl database [for details use: -P show]
ATA Version is: 5
ATA Standard is: Exact ATA specification draft version not indicated
Local Time is: Tue Sep 04 13:50:44 2007 AUSEST
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

=== START OF ENABLE/DISABLE COMMANDS SECTION ===
SMART Enabled.
SMART Disabled. Use option -s with argument 'on' to enable it.
Press any key to continue . . .

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

smartctl -c hda
smartctl version 5.37 [i686-mingw32-98] Copyright (C) 2002-6 Bruce
Allen
Home page is http://smartmontools.sourceforge.net/

SMART Disabled. Use option -s with argument 'on' to enable it.
Press any key to continue . . .
=====================================================================

- Franc Zabkar

 

[Christian Franke]

OK, but why does the Seek Error Rate increase by 10 points when this
happens, if indeed that is what is happening? SmartUDM indicates that
SMART error logging is not supported, nor is SMART self-test, so I
can't see why the drive would be accessing the platters.

"Offline Data Collection" is different from self-test and should be
supported by a drive with "offline" attributes.

Wouldn't the SMART data be written to RAM or flash?

Probably to disk platter itself;-)

In any case it seems to me that
the Seek Error Rate parameter is more like a count than a rate.

Sounds reasonable.

...
I can't get Smartctl to run on my Win98SE box. Everest Home Edition
(Windows) has no problem, nor does SmartUDM (DOS). SMART is enabled in
the BIOS setup.

=====================================================================
smartctl -s on -a hda
smartctl version 5.37 [i686-mingw32-98] Copyright (C) 2002-6 Bruce
Allen
Home page is http://smartmontools.sourceforge.net/

=== START OF INFORMATION SECTION ===
Model Family: Seagate U8 family
Device Model: ST313021A
Serial Number: 6CT0C4JE
Firmware Version: 3.03
User Capacity: 13,022,324,736 bytes
Device is: In smartctl database [for details use: -P show]
ATA Version is: 5
ATA Standard is: Exact ATA specification draft version not indicated
Local Time is: Tue Sep 04 13:50:44 2007 AUSEST
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

=== START OF ENABLE/DISABLE COMMANDS SECTION ===
SMART Enabled.
SMART Disabled. Use option -s with argument 'on' to enable it.

Driver supports SMART, IDENTIFY works and reports SMART enabled, but
apparently SMART STATUS command returns an unexpected error.
The command "smartctl -r ioctl,2 -a hda" prints debug output including
I/O control parameters and return values.

Christian

 

[Folkert Rienstra]

Christian Franke wrote in news:[email protected]

OK, but why does the Seek Error Rate increase by 10 points when this
happens, if indeed that is what is happening? SmartUDM indicates that
SMART error logging is not supported, nor is SMART self-test, so I
can't see why the drive would be accessing the platters.

"Offline Data Collection" is different from self-test and should be
supported by a drive with "offline" attributes.

Wouldn't the SMART data be written to RAM or flash?

Probably to disk platter itself;-)

In any case it seems to me that
the Seek Error Rate parameter is more like a count than a rate.

Sounds reasonable.

...
I can't get Smartctl to run on my Win98SE box. Everest Home Edition
(Windows) has no problem, nor does SmartUDM (DOS). SMART is enabled in
the BIOS setup.

=====================================================================
smartctl -s on -a hda
smartctl version 5.37 [i686-mingw32-98] Copyright (C) 2002-6 Bruce
Allen
Home page is http://smartmontools.sourceforge.net/

=== START OF INFORMATION SECTION ===
Model Family: Seagate U8 family
Device Model: ST313021A
Serial Number: 6CT0C4JE
Firmware Version: 3.03
User Capacity: 13,022,324,736 bytes
Device is: In smartctl database [for details use: -P show]
ATA Version is: 5
ATA Standard is: Exact ATA specification draft version not indicated
Local Time is: Tue Sep 04 13:50:44 2007 AUSEST
SMART support is: Available - device has SMART capability.
SMART support is: Enabled

=== START OF ENABLE/DISABLE COMMANDS SECTION ===
SMART Enabled.
SMART Disabled. Use option -s with argument 'on' to enable it.

 

[Franc Zabkar]

"Offline Data Collection" is different from self-test and should be
supported by a drive with "offline" attributes.



Probably to disk platter itself;-)

I'm assuming that the Identify Drive command retrieves data from solid
state memory and does not entail any seeking. If true, then an
Identify command should not cause any seek statistic to change.

To test this, I dropped back to real DOS, enabled Smartdrv disc
caching, and executed the following sequence of commands:

find-ata p m
smartudm
smartudm
find-ata p m
smartudm

Find-ATA is an old Seagate utility:
http://www.users.on.net/~fzabkar/IDE-identify/FIND-ATA.EXE

The result of the above test was an increase in the Seek Error Rate by
10 points.

I then executed the following:

smartudm
find-ata p m
find-ata p m
...
find-ata p m
smartudm

After 10 repetitions of the Identify command, the Seek Error Rate
still only increased by 10. This would tend to support the idea that
the Seek Error Rate is a genuine seek statistic of some kind, and that
SMART data may be stored on the platters. Having said that, I recall
reading recently that a firmware upgrade clears the SMART data, at
least on some brands of HD. This would suggest that such data are
stored in flash.

Sounds reasonable.



Driver supports SMART, IDENTIFY works and reports SMART enabled, but
apparently SMART STATUS command returns an unexpected error.
The command "smartctl -r ioctl,2 -a hda" prints debug output including
I/O control parameters and return values.

Christian

Here is the output:
http://www.users.on.net/~fzabkar/Smartctl/smart_debug.txt

- Franc Zabkar

 

[maxmillan]

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

- Franc Zabkar
--

Rienstra is just some two-bit dutch patent officer who used to work in
this field who no longer works in this field.

Not that the native dutch were famous for this sort of thing anyway.

 

[Franc Zabkar]

Franc Zabkar wrote:

Probably to disk platter itself;-)

Running SmartUDM on my 120GB Seagate HD causes its Seek Error Rate
statistic to increase by 8 rather than 10. The 120GB drive supports
error logging and self test whereas the 13GB drive does not. This
would suggest that it is the drive's internal activity that is causing
this figure to rise rather than the software.

As my 120GB drive is relatively new, and has no reallocated sectors,
it seems to me that the Seek Error Rate:

(1) is not a rate but a count.
(2) does not count errors.
(3) may represent a cumulative seek count.

If (3), then ...

SER / Power On Seconds = 238390167 / 9981 / 3600
= 6.6 seeks / second

If the average seek time is ~10ms, then the drive would be capable of
100 seeks/sec. Does this result support the idea that SER is a
cumulative seek count? I'm not sure, but in any case, even if it does,
it shouldn't really play a part in determining the error rate. AFAICS,
one should be measuring the rolling average seek error rate, ie the
number of errors in the last 100 or 1000 seeks.

As for why SMART diagnostic I/O results in 8 or 10 seeks, if indeed it
does, I'm guessing that SMART data may be written to 8 or 10 redundant
cylinders, probably at the outer, middle, and inner areas of the disc
???

- Franc Zabkar

 

[Franc Zabkar]

Rienstra is just some two-bit dutch patent officer who used to work in
this field who no longer works in this field.

A desk jockey. I could have guessed.

- Franc Zabkar

 

[Folkert Rienstra]

Franc Zabkar wrote in news:[email protected]

Running SmartUDM on my 120GB Seagate HD causes its Seek Error Rate
statistic to increase by 8 rather than 10. The 120GB drive supports
error logging and self test whereas the 13GB drive does not. This
would suggest that it is the drive's internal activity that is causing
this figure to rise rather than the software.

Congratulations. You finally figured out what "SMART diagnostic I/O" is.
And you know what, it only took you 3 days to do it.
Must be a new australian record.

[rest of bullshit flushed]

 

[Christian Franke]

Here is the output:
http://www.users.on.net/~fzabkar/Smartctl/smart_debug.txt

Thanks for the info.
SMART STATUS returns function not implemented. Probably a driver issue.
smartctl does not continue if this function fails.

Christian

 

[Franc Zabkar]

Thanks for the info.
SMART STATUS returns function not implemented.

According to section 3.1.4 of Seagate's U8 Product Manual, the
"S.M.A.R.T. Return Status" command is implemented for this drive:

http://www.seagate.com/support/disc/manuals/ata/u8pmb.pdf

Probably a driver issue.

If I have read the SMART specs correctly, the Return Status command is
a mandatory component of the SMART feature set. Therefore wouldn't it
be odd if MS omitted this function from its smartvsd driver?

smartctl does not continue if this function fails.

Christian

Could you please indulge me by helping me understand the error report?

I believe the Input line defines the command, features register,
sector number, sector count, cyl low, cyl high for the Return Status
subcommand (0xDA).

But I don't understand the "errno=40". Does this reflect the Status or
Error registers, or is it internal to the program? AIUI, the Status
register should be set to 0x40 if there is no error, otherwise the
Error register should be set to 0x04 "if the device does not support
this command, or if SMART is disabled".

===================================================================
REPORT-IOCTL: DeviceFD=0 Command=SMART STATUS
SMART_SEND_DRIVE_COMMAND failed, Error=1
Input : CMD=0xb0, FR=0xda, SC=0x00, NS=0x00, CL=0x4f, CH=0xc2,
SEL=0x00
REPORT-IOCTL: DeviceFD=0 Command=SMART STATUS returned -1 errno=40
[Function not implemented]
===================================================================

- Franc Zabkar

 

[Christian Franke]

According to section 3.1.4 of Seagate's U8 Product Manual, the
"S.M.A.R.T. Return Status" command is implemented for this drive:

http://www.seagate.com/support/disc/manuals/ata/u8pmb.pdf


If I have read the SMART specs correctly, the Return Status command is
a mandatory component of the SMART feature set. Therefore wouldn't it
be odd if MS omitted this function from its smartvsd driver?

It is implemented in smartvsd.vxd and worked at least on Win98SE with
esdi_506.pdr and smartvsd.vxd from 1999-04-23, MD5:
7e98fae3d43769bcc0ecc0cdc93b7fdd esdi_506.pdr
7809459e198163c9f5b65a728367a65d smartvsd.vxd

Possible issues:

- smartvsd.vxd detects the drive as not fully supporting SMART. AFIAK,
ERROR_INVALID_FUNCTION (1) is returned for all functions except IDENTIFY
regardless whether the real ATA function actually works.

- A third party ATA driver is used.

- There is a bug in the Win9X code of recent smartmontools releases. The
very first smartctl windows release 5.29 from 2004 was actually
developed on 98SE, but I no longer do any regression testing on 9x/ME.
(All releases are still available at sourceforge. If an older release
works, please tell me!)

- The Win98SE has been upgraded from a previous Windows release. It is a
known issue that Windows setup does not reliably install/upgrade
smartvxd.vxd. See smartmontools INSTALL file for details.

...
Could you please indulge me by helping me understand the error report?

I believe the Input line defines the command, features register,
sector number, sector count, cyl low, cyl high for the Return Status
subcommand (0xDA).

Yes.


But I don't understand the "errno=40". Does this reflect the Status or
Error registers, or is it internal to the program? AIUI, the Status
register should be set to 0x40 if there is no error, otherwise the
Error register should be set to 0x04 "if the device does not support
this command, or if SMART is disabled".

===================================================================
REPORT-IOCTL: DeviceFD=0 Command=SMART STATUS
SMART_SEND_DRIVE_COMMAND failed, Error=1
Input : CMD=0xb0, FR=0xda, SC=0x00, NS=0x00, CL=0x4f, CH=0xc2,
SEL=0x00
REPORT-IOCTL: DeviceFD=0 Command=SMART STATUS returned -1 errno=40
[Function not implemented]
===================================================================

The DeviceIoControl() call returns Error 1 (GetLastError() ==
ERROR_INVALID_FUNCTION). It does not return any ATA registers in this
case. There is no evidence whether the actual ATA command was executed
or not.

Smartmontools is a portable app (8 Platforms) with Linux roots,
therefore the error code is mapped to more general (C89, POSIX) form
errno = ENOSYS (40).

Christian

 

[Franc Zabkar]

Franc Zabkar wrote:
It is implemented in smartvsd.vxd and worked at least on Win98SE with
esdi_506.pdr and smartvsd.vxd from 1999-04-23, MD5:
7e98fae3d43769bcc0ecc0cdc93b7fdd esdi_506.pdr
7809459e198163c9f5b65a728367a65d smartvsd.vxd

I'm using the 1999-04-23 versions of both drivers.

Possible issues:

- smartvsd.vxd detects the drive as not fully supporting SMART. AFIAK,
ERROR_INVALID_FUNCTION (1) is returned for all functions except IDENTIFY
regardless whether the real ATA function actually works.

- A third party ATA driver is used.
No.

- There is a bug in the Win9X code of recent smartmontools releases. The
very first smartctl windows release 5.29 from 2004 was actually
developed on 98SE, but I no longer do any regression testing on 9x/ME.
(All releases are still available at sourceforge. If an older release
works, please tell me!)

Version 5.37 appears to have a bug.

Version 5.30 works OK, although I thought I was downloading version
5.29:
http://sourceforge.net/project/showfiles.php?group_id=64297&package_id=61413&release_id=219247

Here is the output:
http://www.users.on.net/~fzabkar/Smartctl/13gb.log

Thanks for your help.

- Franc Zabkar

 

[Christian Franke]

On Sat, 08 Sep 2007 18:41:25 +0200, Christian Franke


Version 5.37 appears to have a bug.

Version 5.30 works OK, although I thought I was downloading version
5.29:
http://sourceforge.net/project/showfiles.php?group_id=64297&package_id=61413&release_id=219247

Here is the output:
http://www.users.on.net/~fzabkar/Smartctl/13gb.log

There is actually a regression that was introduced when 3ware RAID
support was added. I will fix this.

Release 5.36 and the 5.37 test release 5.37-0-20060720 should work on
9x/ME. See http://smartmontools.sourceforge.net/#WindowsInstall for a
download link for test releases.

Thanks for your help.

Thanks for the bug report.

Christian

 

[Franc Zabkar]

There is actually a regression that was introduced when 3ware RAID
support was added. I will fix this.

Release 5.36 and the 5.37 test release 5.37-0-20060720 should work on
9x/ME. See http://smartmontools.sourceforge.net/#WindowsInstall for a
download link for test releases.

Yes, 5.37.0 works.

- Franc Zabkar