Tip: chkdsk will randomly takes forever

Hi All,

Windows 7 Pro x64 SP1

I got a weird one. This customers chkdsk in
her backup routine (I wrote this part) will
occasionally take forever to complete. I
found it after repeating the command about
12 times (on the same drive).

So I ran gsmartcontrol on the drive and her drive is
demonstrably bad. Interesting way of Windows showing
it! Usually something will show up in the Windows
System Log.

I never get the easy ones! :'(

-T



chkdsk b: /x


The type of the file system is NTFS.
Volume label is BACKUP.

CHKDSK is verifying files (stage 1 of 3)...
175616 file records processed.
File verification completed.
0 large file records processed.
0 bad file records processed.
0 EA records processed.
0 reparse records processed.
CHKDSK is verifying indexes (stage 2 of 3)...
175648 index entries processed.
Index verification completed.
0 unindexed files scanned.
0 unindexed files recovered.
CHKDSK is verifying security descriptors (stage 3 of 3)...
175616 file SDs/SIDs processed.
Security descriptor verification completed.
16 data files processed.
Windows has checked the file system and found no problems.

381542399 KB total disk space.
24128 KB in 19 files.
20 KB in 18 indexes.
0 KB in bad sectors.
253227 KB in use by the system.
65536 KB occupied by the log file.
381265024 KB available on disk.

4096 bytes in each allocation unit.
95385599 total allocation units on disk.
95316256 allocation units available on disk.

T wrote:

gsmartcontrol

That's one of many tools that will read the S.M.A.R.T. data from a drive
and look at some important attributes to determine drive health.
Current I leave HD Sentinal running to monitor my drives. I know that I
didn't pay for it but somehow I got a license key from them to register
their Standard version.

I don't recall that chkdsk ever interrogated the S.M.A.R.T attributes in
any version of Windows. That's like relying on MS Paint to draw
something like the Mona Lisa.

T wrote:
Hi All,

Windows 7 Pro x64 SP1

I got a weird one. This customers chkdsk in
her backup routine (I wrote this part) will
occasionally take forever to complete. I
found it after repeating the command about
12 times (on the same drive).

So I ran gsmartcontrol on the drive and her drive is
demonstrably bad. Interesting way of Windows showing
it! Usually something will show up in the Windows
System Log.

I never get the easy ones! :'(

-T



chkdsk b: /x


The type of the file system is NTFS.
Volume label is BACKUP.

CHKDSK is verifying files (stage 1 of 3)...
175616 file records processed.
File verification completed.
0 large file records processed.
0 bad file records processed.
0 EA records processed.
0 reparse records processed.
CHKDSK is verifying indexes (stage 2 of 3)...
175648 index entries processed.
Index verification completed.
0 unindexed files scanned.
0 unindexed files recovered.
CHKDSK is verifying security descriptors (stage 3 of 3)...
175616 file SDs/SIDs processed.
Security descriptor verification completed.
16 data files processed.
Windows has checked the file system and found no problems.

381542399 KB total disk space.
24128 KB in 19 files.
20 KB in 18 indexes.
0 KB in bad sectors.
253227 KB in use by the system.
65536 KB occupied by the log file.
381265024 KB available on disk.

4096 bytes in each allocation unit.
95385599 total allocation units on disk.
95316256 allocation units available on disk.

You should incorporate some basic SMART logic into
your code.

SMART is not perfect, in that if all the disk errors
are concentrated in one zone, SMART tends to miss it.
You can have "Reallocated=0" and still have lousy
performance in the OS area of the disk. I had a drive
do that, and it was promptly retired - not because of
Reallocations, but because the HDTune read transfer curve
was bad *only* in the OS partition area. Sorta like a
wear pattern.

Only a read benchmark can cover that situation, and I'm
not aware of any easy (AI) way to analyze whether
a drive is going bad by using it. When there are dips
in a hard drive transfer curve, that's a general bad
sign. In the case of my hard drive though, it would
have completed CHKDSK, but the time would likely be
double or triple the normal time. It would have
completed, because at that point, the disk still
had data integrity.

In theory, adding a "timer" or "watchdog" to your code,
to verify that CHKDSK "took less than an hour" would
be an idea. Is it safe to terminate CHKDSK ? Maybe
in read-only mode, definitely not if using /F .
Does CHKDSK always listen to a terminate request ?
I can imagine that would be hard to verify if
CHKDSK was in the middle of a kernel call,
while it waited 15 seconds for a bad sector to
come back. You'd need a bad disk to test with, to have
some assurance all your fancy code handled this well.

Paul

Paul wrote on 2/17/2018 6:26 AM:
T wrote:
Hi All,

Windows 7 Pro x64 SP1

I got a weird one.Â* This customers chkdsk in
her backup routine (I wrote this part) will
occasionally take forever to complete.Â* I
found it after repeating the command about
12 times (on the same drive).

So I ran gsmartcontrol on the drive and her drive is
demonstrably bad.Â* Interesting way of Windows showing
it!Â* Usually something will show up in the Windows
System Log.

I never get the easy ones!Â* :'(

-T



chkdsk b: /x


The type of the file system is NTFS.
Volume label is BACKUP.

CHKDSK is verifying files (stage 1 of 3)...
Â* 175616 file records processed.
File verification completed.
Â* 0 large file records processed.
Â* 0 bad file records processed.
Â* 0 EA records processed.
Â* 0 reparse records processed.
CHKDSK is verifying indexes (stage 2 of 3)...
Â* 175648 index entries processed.
Index verification completed.
Â* 0 unindexed files scanned.
Â* 0 unindexed files recovered.
CHKDSK is verifying security descriptors (stage 3 of 3)...
Â* 175616 file SDs/SIDs processed.
Security descriptor verification completed.
Â* 16 data files processed.
Windows has checked the file system and found no problems.

Â*381542399 KB total disk space.
Â*Â*Â*Â* 24128 KB in 19 files.
Â*Â*Â*Â*Â*Â*Â* 20 KB in 18 indexes.
Â*Â*Â*Â*Â*Â*Â*Â* 0 KB in bad sectors.
Â*Â*Â* 253227 KB in use by the system.
Â*Â*Â*Â* 65536 KB occupied by the log file.
Â*381265024 KB available on disk.

Â*Â*Â*Â*Â* 4096 bytes in each allocation unit.
Â* 95385599 total allocation units on disk.
Â* 95316256 allocation units available on disk.

You should incorporate some basic SMART logic into
your code.

SMART is not perfect,
.... REST DELETED .....

According to some things I have read SMART not only is not perfect but
the meaning of the fields and values are not standardized. You may or
may not be able to guess meaning from the surface appearance of a field
name. I do believe it is possible to tell if a value is out of bounds
though. SMART seems to be only a grudgingly accepted Standard. It seemed
to be a good idea at the time.
--
Jeff Barnett

Paul wrote:

T wrote:

Windows 7 Pro x64 SP1

I got a weird one. This customers chkdsk in her backup routine (I
wrote this part) will occasionally take forever to complete. I
found it after repeating the command about 12 times (on the same
drive).

So I ran gsmartcontrol on the drive and her drive is demonstrably
bad. Interesting way of Windows showing it! Usually something will
show up in the Windows System Log.

SMART is not perfect, in that if all the disk errors are concentrated
in one zone, SMART tends to miss it. You can have "Reallocated=0" and
still have lousy performance in the OS area of the disk. I had a
drive do that, and it was promptly retired - not because of
Reallocations, but because the HDTune read transfer curve was bad
*only* in the OS partition area. Sorta like a wear pattern.

Only a read benchmark can cover that situation, and I'm not aware of
any easy (AI) way to analyze whether a drive is going bad by using
it. When there are dips in a hard drive transfer curve, that's a
general bad sign. In the case of my hard drive though, it would have
completed CHKDSK, but the time would likely be double or triple the
normal time. It would have completed, because at that point, the disk
still had data integrity.

That level of testing sounds like chores for specialty disk testing
tools, like SpinRite, HD Sentinel, or HDDRegenerator. Gibson hasn't
updated SpinRite in many years to accomodate newer hardware. It was
king back in its hayday. It's most recent version 6.0 was released back
in 2004. It forté is as an HDD disk surface scanner so it doesn't apply
to SSDs. One feature that I've liked was its refresh (aka reinitialize)
function. All magnetic media suffers dipole stress which causes the
dipoles to unalign over time weakening their differentiation. Most of
the clusters allocated to the OS never change. SpinRite's refresh will
make sure the data is read correctly, move it to another cluster that
was previously tested, and update the file table. That way the data was
read AND WRITTEN to ensure the dipoles were at their strongest
orientation. Defragmentation will not move clusters that are already
contiguous so there is no write of the old cluster.

SMART isn't that great, either, if only looking at a one-time snapshot
of its current status. Monitoring it over time to log fluctuations give
a better indicator of disk health. Backblaze has shown that monitoring
only on the current SMART attributes won't indicate when a disk is going
bad. Plus SMART isn't really a good predictor. You could have a disk
suddenly go bad while an individual snapshot the current state of SMART
attributes showed no problems. For example, a high pending reallocate
isn't bad if it eventually returns to zero.

I'm currently using HDD Sentinel (Standard edition). I don't have to
keep revisiting it to test the current health of the drive (HDD or SDD).
It runs in the background (8 MB working memory) and will alert me of a
drive that exhibits artifacts indicating the drive is going bad. Plus
it adds an overlay to the drive icons in Windows Explorer to show a
checkmark atop a green circle to indicate a drive is okay. I haven't
had a bad disk yet since I usually up the capacity with a new drive
before the old drive goes bad so I don't know what the overlay icon
looks like for a bad drive. Probably an X atop a red circle. However,
disk surface testing is not part of the Standard (or Trial) edition of
HDD Sentinel. You have to buy their Pro version for that. It can do a
read-only test but that is of limited value. The OS will retry a failed
read along with the drive's firmware retrying a failed read. A long
time ago for HDDs, I found the OS retried 3 times and the firmware
retried 5 times so a failed read was after 15 retries. If just one of
them succeeded, the read operation was passed as successful. The
Standard edition can do a write test but that is destructive. The
Standard edition costs $20 (I got it free but don't remember how, maybe
using TrialPay if they offered it back then). Pro costs $30 -- but that
is a lot cheaper than SpinRite's $89 price tag. I use the default level
of diagnosis but you could select a strict level; however, that's really
if you operate a server since HDD Sentinel will warn earlier about an
iffy disk which could reduce the longevity of a drive (i.e., you're told
earlier the drive should be replaced). Although HDD Sentinel adds their
own algorithms for drive health along with logging the drive's status
over time, it uses SMART but weights the attributes. Strict mode merely
changes the weighting.

_Default (light weighting)_
SMART attribute Weight Limit %
Reallocated sectors count 1 70
Seek error rate 0.5 20
Spin retry count 3 60
Reallocation event count 0.6 30
Current pending sectors count 0.6 48
Off line uncorrectable sectors count 1 70

_Strict (heavy weighting)_
SMART attribute Weight Limit %
Reallocated sectors count 6 60
Seek error rate 2 20
Spin retry count 6 40
Reallocation event count 4 30
Current pending sectors count 4 30
Off line uncorrectable sectors count 6 60

HDD Regenerator strikes me as a direct SpinRite alternative both mostly
for disk surface testing and repair. I remember looking at it but
decided to go with HDD Sentinel instead. HDD Regenerator prices itself
($80) at almost the same price as SpinRite ($89). HDD Sentinel Pro is
$30 so if I want more then I'll upgrade from Standard to Pro. Because
my drives get replaced with bigger ones in about 4 years, I'm getting
rid of them before they start going bad. If my builds never changed
then I'd get the Pro edition. If I was a tech for a company then I'd
probably have both HDD Sentinel Pro (monitoring) and Spinrite (one-time
surface testing & repair) in my software toolbox. HDD Regenerator
doesn't explain a lot about their product, like they want to present it
as voodoo magic.

https://www.disktuna.com/looking-for...e-alternative/

Jeff Barnett wrote:
Paul wrote on 2/17/2018 6:26 AM:
T wrote:
Hi All,

Windows 7 Pro x64 SP1

I got a weird one. This customers chkdsk in
her backup routine (I wrote this part) will
occasionally take forever to complete. I
found it after repeating the command about
12 times (on the same drive).

So I ran gsmartcontrol on the drive and her drive is
demonstrably bad. Interesting way of Windows showing
it! Usually something will show up in the Windows
System Log.

I never get the easy ones! :'(

-T



chkdsk b: /x


The type of the file system is NTFS.
Volume label is BACKUP.

CHKDSK is verifying files (stage 1 of 3)...
175616 file records processed.
File verification completed.
0 large file records processed.
0 bad file records processed.
0 EA records processed.
0 reparse records processed.
CHKDSK is verifying indexes (stage 2 of 3)...
175648 index entries processed.
Index verification completed.
0 unindexed files scanned.
0 unindexed files recovered.
CHKDSK is verifying security descriptors (stage 3 of 3)...
175616 file SDs/SIDs processed.
Security descriptor verification completed.
16 data files processed.
Windows has checked the file system and found no problems.

381542399 KB total disk space.
24128 KB in 19 files.
20 KB in 18 indexes.
0 KB in bad sectors.
253227 KB in use by the system.
65536 KB occupied by the log file.
381265024 KB available on disk.

4096 bytes in each allocation unit.
95385599 total allocation units on disk.
95316256 allocation units available on disk.

You should incorporate some basic SMART logic into
your code.

SMART is not perfect,
... REST DELETED .....

According to some things I have read SMART not only is not perfect but
the meaning of the fields and values are not standardized. You may or
may not be able to guess meaning from the surface appearance of a field
name. I do believe it is possible to tell if a value is out of bounds
though. SMART seems to be only a grudgingly accepted Standard. It seemed
to be a good idea at the time.

The "Reallocated" field is the one I trust.

Even though it is thresholded, and "zero doesn't mean zero".

For example, it could read 0..5000 when the actual absolute
error count inside is 100000..105000. It ignores the first
100000 errors, or whatever the "threshold" used at the factory
for acceptance is. The factory allows drives with errors to
leave the factory, and the number of allowed errors is large.

By not giving an honest output, it prevents retail customers
from "returning drives until they get a good one". All the
drives read zero when you get them, so you cannot cherry pick
drives.

Paul

On 02/17/2018 03:33 AM, VanguardLH wrote:
T wrote:

gsmartcontrol

That's one of many tools that will read the S.M.A.R.T. data from a drive
and look at some important attributes to determine drive health.
Current I leave HD Sentinal running to monitor my drives. I know that I
didn't pay for it but somehow I got a license key from them to register
their Standard version.

I don't recall that chkdsk ever interrogated the S.M.A.R.T attributes in
any version of Windows. That's like relying on MS Paint to draw
something like the Mona Lisa.


something wrong on the hard drive making is chkdsk drag.
I have a new drive on order for the customer

On 02/17/2018 05:26 AM, Paul wrote:
You should incorporate some basic SMART logic into
your code.

gsmartcontrol also has a couple of tests. The quick test failed.

On 02/17/2018 11:06 AM, VanguardLH wrote:
That level of testing sounds like chores for specialty disk testing
tools, like SpinRite, HD Sentinel, or HDDRegenerator.

gsmartcontrol has a couple of test in it too. I run
those as well.

Once I find something wrong with a hard drive, I immediately
replace it. Forewarned is forearmed. I am not willing
to risk my or my customer's information to a "repaired" drive
that may fail again. Plus the labor involved in doing
these kinds of repairs is more expensive that replacing the
buggers.