Bad Disk block - but where?

On 2019-09-08 04:30, Philip Herlihy wrote:
Be sure to use some utility that can give you an indication of the SMART
status of the disk. Something like Acronis Drive Monitor (free) or Hard
Disk Sentinel (not).


I have to wonder if not-free utilities (or possible lack of good free
ones) are a symptom of Micro-****s "driver signing" policies...


--
(aka 'Bombastic Bob' in case you wondered)

'Feeling with my fingers, and thinking with my brain' - me

'your story is so touching, but it sounds just like a lie'
"Straighten up and fly right"

In article -
september.org, says...
Be sure to use some utility that can give you an indication of the SMART
status of the disk. Something like Acronis Drive Monitor (free) or Hard
Disk Sentinel (not).


I have a program from the manufacturer, SanDisk SSD
Dashboard. It reports all is well. I ran the longer
version of the SMART test.

In article ToidnfDWHaPc6OjAnZ2dnUU7-
, BigBadBob-at-mrp3-dot-
l says...
And then the SMART utility will tell ya if it's a one time fail or if
you're getting a growing cancer of bad blocks,


Been there/done that

WD has a free SSD Dashboard utility that can perform a
number of diagnostics and has several levels of SMART
testing. I did all of them and it reports that all is
well. It also reports that NO blocks have been re-
assigned, even though chkdsk /f/r found 8 of them. Not
sure what to believe at this point. I would have thought
that the manufacturer's diags could be trusted. Who
knows??

Jason wrote:
In article ToidnfDWHaPc6OjAnZ2dnUU7-
, BigBadBob-at-mrp3-dot-
l says...
And then the SMART utility will tell ya if it's a one time fail or if
you're getting a growing cancer of bad blocks,


Been there/done that

WD has a free SSD Dashboard utility that can perform a
number of diagnostics and has several levels of SMART
testing. I did all of them and it reports that all is
well. It also reports that NO blocks have been re-
assigned, even though chkdsk /f/r found 8 of them. Not
sure what to believe at this point. I would have thought
that the manufacturer's diags could be trusted. Who
knows??

SMART "Reallocated" is thresholded. A large number of
reallocations can happen, yet the SMART indicator raw
value is still "0". That's why the Dashboard cannot show it.

The reason for this, is it prevents hard drives from
being "cherry picked" when brand new. If one drive
had 8 errors, another had 12 errors, you'd be mailing
the 12 error drive back to Newegg and asking for a refund.
This would drive retailers nuts. The only drives they could
sell, would be the drives with "zero errors from the factory".
That never happens! No drive leaves the factory with zero
errors. The thresholding prevents you from seeing this.

To stop that, they fudge the SMART table and only show
errors when a lot of them show up. This is the embodiment
of "factory acceptance threshold".

On a hard drive (like, back in the 9GB era), the
drives were of such low quality, it was considered OK
to release a drive with 100,000 errors leaving the factory.
And if you were the manufacturer, you would ensure that
SMART said "0" when queried. If there were 100,000
errors, SMART would read "0". If there were 100,001
errors, SMART would read "1".

*******

Now, the other thing is CHKDSK. If you do a scan for
"bad clusters", it puts the detected CRC errors in
the $BADCLUS list. That's a sparse file which covers
the entire size of the drive or partition, and thus it
can economically "cover up errors" so the clusters
cannot be reused.

Whether that algorithm works properly with SSDs, who
really knows.

There are a couple issues:

1) How does the user know which file or files is
damaged ? Sure, you can look up $BADCLUS, but do
the filenames get changed to reflect "trouble" ?
On our ancient design at work, a file detected
"bad" in this way, the file name would change to
bad.12345.bad or the like. The number indicating what
part of the storage blew up, and the bad string appended
so you could immediately spot trouble. However, you still
don't know what you lost.

Using NFI.exe, if the filename had not been changed, all the
LBAs are listed, so in theory you could eventually
figure out what a CRC error might have collided with.

2) Is the cluster really isolated ? If the file system operation
is done with pointers, then the answer is "probably".
You would not want the $BADCLUS to be attempted with an
actual write of the cluster contents. Because then the block would
just get spared out (for as long as spares are available).

My first impulse would *not* be scanning the disk with CHKDSK.
I'd do a scan with HDTune and see what sector or sectors are
bad. I've used a screen recorder, to "watch" HDTune and make
a log of the entire scan. And that's how I could tell (roughly)
what block of sectors was involved. That gives you a rough
idea which partition is affected.

Another way to analyze the disk without invoking CHKDSK and
messing it up, is to use ddrescue. Sectors throwing CRC errors
will be placed in the log ddrescue keeps, and so you can
scan the disk that way and generate a list. (You could do
the transfer to /dev/null or NUL so that an actual copy
of the disk is not done, and just a scan for errors is
the result.)

The idea is, to use a technique that "doesn't have side effects".

Paul

On 2019-09-09, Paul wrote:
This would drive retailers nuts. The only drives they could
sell, would be the drives with "zero errors from the factory".
That never happens! No drive leaves the factory with zero
errors. The thresholding prevents you from seeing this.

I would think this is common knowledge. Surely you must remember when new
drives had a bad-block label attached to them so one could map those out
from being used.

The only error-free disks I can recall were disk packs used for early
Unix systems, which required error-free media for the system drive. (Those
were very expensive.)

--
-----------------------------------------------------------------------------
Roger Blake (Posts from Google Groups killfiled due to excess spam.)

NSA sedition and treason -- http://www.DeathToNSAthugs.com
Don't talk to cops! -- http://www.DontTalkToCops.com
Badges don't grant extra rights -- http://www.CopBlock.org
-----------------------------------------------------------------------------

Roger Blake wrote:
On 2019-09-09, Paul wrote:
This would drive retailers nuts. The only drives they could
sell, would be the drives with "zero errors from the factory".
That never happens! No drive leaves the factory with zero
errors. The thresholding prevents you from seeing this.

I would think this is common knowledge. Surely you must remember when new
drives had a bad-block label attached to them so one could map those out
from being used.

The only error-free disks I can recall were disk packs used for early
Unix systems, which required error-free media for the system drive. (Those
were very expensive.)


I played with the first SCSI (SASI?) ones when
they came out, and they had the factory and grown
defect tables. For the drives I played with, I printed
off the defects on about three sheets of paper
and taped the paper inside the PC chassis.

That's how many defects were on the first hard
drives we got.

And when (as an experiment) I reset the grown
defects table, then scanned the drive, all the
grown defects came back.

It was much later, I found out we paid something
like $1400 a piece for those. And they were little
better than fancy floppy drives. (The arm position
controlled by a stepper motor, no voice coil.)

And we did have at least two drive failures, early on.
Nobody considered taking them apart with a screwdriver
for a look or anything. We had too much to do, to be
sitting around "pining over split milk". This is why,
back then, I was making two copies of every file :-)
It's not that I didn't trust the technology or anything.
What could go wrong ?

I don't think the vibration specs were too generous back
then. Probably only a couple G's while the drive was
spinning. They didn't know how to make an arm stiff
back then. They do today.

Paul

On 2019-09-09, Paul wrote:
I played with the first SCSI (SASI?) ones when
they came out, and they had the factory and grown
defect tables. For the drives I played with, I printed
off the defects on about three sheets of paper
and taped the paper inside the PC chassis.

Yes, exactly the kind of thing I was talking about. Of course modern
drives manage their media defects internally and you don't see them at
the OS level unless a threshold is crossed. The point being that a certain
level of media defects have always been normal and expected. Since we
no longer get a defect sheet with modern drives that is something today's
consumers are not necessarily aware of.

I currently have a Seagate Barracuda 1.5TB drive that in SMART has shown
1 bad sector (reallocated) for several years now. I don't keep anything
important on it, though it displays no other symptoms. (I've seen other
drives showing only 1 bad sector die grievous and noisy deaths.)

--
-----------------------------------------------------------------------------
Roger Blake (Posts from Google Groups killfiled due to excess spam.)

NSA sedition and treason -- http://www.DeathToNSAthugs.com
Don't talk to cops! -- http://www.DontTalkToCops.com
Badges don't grant extra rights -- http://www.CopBlock.org
-----------------------------------------------------------------------------

In article ,
lid says...
The idea is, to use a technique that "doesn't have side effects".

That's an excellent suggestion. Taken.