Failing Win XP HDD

[Postal2]

374 errors and errsize 20914 KB.

20914/374~56kB, assumed buffer 64kB and not been decreased. Program is dumb. Data is lost.

This is the cause I wrote my own program for cloning. I used source of "win32diskimager", anybody can try. It's easy.

[PlainName]

PlainName: except that the “smart” cloning software mentioned in this thread so far is failing. Since nobody here knows, what exactly it does and if it’s even capable of working with damaged media, it’s impossible to debug the situation either.

I explained why doing it by partition would likely fail: the boot partition probably wasn't being set active. It's easy to stumble over that because a normal restore to the original disk works because the partition is already set correctly, but to a new disk - which is the case here - the restored partition isn't marked active or bootable so fails.

And, contrary to your assertion, we know exactly what the things do. There are two modes: intelligent (which you presumably refer to as 'magic') and dumb. In the latter case every sector is copied regardless, but that's a waste of space and normally only used if the partition is an unknown format. The 'magic' case just copies sectors that the filesystem is using and is perfectly fine - even if you copied all the unused sectors, the filesystem wouldn't use them anyway so there is no point in doing so.

The only thing we don't know is the compression scheme used to make the copy smaller than the original, for storage purposes.

[m k]

It has now resumed normal speed so if it does not hit any more bad spots I am hoping  it will finish in the next 2 or 3 hours.

If minimum size is smaller than error size/error count, like is an assumption, it goes back to that area.
If and when that happens be brave and stop it.
Save the map file and be ready to go back when needed.

For sector by sector, block by block or something else.
This is rescue, not backup, so least stress for most reads, that can be many things.

For backups 3-2-1 is better if just 3-2, 3 copies and 2 physical locations.
More is also not worse here.

[soldar]

If minimum size is smaller than error size/error count, like is an assumption, it goes back to that area.
If and when that happens be brave and stop it.
Save the map file and be ready to go back when needed.

For sector by sector, block by block or something else.
This is rescue, not backup, so least stress for most reads, that can be many things.

I have no idea what any of this means.

At this point it's been going for 22 hours and it has slowed down again. It done 954 GB and, since the drive is 1000GB, it has about 4.5% to go. But it has slowed down again so it could go on for hours. I will wait and see.

[Postal2]

If minimum size is smaller than error size/error count, ....

Error size is a chunk protected by ECC, it's easy to guess about it size by personal experience.

[soldar]

It's been going for 25 hours now. I just finished Pass 1 (forward) and now it has started Pass 2 (backwards).

I have no idea how long this will take but as I see the count of "rescued" is increasing I will leave it and see how it goes.

I just hope it does not take another 25 hours.

[Monkeh]

374 errors and errsize 20914 KB.

20914/374~56kB, assumed buffer 64kB and not been decreased. Program is dumb. Data is lost.

Program hasn't finished, it has further passes to do which will reduce the loss to the sector level.

Don't run your mouth when you haven't done the work.

[soldar]

OK, program has finished after 26 hours and apparently everything was saved and nothing was lost.

I have booted from the clone and it boots fine. I hope everything works well. The report:

# Rescue Logfile. Created by GNU ddrescue version 1.19
# Command line: ddrescue --force /dev/sda /dev/sdb /home/mint/Desktop/logfile
# Start time:   2024-05-19 16:47:34
# Current time: 2024-05-20 18:35:44
# Finished
# current_pos  current_status
0xDE267BFE00     +
#      pos        size  status
0x00000000  0xE8E0DB6000  +

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

mint@mint ~ $ sudo ddrescue --force /dev/sda /dev/sdb /home/mint/Desktop/logfile
GNU ddrescue 1.19
Press Ctrl-C to interrupt
rescued:     1000 GB,  errsize:       0 B,  current rate:    1031 kB/s
   ipos:   954128 MB,   errors:       0,    average rate:   10767 kB/s
   opos:   954128 MB, run time:    1.07 d,  successful read:       0 s ago
Finished                                     

Thanks to all for your help and encouragement.  If any further problem comes up with this I will post.

I now have a 3TB HDD with on 336 GB OS partition, a 663 storage partition and 2 TB free.

My next project will probably be to move the OS partition to a 500GB SSD, move the storage data to a separate HDD and free this 3TB HDD. But I will do that in the future, when I have recovered from this ordeal.

[Postal2]

it has further passes to do which will reduce the loss to the sector level.

I can do that work with modified win32diskimager in 4 hour or less for 1 TB.

[PlainName]

I have booted from the clone and it boots fine.

Well done. I bet that's a relief

My next project will probably be to move the OS partition to a 500GB SSD

You would probably benefit by using that opportunity to test your backup regime. Usually, putting disaster recovery through its paces isn't too practical, but with a new drive you can do it as many times as it takes to succeed without breaking anything more than temporarily.

[Monkeh]

it has further passes to do which will reduce the loss to the sector level.

I can do that work with modified win32diskimager in 4 hour or less for 1 TB.

Not when the drive is holding you up, you can't. This is clearly an extremely distressed device.

And as you clearly missed it: The 'work' is informing yourself, rather than making guesses as to the function of a program which exists to perform this recovery task, and has done so for nearly 20 years.

[Postal2]

Not when the drive is holding you up, you can't.

I have experience with Samsung HD250HJ in one hour. Fully equal work.

[Monkeh]

Not when the drive is holding you up, you can't.

I have experience with Samsung HD250HJ in one hour. Fully equal work.

And did that drive have as severe a set of surface defects causing it to retry reads for long periods of time? Is it running firmware with identical behaviour?

Your modified tool is surely very nice and capable of bypassing fundamental drive behaviours, but I'll stick with a proven one you apparently couldn't find.

[Postal2]

And did that drive have as severe a set of surface defects causing it to retry reads for long periods of time?

No, disk did not retry because program is not so dumb to send such command. Disk retries himself when lose track, but errors was in data, discovered by ECC.

[Whales]

Congrats soldar   Glad you got a working clone.

[Monkeh]

And did that drive have as severe a set of surface defects causing it to retry reads for long periods of time?

No, disk did not retry because program is not so dumb to send such command. Disk retries himself when lose track, but errors was in data, discovered by ECC.

And how certain are you that the drive in this case actually supports any read without retry command and that they actually behave in the expected manner? They've all been obsolete since the year 2000.

Did you really go to the effort of hacking direct ATA access into a tool to do that?

[soldar]

Congrats soldar   Glad you got a working clone.

It took many tries and a lot of time but I am glad I persevered.

Thanks for your help.

[m k]

Being stubborn is sometimes a virtue.

For the side track,
smartctl scterc
WD TLER

[Postal2]

And how certain are you that the drive in this case actually supports any read without retry command and that they actually behave in the expected manner? They've all been obsolete since the year 2000.
Did you really go to the effort of hacking direct ATA access into a tool to do that?

Of course, I did have a debug session with program and several disks. Disks slows down due intensive ECC calculation, but no sound of mechanical retry. And yes, I tried with shoked disks from notebooks who did retry, but no data has been retrieved from it (from regions with mechanical sound).
I see soldar succeded, I assume first variant with retry by dumb command.
About ATA-commands - I did modify win32diskimager, nothing added, works good. See source.

[golden_labels]

It took many tries and a lot of time but I am glad I persevered.

Congrats!

I’m also glad we persevered in insisting on you using ddrescue.

[nigelwright7557]

I would definitely go for an SSD.
It made a huge difference in speed to my laptop.

[soldar]

I would definitely go for an SSD.
It made a huge difference in speed to my laptop.

Well, that is what i was trying to do initially but it did not work. Finally I managed to clone the entire 1TB HDD to a 3 TB HDD.

Now the problem of sector errors is solved but I still want to do what I was trying to do in the beginning:

I have a 336GB bootable partition with the OS and a 663 GB partition for storage (plus, now, 2 TB free).

I have available a brand new Crucial BX500 480GB SSB. What I would like to do is clone the OS partition to that SSD but it has to be bootable and this is the rub, this is what I could not achieve. It seems for the drive to be bootable I have to clone the entire drive which is precisely what I do not want to do.  I want to have on the SSD only the OS partition. I have managed to copy the partition but then it will not boot.

Now that I have cloned the entire HDD the urgency is gone and I need a rest after this ordeal. I might open a new thread with this issue. Someone mentioned GParted. Maybe I'll have a look at that.

In summary, I want to clone the OS partition from the HDD to the SSD but also the MBR so it will boot.

[PlainName]

What I would like to do is clone the OS partition to that SSD but it has to be bootable and this is the rub, this is what I could not achieve. It seems for the drive to be bootable I have to clone the entire drive which is precisely what I do not want to do.  I want to have on the SSD only the OS partition. I have managed to copy the partition but then it will not boot.

I've spelled out a couple of times how to fix that, and it's expected. There's a simple DOS command which will fix it after copying the partition.

Also as I noted a short while ago, now you have a working system and will soon have a spare disk, why not use that to practice this stuff? You'll learn something useful and it won't break anything.

[Postal2]

.....I have a 336GB bootable partition with the OS and a 663 GB partition for storage (plus, now, 2 TB free).

I have available a brand new Crucial BX500 480GB SSB. What I would like to do is clone the OS partition to that SSD but it has to be bootable and this is the rub, this is what I could not achieve. ........

In summary, I want to clone the OS partition from the HDD to the SSD but also the MBR so it will boot.

No problem, simply start cloning to smaller disk and do to the end of his size. Clone will boot. I do that regularly.

It's taking about 30 min (480GB direct cloning).

[m k]

In summary, I want to clone the OS partition from the HDD to the SSD but also the MBR so it will boot.

If boot partition is from the beginning of the disk you have two possibilities.
One is to clone from the beginning so much that the whole partition is cloned and then clean the rest away.
Other is to first create an empty bootable partition and then clone its content.

ddrescue can do both, cloning from the beginning is probably the easier way.
--size is a self explaining option.

Linux limitations of mangling NTFS partition should be history, but maybe not.
Windows is also picky, but should start its fixes automatically.

Seems that I can't actually remember how XP drops to console mode.
That's possibly needed, being the last resort.

Doing ddrescue with partitions is also very simple, just add those numbers after device names.
Target partition can be bigger, LBA access mode is dynamic, absolute location of blocks is not relevant.
I guess all Linux desktops have some sort of a graphical device manager/viewer with a simple drive list.
You can also create that empty partition using Windows and then change the operating system for cloning.

lshw -class disk -short
lsblk -t