GNU ddrescue

I too have problems with drives stopping responding. ddrescue doesn't notice that the drive is not communicating at all, and proceeds to log the rest of it as errors.

There is a SATA disk reset command, it's a variation on:
echo "- - -" > /sys/class/scsi_host/host*/scan
For me, it makes the drive responsive again.

It would be fantastic if ddrescue could detect if a drive has dropped off, and issue the appropriate command to get it back online.

I highly recommend this project. I would love to see a option to recover only used FS blocks or failing that to make ddrescue use a library for it's core functionality and then get freshmeat.net/projects... to use ddrescue library to attractive this. One other comment some disks seem to stop all data transfer and the only way to resume is to reboot the machine and continue from the ddrescue log file, there has to be some disk ata reset command or some electronics one could do to be able to power down/up the hard disk. I can do the electronics side of it.

Hi guys,
I'm the developer of safecopy ( freshmeat.net/projects... ) (which basically is another dd_rescue clone around for a couple of years). Aside from advertising my project on the comment page of the biggest competitor here *sic* I would encourage you guys to have a look on a little debug library (wrapper for glibc) I wrote and packaged with safecopy, which simulates IO-errors on a virtual file.
Unless you already use something like that internally for debugging and testing of ddrescue it might help your development work for benchmarks or creation of a test suite.

Using ddrescue and SpinRite together
I've created a "Quick and Dirty" Perl program which reads a GNU ddrescue 1.0 "logfile" (sector list file), and generates a DOS .bat file of SpinRite 6.0 commands to try to recover the unrecovered regions. After SpinRite has completed its work, you can run ddrescue 1.0 again to add the recovered data to the recovery image. You can get my little Perl program here:
www.burtonsys.com/down... (www.burtonsys.com/down...)

Note #1: This little Perl program can be run under any system that supports Perl 4 or 5 (Linux, DOS, Windows, etc.).
For tiny compatible Perl 4 implementations for DOS and Windows, see:
www.burtonsys.com/auxi... (www.burtonsys.com/auxi...)

Note #2: ddrescue runs under Linux, and SpinRite runs under MS-DOS or FreeDOS. So (obviously) some rebooting will be required.

Note #3: You must use Antonio Diaz's GNU ddrescue 1.0:
freshmeat.net/projects... (freshmeat.net/projects...)
You cannot use ddrescue 0.9 or earlier, nor can you use Garloff's dd_rescue (which is an entirely different program).

Note #4: Gibson Spinrite will set you back U.S. $89. I think it is well worth it (and I have no connection to Gibson, other than as a satisfied SpinRite user). Get it here:
www.grc.com/sr/spinrit... (www.grc.com/sr/spinrit...)

-Dave

Fast, effective, and flexible!
I compared the speed of Antonio's ddrescue 0.9 to the speed of dd_rescue for copying error-free data on a PC with ATA/EIDE drives, and Antonio's ddrescue ran nearly twice as fast as dd_rescue.

Since dying hard disk drives tend to get sicker and sicker as time goes on, I think speed it is important, to improve the success rate of the rescue. ddrescue's better speed could also help with "freezer-based" hard disk drive recovery. Sometimes a dying hard disk drive can be coaxed into working again, briefly, by sealing it in a plastic bag and freezing it in the freezer overnight. (I'm now 1-for-3 recovering hard disk drives with this technique.) But, of course, when you take it out of the freezer and plug it into your computer it starts warming up. You may only have a few minutes to recover your data before the drive warms up and starts failing again. So a faster ddrescue improves your chances of getting all the data off before the drive gets too warm.

The new 1.0pre1 is a gigantic improvement over 0.9! I've not benchmarked it, but I assume that the speed is about the same. However, the new logfile format has many advantages!

The most important advantage is that it enables you to do multiple ddrescue passes to recover your data. For example:

# first, grab most of the error-free areas in a hurry:

ddrescue -B -n /dev/hdd rescued.ima rescued.log

# then try to recover as much of the dicy areas as possible:

ddrescue -B -r 1 /dev/hdd rescued.ima rescued.log

(the "-r 1" is necessary to make it retry the failed areas)

Then put the drive in the freezer overnight (or run SpinRite on it), and then try again:

ddrescue -B -r 1 /dev/hdd rescued.ima rescued.log

(it won't recopy the already copied areas)

You can also run it on just part of the disk, e.g., to try extra hard to recover the most important areas. This example retries extra hard (10x) to recover any bad areas in the first 50 MB of the drive:

ddrescue -B -r 10 -s 52428800 rescued.ima rescued.log

Each time you run it, ddrescue updates the logfile to show which areas of the disk drive or file were successfully copied, which failed to copy, and which are yet to be attempted. It also updates the logfile when you halt it with Ctrl-C, and also saves it every 30 seconds during normal operation (in case of a crash).

When you are done, the logfile also serves as a nice, concise ASCII record of where the good and bad areas are in the copy of the drive or file. I can envision tools and scripts which use that information to identify the damaged files on a rescued disk drive, or which consolidate the good areas of the two copies of the FAT on a FAT16 or FAT32 drive, or which identify the damaged areas that are in "free space" and just mark them as "good" (because their contents don't matter), etc., etc..

ddrescue 1.0pre1 is a wonderful little utility, and (thanks to the simple, well-documented ASCII logfile format) it is nicely extensible.

Great job, Antonio!

-Dave