On Mon, Jun 05, 2006 at 07:20:04PM +0200, Florian Kulzer wrote: > > Modern hard drives have a built-in self-diagnosis system called SMART. > The corresponding Debian package is called "smartmontools". You can > install this package (or boot from a suitable live CD) and run > > smartctl -a /dev/hdb | less > > to retrieve the stored information. You can also run an extended > self-test with > > smartctl --test=long /dev/hdb > > Maybe that will give you an indication about the seriousness of the > problem. > > Another thing to look at is "testdisk", which is a partition scanner and > disk recovery tool. I have never had to use it myself, but I have heard > good things about it. It cannot do anything, of course, if you really > have a permanent hardware failure.
Useful suggestions - thanks. I was hoping someone could elaborate on my very generic suggestions. (a lot of the hardware I deal with is quite ancient or non-standard, so I tend to stick to a lowest common denominator). One frequent mode of failure I have experienced with defective IDE drives is one where a drive will work for a few minutes after power on, and then stop. Another is one where repeated scans of the disk will fail on different sectors. It strikes me that the most effective was of salvaging data from such drives would be a sector by sector copy from a bad partition to a new one on a replacement disk, keeping a separate bitmap recording which sectors could not be read. The bitmap could then be used to repeat the process, skipping all the correctly transferred sectors. This could continue until all sectors are successfully recovered, or some number of sectors are identified as hard failures. I don't think it would be hard to do, but there is no point in reinventing an existing tool. The trickiest thing would be communicating the information about which sectors were invalid to a filesystem recovery program if there are some that could not be recovered. Does anyone know if something like this exists? Or if not, is there filesystem recovery program that has some way of being given a list of bad blocks whose content is know to be wrong so that files in which they appear can be flagged as damaged? Regards, DigbyT -- Digby R. S. Tarvin digbyt(at)digbyt.com http://www.digbyt.com -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
