hard disc health checks
Ian Malone
ibm21 at cam.ac.uk
Wed Oct 26 16:25:21 UTC 2005
Bruno Wolff III wrote:
> On Wed, Oct 26, 2005 at 13:33:01 +0100,
> Ian Malone <ibm21 at cam.ac.uk> wrote:
>
>>Has anyone any suggestions for checking the physical
>>health of a hard disc? Something analogous to the
>>"surface scan" in Windows Scandisk?
>
>
> You should probably have smartd making regular checks of your disks.
Worth doing, I'll look into it.
>
>
>>I've got a disc which has some sectors marked as bad
>>by Windows scandisk, and has had them for quite a
>>while (for no obvious reason; I think scandisk messed
>>up at some point). smartctl for the drive is fine.
>
>
> They may have been bad at one time and then reallocated by the disk. If a
> disk decides to reallocate a sector it needs to wait until it either gets
> a good read or for the sector to be rewritten. Once reallocation is done,
> the sectir that previously looked bad will be fine.
>
They got marked bad fairly early in its life and (as you
refer to below), there are no reallocated blocks. Coupled
with the fact that my Windows install isn't 100% stable
it makes me suspect they weren't bad in the first place.
But they also make me a bit cautious as to whether
something funny is happening.
>
>>Recently though I've suffered some filesystem
>>problems. This could be a filesystem error, due to
>>a scandisk crash which occured while checking or a
>>bug in the vfat module (since the first I noticed it
>>was after unzipping a file on the disc using linux
>>and then rebooting into Windows).
>>
>>I'm currently taking steps to back up the data (and
>>recover some of the lost stuff), but I'd like to
>>have something solid to check whether I can trust it
>>in future, any thoughts?
>
>
> You can look at the smartctl output to see how many reallocated sectors there
> are and if any of the other stats are below the fail threshold or if the
> disk itself says it is about to fail.
All are above fail threshold, in fact they all look
a little too good for a 2+ year old disc.
>
> If you get the data all copied off, then you might try running badblocks
> with one of the write options and have it make several passes over the disk.
> Then run a long smart self test and see how things look.
badblocks sounds like what I'm looking for, thanks. I
did an extended self test yesterday with no complaints.
--
imalone
More information about the fedora-list
mailing list