Re: strange disk corruption

[Shaul Karl <shaulk@013.net>]

On Thu, Oct 07, 2004 at 07:57:05AM -0700, briand@aracnet.com wrote:
> Whether or not this is related to debian is not clear, but it's possible.
> 
> I've currently set fsck to run pretty much on every other boot.
> 
> And just about every time it runs , it informs me that it fixed file
> system errors and reboots the system.
> 


  Are you shutting down the system properly? I mean, do you use the
shutdown command or one of its derivative? I hope this isn't too dumb.
After all, you could be totally new to the system.


> 
> So this problem has a history.  I had previously been seeing
> corruption in the /dev directory and only the /dev directory.
> Duplicate file names, files which could not be deleted, etc...  The
> only way to fix was to reboot with a rescue disk and wipe the /dev
> directory and start over.  Unfortunately I had to do this several
> times.
> 


  I had such a behavior too a long time ago. In my case this was with a
SCSI disks. I do not know why the problem went away. It could be a 
kernel issue, and I used several kernel versions since then. Maybe the
steps that I took, which are similar to what you have done, managed to
make the problem go away. 


> 
> 1.  the /dev directory seems to be somewhat dynamic, could there be a
>     debian start-up script which is/was somehow corrupting the /dev
>     directory ?


  2.4 had a dynamic device manager. As far as I know it is now obsolete,
and Debian never used it by default. What kernel are you using?


> 
> 2.  Do I have bad blocks on the disk ?  And how would I check this?
>     Again, I've seen no other evidence whatsoever of flaky disk
>     behavior ?


  There is a badblocks utility. I not sure if this is the right name.
Should be close. I believe the modern disks tries to fix those problems
in a transparent manner.


> 
> 3.  Could this be a bug in fsck ?  Why doesn't fsck actually tell me
>     what the errors are !!  It just says "fixed them - rebooting".
>     isn't this a Bad Thing (TM) ?
> 


  This is the default behavior when fsck finds errors in the root fs. I
haven't tried to change that. I guess that there is a way to do it.