question about fsck failing during startup

[Werner Schram]

scar wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> hi, during a routine check of disks during startup, fsck failed due to
> read failures.  it said to run it manually, so i did.  again came the
> read errors.  i just ignored them all.  i think there were 32 in all and
> then fsck finished and the system booted up fine after that.  i decided
> to then reboot into recovery mode and drop to a root shell, unmount the
> root filesystem, and run 'e2fsck -c' in order to update the bad blocks
> list.  i think it found the same 32 sectors and updated the bad blocks
> file.  the system has been running stable since.
>
> so, my question is: is everything ok?  i realize the hard drive should
> probably be replaced soon, but as long as i have marked the bad blocks
> it should be ok for now?  (btw, it is a 1.8" PATA IDE drive, impossible
> to find, probably $100's to replace, so it probably won't get replaced
> soon).
>
> i guess my question arises out of a misunderstanding of what the routine
> check during startup is actually doing.  does it check the whole surface
> of the drive that the filesystem is allocated to?  or does it just check
> the used area?  if it just checks the used area, then that means those
> read failures occurred in places that had data.  if that is the case,
> then how can i find out what data is no longer accessible?  it must not
> be too important as everything has been stable and i don't notice
> anything missing.
You can check the status of your hard disk with palimpsets, which is 
available under System->administration->Disk Utility. If you click on 
your hard disk there, you can see it's SMART status, which indicates the 
health of your disk. If you click on "more information", you get to see 
the specific attributes. It also gives an indication about the urgency 
of the errors found.

The fsck tool checks your filesystem consistency (do the files actually 
exists? Are the reported filesizes correct?), but with the -c option, it 
calls the badblocks tool to do a surface check on the entire partition 
(not the entire disk though) by trying to read every block and see if 
that works. You can do a read/write test with 'fsck -cc', altough I 
don't think that should be necessary.

Werner