[NBLUG/talk] More hard drive problems

[Lincoln Peters]

On Oct 2, 2005, at 8:17 PM, Lincoln Peters wrote:
> When the surface scans finish, I'll review the man pages for  
> reiserfsck and mdadm one more time, and if I don't see any obvious  
> thing to try that I haven't already tried, I'll try the Knoppix CD.

OK, the surface scans finished on all three hard disks.  Just the one  
I mentioned has bad blocks.  Therefore, the RAID-5 array *should* be  
usable, but when I type in:

# mdadm --run /dev/md1

I get the following output (copied by hand; might have typos):

raid5: device hda3 operational as raid disk 0
raid5: device hdg1 operational as raid disk 2
raid5: cannot start dirty degraded array for md1
RAID5 conf printout:
  --- rd:3 wd:2 fd:1
  disk 0, o:1, dev:hda3
  disk 2, o:1, dev:hdg1
raid5: failed to run raid set md1
md: peers->run() failed ...
mdadm: failed to run array /dev/md1: invalid argument


(The damaged disk that would normally appear as raid disk 1 is hde3)

This leads me to doubt that there are any superblock errors on the  
filesystem as I had initially expected (the superblock error was most  
likely generated because the array was inactive).  It looks like  
purely a RAID issue.

I suppose that it would make sense if mdadm was reluctant to start a  
RAID-5 array that was missing one device (since if a second device  
failed, there would be actual data loss), but I thought if I wanted  
to I *could* continue running even if a disk failed!  It's not *that*  
hard to imagine a situation where one might need to start an array  
that's degraded but still functional (I'm in such a situation right  
now!).

If I can't start the array while it's in a degraded state, I suppose  
I could rebuild the partition table on the damaged drive, reattach it  
to the array, and let it rebuild so that I *can* start the array, but  
that option scares me (I'll only try it if reassured by the  
collective wisdom of the group).


This also probably means that the problem isn't anything that would  
be more easily solved from Knoppix.


Lincoln Peters
lmpeters at mac.com