On Mon, 24 Jul 2000, Willy Tarreau wrote:
> Hi all,
>
> this night, my NFS server partially stopped working. One of its soft raid5
> arrays gave many error messages when trying to access it and all involved
> processes stayed in D state. So I rebooted the server and the raid array
> refused to be set up because I have 3 different time stamps on my 3 disks !!!
>
> I don't know how this is possible, and I don't know what to do. I've tried
> to manually change a timestamp, but the raid superblock is after 2G and my
> dd can't seek to it (and I can't remember where I found one which uses
> llseek). So before destroying it totally, I'd like to know if there's a
> good method to recover from sort of these problems, because if not, I wonder
> if finally raid5 is such solid :-/
>
> for info, I'm using a 2.2.17pre11 kernel with ingo's raid-2.2.16, reiserfs
> 3.5.23 and other stuff so I admit this can come from my kernel, but I still
> don't understand how !
First you need to find out whether a drive has indeed failed, or if you were
seeing spurious controller/driver problems.
Try reading from each drive, eg. cat /dev/... > /dev/null. If you can
read all drives completely, they're probably fine. If one has failed, mark
down which one. If two or more drives did in fact fail, well then RAID-5 can't
save you.
If no drives failed: run mkraid -f - it will claim that it's about to
destroy all data, but don't worry, it will only re-write the superblocks and
rebuidl the parity. But this requires that your /etc/raidtab is 100%
consistent with your actual array configration.
If one drive failed: change the raidtab configuration so that the failed drive
is configured as a failed-disk instead of a raid-disk. Then run mkraid -f as
before.
Good luck ! :)
--
................................................................
: jakob@ostenfeld.dtu.dk : And I see the elder races, :
:.........................: putrid forms of man :
: Jakob Østergaard : See him rise and claim the earth, :
: OZ9ABN : his downfall is at hand. :
:.........................:............{Konkhra}...............:
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jul 31 2000 - 21:00:17 EST