At 15:48 23-5-2003 +0200, Dariush Pietrzak wrote:
Hello,
Summary: One cannot use the same filesystem filtered through MD code and
without it ( filesystem is on /dev/sdaX, you can't mount it via /dev/md0,
or - filesystem was created on /dev/md0, you can't mount it via /dev/sdaX )
What I think you are describing is what I did myself last week with an
unbootable software raid 1 array.
Something resembling what I did:
System stuck on the boot line with "LI"
sda1 and sdb1 are a raid1 array and the / fs.
Removed sdb1 from the raid config to prevent problems.
Boot the redhat installer in rescue mode.
mount /dev/sda1 under /mnt/sda1
edit /mnt/sda1/etc/lilo.conf (boot=/dev/sda)
chroot /sbin/lilo
This reinstalled lilo with the correct geometry on sda.
umount /mnt/sda1
Reboot the machine.
raidhotadd /dev/md0 /dev/sdb1 to resync the raid 1 array with the first disk.
This works. I have done this a number of times before on IDE raid
configurations where partitiontables mismatches between the disks (even
same) are common.
You have to be very carefull to at least break the mirror and resync it's
content since even mounting the disk once will trigger the log replay and
thus the data between the disks will change.
Linux software raid 1 uses read balancing to give higher read speeds (raid
0 like), however the data on both disks must be the _exact_ same to the
bit. Mounting the disk without the raid layer once will make them different
and havoc can ensue.
Cheers
--
Seth
It might just be your lucky day, if you only knew.
|