On 03/13/2011 07:46 PM, Tom H wrote:
On Sun, Mar 13, 2011 at 1:33 PM, Urs Beyerle<[email protected]> wrote:
As part of upgrade planning for my main workstation at home (an
updated SL 5.2 system), I booted from the live CD just to see if all
the major devices worked. Everything worked :). Unfortunately, that
act rendered my existing system unbootable.
When I booted from the live CD it found and started my md RAID1
partitions, including my system root partition. Nice, I had all my
data there to play with. I noticed that where I had previously used
/dev/md0, /dev/md1, etc..., the live CD system had created them as
/dev/md122, /dev/md123, and so on, and not in the original order. I
didn't think much of this, figuring my regular system would start them
back in the original configuration, but apparently some metadata
somewhere got changed by the live CD system and now the original
system would not get past switching to the root file system with a 'no
such device' error. I could see, just before going off the screen,
where it had started my root partition as /dev/md125 rather than
/dev/md0.
I tried a couple of things to recover from this: I tried stopping and
reassembling the RAID sets with the desired device names from the live
CD system, and repackaging my initrd with device files for the
/dev/md12* devices, but no luck. At that point I decided that, since I
had already done my backups and was planning on eventually going to
SL6, that I'd just push forward with a fresh install of SL6. It took a
few hours that I hadn't planned on to get everything back up to speed,
and I have one issue yet to work and a couple of minor things to
configure.
I have another retired box with the same RAID setup and it too, got
hosed by the live CD. I'm going to play with this box to see if I can
find a way to recover from this.
Beware.
Thanks for letting us know!
I would be very interesting in a recovery procedure.
I guess that it should be possible to re-assemble the raid device. So in
case /dev/md0 was renamed to /dev/125 and was setup with /dev/sda1
/dev/sdb1, you may get /dev/md0 back, if you do
mdadm --stop /dev/125
mdadm --assemble /dev/md0 /dev/sda1 /dev/sdb1
Are your raid partition of the type 0xfd (Linux RAID autodetect) or 0x83
(Linux)?
If this is a general problem with the LiveCD and software raids, I would
like to fix it on the LiveCD.
Usually md devices are renamed/renumbered when "/etc/mdadm.conf" has
"HOMEHOST<system>" set and the metadata in the superblock has a
different homehost value.
The file /etc/mdadm.conf does not exist on the LiveCD. Therefore the md devices
should not be renamed...?
Bluejay, do you see /etc/mdadm.conf, if you boot your system with the LiveCD?