Thanks for the replys,
It looks as though it should be easy.
I want to be able to reboot using md0 for the root boot partition,
then I can dispose of the ext2 fs on the /dev/sda1 and raidhotadd
sda1 to md0.
Is this possible, what does this lilo error message mean?
[root@otherweb /root]# df
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/sda1 127902 40984 80314 34% /
/dev/md1 10079980 37028 9530908 0% /home
/dev/md2 4032448 416792 3410812 11% /usr
/dev/md4 254699 15 241532 0% /tmp
/dev/md5 254699 6844 234703 3% /var
/dev/md3 2016656 1992 1912220 0% /usr/local
/dev/md0 127790 40988 80204 34% /stage
[root@otherweb /]# lilo -V
LILO version 21.4-3
[root@otherweb /]# lilo -r /stage
boot = /dev/sdb, map = /boot/map.0811
Added bzImage *
Syntax error near line 2 in file /etc/lilo.conf
[root@otherweb /]# cat /stage/etc/lilo.conf
boot=/dev/md0
default=Linux
delay=50
vga=normal
read-only
image=/boot/bzImage
label=Linux
root=/dev/md0
image=/boot/vmlinuz
label=lin
root=/dev/md0
[root@otherweb boot]# head -9 /etc/raidtab
raiddev /dev/md0
raid-level 1
nr-raid-disks 2
persistent-superblock 1
chunk-size 4
device /dev/sda1
failed-disk 0
device /dev/sdb1
raid-disk 1
I want to be able to reboot using md0 for the root boot partition,
then I can dispose of the files on the /dev/sda1 and raidhotadd
sda1 to md0.
Is this possible, what does the lilo error message mean?
-------- Original Message --------
Subject: RE: root+boot on Raid1 + lilo
Date: Wed, 28 Jun 2000 09:46:44 -0500
From: "Diegmueller, Jason (I.T. Dept)" <[EMAIL PROTECTED]>
To: 'Hugh Bragg' <[EMAIL PROTECTED]>
Hugh--
It's actually relatively easy to move a root filesystem over to a
RAID. I don't have a HOWTO or step-by-step, but here's what I did
just one week ago for a 3-disk RAID1:
- Install System as usual
- Establish /dev/md0, but mark the used disk as "failed-disk" in
the /etc/raidtab (as opposed to raid-disk). This will have the
software RAID mark this disk as bad, but since it's RAID1, it will
continue in degraded mode. Also, make sure you ARE using the
persistent-superblocks.
- Format the filesystem (I use 4096-byte block e2fs, benchmarks show
that to be quicker)
- Mount the /dev/md0 filesystem. I always use "/newraid".
- Copy the current filesystem over to the new filesystem (I used
cp -a /bin /dev /usr /var etc etc etc /newraid". No need to
copy /proc.
- mkdir a /newraid/proc (so there is a /proc under the new filesystem
for when it boots up)
- Modify the /newraid/etc/fstab to utilize /dev/md0 as the root
(as opposed to /dev/sda1 or whatever)
Here's where it gets subjective. Some people use the RedHat-patched
RAID1-bootable lilo, me personally .. I do this:
- I have separate 16MB boot partitions, /mnt/boot1, /mnt/boot2, and
/mnt/boot3. These are exactly identical (/tmp, /etc, /boot, /dev)
with /etc/lilo.conf's underneath which tell lilo to install on it's
corresponding drive (ie, /mnt/boot1 == /dev/sda1, and has a
"boot = /dev/sda" line, /mnt/boot2 == /dev/sdb1, and has a
"boot = /dev/sdb" line, etc). This way, any drive can die, and
booting
will occur.
- I then run "lilo -r /mnt/boot1", "lilo -r /mnt/boot2", and
"lilo -r /mnt/boot3". The -r tells lilo to chroot, which is why you
must have a directory structure under /mnt/boot1, etc. You must have
a /mnt/boot1/dev with the devices referred to in
/mnt/boot1/etc/lilo.conf
(such as /mnt/boot1/dev/md0, /mnt/boot1/dev/sda
Also ..
- Make sure the RAID'd partitions are marked type "fd", for RAID
autodetect. Otherwise you'll have trouble booting.
Reboot.
- If you're up and running on the /dev/md0 root, you now need to
integrate
the "old root" in to your RAID. Modify /etc/raidtab and change the
"failed-disk" to "raid-disk", then "raidhotadd /dev/md0 /dev/sda5"
replacing
sda5 with the old root. It might make you --force it, I can't
remember.
When it finishing rebuilding, mark the old root partition as fd as well
so
it will autodetect. You should then be good.
I realize this is a long diatrabe and not a simple HOWTO, but hopefully
it
will help. Feel free to email ANY questions ... I'd love to help. I'm
not
a RAID maintainer, just a happy user. =) I'm doing root-RAID5,
root-RAID1,
and even root-RAID0 on 5 different machines, 3 of those I'd consider
"production" machines .. one of those being up over 300 days now without
a
problem.
