Software RAID1 install notes - Debian Sarge
Hi guys,
I am following software RAID1 installation instructions posted to debian-user
in August of last year. I find problems with his instructions as I progress,
and have finally gotten stuck, see "STUCK HERE" bellow.
Any suggestions where to find good software RAID1 install instructions?
Any idea how I can get past the "STUCK HERE" point bellow?
Here's a copy of Giuseppe's instructions with my install notes included in
double paretheses.
Thanks
Roger
TEFLChina.org
**************************
Following Giuseppe's
Software-Raid1 Root in Woody
http://lists.debian.org/debian-user/2003/debian-user-200308/msg01507.html
To: debian-user@lists.debian.org
Subject: Software-Raid1 Root in Woody
From: giuseppe bonacci <g.bonacci@libero.it>
Date: Fri, 8 Aug 2003 18:11:07 +0200
(( Roger
Roger's notes appear within double parens. Thus "I" means Roger in here and
Giuseppe elsewhere.
I, Roger, installed Woody on RAID1 following as best I could Giuseppe's
footsteps.
Hardware (Roger's)
40G IDE drives (2) hda, hdc
CDROM drive, hdd
Floppy drive, fd0
In BIOS set as bootable, and in this order, these:
fd0
hdd
hda
hdc
I worry that having one of my 40G drives on the same controller (same ribbon
cable) as the CDROM may slow down access to it. I will check later and report
here: XXXXXXXXXXX
Use 80 wire 40 pin IDE cables. The courser, 40 wire 40 pin cables are slower.
))
Hi all.
This document briefly describes the steps needed to install a Debian
Woody GNU/Linux system with root on a software raid1 device.
It took me a fair amount of trials and errors before I got it right, so
I would like to share my current knowledge.
<snip (disclaimer)>
Since this is a test setup, the environment is going to be small, and
supported by the debian vanilla kernel:
- Pentium III
- SCSI disk 0:0, sda, 160 Mb
- SCSI disk 0:1, sdb, 160 Mb
I used SCSI disks because I had them easily available, but the procedure
should work for IDE disks on (e.g.) hda and hdc as well.
1. Install Debian Woody.
I started from floppies (vanilla kernel) + network, and partitioned
the disks as follows:
(( Roger
- Boot from Woody CD#1, bf24
- Partition hda and hdc
hda1 00098 MB spare Linux partition (in case I need it)
hda2 01998 MB Swap for swap tempororily (later md0 for swap)
hda3 37959 MB Linux for / temporarily (later md1 for /)
-----
hdc1 00098 MB spare Linux partition (in case I need it)
hdc2 01998 MB md0 for swap
hdc3 37959 MB md1 for /
hdc4 01053 MB spare Linux partition (this drive is this much bigger than the
other one. So this is spare space outside the RAID and I might as well have
it available for what not)
- mount swap and /
- Installed kernel and kernel modules
- Configure Device Driver Modules:
:: Drivers for Network Devices > VIA Rine Support.
and
:: RAID1 (I loaded all the RAID drivers because I don't know exactly what I
need)
- Tasksel > http > "ucberkeley" > Standard UNIX Server
))
# fdisk -l /dev/sda
Disk /dev/sda: 64 heads, 32 sectors, 160 cylinders
Units = cylinders of 2048 * 512 bytes
Device Boot Start End Blocks Id System
/dev/sda1 * 1 160 163824 83 Linux
# fdisk -l /dev/sdb
Disk /dev/sdb: 64 heads, 32 sectors, 160 cylinders
Units = cylinders of 2048 * 512 bytes
Device Boot Start End Blocks Id System
/dev/sdb1 * 1 160 163824 fd Linux raid autodetect
Then I made an ext2 filesystem and installed a minimal system on /dev/sda1.
Better avoid installing further packages until the raid is set up, while
it might be a good idea to update the system with
(( Roger
I wanted to upgrade the distribution to Testing, so did,
# vi sources.list
(edit out security sources and do :%s/stable/testing/)
# apt-get update
# apt-get dist-upgrade
))
# apt-get update
# apt-get upgrade
2. Switch to kernel 2.4.
(( Roger
I already had kernel 2.4.18 because I booted install CD with 'bf24'.
So I followed Giuseppe's footsteps but choose an appropriate more recent
kernel,
# dpkg -P lilo
# apt-cache seache kernel-image
# apt-get install grub kernel-image-2.4.24-1-686
# grub-install '(hd0)'
))
I prefer to use raidtools2 with a stock 2.4 series kernel.
Moreover, despite later versions of LILO being easier to setup, I
decided to stick on GRUB. So the next steps were:
# dpkg -P lilo
# apt-get install grub kernel-image-2.4.18-686
# grub-install '(hd0)'
Notice that for processors different from Pentium III you should pick a
flavour different from '-686'.
(( Roger
# cat > /boot/grub/menu.lst
title Debian
root (hd0,2)
kernel /vmlinuz root=/dev/hda3 ro
initrd /initrd.img
IMPORTANT:
- I believe Giuseppe meant to put ">" there in the 'cat' sample line. Clearly
it is needed.
- my 'root' is (hd0,2), not (hd0,0), because my root is on partition hda3, not
hda1 (and Grub counts partitions starting from 0, so 3rd partition is 2).
- In the "kernel" line I put /dev/hda3, not /dev/sda1, because my root is on
3rd partition, not 1st partition, and mine drives are IDE, not SCSI.
))
Create /boot/grub/menu.lst like this:
# cat /boot/grub/menu.lst
title Debian
root (hd0,0)
kernel /vmlinuz root=/dev/sda1 ro
initrd /initrd.img
WARNING: Don't install devfsd. Device paths change and initrd-tools
get confused.
Reboot.
3. Create and populate the Raid 1 structure in degraded mode.
Install raidtools2:
(( Roger
Durring the following 'apt-get install raidtools2', I chose 'no' when asked if
I wanted to have raidtools2 put RAID startup into the system initialization
script, because I configured RAID into the kernel when I installed, and so
things should work this way I think.
))
# modprobe -k raid1 # needed to prevent preinst from complaining
# apt-get install raidtools2
(( Roger
- Again, I think Giuseppe meant to 'cat > /etc/raidtab', not 'cat /etc/
raidtab'.
- Giuseppe has sdb as raid-disk 0, sda as raid-disk 1. That seems backwards to
me, but I tried it the other way and 'mkraid' didn't like it. So I guess for
this goofy procedure the order needs to be "good" raid-disk first, "broken"
raid-disk second.
- I have have two RAID partitions, Giuseppe has only one Raid partition. So I
have these:
md0 (hda2 & hdc2) for swap
md1 (hda3 & hdc3) for /
- My /etc/raidtab I created like this:
# cat > /etc/raidtab
raiddev /dev/md0
nr-raid-disks 2
raid-level 1
persistent-superblock 1
chunk-size 4
device /dev/hdc2
raid-disk 0
device /dev/hda2
raid-disk 1
failed-disk 1
raiddev /dev/md1
nr-raid-disks 2
raid-level 1
persistent-superblock 1
chunk-size 4
device /dev/hdc3
raid-disk 0
device /dev/hda3
raid-disk 1
failed-disk 1
))
Create /etc/raidtab like this:
# cat /etc/raidtab
raiddev /dev/md0
nr-raid-disks 2
raid-level 1
persistent-superblock 1
chunk-size 4
device /dev/sdb1
raid-disk 0
device /dev/sda1
raid-disk 1
failed-disk 1
Notice the failed-disk directive at the end of the file. It specifies
that the current root disk, /dev/sda1, is to be ignored for the moment,
and marked `failed' in the raid superblock.
Now make up the raid:
(( Roger
# mkraid /dev/md0
# mkraid /dev/md1
))
# mkraid /dev/md0
If this is your n-th trial with n>1, you might be forced to
use the '-f' option to mkraid. Read carefully the warning before
proceeding.
(( Roger
# less /proc/mdstat
(NOT /proc/mdstatus)
))
You can check the status by looking at /proc/mdstatus
Build an ext2 filesystem on /dev/md0 and copy the whole system on it
(that's why it's useful to keep it minimal)
(( Roger
# mke2fs -O sparse_super,filetype /dev/md1
# mount /dev/md1 /mnt <------------------------- STUCK ON THIS----
mount: wrong fs type, bad option, bad superblock on /dev/md1, <-------------
or too many mounted file systems
# find / -xdev -depth | cpio -pmdu /mnt
))
# mke2fs -O sparse_super,filetype /dev/md0
# mount /dev/md0 /mnt
# find / -xdev -depth | cpio -pmdu /mnt
Modify /mnt/etc/fstab and /mnt/boot/grub/menu.lst to mention md0 or sdb in
place of sda1:
# diff /etc/fstab /mnt/etc/fstab
4c4
< /dev/sda1 / ext2 errors=remount-ro 0 1
---
> /dev/md0 / ext2 errors=remount-ro 0 1
# diff /boot/grub/menu.lst /mnt/boot/grub/menu.lst
2,3c2,3
< root (hd0,0)
< kernel /vmlinuz root=/dev/sda1 ro
---
> root (hd1,0)
> kernel /vmlinuz root=/dev/md0 ro
Now build an initrd for the new setup:
# mkinitrd -o /mnt/boot/initrd.img-2.4.18-"your flavour" -k -r /dev/md0
The option '-k' instructs mkinitrd to keep the expanded image under
/tmp/mkinitrd.XXX/initrd instead of deleting it.
Notice that /mnt/initrd.img -> /boot/initrd.img-YYYY, so don't try to be
too smart.
4. Reboot with root on /dev/md0 and synchronize the disks.
Reboot. So far, the system is still configured to boot from the first
disk and leave the second alone.
Now we have to manually interrupt grub's boot sequence in order to
launch the second disk instead of the first:
Ask grub for a command line (press 'c') and at the 'grub>' prompt
issue the command:
grub> configfile (hd1,0)/boot/grub/menu.lst
You should get the same menu item as before, but meaning a different sequence
(use 'e' to examine it, then ESC to return to the menu). Press RETURN
to boot.
You should end up with linux running and /dev/md0 mounted as /.
Now it's time to change the partition type of the old root disk from 83 to
FD:
# fdisk -l /dev/sda
Disk /dev/sda: 64 heads, 32 sectors, 160 cylinders
Units = cylinders of 2048 * 512 bytes
Device Boot Start End Blocks Id System
/dev/sda1 * 1 160 163824 fd Linux raid autodetect
Now modify /etc/raidtab to remove the last line ('failed-disk...'), and
attach the old root partition as a plex of the raid1 structure:
# raidhotadd /dev/md0 /dev/sda1
The md driver starts synchronization, that can be checked by looking at
/proc/mdstat.
Rebuild the initial ramdisk with a clean raidtab (paranoid mode on):
# mkinitrd -k -o /initrd.img
Setup grub on both disks:
# (echo 'root (hd1,0)'; echo 'setup (hd1)'; echo quit) | grub
# (echo 'root (hd0,0)'; echo 'setup (hd0)'; echo quit) | grub
Done.
Now you can reboot (just to test the setup works), play with dselect
to your pleasure, etc.
Again: if you happen to install devfsd you might end up with an unbootable
system, and no way to recover.
As an exercise to the reader, try disabling one of the disks and test
your ability to recover... If you find a way out of the Kernel panic,
I'll be glad to know. ;-)
best regards
-- gb
*************************
Reply to: