Hi all,
Could you help please? Everything was going so well, then....
I have just (nearly) completed setting up a Linux box to
do RAID-1 "across the board" as follows:
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/md4 497603 62994 434609 13% /
/dev/md0 4059307 12227 4047080 0% /var
/dev/md1 7956242 3468 7952774 0% /home
/dev/md2 2035507 1 2035506 0% /tmp
/dev/md3 4059307 394260 3665047 10% /usr
/dev/md5 497603 4461 493142 1% /boot
Personalities : [raid1]
read_ahead 1024 sectors
md5 : active raid1 hda1[1] hdc1[0] 513984 blocks [2/2] [UU]
md3 : active raid1 hdc5[1] hda5[0] 4200896 blocks [2/2] [UU]
md0 : active raid1 hdc6[1] hda6[0] 4200896 blocks [2/2] [UU]
md2 : active raid1 hdc7[1] hda7[0] 2104384 blocks [2/2] [UU]
md4 : active raid1 hdc8[0] hda8[1] 513984 blocks [2/2] [UU]
md1 : active raid1 hdc11[1] hda11[0] 8249280 blocks [2/2] [UU]
unused devices: <none>
The last step in this, was to turn the /boot partition (/dev/hda1)
into a RAID-1 array (/dev/md5) of /dev/hda1 + /dev/hdc1, which I did
using the "failed-disk" method described in Michael Robinton's excellent
"Root Raid + Lilo mini HOWTO". All OK so far.
I wrote two lilo.conf files, with the intention of running lilo twice -
once to write the boot sector of /dev/hda1 and once to write the boot
sector
of /dev/hdc1.
(I ASSUME that's the plan in the HOWTO? I couldn't think what else to do
with a pair of lilo.confs!)
So, I wrote two files /etc/A.lilo.conf and /etc/C.lilo.conf that look like:
# /etc/A.lilo.conf
disk=/dev/md4
bios=0x80
sectors=63
heads=255
cylinders=1027
# Dummy only - any md device will do!
partition=/dev/md1
start=63
boot=/dev/hda <----<<< C.lilo.conf has boot=/dev/hdc here.
map=/boot/map
install=/boot/boot.b
prompt
linear
timeout=50
image="/boot/bzImage-2.2.14#1"
label=raid-md4
root=/dev/md4
read-only
I did '/sbin/lilo -v -C /etc/A.lilo.conf', and I got:
LILO version 21, Copyright 1992-1998 Werner Almesberger
Reading boot sector from current root.
Merging with /boot/boot.b
Sorry, don't know how to handle device 0x0905
Device 0x0905 is /dev/md5.
I have a number of questions:
Am I right in assuming the intention is to have lilo write "underneath" the
raid array - directly onto /dev/hd[a|c]?
Why does lilo think it needs to "handle" /dev/md5 at all? I've asked it to
write to /dev/hda's boot sector. OK, it needs to go look at /boot/boot.b,
/boot/map, etc, but that should be accessible using normal read/write
stuff.
No?
Why can't lilo use the sectors/heads/cylinders info I gave it. Isn't the
whole idea that it should use these INSTEAD of the md device?
What next? I'm booting from floppy now :(
I'm probably being completely dense here - any help would be greatly
appreciated. I could go back and split /dev/md5 into its original parts,
but that would be admiting defeat.
BTW, kernel=2.2.14, patched with mingo's 2.2.14-B1 with
raidtools-19990824-0_90_tar.gz. The rest (including lilo) is vanilla Redhat
6.0
Thanks & Regards,
Martin Munt
[EMAIL PROTECTED] or [EMAIL PROTECTED]
