Re: LILO with root-raid nightmare - please help.
hi ya drbob
On Mon, 16 Jan 2006, drbob wrote:
> I use mdadm to manage my array. The command was
>
> mdadm --set-faulty /dev/md1 /dev/hda1 --remove /dev/md1 /dev/hda1
i'd think there might be a short cut version, but i'm being
lazy to go look it up
- i'd set it faulty first, if the system didn't already
flag it
- than remove it after the the set-faulty
--- insert new disk
- than (hot-swap-ADD) the new disk
> My system wouldn't boot at all with the new drive in place until I went
> into the bios and explicitly told it to boot off the drive on the ide
> card.
lilo takes care of writing the mbr to both disks without the extra-boot
> The drives attached to the card were listed in the bios setup
> screen as boot options so I'm pretty sure my bios is compatible with
> booting off raid.
being visible to the bios boot sequence does not guarantee it is bootable
> Since Windows will only boot off the drive with bios
> device code 0x80 I believe that the bios may be switching the codes
> around and this is what is confusing lilo.
lilo has NOTHING to do with windoze and windows cannot change the bios
windoze can boot off of D: or E: or any other disk or partitions, but
its happiest as C:
> It's the version that comes with sarge. I set up the array with the LILO
> that came with woody, surely upgrading isn't necessary? I'll look for a
> backport.
for raid ... lilo-22.7.1 is best to fix boot problems
> >>raid-extra-boot=/dev/hda,/dev/hde
>
> AFAIK that line is essential. It tells lilo to write the mbr to both
> members of the raid array,
see above
> That is correct, however this was my first Linux system and I went a bit
> overboard on the partitioning: tmp, usr, var, and home each have their
> own partition.
nope ... personally, i require those partitions on any system that
i maintain or the customers are billed extra time and $$ for me to
sit and wait for it to fsck the whole 500GB instead of just 256MB
to get into single user mode to fix things
> /dev/md8 /home ext3 defaults 0 2
in the prev posts, one disk had /dev/hda8 listed while /dev/hde did NOT
have /dev/hde8
- i don't know if you inadvertantly deleted partition-8 in
the prev examples
> :~# cat /etc/mdadm/mdadm.conf
looks good
> :~# cat /proc/mdstat
looks good ... all shows as aactive with "UU"
now to fix booting problems ...
- try lilo-22.7.1
- copy all the data on the disk off to another set of disks
and break the normal sarge setup and use a newer lilo
and i'd bet it will boot off either disk
- other test ...
- move the cd or ?? in /dev/hdc and /dev/hdd onto the
pci-based ide controller
- boot into single user with a cdrom and change mdadm.conf to show
/dev/hdc instead of /dev/hde and power off
- move the disk on /dev/hde into /dev/hdc
- reboot and keep your thumbs crossed
c ya
alvin
Reply to: