I'm using RedHat 6.2 to set up RAID1 root directories for dual IDE systems(1).
Neat idea to spot the `boot=/dev/md*' and turn it into a `foreach' loop over
the disks in the partition.
However, it appears rather fragile in various ways.
If I don't include a `default=' line, the labels are not available when TAB is
pressed at the LILO prompt, and it tries to boot from a null label and fails.
I can kind of work round that one by giving an explicit default, but I find
that I frequently get the initrd (which contain raid1.o) reported to have
a checksum error, and mounting root (0903) fails. Running `sync' several times
after running LILO and before rebooting seems to help, but it's a bit worrying.
My main concern is that it replaces the `boot=/dev/md*' with `boot=/dev/hdX',
and thus puts LILO on the MBR, rather than as I usually do, in the Linux
partition. I prefer to keep the MBR free to be trampled by M$ OSes etc.
Has anyone any idea if there is some basic requirement to use the MBR, or
whether it's just that there is no way to pass over the information as to where
the boot sector is to be written ?
Should it be that case that the data written to the MBR can be copied to the
boot sector of the Linux partition, the MBR re-instated, the Linux partition
marked as bootable, and things should then work ? (it seems not to be the case)
I tried using a (read-only) mounted copy of the `raw' disks in the image paths
so that LILO can find the block numbers, but it has to have read-write access
to /boot/map, so I couldn't work out how to trick it (without losing RAID1
consistancy) :-(
What do otthers do -- just use the MBR ?
(1) can anyone tell me how to get RH kickstart to say `set up / on this RAID' ?
I hoped that having created hda3 and hdc3 as 0xfd, I could use:
part raid.01 -onpart hda3
part raid.02 -onpart hdc3
raid / --level 1 --device md3 raid.01 raid.02
but kickstart doesn't like it, and as the kickstart method doesn't offer a
`make RAID array' option, I can't use kickstart :-(
