A rant regarding LILO and various other related issues

[Chris Adams]

Once upon a time, John Morris <jmorris at beau.org> said:
> Because if grub is KNOWN
> to be broken in RHEL3 (guess ENTERPRISE environments don't use RAID
> ....) almost certainly still broken in RHEL4 since it is also broken in
> FC3

GRUB is not "broken" in a RAID1 environment (broken to me means it
doesn't work at all).  It doesn't do what you'd like, but that is
because it works differently.  All it takes to install GRUB on multiple
members of a software RAID1 /boot is (this is from a kickstart I have
around, although I haven't tried it in a while):

# If we are doing RAID for /boot, set up GRUB on all disks
if grep "^/dev/md0.* /boot " /etc/fstab; then
	grep "^md0" /proc/mdstat | tr ' ' '\n' | grep '\[' | cut -d'[' -f1 |\
	    while read dev; do
		drive=/dev/${dev%[0-9]}
		grub <<XYZZY
device (hd0) $drive
root (hd0,0)
setup --prefix=/grub (hd0)
XYZZY
	done
fi

It is a one-time thing, since GRUB (unlike LILO) only has to be
installed once (so I've had the above in my standard kickstart file for
a while now; I don't have to think about it normally).  Now, hopefully
with FC4, the above will no longer be necessary.  I would test that out
in FC4t2 myself, but I don't have a software RAID1 /boot on a system I
can test on at the moment.

> it is perhaps, maybe, we really really hope fixed for FC4 means
> this is exactly the WRONG time to be thinking about removing the only
> boot loader known to work cleanly in a software RAID1 environment.

No, it is the time for those that think there is a problem to test the
fix (that is the point of test releases and this mailing list).  If
nobody reports any problems with the test release, then the developers
will assume that the problem is fixed and move on.  If LILO users are
doing nothing but ranting and calling people names, they'll never know
if the fix works for them, so the developers won't find out if there are
still corner cases that need work.

If instead it turns out that the changes to GRUB still don't fix the
problems in a significant number of cases (and it can't be fixed by
test3 for example), LILO may return once again.  If LILO wasn't removed
though, people that have had problems with GRUB in the past will just
keep using LILO (so GRUB fixes never get tested).

> Even
> in a perpetual beta release like Fedora.  When Grub accumulates a year
> or two of real world successes, then it will be time to pull the plug on
> lilo and I can't imagine anyone objecting then.

Fedora is not a "perpetual beta release".  If GRUB works today however,
Fedora is not going to maintain another solution just because at some
point in the past GRUB didn't do everything LILO did.

-- 
Chris Adams <cmadams at hiwaay.net>
Systems and Network Administrator - HiWAAY Internet Services
I don't speak for anybody but myself - that's enough trouble.