Will Woods wrote:
On Fri, 2008-11-07 at 17:40 +0000, Camilo Mesias wrote:
A quick update on this. I successfully upgraded one machine about a
week ago, so I tried another machine today and it had a problem
booting during the install.

After all the packages had been downloaded and the 'reboot' button was
pressed, grub was no longer able to come up. The word 'GRUB' was
displayed and nothing else.

I booted with the F9 Live USB and did the following (sorry it doesn't
help with diagnosis but it might help if anyone else has this problem)

mkdir /tmp/rootdir
mount /dev/mapper/VolGroup00-LogVol00 /tmp/rootdir
mount /dev/sda1 /tmp/rootdir/boot
/tmp/rootdir/sbin/grub-install --root-directory=/tmp/rootdir /dev/sda

If there are any logs that might help diagnose the problem, let me
know. I have one more desktop and one more laptop running F9 which I
plan to preupgrade in the same way, Please let me know if there is
anything I should note to help improve the process.

This is a GRUB bug that we've been unable to reproduce reliably, and
therefore haven't traced fully. It's often triggered by using the grub
--once flag, which is used by both preupgrade and suspend-to-disk
(hibernate). We've seen it happen with both of those things, but the
preupgrade case seems to be more frequent because it tends to also
involve upgrading GRUB at the same time.

As far as we've been able to trace it, *something* causes GRUB stage2
(or stage 1.5) to move its physical on-disk location. But we don't know
what. Nothing that we're doing to GRUB *should* cause that. But
something does, and then stage1 can't find the rest of GRUB, and we get
stuck with "GRUB" on-screen and an unbootable system.

Anyway. I've spent *weeks* trying to track that one down, bugging pjones
(our GRUB maintainer) and esandeen (our mad ext3/ext4 hacker) endlessly,
and never made any real progress. Reinstalling GRUB, as you did, is the
proper fix when this happens. But we still don't know what causes it,
and therefore how to keep it from happening in the first place.

Any further help or theories on this would be greatly appreciated.


Ok, I just ran into this nasty little problem when I did my F9 => F10 preupgrade.  All I get at the first reboot is the GRUB prompt.  Has any progress been made on tracking this down?  This will be a real pain if I have to go through this on 14 machines.  Is there any workaround to avoid this problem?