On Fri, Mar 28, 2014 at 08:57:09AM -0700, Lance Albertson via RT wrote:
On Fri, Mar 28, 2014 at 3:37 AM, Martin Spott via RT <support@osuosl.org>wrote:
> As I can't look into the details of the particular KVM config, I'll try
> a guess. "uname -r" says:
>
> 3.4.6-gentoo-osl-guest-x86_64-1
>
> Therefore I suspect that someone (TM 
has hooked a Gentoo boot
> disk image into the VM setup. The Ganeti page says:
>
> Kernel Path /boot/guest/vmlinuz-x86_64
That's odd. We normally don't do that for non-gentoo hosts and I'd prefer
to stay away from that as we're moving away from maintaining those kernel
images. But I agree it makes it easier to deploy.
Has the problem only been when upgrading the VM to grub2? We haven't
updated the images in a long time, maybe we should focus on fixing that as
well?
Using the Gentoo boot image is a rather pragmatic solution and
personally I'm fine with it, as long as it works. At least it's by
magnitudes better than a host not being reboot-safe 
Note that since I don't have access to the GRUB error messages on our
OSGeo VM's, my opinion is based on "educated guesses" only. Now, as
far as I can tell, the breakage relates to the embedding area being too
small for GRUB2, an issue I've experienced on various machines which
had initially been set up with Debian5.
I assume that's a general birth defect affecting all of our VM's and
can be fixed by prepending a few, let's say 10 MBytes at the
*beginning* of each virtual boot disk, followed by adjusting the
partition table accordingly and re-installing the boot loader.
Cheers,
Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------