On Tue, 2006-09-26 at 09:16 +0100, Mark McLoughlin wrote: > On Wed, 2006-09-13 at 12:52 -0600, Jeffrey Law wrote: > > [ I know you're on vacation -- this isn't urgent, but I didn't want > > to forget the question/issue. ] > > > > Is there a good reason why stacaccli-install requests a kernel from > > the server independently of the OS install image? > > > > ISTM that stacaccli-install should request the OS install image > > and extract a kernel from within that OS install image. Whatever > > version the client is going to be booting must exist in the > > OS install image anyway, so why have an extra copy outside the > > OS install image that has to be installed and maintained. > > > > If there is a good reason for keeping the extra copy, then we > > should seriously consider having the code which creates the > > OS install images also install this extra copy of the kernel. > > It probably came from the fact that I used Xen to create the image ... > so I would have had an image file with boot, swap and LVM partitions and > the root volume contained in the LVM volume group. When I extracted the > root volume as a standalone image file, I would have had to separately > extract the kernel from the boot partition. Ahhh. Your Xen based install created separate root & boot partitions. Makes perfect sense. > > But yes - perhaps we should require that stateless images have a /boot > directory with the correct kernel. It would make the image repository > easier to manage. I think this is the right way to go -- it simplifies things slightly for stacaccli-install for fully caching clients and it also simplifies things for NFS root clients (who still want a /boot directory, even though it may not be a separately mounted filesystem). Jeff