Subject: Re: NetBSD/Xen How to - suggestions
To: Greg Troxel <gdt@ir.bbn.com>
From: Sarton O'Brien <bsd-xen@roguewrt.org>
List: port-xen
Date: 02/06/2008 12:33:52
Greg Troxel wrote:
> "Sarton O'Brien" <bsd-xen@roguewrt.org> writes:
> 
>> Hubert Feyrer wrote:
>>> On Wed, 6 Feb 2008, Sarton O'Brien wrote:
>>>> Has anyone played with grub2?
>>> No, but while on the subject: has anyone played with pygrub (or is
>>> there an alternative), to not require domU kernels sit on the dom0
>>> file system?
>> Being part of xentools I'd imagine there are probably only two people
>> on this list that could even attempt it's use :)
>>
>> I'm curious of the perceived benefits ... other than manual config
>> changes and the consolidation of the running kernel and the domu
>> reference kernel ... Is the benefit mainly time saving?
> 
> To me it's broken that the domU kernel isn't taken from /netbsd in the
> domU filesystem.
> 
> When i update a machine, I update kernel and user, and it would be good
> for a domU to be able to do this in one operation without dom0 access,
> just like a real machine.
> 
> So, I'd like to see 'xm start' be able to mount the domU vnd from a file
> read-only, grab the kernel, unmount, and then boot with the kernel file
> and the vnds configured normally.  But I have not liked this enough to
> actually do it :-)
> 
> So I think you're right about the benefits being having the real and
> reference kernel being the same, and not having to update twice, but I
> think that's pretty important.

Thinking about it a bit more, I agree. The independence of the kernel 
and the ability to make the update process autonomous would be very 
beneficial. It would also make staging an update less annoying.

I'd imagine the biggest benefit would be to people hosting, as it would 
allow the client to do their own thing without any intervention, 
assuming all components are known to work prior to any modification. 
You'd want to document what can and can't be done.