On Thu, 2007-02-22 at 21:25 +0530, Ligesh wrote:
> On Thu, Feb 22, 2007 at 11:29:13PM +0800, Tim Post wrote:
> > On Thu, 2007-02-22 at 20:30 +0530, Ligesh wrote:
> > > On Thu, Feb 22, 2007 at 02:54:46PM +0100, Henning Sprang wrote:
> > > > Not enough doc reading about Xen and it's architecture, and the config
> > > > parameters :)
> > > >
> > Xen (open source) is a set of very powerful tools that need somewhat
> > skilled hands to use them. I think if the project shifted focus such as
> > you suggest, it would lose some of its support and community.
> Hey, I had said: "It is not part of xen the hypervisor, but xen as a
> For instance, xenEnterprise. Or any good management tool built on Xen should
> the means to do a little more than mere create/delete. Your solution is
> actually nice.
> The problem with your templated osimages is that the scripts can be run only
> ONce you have setup the domU, then the placeholder variables all disappear
> and then
> a _reconfiguration_ is not possible. Other than that, the idea is quite
I use quite a bit more 'helpers', reconfiguration is very easy. I was
just giving a rough idea of how I initially "pave" guests.
> A better idea would be create template files separately:
> So a generic script can scan '/etc' for all files ending with xentemplate
> then configure the values
> and save them without the suffix.
> for i in `find -name '*.xentemplate' /mount/domU/etc/` ; do
> sed 's/$IPADDRESS/126.96.36.199/g' < $i > `basename $i .xentempate`
> sed 's/$HOSTNAME/host.domain.com/g' < $i > `basename $i .xentempate`
> This is actually a wonderful idea. That way, the script remains generic, and
> each distro can prepare
> itself Xen with the proper template files.
I agree with everything but sed, but I'm ultra paranoid about line
editors that malloc() way more than they need to because they have no
idea whats in your file :) On dom-0 contiguous cache is prime
real-estate, I treat it like I would any embedded / small memory model
system and code accordingly.
In fact, I helped write grawk, our utility that combines some of the
most frequently grep + awk pipes together into something quite a bit
smaller : http://dev1.netkinetics.net/grawk/ for log sawing and easy
string extraction from delimited files / output scraping. Feel free to
use it instead of sed for such a simple task. Patches for additional
regex are of course welcome, provided they do not significantly increase
the size of the executable and malloc() wisely.
That being said, I do in fact realize that I am *over paranoid* about
it :) I freely admit that I suffer from many odd 'ticks' that cause me
to obsess about every single byte of RAM in every single system I
maintain. Its a wonder I get anything done.
Xen-users mailing list