On 20 Mar 2007, at 12:00,David Lutterkort wrote:
on "reverse engineering" configuration changes ...
Much of that relies heavily on puppet, which makes figuring out those
changes relatively easy. The one important bit where some elbow grease
is needed is pacakges ... I am about 80% done with adding package
support to cft, and plan to release a new version with it RSN.
I guess the ability to "reverse engineer" configurations like this is
useful in allowing you to hack a configuration up on a machine, and
then fold it back into a configuration tool. It is worth mentioning
though, that it only captures "how" an individual machine has been
modified, and not "why". If you are trying to put these results back
into a large scale configuration description, you need to understand
"why" each change was made to attach it to the right "machine class"
- is this something that has too happen on all "web servers?" or only
specific ones? On all FC5 machines?
The whole "specification of virtual machine instances" seems to be
very
similar indeed to some of the work John (Sechrest) talked about at
that
same workshop.
Do you have a pointer to that ? I'd be very interested in reading more
about that. And of course hear more feedback on cft or any of hte
other
projects we are working on.
David
I think this was the work he did with Kyrre - there was a new paper
on this at LISA last year.
Kyrre's papers are here - http://www.iu.hio.no/~kyrre/
Paul
_______________________________________________
lssconf-discuss mailing list
lssconf-discuss@inf.ed.ac.uk
http://lists.inf.ed.ac.uk/mailman/listinfo/lssconf-discuss