On Mon, Mar 31, 2008 at 11:27 PM, Ira Abramov < [EMAIL PROTECTED]> wrote:
> Quoting Marc A. Volovic, from the post of Mon, 31 Mar: > > poppet > > Thanks. Took me 5 minutes to discover it's spelled Puppet, and 20 more > of reading through all the FAQs and manuals to realize it does > management, not provisioning. You can look at it both ways. We use puppet (still learning it) to provision a few Xen guests remotely. Right now we base the install on an existing Xen image (because cpan install is such a mess that the external software provider just dropped us an image). Just yesterday I noticed something called "rollout" ( http://dparrish.com/category/projects/rollout/). It's a rip-off of provisioning software developed in my previous workplace which is used to provision 500+ physical RHEL servers (compiled locally from source). The original code is a bit horrendous (being developed by system admins, not programmers) but does the job extremely well. (to clarify, my experience is with the original code, yesterday I just noticed this web site and from the description (and having heard the name of the author before) I'm sure it's just a copy of what I used there over a year ago). The idea is that you sort of "assert" what software should be installed on the server (be it rpm's, cvs checkouts or whatever) using a giant Perl Hash to describe individual machines, classes of machines and software packages. It can also control any bit of the system configuration and the idea is that you can just kickstart a machine and it will automatically pull down the perl script and configuration at the end of the kickstart process and install everything from there. Individual software package have an opportunity to plugin their own iondividual configuration into the mix and since it's all in perl you have full flexibility to do anything you like (including hacking your foot off with a Swiss Army Chainsaw, of course). The idea is that you should be able to just turn on the machine and forget about it - remember the context it was developed in - 500+ servers which could be literally on the other side of the continent and you want to allow the ops people to just kick-start a replacement server during the night until someone can come over to look at the problem in the morning. You can still update configuration from it later (e.g. add another package or change a config and re-run rollout to apply the change) but this is used mostly during development. For production or staging use it is expected to be used from kickstart, as lint will accumulate over time (it doesn't know about removing unused packages left behind, for instance, or removing old version of the configuration). I'll make it clearer: I'm looking for a product that will allow me to > remote-install blades and tower machines via PXE from a smart kickstart > or other type of image server. Management after provisioning is a bonus, > not a must. Sounds like rollout is just what you want. Cheers, --Amos
