Re: master server and other ideas

On Wed, 23 Jul 2003 09:36:11 -0500
Matt Tesauro <[email protected]> and others wrote:

[snip]
> - I love the idea of a master install server with machine
> profiles.  For those with older hardware, perhaps a mechanism for
> allowing compiled binaries to be stored there and transerfered
> over to slave servers would be a good idea.  Then you wouldn't
> have to put the hurt on your production server while recompiling a
> package (especially for updates).  At least the design should
> accomodate this as a future feature.
[snip]

In reading these related threads, which certainly sounds like the right direction, I think we really need to take a step back here and look at the "big picture" requirements.  I.e., not just roll-outs and building new boxes, but the whole lifecycle (or TCO) thing.

Last night on #gentoo-server, spoob brought up this link:

http://www.infrastructures.org/papers/bootstrap/bootstrap.html

which seems like exactly what we've been discussing here (more or less).  I only read through the first few parts last night, but I recommend everyone take a look and see what they think.

Their "gold server" concept seems like what you guys have been talking about (in terms of a master server) but has the added benefit of being a successfully employed concept (ie, already tested in the real world).  Their ideas are certainly not specific to any one vendor or OS, and most of them map directly to the goals of this group.

I'm even thinking of employing this concept for the dozen or so machines I have here at home (but the whole point is that it scales well up to thousands of machines).

So I think if we develop our requirements a little, we can build a usable concept for large-scale deployments/maintenance of both gentoo and other distros...

That shouldn't be too hard, eh?

On a related note, the offer of hosting at TAMU sounds really good, but I'm really not that fond of the existing forum/wiki software (mostly because it seems to narrowly defined in terms of features, management, etc).

I would really like to see something designed for a more flexible, distributed manager/publisher/user arrangement, with more delegation of roles and permissions, as well as more application support, publishing features, even workflow.  Hmmm, that sounds suspiciously like Zope with CMF, zwikis, etc...

I have a slow connection, but you can see some of my experimental zope stuff here, including some example CMF portals (the PHHS portal actually has members, and some example content in my home folder):

htpp://arnolds.dhs.org

htpp://arnolds.dhs.org/phhs_portal/Members/steve

Feel free to join the technology portal and try it out; it's pretty cool.  I recently migrated to some emergency hardware (and I may have not have fixed everything) so let me know if anything seems broken :/

It's an older server (with slightly older zope/python/apache), and I still haven't had the pleasure of setting up a nice gentoo zope server.  However, all the latest zope stuff (including CMF and many other products) is in portage.  And if you're not familiar with Zope, it uses python for most of the implementation (except some C in performance-dritical areas) which is way nicer for web apps and other stuff than php...

If there's any interest, I could setup the zope stuff on the new server (wherever and whatever that might be).

Zope aside, I think the infrastructures concept is exactly what we need; looking forward to lots of feedback :)

Steve


--
[email protected] mailing list