[Mageia-sysadm] questions about our infrastructure setup & costs

Michael Scherer misc at zarb.org
Tue Apr 3 14:38:48 CEST 2012


Le mardi 03 avril 2012 à 10:34 +0200, Romain d'Alverny a écrit :
> On Mon, Apr 2, 2012 at 21:00, Michael scherer <misc at zarb.org> wrote:
> > Well, you have the hardware we paid, no ?
> > The accounting is on the bank account, and despites requiring searching, it
> > was published.
> >
> > If iyou need information on the hardware we got at the beggining, I can send
> > on this list a partial history of where
> > does every piece of hardware come from, if that what you need, but I prefer to
> > be sure that's really what you need before spending a afternoon on this.
> 
> It's more important, looking ahead, to know what is needed, because
> the costs of the infrastructure before and today (offered/acquired
> hardware, hosting, etc.) are not the same for hardware/services
> available today or tomorrow. For a given service, the needs are known
> and stable (or can be planned). The cost (and options) of that
> infrastructure, on the other hand, may change more rapidly with time
> (up or down).
> 
> > Basically : [...]
> 
> > I do not say that's not a good idea to document ( I started to write
> > various SOP on the wiki ), just that I fear that no one will
> > be motivated to do it and to keep it up to date, especially for such a huge
> > document.
> 
> Thanks for the tips anyway, I'll see what I can prepare with that,
> although I'm afraid it's useless, given some echoes to this thread
> (here and elsewhere). I won't be able to plan for your (sysadmin)
> capacities and needs (that means, looking 2/3 years ahead), if you
> don't tell me in a documented, maintained and clear way.

To tell the truth, I think we have more than what we need in term of
hardware. While we can always be ok with faster builders and so on, as I
say, we upgrade, stuff are seen as fast, and after 1 week, that's
becoming the reference speed and so people do not see it anymore.

So I propose to rather take in account than in X years, we will have
lost Y servers, Z harddrive, etc, and plan this.

( of course, we do not have X Y Z, but at least, that's the way i would
pursue for planning )

> Doesn't need to be a huge doc, especially for a start. But it needs to
> revolve around something logical for deployment and understandable for
> most, non technical people, so service-centered didn't seem absurd to
> me, and a graphic-based format neither, so that a sense of priority
> can be shared.

Service centered is not absurd, just it doesn't map easily to physical
hardware for now.

I would be in favor of moving from "one big server" to "several system"
for alamut ( and also for valstar maybe ). This would permit several
improvements :
- delayed upgrade ( ie, we will upgrade one service at a time )

- easier duplication for testing ( clone vm, upgrade vm, if it doesn't
work, we fix and try again )

- remote upgrade ( ie, done by the confort of my chair in Paris, not in
a noisy server room )

- better service isolation, and the path toward less centralized
management ( ie, we could let someone play with tx without touching to
the rest, etc ).

- better resilience, by moving vm accross server in case of upgrade, etc

an,d from a network point of view, a better layout ( ie, server are not
directly exposed to the internet, we can put a reverse proxy with
various goodies like a page to say "we are upgrading" when the virtual
server is down for maintainance, etc )

Until then, that's hard to say "this service requires that amount of
ressources".

-- 
Michael Scherer



More information about the Mageia-sysadm mailing list