RPM Fusion Infrastructure changes

[Matthias Saou]

Hi everyone,

Quite a productive 2011 FOSDEM from my side. First of all, let me
apologize for being so absent since... well... quite some time.

A few of us RPM Fusion participants have met yesterday : Hans, kwizart
and me. I would have liked it to be more, maybe next year? If anyone is
still at FOSDEM today (Sunday), reading this in time and would like to
meet up, do reply.

Here's what came out from the discussion : The "infrastructure" needs
more love, and as a consequence people are getting tired of it all.
There is of course more to it than just that, but it seems to be a good
summary of where RPM Fusion currently stands.

So, here's what I'll commit to in order to try and make it up to the
project : Clean up all of the existing infrastructure, starting with
all the basic services and moving towards the more complex ones, while
at the same time fulfilling these important requirements :
 * Make sure everything is documented in detail, so that everyone knows
   how everything is architectureed and organized (I had actually
   already done this for the initial architecture in the Infrastructure
   wiki page, but without detailing much).
 * Make sure all services can be managed by at least 3 people. This
   includes "physical" server management (remote reboot etc.), domain
   names, email/lists etc. to ensure that no human single point of
   failure exists. Currently I'm the only one in control of the
   physical servers and the domain names, which is bad.

Here's how I intend to proceed :
 * Rent a dedicated server. This will require the project to take
   donations in order to live : Getting hardware "lent" from here or
   there is really nice, and will still be fine for satellite services
   like mirrors and builders, but for the central vital infrastructure
   (dns, mail, buildsys, bugzilla), I think this is best in order to
   be independent from individuals. I've actually already ordered a
   server at OVH Spain which will cost 60€/month (I've pre-paied install
   + 3 months), which I think is more than reasonable.
 * Move all existing services "mostly as-is" (i.e. switching to RHEL6
   but not to git or koji for now). I already have a detailed plan
   which I'll try to post shortly. Here, I hope to have all of the
   current sysadmins chime in and participate of course :-) The idea is
   to use RHEL6, kvm and 100% puppet. Also implement IPv6 from day one.
   I've already got lots of reusable stuff here and there, and anyone
   who also does is welcome to share!

Once all this is done, the next steps will be to go for the more
specific architecture : VCS, buildsys... having being disconnected from
the project a lot, I don't really know what the current plans for the
future are, but I'm quite sure everyone would like to move to koji+git.
Having python 2.6, clean VMs, ease of granting more accesses etc.
should help a lot getting it done will the least frustration possible.

Let me know what you think!

Matthias