Soliciting Infrastructure feedback on Fedora ARM primary arch proposal
On Thu, 22 Mar 2012 13:11:49 -0500
Dennis Gilmore <firstname.lastname@example.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> Hi everybody.
> This Monday FESCo did an initial review of the ARM team's ARM PA
> Feature proposal. As part of that review, they requested we get in
> touch with affected groups including Infrastructure. What we need is
> to get feedback on what's been written and what still needs to be
> written as it pertains to Infrastructure. I know we have had some
> brief discussions already but id like to get everything together in
> one place to make sure we capture all of the requirements from infra,
> and that we have a workable, scalable, manageable solution. I do
> realise some of its a bit hard right now as we are not 100% sure how
> the hardware will look and work in practice.
Yeah, I think thats the biggest ? in my mind... since we don't know how
fast the hardware will be or how many nodes we will be dealing with
it's hard to say if it will keep up or be managable, how we boot or
load them, etc.
I think we've made some good steps recently about this with skvidal's
work to automate builder installs. We are going to need something like
that outside of puppet, but assured that it's installed and configured
correctly to manage the increased nodes.
> The feature page in question is:
> As you might have already ready on devel@, this is a work in
> progress and has some known deficits. We'll keep updating it based on
> the feedback we receive until we have a plan that works for all the
> stakeholders. If you have some feedback that you haven't already
> shared on devel@, or that you want to share again because it's really
> important and infra related please reply to this message and let us
So, we are getting some koji backup netapp space before too long
hopefully. (Although we might want to use some of this new space for
our private cloud instance too) and hopefully that will help us with
disk issues, but...
Do we know how big disk wise a arm tree is? Ie, currently how big is
f17 and rawhide for arm on disk?
All of it will need to be in a unified /mnt/koji right?
Does mash need to be modified any to mash arm stuff correctly?
(I assume that work is already done).
How about bodhi?
I don't think there's anything insurmountable in this, but we will have
to figure out how to make things scale a good deal more than they have
in the past.
infrastructure mailing list