[Hosting] my understanding of where we are going

[Christian J Hedemark]

Tom asked:
> I couldn't come because my boss asked me to go into work on Saturday.
Could
> someone post a summary of what was discussed/accomplished?

Well you missed out.  I've worked in some pretty cool data centers in my
life but Inflow takes the cake.  Very slick.  It's not a concrete bunker so
if you are worried about bombs dropping you will need to spend more money.
But they *are* a pretty well planned out data center.

We got to see TriLUG's cabinet, and boy is it empty.  Fatalpha is all alone
in there with a cheapo little desktop hub.

There was a lot more camaraderie and BS'ing that went on, which was fine,
but we didn't get to talk too much about what we're going to do.  We did get
a sense for general direction.  Here are some random recollections of what
we talked about:

* We're going to set up a private network within the cabinet, so that any
servers that will be publicly visible will need two NICs.  Not all hosts
will be publicly available.

* We have one, maybe two, MIPS boxes that will be added to the rack.  The
Cobalt Raq that Kevin has will be our Kerberos server.  It's not very fast
so it might not do much more than this.

* John Carnes is giving Kevin Sonney a Cobalt Cube.  This is a very slow
MIPS box also.  Kevin is going to play with it and if it seems useful, it is
up to him whether he wants to donate it to TriLUG or not.  I think there was
talk of using this as a MIPS build box.

* We have another Alpha box coming, courtesy Ed Hill.  Think "Alpha Build
Box".

* We either have or will scrape together the parts to build an X86 box.  If
memory serves this is going to be a build box, AND it will serve up NFS to
the other hosts.  We have at least one 18GB drive and some 9GB drives (the 9
giggers are in fatalpha now but will move) that are all SCSI.  I think I
have a host adapter to donate for this cause.  There was also talk about
buying a BIG but CHEAP IDE drive for archiving stuff.  Use the SCSI drives
for live builds and then shuffle them off to the big IDE drive for archival
and retreival by remote clients.

* I got the sense that we want to do some stuff *like* sourceforge without
actually using sourceforge software.  There was general agreement from what
I can tell that we will run CVS, and likely extend the use of Gnu Mailman to
cover development projects.  It looks like we'll likely use Gnats for bug
tracking.  And also we will need a series of scripts & semaphores to run a
build process across all of the build boxes.  Project maintainer kicks off
the chain of events, and maybe gets an email or something to tell him when
the builds are done.  THOUGHT:  Maybe we can use Tinderbox to help automate
this.

* Since this is a volunteer effort, we're going to use the software that we
already know and feel comfortable volunteering time to maintain rather than
acting like an IT shop and trying to find the "best" software for the job.
Some folks may question, for example, why we are using Gnats rather than
Bugzilla.  It's simply a function of using what we know in order to make the
best use of our volunteers.  This has NOTHING to do with saying that one
package is better than another or anything like that.

* I think I was the one that said something about a fundraiser to buy some
odd hardware to make this happen if we couldn't get everything donated.  For
example, a big honking IDE drive (like 100GB or so).  Kevin responded (and
John C. confirmed) that if TriLUG holds a fund raiser, it is going to be for
the pizza fund because we're running dry.  We've got about two months of
free pizza left.

* Someone had mentioned that a Snap server (NAS device) would be perfect for
this, and there was a lot of head nodding going on.  Did anyone say they
would follow up with the manufacturer to try to procure a donated NAS
device?

* I think there was some lamenting going on (maybe it was just me) that we
don't have any Sparc hardware in the donation farm yet.  Since there is a
nice size niche of Linux/Sparc users, it would be nice to be able to provide
binary builds to that community.

* Someone had mentioned, I think, that the Linux kernel can now
automatically build a spec file for constructing RPM's.  I missed a good
part of that and wanted to experiment with it.  If I heard that right, can
you please point me in the right direction for more info?  Or if I didn't
get that right, please correct me.

What did I miss?