lab hardware move/reorg + moving forward

[Dave Taht]

On Fri, Mar 30, 2012 at 10:53 AM, Francis Dupont <fdupont at isc.org> wrote:

> > I have a need to re-organize and rewire the lab and return at least one
> box
> > to it's original usage.
> >
> > The x86 gear:
> >
> > are io.lab.bufferbloat.net (the aftr box)
> > and ida (one of my laptops) (the sdnat box)
> >
> > I can live without io for a while, or the laptop, but not both. (they are
> > the only gigE capability I presently have)
>
> => I have saved the configs, even fixed the DHCPv6 server one which was
> bad, so if you can wait Alistair is back with 3 boxes I have no concern.
> BTW I have to install my old iMac and primary DNS server replacement
> so I don't expect to be active until Monday. I have the HSM mess
> on my TODO list next week too (even the critical point was fixed
> a long time ago and the customer now knows this :-).
>

I am figuring I will diff /etc when the gear returns to update the basic
firmware.


>
> > I'm curious enough about aftr to configure a cerowrt box as an aftr box
> as
> > well, so I can leave up two wndrs for that purpose,
> > or leave a wndr box off of io for now and re-re-re-purpose that laptop.
> >
> > So A) are both/either these boxes still needed during the rest of ietf?
> (so
> > I can at least rewire)
> > and B)...
>
> => IETF is nearly finished so nothing technical should be done.
>
> As you say rewire I still believe we should not use the wireless in the lab
> (hard to debug, basically insecure with easy to use config, so in most
> cases a bad idea. BTW please document how to get rid of wireless on WNDRs,
> I know only for WRT54G* (remove the antenna)).
>

Well, I need wireless for the bloatlab! You can do whatever you want. :)

I can produce a build with wireless expressly disabled, but frankly I felt
it was important to stress the code out with more users, which is easy to
do with all the wireless clients floating around. It very rapidly shows
that 500 ports is not enough.

So: I could leave wireless up with crypto enabled? or disable it entirely,
that's a single line per radio in the /etc/config/wireless file:

option disabled 1

Another reason why wireless is up by default is that I update machines that
are connected wirelessly only and after a reflash, need to be able to get
to them to finish configuration.


>
> > My plan would be to rebuild the aftr/sdnat lab out of the demo gear
> > returning, which I assume will be monday(ish)?
>
> => ask Alistair (as far as I know he won't rush on the first plane)
>
>
battlemesh.org is worth attending this week too.


> > So a build system needent be a dedicated box, it just requires a
> reasonably
> > modern Linux operating system, about 2GB of disk for the sources,
> > and about 40MB per preserved build. It takes about 3 1/2 hours for a full
> > build on the 8 core system that is huchra, under 8 minutes for
> incremental
> > builds.
>
> => BTW 3 1/2 hours are a lot
>
>
Has to build toolchains, pull current sources, etc, and build up everything
from scratch. It's no fun.

However I remember when we used to have to write everything in assembly,
test, create a rom mask, test, ship it off to a factory to get built, wait
a month for something to come back...

and in that context it isn't all that bad!

I would hope with more modern hardware than huchra the bootstrap phase of
the build could be cut down, however parts of it are very resistant to
parallization.

The other problem is that it's far more efficient to do incremental builds
but most automated systems like jenkins want to do the whole thing from
scratch every time.

-- 
Dave Täht
SKYPE: davetaht
US Tel: 1-239-829-5608
http://www.bufferbloat.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/sdcpe-devel/attachments/20120330/8760b0a2/attachment.html>