[Tech] PXE project - still alive

[Tai Kedzierski]

Thanks for the notes Michael -- some responses below :-)



On 15 February 2016 at 14:14, Michael Paoli <Michael.Paoli at cal.berkeley.edu>
wrote:

> Cool,
>
> I've not forgotten about it ... but I've also been relatively busy with
> other
> stuff too.
>
> A few thoughts/suggestions:
>
> Debian uses Fully Automated Installation (FAI).
> That doesn't necessarily mean Debian doesn't *also* use/support Kickstart,
> but the (fast) majority on Debian would use FAI rather than Kickstart, so
> FAI is likely much better documented and supported on Debian than
> Kickstart.
>
>
Never heard of FAI, will research - thanks.


> Kickstart on Lubuntu/Ubuntu/*buntu?  Hmmm, I wasn't even aware of *buntu
> using/supporting Kickstart - but as it's a Debian derivative, wouldn't
> surprise me to find that it (also) supports Kickstart.
>
>
There's sparse documentation on its support here and there, and many forum
suggesting it can be done - but I have a suspicion it's all "hypothetical
musings." On this topic there's A LOT of unsubstantiated fluff out there,
but no solid evidence anybody has verified their writings by testing them
against an actual workflow.



> *buntu has both their "Live" DVDs, and (unless it's changed?), their
> "alternative installer" DVDs - the latter "just" use the Debian installer
> (well, basically ... that and some color scheme changed to make it look
> *buntu-ish, and descriptions changed from Debian to *buntu - but otherwise
> very much the Debian installer).  Anyway, in the land of *buntu, I don't
> know which is necessarily better supported and/or documented for doing
> Kickstart or FAI type installs of *buntu.
>

I have heard numerous times of these fabled "alternative installers". Pray
tell what are they? I only ever manage to click through to "alternative" as
in "here's something other than Unity"

Ubuntu desktop flavours uses "casper" which as far as I can tell is in
charge of loading the actual install image environment (pain in the baba,
it led me to needing to switch to NFS in the end), to launch the "ubiquity"
installer, which supports some Debian options, but has many more custom
ones. None of the preseed options are properly documented in an intelligent
manner, neither on the Debian installer side, nor on the Ubiquity side.

The Ubuntu Server installer uses a basic Debian installer from what I've
seen (Ubuntu Server is the only one without the required folder structure
for casper, and cannot be used in this PXE deployment of mine) and has an
alternative installer mini.iso which is just a net installer - it
specifically pulls form Ubuntu's servers so is a bit pointless in our use
case. I have documented this.


>
> And, other thing I was thinking about (part technical, part not) ...
> if I understand correctly, at persent, Partimus isn't in schools?
> (but hopes to soon be again).  As I understand it, teacher that was
> using/"supporting" it in school left, and that whomever/whatever
> came in to replace didn't want (at least at that time) to use/continue
> the Partimus/*buntu thing.
>
>
The Ascend school project fell through apparently, but that's not to say
Partimus won't need PXE in other contexts, so let's soldier on :-)


> Anyway, in regards to all that, I was thinking perhaps (and maybe on IRC,
> or ???)
> do a "post-mortem" meeting to go over what did - and didn't work - in that
> environment, and what we can (at least feasibly) do to improve future
> successes (e.g. not only get in more schools again, but be the great
> resource
> that neither schools nor teachers would want to see dropped or removed from
> their schools).
>

Would be interesting to do. It would mean morning time for you and evening
time for me if I'm to join though... we could try to use a Mumble setup for
voice if people are averse to Skype; I'm looking into self-hosted chat
solutions at the moment so could provide a throw-away server for this
purpose.

My take is - we hoped (I hoped) that the PXE setup would be a matter of a
couple months work, and it would be deployable in time, but since it
wasn't, the school probably wasn't interested in waiting. In a sense, *mea
culpa* for not staying on top of managing the time frames etc, but one man
across the pond can only do so much.

But I would certainly like to hear what discussions went down regarding the
interaction with Ascend and the decision to move away from the intended
Lubuntu deployments, and what would help push Partimus's solution in future
for other projects.

For this project, ideally I'd like to see something demo-able to
prospective "clients", and if it is limited to deploying official *buntu
flavours *ab initio*, so be it. Something like, here's two boxes we bring
in, look this one has PXE, and this other one is booting off of it, use any
Ubuntu desktop flavour you want. Demo done.


> And, bit more (semi-)technical ... the earlier relative commitment to
> Lubuntu,
> that was mostly based upon what teacher was presently using with Partimus
> in
> classroom, and lesson plans they had developed around it, etc. - and
> didn't want
> disruptive change to occur (e.g. plans to possibly substantially not match
> what was installed and set up on the computers).  Anyway, with the
> installation
> not in school at present, perhaps an opportune time to revisit the
> question of
> what distribution.
>

I think Lubuntu was Christian's choice (confirm please?) so as to be able
to make the most of old donated hardware that the school was going to be
using.


>
> And ... just teensy bit of my thoughts/opinion on that too ...
> DebianEdu / Skolelinux - might be an *excellent* fit for schools (probably
> at least
> worth some serious investigation).  Most notably, a whole lot of what
> Partimus and
> schools would generally want, is already very well worked out for
> DebianEdu / Skolelinux - and there are lots of resource around it for
> Linux in schools
> and particularly DebianEdu / Skolelinux itself in schools.  Anyway, might
> potentially
> save us a whole lot of work (e.g. avoid reinventing the wheel ... and
> perhaps
> poorly or repeating many of the mistakes that folks have already made and
> figured
> out and fixed/improved for use of Linux in schools).  Anyway, just a
> thought on
> that.
>

The solution I put together should work for any Ubuntu-based distro. I can
give Skolelinux and others a try in the build I've done.

After I've figured out the auto-install for Lubuntu, any additional
customized preseed files for alternate distros will have to be someone
else, as well as customizing the PXE setup for other distros if needed,
cause I'm burning out on that front.... though I can assist in future as I
feel I've learnt a fair bit in this whole exercise....



>
> On the other hand - another thought ... Partimus ... Linux terminal/kiosk
> in some
> community (shelter/resource?) location.  That can be all fine and good,
> but then
> I wonder - what exactly is the Partimus mission and how does that fit in?
> And if we do DebianEdu / Skolelinux ... or even Lubuntu well customized for
> installation/use in schools - those are quite substantially different
> target
> environments, and I'd think how we'd want to configure and set up the
> systems would
> be quite different.  so, ... I might wonder, what target environment are we
> optimizing/configuring the systems for?  Targeting multiple substantially
> distinct
> environments does multiply the amount of work and resources required.
>
>
The PXE setup itself is one-off mostly per organisation/deployment; loading
a new ISO and re-running the setup script should also not be onerous.


As regards the distro, a one-size-fits-all medium even if targeting, say,
"only schools" would likely not cut it, as different classrooms will have
different goals and needs. Any kid can download a distro and slap it on a
box. We're trying to do some extra leg work here.

I've tried to keep the customization task as straightforward (and more
importantly, scriptable!) as possible for anybody coming in afterwards,
working from the start on the assumption that each deployment would
probably want its own customization. The same single resulting ISO can be
delivered over network (PXE), on DVD, or be written to USB, as any of three
"deliverables" to the end client; so a once-only production of medium per
project is possible. Unless you're doing more than one full project per
day, this should not be a problem in the slightest.

Kiosk - a custom Ubuntu with the installer removed might be the way to go,
this can simply be served over the network by PXE server and voila presto.
I can also have a go at trying to serve an image like Porteus or
Webconverger over the solution as it currently stands.


> Anyway, good work on the PXE etc. stuff


Thanks :-)


> most of that quite applies pretty much
> regardless of the specific distribution and target environment


Trust me on this one - nope.

PXE setups are increasingly distro-specific due to bootstrap binaries
behaving slightly differently all over. I have been through this pain first
hand, hence the delay and frustrations,  and have painstakingly worked to
resolve them and as fully as possible.

Debian PXE requirements will be different from Ubuntu PXE requirements, I'm
95% certain, enough at least for tweaking to require more than passing
attention; CentOS booting is different enough (had to throw out some of my
CentOS-based methods to get the Ubuntu method working), and I can imagine
other independent distros will have their own many quirks to work around.

As said, I'll have a go using the PXE setup script as it currently stands
to deploy Skolelinux, Webconverger and Porteus, as well as CentOS, and I'll
see how that goes and let you know. If it turns out to be as easy as you
think I will be VERY HAPPY.


> - I think the only
> especially major difference may be Kickstart "vs." FAI - and even there
> there's
> probably a fair bit of overlap.  I'm hoping to have some time to test it
> out,
>

Please do - for the PXE-only part, aside from downloading the ISO, and
setting up a test environment (10-20min in the background hopefully,
depending on disk speeds); running the PXE setup script should take all of
30sec, at which point you should be ready to launch a PXE client machine!

Testing the Customization script would take longer, more like 30min to
customize to any significant point, then 40min or so to build the ISO
depending on the build environment.


> etc., but ... time, ... priorities, ... we shall see.  (E.g. some fair
> bits of my
> time have gotten sucked into things like keeping some LUG(s) from
> disappearing off
> the Internet ... more resources for, e.g. Partimus, to draw upon, if
> things like
> SF-LUG exist (and where's SF-LUG running?  All but the list stuff is on a
> virtual
> machine running on hardware where I live, and the list stuff is running
> out of
> Rick Moen's home - an installation which bit over a year ago I very
> substantially
> assisted in getting back on-line and operational again ... and yesterday,
> just
> applied the nudges to get BALUG's hosting back on-line (after a billing
> payment
> SNAFU by the person who's been covering that)).
>

Good luck with all of that - sounds like you've got a lot on your plate too.



A presto,

Tai



>
>
> From: "Tai Kedzierski" <dch.tai at gmail.com>
>> Subject: [Tech] PXE project - still alive
>> Date: Mon, 15 Feb 2016 13:15:34 +0000
>>
>
> Hi all,
>>
>> Happy new year!
>>
>> I have a little update on the PXE project which - thought I'd forgotten
>> about it? :-)
>>
>> The full project collaterals have been significantly reduced in number,
>> and
>> the steps are now much more guided and fully bash-scripted!
>>
>> https://github.com/taikedz/our-pxe
>>
>> At this point we have 2 things:
>>
>> *A full script* to create a custom DVD/ISO
>>
>> This requires command line knowledge, so would require a Partimus
>> technical
>> volunteer to create the custom image for the target organization. It needs
>> only be produced once per organization/release.
>>
>> *A full script* to create a fully working PXE server
>>
>> Tested in VirtualBox and working a charm.
>>
>> I'll be testing it on a set of home machines shortly over the home
>> network...
>>
>> *The last component* to make this 100% automatic is kickstart/preseeding
>>
>> This would enable a way to bypass the setup dialogs, automatically set up
>> an administrative user, and go straight to installation.
>>
>> I have been hitting hurdles here, and am waiting for feedback from the
>> askubuntu forum
>>
>> https://askubuntu.com/questions/734733/preseeding-kickstarting-lubuntu-example-or-documentation
>>
>> Once I get it figured out, and I've been banging my head against this for
>> a
>> while now, it'll be all wrapped up.
>>
>> Could anyone please give these scripts a rev and feed back to me, I'd be
>> most appreciative.
>>
>> If anyone knows anything about kickstart/preseeding Ubuntu, now's your
>> time
>> to shine and close in on that last 5%! (and yes, I have spent countless
>> hours on Google over this already :-P )
>>
>> All the best
>>
>> Tai
>>
>>
>> ===
>> Tai Kedzierski
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.partimus.org/pipermail/tech-partimus.org/attachments/20160215/4361a7ab/attachment-0001.htm>