[debian-lan-devel] disk size and other issues

Andreas B. Mundt andi.mundt at web.de
Wed Feb 18 13:55:20 UTC 2015


Hi Ross,

thanks for your comments.

On Tue, Feb 17, 2015 at 08:08:58PM +0000, Boylan, Ross wrote:
>
> DISK SIZE
> It would be helpful if the amount of disk space required for various
> installations was documented somewhere; I poked around and couldn't
> find it.  As a result I had several false starts trying to do a
> mainserver install (combined) in a VM.  Oddly, although it said it
> needed about 10G, even 11G was insufficient and I had to go with 12
> before things worked.

Yes, the size went up a bit from release to release.  Now in jessie, I
use 16GB for the main server.  I added this at the bottom of:

    https://wiki.debian.org/DebianLAN/bootstrap/jessie

>
> I used http://www.das-netzwerkteam.de/~andi.mundt/debian-lan_wheezy_20140302_amd64.iso.

Oh, I should add the image for jessie.

> Maybe virtual box, which I was using, does disk sizes in powers of
> 10 while the installer does powers of 2?  Even so, there seems to be
> some unaccounted for overhead.
>
> I was also a bit puzzled about how disk size was handled after
> seeing the talk.  You mentioned that each class can have its own
> disk partition section, but that seems to be something that can only
> occur once.  Yet the necessary size depends on all the classes
> selected.  How does that work?  Do you adopt a convention of only
> specifying the disk layout in the classes that bring all the others
> together, e.g., SERVER_A?

Yes, it works in the following way:  If a machines belongs to several
classes that define a disk layout in disk_config/ (like the mainserver
which belongs to FAIBASE, FAISERVER and LVM8_A), then the last class
listed in class/50-host-classes takes precedence.  For the mainserver,
this is LVM8_A.

So it's not that a new disk layout is computed from several classes,
only the last one listed is taken as it is.

> WEBSITE
> https://wiki.debian.org/DebianLAN links to
> http://penta.debconf.org/dc13_schedule/events/962.en.html.  The
> first link at the bottom, under attachments, is
> http://penta.debconf.org/dc13_schedule/attachments/265_DebianLAN.pdf,
> which gives me a not found error.  The last link on the page, to
> slide (http://people.debian.org/~andi/Vaumarcus2013.pdf) does work,
> and I suspect is substantially the same material since it matches
> what was in the .ogv files.

Yes, it's the same.  There is a bug in penta.debconf.org, so I added
the slides again at http://people.debian.org/~andi

> INSTALLATION FAILURE
> After I ran through the installation there were some error messages
> (can't recall the details) and when I rebooted no boot loader came
> up.  I'll try again to get a better report, but I know my network
> setup was screwy.  Specifically, the machine was running on a
> virtual LAN in the 192.168.... range, but I know the machine gets
> set up with 10... as the IP.  Also, the machine is set up to be a
> DHCP server, but there are already 2 DHCP servers on the LAN: one
> being run by virtual box for the LAN, and the other on a virtual
> machine on the LAN (as the result of installing debian-lan-config on
> it.  Could these account for the failure?

I can explain how I usually test the setup (combined setup, mainserver
includes gateway).  The first thing you need is an isolated virtual
network with no DHCP server connected.  I use virt-manager and kvm,
and I add a network 'intern' which shows up as virbr1 and in the GUI
as:

'Network 10.0.0.0/16'
'DHCP range: Disabled'
'Forwarding: Isolated network internal and host routing only.'

This network 'intern' is going to be my Debian-LAN network.  Now I
create a VM 'mainserver' with two nics (both connected to the virtual
lan 'default' for now which is provided by libvirt as virbr0 with DHCP
and NAT to the host's network).  I install the mainserver.  After that,
I check which nic has the static IP address 10.0.0.1 and I switch that
virtual card to the 'intern' network.  The rebooted mainserver serves
now as router from 'intern' (10.0.0.0/16) to 'default'
(192.168.0.0/16).

Now, I create another machine 'workstation00' which I connect to the
'intern' network.  If I pxe-boot that machine, it is served by the
mainserver.

I hope this helps.  Many thanks again for your pointers!

Best regards,

     Andi



More information about the debian-lan-devel mailing list