[Gllug] Problem upgrading disk-less system to Etch

[John Winters]

I have a little server (EPIA 5000) which runs very happily in a corner
doing several small tasks.  The only thing is it's running Sarge and
it's probably time it was upgraded.

I took a copy of its root directory and used a similar box to do a test
upgrade to Etch.  All seemed to go well, including booting a custom
2.6.18 kernel but I have one problem.

When the system gets to the part of the boot sequence where it would
configure eth0 it fails.  It brings eth0 down and then gets totally
confused because it can no longer access its root disk and so can't
access the utilities which it needs to bring the link up again.

Now I can work around this by commenting out these two lines:

auto eth0
iface eth0 inet dhcp

in /etc/network/interfaces.  If I do that then the boot sequence runs
just fine.  However this has a knock-on effect.  Because eth0 is not
brought up as part of the boot sequence there is no "interface up" event
and so the mountnfs script in ifup.d never gets run.  This means that
the /home partition (on an nfs server) never gets mounted.

Clearly I can work around this too by putting a "mount -a" in rc.local
but it seems to me that this is getting too messy.  Surely Etch should
be able to boot on a disk-less workstation?

I've looked at the documentation of /etc/network/interfaces but there's
no mention of settings which would stop it bouncing the link.  Does
anyone know of a cleaner way of getting it to work?

TIA,
John

P.S.  Sarge boots just fine with those lines in place.
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug