On 10 July 2010 20:36, Chris Rouch <chris.rouch(a)gmail.com> wrote:
I have a heavily customised installation (lots of locally built
rpms)
that i'm trying to install from kickstart and nfs. If I use a netinst
cdrom and pass the nfs location of the kickstart file without using
dhcp then this all works well. But If I use pxe/dhcp/tftp instead, it
gives me the "retrieving installation repo" message and then complains
about missing packages and installation groups.
If I switch the VT2 I can see that on the working installation I have
the remote nfs directory mounted on /mnt/source. However on the PXE
installation, this directory is mounted on /mnt/isodir and /dev/loop1
is mounted on /mnt/source. This latter contains a completely
different set of packages. This is completely repeatable - if I
comment out the dhcpd.conf entry and restart dhcpd everything works.
If I leave it in (even when booting from the cdrom) then it always
fails
I'm seeing this from a vmware client, but I also see the PXE error
from a real laptop.
How do I stop this happening?
I'd put the dvd.iso file in the top level directory. Anaconda
"helpfully" decided that this, rather than the underlying filesystem
was what I really wanted to use. I'm not sure if this is a bug or user
error. in any case I have a working installation system now that i've
moved the iso somewhere else.