Zack Perry <zack.perry <at> sbcglobal.net> writes:
If I really want to see what's going on, I would run tcpdump on
the
node that runs cobbler, and records the traffic from the inception
of the PXE booting process, save it to a file, and review it using
wireshark UI (on the same node or your desktop). That way, I can
use wireshark's filter to do drill-down whenever I need. Just a
suggestion. You may prefer to use tcpdump directly. Some do.
I did exactly that right after posting my last message :) I found some
interesting stuff.
tcpdump saw no packets at all on eth1 until I backed out of the eth0
configuration and selected eth1 as the install device. I see the following in
the cobbler-generated PXE configurations for BOTH interface MAC addresses:
label linux
kernel /images/CentOS_5.8-x86_64/vmlinuz
ipappend 2
append initrd=/images/CentOS_5.8-x86_64/initrd.img ksdevice=bootif
lang= kssendmac nostorage text ks=http://<CBLR_IP>/cblr/svc/op/ks/system/kvm2
Then I realized I actually used koan to install a grub boot record on this host
(!), so it wouldn't use PXE at all. I guess the question then is why koan
wouldn't pick up the boot device from the Cobbler "Management Interface"
property, rather than using ksdevice=link. I _think_ that getting that set
correctly might just solve my problem!