[libvirt-users] gPXE fails automatically, succeeds with manual autoboot

[Paul Lussier]

Hi Folks,

I'm trying to auto-create a VM using a kickstart server and gPXE with
my VMs.  The underlying physical host is CentOS 6.0 running
libvirt-0.8.1-27, and libvirt-python 0.8.1-27, virt-manager 0.8.4-8.

The underlying network config on the physical host is:
# ip -o -4 addr show
1: lo    inet 127.0.0.1/8 scope host lo
6: vnet0    inet 10.8.31.61/23 brd 10.8.31.255 scope global vnet0
7: vnet1    inet 10.0.4.249/24 brd 10.0.4.255 scope global vnet1
8: virbr0    inet 192.168.122.1/24 brd 192.168.122.255 scope global virbr0


I'm attempting to create the new VM with the following command-line:

# /usr/bin/virt-install -n test -r 2048 --vcpus=2 --disk
path=/var/lib/libvirt/images/test.img,size=25 --pxe --os-type=linux
--os-variant=rhel6 --network bridge=vnet0,mac=52:54:00:07:51:1e
--autostart

So, I'm trying to create the VM with one NIC, bound to the vnet0
interface on the physical host.  The physical host is both the
PXE/DHCP server (using cobbler) and the kickstart server.

When I run the above virt-install command, while simultaneously
running 'tcpdump -i vnet0 -n port 67 or port 68' on the physical host,
the VM powers on, boots up, and attempts to PXE boot.  However, I
never see any traffic on the physical host's vnet0 interface, nor does
the DHCP server log any requests.  If I boot the VM, then hit Ctrl-B
to drop into gPXE's prompt and type autoboot, the first invocation
fails, but the second one succeeds.

Has anyone seen this kind of behavior ?  Is this indicative of a
configuration issue on the physical host? An underlying network
problem ? Or am I missing an option on my command-line ?

I'm a little stumped here, and hoping someone can point me in the
right direction.  I'm more than happy to RTFM if someone can tell me
which M to F'ing R :)

--
Thanks
Paul