kvm hypervisor host network problem during VM startup/shutdown

[Stephen Carpenter <sjc@carpanet.net>]

(please CC me on replies)

I have been running kvm on this host for almost two years now. Using
bridged networking, rather simple setup, with about 4 VMs and the
hypervisor.

There is one bridge, br1 (there is an unused br0 which has no connection
out) which is connected to the physical ethernet, one of the VMs is
providing DHCP to the whole network.

Things work fine, except, when I attempt to create a new host, or
destroy a running host. I am playing with FAI profiles now, and so have
"tested" this many times. Every time, the hypervisor briefly (5-10
seconds or so) stopps responding on the network, though, other systems
which use the same bridge, have no issue.

Script used to start the new VM:
virt-install \
        --connect qemu:///system \
        --pxe \
        -n test \
        -r 512 \
        --vcpus=1 \
        --disk path=/dev/vg0/test \
        --vnc \
        --noautoconsole \
        --os-type linux \
        --os-variant debianLenny \
        --accelerate \
        --network=bridge:br1 \
        --hvm

I run this and get:
Starting install...
Creating domain...                                                 0 B
00:00
Domain installation still in progress. You can reconnect to
the console to complete the installation process.
Timeout, server not responding.

Then I am back on my local host (I ssh in to the hypervisor). I connect
back in, and I can now use virt-viewer to see the console and watch the
install (already in progress).

The same thing happens when I "destroy" test via virsh (why graceful
shutdown a system being blown away?)

I have found that other systems (like the DHCP VM) still respond to ping
during this time. In fact, one of them is providing the VPN end point
which I am using to get to the hypervisor, and the VPN connection never
drops.

In the logs, at the same time, I see only the following, though, looking
around indicates that these are common and probably not indicitive of an
error:
May 10 15:27:38 hyper kernel: [492763.100101] device vnet5 entered
promiscuous mode
May 10 15:27:38 hyper kernel: [492763.104224] br1: port 6(vnet5)
entering listening state
May 10 15:27:47 hyper kernel: [492772.133560] br1: port 6(vnet5)
entering learning state
May 10 15:27:56 hyper kernel: [492781.146314] br1: topology change
detected, propagating
May 10 15:27:56 hyper kernel: [492781.150297] br1: port 6(vnet5)
entering forwarding state

Bridge is setup in /etc/network/interfaces as follows:
auto br1
iface br1 inet static
   address 192.168.0.6
   network 192.168.0.0
   netmask 255.255.255.0
   gateway 192.168.0.1
   bridge_ports eth100
   bridge_hello 2
   bridge_fd 9
   bridge_stp on

Anyone else run into this? Any ideas?