Comment 25 for bug 1246236

> When we start transferring the kernel over, the machine on the other side stops responding

This is indeed what seems to be causing the problem. Note that we're seeing (a worrying number of) retries in some of the earlier (and successful) transfers.

What kind of bridge is being used here? (virbr0 -the bridge created by virsh-, br0, ?). Can you please paste the characteristics of the bridge (by running something like "cd /sys/class/net/virbr0/bridge; for f in `ls *`; do echo -n "$f "; cat $f; done).

As described on https://bugs.launchpad.net/maas/+bug/1155556, the default timeout behavior might be the problem here. Can you try changing the RemoteOriginReadSession class in /usr/share/pyshared/tftp/bootstrap.py: change the timeout = (1, 1, 1, 1, 1, 1) to timeout = (1, 3, 7). and then restart the TFTP server?