F19 source and destination host (I'm using a VM for the destination but I don't think that matters). Both hosts have qemu-1.4.2-1, which has the migration cleanup series from qemu-1.5 backported, including this commit which sounds like it could have been related: commit 45d723c48255328fbd5128c4c70c91103e420590 Author: Michael Roth <mdroth.ibm.com> Date: Thu May 16 16:25:44 2013 -0500 Revert "migration: don't account sleep time for calculating bandwidth" I'm migrating with: sudo virsh migrate --p2p --tunnelled --verbose --live --copy-storage-all f19-migrate qemu+ssh://root.123.99/system After progress meter reaches 99% the migration fails. This is in the destination qemu log: Receiving block device images qemu: warning: error while loading state section id 2 load of migration failed Full qemu command line and logs coming up.
Created attachment 753190 [details] qemu log on the source host
Created attachment 753191 [details] qemu log on the destination host
FWIW I was more or less following the steps here: https://fedoraproject.org/wiki/QA:Testcase_Live_Migration_using_libvirt/virsh
Can you try without --p2p and --tunnelled?
Yeah I think this was some side effect of libvirt silently falling back to the old storage migration method, which current F19 libvirt doesn't do anymore. Since we don't really care about that method, closing this as WONTFIX. I'll reopen if I can reproduce with the nbd migration method.