We want to get accurate results from the rest of the upgrade test and we know about this bug, so let's make it a soft fail so we can see how the test is functioning otherwise. Signed-off-by: Adam Williamson <awilliam@redhat.com>