The current behaviour of live evacuate is to naively request that all
servers on the hypervisor be evacuated. The more VMs are migrated
simultaneously, the more bandwidth is required. Once a certain number
of migrating VMs is reached, there is not enough bandwidth to cope
with the rate at which they dirty their memory. The migrations will
thus never complete.
The correct solution to this would be a whole lot of work on the Nova
side. As a stopgap measure, this patch introduces a --max-servers
option to host-evacuate-live. With this option, the user can control
the number of VMs that are live-migrated at the same time, thus
allowing the user to avoid the dirty memory scenario described above.
Change-Id: I17bad5f3253d6657fc1e6610159fc8e3921e6ea4
3d9c01b2d3