Conclusion: The live migration completed successfully in both cases, and it looks to me like the cpuset attribute was updated correctly to match the destination host policy, but perhaps I've misunderstood & failed to simulate a case that would trigger a migration failure or policy conflict.
Hi folks:
As a learning exercise[1], I tried to reproduce this bug.
Here are my notes from two test cases I tried using a multi-node devstack sandbox.
1. Source host vcpu_pin_set is larger than the destination host vcpu_pin_set:
http:// paste.openstack .org/show/ 473274/
2. Source host vcpu_pin_set is the same size as the destination host vcpu_pin_set, but with completely different values:
http:// paste.openstack .org/show/ 473286/
Conclusion: The live migration completed successfully in both cases, and it looks to me like the cpuset attribute was updated correctly to match the destination host policy, but perhaps I've misunderstood & failed to simulate a case that would trigger a migration failure or policy conflict.
Thoughts?
[1] I'm new to OpenStack.