Comment 23 for bug 1341420

Hmm... rephrased because I lost the code snippet[1].

But the same problem still exists because of the line [2], why host state should be updated when it is already updated to the db state with the same timestamp? This leads to the problem that the concurrent schedule request can not benefit from host state cache, so there are conflicts.

[1] https://github.com/openstack/nova/blob/master/nova/scheduler/host_manager.py#L162-L164
[2] https://github.com/openstack/nova/blob/master/nova/scheduler/host_manager.py#L163