But again, those claims aren't (yet) made for live migration and interface attach/detach operations. So I'm a bit confused on what you're saying is taking a long time.
Regarding the _sync_power_states periodic task I've seen a few vcenter-specific patches to deal with a large number of instances when nova-compute is managing a large vcenter cluster, e.g.:
> cause a Nova sync power process to be kicked off.
I'm not sure if you're referring to the _sync_power_states periodic task:
https:/ /github. com/openstack/ nova/blob/ ff0feed25d56c8c cd2298d5b5b82e6 36880fa986/ nova/compute/ manager. py#L7907
If you are, that's a mischaracterization of what happens when you create or live migrate a server. Since it's a periodic task, it runs like a cron.
Maybe you mean "claims" performed in the resource tracker, like during instance create?
https:/ /github. com/openstack/ nova/blob/ ff0feed25d56c8c cd2298d5b5b82e6 36880fa986/ nova/compute/ manager. py#L2223
But again, those claims aren't (yet) made for live migration and interface attach/detach operations. So I'm a bit confused on what you're saying is taking a long time.
Regarding the _sync_power_states periodic task I've seen a few vcenter-specific patches to deal with a large number of instances when nova-compute is managing a large vcenter cluster, e.g.:
https:/ /review. opendev. org/#/c/ 575034/
Also see bug 1513735.