Unnecessarily use of a list and extra loop call to offload the shelved instances whose shelved time passed the shelved offload time

[lucky]

Unnecessarily use of a list and extra loop call to offload the shelved instances whose shelved time passed the shelved offload time.

To offload the shelved instances whose shelved time passed the shelved offload time, first list is created and then this list of shelved instances is processed in extra loop. Instead of creating a list of offload ready shelved instances and processing them later, same can be done in one loop.

Code should be refactored. Instead of creating a list of offload ready shelved instances and processing them later, same can be done in one loop. Instances can be offloaded the time when their shelved time is checked. (Reference File: nova/compute/manger.py, Function: _poll_shelved_instances, Line No: 5847)

[Matt Riedemann (mriedem)]

So this:

https://github.com/openstack/nova/blob/0e48e7b1d2ee914152958d10f9725aa86daafbb6/nova/compute/manager.py#L6809-L6812

[Matt Riedemann (mriedem)]

Well, wait. In looking at this, I assume it's done in two loops because the first loop is time-sensitive in collecting the list of instances that should be offloaded, the 2nd loop then offloads those instances from the first loop. If we merge the loops, we could end up shelving things that we might not before because the offload operation slides the window. So I'm going to mark this as invalid.