suspending a paused instance

[Loganathan Parthipan]

Is there a compelling reason why we don't support suspending a paused instance? At the moment we only allow 'active' and 'rescued' states to be suspended.

In compute/api:

@check_instance_state(vm_state=[vm_states.ACTIVE, vm_states.RESCUED])

Trying to suspend a paused instance results in:

nova suspend b10f8175-1663-41b2-8533-0b0606f369ff
ERROR: Cannot 'suspend' while instance is in vm_state paused (HTTP 409) (Request-ID: req-7349d554-ff48-4155-a62b-967f0813c59c)

I haven't tested with other hypervisors, but as far as libvirt/kvm is concerned it suspends (virsh managedsave) both states ('running', 'paused') and resumes (virsh start) a suspended instance to the pre-suspended state.

A use-case for this is to be able to quickly pause a bunch of instances for security reasons and then suspend one of them conveniently using the API rather than virsh. Taking a suspicious vm to ‘active’ state so that it can be suspended doesn’t seem like a good idea.

[John Garbutt (johngarbutt)]

Seems like a reasonable relaxation of the state checking in the compute manager.

It would be good to see if this causes issues on other hypervisors, I guess not.

[Loganathan Parthipan (parthipan)]

Also as part of this we should make sure that on resume nova sets vm_state to the correct state based on the power_state. Right now it assumes a successful resume to result in ACTIVE. But if it was suspended from PAUSED, at least as far as libvirt/kvm is concerned it will come back to 'paused' instead of 'running'. So we might need to do something like

+++ b/nova/compute/manager.py

         instance.power_state = self._get_power_state(context, instance)
- instance.vm_state = vm_states.ACTIVE
+
+ if instance.power_state == power_state.PAUSED:
+ instance.vm_state = vm_states.PAUSED
+ else:
+ instance.vm_state = vm_states.ACTIVE
         instance.task_state = None