Merge lp://qastaging/~rackspace-titan/nova/xen_vmops_rescue_fixes into lp://qastaging/~hudson-openstack/nova/trunk

Small typo on line 9: "We try use the second"

And hm. Doesn't this destroy the instance's disks? Seems like just shutting down rescue and calling destroy on all of the VDIs will remove the disk we want to keep around. I'd figure we'd still want to unplug the disk... but maybe do it after shutdown.

> And hm. Doesn't this destroy the instance's disks? Seems like just shutting
> down rescue and calling destroy on all of the VDIs will remove the disk we
> want to keep around. I'd figure we'd still want to unplug the disk... but
> maybe do it after shutdown.

Chris, Looking into this now. I don't think I had an issue with it but it probably won't hurt to try that.

Okay Chris. Looks like you were right. I can't just not unplug them. Leaving like it was causes me 'device in use' errors. We spoke about this on IRC a bit... Do you have any other ideas?

Hm. Did you try doing the unplugging/VBDs destroy after the rescue instance is shutdown? If that still gives 'device in use'.. that'd be pretty strange.

Ie, I was thinking if you restore _destroy_rescue_vbds() and just change the order in _destroy_rescue_instance() to the below, it should work?

def _destroy_rescue_instance(self, rescue_vm_ref):
    """Destroy a rescue instance."""
    self._shutdown_rescue(rescue_vm_ref)
    self._destroy_rescue_vbds(rescue_vm_ref)
    self._destroy_rescue_vdis(rescue_vm_ref)

I also noticed that _destroy_vdis() and _destroy_rescue_vdis() are very similar (I know you didn't modify them)... the former waits for the vdi destroys to complete. I wonder if those could be consolidated as a part of this, even though it's somewhat unrelated.