Description of problem: Power-off should be almost immediate. Indeed, when calling the 'destroy' verb we immediately get an event as a result but it is ignored because before calling the 'destroy' verb we acquire a monitoring lock. So after this event is ignored, we wait for the next polling cycle that could be in 15sec in the worst case. Version-Release number of selected component (if applicable): How reproducible: depends on timing, but high reproduction rate Steps to Reproduce: 1. Power off a running VM 2. 3. Actual results: The VM status is switched to powering-down (which is also wrong because the VM is not powering down) and it takes several seconds until the next polling cycle that detects the VM is not reported from VDSM and thus destroyed. Expected results: The VM should switch to down very quick as a result of the event we get from VDSM. Additional info:
targeting to 4.0.4 since it may have high usability impact.
Verified on version: 4.1.1.3-0.1.el7