Description of problem: Having the instances running when detaching volume generates issues as described in these BZ: bug 1548053 bug 1546826 Version-Release number of selected component (if applicable): openstack-heat-common-9.0.1-3 How reproducible: 100% of the time Steps to Reproduce: 1. Create a stack with running instances 2. Delete stack with running instances Actual results: If the volumes are in use in the instances, libvirt is unable to cleanly detach them, thus failing the stack-delete Expected results: stack-delete shouldn't fail Additional info: If heat would shutdown the instance before, all the volume detachment would work
So IIUC this is believed to be a regression in libvirt, but you want Heat to work around it? Seeing the template would help. Are you using a separate VolumeAttachment resource, or attaching it directly to the server? If the former then there's definitely nothing Heat can do. If the latter then it might be possible, I guess, but I'd still be concerned that it's a fairly substantial behaviour change to address a problem that is not of Heat's making, and which will in any event continue to exist for any users of Cinder that either don't use Heat or use the VolumeAttachment resource.
Created attachment 1404407 [details] vm with volume as root disk