libvirt: hard reboot incorrectly create nova/instances/<uuid>/<uuid>_disk with lvm backend
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Confirmed
|
Medium
|
Unassigned |
Bug Description
Configure Nova compute to use lvm backend, spawn an instance based on glance image, under folder nova/instances/
Steps to reproduce under devstack:
* On compute node, create LVM volume group "image-lvm-local"
* On compute node, modify nova config file under [libvirt] to enable lvm backend with
images_type = "lvm" and images_volume_group = "image-lvm-local"
* Start an instance no file nova/instances/
* Do hard reboot on the instance and nova/instances/
Changed in nova: | |
assignee: | nobody → vu tran (vu-tran) |
tags: | added: libvirt |
Changed in nova: | |
assignee: | vu tran (vu-tran) → nobody |
I just checked the path of hard_reboot in the libvirt driver, and it shares the same logic with pre-live-migration for creating the images.
By looking at _create_ images_ and_backing( ) helper method, I think the logic is wrong by verifying the backing file and not using the imagebackend.
So, I think the problem is for both API actions (live-migrate and hard reboot).