How to use Nova or download instances with no controller

asked 2019-01-04 08:54:44 -0500

schap gravatar image

Hello so we had a Mitaka controller that became hosed when we tried updating things. So now we have no controller. All of our instances are under /var/lib/nova/instances on the compute.... How can I download them without being able to source my environment and not being able to use nova on the compute? Any suggestions? Nova list asks for all of the information but I cannot source anything bcause we do not have a controller anymore. We would like to move all of the instances to a new Rocky openstack cluster.

edit retag flag offensive close merge delete


It depends on backend storage type you are using. If it is located on external cinder storage you there is nothing to in compute nodes, just find them and move over. First of all you should find the disk files.

Am Ab gravatar imageAm Ab ( 2019-01-05 00:37:12 -0500 )edit

Assumption: Your hypervisor is KVM, your VM disks are qcow2 files. Example: Shut the VM down with virsh. Use qemu-img to convert the disk image to a single qcow2 (it might have a base image) or raw image. Use the result to launch another VM in a different cloud.

Bernd Bausch gravatar imageBernd Bausch ( 2019-01-05 05:01:45 -0500 )edit

So it looks like it is already a qcow 2

# qemu-img info 08cf937f-6054-4b3f-8660-7fe63d2e7f95/disk
image: 08cf937f-6054-4b3f-8660-7fe63d2e7f95/disk
file format: qcow2
virtual size: 40G (42949672960 bytes)
disk size: 1.9G
cluster_size: 65536
backing file: /var/lib
schap gravatar imageschap ( 2019-01-05 07:07:40 -0500 )edit

How do I download that file? And Do I need to move anything else over with it that is under the instance directory? Like, or libvirt.xml? Or do I need the backing file?

schap gravatar imageschap ( 2019-01-05 07:09:39 -0500 )edit

Can I move the whole instance folder over to another compute and resync database some how?

schap gravatar imageschap ( 2019-01-05 07:35:32 -0500 )edit