Hi Derek,

When you select the "Create new volume" in Horizon (assumption), there is a cinder volume created which contains the image (bootable) and the instance is backed by it.

When you perform the snapshot creation operation on that instance, a cinder snapshot of that volume is created and a glance image is registered pointing to it.

Based on the scenario you described, the cinder snapshot operation seems to be failing. I would suggest you to check the following:

1) Which OpenStack version are you using?

1) Which cinder backend are you using?

2) Checking the nova logs and cinder logs (api, sch, vol) for any possible errors

3) Try to create a cinder volume, attach it, and try to snapshot it (Note that we've removed the requirement of a force flag since xena so you will need to provide force=True if you're on a pre-xena version).

We've faced a recent issue related to snapshot create operation[1], maybe that's related.

[1] https://bugs.launchpad.net/python-cinderclient/+bug/1995883

Rajat Dhasmana

On Thu, Nov 10, 2022 at 3:46 PM Derek O keeffe <derekokeeffe85@yahoo.ie> wrote:

Hi all,

When we create an instance and leave the "Create new volume" option as no then we can manage the instance with no issues (migrate, snapshot, etc..) These instances are saved locally on the compute nodes.

When we create an instance and select "Create new volume" yes the instance is spun up fine on our backend storage with no obvious issues (reachable with ping & ssh. shutdown, restart, networking, etc.. all fine) however, when we try to snapshot it or migrate it it fails. We can however take volume snapshots of volumes that we have created and are stored on the same shared backend.

Has anyone came across this or maybe a pointer as to what they may think is causing it? It sounds to us as if nova try's to create a snapshot of the VM but thinks it's a volume maybe?

Any help greatly appreciated.

Regards,
Derek