libvirt image cache manager does not handle shared storage in folsom

[Joe T]

I have found that the image cache manager is deleting needed images under _base when the following scenario happens:

1. Using shared storage
2. All instances (1 or more) of a certain image or snapshots are running on one compute node
3. That compute node is taken down for 10 minutes or so (reboot, maintenance, etc)
4. Other compute nodes see that the _base file is not in use and deletes it
5. The compute node comes back online and the instance(s) running on that compute node can no longer start due to the _base image being removed.

Please let me know if you need further information or clarification on the above.

Thanks,
Joe

[Michael Still (mikal)]

Was this with grizzly?

[Joe T (joe-topjian-v)]

This was in Folsom.

[Michael Still (mikal)]

The storage topology detection code wasn't back ported to folsom, it was deemed to be too big by the reviewers at the time. We instead just disabled automatic cleanup. However, grizzly has this code enabled. I'd be interested in results of testing with grizzly, but unfortunately this folsom bug is unlikely to get fixed at this stage.