Driver concept

Unsolved problems

The nova scheduler unaware of available storage space
- Instances may repeatedly not be created/migrated
- Custom Scheduler filter could take available space into account
Volume not attachable if it exceeds available space on instance’s host
- No transparent communication to user
- Only mitigatable on operational level
  - Monitor disk space and extended or move instances once threshold is reached
  - Limit volume size for this driver
    - per_volume_gigabytes not available per volume type
Some volume actions will fail while a volume is being hydrated, e.g.:
- Attaching the volume on another host (e.g. attachment/detachment to/from another instance)
  - would require chaining of multiple clone targets
- Extending volume
  - Unsupported by device mapper clone target
- live-migration
- …
- No transparent communication to user
- Set volume to maintenance until hydration finishes?
  - Cinder sets status after create/attach
  - Detachment would not be allowed, although technically possible
    - expectation is that detachment is always possible, e.g. server rebuild (?)
- Currently volume is set to maintenance on first unsupported action