Bare metal volumes
It would be nice to be able to work with the locally attached storage on bare metal nodes in a similar fashion to working with volumes for virtualised instances. Snapshotting would be hard, but a first cut of being able to reserve persistent storage on a bare metal node would enable the same separation of persistent data and image content that is possible with virtual instances.
Current thoughts:
- creating a bare metal volume creates a uuid to return and a reservation in the bare metal system
- nova boot --with-volume will combine a volume reservation with the cpu and memory constraints from the flavor to decide what bare metal node to put the volume on
- change the nova bare metal deploy code to consult with cinder for such volumes and only permit an attached instance to land on that node
- deploy will only wipe the disks on the first deploy to a node with a given volume
Blueprint information
- Status:
- Complete
- Approver:
- John Griffith
- Priority:
- Low
- Drafter:
- Robert Collins
- Direction:
- Approved
- Assignee:
- None
- Definition:
- Obsolete
- Series goal:
- Accepted for future
- Implementation:
-
Not started
- Milestone target:
- None
- Started by
- Completed by
- Sean McGinnis
Related branches
Sprints
Whiteboard
(EMH) Is this supported by the block_device driver?
2015-10-4 (smcginnis): I believe this is covered by the block_device driver as suggested. Marking as obsolete to clear it out. If something is still needed here please let me know.
Work Items
Dependency tree
* Blueprints in grey have been implemented.
