Bug 1649620

Description of problem:
Need a playbook for migration of osd objectstore from FileStore to BlueStore for 4.0.

This playbook should have the ability to do full OSD nodes as well as a single OSD.

Playbook should check cluster for availability capacity to ensure that when a node is brought down and backfilled to cluster there is enough avail space to perform the backfill.

Steps Playbook can take:

1. Se osd_objectstore to Bluestore if set to FileStore. 

2. Check number of OSD nodes in the cluster decide on full node migration of single OSD migration. If under a certain number of nodes (4 for example) in a CRUSH root then will default to single OSD migration at a time.

3. Check recovery/backfill setting and validate throttled. Defaults should be fine, but should verify these are set to default.

4. Drop OSD (or OSD's) and mark out.

5. Monitor cluster for complete backfill of the data (PG's Active+Clean).

6. Remove OSD's, await active+clean as will likely see repeer due to map changes (possible).

7. Redeploy OSD (OSD's) on lvm and Bluestore.

8. Await active+clean.

9. Move to next OSD (OSD's)


Version-Release number of selected component (if applicable):
4.0