When the Cinder volumes for tools-db-1 and tools-db-2 were created in T329970 and T329521, they were created using Cinder snapshots (which in turn use Ceph RBD snapshots). This means we now have a volume chain of copy-on-write layers that cannot be easily flattened:
```
tools-db-basevolume1 -> tools-db-1-basesnapshot1 -> tools-db-1-basevolume2 -> tools-db-1-basesnapshot2 -> ( tools-db-1 | tools-db-2 )
```
The volumes `tools-db-1` and `tools-db-2` are the ones currently in use, but they are both layers on top of tools-db-1-basesnapshot2 and its parent layers. All the parent layers cannot be deleted because they are part of the same copy-on-write volume chain.
This should not cause any issue, but it would be cleaner to remove this long layer hierarchy. The easiest way I can think of is as follows:
1. Create a new snapshot of `tools-db-1` and name it `tools-db-1-tmpsnapshot`
2. Create a new volume based on that snapshot and name it `tools-db-1-fromsnapshot`
3. Create two new empty volumes `tools-db-3` and `tools-db-4`
4. Rsync all data from `tools-db-1-fromsnapshot` to `tools-db-3`
5. Create two new instances `tools-db-3` and `tools-db-4` and attach the new volumes `tools-db-3` and `tools-db-4`
6. Set up `tools-db-3` to replicate from `tools-db-2`, and set up `tools-db-4` to replicate from `tools-db-3`
7. When they are in sync, promote `tools-db-3` to become the new primary
8. Delete the old instances `tools-db-1` and `tools-db-2`
9. Delete the snapshot `tools-db-1-tmpsnapshot` and the volume `tools-db-1-fromsnapshot`