Cinder Enhancements for Replication (and more) using Stateless Snapshots
Post-summit proposed plan:
We will require a consensus on how to specify the target of a migration/replication/passive-end-of-a-mirror. This is already under discussion. We prefer to have the Volume Driver return an array of targets, but could work with any of the alternatives that have been discussed, but we need one to be picked.
- Refactor volume migration to use snapshot replication.
- Support High Availability Mirroring using snapshot replication.
- Support a "roll-your-own" Warm standby solution where snapshot replication can prepare a target for rapid failover without datacenter admin support.
Leftovers from the pre-summit description: