This is the next step towards disaster recovery for rbd. The idea is to record all writes (data and metadata changes) to a journal of rados objects, preserving a consistent point-in-time stream that can be mirrored to other sites.
Jason Dillaman (Red Hat)
Josh Durgin (Red Hat)
A new feature bit is added for enabling journalling for an image, and a general journaling mechanism is almost ready.
Updates since then: active sets of objects are switched out wholesale once any current object becomes full, so that replay is bounded to an active set of objects as a maximum necessary amount of data to read at once.