Bug 1550704

The current v2 driver uses a journaling DB table/threads in order to handle the possibility of out-of-order events occurring from different neutron workers to update the same/dependent resources.  There can also be out-of-order updates/events if there is a temporary network outage to one of the OpenDaylights or an OpenDaylight serving a request goes down.  There are also issues with journaling performance in cases where there are high number of journal threads or pending transactions.

This RFE proposes a v3 driver which will abandon the journaling design for a revision based synchronous system.  Where every update to a resource increments a revision number associated with that resource.  Therefore out of order updates are handled by ODL as it compares the revision number of incoming updates and there is no need for the networking-odl driver to queue transactions in a journal.  There will still need to be some type of retry mechanism in networking-odl in case ODL detects that a revision is too old of an incoming request and rejects it.  There will also still need to be the ability to offer a full-sync of the database between networking-odl and ODL in order to copy any missing out-of-date revisions to the ODL DB.