Slave mount point registration hardening/resiliency
(cherry-picked from https://git.opendaylight.org/gerrit/#/c/71898/)
- It's possible for a message to the master to timeout in which
case the slave mount point doesn't get registered so add retries
for the AskForMasterMountPoint message in NetconfNodeManager and
schema resolution in NetconfNodeActor. The AskForMasterMountPoint
and RegisterMountPoint messages were changed to pass the
slave/master ActorRefs in order to send a success/failure reply to
the original caller.
- In NetconfNodeActor, preserve actor safety by executing
ListenableFuture callbacks via the actor's dispatch queue
(using executeInSelf) if the callback accesses actor state.
- Hardened synchronization in NetconfNodeManager and NetconfNodeManager.
- Added UTs for NetconfNodeManager to cover the slave mount point
registration scenarios.
- Added more logging where appropriate
Change-Id: I084115e5c8945670f1d99243ea6de947a757a9b2
Signed-off-by: Tom Pantelis <tompantelis@gmail.com>
13 files changed: