Description of problem: create a volume with auth.allow set to <ip_address/subnet> of client nodes. Perform I/O operations from the mounts on client nodes. set auth.reject <ip_address/subnet> for same client nodes. Since the client has mounted to volume before the auth.reject option is set, client nodes are still able to access data and perform operations on data from the mount points. Ideally, the auth.reject or auth.allow volume set operations should re establish the client connections.
This is the expected behavior. Antecedent connections should not be reset.
This bug has to be documented.