There is no such thing ...
As a workaround, you can deactivate use the topology before killing it. This ensures that Spout.deactivate() called.
If you need to call the bolted method, use Spout.deactivate() to send a "tuple of notifications" (which does not contain data to be processed) through the entire topology. And in each bolt, call your special method if a "tuple of notifications" is received.
In addition, this “tuple of notifications” should be redirected by a bolt to all its predecessors. You must make sure that the "notification tuples" are sent to all parallel executors of each bolt. To do this, use the dedicated "notification stream" and sign each bolt through allGrouping() to this pair (in addition to the usual input streams). In each bolt, you need to check if the tuple is a tuple of notifications or not (for example, through Tuple.getSourceStreamId() )
After cleaning is complete, you can permanently destroy the topology.
source share