Description
Currently a request to force-stop one or more datafeeds ignores datafeeds that are in the stopping state. The rationale for this is that the datafeed will soon stop by itself. However, there is a situation where this will not happen: when the node that the datafeed was running on is no longer in the cluster. A datafeed can only stop "normally" by redirecting the request to the node on which it is running, so if this doesn't exist then it gets stuck in the stopping state.
Certainly for stopping datafeeds that are unassigned or that have stale assignments, force-stop should remove the persistent task.
Possibly the logic for force-stop could be changed so that persistent tasks are removed unconditionally for all datafeeds listed in the request, regardless of current state. But this requires a little more thought.