是否（或计划有）在流式Google Cloud Dataflow管道上执行手动扩展的方法？

If you are running a streaming pipeline on Google Cloud Dataflow, is there (or is there planned to be at some point in the future) a method of manually scaling the number of requested worker instances up and down based on some form of custom metric or API call? Or does such scaling require performing a clean shutdown of the pipeline, a difficult task in and of itself even in the most well-controlled circumstances, and then restarting it completely with a different number of instances?If you are running a streaming pipeline on Goog