This is a dedicated watch page for a single video.
You're managing a streaming Cloud Dataflow pipeline. Your engineers have developed a new version of the pipeline featuring a different windowing algorithm and triggering strategy. You aim to update the running pipeline with the new version while ensuring that no data is lost. What is the recommended approach?