mxm commented on PR #728: URL: https://github.com/apache/flink-kubernetes-operator/pull/728#issuecomment-1864824444
> @gyfora @mxm by the way, I'd like to discuss the logic of scaling effectiveness evaluation with you. > > 1. Now it's controlled by two config `scaling.effectiveness.detection.enabled` and `scaling.effectiveness.threshold` and we evaluate the effectiveness under the condition `last scaling is scale up` and only refer to the last scaling effectiveness. > 2. Image the following scenario: scale up double parallism first, then scale down to 0.8 parallism, then scale up double scale down 0.8, the effectiveness detection will be invalid in this scenario, even scale up is ineffecive, we'll continue scale up > 3. Maybe we can add a new config like `scaling.effectiveness.history.reference.num` and set a default value, then we can evaluate based on the last `scaling.effectiveness.history.reference.num` scale up summaries. > > Looking forward to your reply. What yo you describe is an edge case we hadn't considered. We were more concerned about a continuous increase in parallelism. If there is any scale down we are currently assuming that the algorithm is not completely broken. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
