mxm opened a new pull request, #710: URL: https://github.com/apache/flink-kubernetes-operator/pull/710
The state store implementation should take care of optimizing flush operations. That is already the case for the Kubernetes-based implementation. The in-memory state store does not require flushing. The current logic tries to save on API calls but there is no actual savings because the ConfigMap needs to be retrieved at least once to check whether its empty. So instead of checking for emptiness, we can simply clear it. Further, I've removed the optimization to keep some of the data when disabling the autoscaler. The metric collection contract requires that the entire metric window contains data. By removing some of the data, once turned on again, we could add metric gaps to the metric history if there are still metrics left from the last metric window which overlap with the current. The reason is that the metric window is counted from the oldest timestamp found in the metric history. -- 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
