[ https://issues.apache.org/jira/browse/FLINK-33801?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17796082#comment-17796082 ]
Gyula Fora commented on FLINK-33801: ------------------------------------ This is basically similar to the rollback feature in the operator but with regards to throughput / processing rate. > Revert to last scaling when job becomes unhealthy after scaling > --------------------------------------------------------------- > > Key: FLINK-33801 > URL: https://issues.apache.org/jira/browse/FLINK-33801 > Project: Flink > Issue Type: New Feature > Components: Autoscaler > Reporter: Rui Fan > Assignee: Rui Fan > Priority: Major > > Revert to last scaling when job becomes unhealthy after scaling, such as: the > memory is under pressure. > In the first version, the unhealthy only for memory is under pressure. We can > improve more unhealthy cases in the future. > This Jira is created when review this PR: > https://github.com/apache/flink-kubernetes-operator/pull/726#discussion_r1422456112 -- This message was sent by Atlassian Jira (v8.20.10#820010)