Hello team, I have created a Jira issue and a PR to fix Jemalloc misconfiguration that causes higher memory utilisation and sometimes OOMKills. This primarily impact pods with CPU limit configured to be lower than host machine's CPU count. In my verification, the fix reduces average process memory usage by 13.7% and improves average throughput by 4.4 %. The Jira and PR have reproduction and verification details. PR: https://github.com/apache/flink-docker/pull/266 Jira: https://issues.apache.org/jira/browse/FLINK-39924 Would appreciate reviews here. I will follow up with back-port PRs if approved. Thank you