Hi All, We are currently using a stateful cluster of flink in HA with 2 job managers and 5 task managers. We are facing a challenge where after restoring a job from savepoint we have to manually restart all the task managers once first checkpoint is created, if not done memory on the task managers spikes causing the process to fail at one time. Once the restart is completed the cluster if behaving perfectly. Can anyone point me on any configuration changes related to GC or flink which can be done to make sure after restoring from savepoint and creating first checkpoint, flink task managers automatically flush out the memory. Though we have a temporary solution but restarting task managers everytime makes a challenge for us as the job itself takes some time to restart which causing some additional downtime on production systems.