Hello, trying to better understand Managed Memory maxing out for our Kafka -> S3 Flink Application. We have an app's Task Manager set with 4 Task Slots and 16Gb of memory. It seems the managed memory is getting overwhelmed (we are using RocksDB, which may explain this). The Flink docs seem to state that the default fraction value for managed memory is

0.4

but i'm seeing 28% memory allocation to managed memory. Would we be better off managing this via a fraction or via an explicit

taskmanager.memory.managed.size

setting