We have a Kafka stream aggregation topology. We need to keep the size of the changeLog topic in check to reduce the Kafka storage costs. So we use transformer (DSL API) in the topology to schedule a punctuation that deletes the old records from the stateStore using keyValueStore.delete().
I am able to verify that after a delete, on further scheduled triggers of the punctuation, the deleted key is not there in the state store. But does it remove the record from the changeLog topic as well? More importantly, does it reduce the size of the changeLog topic as well so that the Kafka storage cost is in check??