tier storage uploaded segments not recycled

[gaobosince1987]

I tried latest tier storage from apache 3.6 branch, one thing I notice is broker local segments are not cleaned up after uploaded to remote storage. This is show as local disk usage is keep increasing and local segment never get deleted. Anyone seeing similar behavior? I wonder if there are some new configurations I should add, thanks

topic configuration: bin/kafka-topics.sh --bootstrap-server kafka-test-51-kafka-bootstrap:9092 --topic dfi-test6 --create --config remote.storage.enable=true --config local.retention.bytes=104857600

local disk usage: [kafka@kafka-test-51-kafka-7 /]$ du -sh /var/lib/kafka/data-0/kafka-log7/dfi*
12G /var/lib/kafka/data-0/kafka-log7/dfi-test-10

offset 0 still exist and disk available is decreasing with incoming traffic.