Hi Mariam, 1. The expired state object will be cleaned up when it is accessed in a read operation. And for the expired state that is never accessed again, there are three strategies to remove it in the background: FULL_STATE_SCAN_SNAPSHOT, INCREMENTAL_CLEANUP, ROCKSDB_COMPACTION_FILTER. 2. FULL_STATE_SCAN_SNAPSHOT: When a full snapshot for checkpoint or savepoint is taken, the expired state will be evicted automatically. The strategy is supported in 1.6.0. 3. INCREMENTAL_CLEANUP: This strategy is specific to Heap state backends. If enabled it, every state access triggers a cleanup step. For every clean up step, a certain number of state entries are checked for expiration. 4. ROCKSDB_COMPACTION_FILTER: This strategy is specific to Rocksdb statebackend. The strategy will register a Flink specific compaction filter to RocksDB, and the expired state data will be cleaned up during the compaction process of RocksDB.
For more information, you can refer to this page: https://flink.apache.org/2019/05/19/state-ttl.html Best regards, Jinzhong, Li On 2022/01/09 08:41:08 Mariam Walid wrote: > Dear All, > > I am writing to inquire about the state time to live lazy removal. I guess > that even if the state has expired it i still in the memory and not garbage > collected. The question is how to get rid of those expired states. > > Best regards, > Mariam >
