RocksDB Compaction filter to clean up state with TTL

["Seth, Abhilasha" <ab...@amazon.com.INVALID>]

Hello,
Flink 1.8 introduces the config ‘state.backend.rocksdb.ttl.compaction.filter.enabled’ to enable or disable the compaction filter to cleanup state with TTL. I was curious why its disabled by default. Are there any performance implications of turning it ON by default?
Thanks,
Abhilasha

Re: RocksDB Compaction filter to clean up state with TTL

[Yu Li <ca...@gmail.com>]

Hi Abhilasha,

We were conservative about enabling this by default to prevent any
unexpected problems. However, since there has been no reported issues so
far, we will enable this by default in 1.10.0 release. Please refer to
FLINK-14898 [1], FLINK-15506 [2] and (currently drafted) 1.10.0 release
note [3] for more details

Best Regards,
Yu

[1] https://issues.apache.org/jira/browse/FLINK-14898
[2] https://issues.apache.org/jira/browse/FLINK-15506
[3]
https://ci.apache.org/projects/flink/flink-docs-master/release-notes/flink-1.10.html#state

On Tue, 4 Feb 2020 at 03:32, Seth, Abhilasha <ab...@amazon.com.invalid>
wrote:

> Hello,
> Flink 1.8 introduces the config
> ‘state.backend.rocksdb.ttl.compaction.filter.enabled’ to enable or disable
> the compaction filter to cleanup state with TTL. I was curious why its
> disabled by default. Are there any performance implications of turning it
> ON by default?
> Thanks,
> Abhilasha
>
>