Hi Febian, With watermark, I understand it could only write those that are smaller than the received watermark, but could I know why it would also need to maintain a write ahead log of all received rows? When an event received, it just compare time with current watermark, write it to correct bucket if smaller then watermark, otherwise drop it.
With with assumption, BucketingSink could close all bucket that is older than current watermark, I think it make sense as those bucket data won't change anymore. The close action could be done in checkpoint callback or when every event receive. It implemented the BucketReady mechanism @Minglei suggested in https://issues.apache.org/jira/browse/FLINK-9609 using current watermark mechanism. And I think we don't need BucketWatermark concept, as it confuse with current watermark. Thanks, Xilang -- Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
