Ok, then may be I have misunderstanding about checkpoint. I thought flink use checkpoint to store offset, but when kafka connector making a checkpoint, it doesn't know whether data is in in-progress file or in pending-file so a whole offset is saved in checkpoint. I used to guess, the data in in-progress file may be lost when checkpoint requested.
-- Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
