Hi, We fixed a critical bug in the ZooKeeper checkpoint store but never released a 1.3.x version of Flink with the fix. I'm proposing to release 1.3.3 with only fixes for this issue so that users can benefit from the fix and we have the least amount of overhead in release verification as possible.
The fixes in question are: - https://issues.apache.org/jira/browse/FLINK-7783: Don't always remove checkpoints in ZooKeeperCompletedCheckpointStore#recover() - https://issues.apache.org/jira/browse/FLINK-8807: ZookeeperCompleted checkpoint store can get stuck in infinite loop - https://issues.apache.org/jira/browse/FLINK-8890: Compare checkpoints with order in CompletedCheckpoint.checkpointsMatch() - https://issues.apache.org/jira/browse/FLINK-8487: State loss after multiple restart attempts What do you think? Best, Aljoscha