There is no such concept in Flink. Flink tracks offsets in its checkpoints. It can optionally commit offsets to Kafka, but that is only for reporting purposes. If you wish to lower the number of records that get reprocessed in the case of a restart, then you must lower the checkpoint interval.
On Tue, Jun 4, 2019 at 10:47 AM wang xuchen <ben....@gmail.com> wrote: > > Hi Flink users, > > When # of Kafka consumers = # of partitions, and I use > setParallelism(>1), something like this > > 'messageSteam.rebalance().map(lamba).setParallelism(3).print()' > > How do I tune # of outstanding uncommitted offset? Something similar to > > https://storm.apache.org/releases/1.1.2/storm-kafka-client.html in Storm. > > Thanks > Ben >