Github user tzulitai commented on the issue:

    https://github.com/apache/flink/pull/5282
  
    @aljoscha this is the currently the case for any startup mode. Any 
partition discovered after the initial batch fetch is considered a new 
partition due to Kafka scale outs, and is therefore read from the record 
horizon. Startup modes apply only to existing partitions.


---

Reply via email to