[FLINK-26018] Unnecessary late events when using the new KafkaSource

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.14.3, 1.14.4
• Fix Version/s: 1.14.5, 1.15.0
• Component/s: Connectors / Common
• Labels: pull-request-available
• Assignee: Unassigned
• Reporter: Jun Qin
• Created: 08/Feb/22 20:37
• Updated: 31/Mar/22 07:02
• Resolved: 31/Mar/22 06:56

Description

There is an issue with the new KafkaSource connector in Flink 1.14: when one task consumes messages from multiple topic partitions (statically created, timestamp are in order), it may start with one partition and advances watermarks before the data from other partitions come. In this case, the early messages in other partitions may unnecessarily be considered  as late ones.

I discussed with renqs, it seems that the new KafkaSource only adds a partition into WatermarkMultiplexer when it receives data from that partition. In contrast, FlinkKafkaConsumer adds all known partition before it fetch any data. 

Attached two files: the messages in Kafka and the corresponding TM logs.

Issue Links

• is related to: FLINK-21871 Support watermark for Hive and Filesystem streaming source
• links to: GitHub Pull Request #18790