The doc tells us that the kafka sink uses fixed partitioner by default. However, in my local test, the it uses sticky partitioner to get the record partition id if key is not set.
You can add the test in the KafkaTableITCase, the code follows
The test will use the kafka default partitioner and sends record to kafka topic. After insert, we can read the record with the parititon id. If it uses the fixed partitioner, all records will has the same partition id. I repeat the test 3 times and the results are
The last column is the partition-id and we have 3 partitions in the test. The results show the default partitioner is sticky paritioner rather than fixed partitioner.
By the way, the sink partitioning section in the doc only works when the key is null. If we set the key fields, the round-robin strategy will not work.