Kafka Sink connector message order processing not guaranteed

[leonpuljic]

Hi,
We have installed early access version since we needed JSON type Kafka Redis connector which supports handling tombstones (i.e., null records). This part works fine, but unfortunately we've stumbled on another issue.
As you might already know, by default key for writing kafka records to Redis is topic name + record key. Kafka connector is polling records from topic(s) in batches, and in case in one batch both non-null record and null record are polled from the same topic with same key, there is no guarantee that order will be preserved since due to current implementations delete will always be processed first, thus potentially making Redis data inconsistent with Kafka topic.
Can you please take a look into this issue?