问题
I am building a Spring Cloud Stream Kafka processor app that will consume raw data with a String key and sometimes a null payload from a Kafka topic. I want to produce to another topic a String key and the null payload (known as a tombstone within Kafka). In order to use raw headers on the message, I need to output a byte[], but if I encode KafkaNull.INSTANCE into a byte[] it will literally output a String of the object hashcode.
If I try to send anything other than a byte[], I can't use raw headers.
What is the correct way to do this? Where would the headers even be if the payload is null? I set producer.headerMode=embeddedHeaders on deploy and that didn't seem to make a difference, still have the hashcode as the payload.
回答1:
You can use the latter introduced useNativeEncoding property (which overlaps the raw headerMode in that it skips headers) but allows you use a native Kafka serializer.
See http://docs.spring.io/autorepo/docs/spring-cloud-stream-docs/Brooklyn.BUILD-SNAPSHOT/reference/htmlsingle/#_producer_properties or http://docs.spring.io/autorepo/docs/spring-cloud-stream-docs/Chelsea.BUILD-SNAPSHOT/reference/htmlsingle/#_producer_properties
来源：https://stackoverflow.com/questions/42960237/publish-null-tombstone-message-with-raw-headers