Flink CDC里binlogs日志读到kafka.每修改一次为什么会丢字段值？

Flink CDC里binlogs日志读到kafka.每修改一次会读到3条日志，而且顺序可能不通，字段还有丢失是为啥？如果最后一条没有丢，数据正确，但是如果没丢的在前面，那么字段值就丢了。{"id":1,"long_txt_value":"d","name":"1","varchar_value":"9"}
{"id":1,"long_txt_value":"d","name":"1"}
{"id":1,"long_txt_value":"d","name":"1"}

{"id":1,"long_txt_value":"d","name":"2","varchar_value":"9"}
{"id":1,"long_txt_value":"d","name":"2"}
{"id":1,"long_txt_value":"d","name":"2"}
{"id":1,"long_txt_value":"d","name":"3"}
{"id":1,"long_txt_value":"d","name":"3"}
{"id":1,"long_txt_value":"d","name":"3","varchar_value":"9"}进kafka之前初始化的时候没有问题，就是增量有问题。