| 流内 | 窗口去重取首条/末条 |
|---|---|
| 入库 | 主键 upsert |
| 水位 | 进度可恢复 |
Flink 流内去重
Flink 流内去重
SELECT msg_id, order_id, amount, event_time
FROM (
SELECT *, ROW_NUMBER() OVER (PARTITION BY msg_id ORDER BY event_time ASC) AS rn
FROM kafka_order_source
) WHERE rn = 1;去重窗口要有界:无界状态会无限增长,按消息保留期设置状态 TTL。