Flink报错: AppendStreamTableSink requires that Table has only insert changes

出错程序:

import java.sql.Timestamp
import java.text.SimpleDateFormat

import org.apache.flink.api.common.typeinfo.Types
import org.apache.flink.streaming.api.TimeCharacteristic
import org.apache.flink.streaming.api.functions.sink.SinkFunction
import org.apache.flink.streaming.api.scala.StreamExecutionEnvironment
import org.apache.flink.table.api.{EnvironmentSettings, Table}
import org.apache.flink.table.api.scala.StreamTableEnvironment
import org.apache.flink.table.descriptors.{Json, Kafka, Rowtime, Schema}
import org.apache.flink.table.functions.ScalarFunction
import org.apache.flink.types.Row


/**
  * 举两个UV计算的场景：
  *
  * 1. 实时计算当天零点起，到当前时间的uv。
  * 2. 实时计算当天每个小时的UV。0点...12点...24点
  *
  * 代码里面的案例，是可以用于生产中的吗？
  * 假如数据量小可以直接使用，每秒数据量大的话，就比较麻烦。
  * 因为你看group by后面的维度，只有当天date 这个维度，
  * 这样就会导致计算状态超级集中而使得内存占用超大进而引发oom。
  *
  * 这种情况解决办法就是将状态打散，然后再次聚合即可，典型的分治思想。
  *
  */
object ComputeUVDay {

  def main(args: Array[String]): Unit = {
    val fsSettings: EnvironmentSettings = EnvironmentSettings.newInstance().useOldPlanner().inStreamingMode().build()
    val fsEnv: StreamExecutionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment
    fsEnv.setStreamTimeCharacteristic(TimeCharacteristic.EventTime)

    val tEnv: StreamTableEnvironment = StreamTableEnvironment.create(fsEnv,fsSettings)

    //注册自定义函数
    tEnv.registerFunction("DateUtil", new DateUtil)

    // 注册source
    tEnv.connect(   //连接信息
      new Kafka()
        .version("0.11")
        .topic("test")
        .property("bootstrap.servers", "localhost:9092")
        .property("group.id","test")
        .startFromLatest()
    )
      .withFormat(
      new Json()
        .failOnMissingField(false)
        .deriveSchema()
    )
      .withSchema(
        new Schema()
          .field("rowtime", Types.SQL_TIMESTAMP)
          .rowtime(new Rowtime()
            .timestampsFromField("eventtime")
            .watermarksPeriodicBounded(2000)
          )
          .field("fruit", Types.STRING)
          .field("number", Types.LONG)
      )
      .inAppendMode()  
      .registerTableSource("source")

    // 注册sink
    tEnv.connect(
      new Kafka()
        .version("0.11")
        .topic("test_22")
        .property("acks", "all")
        .property("retries", "0")
        .property("batch.size", "16384")
        .property("linger.ms", "10")
        .property("bootstrap.servers", "localhost:9092")
        .sinkPartitionerFixed()
    ).inAppendMode()      // 仅交互INSERT操作更新数据
      .withFormat(
        new Json().deriveSchema()
      )
      .withSchema(
        new Schema()
          .field("time1",Types.STRING)
          .field("total", Types.LONG)
      )
      .registerTableSink("sink")



    //计算天级别的uv
    tEnv.sqlUpdate("insert into sink select DateUtil(rowtime) time1,count(distinct fruit) total from source group by DateUtil(rowtime)")

    //计算小时级别的uv
    // val table: Table = tEnv.sqlQuery("select  DateUtil(rowtime,'yyyyMMddHH'),count(distinct fruit) from source group by DateUtil(rowtime,'yyyyMMddHH')")

    //将table通过toRetractStream方法转换成Row格式的DataStream
    //返回结果类型为(Boolean, Row)
    //可以根据第一个字段是否为True判断是插入还是删除引起的更新数据
    //tEnv.toRetractStream[Row](table).addSink(new MySink)

    fsEnv.execute(this.getClass.getSimpleName)

  }

}

/**
  * 自定义UDF
  */
class DateUtil extends ScalarFunction {

  def eval(timestamp:Long): String = {
    val sdf = new SimpleDateFormat("yyyy-MM-dd")
    sdf.format(new Timestamp(timestamp))
  }

  def eval(timestamp:Long, format:String): Unit = {
    val sdf = new SimpleDateFormat(format)
    sdf.format(new Timestamp(timestamp))
  }

}

/**
  * 自定义输出sink
  */
class MySink extends SinkFunction[(Boolean, Row)] {

  override def invoke(value: (Boolean, Row), context: SinkFunction.Context[_]): Unit = {
    println(value._1.toString)
  }
}

就是一个kafka的数据，聚合计算之后往另一个kafka写的操作

报错详细信息:

Exception in thread "main" org.apache.flink.table.api.TableException: AppendStreamTableSink requires that Table has only insert changes.
	at org.apache.flink.table.planner.StreamPlanner.writeToAppendSink(StreamPlanner.scala:329)
	at org.apache.flink.table.planner.StreamPlanner.org$apache$flink$table$planner$StreamPlanner$$writeToSink(StreamPlanner.scala:285)
	at org.apache.flink.table.planner.StreamPlanner$$anonfun$2.apply(StreamPlanner.scala:169)
	at org.apache.flink.table.planner.StreamPlanner$$anonfun$2.apply(StreamPlanner.scala:155)
	at scala.Option.map(Option.scala:146)
	at org.apache.flink.table.planner.StreamPlanner.org$apache$flink$table$planner$StreamPlanner$$translate(StreamPlanner.scala:155)
	at org.apache.flink.table.planner.StreamPlanner$$anonfun$translate$1.apply(StreamPlanner.scala:127)
	at org.apache.flink.table.planner.StreamPlanner$$anonfun$translate$1.apply(StreamPlanner.scala:127)
	at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:234)
	at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:234)
	at scala.collection.Iterator$class.foreach(Iterator.scala:891)
	at scala.collection.AbstractIterator.foreach(Iterator.scala:1334)
	at scala.collection.IterableLike$class.foreach(IterableLike.scala:72)
	at scala.collection.AbstractIterable.foreach(Iterable.scala:54)
	at scala.collection.TraversableLike$class.map(TraversableLike.scala:234)
	at scala.collection.AbstractTraversable.map(Traversable.scala:104)
	at org.apache.flink.table.planner.StreamPlanner.translate(StreamPlanner.scala:127)
	at org.apache.flink.table.api.internal.TableEnvironmentImpl.translate(TableEnvironmentImpl.java:439)
	at org.apache.flink.table.api.internal.TableEnvironmentImpl.sqlUpdate(TableEnvironmentImpl.java:348)
	at yunqing.table.daywindow.ComputeUVDay$.main(ComputeUVDay.scala:98)
	at yunqing.table.daywindow.ComputeUVDay.main(ComputeUVDay.scala)

意思就是 AppendStreamTableSink 需要表只有插入（不能update），去掉表上面的groupBy()，就不会报错了。。。

把sql调整为:

tEnv.sqlUpdate("insert into sink select DateUtil(rowtime) time1, fruit total from source")

程序运行正常

注意: group 字段的 count 值变化的时候，会产生两条数据，一条是旧数据，带着false标示，一条是新数据，带着true标示

但是如果就是必须要执行groupBy操作，该怎么办呢？

换个sink端，要支持更新的sink
kafka是一个消息队列，仅仅支持append, 不支持更新和删除