apache-spark - 解释 Spark Structured Streaming executor 和 Kafka partitions 之间的映射

Question

我已经在具有 4 个分区的 Kafka 主题上部署了一个具有 4 个工作人员的结构化流。

我假设将为 4 个分区部署 4 个工作人员，并在工作人员<->分区之间进行一对一映射。

但是，事实并非如此。所有分区都由同一个 Executor 提供服务。我通过检查线程 ID 并登录执行程序来确认这一点。

是否有任何文档显示 Kafka 分区和 Spark Structured Streams 之间的相关性。此外，是否有任何我们可以调整的旋钮。

Answer 1

相关性是“1:n(executor:partitions)”:一个Kafka分区只能被一个executor消费，一个executor可以消费多个Kafka分区。

这与 Spark Streaming 一致。

对于Structured Streaming，默认模型为“微批处理模型”，“连续处理模型”仍处于“实验”状态。

对于“微批处理模型”，在“KafkaSource.scala”中，有

 *   - The DF returned is based on [[KafkaSourceRDD]] which is constructed such that the
 *     data from Kafka topic + partition is consistently read by the same executors across
 *     batches, and cached KafkaConsumers in the executors can be reused efficiently. See the
 *     docs on [[KafkaSourceRDD]] for more details.

在“KafkaSourceRDD”中

/**
 * An RDD that reads data from Kafka based on offset ranges across multiple partitions.
 * Additionally, it allows preferred locations to be set for each topic + partition, so that
 * the [[KafkaSource]] can ensure the same executor always reads the same topic + partition
 * and cached KafkaConsumers (see [[KafkaDataConsumer]] can be used read data efficiently.
 *
 * ...
 */
private[kafka010] class KafkaSourceRDD(

我们知道默认位置策略是 LocationStrategies.PreferConsistent .

对于“连续处理模型”，在“KafkaContinuousReader.scala”中

  override def createUnsafeRowReaderFactories(): ju.List[DataReaderFactory[UnsafeRow]] = {
    ...
    startOffsets.toSeq.map {
      case (topicPartition, start) =>
        KafkaContinuousDataReaderFactory(
          topicPartition, start, kafkaParams, pollTimeoutMs, failOnDataLoss)
          .asInstanceOf[DataReaderFactory[UnsafeRow]]
    }.asJava
  }

/**
 * A data reader factory for continuous Kafka processing. This will be serialized and transformed
 * into a full reader on executors.
 *
 * @param topicPartition The (topic, partition) pair this task is responsible for.
 * ...
 */
case class KafkaContinuousDataReaderFactory(
    topicPartition: TopicPartition,
    startOffset: Long,
    kafkaParams: ju.Map[String, Object],
    pollTimeoutMs: Long,
    failOnDataLoss: Boolean) extends DataReaderFactory[UnsafeRow] {
  override def createDataReader(): KafkaContinuousDataReader = {
    new KafkaContinuousDataReader(
      topicPartition, startOffset, kafkaParams, pollTimeoutMs, failOnDataLoss)
  }
}

我们可以知道每个 (topic, partition)将包含在一个工厂中，然后将包含在一个执行器中。