/**
* :: DeveloperApi ::
* Base class for dependencies where each partition of the child RDD depends on a small number
* of partitions of the parent RDD. Narrow dependencies allow for pipelined execution.
*/
@DeveloperApi
abstract class NarrowDependency[T](_rdd: RDD[T]) extends Dependency[T] {
/**
* Get the parent partitions for a child partition.
* @param partitionId a partition of the child RDD
* @return the partitions of the parent RDD that the child partition depends upon
*/
def getParents(partitionId: Int): Seq[Int]
override def rdd: RDD[T] = _rdd
}
如上源码注释所讲,这些最基本的依赖关系,其中子RDD的每个分区依赖于父RDD的少量分区,并且窄依赖允许通过一个管道来执行。
总结来说:父RDD的一个分区只会被子RDD的1个分区所依赖,并且其中是不会产生shuffle过程的。
下图可以帮助理解
/**
* :: DeveloperApi ::
* Represents a dependency on the output of a shuffle stage. Note that in the case of shuffle,
* the RDD is transient since we don't need it on the executor side.
*
* @param _rdd the parent RDD
* @param partitioner partitioner used to partition the shuffle output
* @param serializer [[org.apache.spark.serializer.Serializer Serializer]] to use. If not set
* explicitly then the default serializer, as specified by `spark.serializer`
* config option, will be used.
* @param keyOrdering key ordering for RDD's shuffles
* @param aggregator map/reduce-side aggregator for RDD's shuffle
* @param mapSideCombine whether to perform partial aggregation (also known as map-side combine)
*/
@DeveloperApi
class ShuffleDependency[K: ClassTag, V: ClassTag, C: ClassTag](
@transient private val _rdd: RDD[_ <: Product2[K, V]],
val partitioner: Partitioner,
val serializer: Serializer = SparkEnv.get.serializer,
val keyOrdering: Option[Ordering[K]] = None,
val aggregator: Option[Aggregator[K, V, C]] = None,
val mapSideCombine: Boolean = false)
extends Dependency[Product2[K, V]] {
if (mapSideCombine) {
require(aggregator.isDefined, "Map-side combine without Aggregator specified!")
}
override def rdd: RDD[Product2[K, V]] = _rdd.asInstanceOf[RDD[Product2[K, V]]]
private[spark] val keyClassName: String = reflect.classTag[K].runtimeClass.getName
private[spark] val valueClassName: String = reflect.classTag[V].runtimeClass.getName
// Note: It's possible that the combiner class tag is null, if the combineByKey
// methods in PairRDDFunctions are used instead of combineByKeyWithClassTag.
private[spark] val combinerClassName: Option[String] =
Option(reflect.classTag[C]).map(_.runtimeClass.getName)
val shuffleId: Int = _rdd.context.newShuffleId()
val shuffleHandle: ShuffleHandle = _rdd.context.env.shuffleManager.registerShuffle(
shuffleId, _rdd.partitions.length, this)
_rdd.sparkContext.cleaner.foreach(_.registerShuffleForCleanup(this))
}
源码注释可以理解为:表示对shuffle阶段的输出的依赖关系。注意,在shuffle的情况下,RDD是暂时的,因为我们不需要在executor端使用它。其实看dependency名字就知道了,顾名思义,只有发生了shuffle才可以称之为宽依赖
可以总结如下:父RDD的一个分区会被子RDD的多个分区所依赖
下图可以帮助理解
Spark的DAG:就是spark任务/程序执行的流程图,
DAG的开始:从创建RDD开始
DAG的结束:到Action结束
一个Spark程序中有几个Action操作就有几个DAG,如何区分算子是否是action还是transformation,如果api返回的是RDD,则这个api算子一定是transformation,反之则为action。
Stage:是DAG中根据shuffle划分出来的阶段,前面的阶段执行完才可以执行后面的阶段,同一个阶段中的各个任务可以并行执行无需等待。
如图:
总结:
窄依赖: 并行化+容错
宽依赖: 进行阶段划分(shuffle后的阶段需要等待shuffle前的阶段计算完才能执行)
map、flatMap、filter等等常规情况下都是窄依赖,不会产生shuffle
reduceBykey、groupByKey等等常规情况下都是宽依赖,会产生shuffle
但是其实宽窄依赖不能光靠算子名称来划分,需要根据定义来划分,因为有的时候,join操作可能是窄依赖,有的时候就是宽依赖。具体的还是要看stage具体执行的时候划分。
当join和前一个父RDD被划分到同一个Stage中的时候,就可以认为这是一个窄依赖
SparkConf conf = new SparkConf().setAppName("Java-Test-WordCount").setMaster("local[*]");
JavaSparkContext jsc = new JavaSparkContext(conf);
List> tuple2List1 = Arrays.asList(new Tuple2<>("Alice", 15), new Tuple2<>("Bob", 18), new Tuple2<>("Thomas", 20), new Tuple2<>("Catalina", 25));
List> tuple3List = Arrays.asList(new Tuple3<>("Alice", "Female", "NanJ"), new Tuple3<>("Thomas", "Male", "ShangH"), new Tuple3<>("Tom", "Male", "BeiJ"));
//通过parallelize构建第一个RDD
JavaRDD> javaRDD1 = jsc.parallelize(tuple2List1);
//通过parallelize构建第二个RDD
JavaRDD> javaRDD2 = jsc.parallelize(tuple3List);
//通过mapToPair根据第一个RDD构建第三个RDD
JavaPairRDD javaRDD3 = javaRDD1.mapToPair(new PairFunction, String, Integer>() {
@Override
public Tuple2 call(Tuple2 tuple2) {
return tuple2;
}
});
//通过partitionBy根据第三个RDD构建第五个RDD
JavaPairRDD javaRDD31 = javaRDD3.partitionBy(new HashPartitioner(2));
//通过mapToPair根据第二个RDD构建第四个RDD
JavaPairRDD> javaRDD4 = javaRDD2.mapToPair(new PairFunction, String, Tuple2>() {
@Override
public Tuple2> call(Tuple3 tuple3) {
return new Tuple2<>(tuple3._1(), new Tuple2<>(tuple3._2(), tuple3._3()));
}
});
//通过partitionBy根据第四个RDD构建第六个RDD
JavaPairRDD> javaRDD41 = javaRDD4.partitionBy(new HashPartitioner(2));
//通过join 根据第五和第六个RDD构建出第七个RDD
JavaPairRDD>> javaRDD6 = javaRDD31.join(javaRDD41);
javaRDD6.foreach(new VoidFunction>>>() {
@Override
public void call(Tuple2>> stringTuple2Tuple2) throws Exception {
System.out.print(stringTuple2Tuple2);
}
});
查看Spark Web UI
Note:通过如上UI显示,可以看出,Stage5中,partitionBy 和 join在同一个Stage中,并且join是子RDD的算子,故而可以得出结论,在此Stage中,join就是一个窄依赖
SparkConf conf = new SparkConf().setAppName("Java-Test-WordCount").setMaster("local[*]");
JavaSparkContext jsc = new JavaSparkContext(conf);
List> tuple2List1 = Arrays.asList(new Tuple2<>("Alice", 15), new Tuple2<>("Bob", 18), new Tuple2<>("Thomas", 20), new Tuple2<>("Catalina", 25));
List> tuple3List = Arrays.asList(new Tuple3<>("Alice", "Female", "NanJ"), new Tuple3<>("Thomas", "Male", "ShangH"), new Tuple3<>("Tom", "Male", "BeiJ"));
//通过parallelize构建第一个RDD
JavaRDD> javaRDD1 = jsc.parallelize(tuple2List1);
//通过parallelize构建第二个RDD
JavaRDD> javaRDD2 = jsc.parallelize(tuple3List);
//通过mapToPair根据第一个RDD构建第三个RDD
JavaPairRDD javaRDD3 = javaRDD1.mapToPair(new PairFunction, String, Integer>() {
@Override
public Tuple2 call(Tuple2 tuple2) {
return tuple2;
}
});
//通过mapToPair根据第二个RDD构建第四个RDD
JavaPairRDD> javaRDD4 = javaRDD2.mapToPair(new PairFunction, String, Tuple2>() {
@Override
public Tuple2> call(Tuple3 tuple3) {
return new Tuple2<>(tuple3._1(), new Tuple2<>(tuple3._2(), tuple3._3()));
}
});
//通过join 根据第三个RDD和第四个RDD构建得出第五个RDD
JavaPairRDD>> javaRDD5 = javaRDD3.join(javaRDD4);
javaRDD5.foreach(new VoidFunction>>>() {
@Override
public void call(Tuple2>> stringTuple2Tuple2) throws Exception {
System.out.print(stringTuple2Tuple2);
}
});
查看Spark Web UI
Note:通过如上UI显示,可以看出Stage2中就只有join一个算子操作,故而此处的join算子就是一个宽依赖。