【Spark三十七】Spark Cache机制

今天状态很差，很困，无精打采。学到的Spark知识，没有连贯起来，很多知识点有印象但是很模糊，说不出个123来。本来今天要看看cache，checkpoint和broadcast，结果今天到现在为止已经是5点了，还没有任何的进展。开始硬着头皮把Spark的Cache机制搞一搞吧，发现，cache机制比想象中的难驾驭。

 

 

 调用reduceByKey对应的ShuffledRDD对应的cache

 

cache不起作用

 

package spark.examples

import org.apache.spark.SparkConf
import org.apache.spark.SparkContext

import org.apache.spark.SparkContext._

object SparkWordCountCache {
  def main(args: Array[String]) {
    System.setProperty("hadoop.home.dir", "E:\\devsoftware\\hadoop-2.5.2\\hadoop-2.5.2");
    val conf = new SparkConf()
    conf.setAppName("SparkWordCount")
    conf.setMaster("local[3]")
    conf.set("spark.shuffle.manager", "hash"); ///hash是否有影响?
    val sc = new SparkContext(conf)
    val rdd1 = sc.textFile("file:///D:/word.in.3");
    val rdd2 = rdd1.flatMap(_.split(" "))
    val rdd3 = rdd2.map((_, 1))
    val rdd4 = rdd3.reduceByKey(_ + _, 3);
    rdd4.cache();
    rdd4.saveAsTextFile("file:///D:/wordout" + System.currentTimeMillis());
    val result = rdd4.collect; ///没有触发ShuffleMapTask执行，但是依然需要从ShuffleMapTask产生的结果拉取数据
    result.foreach(println(_));
    sc.stop
  }
}

 

以上代码调用rdd3.cache(),而rdd3是一个ShuffleMapRDD，也就是说，保存的是Stage2里面的RDD结果。此时调用cache.collect时，产生的Task都是ResultTask，也就是说，由于cache作用，最后一个Job并没有从前面从头计算？

感觉不对，即使不用cache，也应该不会从头计算吧

 

经验证，感觉是对的，将上面的代码做如下修改，结果一样，最后也不会调用ShuffleMapTask，但是在执行ResultTask时，还是会从MapTask的输出中拉取数据，所以并没有对Shuffle读过程进行简化。

 

    rdd3.saveAsTextFile("file:///D:/wordout" + System.currentTimeMillis());
    val result = rdd3.collect;
    result.foreach(println(_));

 

上来就踩了个cache的坑！Spark是不支持ShuffleMapRDD的cache的，虽然上面不需要ShuffleMapTask，但是ResultTask运行时，依然需要从MapTask的结果中拉取数据

 

 

调用groupByKey对应的ShuffledRDD对应的cache

 

结果rdd.cache起作用了

 

package spark.examples

import org.apache.spark.SparkContext
import org.apache.spark.SparkContext._
import org.apache.spark.SparkConf

object SparkGroupByExample {

  def main(args: Array[String]) {
    val conf = new SparkConf().setAppName("GroupByKey").setMaster("local")
    val sc = new SparkContext(conf)
    sc.setCheckpointDir("/tmp/checkpoint/" + System.currentTimeMillis())

    val data = Array[(Int, Char)]((1, 'a'), (2, 'b'),
      (3, 'c'), (4, 'd'),
      (5, 'e'), (3, 'f'),
      (2, 'g'), (1, 'h')
    )
    val pairs = sc.parallelize(data)
    val rdd = pairs.groupByKey(2)
    rdd.cache
    rdd.count;
    rdd.collect.foreach(println(_));
  }
}

 

 

 

调用textFile对应的MappedRDD对应的cache操作

 

基本流程：假如在一个程序中有两个Job。第一个Job运行时，，对于调用了cache的RDD首先计算它的数据，然后写入cache。第二个job在运行时，会直接从cache中读取。

这对于迭代计算的Job，会非常适合，将上个任务的结果缓存，供第二个任务使用，然后依次类推

 

 

package spark.examples

import org.apache.spark.SparkConf
import org.apache.spark.SparkContext

import org.apache.spark.SparkContext._

object SparkWordCountCache {
  def main(args: Array[String]) {
    System.setProperty("hadoop.home.dir", "E:\\devsoftware\\hadoop-2.5.2\\hadoop-2.5.2");
    val conf = new SparkConf()
    conf.setAppName("SparkWordCount")
    conf.setMaster("local")
    //Hash based Shuffle；
    conf.set("spark.shuffle.manager", "hash");
    val sc = new SparkContext(conf)
    val rdd1 = sc.textFile("file:///D:/word.in.3");
    rdd1.cache() ///数据读取后即做cache，第一个job运行后，就会缓存
    val rdd2 = rdd1.flatMap(_.split(" "))
    val rdd3 = rdd2.map((_, 1))
    val result = rdd3.collect; ///打印rdd3的内容
    result.foreach(println(_));
    val rdd4 = rdd3.reduceByKey(_ + _); ///对rdd3做reduceByKey操作
    rdd4.saveAsTextFile("file:///D:/wordout" + System.currentTimeMillis());
    sc.stop
  }
}

 

 

源代码基本流程：

 

• 调用RDD的iterator方法，计算RDD的数据集合(得到的是一个可迭代的集合)
• 在RDD的iterator方法中，检查RDD的storage level，如果设置了storage level，那么调用SparkEnv.get.cacheManager.getOrCompute(this, split, context, storageLevel)
• 在CacheManager的getOrCompute方法中，

           a.首先判断是否存在于cache中，如果存在则直接返回，

           b.如果不存在，则调用  val computedValues = rdd.computeOrReadCheckpoint(partition, context)进行计算。

           c.计算结束后，调用CacheManager自身的putInBlockManager将计算得到的数据缓存

           d. 数据放入BlockManager后，还需要更新这个RDD和BlockManager之间的对应关系，以便下次再计算这个RDD时，检查RDD数据是否已经缓存

 

主要源代码

 

 1. getOrCompute方法

 

/** Gets or computes an RDD partition. Used by RDD.iterator() when an RDD is cached. */
  def getOrCompute[T](
      rdd: RDD[T],
      partition: Partition,
      context: TaskContext,
      storageLevel: StorageLevel): Iterator[T] = {

    val key = RDDBlockId(rdd.id, partition.index) //RDD的id和partition的index构造RDDBlockId，一个RDD可以有多个partition
    logDebug(s"Looking for partition $key")
    blockManager.get(key) match { ///从blockManger中根据key查找，key最后会存入BlockManager么吗？BlockManager管理Spark的块信息
      case Some(blockResult) =>
        // Partition is already materialized, so just return its values
        context.taskMetrics.inputMetrics = Some(blockResult.inputMetrics)
        new InterruptibleIterator(context, blockResult.data.asInstanceOf[Iterator[T]])

      case None =>
        // Acquire a lock for loading this partition
        // If another thread already holds the lock, wait for it to finish return its results
        val storedValues = acquireLockForPartition[T](key) ///根据Key获取缓存的数据，acquireLockForPartition名字起得不好
        if (storedValues.isDefined) { ///找到数据
          return new InterruptibleIterator[T](context, storedValues.get)
        }

        // Otherwise, we have to load the partition ourselves
        ///为找到缓存的数据，表明是job第一次运行
        try { 
          logInfo(s"Partition $key not found, computing it")
          val computedValues = rdd.computeOrReadCheckpoint(partition, context) ///计算RDD数据

          // If the task is running locally, do not persist the result
          if (context.isRunningLocally) { ///如果数据在本地，就不需要缓存了？
            return computedValues
          }

          // Otherwise, cache the values and keep track of any updates in block statuses
          ///缓存数据
          val updatedBlocks = new ArrayBuffer[(BlockId, BlockStatus)]
          
          ///将数据存入BlockManager，注意四个参数
          val cachedValues = putInBlockManager(key, computedValues, storageLevel, updatedBlocks)
          
          ///这是什么意思?任务的metrics，任务的
          val metrics = context.taskMetrics
          val lastUpdatedBlocks = metrics.updatedBlocks.getOrElse(Seq[(BlockId, BlockStatus)]())
          metrics.updatedBlocks = Some(lastUpdatedBlocks ++ updatedBlocks.toSeq)
          new InterruptibleIterator(context, cachedValues)

        } finally {
          loading.synchronized {
            loading.remove(key)
            loading.notifyAll()
          }
        }
    }
  }

 

2. putInBlockManager方法

/**
   * Cache the values of a partition, keeping track of any updates in the storage statuses of
   * other blocks along the way.
   *
   * The effective storage level refers to the level that actually specifies BlockManager put
   * behavior, not the level originally specified by the user. This is mainly for forcing a
   * MEMORY_AND_DISK partition to disk if there is not enough room to unroll the partition,
   * while preserving the the original semantics of the RDD as specified by the application.
   */
  private def putInBlockManager[T](
      key: BlockId,
      values: Iterator[T],
      level: StorageLevel,
      updatedBlocks: ArrayBuffer[(BlockId, BlockStatus)],
      effectiveStorageLevel: Option[StorageLevel] = None): Iterator[T] = {

    val putLevel = effectiveStorageLevel.getOrElse(level)
    if (!putLevel.useMemory) {
      /*
       * This RDD is not to be cached in memory, so we can just pass the computed values as an
       * iterator directly to the BlockManager rather than first fully unrolling it in memory.
       */
      updatedBlocks ++=
        blockManager.putIterator(key, values, level, tellMaster = true, effectiveStorageLevel)
      blockManager.get(key) match {
        case Some(v) => v.data.asInstanceOf[Iterator[T]]
        case None =>
          logInfo(s"Failure to store $key")
          throw new BlockException(key, s"Block manager failed to return cached value for $key!")
      }
    } else {
      /*
       * This RDD is to be cached in memory. In this case we cannot pass the computed values
       * to the BlockManager as an iterator and expect to read it back later. This is because
       * we may end up dropping a partition from memory store before getting it back.
       *
       * In addition, we must be careful to not unroll the entire partition in memory at once.
       * Otherwise, we may cause an OOM exception if the JVM does not have enough space for this
       * single partition. Instead, we unroll the values cautiously, potentially aborting and
       * dropping the partition to disk if applicable.
       */
      blockManager.memoryStore.unrollSafely(key, values, updatedBlocks) match {
        case Left(arr) =>
          // We have successfully unrolled the entire partition, so cache it in memory
          updatedBlocks ++=
            blockManager.putArray(key, arr, level, tellMaster = true, effectiveStorageLevel)
          arr.iterator.asInstanceOf[Iterator[T]]
        case Right(it) =>
          // There is not enough space to cache this partition in memory
          val returnValues = it.asInstanceOf[Iterator[T]]
          if (putLevel.useDisk) {
            logWarning(s"Persisting partition $key to disk instead.")
            val diskOnlyLevel = StorageLevel(useDisk = true, useMemory = false,
              useOffHeap = false, deserialized = false, putLevel.replication)
            putInBlockManager[T](key, returnValues, level, updatedBlocks, Some(diskOnlyLevel))
          } else {
            returnValues
          }
      }
    }
  }

 

 

评论

1 楼 jchubby 2017-02-20  

关于第一个reduceByKey对应的cache，shuffle之后的rdd，spark会自动保存该结果，直到这个rdd短时间内不会再被使用或者被垃圾回收了，所以用不用cache，collect这个job的第一个map stage都会被skip