算子练习总结文档
发生shuffle的算子:
1.去重 distinct
/**
* Return a new RDD containing the distinct elements in this RDD.
*/
def distinct(numPartitions: Int)(implicit ord: Ordering[T] = null): RDD[T] = withScope {
map(x => (x, null)).reduceByKey((x, y) => x, numPartitions).map(_._1)
}
/**
* Return a new RDD containing the distinct elements in this RDD.
*/
def distinct(): RDD[T] = withScope {
distinct(partitions.length)
}
例子:
val c = sc.parallelize(List("Gnu", "Cat", "Rat", "Dog", "Gnu", "Rat"), 2)
c.distinct.collect
res6: Array[String] = Array(Dog, Gnu, Cat, Rat)
val a = sc.parallelize(List(1,2,3,4,5,6,7,8,9,10))
a.distinct(2).partitions.length
res16: Int = 2
a.distinct(3).partitions.length
res17: Int = 3
2.聚合
reduceByKey
/**
* Merge the values for each key using an associative and commutative reduce function. This will
* also perform the merging locally on each mapper before sending results to a reducer, similarly
* to a "combiner" in MapReduce.
*/
def reduceByKey(partitioner: Partitioner, func: (V, V) => V): RDD[(K, V)] = self.withScope {
combineByKeyWithClassTag[V]((v: V) => v, func, func, partitioner)
}
/**
* Merge the values for each key using an associative and commutative reduce function. This will
* also perform the merging locally on each mapper before sending results to a reducer, similarly
* to a "combiner" in MapReduce. Output will be hash-partitioned with numPartitions partitions.
*/
def reduceByKey(func: (V, V) => V, numPartitions: Int): RDD[(K, V)] = self.withScope {
reduceByKey(new HashPartitioner(numPartitions), func)
}
/**
* Merge the values for each key using an associative and commutative reduce function. This will
* also perform the merging locally on each mapper before sending results to a reducer, similarly
* to a "combiner" in MapReduce. Output will be hash-partitioned with the existing partitioner/
* parallelism level.
*/
def reduceByKey(func: (V, V) => V): RDD[(K, V)] = self.withScope {
reduceByKey(defaultPartitioner(self), func)
}
例子:
val a = sc.parallelize(List("dog", "cat", "owl", "gnu", "ant"), 2)
val b = a.map(x => (x.length, x))
b.reduceByKey(_ + _).collect
res86: Array[(Int, String)] = Array((3,dogcatowlgnuant))
val a = sc.parallelize(List("dog", "tiger", "lion", "cat", "panther", "eagle"), 2)
val b = a.map(x => (x.length, x))
b.reduceByKey(_ + _).collect
res87: Array[(Int, String)] = Array((4,lion), (3,dogcat), (7,panther), (5,tigereagle))
groupBy
/**
* Return an RDD of grouped items. Each group consists of a key and a sequence of elements
* mapping to that key. The ordering of elements within each group is not guaranteed, and
* may even differ each time the resulting RDD is evaluated.
*
* @note This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using `PairRDDFunctions.aggregateByKey`
* or `PairRDDFunctions.reduceByKey` will provide much better performance.
*/
def groupBy[K](f: T => K)(implicit kt: ClassTag[K]): RDD[(K, Iterable[T])] = withScope {
groupBy[K](f, defaultPartitioner(this))
}
/**
* Return an RDD of grouped elements. Each group consists of a key and a sequence of elements
* mapping to that key. The ordering of elements within each group is not guaranteed, and
* may even differ each time the resulting RDD is evaluated.
*
* @note This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using `PairRDDFunctions.aggregateByKey`
* or `PairRDDFunctions.reduceByKey` will provide much better performance.
*/
def groupBy[K](
f: T => K,
numPartitions: Int)(implicit kt: ClassTag[K]): RDD[(K, Iterable[T])] = withScope {
groupBy(f, new HashPartitioner(numPartitions))
}
/**
* Return an RDD of grouped items. Each group consists of a key and a sequence of elements
* mapping to that key. The ordering of elements within each group is not guaranteed, and
* may even differ each time the resulting RDD is evaluated.
*
* @note This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using `PairRDDFunctions.aggregateByKey`
* or `PairRDDFunctions.reduceByKey` will provide much better performance.
*/
def groupBy[K](f: T => K, p: Partitioner)(implicit kt: ClassTag[K], ord: Ordering[K] = null)
: RDD[(K, Iterable[T])] = withScope {
val cleanF = sc.clean(f)
this.map(t => (cleanF(t), t)).groupByKey(p)
}
例子:
val a = sc.parallelize(1 to 9, 3)
a.groupBy(x => { if (x % 2 == 0) "even" else "odd" }).collect
res42: Array[(String, Seq[Int])] = Array((even,ArrayBuffer(2, 4, 6, 8)), (odd,ArrayBuffer(1, 3, 5, 7, 9)))
val a = sc.parallelize(1 to 9, 3)
def myfunc(a: Int) : Int =
{
a % 2
}
a.groupBy(myfunc).collect
res3: Array[(Int, Seq[Int])] = Array((0,ArrayBuffer(2, 4, 6, 8)), (1,ArrayBuffer(1, 3, 5, 7, 9)))
val a = sc.parallelize(1 to 9, 3)
def myfunc(a: Int) : Int =
{
a % 2
}
a.groupBy(x => myfunc(x), 3).collect
a.groupBy(myfunc(_), 1).collect
res7: Array[(Int, Seq[Int])] = Array((0,ArrayBuffer(2, 4, 6, 8)), (1,ArrayBuffer(1, 3, 5, 7, 9)))
import org.apache.spark.Partitioner
class MyPartitioner extends Partitioner {
def numPartitions: Int = 2
def getPartition(key: Any): Int =
{
key match
{
case null => 0
case key: Int => key % numPartitions
case _ => key.hashCode % numPartitions
}
}
override def equals(other: Any): Boolean =
{
other match
{
case h: MyPartitioner => true
case _ => false
}
}
}
val a = sc.parallelize(1 to 9, 3)
val p = new MyPartitioner()
val b = a.groupBy((x:Int) => { x }, p)
val c = b.mapWith(i => i)((a, b) => (b, a))
c.collect
res42: Array[(Int, (Int, Seq[Int]))] = Array((0,(4,ArrayBuffer(4))), (0,(2,ArrayBuffer(2))), (0,(6,ArrayBuffer(6))), (0,(8,ArrayBuffer(8))), (1,(9,ArrayBuffer(9))), (1,(3,ArrayBuffer(3))), (1,(1,ArrayBuffer(1))), (1,(7,ArrayBuffer(7))), (1,(5,ArrayBuffer(5))))
groupByKey
/**
* Group the values for each key in the RDD into a single sequence. Allows controlling the
* partitioning of the resulting key-value pair RDD by passing a Partitioner.
* The ordering of elements within each group is not guaranteed, and may even differ
* each time the resulting RDD is evaluated.
*
* @note This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using `PairRDDFunctions.aggregateByKey`
* or `PairRDDFunctions.reduceByKey` will provide much better performance.
*
* @note As currently implemented, groupByKey must be able to hold all the key-value pairs for any
* key in memory. If a key has too many values, it can result in an `OutOfMemoryError`.
*/
def groupByKey(partitioner: Partitioner): RDD[(K, Iterable[V])] = self.withScope {
// groupByKey shouldn't use map side combine because map side combine does not
// reduce the amount of data shuffled and requires all map side data be inserted
// into a hash table, leading to more objects in the old gen.
val createCombiner = (v: V) => CompactBuffer(v)
val mergeValue = (buf: CompactBuffer[V], v: V) => buf += v
val mergeCombiners = (c1: CompactBuffer[V], c2: CompactBuffer[V]) => c1 ++= c2
val bufs = combineByKeyWithClassTag[CompactBuffer[V]](
createCombiner, mergeValue, mergeCombiners, partitioner, mapSideCombine = false)
bufs.asInstanceOf[RDD[(K, Iterable[V])]]
}
/**
* Group the values for each key in the RDD into a single sequence. Hash-partitions the
* resulting RDD with into `numPartitions` partitions. The ordering of elements within
* each group is not guaranteed, and may even differ each time the resulting RDD is evaluated.
*
* @note This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using `PairRDDFunctions.aggregateByKey`
* or `PairRDDFunctions.reduceByKey` will provide much better performance.
*
* @note As currently implemented, groupByKey must be able to hold all the key-value pairs for any
* key in memory. If a key has too many values, it can result in an `OutOfMemoryError`.
*/
def groupByKey(numPartitions: Int): RDD[(K, Iterable[V])] = self.withScope {
groupByKey(new HashPartitioner(numPartitions))
}
例子:
val a = sc.parallelize(List("dog", "tiger", "lion", "cat", "spider", "eagle"), 2)
val b = a.keyBy(_.length)
b.groupByKey.collect
res11: Array[(Int, Seq[String])] = Array((4,ArrayBuffer(lion)), (6,ArrayBuffer(spider)), (3,ArrayBuffer(dog, cat)), (5,ArrayBuffer(tiger, eagle)))
aggregateByKey
/**
* Aggregate the values of each key, using given combine functions and a neutral "zero value".
* This function can return a different result type, U, than the type of the values in this RDD,
* V. Thus, we need one operation for merging a V into a U and one operation for merging two U's,
* as in scala.TraversableOnce. The former operation is used for merging values within a
* partition, and the latter is used for merging values between partitions. To avoid memory
* allocation, both of these functions are allowed to modify and return their first argument
* instead of creating a new U.
*/
def aggregateByKey[U: ClassTag](zeroValue: U, partitioner: Partitioner)(seqOp: (U, V) => U,
combOp: (U, U) => U): RDD[(K, U)] = self.withScope {
// Serialize the zero value to a byte array so that we can get a new clone of it on each key
val zeroBuffer = SparkEnv.get.serializer.newInstance().serialize(zeroValue)
val zeroArray = new Array[Byte](zeroBuffer.limit)
zeroBuffer.get(zeroArray)
lazy val cachedSerializer = SparkEnv.get.serializer.newInstance()
val createZero = () => cachedSerializer.deserialize[U](ByteBuffer.wrap(zeroArray))
// We will clean the combiner closure later in `combineByKey`
val cleanedSeqOp = self.context.clean(seqOp)
combineByKeyWithClassTag[U]((v: V) => cleanedSeqOp(createZero(), v),
cleanedSeqOp, combOp, partitioner)
}
/**
* Aggregate the values of each key, using given combine functions and a neutral "zero value".
* This function can return a different result type, U, than the type of the values in this RDD,
* V. Thus, we need one operation for merging a V into a U and one operation for merging two U's,
* as in scala.TraversableOnce. The former operation is used for merging values within a
* partition, and the latter is used for merging values between partitions. To avoid memory
* allocation, both of these functions are allowed to modify and return their first argument
* instead of creating a new U.
*/
def aggregateByKey[U: ClassTag](zeroValue: U, numPartitions: Int)(seqOp: (U, V) => U,
combOp: (U, U) => U): RDD[(K, U)] = self.withScope {
aggregateByKey(zeroValue, new HashPartitioner(numPartitions))(seqOp, combOp)
}
/**
* Aggregate the values of each key, using given combine functions and a neutral "zero value".
* This function can return a different result type, U, than the type of the values in this RDD,
* V. Thus, we need one operation for merging a V into a U and one operation for merging two U's,
* as in scala.TraversableOnce. The former operation is used for merging values within a
* partition, and the latter is used for merging values between partitions. To avoid memory
* allocation, both of these functions are allowed to modify and return their first argument
* instead of creating a new U.
*/
def aggregateByKey[U: ClassTag](zeroValue: U)(seqOp: (U, V) => U,
combOp: (U, U) => U): RDD[(K, U)] = self.withScope {
aggregateByKey(zeroValue, defaultPartitioner(self))(seqOp, combOp)
}
例子:
val pairRDD = sc.parallelize(List( ("cat",2), ("cat", 5), ("mouse", 4),("cat", 12), ("dog", 12), ("mouse", 2)), 2)
// lets have a look at what is in the partitions
def myfunc(index: Int, iter: Iterator[(String, Int)]) : Iterator[String] = {
iter.map(x => "[partID:" + index + ", val: " + x + "]")
}
pairRDD.mapPartitionsWithIndex(myfunc).collect
res2: Array[String] = Array([partID:0, val: (cat,2)], [partID:0, val: (cat,5)], [partID:0, val: (mouse,4)], [partID:1, val: (cat,12)], [partID:1, val: (dog,12)], [partID:1, val: (mouse,2)])
pairRDD.aggregateByKey(0)(math.max(_, _), _ + _).collect
res3: Array[(String, Int)] = Array((dog,12), (cat,17), (mouse,6))
pairRDD.aggregateByKey(100)(math.max(_, _), _ + _).collect
res4: Array[(String, Int)] = Array((dog,100), (cat,200), (mouse,200))
combineByKey
/**
* Simplified version of combineByKeyWithClassTag that hash-partitions the resulting RDD using the
* existing partitioner/parallelism level. This method is here for backward compatibility. It
* does not provide combiner classtag information to the shuffle.
*
* @see `combineByKeyWithClassTag`
*/
def combineByKey[C](
createCombiner: V => C,
mergeValue: (C, V) => C,
mergeCombiners: (C, C) => C): RDD[(K, C)] = self.withScope {
combineByKeyWithClassTag(createCombiner, mergeValue, mergeCombiners)(null)
}
例子:
val pairRDD = sc.parallelize(List( ("cat",2), ("cat", 5), ("mouse", 4),("cat", 12), ("dog", 12), ("mouse", 2)), 2)
// lets have a look at what is in the partitions
def myfunc(index: Int, iter: Iterator[(String, Int)]) : Iterator[String] = {
iter.map(x => "[partID:" + index + ", val: " + x + "]")
}
pairRDD.mapPartitionsWithIndex(myfunc).collect
res2: Array[String] = Array([partID:0, val: (cat,2)], [partID:0, val: (cat,5)], [partID:0, val: (mouse,4)], [partID:1, val: (cat,12)], [partID:1, val: (dog,12)], [partID:1, val: (mouse,2)])
pairRDD.aggregateByKey(0)(math.max(_, _), _ + _).collect
res3: Array[(String, Int)] = Array((dog,12), (cat,17), (mouse,6))
pairRDD.aggregateByKey(100)(math.max(_, _), _ + _).collect
res4: Array[(String, Int)] = Array((dog,100), (cat,200), (mouse,200))
自己调试理解textFile分片过程
def textFile(
path: String,
minPartitions: Int = defaultMinPartitions): RDD[String] = withScope {
assertNotStopped()
hadoopFile(path, classOf[TextInputFormat], classOf[LongWritable], classOf[Text],
minPartitions).map(pair => pair._2.toString).setName(path)
}
如果给了分区数,就按照hadoopRDD来,这里用来getSplits所以参照hadoop的源码 FileInputFormat
如果没有给分区数,就按默认的来,先读配置文件dspark.default.parallelism ,有就按配置的来,没有就按默认给的的cores,然后在与2相比 拿出最大值。
注意:在实际过程中,当我们加载一个包含2以上的blocks个数的文件时,实际得到的分区个数并不是上面这个最大的默认分区数2。原因在官网上 已经说明,文件的分区数不能低于文件的blocks个数,跟踪到hadoopRDD可见。
重分区算子的理解
repartition(大佬说的翻译过来就是coalesce的简化版)
def repartition(numPartitions: Int)(implicit ord: Ordering[T] = null): RDD[T] = withScope {
coalesce(numPartitions, shuffle = true)
}
coalesce(shuffle默认为false,不传入true时,就不能多分区)
def coalesce(numPartitions: Int, shuffle: Boolean = false,
partitionCoalescer: Option[PartitionCoalescer] = Option.empty)
(implicit ord: Ordering[T] = null)
: RDD[T] = withScope {
require(numPartitions > 0, s"Number of partitions ($numPartitions) must be positive.")
if (shuffle) {
/** Distributes elements evenly across output partitions, starting from a random partition. */
val distributePartition = (index: Int, items: Iterator[T]) => {
var position = (new Random(index)).nextInt(numPartitions)
items.map { t =>
// Note that the hash code of the key will just be the key itself. The HashPartitioner
// will mod it with the number of total partitions.
position = position + 1
(position, t)
}
} : Iterator[(Int, T)]
// include a shuffle step so that our upstream tasks are still distributed
new CoalescedRDD(
new ShuffledRDD[Int, T, T](mapPartitionsWithIndex(distributePartition),
new HashPartitioner(numPartitions)),
numPartitions,
partitionCoalescer).values
} else {
new CoalescedRDD(this, numPartitions, partitionCoalescer)
}
}
checkpoint的应用场景
checkpoint的意思就是建立检查点,类似于快照,例如在spark计算里面 计算流程DAG特别长,服务器需要将整个DAG计算完成得出结果,但是如果在这很长的计算流程中突然中间算出的数据丢失了,spark又会根据RDD的依赖关系从头到尾计算一遍,这样子就很费性能,当然我们可以将中间的计算结果通过cache或者persist放到内存或者磁盘中,但是这样也不能保证数据完全不会丢失,存储的这个内存出问题了或者磁盘坏了,也会导致spark从头再根据RDD计算一遍,所以就有了checkpoint,其中checkpoint的作用就是将DAG中比较重要的中间数据做一个检查点将结果存储到一个高可用的地方(通常这个地方就是HDFS里面)
怎么建立checkpoint
首先需要用sparkContext设置hdfs的checkpoint的目录(如果不设置使用checkpoint会抛出异常:throw new SparkException(“Checkpoint directory has not been set in the SparkContext”):
scala> sc.setCheckpointDir("hdfs://spark01:8020/checkpoint")
//执行checkpoint
scala> val rdd1 = sc.parallelize(1 to 122)
scala> rdd1.checkpoint
现hdfs中还是没有数据,通过collect然后hdfs就有数据了,说明checkpoint也是个transformation的算子
但是执行的时候相当于走了两次流程,sum的时候前面计算了一遍,然后checkpoint又会计算一次,所以一般我们先进行cache然后做checkpoint就会只走一次流程,checkpoint的时候就会从刚cache到内存中取数据写入hdfs中,如下:
rdd.cache()
rdd.checkpoint()
rdd.collect
集群启动流程
1.调用start-all.sh脚本开始启动集群,首先启动Master服务
2.Master服务启动后,会调用一个定时器,定时检查超时的worker
3.通过解析slaves配置文件,开始在相应的节点启动Worker服务
4.worker服务启动后,开始向Master发送注册消息
5.Master收到注册信息并保存,开始向Worker相应成功注册的信息
6.Worker收到masterUrl后并保存,开始向Master发送心跳
任务提交流程
1.Driver向Master进行注册 (发送任务描述消息)
2.Master收到任务描述,开始生成任务信息,并放到队列中
3.Master开始通知Worker启动Executor并发送任务消息
4.Worker开始启动Executor
5.Eexecutor向Driver注册
6.Driver把生成的task发送给相应的Executor
基站停留时长需求的理解和实现过程
import org.apache.spark.rdd.RDD
import org.apache.spark.{SparkConf, SparkContext}
object MobileLocation {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setAppName("MobileLocation").setMaster("local")
val sc = new SparkContext(conf)
//获取用户访问信息
val files: RDD[String] = sc.textFile("D:\\ceshi/lacduration/log")
val splitedUserInfo: RDD[((String, String), Long)] = files.map(lines => {
val fields = lines.split(",")
val phone = fields(0) //手机号
val time = fields(1).toLong //时间戳
val lac = fields(2) //基站id
val eventType = fields(3).toInt
val time_long = if (eventType.equals("1")) -time else time
((phone, lac), time_long)
})
//计算用户在基站停留的总时长
val aggred: RDD[((String, String), Long)] = splitedUserInfo.reduceByKey(_ + _)
//为了方便和基站进行join,需要把数据重新整合
val lacAndPhoneAndTime = aggred.map(tup => {
val phone = tup._1._1 //手机号
val lac = tup._1._2 //基站id
val time = tup._2 //用户在某个基站停留的总时长
(lac, (phone, time))
})
//获取基站信息
val lacInfo = sc.textFile("D:\\ceshi/lac_info.txt")
//切分基站信息
val splitLacInfo: RDD[(String, (String, String))] = lacInfo.map(line => {
val fields = line.split(",")
val lac = fields(0) //基站id
val x = fields(1) //经度
val y = fields(2) //维度
(lac, (x, y))
})
//用户信息和基站信息进行join
val joined: RDD[(String, ((String, Long), (String, String)))] = lacAndPhoneAndTime.join(splitLacInfo)
//整合join后的数据,便于计算
val lacAndPhoneAndTimeAndXY: RDD[(String, String, Long, (String, String))] = joined.map(tup => {
val lac = tup._1 //基站
val phone = tup._2._1._1 //手机号
val time = tup._2._1._2 //基站停留时长
val xy = tup._2._2 //经纬度
(lac, phone, time, xy)
})
//为了便于统计用户在基站的信息 生成用户粒度的数据
val grouped: RDD[(String, Iterable[(String, String, Long, (String, String))])] = lacAndPhoneAndTimeAndXY.groupBy(_._2)
//按照时长进行分组排序
val sorted: RDD[(String, List[(String, String, Long, (String, String))])] = grouped.map(x => (x._1, x._2.toList.sortBy(_._3).reverse))
//获取top2
val res: RDD[(String, List[(String, String, Long, (String, String))])] = sorted.mapValues(_.take(2))
println(res.collect.toBuffer)
sc.stop()
}
}