Spark 广播变量和计数器

spark广播变量

	将外部变量发送到executor中使用。

注意事项
1、不能，因为RDD是不存储数据的。可以将RDD的结果广播出去。

2、 广播变量只能在Driver端定义，不能在Executor端定义。

3、 在Driver端可以修改广播变量的值，在Executor端无法修改广播变量的值。

4、如果executor端用到了Driver的变量，如果不使用广播变量在Executor有多少task就有多少Driver端的变量副本。

5、如果Executor端用到了Driver的变量，如果使用广播变量在每个Executor中只有一份Driver端的变量副本。

静态广播变量使用

		val conf = new SparkConf()
		conf.setMaster("local").setAppName("brocast")
		val sc = new SparkContext(conf)
		val list = List("hello xasxt")
		val broadCast = sc.broadcast(list)
		val lineRDD = sc.textFile("./words.txt")
		lineRDD.filter { x => broadCast.value.contains(x) }.foreach { println}
		sc.stop()

动态广播变量使用

在实际项目中,有时候我们的广播变量是动态的,比如需要一分钟更新一次,这个也是可以实现的,我们知道广播变量是在driver端初始化,在excetors端获取这个变量,但是不能修改,所以,我们可以在driver端进行更新这个变量

package com.unionpay.ysf

import java.sql.{Connection, DriverManager, ResultSet, Statement}
import java.text.SimpleDateFormat
import java.util.Date
import org.apache.kafka.clients.consumer.ConsumerRecord
import org.apache.kafka.common.serialization.StringDeserializer
import org.apache.log4j.{Level, Logger}
import org.apache.spark.broadcast.Broadcast
import org.apache.spark.streaming.dstream.InputDStream
import org.apache.spark.streaming.kafka010.{ConsumerStrategies, HasOffsetRanges, KafkaUtils, LocationStrategies}
import org.apache.spark.streaming.{Seconds, StreamingContext}
import org.apache.spark.{SparkConf, SparkContext}


object test3 {
  @volatile private var instance: Broadcast[Map[String, Double]] = null
  var kafkaStreams: InputDStream[ConsumerRecord[String, String]] = null
  val sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss:SSS")
  def main(args: Array[String]): Unit = {
    Logger.getLogger("org.apache.spark").setLevel(Level.INFO)
    Logger.getLogger("org.eclipse.jetty.server").setLevel(Level.INFO)
    Logger.getLogger("org.apache.kafka.clients.consumer").setLevel(Level.INFO)
    val conf = new SparkConf().setAppName("Spark Streaming TO ES TOPIC")
    conf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
    @transient
    val scc = new StreamingContext(conf, Seconds(1))
    val topic ="topic_combine"
    val topicSet = Set(topic) //设置kafka的topic;
    val kafkaParams = Map[String, Object](
      "auto.offset.reset" -> "earliest",   //latest;earliest
      "value.deserializer" -> classOf[StringDeserializer] //key,value的反序列化;
      , "key.deserializer" -> classOf[StringDeserializer]
      , "bootstrap.servers" -> "192.168.38.12:9092"
      , "group.id" -> "groupId_es"
      , "enable.auto.commit" -> (false: java.lang.Boolean)
    )
    //初始化instance;
    getInstance(scc.sparkContext)
    kafkaStreams = KafkaUtils.createDirectStream[String, String](
      scc,
      LocationStrategies.PreferConsistent,
      ConsumerStrategies.Subscribe[String, String](topicSet, kafkaParams))
    kafkaStreams.foreachRDD(rdd => {
      val current_time = sdf.format(new Date())
      val new_time = current_time.substring(14,16).toLong
      if(new_time % 5 == 0){
        update(rdd.sparkContext,true) //五分钟更新一次广播变量的内容;
      }
      if (!rdd.isEmpty()) {
        val offsetRanges = rdd.asInstanceOf[HasOffsetRanges].offsetRanges //获得偏移量对象数组
        rdd.foreachPartition(pr => {
          pr.foreach(pair => {
            val d = pair.value()
            if(instance.value.contains(d)){
              //自己的处理逻辑;
            }
          })
        })
      }
    })
    scc.start()
    scc.awaitTermination()
  }

  /**
   * 从sqlserver获取数据放到一个map里;
   * @return
   */
  def getSqlServerData(): Map[String,Double] = {
    val time = sdf.format(new Date())
    val enter_time = time.substring(0,10)
    var map = Map[String,Double]()
    var conn:Connection = null
    var stmt:Statement = null
    var rs:ResultSet = null
    val url = ""
    val user_name = ""
    val password = ""
    val sql = ""
    try {
      conn = DriverManager.getConnection(url,user_name,password)
      stmt = conn.createStatement
      rs = stmt.executeQuery(sql)
      while (rs.next) {
        val url = rs.getString("url")
        val WarningPrice = rs.getString("WarningPrice").toDouble
        map += (url -> WarningPrice)
      }
      if (rs != null) {
        rs.close
        rs = null
      }
      if (stmt != null) {
        stmt.close
        stmt = null
      }
      if (conn != null) {
        conn.close
        conn = null
      }
    } catch {
      case e: Exception => e.printStackTrace()
        println("sqlserver连接失败:" + e)
    }
    map
  }

  /**
   * 更新instance;
   * @param sc
   * @param blocking
   */
  def update(sc: SparkContext, blocking: Boolean = false): Unit = {
    if (instance != null){
      instance.unpersist(blocking)
      instance = sc.broadcast(getSqlServerData())
    }
  }

  /**
   * 初始化instance;
   * @param sc
   * @return
   */
  def getInstance(sc: SparkContext): Broadcast[Map[String,Double]] = {
    if (instance == null) {
      synchronized {
        if (instance == null) {
          instance = sc.broadcast(getSqlServerData())
        }
      }
    }
    instance
  }
}

spark累加器

	相当于统筹大变量，常用于计数，统计。

import org.apache.spark.{SparkConf, SparkContext}
 
object AccumulatorOperator {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf()
    conf.setMaster("local").setAppName("accumulator")
    val sc = new SparkContext(conf)
    val accumulator = sc.accumulator(0)
    sc.textFile("./records.txt",2).foreach {//两个变量
      x =>{accumulator.add(1)
      println(accumulator)}}
    println(accumulator.value)
    sc.stop()
  }
}

package com.spark.spark.others;

import org.apache.spark.Accumulator;
import org.apache.spark.SparkConf;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.api.java.function.VoidFunction;
/**
 * 累加器在Driver端定义赋初始值和读取，在Executor端累加。
 * @author root
 *
 */
public class AccumulatorOperator {
    public static void main(String[] args) {
        SparkConf conf = new SparkConf();
        conf.setMaster("local").setAppName("accumulator");
        JavaSparkContext sc = new JavaSparkContext(conf);
        final Accumulator<Integer> accumulator = sc.accumulator(0);
//        accumulator.setValue(1000);
        sc.textFile("./words.txt",2).foreach(new VoidFunction<String>() {
            
            /**
             * 
             */
            private static final long serialVersionUID = 1L;

            @Override
            public void call(String t) throws Exception {
                accumulator.add(1);
//                System.out.println(accumulator.value());
                System.out.println(accumulator);
            }
        });
        System.out.println(accumulator.value());
        sc.stop();
        
    }
}

注意事项
累加器在Driver端定义赋初始值，累加器只能在Driver端读取最后的值，在Excutor端更新。