1、基本数据源
1.1、文件流
在spark Shell 下运行:
[lyh@hadoop102 spark-yarn-3.2.4]$ spark-shell
Setting default log level to "WARN".
To adjust logging level use sc.setLogLevel(newLevel). For SparkR, use setLogLevel(newLevel).
2022-09-08 08:56:21,875 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Spark context Web UI available at http://hadoop102:4040
Spark context available as 'sc' (master = local[*], app id = local-1662598583370).
Spark session available as 'spark'.
Welcome to
____ __
/ __/__ ___ _____/ /__
_\ \/ _ \/ _ `/ __/ '_/
/___/ .__/\_,_/_/ /_/\_\ version 3.2.4
/_/
Using Scala version 2.12.15 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_241)
Type in expressions to have them evaluated.
Type :help for more information.
scala> import org.apache.spark.streaming._
import org.apache.spark.streaming._
scala> val ssc = new StreamingContext(sc,Seconds(20))
ssc: org.apache.spark.streaming.StreamingContext = org.apache.spark.streaming.StreamingContext@379899f4
scala> val lines = ssc.textFileStream("file:///home/lyh/streaming/logfile")
lines: org.apache.spark.streaming.dstream.DStream[String] = org.apache.spark.streaming.dstream.MappedDStream@531245fe
scala> val kv = lines.map((_,1)).reduceByKey(_+_)
kv: org.apache.spark.streaming.dstream.DStream[(String, Int)] = org.apache.spark.streaming.dstream.ShuffledDStream@c207c10
scala> kv.print()
scala> ssc.start()
------------------------------------------
Time: 1662598860000 ms
-------------------------------------------
-------------------------------------------
Time: 1662598880000 ms
-------------------------------------------
-------------------------------------------
Time: 1662598900000 ms
-------------------------------------------
(c#,1)
(hh,1)
(h,1)
(javafx,1)
(spark,1)
(hadoop,1)
(js,1)
(java,1)
(s,1)
(c,1)
执行后立即新建终端在 /home/lyh/streaming/logfile 目录下创建文件并写入数据
1.2、Socket 套接字流
// todo 创建环境对象
val conf = new SparkConf()
conf.setAppName("word count").setMaster("local[*]")
val ssc = new StreamingContext(conf,Seconds(3))
// todo 逻辑处理
// 获取端口数据(Socket)
val lines: ReceiverInputDStream[String] = ssc.socketTextStream("localhost", 9999)
val words: DStream[String] = lines.flatMap(_.split(" "))
val word: DStream[(String,Int)] = words.map((_, 1))
val wordCount: DStream[(String,Int)] = word.reduceByKey(_ + _)
wordCount.print()
// todo 关闭环境
// 由于SparkStreaming的采集器是长期运行的,所以不能直接关闭
// 而且main方法的关闭也会使SparkStreaming的采集器关闭
ssc.start()
// 等待采集器关闭
ssc.awaitTermination()
启动 NetCat
> nc -lp 9999
> hello world
> hello spark
> ...
运行结果:
1.3、自定义 Socket 数据源
通过自定义 Socket 实现数据源不断产生数据
import java.io.PrintWriter
import java.net.{ServerSocket, Socket}
import scala.io.Source
/**
* 通过自定义的Socket来不断给客户端发送数据
*/
object MySocketReceiver {
def index(length: Int): Int = {
val rdm = new java.util.Random()
rdm.nextInt(length)
}
def main(args: Array[String]): Unit = {
val fileName = "input/1.txt"
val lines: List[String] = Source.fromFile(fileName).getLines().toList
val listener: ServerSocket = new ServerSocket(9999)
while(true){
val socket: Socket = listener.accept()
new Thread(){
override def run(){
val out: PrintWriter = new PrintWriter(socket.getOutputStream,true)
while (true){
Thread.sleep(1000)
val content = lines(index(lines.length)) // 源源不断,每次打印list的第(1~length)随机行
println(content)
out.write(content + '\n')
out.flush()
}
socket.close()
}
}.start()
}
}
}
定义一个处理器接收自定义数据源端口发送过来的数据。
def main(args: Array[String]): Unit = {
// todo 创建环境对象
val conf = new SparkConf()
conf.setAppName("word count").setMaster("local[*]")
val ssc = new StreamingContext(conf,Seconds(3))
// todo 逻辑处理
// 获取端口数据(Socket)
val lines: ReceiverInputDStream[String] = ssc.socketTextStream("localhost", 9999)
val words: DStream[String] = lines.flatMap(_.split(" "))
val word: DStream[(String,Int)] = words.map((_, 1))
val wordCount: DStream[(String,Int)] = word.reduceByKey(_ + _)
wordCount.print()
// todo 关闭环境
// 由于SparkStreaming的采集器是长期运行的,所以不能直接关闭
ssc.start()
// 等待采集器关闭
ssc.awaitTermination()
}
先运行我们的数据源,再运行处理器:
处理器:
1.4、RDD 队列流
import org.apache.spark.SparkConf
import org.apache.spark.rdd.RDD
import org.apache.spark.streaming.dstream.{DStream, InputDStream}
import org.apache.spark.streaming.{Seconds, StreamingContext}
import scala.collection.mutable
object SparkStreaming02_RDDStream {
def main(args: Array[String]): Unit = {
// 1. 初始化配置信息
val conf = new SparkConf()
conf.setAppName("rdd Stream").setMaster("local[*]")
// 2.初始化SparkStreamingContext
val ssc = new StreamingContext(conf,Seconds(4))
// 3.创建RDD队列
val rddQueue: mutable.Queue[RDD[Int]] = new mutable.Queue[RDD[Int]]()
// 4.创建QueueInputStream
// oneAtATime = true 默认,一次读取队列里面的一个数据
// oneAtATime = false, 按照设定的时间,读取队列里面数据
val inputStream: InputDStream[Int] = ssc.queueStream(rddQueue,oneAtATime = false)
// 5. 处理队列中的RDD数据
val sumStream: DStream[Int] = inputStream.reduce(_ + _)
// 6. 打印结果
sumStream.print()
// 7.启动任务
ssc.start()
// 8.向队列中放入RDD
for(i <- 1 to 5){
rddQueue += ssc.sparkContext.makeRDD(1 to 5)
Thread.sleep(2000)
}
// 9. 等待数据源进程停止后关闭
ssc.awaitTermination()
}
}
2、高级数据源
2.1、Kafka 数据源
2.1.1、消费者程序处理流数据
import org.apache.kafka.clients.consumer.{ConsumerConfig, ConsumerRecord}
import org.apache.kafka.common.serialization.StringDeserializer
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.{DStream, InputDStream}
import org.apache.spark.streaming.kafka010.{ConsumerStrategies, KafkaUtils, LocationStrategies}
import org.apache.spark.streaming.{Seconds, StreamingContext}
object SparkStreaming03_Kafka {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setMaster("local[*]").setAppName("kafka source")
val ssc = new StreamingContext(conf,Seconds(3))
// 定义Kafka参数: kafka集群地址、消费者组名称、key序列化、value序列化
val kafkaPara: Map[String,Object] = Map[String,Object](
ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG -> "hadoop102:9092,hadoop103:9092,hadoop104:9092",
ConsumerConfig.GROUP_ID_CONFIG ->"lyh",
ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG -> "org.apache.kafka.common.serialization.StringDeserializer",
ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG -> classOf[StringDeserializer]
)
// 读取Kafka数据创建DStream
val kafkaDStream: InputDStream[ConsumerRecord[String,String]] = KafkaUtils.createDirectStream[String,String](
ssc,
LocationStrategies.PreferConsistent, //优先位置
ConsumerStrategies.Subscribe[String,String](Set("testTopic"),kafkaPara) // 消费策略:(订阅多个主题,配置参数)
)
// 将每条消息的KV取出
val valueDStream: DStream[String] = kafkaDStream.map(_.value())
// 计算WordCount
valueDStream.flatMap(_.split(" "))
.map((_,1))
.reduceByKey(_+_)
.print()
// 开启任务
ssc.start()
ssc.awaitTermination()
}
}
2.1.2、生产者生产数据
(1)kafka 端生产数据
启动 Kafka 集群
创建 Topic(指定一个分区三个副本):
kafka-topics.sh --bootstrap-server hadoop102:9092 --topic <topic名称> --create --partitions 1 --replication-factor 3
查看是否生成 Topic:
kafka-topics.sh --bootstrap-server hadoop102:9092 --list
生产者生产数据:
> kafka-console-producer.sh --bootstrap-server hadoop102:9092 --topic <topic名称>
> hello world
> hello spark
> ...
(2)编写生产者程序
package com.lyh
import org.apache.kafka.clients.consumer.{ConsumerConfig, ConsumerRecord}
import org.apache.kafka.common.serialization.StringDeserializer
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.{DStream, InputDStream}
import org.apache.spark.streaming.kafka010.{ConsumerStrategies, KafkaUtils, LocationStrategies}
import org.apache.spark.streaming.{Seconds, StreamingContext}
object SparkStreaming03_Kafka {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setMaster("local[*]").setAppName("kafka source")
val ssc = new StreamingContext(conf,Seconds(3))
// 定义Kafka参数: kafka集群地址、消费者组名称、key序列化、value序列化
val kafkaPara: Map[String,Object] = Map[String,Object](
ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG -> "hadoop102:9092,hadoop103:9092,hadoop104:9092",
ConsumerConfig.GROUP_ID_CONFIG ->"lyh",
ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG -> "org.apache.kafka.common.serialization.StringDeserializer",
ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG -> classOf[StringDeserializer]
)
// 读取Kafka数据创建DStream
val kafkaDStream: InputDStream[ConsumerRecord[String,String]] = KafkaUtils.createDirectStream[String,String](
ssc,
LocationStrategies.PreferConsistent, //优先位置
ConsumerStrategies.Subscribe[String,String](Set("testTopic"),kafkaPara) // 消费策略:(订阅多个主题,配置参数)
)
// 将每条消息的KV取出
val valueDStream: DStream[String] = kafkaDStream.map(_.value())
// 计算WordCount
valueDStream.flatMap(_.split(" "))
.map((_,1))
.reduceByKey(_+_)
.print()
// 开启任务
ssc.start()
ssc.awaitTermination()
}
}
3、转换操作
3.1、无状态转换操作
3.2、有状态转换操作
3.1.1、滑动窗口转换操作
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.DStream
import org.apache.spark.streaming.{Seconds, StreamingContext}
object SparkStreaming05_Window {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setMaster("local[*]").setAppName("sparkStreaming window")
val ssc = new StreamingContext(conf,Seconds(3))
val lines:DStream[String] = ssc.socketTextStream("localhost", 9999)
val word_kv = lines.map((_, 1))
/**
* 收集器收集RDD合成DStream: 3s 窗口范围: 12s 窗口滑动间隔: 6s/次
* 1. windowLength:表示滑动窗口的长度,即窗口内包含的数据的时间跨度。它是一个Duration对象,用于指定窗口的时间长度。
* 2. slideInterval:表示滑动窗口的滑动间隔,即每隔多长时间将窗口向右滑动一次。同样是一个Duration对象。
* 返回一个新的 DStream
**/
val wordToOneByWindow:DStream[(String,Int)] = word_kv.window(Seconds(12), Seconds(6))
// 窗口每滑动一次(6s),对窗口内的数据进行一次聚合操作.
val res: DStream[(String,Int)] = wordToOneByWindow.reduceByKey(_ + _)
res.print()
ssc.start()
ssc.awaitTermination()
}
}
3.1.2、updateStateByKey
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.DStream
import org.apache.spark.streaming.{Seconds, StreamingContext}
/**
* DStream 有状态转换操作之 updateStateByKey(func) 转换操作
*/
object SparkStreaming04_State {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setMaster("local[*]").setAppName("kafka state")
val ssc = new StreamingContext(conf,Seconds(3))
/**
* 设置检查点目录的作用是为了确保Spark Streaming应用程序的容错性和可恢复性。
* 在Spark Streaming应用程序运行过程中,它会将接收到的数据分成一批批进行处理。
* 通过设置检查点目录,Spark Streaming会定期将当前的处理状态、接收到的数据偏移量等信息保存到可靠的存储系统中,
* 比如分布式文件系统(如HDFS)或云存储服务(如Amazon S3)。
* 一旦应用程序出现故障或崩溃,它可以从最近的检查点中恢复状态,并从上次处理的位置继续处理数据,从而确保数据的完整性和一致性。
*/
//检查点的路径如果是本地路径要+ file:// 否则认为是 hdfs路径 / 开头
ssc.checkpoint("file:///D://IdeaProject/SparkStudy/data/") //设置检查点,检查点具有容错机制
val lines: DStream[String] = ssc.socketTextStream("localhost",9999)
val word_kv = lines.map((_, 1))
val stateDStream: DStream[(String, Int)] = word_kv.updateStateByKey(
/** 参数是一个函数
1. Seq[Int]: 当前key对应的所有value值的集合,因为我们的value是Int,所以这里也是Int
2. Option[Int]: 当前key的历史状态,对于wordCount,历史值就是上一个DStream中这个key的value计算结果(求和结果)
Option 是 Scala 中用来表示可能存在或可能不存在的值的容器,是一种避免空引用(null reference)问题的模式。
Option[Int] 有两个可能的实例:
(1) Some(value: Int):表示一个包含 Int 类型值的 Option。
(2) None:表示一个空的 Option,不包含任何值。
**/
(values: Seq[Int], state: Option[Int]) => {
val currentCount = values.foldLeft(0)(_ + _)
val previousCount = state.getOrElse(0)
Option(currentCount + previousCount)
}
)
stateDStream.print()
stateDStream.saveAsTextFiles("./out") //输出结果保存到 文本文件中
ssc.start()
ssc.awaitTermination()
}
}
4、输出操作
4.1、输出到文本文件
上面 3.1.2 中就保存DStream输出到了本地:
stateDStream.saveAstextFiles("./out")
4.2、输出到MySQL数据库
import org.apache.spark.SparkConf
import org.apache.spark.rdd.RDD
import org.apache.spark.streaming.dstream.{DStream, ReceiverInputDStream}
import org.apache.spark.streaming.{Seconds, StreamingContext}
import java.sql.{Connection, PreparedStatement}
object NetWorkWordCountStateMySQL {
def main(args: Array[String]): Unit = {
val updateFunc = (values: Seq[Int],state: Option[Int]) => {
val currentCount = values.foldLeft(0)(_+_)
val previousCount = state.getOrElse(0)
Some(currentCount + previousCount)
}
val conf = new SparkConf().setMaster("local[*]").setAppName("state mysql")
val ssc = new StreamingContext(conf,Seconds(5))
// file:\\ 代表本地文件系统 如果用的是 /user/... 这种形式是 HDFS 文件系统 需要启动Hadoop集群
ssc.checkpoint("file:\\D:\\IdeaProjects\\SparkStudy\\data\\state")
val lines: ReceiverInputDStream[String] = ssc.socketTextStream("localhost", 9999)
val word_kv: DStream[(String, Int)] = lines.flatMap(_.split(" ").map((_, 1))).reduceByKey(_ + _)
val stateDStream: DStream[(String, Int)] = word_kv.updateStateByKey[Int](updateFunc)
stateDStream.print()
stateDStream.foreachRDD( rdd=> {
def func(records: Iterator[(String,Int)]): Unit ={
var conn: Connection = null
var stmt: PreparedStatement = null
try{
conn = DBUtils.getConnection("jdbc:mysql://127.0.0.1:3306/spark","root","Yan1029.")
records.foreach(p=>{
val sql = "insert into wordcount values (?,?)"
stmt = conn.prepareStatement(sql)
stmt.setString(1,p._1.trim)
stmt.setInt(2,p._2)
stmt.executeUpdate() //不executeUpdate就不会写入数据库
})
}catch {
case e: Exception => e.printStackTrace()
}finally {
// if (stmt!=null) stmt.close()
// DBUtils.close()
}
}
val repartitionedRDD: RDD[(String,Int)] = rdd.repartition(3) //扩大分区用 repartition
repartitionedRDD.foreachPartition(func)
})
ssc.start()
ssc.awaitTermination()
}
}
运行结果:
5、优雅的关闭和恢复数据
5.1、关闭SparkStreaming
流式任务通常都需要7*24小时执行,但是有时涉及到升级代码需要主动停止程序,但是分布式程序,没办法做到一个个进程去杀死,所以配置优雅的关闭就显得至关重要了。
关闭方式:我们通常使用外部文件系统来控制内部程序关闭。
package com.lyh
import org.apache.hadoop.conf.Configuration
import org.apache.hadoop.fs.{FileSystem, Path}
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.DStream
import org.apache.spark.streaming.{Seconds, StreamingContext, StreamingContextState}
import java.net.URI
object SparkStreaming06_Close {
def main(args: Array[String]): Unit = {
val conf = new SparkConf().setMaster("local[*]").setAppName("sparkStreaming close")
val ssc = new StreamingContext(conf,Seconds(3))
val lines:DStream[String] = ssc.socketTextStream("localhost", 9999)
val word_kv = lines.map((_, 1))
word_kv.print()
ssc.start()
// 再创建一个线程去关闭
new Thread(new MonitorStop(ssc)).start()
ssc.awaitTermination() //阻塞当前main线程
}
}
class MonitorStop(ssc: StreamingContext) extends Runnable{
override def run(): Unit = {
while (true){ // 一直轮询判断
Thread.sleep(5000) //每5s检查一遍
val fs: FileSystem = FileSystem.get(new URI("hdfs://hadoop102:9000"),new Configuration(),"lyh")
val exists: Boolean = fs.exists(new Path("hdfs://hadoop102:9000/stopSpark"))
if (exists) { //如果比如(MySQL出现了一行数据、Zookeeper的某个节点出现变化、hdfs是否存在某个目录...)就关闭
val state: StreamingContextState = ssc.getState()
if (state == StreamingContextState.ACTIVE){
// 优雅地关闭-处理完当前的数据再关闭
// 计算节点不再接受新的数据,而是把现有的数据处理完毕,然后关闭
ssc.stop(true,true)
System.exit(0)
}
}
}
}
}
5.2、恢复检查点的数据
使用 getActiveOrCreate 的方法来对上一个失败的 Spark 任务进行数据恢复(通过检查点来进行恢复)
方法说明:
若Application为首次重启,将创建一个新的StreamingContext实例;如果Application从失败中重启,从checkpoint目录导入checkpoint数据来重新创建StreamingContext实例。
import org.apache.spark.SparkConf
import org.apache.spark.streaming.dstream.DStream
import org.apache.spark.streaming.{Seconds, StreamingContext, StreamingContextState}
import java.net.URI
object SparkStreaming07_Resume {
def main(args: Array[String]): Unit = {
//好处:若Application为首次重启,将创建一个新的StreamingContext实例;如果Application从失败中重启,从checkpoint目录导入checkpoint数据来重新创建StreamingContext实例。
val ssc: StreamingContext = StreamingContext.getActiveOrCreate("file:\\D:\\IdeaProjects\\SparkStudy\\data\\state", () => {
val conf = new SparkConf().setMaster("local[*]").setAppName("sparkStreaming resume")
val ssc = new StreamingContext(conf, Seconds(3))
val lines: DStream[String] = ssc.socketTextStream("localhost", 9999)
val word_kv = lines.map((_, 1))
word_kv.print()
ssc
})
// 依然设置检查点 防止application失败后丢失数据
ssc.checkpoint("file:\\D:\\IdeaProjects\\SparkStudy\\data\\state")
ssc.start()
ssc.awaitTermination() //阻塞当前main线程
}
}