spark streaming集成kafka

Kakfa起初是由LinkedIn公司开发的一个分布式的消息系统，后成为Apache的一部分，它使用Scala编写，以可水平扩展和高吞吐率而被广泛使用。目前越来越多的开源分布式处理系统如Cloudera、Apache Storm、Spark等都支持与Kafka集成。

Spark streaming集成kafka是企业应用中最为常见的一种场景。

一、安装kafka

参考文档：

http://kafka.apache.org/quickstart#quickstart_createtopic

1、安装java

略

2、安装zookeeper集群

参考：http://www.cnblogs.com/wcwen1990/p/6652105.html

3、安装scala

略

4、安装kafka

下载kafka安装文件：

https://archive.apache.org/dist/kafka/0.8.2.1/kafka_2.10-0.8.2.1.tgz

解压kafka安装包：

# tar -zxvf kafka_2.10-0.8.2.1.tgz -C /opt/cdh-5.3.6/

# chown -R hadoop:hadoop /opt/cdh-5.3.6/kafka_2.10-0.8.2.1/

删除kafka libs/zookeeper jar包,拷贝自己安装集群zookeeper jar包到kafka libs目录下：

$ rm libs/zookeeper-3.4.6.jar –rf

$ cp /opt/cdh-5.3.6/zookeeper-3.4.5-cdh5.3.6/zookeeper-3.4.5-cdh5.3.6.jar libs/

5、定义kafka配置文件

5.1）定义server.properties：

host.name=chavin.king

log.dirs=/opt/cdh-5.3.6/kafka_2.10-0.8.2.1/kafka-logs

zookeeper.connect=chavin.king:2181

定义producer.properties：

metadata.broker.list=chavin.king:9092

定义consumer.properties：

zookeeper.connect=chavin.king:2181

5.2）启动kafka server

$ bin/kafka-server-start.sh config/server.properties

$ jps

14020 NameNode

57749 Jps

14776 QuorumPeerMain

57690 Kafka

14507 NodeManager

14235 ResourceManager

14093 DataNode

14686 JobHistoryServer

57663 ZooKeeperMain

[zk: localhost:2181(CONNECTED) 3] ls /

[controller, controller_epoch, brokers, zookeeper, admin, consumers, config, hbase]

5.3）创建一个topic

$ bin/kafka-topics.sh --create --zookeeper chavin.king:2181 --replication-factor 1 --partitions 1 --topic test

$ bin/kafka-topics.sh --list --zookeeper chavin.king:2181

5.4）创建一个生产者，产生数据

$ bin/kafka-console-producer.sh --broker-list chavin.king:9092 --topic test

5.5）创建一个消费者，消费数据

$ bin/kafka-console-consumer.sh --zookeeper chavin.king:2181 --topic test --from-beginning

在生产者shell窗口输入数据，在消费者窗口可以看到数据输出到界面上。

二、spark streaming与kafka集成

参考文档：http://spark.apache.org/docs/1.3.0/streaming-kafka-integration.html

一）准备工作

1、编译spark，获得集成kafka jar包：

参考文档：http://www.cnblogs.com/wcwen1990/p/7688027.html

说明：spark streaming集成flume或者kafka需要一些支持jar包，这些jar包在编译spark过程中会自动在external目录下生成相应的jar文件，因此，这里需要编译spark来获得这些jar包。

Spark streaming集成kafka主要需要：spark-streaming-kafka_2.10-1.3.0.jar包。

2、集成相关jar包

$ cp external/kafka/target/spark-streaming-kafka_2.10-1.3.0.jar /opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/

$ cp libs/kafka_2.10-0.8.2.1.jar libs/kafka-clients-0.8.2.1.jar libs/zkclient-0.3.jar libs/metrics-core-2.2.0.jar /opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/

[externalLibs]$ ls

kafka_2.10-0.8.2.1.jar

kafka-clients-0.8.2.1.jar

metrics-core-2.2.0.jar

spark-streaming-kafka_2.10-1.3.0.jar

zkclient-0.3.jar

二）集成方式1：Receiver-based Approach

1、编写spark streaming集成kafka的wordcount

import java.util.HashMap

import org.apache.spark._

import org.apache.spark.streaming._

import org.apache.spark.streaming.StreamingContext._

import org.apache.spark.streaming.kafka._

val ssc = new StreamingContext(sc, Seconds(5))

val topicMap = Map("test" -> 1)

// read data

val lines = KafkaUtils.createStream(ssc, "chavin.king:2181", "testWordCountGroup", topicMap).map(_._2)

val words = lines.flatMap(_.split(" "))

val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _)

wordCounts.print()

ssc.start() // Start the computation

ssc.awaitTermination() // Wait for the computation to terminate

2、spark-shell local模式启动，并运行步骤1程序

bin/spark-shell --master local[2] --jars \

/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/spark-streaming-kafka_2.10-1.3.0.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka_2.10-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka-clients-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/zkclient-0.3.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/metrics-core-2.2.0.jar

scala> import java.util.HashMap

import java.util.HashMap

scala> import org.apache.spark._

import org.apache.spark._

scala> import org.apache.spark.streaming._

import org.apache.spark.streaming._

scala> import org.apache.spark.streaming.StreamingContext._

import org.apache.spark.streaming.StreamingContext._

scala> import org.apache.spark.streaming.kafka._

import org.apache.spark.streaming.kafka._

scala> val ssc = new StreamingContext(sc, Seconds(5))

ssc: org.apache.spark.streaming.StreamingContext = org.apache.spark.streaming.StreamingContext@1a28f9a0

scala> val topicMap = Map("test" -> 1)

topicMap: scala.collection.immutable.Map[String,Int] = Map(test -> 1)

scala> val lines = KafkaUtils.createStream(ssc, "chavin.king:2181", "testWordCountGroup", topicMap).map(_._2)

lines: org.apache.spark.streaming.dstream.DStream[String] = org.apache.spark.streaming.dstream.MappedDStream@27267641

scala>

scala> val words = lines.flatMap(_.split(" "))

words: org.apache.spark.streaming.dstream.DStream[String] = org.apache.spark.streaming.dstream.FlatMappedDStream@169b0639

scala> val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _)

wordCounts: org.apache.spark.streaming.dstream.DStream[(String, Int)] = org.apache.spark.streaming.dstream.ShuffledDStream@14f2b1ba

scala> wordCounts.print()

scala> ssc.start()

scala>ssc.awaitTermination()

3、测试

在kafka生产者shell端输入：

hadoop oracle mysql mysql mysql

这是我们在kafka消费者端可以看到如下输出：

hadoop oracle mysql mysql mysql

同时在spark streaming端也可以看到如下输出：

-------------------------------------------

Time: 1500021590000 ms

-------------------------------------------

(mysql,3)

(oracle,1)

(hadoop,1)

三）集成方式2：Direct Approach (No Receivers)

1、编写spark streaming集成kafka的wordcount

import kafka.serializer.StringDecoder

import org.apache.spark._

import org.apache.spark.streaming._

import org.apache.spark.streaming.StreamingContext._

import org.apache.spark.streaming.kafka._

val ssc = new StreamingContext(sc, Seconds(5))

val kafkaParams = Map[String, String]("metadata.broker.list" -> "chavin.king:9092")

val topicsSet = Set("test")

// read data

val messages = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topicsSet)

val lines = messages.map(_._2)

val words = lines.flatMap(_.split(" "))

val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _)

wordCounts.print()

ssc.start() // Start the computation

ssc.awaitTermination() // Wait for the computation to terminate

2、spark-shell local模式启动，并运行步骤1程序

bin/spark-shell --master local[2] --jars \

/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/spark-streaming-kafka_2.10-1.3.0.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka_2.10-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka-clients-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/zkclient-0.3.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/metrics-core-2.2.0.jar

scala> import kafka.serializer.StringDecoder

import kafka.serializer.StringDecoder

scala> import org.apache.spark._

import org.apache.spark._

scala> import org.apache.spark.streaming._

import org.apache.spark.streaming._

scala> import org.apache.spark.streaming.StreamingContext._

import org.apache.spark.streaming.StreamingContext._

scala> import org.apache.spark.streaming.kafka._

import org.apache.spark.streaming.kafka._

scala>

scala> val ssc = new StreamingContext(sc, Seconds(5))

ssc: org.apache.spark.streaming.StreamingContext = org.apache.spark.streaming.StreamingContext@2d05daca

scala>

scala> val kafkaParams = Map[String, String]("metadata.broker.list" -> "chavin.king:9092")

kafkaParams: scala.collection.immutable.Map[String,String] = Map(metadata.broker.list -> chavin.king:9092)

scala> val topicsSet = Set("test")

topicsSet: scala.collection.immutable.Set[String] = Set(test)

scala>

scala> // read data

scala> val messages = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topicsSet)

17/07/14 16:59:31 INFO VerifiableProperties: Verifying properties

17/07/14 16:59:31 INFO VerifiableProperties: Property group.id is overridden to

17/07/14 16:59:31 INFO VerifiableProperties: Property zookeeper.connect is overridden to

messages: org.apache.spark.streaming.dstream.InputDStream[(String, String)] = org.apache.spark.streaming.kafka.DirectKafkaInputDStream@375c2870

scala>

scala> val lines = messages.map(_._2)

lines: org.apache.spark.streaming.dstream.DStream[String] = org.apache.spark.streaming.dstream.MappedDStream@1dda179e

scala> val words = lines.flatMap(_.split(" "))

words: org.apache.spark.streaming.dstream.DStream[String] = org.apache.spark.streaming.dstream.FlatMappedDStream@996294c

scala> val wordCounts = words.map(x => (x, 1)).reduceByKey(_ + _)

wordCounts: org.apache.spark.streaming.dstream.DStream[(String, Int)] = org.apache.spark.streaming.dstream.ShuffledDStream@19cd9e6a

scala> wordCounts.print()

scala> ssc.start()

scala>ssc.awaitTermination()

3、测试

在kafka生产者shell端输入：

hadoop oracle mysql mysql mysql

这是我们在kafka消费者端可以看到如下输出：

hadoop oracle mysql mysql mysql

同时在spark streaming端也可以看到如下输出：

-------------------------------------------

Time: 1500021590000 ms

-------------------------------------------

(mysql,3)

(oracle,1)

(hadoop,1)

至此，spark streaming集成kafka两种方式演示OK。但是通过上述案例我们可以发现，目前的spark streaming仅仅对每次的输入值进行一次计算，而企业应用中，可能更需要将多次的输入值进行累加，那么该怎么实现呢？看下面案例？

四）使用UpdataStateByKey实现spark streaming多次输入值的累加操作

1、创建文件udsb.scala文件，输入如下内容：

$ cat udsb.scala

import kafka.serializer.StringDecoder

import org.apache.spark._

import org.apache.spark.streaming._

import org.apache.spark.streaming.StreamingContext._

import org.apache.spark.streaming.kafka._

val ssc = new StreamingContext(sc, Seconds(5))

ssc.checkpoint(".")

val kafkaParams = Map[String, String]("metadata.broker.list" -> "chavin.king:9092")

val topicsSet = Set("test")

val updateFunc = (values: Seq[Int], state: Option[Int]) => {

val currentCount = values.sum

val previousCount = state.getOrElse(0)

Some(currentCount + previousCount)

}

// read data

val messages = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topicsSet)

val lines = messages.map(_._2)

val words = lines.flatMap(_.split(" "))

val wordDstream = words.map(x => (x, 1))

val stateDstream = wordDstream.updateStateByKey[Int](updateFunc)

stateDstream.print()

ssc.start()

ssc.awaitTermination()

2、spark-shell local模式启动，并运行步骤1程序

bin/spark-shell --master local[2] --jars \

/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/spark-streaming-kafka_2.10-1.3.0.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka_2.10-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/kafka-clients-0.8.2.1.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/zkclient-0.3.jar,/opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/externalLibs/metrics-core-2.2.0.jar

scala> :load /opt/cdh-5.3.6/spark-1.3.0-bin-2.5.0-cdh5.3.6/udsb.scala

3、测试

在kafka生产者shell端输入：

3.1）第一次输入：hadoop oracle mysql

Spark streaming端可以看到如下输出：

-------------------------------------------

Time: 1500023985000 ms

-------------------------------------------

(mysql,1)

(oracle,1)

(hadoop,1)

3.2）第二次输入：hadoop oracle mysql

Spark streaming端可以看到如下输出：

-------------------------------------------

Time: 1500023985000 ms

-------------------------------------------

(mysql,2)

(oracle,2)

(hadoop,2)

3.3）第三次输入：hadoop oracle mysql

Spark streaming端可以看到如下输出：

-------------------------------------------

Time: 1500023985000 ms

-------------------------------------------

(mysql,3)

(oracle,3)

(hadoop,3)