基于Spark通用计算平台,可以很好地扩展各种计算类型的应用,尤其是Spark提供了内建的计算库支持,像Spark Streaming、Spark SQL、MLlib、GraphX,这些内建库都提供了高级抽象,可以用非常简洁的代码实现复杂的计算逻辑、这也得益于Scala编程语言的简洁性。这里,我们基于1.3.0版本的Spark搭建了计算平台,实现基于Spark Streaming的实时计算。
我们的应用场景是分析用户使用手机App的行为,描述如下所示:
- 手机客户端会收集用户的行为事件(我们以点击事件为例),将数据发送到数据服务器,我们假设这里直接进入到Kafka消息队列
- 后端的实时服务会从Kafka消费数据,将数据读出来并进行实时分析,这里选择Spark Streaming,因为Spark Streaming提供了与Kafka整合的内置支持
- 经过Spark Streaming实时计算程序分析,将结果写入Redis,可以实时获取用户的行为数据,并可以导出进行离线综合统计分析
Spark Streaming介绍
Spark Streaming提供了一个叫做DStream(Discretized Stream)的高级抽象,DStream表示一个持续不断输入的数据流,可以基于Kafka、TCP Socket、Flume等输入数据流创建。在内部,一个DStream实际上是由一个RDD序列组成的。Sparking Streaming是基于Spark平台的,也就继承了Spark平台的各种特性,如容错(Fault-tolerant)、可扩展(Scalable)、高吞吐(High-throughput)等。
在Spark Streaming中,每个DStream包含了一个时间间隔之内的数据项的集合,我们可以理解为指定时间间隔之内的一个batch,每一个batch就构成一个RDD数据集,所以DStream就是一个个batch的有序序列,时间是连续的,按照时间间隔将数据流分割成一个个离散的RDD数据集,如图所示(来自官网):
我们都知道,Spark支持两种类型操作:Transformations和Actions。Transformation从一个已知的RDD数据集经过转换得到一个新的RDD数据集,这些Transformation操作包括map、filter、flatMap、union、join等,而且Transformation具有lazy的特性,调用这些操作并没有立刻执行对已知RDD数据集的计算操作,而是在调用了另一类型的Action操作才会真正地执行。Action执行,会真正地对RDD数据集进行操作,返回一个计算结果给Driver程序,或者没有返回结果,如将计算结果数据进行持久化,Action操作包括reduceByKey、count、foreach、collect等。关于Transformations和Actions更详细内容,可以查看官网文档。
同样、Spark Streaming提供了类似Spark的两种操作类型,分别为Transformations和Output操作,它们的操作对象是DStream,作用也和Spark类似:Transformation从一个已知的DStream经过转换得到一个新的DStream,而且Spark Streaming还额外增加了一类针对Window的操作,当然它也是Transformation,但是可以更灵活地控制DStream的大小(时间间隔大小、数据元素个数),例如window(windowLength, slideInterval)、countByWindow(windowLength, slideInterval)、reduceByWindow(func, windowLength, slideInterval)等。Spark Streaming的Output操作允许我们将DStream数据输出到一个外部的存储系统,如数据库或文件系统等,执行Output操作类似执行Spark的Action操作,使得该操作之前lazy的Transformation操作序列真正地执行。
Kafka+Spark Streaming+Redis编程实践
下面,我们根据上面提到的应用场景,来编程实现这个实时计算应用。首先,写了一个Kafka Producer模拟程序,用来模拟向Kafka实时写入用户行为的事件数据,数据是JSON格式,示例如下:
1 | {"uid":"068b746ed4620d25e26055a9f804385f","event_time":"1430204612405","os_type":"Android","click_count":6} |
一个事件包含4个字段:
- uid:用户编号
- event_time:事件发生时间戳
- os_type:手机App操作系统类型
- click_count:点击次数
下面是我们实现的代码,如下所示:
package org.shirdrn.spark.streaming.utils
import java.util.Properties
import scala.util.Properties
import org.codehaus.jettison.json.JSONObject
import kafka.javaapi.producer.Producer
import kafka.producer.KeyedMessage
import kafka.producer.KeyedMessage
import kafka.producer.ProducerConfig
import scala.util.Random
object KafkaEventProducer {
private val users = Array(
"4A4D769EB9679C054DE81B973ED5D768", "8dfeb5aaafc027d89349ac9a20b3930f",
"011BBF43B89BFBF266C865DF0397AA71", "f2a8474bf7bd94f0aabbd4cdd2c06dcf",
"068b746ed4620d25e26055a9f804385f", "97edfc08311c70143401745a03a50706",
"d7f141563005d1b5d0d3dd30138f3f62", "c8ee90aade1671a21336c721512b817a",
"6b67c8c700427dee7552f81f3228c927", "a95f22eabc4fd4b580c011a3161a9d9d")
private val random = new Random()
private var pointer = -1
def getUserID() : String = {
pointer = pointer + 1
if(pointer >= users.length) {
pointer = 0
users(pointer)
} else {
users(pointer)
}
}
def click() : Double = {
random.nextInt(10)
}
// bin/kafka-topics.sh --zookeeper zk1:2181,zk2:2181,zk3:2181/kafka --create --topic user_events --replication-factor 2 --partitions 2
// bin/kafka-topics.sh --zookeeper zk1:2181,zk2:2181,zk3:2181/kafka --list
// bin/kafka-topics.sh --zookeeper zk1:2181,zk2:2181,zk3:2181/kafka --describe user_events
// bin/kafka-console-consumer.sh --zookeeper zk1:2181,zk2:2181,zk3:22181/kafka --topic test_json_basis_event --from-beginning
def main(args: Array[String]): Unit = {
val topic = "user_events"
val brokers = "10.10.4.126:9092,10.10.4.127:9092"
val props = new Properties()
props.put("metadata.broker.list", brokers)
props.put("serializer.class", "kafka.serializer.StringEncoder")
val kafkaConfig = new ProducerConfig(props)
val producer = new Producer[String, String](kafkaConfig)
while(true) {
// prepare event data
val event = new JSONObject()
event
.put("uid", getUserID)
.put("event_time", System.currentTimeMillis.toString)
.put("os_type", "Android")
.put("click_count", click)
// produce event message
producer.send(new KeyedMessage[String, String](topic, event.toString))
println("Message sent: " + event)
Thread.sleep(200)
}
}
}
通过控制上面程序最后一行的时间间隔来控制模拟写入速度。下面我们来讨论实现实时统计每个用户的点击次数,它是按照用户分组进行累加次数,逻辑比较简单,关键是在实现过程中要注意一些问题,如对象序列化等。先看实现代码,稍后我们再详细讨论,代码实现如下所示:
object UserClickCountAnalytics {
def main(args: Array[String]): Unit = {
var masterUrl = "local[1]"
if (args.length > 0) {
masterUrl = args(0)
}
// Create a StreamingContext with the given master URL
val conf = new SparkConf().setMaster(masterUrl).setAppName("UserClickCountStat")
val ssc = new StreamingContext(conf, Seconds(5))
// Kafka configurations
val topics = Set("user_events")
val brokers = "10.10.4.126:9092,10.10.4.127:9092"
val kafkaParams = Map[String, String](
"metadata.broker.list" -> brokers, "serializer.class" -> "kafka.serializer.StringEncoder")
val dbIndex = 1
val clickHashKey = "app::users::click"
// Create a direct stream
val kafkaStream = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topics)
val events = kafkaStream.flatMap(line => {
val data = JSONObject.fromObject(line._2)
Some(data)
})
// Compute user click times
val userClicks = events.map(x => (x.getString("uid"), x.getInt("click_count"))).reduceByKey(_ + _)
userClicks.foreachRDD(rdd => {
rdd.foreachPartition(partitionOfRecords => {
partitionOfRecords.foreach(pair => {
val uid = pair._1
val clickCount = pair._2
val jedis = RedisClient.pool.getResource
jedis.select(dbIndex)
jedis.hincrBy(clickHashKey, uid, clickCount)
RedisClient.pool.returnResource(jedis)
})
})
})
ssc.start()
ssc.awaitTermination()
}
}
上面代码使用了Jedis客户端来操作Redis,将分组计数结果数据累加写入Redis存储,如果其他系统需要实时获取该数据,直接从Redis实时读取即可。RedisClient实现代码如下所示:
object RedisClient extends Serializable {
val redisHost = "10.10.4.130"
val redisPort = 6379
val redisTimeout = 30000
lazy val pool = new JedisPool(new GenericObjectPoolConfig(), redisHost, redisPort, redisTimeout)
lazy val hook = new Thread {
override def run = {
println("Execute hook thread: " + this)
pool.destroy()
}
}
sys.addShutdownHook(hook.run)
}
上面代码我们分别在local[K]和Spark Standalone集群模式下运行通过。
如果我们是在开发环境进行调试的时候,也就是使用local[K]部署模式,在本地启动K个Worker线程来计算,这K个Worker在同一个JVM实例里,上面的代码默认情况是,如果没有传参数则是local[K]模式,所以如果使用这种方式在创建Redis连接池或连接的时候,可能非常容易调试通过,但是在使用Spark Standalone、YARN Client(YARN Cluster)或Mesos集群部署模式的时候,就会报错,主要是由于在处理Redis连接池或连接的时候出错了。我们可以看一下Spark架构,如图所示(来自官网):
无论是在本地模式、Standalone模式,还是在Mesos或YARN模式下,整个Spark集群的结构都可以用上图抽象表示,只是各个组件的运行环境不同,导致组件可能是分布式的,或本地的,或单个JVM实例的。如在本地模式,则上图表现为在同一节点上的单个进程之内的多个组件;而在YARN Client模式下,Driver程序是在YARN集群之外的一个节点上提交Spark Application,其他的组件都运行在YARN集群管理的节点上。
在Spark集群环境部署Application后,在进行计算的时候会将作用于RDD数据集上的函数(Functions)发送到集群中Worker上的Executor上(在Spark Streaming中是作用于DStream的操作),那么这些函数操作所作用的对象(Elements)必须是可序列化的,通过Scala也可以使用lazy引用来解决,否则这些对象(Elements)在跨节点序列化传输后,无法正确地执行反序列化重构成实际可用的对象。上面代码我们使用lazy引用(Lazy Reference)来实现的,代码如下所示:
// lazy pool reference
lazy val pool = new JedisPool(new GenericObjectPoolConfig(), redisHost, redisPort, redisTimeout)
...
partitionOfRecords.foreach(pair => {
val uid = pair._1
val clickCount = pair._2
val jedis = RedisClient.pool.getResource
jedis.select(dbIndex)
jedis.hincrBy(clickHashKey, uid, clickCount)
RedisClient.pool.returnResource(jedis)
})
另一种方式,我们将代码修改为,把对Redis连接的管理放在操作DStream的Output操作范围之内,因为我们知道它是在特定的Executor中进行初始化的,使用一个单例的对象来管理,如下所示:
package org.shirdrn.spark.streaming
import org.apache.commons.pool2.impl.GenericObjectPoolConfig
import org.apache.spark.SparkConf
import org.apache.spark.streaming.Seconds
import org.apache.spark.streaming.StreamingContext
import org.apache.spark.streaming.dstream.DStream.toPairDStreamFunctions
import org.apache.spark.streaming.kafka.KafkaUtils
import kafka.serializer.StringDecoder
import net.sf.json.JSONObject
import redis.clients.jedis.JedisPool
object UserClickCountAnalytics {
def main(args: Array[String]): Unit = {
var masterUrl = "local[1]"
if (args.length > 0) {
masterUrl = args(0)
}
// Create a StreamingContext with the given master URL
val conf = new SparkConf().setMaster(masterUrl).setAppName("UserClickCountStat")
val ssc = new StreamingContext(conf, Seconds(5))
// Kafka configurations
val topics = Set("user_events")
val brokers = "10.10.4.126:9092,10.10.4.127:9092"
val kafkaParams = Map[String, String](
"metadata.broker.list" -> brokers, "serializer.class" -> "kafka.serializer.StringEncoder")
val dbIndex = 1
val clickHashKey = "app::users::click"
// Create a direct stream
val kafkaStream = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topics)
val events = kafkaStream.flatMap(line => {
val data = JSONObject.fromObject(line._2)
Some(data)
})
// Compute user click times
val userClicks = events.map(x => (x.getString("uid"), x.getInt("click_count"))).reduceByKey(_ + _)
userClicks.foreachRDD(rdd => {
rdd.foreachPartition(partitionOfRecords => {
partitionOfRecords.foreach(pair => {
/**
* Internal Redis client for managing Redis connection {@link Jedis} based on {@link RedisPool}
*/
object InternalRedisClient extends Serializable {
@transient private var pool: JedisPool = null
def makePool(redisHost: String, redisPort: Int, redisTimeout: Int,
maxTotal: Int, maxIdle: Int, minIdle: Int): Unit = {
makePool(redisHost, redisPort, redisTimeout, maxTotal, maxIdle, minIdle, true, false, 10000)
}
def makePool(redisHost: String, redisPort: Int, redisTimeout: Int,
maxTotal: Int, maxIdle: Int, minIdle: Int, testOnBorrow: Boolean,
testOnReturn: Boolean, maxWaitMillis: Long): Unit = {
if(pool == null) {
val poolConfig = new GenericObjectPoolConfig()
poolConfig.setMaxTotal(maxTotal)
poolConfig.setMaxIdle(maxIdle)
poolConfig.setMinIdle(minIdle)
poolConfig.setTestOnBorrow(testOnBorrow)
poolConfig.setTestOnReturn(testOnReturn)
poolConfig.setMaxWaitMillis(maxWaitMillis)
pool = new JedisPool(poolConfig, redisHost, redisPort, redisTimeout)
val hook = new Thread{
override def run = pool.destroy()
}
sys.addShutdownHook(hook.run)
}
}
def getPool: JedisPool = {
assert(pool != null)
pool
}
}
// Redis configurations
val maxTotal = 10
val maxIdle = 10
val minIdle = 1
val redisHost = "10.10.4.130"
val redisPort = 6379
val redisTimeout = 30000
val dbIndex = 1
InternalRedisClient.makePool(redisHost, redisPort, redisTimeout, maxTotal, maxIdle, minIdle)
val uid = pair._1
val clickCount = pair._2
val jedis =InternalRedisClient.getPool.getResource
jedis.select(dbIndex)
jedis.hincrBy(clickHashKey, uid, clickCount)
InternalRedisClient.getPool.returnResource(jedis)
})
})
})
ssc.start()
ssc.awaitTermination()
}
}
上面代码实现,得益于Scala语言的特性,可以在代码中任何位置进行class或object的定义,我们将用来管理Redis连接的代码放在了特定操作的内部,就避免了瞬态(Transient)对象跨节点序列化的问题。这样做还要求我们能够了解Spark内部是如何操作RDD数据集的,更多可以参考RDD或Spark相关文档。
在集群上,以Standalone模式运行,执行如下命令:
1 | cd /usr/ local /spark |
2 | ./bin/spark-submit --class org.shirdrn.spark.streaming.UserClickCountAnalytics --master spark://hadoop1:7077 --executor-memory 1G --total-executor-cores 2 ~/spark-0.0.SNAPSHOT.jar spark://hadoop1:7077 |
可以查看集群中各个Worker节点执行计算任务的状态,也可以非常方便地通过Web页面查看。
下面,看一下我们存储到Redis中的计算结果,如下所示:
01 | 127.0.0.1:6379[1]> HGETALL app::users::click |
02 | 1) "4A4D769EB9679C054DE81B973ED5D768" |
03 | 2) "7037" |
04 | 3) "8dfeb5aaafc027d89349ac9a20b3930f" |
05 | 4) "6992" |
06 | 5) "011BBF43B89BFBF266C865DF0397AA71" |
07 | 6) "7021" |
08 | 7) "97edfc08311c70143401745a03a50706" |
09 | 8) "6874" |
10 | 9) "d7f141563005d1b5d0d3dd30138f3f62" |
11 | 10) "7057" |
12 | 11) "a95f22eabc4fd4b580c011a3161a9d9d" |
13 | 12) "7092" |
14 | 13) "6b67c8c700427dee7552f81f3228c927" |
15 | 14) "7266" |
16 | 15) "f2a8474bf7bd94f0aabbd4cdd2c06dcf" |
17 | 16) "7188" |
18 | 17) "c8ee90aade1671a21336c721512b817a" |
19 | 18) "6950" |
20 | 19) "068b746ed4620d25e26055a9f804385f" |
有关更多关于Spark Streaming的详细内容,可以参考官方文档。
附录
这里,附上前面开发的应用所对应的依赖,以及打包Spark Streaming应用程序的Maven配置,以供参考。如果使用maven-shade-plugin插件,配置有问题的话,打包后在Spark集群上提交Application时候可能会报错Invalid signature file digest for Manifest main attributes。参考的Maven配置,如下所示:
<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<groupId>org.shirdrn.spark</groupId>
<artifactId>spark</artifactId>
<version>0.0.1-SNAPSHOT</version>
<dependencies>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-core_2.10</artifactId>
<version>1.3.0</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-streaming_2.10</artifactId>
<version>1.3.0</version>
</dependency>
<dependency>
<groupId>net.sf.json-lib</groupId>
<artifactId>json-lib</artifactId>
<version>2.3</version>
</dependency>
<dependency>
<groupId>org.apache.spark</groupId>
<artifactId>spark-streaming-kafka_2.10</artifactId>
<version>1.3.0</version>
</dependency>
<dependency>
<groupId>redis.clients</groupId>
<artifactId>jedis</artifactId>
<version>2.5.2</version>
</dependency>
<dependency>
<groupId>org.apache.commons</groupId>
<artifactId>commons-pool2</artifactId>
<version>2.2</version>
</dependency>
</dependencies>
<build>
<sourceDirectory>${basedir}/src/main/scala</sourceDirectory>
<testSourceDirectory>${basedir}/src/test/scala</testSourceDirectory>
<resources>
<resource>
<directory>${basedir}/src/main/resources</directory>
</resource>
</resources>
<testResources>
<testResource>
<directory>${basedir}/src/test/resources</directory>
</testResource>
</testResources>
<plugins>
<plugin>
<artifactId>maven-compiler-plugin</artifactId>
<version>3.1</version>
<configuration>
<source>1.6</source>
<target>1.6</target>
</configuration>
</plugin>
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-shade-plugin</artifactId>
<version>2.2</version>
<configuration>
<createDependencyReducedPom>true</createDependencyReducedPom>
</configuration>
<executions>
<execution>
<phase>package</phase>
<goals>
<goal>shade</goal>
</goals>
<configuration>
<artifactSet>
<includes>
<include>*:*</include>
</includes>
</artifactSet>
<filters>
<filter>
<artifact>*:*</artifact>
<excludes>
<exclude>META-INF/*.SF</exclude>
<exclude>META-INF/*.DSA</exclude>
<exclude>META-INF/*.RSA</exclude>
</excludes>
</filter>
</filters>
<transformers>
<transformer
implementation="org.apache.maven.plugins.shade.resource.ServicesResourceTransformer" />
<transformer
implementation="org.apache.maven.plugins.shade.resource.AppendingTransformer">
<resource>reference.conf</resource>
</transformer>
<transformer
implementation="org.apache.maven.plugins.shade.resource.DontIncludeResourceTransformer">
<resource>log4j.properties</resource>
</transformer>
</transformers>
</configuration>
</execution>
</executions>
</plugin>
</plugins>
</build>
</project>
参考链接
- http://spark.apache.org/docs/1.3.0/index.html
- http://spark.apache.org/docs/1.3.0/cluster-overview.html
- http://spark.apache.org/docs/1.3.0/job-scheduling.html
- http://spark.apache.org/docs/1.3.0/streaming-programming-guide.html
- http://stackoverflow.com/questions/28006517/redis-on-sparktask-not-serializable