spark-streaming-[9]-SparkStreaming消费Kafka-Direct Approach

时间:2021-06-14 20:48:41


spark-streaming-[8]-Spark Streaming + Kafka Integration Guide0.8.2.1学习笔记
中已知:

There are two approaches to this - the old approach using Receivers and Kafka’s high-level API, and a new approach (introduced in Spark 1.3) without using Receivers. 

Spark-Streaming获取kafka数据的两种方式-Receiver与Direct的方式,可以从代码中简单理解成Receiver方式是通过zookeeper来连接kafka队列,Direct方式是直接连接到kafka的节点上获取数据了。
一、基于Receiver的方式
这种方式使用Receiver来获取数据。Receiver是使用Kafka的高层次Consumer API来实现的。receiver从Kafka中获取的数据都是存储在Spark Executor的内存中的,然后Spark Streaming启动的job会去处理那些数据。
然而,在默认的配置下,这种方式可能会因为底层的失败而丢失数据。如果要启用高可靠机制,让数据零丢失,就必须启用Spark Streaming的预写日志机制(Write Ahead Log,WAL)。该机制会同步地将接收到的Kafka数据写入分布式文件系统(比如HDFS)上的预写日志中。所以,即使底层节点出现了失败,也可以使用预写日志中的数据进行恢复。
注意
1、Kafka中的topic的partition,与Spark中的RDD的partition是没有关系的。所以,在KafkaUtils.createStream()中,提高partition的数量,只会增加一个Receiver中,读取partition的线程的数量。不会增加Spark处理数据的并行度。
2、可以创建多个Kafka输入DStream,使用不同的consumer group和topic,来通过多个receiver并行接收数据。
3、如果基于容错的文件系统,比如HDFS,启用了预写日志机制,接收到的数据都会被复制一份到预写日志中。因此,在KafkaUtils.createStream()中,设置的持久化级别是StorageLevel.MEMORY_AND_DISK_SER。
二、基于Direct的方式
这种新的不基于Receiver的直接方式,是在Spark 1.3中引入的,从而能够确保更加健壮的机制。替代掉使用Receiver来接收数据后,这种方式会周期性地查询Kafka,来获得每个topic+partition的最新的offset,从而定义每个batch的offset的范围。当处理数据的job启动时,就会使用Kafka的简单consumer api来获取Kafka指定offset范围的数据。
这种方式有如下优点:
1、简化并行读取:如果要读取多个partition,不需要创建多个输入DStream然后对它们进行union操作。Spark会创建跟Kafka partition一样多的RDD partition,并且会并行从Kafka中读取数据。所以在Kafka partition和RDD partition之间,有一个一对一的映射关系。
2、高性能:如果要保证零数据丢失,在基于receiver的方式中,需要开启WAL机制。这种方式其实效率低下,因为数据实际上被复制了两份,Kafka自己本身就有高可靠的机制,会对数据复制一份,而这里又会复制一份到WAL中。而基于direct的方式,不依赖Receiver,不需要开启WAL机制,只要Kafka中作了数据的复制,那么就可以通过Kafka的副本进行恢复。
3、一次且仅一次的事务机制:
基于receiver的方式,是使用Kafka的高阶API来在ZooKeeper中保存消费过的offset的。这是消费Kafka数据的传统方式。这种方式配合着WAL机制可以保证数据零丢失的高可靠性,但是却无法保证数据被处理一次且仅一次,可能会处理两次。因为Spark和ZooKeeper之间可能是不同步的。
基于direct的方式,使用kafka的简单api,Spark Streaming自己就负责追踪消费的offset,并保存在checkpoint中。Spark自己一定是同步的,因此可以保证数据是消费一次且仅消费一次

以上参考: Spark-Streaming获取kafka数据的两种方式-Receiver与Direct的方式 感谢

Approach 1: Receiver-based Approach

参见: spark-streaming-[6]-KafkaWordCount和KafkaWordCountProducer(Receiver-based Approach)

Approach 2: Direct Approach (No Receivers)

示例

第一步:kafka中建立topic test2,和producer发送数据给topic test2;
第二步:spark streaming 以Direct 方式消费数据。

第一步具体如下:

具体参见: Kafka-[2]-Documentation-单机QuickStart

Start the server

Kafka uses ZooKeeper so you need to first start a ZooKeeper server if you don't already have one. You can use the convenience script packaged with kafka to get a quick-and-dirty single-node ZooKeeper instance.

> bin/zookeeper-server-start.sh config/zookeeper.properties
 

Now start the Kafka server:

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

Create a topic

Let's create a topic named "test" with a single partition and only one replica:

> bin/kafka-topics.sh --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1 --topic test2

We can now see that topic if we run the list topic command:

> bin/kafka-topics.sh --list --zookeeper localhost:2181
test2

Send some messages

Kafka comes with a command line client that will take input from a file or from standard input and send it out as messages to the Kafka cluster. By default, each line will be sent as a separate message.

Run the producer and then type a few messages into the console to send to the server.

> bin/kafka-console-producer.sh --broker-list localhost:9092 --topic test2
将以下数据粘到console中,粘贴3次,即模拟有3次数据写入Kafka中。(“WWW”出现33次
[2017-05-07 13:11:24,645] WWW Server environment:java.io.tmpdir=sw8vdr0000gn/T/ (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:java.compiler=<NA> (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:os.name=Mac OS X (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:os.arch=x86_64 (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:os.version=10.10.5 (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:user.name=hjw (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:user.home=/Users/hjw (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,645] WWW Server environment:user.dir=ka/kafka_2.11-0.10.2.1 (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,653] WWW tickTime set to 3000 (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,653] WWW minSessionTimeout set to -1 (org.apache.zookeeper.server.ZooKeeperServer)
[2017-05-07 13:11:24,653] WWW maxSessionTimeout set to -1 (org.apache.zookeeper.server.ZooKeeperServer)

第二步运行sparkstreaming DirectKafkaWordCount

参数:localhost:9092  test2

其中Kafka的offset为从头开始读取数据

val kafkaParams = Map[String, String]("metadata.broker.list" -> brokers,"auto.offset.reset"->"smallest")

输出:

-------------------------------------------
Time: 1494141102000 ms
-------------------------------------------
(-1,6)
(environment:os.name=Mac,3)
(X,3)
(environment:java.compiler=<NA>,3)
(OS,3)
([2017-05-07,33)
(3000,3)
((org.apache.zookeeper.server.ZooKeeperServer),33)
(13:11:24,653],9)
...
17/05/07 15:11:42 INFO VerifiableProperties: Verifying properties
17/05/07 15:11:42 INFO VerifiableProperties: Property auto.offset.reset is overridden to smallest
17/05/07 15:11:42 INFO VerifiableProperties: Property group.id is overridden to 
17/05/07 15:11:42 INFO VerifiableProperties: Property zookeeper.connect is overridden to 
-------------------------------------------
Time: 1494141102000 ms
-------------------------------------------
(WWW,33)

再次粘贴测试数据一次进入producer:

输出:

-------------------------------------------
Time: 1494142328000 ms
-------------------------------------------
(-1,2)
(environment:os.name=Mac,1)
(X,1)
(environment:java.compiler=<NA>,1)
(OS,1)
([2017-05-07,11)
(3000,1)
((org.apache.zookeeper.server.ZooKeeperServer),11)
(13:11:24,653],3)
...
17/05/07 15:32:08 INFO VerifiableProperties: Verifying properties
17/05/07 15:32:08 INFO VerifiableProperties: Property auto.offset.reset is overridden to smallest
17/05/07 15:32:08 INFO VerifiableProperties: Property group.id is overridden to 
17/05/07 15:32:08 INFO VerifiableProperties: Property zookeeper.connect is overridden to 
-------------------------------------------
Time: 1494142328000 ms
-------------------------------------------
(WWW,44)

具体代码

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *    http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

// scalastyle:off println
/**
  * Created by hjw on 17/5/7.
  */

package com.dt.spark.main.Streaming.Kafka
package com.dt.spark.main.Streaming.Kafka.org.apache.spark.examples.streaming

import kafka.serializer.StringDecoder
import org.apache.log4j.{Level, Logger}
import org.apache.spark.{HashPartitioner, SparkConf}
import org.apache.spark.streaming._
import org.apache.spark.streaming.kafka._

/**
  * Consumes messages from one or more topics in Kafka and does wordcount.
  * Usage: DirectKafkaWordCount <brokers> <topics>
  *   <brokers> is a list of one or more Kafka brokers
  *   <topics> is a list of one or more kafka topics to consume from
  *
  * Example:
  *    $ bin/run-example streaming.DirectKafkaWordCount broker1-host:port,broker2-host:port \
  *    topic1,topic2
  */
object DirectKafkaWordCount {

  ///函数常量定义,返回类型是Some(Int),表示的含义是最新状态
  ///函数的功能是将当前时间间隔内产生的Key的value集合,加到上一个状态中,得到最新状态
  val updateFunc = (values: Seq[Int], state: Option[Int]) => {
    val currentCount = values.sum
    val previousCount = state.getOrElse(0)
    Some(currentCount + previousCount)
  }

  ///入参是三元组遍历器,三个元组分别表示Key、当前时间间隔内产生的对应于Key的Value集合、上一个时间点的状态
  ///newUpdateFunc的返回值要求是iterator[(String,Int)]类型的
  val newUpdateFunc = (iterator: Iterator[(String, Seq[Int], Option[Int])]) => {
    ///对每个Key调用updateFunc函数(入参是当前时间间隔内产生的对应于Key的Value集合、上一个时间点的状态)得到最新状态
    ///然后将最新状态映射为Key和最新状态
    iterator.flatMap(t => updateFunc(t._2, t._3).map(s => (t._1, s)))
  }

  def main(args: Array[String]) {
    Logger.getLogger("org").setLevel(Level.ERROR)
    if (args.length < 2) {
      System.err.println(s"""
                            |Usage: DirectKafkaWordCount <brokers> <topics>
                            |  <brokers> is a list of one or more Kafka brokers
                            |  <topics> is a list of one or more kafka topics to consume from
                            |
        """.stripMargin)
      System.exit(1)
    }

    val Array(brokers, topics) = args

    // Create context with 2 second batch interval
    val sparkConf = new SparkConf().setAppName("DirectKafkaWordCount").setMaster("local[4]")
    val ssc = new StreamingContext(sparkConf, Seconds(2))

    // Create direct kafka stream with brokers and topics
    val topicsSet = topics.split(",").toSet
    val kafkaParams = Map[String, String]("metadata.broker.list" -> brokers,"auto.offset.reset"->"smallest")
    val messages = KafkaUtils.createDirectStream[String, String, StringDecoder, StringDecoder](
      ssc, kafkaParams, topicsSet)

    // Get the lines, split them into words, count the words and print
    val lines = messages.map(_._2)
    val words = lines.flatMap(_.split(" "))
    val wordCounts = words.map(x => (x, 1L)).reduceByKey(_ + _)
    wordCounts.print()

    // updateStateByKey 计算包含"WWW"单次的出现次数
    ssc.checkpoint("DirectKafkaWordCount-checkpoint")
    // Initial RDD input to updateStateByKey
    val initialRDD = ssc.sparkContext.parallelize(List(("WWW", 0)))
    val WWWwords = lines.filter(_.contains("WWW")).map(x=>("WWW",1))
    val stateDstream = WWWwords.updateStateByKey[Int](newUpdateFunc,
      new HashPartitioner(ssc.sparkContext.defaultParallelism), true, initialRDD)
    stateDstream.print()

    // Start the computation
    ssc.start()
    ssc.awaitTermination()
  }
}
// scalastyle:on println