1 目的

Spark Streaming整合Flume。参考官方整合文档（http://spark.apache.org/docs/2.2.0/streaming-flume-integration.html）

2 整合方式一：基于推

2.1 基本要求

• flume和spark一个work节点要在同一台机器上，flume会在本机器上通过配置的端口推送数据
• streaming应用必须先启动，receive必须要先监听推送数据的端口后，flume才能推送数据
• 添加如下依赖

groupId = org.apache.spark
 artifactId = spark-streaming-flume_2.11
 version = 2.2.0

2.2 配置Flume

我们知道flume 的使用就是如何配置它的配置文件，使用本地的netcat source来模拟数据，本次配置如下：

# Name the components on this agent
a1.sources = r1
a1.sinks = k1
a1.channels = c1

# Describe/configure the source
a1.sources.r1.type = netcat
a1.sources.r1.bind = hadoop
a1.sources.r1.port = 5900

# Describe the sink
a1.sinks.k1.type = avro
a1.sinks.k1.hostname = hadoop
a1.sinks.k1.port = 5901
#a1.sinks.k1.type = logger

# Use a channel which buffers events in memory
a1.channels.c1.type = memory
#a1.channels.c1.capacity = 1000
#a1.channels.c1.transactionCapacity = 100

# Bind the source and sink to the channel
a1.sources.r1.channels = c1
a1.sinks.k1.channel = c1

2.3 在服务器上运行

思路如下：

• 用maven打包工程
• 使用saprk-submit提交
• 开启flume
• 发送模拟数据
• 验证

验证代码如下：功能简单的做一个单词统计：

package flume_streaming

import org.apache.spark.SparkConf
import org.apache.spark.streaming.flume.FlumeUtils
import org.apache.spark.streaming.{Durations, StreamingContext}

/**
 * @Author: SmallWild
 * @Date: 2019/11/2 9:42
 * @Desc: 基于flumePushWordCount
 */
object flumePushWordCount {
  def main(args: Array[String]): Unit = {
    if (args.length != 2) {
      System.err.println("错误参数，用法：flumePushWordCount <hostname> <port>")
      System.exit(1)
    }
    //传入参数
    val Array(hostname, port) = args
    //一定不能使用local[1]
    val sparkConf = new SparkConf() //.setMaster("local[2]").setAppName("kafkaDirectWordCount")
    val ssc = new StreamingContext(sparkConf, Durations.seconds(5))
    //设置日志级别
    ssc.sparkContext.setLogLevel("WARN")
    //TODO 简单的进行单词统计
    val flumeStream = FlumeUtils.createStream(ssc, hostname, port.toInt)
    flumeStream.map(x => new String(x.event.getBody.array()).trim)
      .flatMap(_.split(" ")).map((_, 1)).reduceByKey(_ + _).print()
    ssc.start()
    ssc.awaitTermination()
  }
}

验证具体步骤如下：

1）打包工程
 mvn clean package -DskipTest
 2)spark-submit提交（这里使用local模式）
 ./spark-submit --class flume_streaming.flumePushWordCount /
 --master local[2] /
 --packages org.apache.spark:spark-streaming-flume_2.11:2.2.0 /
 /smallwild/app/SparkStreaming-1.0.jar hadoop 5901
 3）开启flume
 flume-ng agent --name simple-agent --conf $FLUME_HOME/conf --conf-file $FLUME_HOME/conf -Dflume.root.logger=INFO,console
 4）发送模式数据
 这里使用本地5900端口发送数据
 telnet hadoop 5900
 5）验证
 查看streaming应用程序是否能出现对应的单词计数字样

验证结果：能正确统计从端口发送过来的某一批次的单词的数量

3 整合方式二：基于拉（常用）

这种方式和上面基本一致

3.1 注意事项

• 先启动flume
• 使用自定义的sink，streaming主动去拉取数据，数据会先存放在缓冲区
• 事务保障机制，副本机制和数据被接收（Transactions succeed only after data is received and replicated by Spark Streaming.）
• 高容错保证
• 添加如下依赖

  groupId = org.apache.spark
   artifactId = spark-streaming-flume-sink_2.11
   version = 2.2.0
  
   groupId = org.scala-lang
   artifactId = scala-library
   version = 2.11.8
  
   groupId = org.apache.commons
   artifactId = commons-lang3
   version = 3.5

3.2 配置Flume

和前面差别在配置sink，需要使用自定义的sink

# Name the components on this agent
a1.sources = r1
a1.sinks = k1
a1.channels = c1

# Describe/configure the source
a1.sources.r1.type = netcat
a1.sources.r1.bind = hadoop
a1.sources.r1.port = 5900

# Describe the sink
a1.sinks.k1.type = org.apache.spark.streaming.flume.sink.SparkSink
a1.sinks.k1.hostname = hadoop
a1.sinks.k1.port = 5901
#a1.sinks.k1.type = logger

# Use a channel which buffers events in memory
a1.channels.c1.type = memory
#a1.channels.c1.capacity = 1000
#a1.channels.c1.transactionCapacity = 100

# Bind the source and sink to the channel
a1.sources.r1.channels = c1
a1.sinks.k1.channel = c1

3.3 在服务上运行

业务逻辑大致和前面一样，这里使用下面的类

import org.apache.spark.streaming.flume._

 val flumeStream = FlumeUtils.createPollingStream(streamingContext, [sink machine hostname], [sink port])

3.4 提交验证

思路如下：

• 用maven打包工程
• 开启flume
• 使用saprk-submit提交
• 发送模拟数据
• 验证

和前面基本一致

4 总结

整理两种整合flume的实践。