Hadoop-(Flume)

Hadoop-(Flume)

1. Flume 介绍

1.1. 概述

  • Flume是一个分布式、可靠、和高可用的海量日志采集、聚合和传输的系统。
  • Flume可以采集文件，socket数据包、文件、文件夹、kafka等各种形式源数据，又可以将采集到的数据(下沉sink)输出到HDFS、hbase、hive、kafka等众多外部存储系统中
  • 一般的采集需求，通过对flume的简单配置即可实现
  • Flume针对特殊场景也具备良好的自定义扩展能力， 
    因此，flume可以适用于大部分的日常数据采集场景

1.2. 运行机制

  1. Flume分布式系统中最核心的角色是agent，flume采集系统就是由一个个agent所连接起来形成

  2. 每一个agent相当于一个数据传递员，内部有三个组件：

    1. Source：采集组件，用于跟数据源对接，以获取数据

    2. Sink：下沉组件，用于往下一级agent传递数据或者往最终存储系统传递数据

    3. Channel：传输通道组件，用于从source将数据传递到sink

      1566284693606

1.3. Flume 结构图

简单结构

单个 Agent 采集数据

1566284704254

复杂结构

多级 Agent 之间串联

1566091210646

1566284714483

2. Flume 实战案例

案例：使用网络telent命令向一台机器发送一些网络数据，然后通过flume采集网络端口数据

1566284725926

2.1. Flume 的安装部署

Step 1: 下载解压修改配置文件

下载地址：

http://archive.apache.org/dist/flume/1.8.0/apache-flume-1.8.0-bin.tar.gz

Flume的安装非常简单，只需要解压即可，当然，前提是已有hadoop环境

上传安装包到数据源所在节点上

这里我们采用在第三台机器来进行安装

1566091372163

  1. cd /export/softwares/
  2. tar -zxvf apache-flume-1.8.0-bin.tar.gz -C ../servers/
  3. cd /export/servers/apache-flume-1.8.0-bin/conf
  4. cp  flume-env.sh.template flume-env.sh
  5. vim flume-env.sh
  6. export JAVA_HOME=/export/servers/jdk1.8.0_141
Step 2: 开发配置文件

根据数据采集的需求配置采集方案，描述在配置文件中(文件名可任意自定义)

配置我们的网络收集的配置文件 
在flume的conf目录下新建一个配置文件（采集方案）

  1. vim   /export/servers/apache-flume-1.8.0-bin/conf/netcat-logger.conf
  1. # 定义这个agent中各组件的名字
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  6. # 描述和配置source组件：r1
  7. a1.sources.r1.type = netcat
  8. a1.sources.r1.bind = 192.168.174.
  9. a1.sources.r1.port = 44444
  11. # 描述和配置sink组件：k1
  12. a1.sinks.k1.type = logger
  14. # 描述和配置channel组件，此处使用是内存缓存的方式
  15. a1.channels.c1.type = memory
  16. a1.channels.c1.capacity = 1000
  17. a1.channels.c1.transactionCapacity = 100
  19. # 描述和配置source  channel   sink之间的连接关系
  20. a1.sources.r1.channels = c1
  21. a1.sinks.k1.channel = c1

1566092795186

Step 3: 启动配置文件

指定采集方案配置文件，在相应的节点上启动flume agent

先用一个最简单的例子来测试一下程序环境是否正常 
启动agent去采集数据

  1. bin/flume-ng agent -c conf -f conf/netcat-logger.conf -n a1  -Dflume.root.logger=INFO,console
  • -c conf 指定flume自身的配置文件所在目录
  • -f conf/netcat-logger.con 指定我们所描述的采集方案
  • -n a1 指定我们这个agent的名字
Step 4: 安装 Telnet 准备测试

在node02机器上面安装telnet客户端，用于模拟数据的发送

  1. yum -y install telnet
  2. telnet  node03  44444 # 使用telnet模拟数据发送

2.2. 采集案例

2.2.3. 采集目录到 HDFS

1566284746540

需求

某服务器的某特定目录下，会不断产生新的文件，每当有新文件出现，就需要把文件采集到HDFS中去

思路

根据需求，首先定义以下3大要素

  1. 数据源组件，即source ——监控文件目录 : spooldir
    1. 监视一个目录，只要目录中出现新文件，就会采集文件中的内容
    2. 采集完成的文件，会被agent自动添加一个后缀：COMPLETED
    3. 所监视的目录中不允许重复出现相同文件名的文件,否则报错\罢工
  2. 下沉组件，即sink——HDFS文件系统 : hdfs sink
  3. 通道组件，即channel——可用file channel 也可以用内存channel
Step 1: Flume 配置文件
  1. cd  /export/servers/apache-flume-1.8.0-bin/conf
  2. mkdir -p /export/servers/dirfile
  3. vim spooldir.conf
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  7. # Describe/configure the source
  8. ##注意：不能往监控目中重复丢同名文件
  9. a1.sources.r1.type = spooldir
  10. a1.sources.r1.spoolDir = /export/servers/dirfile
  11. a1.sources.r1.fileHeader = true
  14. # Describe the sink
  15. a1.sinks.k1.type = hdfs
  16. a1.sinks.k1.channel = c1
  17. a1.sinks.k1.hdfs.path = hdfs://node01:8020/spooldir/files/%y-%m-%d/%H%M/
  18. a1.sinks.k1.hdfs.filePrefix = events-
  20. #控制文件夹以多少时间滚动     10分钟
  21. a1.sinks.k1.hdfs.round = true
  22. a1.sinks.k1.hdfs.roundValue = 10
  23. a1.sinks.k1.hdfs.roundUnit = minute
  24. #roll控制写入hdfs文件,以何种方式滚动
  25. #时间间隔
  26. a1.sinks.k1.hdfs.rollInterval = 3
  27. #文件大小
  28. a1.sinks.k1.hdfs.rollSize = 20
  29. #even数量
  30. a1.sinks.k1.hdfs.rollCount = 5
  31. a1.sinks.k1.hdfs.batchSize = 1
  32. #不想滚动,设置0
  34. a1.sinks.k1.hdfs.useLocalTimeStamp = true
  36. #生成的文件类型，默认是Sequencefile，可用DataStream，则为普通文本
  38. a1.sinks.k1.hdfs.fileType = DataStream
  40. # Use a channel which buffers events in memory
  41. a1.channels.c1.type = memory
  42. #胶囊容量
  43. a1.channels.c1.capacity = 1000
  44. #一次向sink运输多少个event
  45. a1.channels.c1.transactionCapacity = 100
  47. # Bind the source and sink to the channel
  48. a1.sources.r1.channels = c1
  49. a1.sinks.k1.channel = c1

Channel参数解释

capacity：默认该通道中最大的可以存储的event数量 
trasactionCapacity：每次最大可以从source中拿到或者送到sink中的event数量 
keep-alive：event添加到通道中或者移出的允许时间

Step 2: 启动 Flume
  1. bin/flume-ng agent -c ./conf -f ./conf/spooldir.conf -n a1 -Dflume.root.logger=INFO,console
  2. #命令的精简版
Step 3: 上传文件到指定目录

将不同的文件放到下面目录里面去，注意文件不能重名

  1. cd /export/servers/dirfile

2.2.4. 采集文件到 HDFS

需求

比如业务系统使用log4j生成的日志，日志内容不断增加，需要把追加到日志文件中的数据实时采集到hdfs

分析

根据需求，首先定义以下3大要素

  • 采集源，即source——监控文件内容更新 : exec ‘tail -F file’
  • 下沉目标，即sink——HDFS文件系统 : hdfs sink
  • Source和sink之间的传递通道——channel，可用file channel 也可以用 内存channel
Step 1: 定义 Flume 配置文件
  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim tail-file.conf
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  6. # Describe/configure the source
  7. a1.sources.r1.type = exec
  8. a1.sources.r1.command = tail -F /root/logs/test.log
  9. a1.sources.r1.channels = c1
  11. # Describe the sink
  12. a1.sinks.k1.type = hdfs
  13. a1.sinks.k1.channel = c1
  14. a1.sinks.k1.hdfs.path = /flume/tailout/%y-%m-%d/%H-%M/
  15. a1.sinks.k1.hdfs.filePrefix = events-
  16. a1.sinks.k1.hdfs.round = true
  17. a1.sinks.k1.hdfs.roundValue = 10
  18. a1.sinks.k1.hdfs.roundUnit = minute
  19. a1.sinks.k1.hdfs.rollInterval = 3
  20. a1.sinks.k1.hdfs.rollSize = 20
  21. a1.sinks.k1.hdfs.rollCount = 5
  22. a1.sinks.k1.hdfs.batchSize = 1
  23. a1.sinks.k1.hdfs.useLocalTimeStamp = true
  24. #生成的文件类型，默认是Sequencefile，可用DataStream，则为普通文本
  25. a1.sinks.k1.hdfs.fileType = DataStream
  27. # Use a channel which buffers events in memory
  28. a1.channels.c1.type = memory
  29. a1.channels.c1.capacity = 1000
  30. a1.channels.c1.transactionCapacity = 100
  32. # Bind the source and sink to the channel
  33. a1.sources.r1.channels = c1
  34. a1.sinks.k1.channel = c1
Step 2: 启动 Flume
  1. cd  /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -c conf -f conf/tail-file.conf -n agent1  -Dflume.root.logger=INFO,console
Step 3: 开发 Shell 脚本定时追加文件内容
  1. mkdir -p /export/servers/shells/
  2. cd  /export/servers/shells/
  3. vim tail-file.sh
  1. #!/bin/bash
  2. while true
  3. do
  4. date >> /export/servers/taillogs/access_log;
  5.   sleep 0.5;
  6. done
Step 4: 启动脚本
  1. # 创建文件夹
  2. mkdir -p /export/servers/taillogs
  3. # 启动脚本
  4. sh /export/servers/shells/tail-file.sh

2.2.5. Agent 级联

1566284897069

分析

第一个agent负责收集文件当中的数据，通过网络发送到第二个agent当中去 
第二个agent负责接收第一个agent发送的数据，并将数据保存到hdfs上面去

Step 1: Node02 安装 Flume

将node03机器上面解压后的flume文件夹拷贝到node02机器上面去

  1. cd  /export/servers
  2. scp -r apache-flume-1.8.0-bin/ node02:$PWD
Step 2: Node02 配置 Flume

在node02机器配置我们的flume

  1. cd /export/servers/ apache-flume-1.8.0-bin/conf
  2. vim tail-avro-avro-logger.conf
  1. ##################
  2. # Name the components on this agent
  3. a1.sources = r1
  4. a1.sinks = k1
  5. a1.channels = c1
  6. # Describe/configure the source
  7. a1.sources.r1.type = exec
  8. a1.sources.r1.command = tail -F /export/servers/taillogs/access_log
  9. a1.sources.r1.channels = c1
  10. # Describe the sink
  11. ##sink端的avro是一个数据发送者
  12. a1.sinks = k1
  13. a1.sinks.k1.type = avro
  14. a1.sinks.k1.channel = c1
  15. a1.sinks.k1.hostname = node03
  16. a1.sinks.k1.port = 4141
  17. a1.sinks.k1.batch-size = 10
  18. # Use a channel which buffers events in memory
  19. a1.channels.c1.type = memory
  20. a1.channels.c1.capacity = 1000
  21. a1.channels.c1.transactionCapacity = 100
  22. # Bind the source and sink to the channel
  23. a1.sources.r1.channels = c1
  24. a1.sinks.k1.channel = c1
Step 3: 开发脚本向文件中写入数据

直接将node03下面的脚本和数据拷贝到node02即可，node03机器上执行以下命令

  1. cd  /export/servers
  2. scp -r shells/ taillogs/ node02:$PWD
Step 4: Node03 Flume 配置文件

在node03机器上开发flume的配置文件

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim avro-hdfs.conf
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  5. # Describe/configure the source
  6. ##source中的avro组件是一个接收者服务
  7. a1.sources.r1.type = avro
  8. a1.sources.r1.channels = c1
  9. a1.sources.r1.bind = node03
  10. a1.sources.r1.port = 4141
  11. # Describe the sink
  12. a1.sinks.k1.type = hdfs
  13. a1.sinks.k1.hdfs.path = hdfs://node01:8020/av/%y-%m-%d/%H%M/
  14. a1.sinks.k1.hdfs.filePrefix = events-
  15. a1.sinks.k1.hdfs.round = true
  16. a1.sinks.k1.hdfs.roundValue = 10
  17. a1.sinks.k1.hdfs.roundUnit = minute
  18. a1.sinks.k1.hdfs.rollInterval = 3
  19. a1.sinks.k1.hdfs.rollSize = 20
  20. a1.sinks.k1.hdfs.rollCount = 5
  21. a1.sinks.k1.hdfs.batchSize = 1
  22. a1.sinks.k1.hdfs.useLocalTimeStamp = true
  23. #生成的文件类型，默认是Sequencefile，可用DataStream，则为普通文本
  24. a1.sinks.k1.hdfs.fileType = DataStream
  25. # Use a channel which buffers events in memory
  26. a1.channels.c1.type = memory
  27. a1.channels.c1.capacity = 1000
  28. a1.channels.c1.transactionCapacity = 100
  30. # Bind the source and sink to the channel
  31. a1.sources.r1.channels = c1
  32. a1.sinks.k1.channel = c1
Step 5: 顺序启动

node03机器启动flume进程

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -c conf -f conf/avro-hdfs.conf -n a1  -Dflume.root.logger=INFO,console

node02机器启动flume进程

  1. cd /export/servers/apache-flume-1.8.0-bin/
  2. bin/flume-ng agent -c conf -f conf/tail-avro-avro-logger.conf -n a1  -Dflume.root.logger=INFO,console

node02机器启shell脚本生成文件

  1. cd  /export/servers/shells
  2. sh tail-file.sh

3. flume的高可用方案-failover

在完成单点的Flume NG搭建后，下面我们搭建一个高可用的Flume NG集群，架构图如下所示：

3.1. 角色分配

Flume的Agent和Collector分布如下表所示：

名称HOST角色
Agent1node01Web Server
Collector1node02AgentMstr1
Collector2node03AgentMstr2

图中所示，Agent1数据分别流入到Collector1和Collector2，Flume NG本身提供了Failover机制，可以自动切换和恢复。在上图中，有3个产生日志服务器分布在不同的机房，要把所有的日志都收集到一个集群中存储。下 面我们开发配置Flume NG集群

3.2. Node01 安装和配置

将node03机器上面的flume安装包以及文件生产的两个目录拷贝到node01机器上面去

node03机器执行以下命令

  1. cd /export/servers
  2. scp -r apache-flume-1.8.0-bin/ node01:$PWD
  3. scp -r shells/ taillogs/ node01:$PWD

node01机器配置agent的配置文件

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim agent.conf
  1. #agent1 name
  2. agent1.channels = c1
  3. agent1.sources = r1
  4. agent1.sinks = k1 k2
  5. #
  6. ##set gruop
  7. agent1.sinkgroups = g1
  8. #
  10. agent1.sources.r1.channels = c1
  11. agent1.sources.r1.type = exec
  12. agent1.sources.r1.command = tail -F /export/servers/taillogs/access_log
  13. #
  14. ##set channel
  15. agent1.channels.c1.type = memory
  16. agent1.channels.c1.capacity = 1000
  17. agent1.channels.c1.transactionCapacity = 100
  18. #
  19. ## set sink1
  20. agent1.sinks.k1.channel = c1
  21. agent1.sinks.k1.type = avro
  22. agent1.sinks.k1.hostname = node02
  23. agent1.sinks.k1.port = 52020
  24. #
  25. ## set sink2
  26. agent1.sinks.k2.channel = c1
  27. agent1.sinks.k2.type = avro
  28. agent1.sinks.k2.hostname = node03
  29. agent1.sinks.k2.port = 52020
  30. #
  31. ##set sink group
  32. agent1.sinkgroups.g1.sinks = k1 k2
  33. #
  34. ##set failover
  35. agent1.sinkgroups.g1.processor.type = failover
  36. agent1.sinkgroups.g1.processor.priority.k1 = 10
  37. agent1.sinkgroups.g1.processor.priority.k2 = 1
  38. agent1.sinkgroups.g1.processor.maxpenalty = 10000
  1. #agent1 name
  2. agent1.channels = c1
  3. agent1.sources = r1
  4. agent1.sinks = k1 k2
  6. #set gruop
  7. agent1.sinkgroups = g1
  9. #set channel
  10. agent1.channels.c1.type = memory
  11. agent1.channels.c1.capacity = 1000
  12. agent1.channels.c1.transactionCapacity = 100
  14. agent1.sources.r1.channels = c1
  15. agent1.sources.r1.type = exec
  16. agent1.sources.r1.command = tail -F /root/logs/456.log
  18. # set sink1
  19. agent1.sinks.k1.channel = c1
  20. agent1.sinks.k1.type = avro
  21. agent1.sinks.k1.hostname = node02
  22. agent1.sinks.k1.port = 52020
  24. # set sink2
  25. agent1.sinks.k2.channel = c1
  26. agent1.sinks.k2.type = avro
  27. agent1.sinks.k2.hostname = node03
  28. agent1.sinks.k2.port = 52020
  30. #set sink group
  31. agent1.sinkgroups.g1.sinks = k1 k2
  33. #set failover
  34. agent1.sinkgroups.g1.processor.type = failover
  35. agent1.sinkgroups.g1.processor.priority.k1 = 10
  36. agent1.sinkgroups.g1.processor.priority.k2 = 1
  37. agent1.sinkgroups.g1.processor.maxpenalty = 10000

3.3. Node02 与 Node03 配置 FlumeCollection

node02机器修改配置文件

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim collector.conf
  1. #set Agent name
  2. a1.sources = r1
  3. a1.channels = c1
  4. a1.sinks = k1
  5. #
  6. ##set channel
  7. a1.channels.c1.type = memory
  8. a1.channels.c1.capacity = 1000
  9. a1.channels.c1.transactionCapacity = 100
  10. #
  11. ## other node,nna to nns
  12. a1.sources.r1.type = avro
  13. a1.sources.r1.bind = node02
  14. a1.sources.r1.port = 52020
  15. a1.sources.r1.channels = c1
  16. #
  17. ##set sink to hdfs
  18. a1.sinks.k1.type=hdfs
  19. a1.sinks.k1.hdfs.path= hdfs://node01:8020/flume/failover/
  20. a1.sinks.k1.hdfs.fileType=DataStream
  21. a1.sinks.k1.hdfs.writeFormat=TEXT
  22. a1.sinks.k1.hdfs.rollInterval=10
  23. a1.sinks.k1.channel=c1
  24. a1.sinks.k1.hdfs.filePrefix=%Y-%m-%d
  25. #
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  6. # Describe/configure the source
  7. a1.sources.r1.type = avro
  8. a1.sources.r1.channels = c1
  9. a1.sources.r1.bind = node02
  10. a1.sources.r1.port = 52020
  12. # Describe the sink
  13. a1.sinks.k1.type = logger
  15. # Use a channel which buffers events in memory
  16. a1.channels.c1.type = memory
  17. a1.channels.c1.capacity = 1000
  18. a1.channels.c1.transactionCapacity = 100
  20. # Bind the source and sink to the channel
  21. a1.sources.r1.channels = c1
  22. a1.sinks.k1.channel = c1

node03机器修改配置文件

  1. cd  /export/servers/apache-flume-1.8.0-bin/conf
  2. vim collector.conf
  1. #set Agent name
  2. a1.sources = r1
  3. a1.channels = c1
  4. a1.sinks = k1
  5. #
  6. ##set channel
  7. a1.channels.c1.type = memory
  8. a1.channels.c1.capacity = 1000
  9. a1.channels.c1.transactionCapacity = 100
  10. #
  11. ## other node,nna to nns
  12. a1.sources.r1.type = avro
  13. a1.sources.r1.bind = node03
  14. a1.sources.r1.port = 52020
  15. a1.sources.r1.channels = c1
  16. #
  17. ##set sink to hdfs
  18. a1.sinks.k1.type=hdfs
  19. a1.sinks.k1.hdfs.path= hdfs://node01:8020/flume/failover/
  20. a1.sinks.k1.hdfs.fileType=DataStream
  21. a1.sinks.k1.hdfs.writeFormat=TEXT
  22. a1.sinks.k1.hdfs.rollInterval=10
  23. a1.sinks.k1.channel=c1
  24. a1.sinks.k1.hdfs.filePrefix=%Y-%m-%d
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  6. # Describe/configure the source
  7. a1.sources.r1.type = avro
  8. a1.sources.r1.channels = c1
  9. a1.sources.r1.bind = node03
  10. a1.sources.r1.port = 52020
  12. # Describe the sink
  13. a1.sinks.k1.type = logger
  15. # Use a channel which buffers events in memory
  16. a1.channels.c1.type = memory
  17. a1.channels.c1.capacity = 1000
  18. a1.channels.c1.transactionCapacity = 100
  20. # Bind the source and sink to the channel
  21. a1.sources.r1.channels = c1
  22. a1.sinks.k1.channel = c1

3.4. 顺序启动

node03机器上面启动flume

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -n a1 -c conf -f conf/collector.conf -Dflume.root.logger=DEBUG,console

node02机器上面启动flume

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -n a1 -c conf -f conf/collector.conf -Dflume.root.logger=DEBUG,console

node01机器上面启动flume

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -n agent1 -c conf -f conf/agent.conf -Dflume.root.logger=DEBUG,console

node01机器启动文件产生脚本

  1. cd  /export/servers/shells
  2. sh tail-file.sh

3.5. Failover 测试

下面我们来测试下Flume NG集群的高可用（故障转移）。场景如下：我们在Agent1节点上传文件，由于我们配置Collector1的权重比Collector2大，所以 Collector1优先采集并上传到存储系统。然后我们kill掉Collector1，此时有Collector2负责日志的采集上传工作，之后，我 们手动恢复Collector1节点的Flume服务，再次在Agent1上次文件，发现Collector1恢复优先级别的采集工作。具体截图如下所 示：

Collector1优先上传

1566284970705

HDFS集群中上传的log内容预览

1566284977902

Collector1宕机，Collector2获取优先上传权限

1566284988161

重启Collector1服务，Collector1重新获得优先上传的权限

4. flume 的负载均衡

负载均衡是用于解决一台机器(一个进程)无法解决所有请求而产生的一种算法。Load balancing Sink Processor 能够实现 load balance 功能，如下图Agent1 是一个路由节点，负责将 Channel 暂存的 Event 均衡到对应的多个 Sink组件上，而每个 Sink 组件分别连接到一个独立的 Agent 上，示例配置，如下所示：

1566285002934

在此处我们通过三台机器来进行模拟flume的负载均衡

三台机器规划如下：

node01：采集数据，发送到node02和node03机器上去

node02：接收node01的部分数据

node03：接收node01的部分数据

第一步：开发node01服务器的flume配置

node01服务器配置：

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim load_banlancer_client.conf
  1. #agent name
  3. a1.channels = c1
  4. a1.sources = r1
  5. a1.sinks = k1 k2
  9. #set gruop
  11. a1.sinkgroups = g1
  13. #set channel
  15. a1.channels.c1.type = memory
  16. a1.channels.c1.capacity = 1000
  17. a1.channels.c1.transactionCapacity = 100
  18. a1.sources.r1.channels = c1
  19. a1.sources.r1.type = exec
  20. a1.sources.r1.command = tail -F /export/servers/taillogs/access_log
  22. # set sink1
  24. a1.sinks.k1.channel = c1
  25. a1.sinks.k1.type = avro
  26. #对接端口
  27. a1.sinks.k1.hostname = node02
  28. a1.sinks.k1.port = 52020
  30. # set sink2
  31. a1.sinks.k2.channel = c1
  32. a1.sinks.k2.type = avro
  33. #对接端口
  34. a1.sinks.k2.hostname = node03
  35. a1.sinks.k2.port = 52020
  36. #set sink group
  38. a1.sinkgroups.g1.sinks = k1 k2
  40. #set failover
  41. #负载均衡
  42. a1.sinkgroups.g1.processor.type = load_balance
  43. a1.sinkgroups.g1.processor.backoff = true
  44. #轮训
  45. a1.sinkgroups.g1.processor.selector = round_robin
  46. a1.sinkgroups.g1.processor.selector.maxTimeOut=10000

第二步：开发node02服务器的flume配置

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  3. vim load_banlancer_server.conf
  1. # Name the components on this agent
  3. a1.sources = r1
  4. a1.sinks = k1
  5. a1.channels = c1
  7. # Describe/configure the source
  9. a1.sources.r1.type = avro
  10. a1.sources.r1.channels = c1
  11. a1.sources.r1.bind = node02
  12. a1.sources.r1.port = 52020
  14. # Describe the sink
  15. a1.sinks.k1.type = logger
  17. # Use a channel which buffers events in memory
  18. a1.channels.c1.type = memory
  19. a1.channels.c1.capacity = 1000
  20. a1.channels.c1.transactionCapacity = 100
  22. # Bind the source and sink to the channel
  23. a1.sources.r1.channels = c1
  24. a1.sinks.k1.channel = c1

第三步：开发node03服务器flume配置

node03服务器配置

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim load_banlancer_server.conf
  1. # Name the components on this agent
  2. a1.sources = r1
  3. a1.sinks = k1
  4. a1.channels = c1
  6. # Describe/configure the source
  7. a1.sources.r1.type = avro
  8. a1.sources.r1.channels = c1
  9. a1.sources.r1.bind = node03
  10. a1.sources.r1.port = 52020
  12. # Describe the sink
  13. a1.sinks.k1.type = logger
  15. # Use a channel which buffers events in memory
  16. a1.channels.c1.type = memory
  17. a1.channels.c1.capacity = 1000
  18. a1.channels.c1.transactionCapacity = 100
  21. # Bind the source and sink to the channel
  22. a1.sources.r1.channels = c1
  23. a1.sinks.k1.channel = c1

第四步：准备启动flume服务

启动node03的flume服务

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -n a1 -c conf -f conf/load_banlancer_server.conf -Dflume.root.logger=DEBUG,console

启动node02的flume服务

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -n a1 -c conf -f conf/load_banlancer_server.conf -Dflume.root.logger=DEBUG,console

启动node01的flume服务

  1. cd /export/servers/apache-flume-1.8.0-bin
  3. bin/flume-ng agent -n a1 -c conf -f conf/load_banlancer_client.conf -Dflume.root.logger=DEBUG,console

第五步：node01服务器运行脚本产生数据

cd /export/servers/shells

sh tail-file.sh

5. Flume 案例-静态拦截器

1. 案例场景

A、B两台日志服务机器实时生产日志主要类型为access.log、nginx.log、web.log

现在要求：

把A、B 机器中的access.log、nginx.log、web.log 采集汇总到C机器上然后统一收集到hdfs中。

但是在hdfs中要求的目录为：

  1. /source/logs/access/20180101/**
  2. /source/logs/nginx/20180101/**
  3. /source/logs/web/20180101/**

2. 场景分析

1566285061121

​ 图一

3. 数据流程处理分析

1566285081174

4、实现

服务器A对应的IP为 192.168.174.100

服务器B对应的IP为 192.168.174.110

服务器C对应的IP为 node03

采集端配置文件开发

node01与node02服务器开发flume的配置文件

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim exec_source_avro_sink.conf
  1. # Name the components on this agent
  2. a1.sources = r1 r2 r3
  3. a1.sinks = k1
  4. a1.channels = c1
  5. # Describe/configure the source
  6. a1.sources.r1.type = exec
  7. a1.sources.r1.command = tail -F /export/servers/taillogs/access.log
  8. a1.sources.r1.interceptors = i1
  9. a1.sources.r1.interceptors.i1.type = static
  10. ##  static拦截器的功能就是往采集到的数据的header中插入自己定## 义的key-value对
  11. a1.sources.r1.interceptors.i1.key = type
  12. a1.sources.r1.interceptors.i1.value = access
  14. a1.sources.r2.type = exec
  15. a1.sources.r2.command = tail -F /export/servers/taillogs/nginx.log
  16. a1.sources.r2.interceptors = i2
  17. a1.sources.r2.interceptors.i2.type = static
  18. a1.sources.r2.interceptors.i2.key = type
  19. a1.sources.r2.interceptors.i2.value = nginx
  21. a1.sources.r3.type = exec
  22. a1.sources.r3.command = tail -F /export/servers/taillogs/web.log
  23. a1.sources.r3.interceptors = i3
  24. a1.sources.r3.interceptors.i3.type = static
  25. a1.sources.r3.interceptors.i3.key = type
  26. a1.sources.r3.interceptors.i3.value = web
  28. # Describe the sink
  29. a1.sinks.k1.type = avro
  30. a1.sinks.k1.hostname = node03
  31. a1.sinks.k1.port = 41414
  32. # Use a channel which buffers events in memory
  33. a1.channels.c1.type = memory
  34. a1.channels.c1.capacity = 20000
  35. a1.channels.c1.transactionCapacity = 10000
  37. # Bind the source and sink to the channel
  38. a1.sources.r1.channels = c1
  39. a1.sources.r2.channels = c1
  40. a1.sources.r3.channels = c1
  41. a1.sinks.k1.channel = c1

服务端配置文件开发

在node03上面开发flume配置文件

  1. cd /export/servers/apache-flume-1.8.0-bin/conf
  2. vim avro_source_hdfs_sink.conf
  1. a1.sources = r1
  2. a1.sinks = k1
  3. a1.channels = c1
  4. #定义source
  5. a1.sources.r1.type = avro
  6. a1.sources.r1.bind = node03
  7. a1.sources.r1.port =41414
  9. #添加时间拦截器
  10. a1.sources.r1.interceptors = i1
  11. a1.sources.r1.interceptors.i1.type = org.apache.flume.interceptor.TimestampInterceptor$Builder
  12. #定义channels
  13. a1.channels.c1.type = memory
  14. a1.channels.c1.capacity = 20000
  15. a1.channels.c1.transactionCapacity = 10000
  16. #定义sink
  17. a1.sinks.k1.type = hdfs
  18. a1.sinks.k1.hdfs.path=hdfs://node01:8020/source/logs/%{type}/%Y%m%d
  19. a1.sinks.k1.hdfs.filePrefix =events
  20. a1.sinks.k1.hdfs.fileType = DataStream
  21. a1.sinks.k1.hdfs.writeFormat = Text
  22. #时间类型
  23. a1.sinks.k1.hdfs.useLocalTimeStamp = true
  24. #生成的文件不按条数生成
  25. a1.sinks.k1.hdfs.rollCount = 0
  26. #生成的文件按时间生成
  27. a1.sinks.k1.hdfs.rollInterval = 30
  28. #生成的文件按大小生成
  29. a1.sinks.k1.hdfs.rollSize  = 10485760
  30. #批量写入hdfs的个数
  31. a1.sinks.k1.hdfs.batchSize = 10000
  32. #flume操作hdfs的线程数（包括新建，写入等）
  33. a1.sinks.k1.hdfs.threadsPoolSize=10
  34. #操作hdfs超时时间
  35. a1.sinks.k1.hdfs.callTimeout=30000
  36. #组装source、channel、sink
  37. a1.sources.r1.channels = c1
  38. a1.sinks.k1.channel = c1

采集端文件生成脚本

在node01与node02上面开发shell脚本，模拟数据生成

  1. cd /export/servers/shells
  2. vim server.sh
  1. #!/bin/bash
  3. while true
  5. do 
  7. date >> /export/servers/taillogs/access.log;
  9. date >> /export/servers/taillogs/web.log;
  11. date >> /export/servers/taillogs/nginx.log;
  13.   sleep 0.5;
  15. done

顺序启动服务

node03启动flume实现数据收集

  1. cd /export/servers/apache-flume-1.8.0-bin
  3. bin/flume-ng agent -c conf -f conf/avro_source_hdfs_sink.conf -name a1 -Dflume.root.logger=DEBUG,console

node01与node02启动flume实现数据监控

  1. cd /export/servers/apache-flume-1.8.0-bin
  3. bin/flume-ng agent -c conf -f conf/exec_source_avro_sink.conf -name a1 -Dflume.root.logger=DEBUG,console

node01与node02启动生成文件脚本

  1. cd /export/servers/shells
  2. sh server.sh

5、项目实现截图

1566285113421

1566285121716

6. Flume 案例二

案例需求：

在数据采集之后，通过flume的拦截器，实现不需要的数据过滤掉，并将指定的第一个字段进行加密，加密之后再往hdfs上面保存

原始数据与处理之后的数据对比

1566285132227

图一 原始文件内容

1566285139361

图二 HDFS上产生收集到的处理数

实现步骤

第一步：创建maven java工程，导入jar包

  1. <?xml version="1.0" encoding="UTF-8"?>
  2. <project xmlns="http://maven.apache.org/POM/4.0.0"
  3.  xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
  4.  xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
  5.  <modelVersion>4.0.0</modelVersion>
  7.  <groupId>cn.le.cloud</groupId>
  8.  <artifactId>example-flume-intercepter</artifactId>
  9.  <version>1.0-SNAPSHOT</version>
  11.  <dependencies>
  12.  <dependency>
  13.  <groupId>org.apache.flume</groupId>
  14.  <artifactId>flume-ng-sdk</artifactId>
  15.  <version>1.8.0</version>
  16.  </dependency>
  17.  <dependency>
  18.  <groupId>org.apache.flume</groupId>
  19.  <artifactId>flume-ng-core</artifactId>
  20.  <version>1.8.0</version>
  21.  </dependency>
  22.  </dependencies>
  24.  <build>
  25.  <plugins>
  26.  <plugin>
  27.  <groupId>org.apache.maven.plugins</groupId>
  28.  <artifactId>maven-compiler-plugin</artifactId>
  29.  <version>3.0</version>
  30.  <configuration>
  31.  <source>1.8</source>
  32.  <target>1.8</target>
  33.  <encoding>UTF-8</encoding>
  34.  <!--    <verbal>true</verbal>-->
  35.  </configuration>
  36.  </plugin>
  37.  <plugin>
  38.  <groupId>org.apache.maven.plugins</groupId>
  39.  <artifactId>maven-shade-plugin</artifactId>
  40.  <version>3.1.1</version>
  41.  <executions>
  42.  <execution>
  43.  <phase>package</phase>
  44.  <goals>
  45.  <goal>shade</goal>
  46.  </goals>
  47.  <configuration>
  48.  <filters>
  49.  <filter>
  50.  <artifact>*:*</artifact>
  51.  <excludes>
  52.  <exclude>META-INF/*.SF</exclude>
  53.  <exclude>META-INF/*.DSA</exclude>
  54.  <exclude>META-INF/*.RSA</exclude>
  55.  </excludes>
  56.  </filter>
  57.  </filters>
  58.  <transformers>
  59.  <transformer implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer">
  60.  <mainClass></mainClass>
  61.  </transformer>
  62.  </transformers>
  63.  </configuration>
  64.  </execution>
  65.  </executions>
  66.  </plugin>
  67.  </plugins>
  68.  </build>
  69. </project>

第二步：自定义flume的拦截器

  1. package cn.le.iterceptor;
  2. import com.google.common.base.Charsets;
  3. import org.apache.flume.Context;
  4. import org.apache.flume.Event;
  5. import org.apache.flume.interceptor.Interceptor;
  6. import java.security.MessageDigest;
  7. import java.security.NoSuchAlgorithmException;
  8. import java.util.ArrayList;
  9. import java.util.List;
  10. import java.util.regex.Matcher;
  11. import java.util.regex.Pattern;
  13. import static cn.le.iterceptor.CustomParameterInterceptor.Constants.*;
  15. public class CustomParameterInterceptor implements Interceptor {
  16.  /** The field_separator.指明每一行字段的分隔符 */
  17.  private final String fields_separator;
  19.  /** The indexs.通过分隔符分割后，指明需要那列的字段 下标*/
  20.  private final String indexs;
  22.  /** The indexs_separator. 多个下标的分隔符*/
  23.  private final String indexs_separator;
  25.  /**
  26.      *
  27.      * @param indexs
  28.      * @param indexs_separator
  29.      */
  30.  public CustomParameterInterceptor( String fields_separator,
  31.  String indexs, String indexs_separator,String encrypted_field_index) {
  32.  String f = fields_separator.trim();
  33.  String i = indexs_separator.trim();
  34.  this.indexs = indexs;
  35.  this.encrypted_field_index=encrypted_field_index.trim();
  36.  if (!f.equals("")) {
  37.             f = UnicodeToString(f);
  38.  }
  39.  this.fields_separator =f;
  40.  if (!i.equals("")) {
  41.             i = UnicodeToString(i);
  42.  }
  43.  this.indexs_separator = i;
  44.  }
  46.  /*
  47.      *
  48.      * \t 制表符 ('\u0009') \n 新行（换行）符 (' ') \r 回车符 (' ') \f 换页符 ('\u000C') \a 报警
  49.      * (bell) 符 ('\u0007') \e 转义符 ('\u001B') \cx  空格(\u0020)对应于 x 的控制符
  50.      *
  51.      * @param str
  52.      * @return
  53.      * @data:2015-6-30
  54.      */
  56.  /** The encrypted_field_index. 需要加密的字段下标*/
  57.  private final String encrypted_field_index;
  58.  public static String UnicodeToString(String str) {
  59.  Pattern pattern = Pattern.compile("(\\\\u(\\p{XDigit}{4}))");
  60.  Matcher matcher = pattern.matcher(str);
  61.  char ch;
  62.  while (matcher.find()) {
  63.             ch = (char) Integer.parseInt(matcher.group(2), 16);
  64.             str = str.replace(matcher.group(1), ch + "");
  65.  }
  66.  return str;
  67.  }
  69.  /*
  70.      * @see org.apache.flume.interceptor.Interceptor#intercept(org.apache.flume.Event)
  71.      * 单个event拦截逻辑
  72.      */
  73.  public Event intercept(Event event) {
  74.  if (event == null) {
  75.  return null;
  76.  }
  77.  try {
  78.  String line = new String(event.getBody(), Charsets.UTF_8);
  79.  String[] fields_spilts = line.split(fields_separator);
  80.  String[] indexs_split = indexs.split(indexs_separator);
  81.  String newLine="";
  82.  for (int i = 0; i < indexs_split.length; i++) {
  83.  int parseInt = Integer.parseInt(indexs_split[i]);
  84.  //对加密字段进行加密
  85.  if(!"".equals(encrypted_field_index)&&encrypted_field_index.equals(indexs_split[i])){
  86.                     newLine+=StringUtils.GetMD5Code(fields_spilts[parseInt]);
  87.  }else{
  88.                     newLine+=fields_spilts[parseInt];
  89.  }
  91.  if(i!=indexs_split.length-1){
  92.                     newLine+=fields_separator;
  93.  }
  94.  }
  95.             event.setBody(newLine.getBytes(Charsets.UTF_8));
  96.  return event;
  97.  } catch (Exception e) {
  98.  return event;
  99.  }
  100.  }
  102.  /*
  103.      * @see org.apache.flume.interceptor.Interceptor#intercept(java.util.List)
  104.      * 批量event拦截逻辑
  105.      */
  106.  public List<Event> intercept(List<Event> events) {
  107.  List<Event> out = new ArrayList<Event>();
  108.  for (Event event : events) {
  109.  Event outEvent = intercept(event);
  110.  if (outEvent != null) {
  111.                 out.add(outEvent);
  112.  }
  113.  }
  114.  return out;
  115.  }
  117.  /*
  118.      * @see org.apache.flume.interceptor.Interceptor#initialize()
  119.      */
  120.  public void initialize() {
  121.  // TODO Auto-generated method stub
  123.  }
  125.  /*
  126.      * @see org.apache.flume.interceptor.Interceptor#close()
  127.      */
  128.  public void close() {
  129.  // TODO Auto-generated method stub
  131.  }
  134.  /**
  135.      * 相当于自定义Interceptor的工厂类
  136.      * 在flume采集配置文件中通过制定该Builder来创建Interceptor对象
  137.      * 可以在Builder中获取、解析flume采集配置文件中的拦截器Interceptor的自定义参数：
  138.      * 字段分隔符，字段下标，下标分隔符、加密字段下标 ...等
  139.      * @author
  140.      *
  141.      */
  142.  public static class Builder implements Interceptor.Builder {
  144.  /** The fields_separator.指明每一行字段的分隔符 */
  145.  private String fields_separator;
  147.  /** The indexs.通过分隔符分割后，指明需要那列的字段 下标*/
  148.  private String indexs;
  150.  /** The indexs_separator. 多个下标下标的分隔符*/
  151.  private String indexs_separator;
  153.  /** The encrypted_field. 需要加密的字段下标*/
  154.  private String encrypted_field_index;
  155.  /*
  156.          * @see org.apache.flume.conf.Configurable#configure(org.apache.flume.Context)
  157.          */
  158.  public void configure(Context context) {
  159.             fields_separator = context.getString(FIELD_SEPARATOR, DEFAULT_FIELD_SEPARATOR);
  160.             indexs = context.getString(INDEXS, DEFAULT_INDEXS);
  161.             indexs_separator = context.getString(INDEXS_SEPARATOR, DEFAULT_INDEXS_SEPARATOR);
  162.             encrypted_field_index= context.getString(ENCRYPTED_FIELD_INDEX, DEFAULT_ENCRYPTED_FIELD_INDEX);
  163.  }
  164.  /*
  165.          * @see org.apache.flume.interceptor.Interceptor.Builder#build()
  166.          */
  167.  public Interceptor build() {
  168.  return new CustomParameterInterceptor(fields_separator, indexs, indexs_separator,encrypted_field_index);
  169.  }
  170.  }
  171.  /**
  172.      * 常量
  173.      *
  174.      */
  175.  public static class Constants {
  176.  /** The Constant FIELD_SEPARATOR. */
  177.  public static final String FIELD_SEPARATOR = "fields_separator";
  179.  /** The Constant DEFAULT_FIELD_SEPARATOR. */
  180.  public static final String DEFAULT_FIELD_SEPARATOR =" ";
  182.  /** The Constant INDEXS. */
  183.  public static final String INDEXS = "indexs";
  185.  /** The Constant DEFAULT_INDEXS. */
  186.  public static final String DEFAULT_INDEXS = "0";
  188.  /** The Constant INDEXS_SEPARATOR. */
  189.  public static final String INDEXS_SEPARATOR = "indexs_separator";
  191.  /** The Constant DEFAULT_INDEXS_SEPARATOR. */
  192.  public static final String DEFAULT_INDEXS_SEPARATOR = ",";
  194.  /** The Constant ENCRYPTED_FIELD_INDEX. */
  195.  public static final String ENCRYPTED_FIELD_INDEX = "encrypted_field_index";
  197.  /** The Constant DEFAUL_TENCRYPTED_FIELD_INDEX. */
  198.  public static final String DEFAULT_ENCRYPTED_FIELD_INDEX = "";
  200.  /** The Constant PROCESSTIME. */
  201.  public static final String PROCESSTIME = "processTime";
  202.  /** The Constant PROCESSTIME. */
  203.  public static final String DEFAULT_PROCESSTIME = "a";
  205.  }
  206.  /**
  207.      * 工具类：字符串md5加密
  208.      */
  209.  public static class StringUtils {
  210.  // 全局数组
  211.  private final static String[] strDigits = { "0", "1", "2", "3", "4", "5",
  212.  "6", "7", "8", "9", "a", "b", "c", "d", "e", "f" };
  213.  // 返回形式为数字跟字符串
  214.  private static String byteToArrayString(byte bByte) {
  215.  int iRet = bByte;
  216.  // System.out.println("iRet="+iRet);
  217.  if (iRet < 0) {
  218.                 iRet += 256;
  219.  }
  220.  int iD1 = iRet / 16;
  221.  int iD2 = iRet % 16;
  222.  return strDigits[iD1] + strDigits[iD2];
  223.  }
  225.  // 返回形式只为数字
  226.  private static String byteToNum(byte bByte) {
  227.  int iRet = bByte;
  228.  System.out.println("iRet1=" + iRet);
  229.  if (iRet < 0) {
  230.                 iRet += 256;
  231.  }
  232.  return String.valueOf(iRet);
  233.  }
  235.  // 转换字节数组为16进制字串
  236.  private static String byteToString(byte[] bByte) {
  237.  StringBuffer sBuffer = new StringBuffer();
  238.  for (int i = 0; i < bByte.length; i++) {
  239.                 sBuffer.append(byteToArrayString(bByte[i]));
  240.  }
  241.  return sBuffer.toString();
  242.  }
  244.  public static String GetMD5Code(String strObj) {
  245.  String resultString = null;
  246.  try {
  247.                 resultString = new String(strObj);
  248.  MessageDigest md = MessageDigest.getInstance("MD5");
  249.  // md.digest() 该函数返回值为存放哈希值结果的byte数组
  250.                 resultString = byteToString(md.digest(strObj.getBytes()));
  251.  } catch (NoSuchAlgorithmException ex) {
  252.                 ex.printStackTrace();
  253.  }
  254.  return resultString;
  255.  }
  256.  }
  258. }

第三步：打包上传服务器

将我们的拦截器打成jar包放到flume的lib目录下

1566285178862

第四步：开发flume的配置文件

第三台机器开发flume的配置文件

  1. cd  /export/servers/apache-flume-1.8.0-bin/conf
  2. vim spool-interceptor-hdfs.conf
  1. a1.channels = c1
  2. a1.sources = r1
  3. a1.sinks = s1
  4. #channel
  6. a1.channels.c1.type = memory
  7. a1.channels.c1.capacity=100000
  8. a1.channels.c1.transactionCapacity=50000
  10. #source
  12. a1.sources.r1.channels = c1
  13. a1.sources.r1.type = spooldir
  14. a1.sources.r1.spoolDir = /export/servers/intercept
  15. a1.sources.r1.batchSize= 50
  16. a1.sources.r1.inputCharset = UTF-8
  20. a1.sources.r1.interceptors =i1 i2
  21. a1.sources.r1.interceptors.i1.type =cn.le.iterceptor.CustomParameterInterceptor$Builder
  22. a1.sources.r1.interceptors.i1.fields_separator=\\u0009
  24. a1.sources.r1.interceptors.i1.indexs =0,1,3,5,6
  25. a1.sources.r1.interceptors.i1.indexs_separator =\\u002c
  26. a1.sources.r1.interceptors.i1.encrypted_field_index =0
  27. a1.sources.r1.interceptors.i2.type = org.apache.flume.interceptor.TimestampInterceptor$Builder
  29. #sink
  31. a1.sinks.s1.channel = c1
  33. a1.sinks.s1.type = hdfs
  35. a1.sinks.s1.hdfs.path =hdfs://node01:8020/flume/intercept/%Y%m%d
  37. a1.sinks.s1.hdfs.filePrefix = event
  38. a1.sinks.s1.hdfs.fileSuffix = .log
  39. a1.sinks.s1.hdfs.rollSize = 10485760
  40. a1.sinks.s1.hdfs.rollInterval =20
  41. a1.sinks.s1.hdfs.rollCount = 0
  42. a1.sinks.s1.hdfs.batchSize = 1500
  44. a1.sinks.s1.hdfs.round = true
  45. a1.sinks.s1.hdfs.roundUnit = minute
  46. a1.sinks.s1.hdfs.threadsPoolSize = 25
  47. a1.sinks.s1.hdfs.useLocalTimeStamp = true
  48. a1.sinks.s1.hdfs.minBlockReplicas = 1
  49. a1.sinks.s1.hdfs.fileType =DataStream
  50. a1.sinks.s1.hdfs.writeFormat = Text
  51. a1.sinks.s1.hdfs.callTimeout = 60000
  52. a1.sinks.s1.hdfs.idleTimeout =60

第五步：上传测试数据

上传我们的测试数据到/export/servers/intercept 这个目录下面去，如果目录不存在则创建

  1. mkdir  -p /export/servers/intercept

测试数据如下

  1. 13601249301 100 200 300 400 500 600 700
  2. 13601249302 100 200 300 400 500 600 700
  3. 13601249303 100 200 300 400 500 600 700
  4. 13601249304 100 200 300 400 500 600 700
  5. 13601249305 100 200 300 400 500 600 700
  6. 13601249306 100 200 300 400 500 600 700
  7. 13601249307 100 200 300 400 500 600 700
  8. 13601249308 100 200 300 400 500 600 700
  9. 13601249309 100 200 300 400 500 600 700
  10. 13601249310 100 200 300 400 500 600 700
  11. 13601249311 100 200 300 400 500 600 700
  12. 13601249312 100 200 300 400 500 600 700

第六步：启动flume

  1. cd /export/servers/apache-flume-1.8.0-bin
  2. bin/flume-ng agent -c conf -f conf/spool-interceptor-hdfs.conf -name a1 -Dflume.root.logger=DEBUG,console

1566146311181

1566146325940

1566146341091

1566146363100156614637714215661463965111566146411253

1566146441320

1566146453275

  1. package cn.le.flumesource;
  3. import org.apache.flume.Context;
  4. import org.apache.flume.Event;
  5. import org.apache.flume.EventDeliveryException;
  6. import org.apache.flume.PollableSource;
  7. import org.apache.flume.conf.Configurable;
  8. import org.apache.flume.event.SimpleEvent;
  9. import org.apache.flume.source.AbstractSource;
  10. import org.slf4j.Logger;
  12. import java.util.ArrayList;
  13. import java.util.HashMap;
  14. import java.util.List;
  16. import static org.slf4j.LoggerFactory.*;
  18. public class MySqlSource extends AbstractSource implements Configurable, PollableSource {
  20.  //打印日志
  21.  private static final Logger LOG = getLogger(MySqlSource.class);
  22.  //定义sqlHelper
  23.  private QueryMySql sqlSourceHelper;
  26.  @Override
  27.  public long getBackOffSleepIncrement() {
  28.  return 0;
  29.  }
  31.  @Override
  32.  public long getMaxBackOffSleepInterval() {
  33.  return 0;
  34.  }
  36.  @Override
  37.  public void configure(Context context) {
  38.  //初始化
  39.         sqlSourceHelper = new QueryMySql(context);
  40.  }
  42.  @Override
  43.  public PollableSource.Status process() throws EventDeliveryException {
  44.  try {
  45.  //查询数据表
  46.  List<List<Object>> result = sqlSourceHelper.executeQuery();
  47.  //存放event的集合
  48.  List<Event> events = new ArrayList<>();
  49.  //存放event头集合
  50.  HashMap<String, String> header = new HashMap<>();
  51.  //如果有返回数据，则将数据封装为event
  52.  if (!result.isEmpty()) {
  53.  List<String> allRows = sqlSourceHelper.getAllRows(result);
  54.  Event event = null;
  55.  for (String row : allRows) {
  56.                     event = new SimpleEvent();
  57.                     event.setBody(row.getBytes());
  58.                     event.setHeaders(header);
  59.                     events.add(event);
  60.  }
  61.  //将event写入channel
  62.  this.getChannelProcessor().processEventBatch(events);
  63.  //更新数据表中的offset信息
  64.                 sqlSourceHelper.updateOffset2DB(result.size());
  65.  }
  66.  //等待时长
  67.  Thread.sleep(sqlSourceHelper.getRunQueryDelay());
  68.  return Status.READY;
  69.  } catch (InterruptedException e) {
  70.             LOG.error("Error procesing row", e);
  71.  return Status.BACKOFF;
  72.  }
  73.  }
  75.  @Override
  76.  public synchronized void stop() {
  77.         LOG.info("Stopping sql source {} ...", getName());
  78.  try {
  79.  //关闭资源
  80.             sqlSourceHelper.close();
  81.  } finally {
  82.  super.stop();
  83.  }
  84.  }
  85. }
  1. package cn.le.flumesource;
  3. import org.apache.flume.Context;
  4. import org.apache.flume.conf.ConfigurationException;
  5. import org.apache.http.ParseException;
  6. import org.slf4j.Logger;
  7. import org.slf4j.LoggerFactory;
  9. import java.sql.*;
  10. import java.util.ArrayList;
  11. import java.util.List;
  12. import java.util.Properties;
  14. public class QueryMySql {
  15.  private static final Logger LOG = LoggerFactory.getLogger(QueryMySql.class);
  17.  private int runQueryDelay, //两次查询的时间间隔
  18.             startFrom, //开始id
  19.             currentIndex, //当前id
  20.             recordSixe = 0, //每次查询返回结果的条数
  21.             maxRow; //每次查询的最大条数
  24.  private String table, //要操作的表
  25.             columnsToSelect, //用户传入的查询的列
  26.             customQuery, //用户传入的查询语句
  27.             query, //构建的查询语句
  28.             defaultCharsetResultSet;//编码集
  30.  //上下文，用来获取配置文件
  31.  private Context context;
  33.  //为定义的变量赋值（默认值），可在flume任务的配置文件中修改
  34.  private static final int DEFAULT_QUERY_DELAY = 10000;
  35.  private static final int DEFAULT_START_VALUE = 0;
  36.  private static final int DEFAULT_MAX_ROWS = 2000;
  37.  private static final String DEFAULT_COLUMNS_SELECT = "*";
  38.  private static final String DEFAULT_CHARSET_RESULTSET = "UTF-8";
  40.  private static Connection conn = null;
  41.  private static PreparedStatement ps = null;
  42.  private static String connectionURL, connectionUserName, connectionPassword;
  44.  //加载静态资源
  45.  static {
  46.  Properties p = new Properties();
  47.  try {
  48.             p.load(QueryMySql.class.getClassLoader().getResourceAsStream("jdbc.properties"));
  49.             connectionURL = p.getProperty("dbUrl");
  50.             connectionUserName = p.getProperty("dbUser");
  51.             connectionPassword = p.getProperty("dbPassword");
  52.  Class.forName(p.getProperty("dbDriver"));
  53.  } catch (Exception e) {
  54.             LOG.error(e.toString());
  55.  }
  56.  }
  58.  //获取JDBC连接
  59.  private static Connection InitConnection(String url, String user, String pw) {
  60.  try {
  61.  Connection conn = DriverManager.getConnection(url, user, pw);
  62.  if (conn == null)
  63.  throw new SQLException();
  64.  return conn;
  65.  } catch (SQLException e) {
  66.             e.printStackTrace();
  67.  }
  68.  return null;
  69.  }
  71.  //构造方法
  72.  QueryMySql(Context context) throws ParseException {
  73.  //初始化上下文
  74.  this.context = context;
  76.  //有默认值参数：获取flume任务配置文件中的参数，读不到的采用默认值
  77.  this.columnsToSelect = context.getString("columns.to.select", DEFAULT_COLUMNS_SELECT);
  78.  this.runQueryDelay = context.getInteger("run.query.delay", DEFAULT_QUERY_DELAY);
  79.  this.startFrom = context.getInteger("start.from", DEFAULT_START_VALUE);
  80.  this.defaultCharsetResultSet = context.getString("default.charset.resultset", DEFAULT_CHARSET_RESULTSET);
  82.  //无默认值参数：获取flume任务配置文件中的参数
  83.  this.table = context.getString("table");
  84.  this.customQuery = context.getString("custom.query");
  85.         connectionURL = context.getString("connection.url");
  86.         connectionUserName = context.getString("connection.user");
  87.         connectionPassword = context.getString("connection.password");
  88.         conn = InitConnection(connectionURL, connectionUserName, connectionPassword);
  90.  //校验相应的配置信息，如果没有默认值的参数也没赋值，抛出异常
  91.         checkMandatoryProperties();
  92.  //获取当前的id
  93.         currentIndex = getStatusDBIndex(startFrom);
  94.  //构建查询语句
  95.         query = buildQuery();
  96.  }
  98.  //校验相应的配置信息（表，查询语句以及数据库连接的参数）
  99.  private void checkMandatoryProperties() {
  100.  if (table == null) {
  101.  throw new ConfigurationException("property table not set");
  102.  }
  103.  if (connectionURL == null) {
  104.  throw new ConfigurationException("connection.url property not set");
  105.  }
  106.  if (connectionUserName == null) {
  107.  throw new ConfigurationException("connection.user property not set");
  108.  }
  109.  if (connectionPassword == null) {
  110.  throw new ConfigurationException("connection.password property not set");
  111.  }
  112.  }
  114.  //构建sql语句
  115.  private String buildQuery() {
  116.  String sql = "";
  117.  //获取当前id
  118.         currentIndex = getStatusDBIndex(startFrom);
  119.         LOG.info(currentIndex + "");
  120.  if (customQuery == null) {
  121.             sql = "SELECT " + columnsToSelect + " FROM " + table;
  122.  } else {
  123.             sql = customQuery;
  124.  }
  125.  StringBuilder execSql = new StringBuilder(sql);
  126.  //以id作为offset
  127.  if (!sql.contains("where")) {
  128.             execSql.append(" where ");
  129.             execSql.append("id").append(">").append(currentIndex);
  130.  return execSql.toString();
  131.  } else {
  132.  int length = execSql.toString().length();
  133.  return execSql.toString().substring(0, length - String.valueOf(currentIndex).length()) + currentIndex;
  134.  }
  135.  }
  137.  //执行查询
  138.  List<List<Object>> executeQuery() {
  139.  try {
  140.  //每次执行查询时都要重新生成sql，因为id不同
  141.             customQuery = buildQuery();
  142.  //存放结果的集合
  143.  List<List<Object>> results = new ArrayList<>();
  144.  if (ps == null) {
  145.  //
  146.                 ps = conn.prepareStatement(customQuery);
  147.  }
  148.  ResultSet result = ps.executeQuery(customQuery);
  149.  while (result.next()) {
  150.  //存放一条数据的集合（多个列）
  151.  List<Object> row = new ArrayList<>();
  152.  //将返回结果放入集合
  153.  for (int i = 1; i <= result.getMetaData().getColumnCount(); i++) {
  154.                     row.add(result.getObject(i));
  155.  }
  156.                 results.add(row);
  157.  }
  158.             LOG.info("execSql:" + customQuery + "\nresultSize:" + results.size());
  159.  return results;
  160.  } catch (SQLException e) {
  161.             LOG.error(e.toString());
  162.  // 重新连接
  163.             conn = InitConnection(connectionURL, connectionUserName, connectionPassword);
  164.  }
  165.  return null;
  166.  }
  168.  //将结果集转化为字符串，每一条数据是一个list集合，将每一个小的list集合转化为字符串
  169.  List<String> getAllRows(List<List<Object>> queryResult) {
  170.  List<String> allRows = new ArrayList<>();
  171.  if (queryResult == null || queryResult.isEmpty())
  172.  return allRows;
  173.  StringBuilder row = new StringBuilder();
  174.  for (List<Object> rawRow : queryResult) {
  175.  Object value = null;
  176.  for (Object aRawRow : rawRow) {
  177.                 value = aRawRow;
  178.  if (value == null) {
  179.                     row.append(",");
  180.  } else {
  181.                     row.append(aRawRow.toString()).append(",");
  182.  }
  183.  }
  184.             allRows.add(row.toString());
  185.             row = new StringBuilder();
  186.  }
  187.  return allRows;
  188.  }
  190.  //更新offset元数据状态，每次返回结果集后调用。必须记录每次查询的offset值，为程序中断续跑数据时使用，以id为offset
  191.  void updateOffset2DB(int size) {
  192.  //以source_tab做为KEY，如果不存在则插入，存在则更新（每个源表对应一条记录）
  193.  String sql = "insert into flume_meta(source_tab,currentIndex) VALUES('"
  194.  + this.table
  195.  + "','" + (recordSixe += size)
  196.  + "') on DUPLICATE key update source_tab=values(source_tab),currentIndex=values(currentIndex)";
  197.         LOG.info("updateStatus Sql:" + sql);
  198.         execSql(sql);
  199.  }
  201.  //执行sql语句
  202.  private void execSql(String sql) {
  203.  try {
  204.             ps = conn.prepareStatement(sql);
  205.             LOG.info("exec::" + sql);
  206.             ps.execute();
  207.  } catch (SQLException e) {
  208.             e.printStackTrace();
  209.  }
  210.  }
  212.  //获取当前id的offset
  213.  private Integer getStatusDBIndex(int startFrom) {
  214.  //从flume_meta表中查询出当前的id是多少
  215.  String dbIndex = queryOne("select currentIndex from flume_meta where source_tab='" + table + "'");
  216.  if (dbIndex != null) {
  217.  return Integer.parseInt(dbIndex);
  218.  }
  219.  //如果没有数据，则说明是第一次查询或者数据表中还没有存入数据，返回最初传入的值
  220.  return startFrom;
  221.  }
  223.  //查询一条数据的执行语句(当前id)
  224.  private String queryOne(String sql) {
  225.  ResultSet result = null;
  226.  try {
  227.             ps = conn.prepareStatement(sql);
  228.             result = ps.executeQuery();
  229.  while (result.next()) {
  230.  return result.getString(1);
  231.  }
  232.  } catch (SQLException e) {
  233.             e.printStackTrace();
  234.  }
  235.  return null;
  236.  }
  238.  //关闭相关资源
  239.  void close() {
  240.  try {
  241.             ps.close();
  242.             conn.close();
  243.  } catch (SQLException e) {
  244.             e.printStackTrace();
  245.  }
  246.  }
  248.  int getCurrentIndex() {
  249.  return currentIndex;
  250.  }
  252.  void setCurrentIndex(int newValue) {
  253.         currentIndex = newValue;
  254.  }
  256.  int getRunQueryDelay() {
  257.  return runQueryDelay;
  258.  }
  260.  String getQuery() {
  261.  return query;
  262.  }
  264.  String getConnectionURL() {
  265.  return connectionURL;
  266.  }
  268.  private boolean isCustomQuerySet() {
  269.  return (customQuery != null);
  270.  }
  272.  Context getContext() {
  273.  return context;
  274.  }
  276.  public String getConnectionUserName() {
  277.  return connectionUserName;
  278.  }
  280.  public String getConnectionPassword() {
  281.  return connectionPassword;
  282.  }
  284.  String getDefaultCharsetResultSet() {
  285.  return defaultCharsetResultSet;
  286.  }
  287. }
posted @ 2019-08-20 17:13  不穿格子衫的徍爺  阅读(186)  评论(0)  编辑  收藏  举报