Flume-1.8.0_部署与常用案例

时间:2022-07-26
本文章向大家介绍Flume-1.8.0_部署与常用案例,主要内容包括其使用实例、应用技巧、基本知识点总结和需要注意事项,具有一定的参考价值,需要的朋友可以参考一下。

该文章是基于 Hadoop2.7.6_01_部署 进行的

Flume官方文档:FlumeUserGuide

常见问题:记flume部署过程中遇到的问题以及解决方法(持续更新)

1. 前言

在一个完整的大数据处理系统中,除了hdfs+mapreduce+hive组成分析系统的核心之外,还需要数据采集、结果数据导出、任务调度等不可或缺的辅助系统,而这些辅助工具在hadoop生态体系中都有便捷的开源框架,如图所示:

2. Flume介绍

2.1. 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数据包等各种形式源数据,又可以将采集到的数据输出到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

2.2. 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:

  注意:Source 到 Channel 到 Sink之间传递数据的形式是Event事件;Event事件是一个数据流单元。

  a) Source:采集源,用于跟数据源对接,以获取数据

  b) Sink:下沉地,采集数据的传送目的,用于往下一级agent传递数据或者往最终存储系统传递数据

  c)Channel:angent内部的数据传输通道,用于从source将数据传递到sink

3. Flume采集系统结构图

3.1. 简单结构

  单个agent采集数据

3.2. 复杂结构

  多级agent之间串联

4. Flume的安装部署

4.1. 软件部署

 1 [yun@mini01 software]$ pwd
 2 /app/software
 3 [yun@mini01 software]$ tar xf apache-flume-1.8.0-bin.tar.gz
 4 [yun@mini01 software]$ mv apache-flume-1.8.0-bin /app/flume-1.8.0
 5 [yun@mini01 software]$ cd /app/
 6 [yun@mini01 ~]$ ln -s flume-1.8.0  flume   # 建立软连接
 7 [yun@mini01 ~]$ ll
 8 total 28
 9 lrwxrwxrwx  1 yun yun    11 Jul 25 21:54 flume -> flume-1.8.0
10 drwxrwxr-x  7 yun yun   187 Jul 25 21:53 flume-1.8.0
11 ………………

4.2. 环境变量

1 [root@mini01 profile.d]# pwd
2 /etc/profile.d
3 [root@mini01 profile.d]# cat flume.sh
4 export FLUME_HOME="/app/flume"
5 export PATH=$FLUME_HOME/bin:$PATH
6 [root@mini01 profile.d]# logout
7 [yun@mini01 ~]$ source /etc/profile  # 环境变量生效

5. 采集案例

5.1. 简单案例——从网络端口接收数据下沉到logger

配置文件

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini01 conf]$ ll
 4 total 20
 5 -rw-r--r-- 1 yun yun 1661 Sep 15  2017 flume-conf.properties.template
 6 -rw-r--r-- 1 yun yun 1455 Sep 15  2017 flume-env.ps1.template
 7 -rw-r--r-- 1 yun yun 1568 Sep 15  2017 flume-env.sh.template
 8 -rw-r--r-- 1 yun yun 3107 Sep 15  2017 log4j.properties
 9 -rw-rw-r-- 1 yun yun  741 Jul 25 22:12 netcat-logger.conf
10 [yun@mini01 conf]$ cat netcat-logger.conf 
11 # Name the components on this agent
12 a1.sources = r1
13 a1.sinks = k1
14 a1.channels = c1
15 
16 # Describe/configure the source
17 a1.sources.r1.type = netcat
18 a1.sources.r1.bind = localhost
19 a1.sources.r1.port = 44444
20 # bind = localhost 绑定的是本地端口
21 
22 # Describe the sink
23 a1.sinks.k1.type = logger
24 
25 # Use a channel which buffers events in memory
26 #下沉的时候是一批一批的, 下沉的时候是一个个eventChannel参数解释:
27 #capacity:默认该通道中最大的可以存储的event数量
28 #trasactionCapacity:每次最大可以从source中拿到或者送到sink中的event数量
29 a1.channels.c1.type = memory
30 a1.channels.c1.capacity = 1000
31 a1.channels.c1.transactionCapacity = 100
32 
33 # Bind the source and sink to the channel
34 a1.sources.r1.channels = c1
35 a1.sinks.k1.channel = c1

flume启动

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 # 其中--conf-file 指定的配置文件可以为相对路径也可以是绝对路径
 4 [yun@mini01 conf]$ flume-ng agent --conf conf --conf-file netcat-logger.conf --name a1 -Dflume.root.logger=INFO,console
 5 ………………
 6 18/07/25 22:19:23 INFO node.Application: Starting Channel c1
 7 18/07/25 22:19:23 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: CHANNEL, name: c1: Successfully registered new MBean.
 8 18/07/25 22:19:23 INFO instrumentation.MonitoredCounterGroup: Component type: CHANNEL, name: c1 started
 9 18/07/25 22:19:23 INFO node.Application: Starting Sink k1
10 18/07/25 22:19:23 INFO node.Application: Starting Source r1
11 18/07/25 22:19:23 INFO source.NetcatSource: Source starting
12 18/07/25 22:19:23 INFO source.NetcatSource: Created serverSocket:sun.nio.ch.ServerSocketChannelImpl[/127.0.0.1:44444]

source端的Telnet输入

 1 [yun@mini01 ~]$ telnet localhost 44444
 2 Trying ::1...
 3 telnet: connect to address ::1: Connection refused
 4 Trying 127.0.0.1...
 5 Connected to localhost.
 6 Escape character is '^]'.
 7 111
 8 OK
 9 222
10 OK
11 334334634geg
12 OK
13 gwegweg
14 OK
15 ^]
16 telnet> quit
17 Connection closed.

当在Telnet端输入,logger显示

1 18/07/25 22:20:09 INFO sink.LoggerSink: Event: { headers:{} body: 31 31 31 0D                                     111. }
2 18/07/25 22:20:10 INFO sink.LoggerSink: Event: { headers:{} body: 32 32 32 0D                                     222. }
3 18/07/25 22:20:13 INFO sink.LoggerSink: Event: { headers:{} body: 33 33 34 33 33 34 36 33 34 67 65 67 0D          334334634geg. }
4 18/07/25 22:20:14 INFO sink.LoggerSink: Event: { headers:{} body: 67 77 65 67 77 65 67 0D                         gwegweg. }

5.2. 监视文件夹——下沉到logger

配置文件

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini01 conf]$ ll
 4 total 20
 5 -rw-r--r-- 1 yun yun 1661 Sep 15  2017 flume-conf.properties.template
 6 -rw-r--r-- 1 yun yun 1455 Sep 15  2017 flume-env.ps1.template
 7 -rw-r--r-- 1 yun yun 1568 Sep 15  2017 flume-env.sh.template
 8 -rw-r--r-- 1 yun yun 3107 Sep 15  2017 log4j.properties
 9 -rw-rw-r-- 1 yun yun  741 Jul 25 22:12 netcat-logger.conf
10 -rw-rw-r-- 1 yun yun  597 Jul 25 22:29 spooldir-logger.conf
11 [yun@mini01 conf]$ cat spooldir-logger.conf 
12 # Name the components on this agent
13 a1.sources = r1
14 a1.sinks = k1
15 a1.channels = c1
16 
17 # Describe/configure the source
18 #监听目录,spoolDir指定目录, fileHeader要不要给文件加前缀名
19 a1.sources.r1.type = spooldir
20 a1.sources.r1.spoolDir = /app/software/flume
21 a1.sources.r1.fileHeader = true
22 
23 # Describe the sink
24 a1.sinks.k1.type = logger
25 
26 # Use a channel which buffers events in memory
27 a1.channels.c1.type = memory
28 a1.channels.c1.capacity = 1000
29 a1.channels.c1.transactionCapacity = 100
30 
31 # Bind the source and sink to the channel
32 a1.sources.r1.channels = c1
33 a1.sinks.k1.channel = c1

flume启动

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini01 conf]$ flume-ng agent --conf conf --conf-file spooldir-logger.conf --name a1 -Dflume.root.logger=INFO,console
 4 ………………
 5 18/07/25 23:01:04 INFO instrumentation.MonitoredCounterGroup: Component type: CHANNEL, name: c1 started
 6 18/07/25 23:01:04 INFO node.Application: Starting Sink k1
 7 18/07/25 23:01:04 INFO node.Application: Starting Source r1
 8 18/07/25 23:01:04 INFO source.SpoolDirectorySource: SpoolDirectorySource source starting with directory: /app/software/flume
 9 18/07/25 23:01:04 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: SOURCE, name: r1: Successfully registered new MBean.
10 18/07/25 23:01:04 INFO instrumentation.MonitoredCounterGroup: Component type: SOURCE, name: r1 started

往/app/software/flume目录加入文件

 1 # 原文件目录
 2 [yun@mini01 hive]$ pwd
 3 /app/software/hive
 4 [yun@mini01 hive]$ ll
 5 total 48
 6 -rw-rw-r-- 1 yun yun  71 Jul 12 21:53 t_sz01.dat
 7 -rw-rw-r-- 1 yun yun  71 Jul 12 21:53 t_sz01.dat2
 8 -rw-rw-r-- 1 yun yun  79 Jul 12 22:15 t_sz02_ext.dat
 9 -rw-rw-r-- 1 yun yun  52 Jul 12 23:09 t_sz03_20180711.dat1
10 -rw-rw-r-- 1 yun yun  52 Jul 12 23:09 t_sz03_20180711.dat2
11 -rw-rw-r-- 1 yun yun  52 Jul 12 23:09 t_sz03_20180712.dat1
12 -rw-rw-r-- 1 yun yun  52 Jul 12 23:09 t_sz03_20180712.dat2
13 -rw-rw-r-- 1 yun yun 753 Jul 14 10:36 t_sz05_buck.dat
14 -rw-rw-r-- 1 yun yun 507 Jul 14 10:07 t_sz05_buck.dat.bak
15 [yun@mini01 hive]$ cp -a t_access_times.dat t_sz01.dat t_sz01.dat2 ../flume/
16 [yun@mini01 hive]$ cp -a t_sz05_buck.dat t_sz05_buck.dat2 ../flume/
17 ############################################
18 # 对应的flume目录   注意文件名不能重复,否则flume会报错,也不能是一个目录
19 [yun@mini01 flume]$ pwd
20 /app/software/flume
21 [yun@mini01 flume]$ ll
22 total 20
23 -rw-rw-r-- 1 yun yun 288 Jul 18 14:20 t_access_times.dat.COMPLETED
24 -rw-rw-r-- 1 yun yun  71 Jul 12 21:53 t_sz01.dat2.COMPLETED
25 -rw-rw-r-- 1 yun yun  71 Jul 12 21:53 t_sz01.dat.COMPLETED
26 -rw-rw-r-- 1 yun yun 753 Jul 14 10:36 t_sz05_buck.dat
27 -rw-rw-r-- 1 yun yun 753 Jul 14 10:36 t_sz05_buck.dat.COMPLETED

有fileHeader配置

有fileHeader配置

5.3. 用tail命令获取数据,下沉到HDFS

配置文件

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini01 conf]$ ll
 4 total 28
 5 -rw-r--r-- 1 yun yun 1661 Sep 15  2017 flume-conf.properties.template
 6 -rw-r--r-- 1 yun yun 1455 Sep 15  2017 flume-env.ps1.template
 7 -rw-r--r-- 1 yun yun 1568 Sep 15  2017 flume-env.sh.template
 8 -rw-r--r-- 1 yun yun 3107 Sep 15  2017 log4j.properties
 9 -rw-rw-r-- 1 yun yun  741 Jul 25 22:12 netcat-logger.conf
10 -rw-rw-r-- 1 yun yun  593 Jul 25 22:30 spooldir-logger.conf
11 -rw-rw-r-- 1 yun yun 1275 Jul 25 23:29 tail-hdfs.conf
12 [yun@mini01 conf]$ cat tail-hdfs.conf 
13 # Name the components on this agent
14 a1.sources = r1
15 a1.sinks = k1
16 a1.channels = c1
17 
18 # Describe/configure the source
19 a1.sources.r1.type = exec
20 a1.sources.r1.command = tail -F /app/webservice/logs/access.log
21 a1.sources.r1.channels = c1
22 
23 # Describe the sink
24 a1.sinks.k1.type = hdfs
25 a1.sinks.k1.channel = c1
26 a1.sinks.k1.hdfs.path = /flume/events/%y-%m-%d/%H%M/
27 a1.sinks.k1.hdfs.filePrefix = events-
28 # 以下3项表示每隔10分钟切换目录存储
29 a1.sinks.k1.hdfs.round = true
30 a1.sinks.k1.hdfs.roundValue = 10
31 a1.sinks.k1.hdfs.roundUnit = minute
32 # 滚动当前文件前等待的秒数
33 a1.sinks.k1.hdfs.rollInterval = 30
34 # 文件大小以字节为单位触发滚动  
35 a1.sinks.k1.hdfs.rollSize = 1024
36 # 在滚动之前写入文件的事件数
37 a1.sinks.k1.hdfs.rollCount = 500
38 # 在它被刷新到HDFS之前写入文件的事件数量。100个事件为一个批次
39 a1.sinks.k1.hdfs.batchSize = 100
40 a1.sinks.k1.hdfs.useLocalTimeStamp = true
41 #生成的文件类型,默认是Sequencefile,可用DataStream,则为普通文本
42 a1.sinks.k1.hdfs.fileType = DataStream
43 
44 # Use a channel which buffers events in memory
45 a1.channels.c1.type = memory
46 a1.channels.c1.capacity = 1000
47 a1.channels.c1.transactionCapacity = 100
48 
49 # Bind the source and sink to the channel
50 a1.sources.r1.channels = c1
51 a1.sinks.k1.channel = c1

flume启动

1 [yun@mini01 conf]$ flume-ng agent -c conf -f tail-hdfs.conf -n a1

启动jar包打印日志

1 [yun@mini01 webservice]$ pwd
2 /app/webservice
3 [yun@mini01 webservice]$ java -jar testlog.jar &

可参见:Hadoop2.7.6_02_HDFS常用操作 ----- 3.3. web日志模拟

浏览器查看flume下沉的数据

5.4. 级联下沉到HDFS

由mini01 的flume发送数据到mini02的flume,然后由mini02的flume下沉到HDFS。

其中mini02的flume安装过程略。

配置文件mini01

 1 [yun@mini01 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini01 conf]$ ll
 4 total 32
 5 -rw-r--r-- 1 yun yun 1661 Sep 15  2017 flume-conf.properties.template
 6 -rw-r--r-- 1 yun yun 1455 Sep 15  2017 flume-env.ps1.template
 7 -rw-r--r-- 1 yun yun 1568 Sep 15  2017 flume-env.sh.template
 8 -rw-r--r-- 1 yun yun 3107 Sep 15  2017 log4j.properties
 9 -rw-rw-r-- 1 yun yun  741 Jul 25 22:12 netcat-logger.conf
10 -rw-rw-r-- 1 yun yun  593 Jul 25 22:30 spooldir-logger.conf
11 -rw-rw-r-- 1 yun yun  789 Jul 26 22:31 tail-avro-avro-logger.conf
12 -rw-rw-r-- 1 yun yun 1283 Jul 25 23:41 tail-hdfs.conf
13 [yun@mini01 conf]$ cat tail-avro-avro-logger.conf 
14 # Name the components on this agent
15 a1.sources = r1
16 a1.sinks = k1
17 a1.channels = c1
18 
19 # Describe/configure the source
20 a1.sources.r1.type = exec
21 a1.sources.r1.command = tail -F /app/webservice/logs/access.log
22 a1.sources.r1.channels = c1
23 
24 # Describe the sink
25 #绑定的不是本机, 是另外一台机器的服务地址, sink端的avro是一个发送端, avro的客户端, 往mini02这个机器上发
26 a1.sinks = k1
27 a1.sinks.k1.type = avro
28 a1.sinks.k1.channel = c1
29 a1.sinks.k1.hostname = mini02
30 a1.sinks.k1.port = 4141
31 a1.sinks.k1.batch-size = 2
32 
33 # Use a channel which buffers events in memory
34 a1.channels.c1.type = memory
35 a1.channels.c1.capacity = 1000
36 a1.channels.c1.transactionCapacity = 100
37 
38 # Bind the source and sink to the channel
39 a1.sources.r1.channels = c1
40 a1.sinks.k1.channel = c1

配置文件mini02

 1 [yun@mini02 conf]$ pwd
 2 /app/flume/conf
 3 [yun@mini02 conf]$ ll
 4 total 20
 5 -rw-rw-r-- 1 yun yun 1357 Jul 26 22:39 avro-hdfs.conf
 6 -rw-r--r-- 1 yun yun 1661 Sep 15  2017 flume-conf.properties.template
 7 -rw-r--r-- 1 yun yun 1455 Sep 15  2017 flume-env.ps1.template
 8 -rw-r--r-- 1 yun yun 1568 Sep 15  2017 flume-env.sh.template
 9 -rw-r--r-- 1 yun yun 3107 Sep 15  2017 log4j.properties
10 [yun@mini02 conf]$ cat avro-hdfs.conf 
11 # Name the components on this agent
12 a1.sources = r1
13 a1.sinks = k1
14 a1.channels = c1
15 
16 # Describe/configure the source
17 #source中的avro组件是接收者服务, 绑定本机
18 a1.sources.r1.type = avro
19 a1.sources.r1.channels = c1
20 a1.sources.r1.bind = 0.0.0.0
21 a1.sources.r1.port = 4141
22 
23 # Describe the sink
24 a1.sinks.k1.type = hdfs
25 a1.sinks.k1.channel = c1
26 a1.sinks.k1.hdfs.path = /flume/new-events/%y-%m-%d/%H%M/
27 a1.sinks.k1.hdfs.filePrefix = events-
28 # 以下3项表示每隔10分钟切换目录存储
29 a1.sinks.k1.hdfs.round = true
30 a1.sinks.k1.hdfs.roundValue = 10
31 a1.sinks.k1.hdfs.roundUnit = minute
32 # 滚动当前文件前等待的秒数
33 a1.sinks.k1.hdfs.rollInterval = 30
34 # 文件大小以字节为单位触发滚动
35 a1.sinks.k1.hdfs.rollSize = 204800
36 # 在滚动之前写入文件的事件数
37 a1.sinks.k1.hdfs.rollCount = 500
38 # 在它被刷新到HDFS之前写入文件的事件数量,每批次事件最大数
39 a1.sinks.k1.hdfs.batchSize = 100
40 a1.sinks.k1.hdfs.useLocalTimeStamp = true
41 #生成的文件类型,默认是Sequencefile,可用DataStream,则为普通文本
42 a1.sinks.k1.hdfs.fileType = DataStream
43 
44 # Use a channel which buffers events in memory
45 a1.channels.c1.type = memory
46 a1.channels.c1.capacity = 1000
47 a1.channels.c1.transactionCapacity = 100
48 
49 # Bind the source and sink to the channel
50 a1.sources.r1.channels = c1
51 a1.sinks.k1.channel = c1

启动flume

1 # 启动mini02的flume
2 
3 
4 # 启动mini01的flume

启动jar包打印日志

1 [yun@mini01 webservice]$ pwd
2 /app/webservice
3 [yun@mini01 webservice]$ java -jar testlog.jar &

可参见:Hadoop2.7.6_02_HDFS常用操作 ----- 3.3. web日志模拟

浏览器查看flume下沉的数据

6. 更多source和sink组件

Flume支持众多的source和sink类型,详细手册可参考官方文档