Flume部署和使用

最新推荐文章于 2025-03-17 18:10:22 发布

你拓哥

最新推荐文章于 2025-03-17 18:10:22 发布

阅读量420

点赞数

分类专栏： Flume

本文链接：https://blog.youkuaiyun.com/nituoge/article/details/104326210

版权

Flume 专栏收录该内容

0 篇文章

订阅专栏

本文详细介绍Flume的部署与使用，包括安装步骤、配置文件编写及多种Source、Channel、Sink类型的实例应用，如Netcat、Exec、Spooling Directory、Taildir、Avro等，适合初学者快速上手。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

Flume部署和使用

官方文档: http://flume.apache.org/
example: WebServer --> Agent[Source–>Channel–>Sink] --> HDFS

一.简介

Flume是一个分布式，可靠的的框架，它能从许多不同的数据源高效地收集、聚合和移动大量日志数据到一个集中的目的进行存储。Flume OG主要是0.9及以前的版本，Flume NG主要是1.X的版本。类似的框架还有Logstash。对应广义的Hadoop而言，Flume是比较常用的。Logstash更轻量级，主要配合ELK使用
我们使用Flume，真要需要开发的代码不多，从使用层面来讲就是写配置文件，掌握常见类型Source、Channel、Sink的配置，配置我们的Agent。这部分主要对应的是Flume User Guide: http://flume.apache.org/releases/content/1.9.0/FlumeDeveloperGuide.html。
需要自己写代码的部分,主要是如何基于Flume进行自定义开发，自定义Source、Sink、和Channel。
常用的Source类型如下(从哪里收集)
- avro (序列化)
- exec (命令行)
- spooling (目录)
- taildir (重要)
- kafka
常见的Channel类型(数据存在哪里)
- memory
- kafka
- file
常见Sink类型(数据输出到哪里)
- hdfs (Hadoop)
- logger (控制台)
- avro
- kafka

二.安装

下载

[ruoze@rzdata001 ~]$ cd software/
[ruoze@rzdata001 software]$ wget http://archive.cloudera.com/cdh5/cdh/5/flume-ng-1.6.0-cdh5.16.2.tar.gz

解压,创建软连接

[ruoze@rzdata001 software]$ tar -zxvf flume-ng-1.6.0-cdh5.16.2.tar.gz -C /home/app/
[ruoze@rzdata001 software]$ cd ~/app/
[ruoze@rzdata001 app]$ ln -s /home/ruoze/app/apache-flume-1.6.0-cdh5.16.2-bin /home/ruoze/app/flume

配置配置文件

[ruoze@rzdata001 ~]$ cd /home/ruoze/app/flume/conf
[ruoze@rzdata001 conf]$ cp flume-env.sh.template flume-env.sh
[ruoze@rzdata001 conf]$ cp flume-conf.properties.template flume-conf.properties
[ruoze@rzdata001 conf]$ vim flume-env.sh
export JAVA_HOME=/usr/java/jdk1.8.0_121

添加环境变量

[ruoze@rzdata001 ~]$ vim .bash_profile
# Flume env
export FLUME_HOME=/home/ruoze/app/flume
export PATH=$FLUME_HOME/bin:$PATH

flume版本

[ruoze@rzdata001 bin]$ 
[ruoze@rzdata001 bin]$ flume-ng version
Flume 1.6.0-cdh5.16.2
Source code repository: https://git-wip-us.apache.org/repos/asf/flume.git
Revision: df92badde3691ee3eb6074a177f0e96682345381
Compiled by jenkins on Mon Jun  3 03:49:33 PDT 2019
From source with checksum 9336bfa3ff8cfb5e20cd9d700135a2c1
[ruoze@rzdata001 bin]$

三.任务配置

Flume agent配置是一个配置在本地的配置文件，是一个遵循java配置文件格式的text文件。
在同一个配置文件中可以配置指定一个或多个agents，同时配置上每个source、sink和channel数据流串联的方式。
目标: 掌握根据官网文档进行配置的方法。
各个类型的source、channel和sink的配置说明中，粗体是必须设置的参数。

例1: 1_netcat_memory_logger.conf

写配置文件：

# conf file name: 1_netcat_memory_logger.conf
# Source type : netcat source
# Channel type: memory channel
# Sink type   : logger sink

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

# Describe/configure the source <== define Source
a1.sources.r1.type = netcat
a1.sources.r1.bind = 0.0.0.0
a1.sources.r1.port = 44444

# Use a channel which buffers events in memory  <== define Channel
a1.channels.c1.type = memory

# Describe the sink <== define Sink
a1.sinks.k1.type = logger

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

启动agent:

$FLUME_HOME/bin/flume-ng agent \
--name a1 \
--conf $FLUME_HOME/conf \
--conf-file $FLUME_HOME/myconf/1_netcat_memory_logger.conf \
-Dflume.root.logger=INFO,console

jps查看新增的服务：

[ruoze@rzdata001 ~]$ jps
7040 SecondaryNameNode
31123 Jps
20724 NameNode
21144 ResourceManager
7897 RunJar
30939 Application          <== 新增agent的信息
20860 DataNode
21260 NodeManager
19215 StandaloneSessionClusterEntrypoint
[ruoze@rzdata001 ~]$

查看服务启动的详细信息：

[ruoze@rzdata001 ~]$ ps -ef | grep 30939 | grep -v grep 
ruoze    30939 31334  0 20:49 pts/0    00:00:01 /usr/java/jdk1.8.0_121/bin/java -Xmx20m -Dflume.root.logger=INFO,console -cp /home/ruoze/app/flume/conf:/home/ruoze/app/flume/lib/*:/home/ruoze/app/hadoop/etc/hadoop:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/common/lib/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/common/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/hdfs:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/hdfs/lib/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/hdfs/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/yarn/lib/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/yarn/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/mapreduce/lib/*:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/share/hadoop/mapreduce/*:/home/ruoze/app/hadoop/contrib/capacity-scheduler/*.jar:/home/ruoze/app/hive/lib/* -Djava.library.path=:/home/ruoze/app/hadoop-2.6.0-cdh5.16.2/lib/native org.apache.flume.node.Application --name a1 --conf-file /home/ruoze/app/flume/myconf/1_netcat_memory_logger.conf
[ruoze@rzdata001 ~]$

agent启动后，向44444端口发送数据。注意，一定要先启动agent。

[ruoze@rzdata001 ~]$ telnet localhost 44444
Trying ::1...
Connected to localhost.
Escape character is '^]'.
123456789     
OK
abcdefg           
OK
1234567890987654321
OK

总共发送了3条，前两条控制台上完整打印，最后一条控制台上只打印了1234567890987654，原因是logger sink的maxBytesToLog参数为设置，默认截断长度是16字节
思考控制台的输出：
Event: { headers:{} body: 31 32 33 34 35 36 37 38 39 0D 123456789. }
Event：Flume数据传输的基本单元，其构成部分中。
header是可选的，其中的内容已k-v的方式存在；
body是字节数组.

例2: 2_exec_memory_hdfs.conf

写配置文件：

# conf file name: 2_exec_memory_hdfs.conf
# Source type : exec source
# Channel type: memory channel
# Sink type   : hdfs sink

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

# Describe/configure the source <== define Source
a1.sources.r1.type = exec
a1.sources.r1.command = tail -F /home/ruoze/data/flume/data.log
a1.sources.r1.shell = /bin/sh -c

# Use a channel which buffers events in memory  <== define Channel
a1.channels.c1.type = memory

# Describe the sink <== define Sink
a1.sinks.k1.type = hdfs
a1.sinks.k1.hdfs.path = hdfs://rzdata001:9000/ruozedata/flume/tail/events/%y-%m-%d/%H%M
a1.sinks.k1.hdfs.filePrefix = events
a1.sinks.k1.hdfs.fileSuffix = .txt
a1.sinks.k1.hdfs.round = true
a1.sinks.k1.hdfs.roundValue = 10
a1.sinks.k1.hdfs.roundUnit = minute
a1.sinks.k1.hdfs.rollInterval = 600
a1.sinks.k1.hdfs.rollSize = 1024000
a1.sinks.k1.hdfs.rollCount = 2000
a1.sinks.k1.hdfs.batchSize = 10
a1.sinks.k1.hdfs.fileType = DataStream
a1.sinks.k1.hdfs.writeFormat = Text
a1.sinks.k1.hdfs.useLocalTimeStamp = true

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

启动agent:

$FLUME_HOME/bin/flume-ng agent \
--name a1 \
--conf $FLUME_HOME/conf \
--conf-file $FLUME_HOME/myconf/2_exec_memory_hdfs.conf \
-Dflume.root.logger=INFO,console

验证数据

[ruoze@rzdata001 ~]$ hadoop fs -ls /ruozedata/flume/tail/events/*/*
20/02/13 22:33:27 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 5 items
-rw-r--r--   1 ruoze supergroup     215039 2020-02-13 22:26 /ruozedata/flume/tail/events/20-02-13/2220/events.1581603956479.txt
-rw-r--r--   1 ruoze supergroup     216644 2020-02-13 22:26 /ruozedata/flume/tail/events/20-02-13/2220/events.1581603956480.txt
-rw-r--r--   1 ruoze supergroup     217824 2020-02-13 22:26 /ruozedata/flume/tail/events/20-02-13/2220/events.1581603956481.txt
-rw-r--r--   1 ruoze supergroup     215789 2020-02-13 22:26 /ruozedata/flume/tail/events/20-02-13/2220/events.1581603956482.txt
-rw-r--r--   1 ruoze supergroup      75874 2020-02-13 22:31 /ruozedata/flume/tail/events/20-02-13/2220/events.1581603956483.txt
Found 1 items
-rw-r--r--   1 ruoze supergroup     104360 2020-02-13 22:31 /ruozedata/flume/tail/events/20-02-13/2230/events.1581604278540.txt.tmp
[ruoze@rzdata001 ~]$

需要注意的是，缓冲区的大小和roll的三个参数需要根据实际业务调整。

例3: 3_spooldir_memory_hdfs.conf

写配置文件：

# conf file name: 3_spooldir_memory_hdfs.conf
# Source type : spooling directory source
# Channel type: memory channel
# Sink type   : hdfs sink

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

# Describe/configure the source <== define Source
a1.sources.r1.type = spooldir
a1.sources.r1.spoolDir = /home/ruoze/data/flume/spool
a1.sources.r1.fileSuffix = .COMPLETED

# Use a channel which buffers events in memory  <== define Channel
a1.channels.c1.type = memory

# Describe the sink <== define Sink
a1.sinks.k1.type = hdfs
a1.sinks.k1.hdfs.path = hdfs://rzdata001:9000/ruozedata/flume/spool/%Y%m%d%H%M
a1.sinks.k1.hdfs.filePrefix = events
a1.sinks.k1.hdfs.fileSuffix = .txt
a1.sinks.k1.hdfs.round = true
a1.sinks.k1.hdfs.roundValue = 10
a1.sinks.k1.hdfs.roundUnit = minute
a1.sinks.k1.hdfs.rollInterval = 600
a1.sinks.k1.hdfs.rollSize = 1024000
a1.sinks.k1.hdfs.rollCount = 2000
a1.sinks.k1.hdfs.batchSize = 10
a1.sinks.k1.hdfs.fileType = DataStream
a1.sinks.k1.hdfs.writeFormat = Text
a1.sinks.k1.hdfs.useLocalTimeStamp = true

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

启动agent:

$FLUME_HOME/bin/flume-ng agent \
--name a1 \
--conf $FLUME_HOME/conf \
--conf-file $FLUME_HOME/myconf/3_spooldir_memory_hdfs.conf \
-Dflume.root.logger=INFO,console

验证数据

[ruoze@rzdata001 ~]$ hadoop fs -ls /ruozedata/flume/spool/*/*
20/02/14 15:58:54 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
-rw-r--r--   1 ruoze supergroup     224236 2020-02-14 13:39 /ruozedata/flume/spool/202002141330/events.1581658686779.txt
-rw-r--r--   1 ruoze supergroup     216625 2020-02-14 13:39 /ruozedata/flume/spool/202002141330/events.1581658686780.txt
-rw-r--r--   1 ruoze supergroup     136405 2020-02-14 13:49 /ruozedata/flume/spool/202002141330/events.1581658686781.txt
-rw-r--r--   1 ruoze supergroup     139660 2020-02-14 13:40 /ruozedata/flume/spool/202002141340/events.1581658800114.txt
-rw-r--r--   1 ruoze supergroup     216048 2020-02-14 13:40 /ruozedata/flume/spool/202002141340/events.1581658800115.txt
-rw-r--r--   1 ruoze supergroup     216894 2020-02-14 13:41 /ruozedata/flume/spool/202002141340/events.1581658800132.txt
-rw-r--r--   1 ruoze supergroup     214981 2020-02-14 13:41 /ruozedata/flume/spool/202002141340/events.1581658800133.txt
-rw-r--r--   1 ruoze supergroup      56282 2020-02-14 13:51 /ruozedata/flume/spool/202002141340/events.1581658800134.txt
[ruoze@rzdata001 ~]$ s

例4: 4_taildir_memory_hdfs.conf

taildir source 在数据采集时会把OFFSET偏移量记录下来，数据不会丢失。
这个方式不仅支持文件，而且支持目录
写配置文件：

# conf file name: 4_taildir_memory_hdfs.conf
# Source type : TAILDIR source
# Channel type: memory channel
# Sink type   : hdfs sink

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

# Describe/configure the source <== define Source
a1.sources.r1.type = TAILDIR
a1.sources.r1.positionFile = /home/ruoze/data/flume/taildir/taildir_position.json
a1.sources.r1.filegroups = f1 f2
a1.sources.r1.filegroups.f1 = /home/ruoze/data/flume/taildir/dir1/hive.log
a1.sources.r1.headers.f1.headerKey1 = value1
a1.sources.r1.filegroups.f2 = /home/ruoze/data/flume/taildir/dir2/.*log.*
a1.sources.r1.headers.f2.headerKey1 = value2
a1.sources.r1.headers.f2.headerKey2 = value2-2

# Use a channel which buffers events in memory  <== define Channel
a1.channels.c1.type = memory

# Describe the sink <== define Sink
a1.sinks.k1.type = hdfs
a1.sinks.k1.hdfs.path = hdfs://rzdata001:9000/ruozedata/flume/taildir/%Y%m%d%H
a1.sinks.k1.hdfs.filePrefix = hive
a1.sinks.k1.hdfs.fileSuffix = .log
a1.sinks.k1.hdfs.round = true
a1.sinks.k1.hdfs.roundUnit = minute
a1.sinks.k1.hdfs.roundValue = 1
a1.sinks.k1.hdfs.rollInterval = 10
a1.sinks.k1.hdfs.rollSize = 102400
a1.sinks.k1.hdfs.rollCount = 500
a1.sinks.k1.hdfs.batchSize = 10
a1.sinks.k1.hdfs.fileType = DataStream
a1.sinks.k1.hdfs.writeFormat = Text
a1.sinks.k1.hdfs.useLocalTimeStamp = true

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

启动agent:

$FLUME_HOME/bin/flume-ng agent \
--name a1 \
--conf $FLUME_HOME/conf \
--conf-file $FLUME_HOME/myconf/4_taildir_memory_hdfs.conf \
-Dflume.root.logger=INFO,console

验证taildir的断点还原功能

# 1.准备数据
[ruoze@rzdata001 taildir]$ pwd
/home/ruoze/data/flume/taildir
[ruoze@rzdata001 taildir]$ 
[ruoze@rzdata001 taildir]$ ll ./*/*
-rw-rw-r-- 1 ruoze ruoze 200000 Feb 14 23:04 ./dir1/hive.log
-rw-rw-r-- 1 ruoze ruoze 120000 Feb 14 22:54 ./dir2/hive.log.1
-rw-rw-r-- 1 ruoze ruoze  20000 Feb 14 23:00 ./dir2/hive.txt
-rw-rw-r-- 1 ruoze ruoze  70000 Feb 14 22:57 ./dir9/hive.log.2
[ruoze@rzdata001 taildir]$ 

# 2.开启4个窗口，验证：
# Step1: 第1个窗口启动flume进程.
# Step2: 第2个窗口操作本地数据.
# Step3: 第3个窗口查看HDFS上sink的数据: hadoop fs -ls /ruozedata/flume/taildir/*/* 
# Step4: 第4个窗口监控taildir_position.json: tail -F /home/ruoze/data/flume/taildir/taildir_position.json 
# Step5: 查看taildir_position.json 中只有两个文件被读取，hive.txt由于匹配规则，未被采集，核查 pos值和文件大小对应
# Step6：窗口1关闭进程，在窗口2中 拷贝 ./dir9/hive.log.2 到 ./dir2 ； cat ./dir2/hive.txt >> ./dir1/hive.log
# Step7：第一个窗口启动Flume进程，查看taildir_position.json变化情况：hive.log 有对应变化，新增 hive.log.2的pos信息 。

# 其他 flume服务运行的时候，删除taildir_position.json后又会重新生成

例5: 5_avro_memory_hdfs.conf

http://flume.apache.org/releases/content/1.9.0/FlumeUserGuide.html#log4j-appender

写配置文件：

# conf file name: 5_avro_memory_hdfs.conf
# Source type : avro source
# Channel type: memory channel
# Sink type   : hdfs sink

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

# Describe/configure the source <== define Source
a1.sources.r1.type = avro
a1.sources.r1.bind = rzdata001
a1.sources.r1.port = 44445

# Use a channel which buffers events in memory  <== define Channel
a1.channels.c1.type = memory

# Describe the sink <== define Sink
a1.sinks.k1.type = hdfs
a1.sinks.k1.hdfs.path = hdfs://rzdata001:9000/ruozedata/flume/avro/%Y%m%d%H
a1.sinks.k1.hdfs.filePrefix = hive
a1.sinks.k1.hdfs.fileSuffix = .log
a1.sinks.k1.hdfs.round = true
a1.sinks.k1.hdfs.roundUnit = minute
a1.sinks.k1.hdfs.roundValue = 1
a1.sinks.k1.hdfs.rollInterval = 10
a1.sinks.k1.hdfs.rollSize = 102400
a1.sinks.k1.hdfs.rollCount = 500
a1.sinks.k1.hdfs.batchSize = 10
a1.sinks.k1.hdfs.fileType = DataStream
a1.sinks.k1.hdfs.writeFormat = Text
a1.sinks.k1.hdfs.useLocalTimeStamp = true

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

启动agent:

$FLUME_HOME/bin/flume-ng agent \
--name a1 \
--conf $FLUME_HOME/conf \
--conf-file $FLUME_HOME/myconf/5_avro_memory_hdfs.conf \
-Dflume.root.logger=INFO,console

Java App 使用往Avro对应端口发送日志：

# 1)添加 maven 依赖
    <dependency>
      <groupId>org.apache.flume.flume-ng-clients</groupId>
      <artifactId>flume-ng-log4jappender</artifactId>
      <version>1.6.0-cdh5.16.2</version>
    </dependency>

# 2)log4j配置文件配置
log4j.rootLogger = INFO,Console,File,Flume
#### 日志输出到Avro Source,后续在主机上使用flume avro source 采集 ###
log4j.appender.Flume=org.apache.flume.clients.log4jappender.Log4jAppender
log4j.appender.Flume.Hostname=rzdata001
log4j.appender.Flume.Port=44445
log4j.appender.Flume.UnsafeMode=true
log4j.appender.Flume.layout=org.apache.log4j.PatternLayout
log4j.appender.Flume.layout.ConversionPattern=%d{yyyy/MM/dd HH:mm:ss.SSS}|%t|[%m]

# 3)代码产生日志
public class SendFlumeAvro {
    private static final Logger logger = LogManager.getLogger(SendFlumeAvro.class.getName());
    public static void main(String[] args) {
        logger.trace("这是一条 trace 信息");
        logger.info("这是一条 info 信息1");
        logger.error("这是一条 error 信息1");
        logger.info("这是一条 info 信息4");
        logger.error("这是一条 error 信息2");
        logger.info("这是一条 info 信息6");
        logger.error("这是一条 error 信息3");
        System.out.println("exit");
    }
}

# 4)运行程序发送日志

# 5)查看hdfs上sink落地的数据
[ruoze@rzdata001 ~]$ hadoop fs -text /ruozedata/flume/avro/2020021512/hive.1581740388601.log
20/02/15 12:27:00 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
2020/02/15 12:19:46.987|main|[这是一条 info 信息1]
2020/02/15 12:19:47.087|main|[这是一条 error 信息1]
2020/02/15 12:19:47.100|main|[这是一条 info 信息4]
2020/02/15 12:19:47.125|main|[这是一条 error 信息2]
2020/02/15 12:19:47.137|main|[这是一条 info 信息6]
2020/02/15 12:19:47.165|main|[这是一条 error 信息3]
[ruoze@rzdata001 ~]$