Flume是一个分布式、可靠、和高可用的海量日志聚合的系统，支持在系统中定制各类数据发送方，用于收集数据；
同时，Flume提供对数据进行简单处理，并写到各种数据接受方（可定制）的能力。

Flume是一个专门设计用来从大量的源，推送数据到Hadoop生态系统中各种各样存储系统中去的，例如HDFS和HBase。

Guide: http://flume.apache.org/FlumeUserGuide.html

体系架构

Flume的数据流由事件(Event)贯穿始终。事件是Flume的基本数据单位，它携带日志数据(字节数组形式)并且携带有头信息，这些Event由Agent外部的Source生成，当Source捕获事件后会进行特定的格式化，然后Source会把事件推入(单个或多个)Channel中。你可以把Channel看作是一个缓冲区，它将保存事件直到Sink处理完该事件。Sink负责持久化日志或者把事件推向另一个Source。

Flume以Flume Agent为最小的独立运行单位。一个Agent就是一个JVM。单agent由Source、Sink和Channel三大组件构成。一个Flume Agent可以连接一个或者多个其他的Flume Agent；一个Flume Agent也可以从一个或者多个Flume Agent接收数据。

注意：在Flume管道中如果有意想不到的错误、超时并进行了重试，Flume会产生重复的数据最终被被写入，后续需要处理这些冗余的数据。

具体可以参考文章：Flume教程(一) Flume入门教程

组件

Source：source是从一些其他产生数据的应用中接收数据的活跃组件。Source可以监听一个或者多个网络端口，用于接收数据或者可以从本地文件系统读取数据。每个Source必须至少连接一个Channel。基于一些标准，一个Source可以写入几个Channel，复制事件到所有或者某些Channel。

Source可以通过处理器 - 拦截器 - 选择器路由写入多个Channel。

Channel：channel的行为像队列，Source写入到channel，Sink从Channel中读取。多个Source可以安全地写入到相同的Channel，并且多个Sink可以从相同的Channel进行读取。

可是一个Sink只能从一个Channel读取。如果多个Sink从相同的Channel读取，它可以保证只有一个Sink将会从Channel读取一个指定特定的事件。

Flume自带两类Channel：Memory Channel和File Channel。Memory Channel的数据会在JVM或者机器重启后丢失；File Channel不会。

Sink: sink连续轮询各自的Channel来读取和删除事件。

拦截器：每次Source将数据写入Channel，它是通过委派该任务到其Channel处理器来完成，然后Channel处理器将这些事件传到一个或者多个Source配置的拦截器中。

拦截器是一段代码，基于某些标准，如正则表达式，拦截器可以用来删除事件，为事件添加新报头或者移除现有的报头等。每个Source可以配置成使用多个拦截器，按照配置中定义的顺序被调用，将拦截器的结果传递给链的下一个单元。一旦拦截器处理完事件，拦截器链返回的事件列表传递到Channel列表，即通过Channel选择器为每个事件选择的Channel。

组件

功能

Agent

使用JVM运行Flume。每台机器运行一个agent，但是可以在一个agent中包含多个sources和sinks。

Client

生产数据，运行在一个独立的线程。

Source

从Client收集数据，传递给Channel。

Sink

从Channel收集数据，运行在一个独立线程。

Channel

连接sources和sinks，这个有点像一个队列。

Events

可以是日志记录、avro对象等。

配置文件

Flume Agent使用纯文本配置文件来配置。Flume配置使用属性文件格式，仅仅是用换行符分隔的键值对的纯文本文件，如：key1 = value1；当有多个的时候：agent.sources = r1 r2

参考 flume配置介绍

1. 从file source 到 file sink的配置文件

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1

agent.sources.r1.interceptors = i1

agent.sources.r1.interceptors.i1.type = Inteceptor.DemoInterceptor$Builder

# ========= Describe the source =============

agent.sources.r1.type = spooldir

agent.sources.r1.spoolDir = /home/lintong/桌面/data/input

# ========= Describe the channel =============

# Use a channel which buffers events in memory

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

# ========= Describe the sink =============

agent.sinks.s1.type = file_roll

agent.sinks.s1.sink.directory = /home/lintong/桌面/data/output

agent.sinks.s1.sink.rollInterval = 0

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel = c1

2. 从kafka source 到 file sink的配置文件，kafka使用zookeeper，但是建议使用bootstrap-server

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1

agent.sources.r1.interceptors = i1

agent.sources.r1.interceptors.i1.type = Inteceptor.DemoInterceptor$Builder

# ========= Describe the source =============

agent.sources.r1.type=org.apache.flume.source.kafka.KafkaSource

agent.sources.r1.zookeeperConnect=127.0.0.1:2181

agent.sources.r1.topic=test #不能写成topics

#agent.sources.kafkaSource.groupId=flume

agent.sources.kafkaSource.kafka.consumer.timeout.ms=100  

# ========= Describe the channel =============

# Use a channel which buffers events in memory

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

# ========= Describe the sink =============

agent.sinks.s1.type = file_roll

agent.sinks.s1.sink.directory = /home/lintong/桌面/data/output

agent.sinks.s1.sink.rollInterval = 0

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel = c1

3.kafka source到kafka sink的配置文件

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1 s2

agent.sources.r1.interceptors = i1

agent.sources.r1.interceptors.i1.type = com.XXX.interceptor.XXXInterceptor$Builder

# ========= Describe the source =============

agent.sources.r1.type = org.apache.flume.source.kafka.KafkaSource

agent.sources.r1.channels = c1

agent.sources.r1.zookeeperConnect = localhost:2181

agent.sources.r1.topic = input

# ========= Describe the channel =============

# Use a channel which buffers events in memory

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

# ========= Describe the sink =============

agent.sinks.s1.type = org.apache.flume.sink.kafka.KafkaSink

agent.sinks.s1.topic = test

agent.sinks.s1.brokerList = localhost:9092

# 避免死循环

agent.sinks.s1.allowTopicOverride = false

agent.sinks.s2.type = file_roll

agent.sinks.s2.sink.directory = /home/lintong/桌面/data/output

agent.sinks.s2.sink.rollInterval = 0

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel = c1

#agent.sinks.s2.channel = c1

4.file source到hbase sink的配置文件

从文件读取实时消息，不做处理直接存储到Hbase

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1

# ========= Describe the source =============

agent.sources.r1.type = exec

agent.sources.r1.command = tail -f /home/lintong/桌面/test.log

agent.sources.r1.checkperiodic = 50

# ========= Describe the sink =============

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

# agent.channels.file-channel.type = file

# agent.channels.file-channel.checkpointDir = /data/flume-hbase-test/checkpoint

# agent.channels.file-channel.dataDirs = /data/flume-hbase-test/data

# ========= Describe the sink =============

agent.sinks.s1.type = org.apache.flume.sink.hbase.HBaseSink

agent.sinks.s1.zookeeperQuorum=master:2183

#HBase表名

agent.sinks.s1.table=mikeal-hbase-table

#HBase表的列族名称

agent.sinks.s1.columnFamily=familyclom1

agent.sinks.s1.serializer = org.apache.flume.sink.hbase.SimpleHbaseEventSerializer

#HBase表的列族下的某个列名称

agent.sinks.s1.serializer.payloadColumn=cloumn-1

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel=c1

5.source是http，sink是kafka

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1 s2

# ========= Describe the source =============

agent.sources.r1.type=http

agent.sources.r1.bind=localhost

agent.sources.r1.port=50000

agent.sources.r1.channels=c1

# ========= Describe the channel =============

# Use a channel which buffers events in memory

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

# ========= Describe the sink =============

agent.sinks.s1.type = org.apache.flume.sink.kafka.KafkaSink

agent.sinks.s1.topic = test_topic

agent.sinks.s1.brokerList = master:9092

# 避免死循环

agent.sinks.s1.allowTopicOverride = false

agent.sinks.s2.type = file_roll

agent.sinks.s2.sink.directory = /home/lintong/桌面/data/output

agent.sinks.s2.sink.rollInterval = 0

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel = c1

#agent.sinks.s2.channel = c1

如果在启动flume的时候遇到

java.lang.NoClassDefFoundError: org/apache/hadoop/hbase/***

解决方案，在 ~/software/apache/hadoop-2.9.1/etc/hadoop/hadoop-env.sh 中添加

HADOOP_CLASSPATH=/home/lintong/software/apache/hbase-1.2.6/lib/*

5.kafka source到hdfs sink的配置文件

# 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.

# The configuration file needs to define the sources,

# the channels and the sinks.

# Sources, channels and sinks are defined per agent,

# in this case called 'agent'

# ========= Name the components on this agent =========

agent.sources = r1

agent.channels = c1

agent.sinks = s1

agent.sources.r1.interceptors = i1

agent.sources.r1.interceptors.i1.type = Util.HdfsInterceptor$Builder

# ========= Describe the source =============

agent.sources.r1.type = org.apache.flume.source.kafka.KafkaSource

agent.sources.r1.channels = c1

agent.sources.r1.zookeeperConnect = localhost:2181

agent.sources.r1.topic = topicB

#agent.sources.r1.kafka.consumer.max.partition.fetch.bytes = 409600000

# ========= Describe the channel =============

# Use a channel which buffers events in memory

agent.channels.c1.type = memory

agent.channels.c1.capacity = 100000

agent.channels.c1.transactionCapacity = 1000

#agent.channels.c1.keep-alive = 60

# ========= Describe the sink =============

agent.sinks.s1.type = hdfs

agent.sinks.s1.hdfs.path = /user/lintong/logs/nsh/json/%{filepath}/ds=%{ds}

agent.sinks.s1.hdfs.filePrefix = test

agent.sinks.s1.hdfs.fileSuffix = .log

agent.sinks.s1.hdfs.fileType = DataStream

agent.sinks.s1.hdfs.useLocalTimeStamp = true

agent.sinks.s1.hdfs.writeFormat = Text

agent.sinks.s1.hdfs.rollCount = 0

agent.sinks.s1.hdfs.rollSize = 10240

agent.sinks.s1.hdfs.rollInterval = 600

agent.sinks.s1.hdfs.batchSize = 500

agent.sinks.s1.hdfs.threadsPoolSize = 10

agent.sinks.s1.hdfs.idleTimeout = 0

agent.sinks.s1.hdfs.minBlockReplicas = 1

agent.sinks.s1.channel = fileChannel

# ========= Bind the source and sink to the channel =============

agent.sources.r1.channels = c1

agent.sinks.s1.channel = c1

hdfs sink的配置参数参考：Flume中的HDFS Sink配置参数说明

因为写HDFS的速度很慢，当数据量大的时候会出现一下问题

org.apache.flume.ChannelException: Take list for MemoryTransaction, capacity 1000 full, consider committing more frequently, increasing capacity, or increasing thread count

可以将内存channel改成file channel或者改成kafka channel

当换成kafka channel的时候，数据量大的时候，依然会问题

16:07:48.615 ERROR org.apache.kafka.clients.consumer.internals.ConsumerCoordinator:550 - Error ILLEGAL_GENERATION occurred while committing offsets for group flume

16:07:48.617 ERROR org.apache.flume.source.kafka.KafkaSource:317 - KafkaSource EXCEPTION, {}

org.apache.kafka.clients.consumer.CommitFailedException: Commit cannot be completed due to group rebalance

	at org.apache.kafka.clients.consumer.internals.ConsumerCoordinator$OffsetCommitResponseHandler.handle(ConsumerCoordinator.java:552)

或者

ERROR org.apache.kafka.clients.consumer.internals.ConsumerCoordinator:550 - Error UNKNOWN_MEMBER_ID occurred while committing offsets for group flume

参考：flume1.7使用KafkaSource采集大量数据

Flume官方使用kafka channel的Demo

修改增大以下两个参数

agent.sources.r1.kafka.consumer.max.partition.fetch.bytes = 409600000

agent.sources.r1.kafka.consumer.timeout.ms = 100

kafka channel 爆了

ERROR org.apache.kafka.clients.consumer.internals.ConsumerCoordinator:550 - Error UNKNOWN_MEMBER_ID occurred while committing offsets for group flume

添加参数

agent.channels.c1.kafka.consumer.session.timeout.ms=100000

agent.channels.c1.kafka.consumer.request.timeout.ms=110000

agent.channels.c1.kafka.consumer.fetch.max.wait.ms=1000

命令

启动

bin/flume-ng agent -c conf -f conf/flume-conf.properties -n agent -Dflume.root.logger=INFO,console

flume学习笔记——安装和使用的更多相关文章

  1. SystemTap 学习笔记 - 安装篇

    https://segmentfault.com/a/1190000000671438 在安装前,需要知道下自己的系统环境,我的环境如下: uname -r 2.6.18-308.el5 Linux ...

  2. alfs学习笔记-安装和使用blfs工具

    我的邮箱地址:zytrenren@163.com欢迎大家交流学习纠错! 一名linux爱好者,记录构建Beyond Linux From Scratch的过程 经博客园-骏马金龙前辈介绍,开始接触学习 ...

  3. Flume 学习笔记之 Flume NG概述及单节点安装

    Flume NG概述: Flume NG是一个分布式,高可用,可靠的系统,它能将不同的海量数据收集,移动并存储到一个数据存储系统中.轻量,配置简单,适用于各种日志收集,并支持 Failover和负载均 ...

  4. cocos2d-js 学习笔记 --安装调试（1）

    目前从国内来看,比较知名的,适合前端程序员使用的游戏框架,基本是 egret 和cocos2d-js;刚好两个框架都学习了下, egret的是最近兴起的一个框架主要使用TypeScript.js语言做 ...

  5. [TenserFlow学习笔记]——安装

    最近人工智能.深度学习.机器学习等词汇很是热闹,所以想进一步学习一下.不一定吃这口饭,但多了解一下没有坏处.接下来将学习到的一些知识点做一下记录. 1.安装环境 在VMWare虚拟机中安装最新版本的U ...

  6. elasticsearch学习笔记——安装，初步使用

    前言 久仰elasticsearch大名,近年来,fackbook,baidu等大型网站的搜索功能均开始采用elasticsearch,足见其在处理大数据和高并发搜索中的卓越性能.不少其他网站也开始将 ...

  7. flume学习笔记

    #################################################################################################### ...

  8. Python学习笔记——安装

    最近打算使用下GAE,便准备学习一下python.我对python是一窍不通,因此这里将我的学习历程记录下来,方便后续复习. 安装python: 可以从如下地址:http://www.python.o ...

  9. Nodejs 学习笔记 --- 安装与环境配置

    一.安装Node.js步骤 1.下载对应自己系统对应的 Node.js 版本,地址:https://nodejs.org/zh-cn/      2.选安装目录进行安装      3.环境配置    ...

随机推荐

  1. egret.Shape渲染集合图形

    代码: class Main extends egret.DisplayObjectContainer { public constructor() { super(); this.addEventL ...

  2. python之面向对象的程序设计

    一.什么是面向对象的程序设计 面向过程的程序设计的核心是过程,过程即解决问题的步骤,面向过程的设计就好比精心设计好一条流水线,考虑周全什么时候处理什么东西. 优点是:极大的降低了程序的复杂度 缺点是: ...

  3. mvc中使用Pagination，对其进行再封装

    对其进行再次封装: (function($) { $["fn"]["easyPaging"] = function(o) { if (!o.pageSelect ...

  4. 论YUV422(YUYV)与YUV420相互转换

    Example 2.13. V4L2_PIX_FMT_YUYV 4 × 4 pixelimage start + 0: Y'00 Cb00 Y'01 Cr00 Y'02 Cb01 Y'03 Cr01 ...

  5. beego orm 忽略字段

    忽略字段 设置 - 即可忽略 struct 中的字段 type User struct { ... AnyField string `orm:"-"` ... } beego or ...

  6. [Web 前端] qs.parse()、qs.stringify()使用方法

    cp from : https://blog.csdn.net/suwu150/article/details/78333452 qs是一个npm仓库所管理的包,可通过npm install qs命令 ...

  7. 毕业论文Word格式技巧

    word利用多级列表功能实现章节标题自动编号 坑:在选择级别1后,直接在当前对话框直接继续设置级别2:然后选中级别2,设置相应的格式,主要在将级别链接到样式处选择为标题2,然后要在库中显示的级别选择级 ...

  8. Git远程仓库地址变更本地如何修改

    以项目test为例: 老地址:http://192.168.1.1:9797/john/test.git 新地址:http://git.xxx.xxx/john/test.git 远程仓库名称: or ...

  9. 用R语言分析与预測员工离职

    版权声明:本文为博主原创文章.未经博主同意不得转载. https://blog.csdn.net/kMD8d5R/article/details/83542978 https://mmbiz.qpic ...

  10. A Sample Linker Script

    from:http://www.hertaville.com/a-sample-linker-script.html A sample script file that will work with ...