1、Flume
概述:Flume是一种分布式,可靠且可用的服务,用于有效地收集,聚合和移动大量日
志数据。它具有基于流数据流的简单灵活的架构。它具有可靠的可靠性机制和许多故障
转移和恢复机制,具有强大的容错性。它使用简单的可扩展数据模型,允许在线分析应
用程序。
1)数据采集(爬虫\日志数据\flume)
2)数据存储(hdfs/hive/hbase(nosql))
3)数据计算(mapreduce/hive/sparkSQL/sparkStreaming/flink)
4)数据可视化
2、Flume角色
1)source
数据源,用户采集数据,source产生数据流,同时会把产生的数据流传输到channel
2)channel
传输通道,用于桥接source和sink
3)sink
下沉,用于收集channel传输的数据,将数据源传递到目标源
4)agent
在flume中使用事件作为传输的基本单元
3、flume使用
简单易用,只需要写配置文件即可
4、flume安装配置
1)下载flume
2)上传到Linux
3)解压
tar -zxvf apache-flume-1.6.0-bin.tar.gz -C /root/hd
4)重命名
mv apache-flume-1.6.0-bin/ flume
cp flume-env.sh.template flume-env.sh
5)修改配置
vi flume-env.sh
export JAVA_HOME=/root/hd/jdk1.8.0_192
5、flume监听端口
启动命令:
bin/flume-ng agent –conf conf/log4j.properties –name a1 –conf-file conf/flumejob_telnet.conf
我已经排坑了,这里我建议–conf 后面指定的路径建议是全路径,指定到log4j.properties或,我当时老师讲的是直接conf/,我实际操作是有问题的,不能实时的反馈
bin/flume-ng agent 使用ng启动agent--conf conf/log4j.properties 指定配置所在的文件夹--name a1 指定的agent别名--conf-file conf/flumejob_telnet.conf 文件-Dflume.root.logger=INFO,console 日志级别
flumejob_telnet.conf
#smple.conf: A single-node Flume configuration# Name the components on this agent 定义变量方便调用 加s可以有多个此角色a1.sources = r1a1.sinks = k1a1.channels = c1 # Describe/configure the source 描述source角色 进行内容定制# 此配置属于tcp source 必须是netcat类型a1.sources.r1.type = netcat a1.sources.r1.bind = localhosta1.sources.r1.port = 44444 # Describe the sink 输出日志文件a1.sinks.k1.type = logger # Use a channel which buffers events in memory(file) 使用内存 总大小1000 每次传输100a1.channels.c1.type = memorya1.channels.c1.capacity = 1000a1.channels.c1.transactionCapacity = 100 # Bind the source and sink to the channel 一个source可以绑定多个channel # 一个sinks可以只能绑定一个channel ?使用的是图二的模型a1.sources.r1.channels = c1a1.sinks.k1.channel = c1
[root@hsiehchou121 flume]# bin/flume-ng agent \> --conf conf/ \> --name a1 \> --conf-file conf/flumejob_telnet.conf \> -Dflume.root.logger=INFO.console
yum search telnet
yum install telnet.x86_64
6、flume监听本地linux文件采集到hdfs
启动命令:
bin/flume-ng agent –conf conf/log4j.properties –name a1 –conf-file conf/flum
ejob_hdfs.conf
flumejob_hdfs.conf
# Name the components on this agent agent别名设置a1.sources = r1a1.sinks = k1a1.channels = c1 # Describe/configure the source ?设置数据源监听本地文件配置# exec 执行一个命令的方式去查看文件 tail -F 实时查看a1.sources.r1.type = exec# 要执行的脚本command tail -F 默认10行 man tail ?查看帮助a1.sources.r1.command = tail -F /tmp/root/hive.log# 执行这个command使用的是哪个脚本 -c 指定使用什么命令# whereis bash# bash: /usr/bin/bash /usr/share/man/man1/bash.1.gz a1.sources.r1.shell = /usr/bin/bash -c # Describe the sink a1.sinks.k1.type = hdfsa1.sinks.k1.hdfs.path = hdfs://hsiehchou121:9000/flume/%Y%m%d/%H#上传文件的前缀a1.sinks.k1.hdfs.filePrefix = logs-#是否按照时间滚动文件夹a1.sinks.k1.hdfs.round = true#多少时间单位创建一个新的文件夹 ?秒 (默认30s)a1.sinks.k1.hdfs.roundValue = 1#重新定义时间单位(每小时滚动一个文件夹)a1.sinks.k1.hdfs.roundUnit = minute#是否使用本地时间戳a1.sinks.k1.hdfs.useLocalTimeStamp = true#积攒多少个 Event 才 flush 到 HDFS 一次a1.sinks.k1.hdfs.batchSize = 500#设置文件类型,可支持压缩a1.sinks.k1.hdfs.fileType = DataStream#多久生成一个新的文件 秒a1.sinks.k1.hdfs.rollInterval = 30#设置每个文件的滚动大小 字节(最好128M,合理)a1.sinks.k1.hdfs.rollSize = 134217700#文件的滚动与 Event 数量无关a1.sinks.k1.hdfs.rollCount = 0#最小冗余数(备份数 生成滚动功能则生效roll hadoop本身有此功能 无需配置) 1份 不冗余 hdfs已经备份3份a1.sinks.k1.hdfs.minBlockReplicas = 1 # Use a channel which buffers events in memory a1.channels.c1.type = memory a1.channels.c1.capacity = 1000a1.channels.c1.transactionCapacity = 100 # Bind the source and sink to the channela1.sources.r1.channels = c1a1.sinks.k1.channel = c1
[root@hsiehchou121 flume]# bin/flume-ng agent > --conf conf/log4j.properties > --name a1 > --conf-file conf/flumejob_hdfs.conf
7、监听文件夹
flumejob_dir.conf
# 定义别名a1.sources = r1a1.sinks = k1a1.channels = c1 # Describe/configure the sourcea1.sources.r1.type = spooldir# 监控的文件夹a1.sources.r1.spoolDir = /root/testdir# 上传成功后显示后缀名 a1.sources.r1.fileSuffix = .COMPLETED# 如论如何 加绝对路径的文件名 默认falsea1.sources.r1.fileHeader = true#忽略所有以.tmp 结尾的文件(正在被写入),不上传# ^以任何开头 出现无限次 以.tmp结尾的a1.sources.r1.ignorePattern = ([^ ]*\.tmp) # Describe the sink 下沉到hdfsa1.sinks.k1.type = hdfsa1.sinks.k1.hdfs.path = hdfs://hsiehchou121:9000/flume/testdir/%Y%m%d/%H#上传文件的前缀a1.sinks.k1.hdfs.filePrefix = testdir-#是否按照时间滚动文件夹a1.sinks.k1.hdfs.round = true#多少时间单位创建一个新的文件夹a1.sinks.k1.hdfs.roundValue = 1#重新定义时间单位a1.sinks.k1.hdfs.roundUnit = hour#是否使用本地时间戳a1.sinks.k1.hdfs.useLocalTimeStamp = true#积攒多少个 Event 才 flush 到 HDFS 一次a1.sinks.k1.hdfs.batchSize = 100#设置文件类型,可支持压缩a1.sinks.k1.hdfs.fileType = DataStream#多久生成一个新的文件a1.sinks.k1.hdfs.rollInterval = 600#设置每个文件的滚动大小大概是 128M a1.sinks.k1.hdfs.rollSize = 134217700#文件的滚动与 Event 数量无关a1.sinks.k1.hdfs.rollCount = 0#最小副本数a1.sinks.k1.hdfs.minBlockReplicas = 1 # Use a channel which buffers events in memory a1.channels.c1.type = memory a1.channels.c1.capacity = 1000a1.channels.c1.transactionCapacity = 100 # Bind the source and sink to the channela1.sources.r1.channels = c1 a1.sinks.k1.channel = c1
[root@hsiehchou121 conf]# bin/flume-ng agent –conf conf/log4j.properties –name a1 –conf-file conf/flumejob_dir.conf
[root@hsiehchou121 flume]# bin/flume-ng agent > --conf conf/log4j.properties > --name a1 > --conf-file conf/flumejob_dir.conf
8、多个channel/sink
需求:监控hive.log文件,用同时产生两个channel,一个channel对应的sink存储到hdfs中,另外一个channel对应的sink存储到本地
flumejob_1.conf
# name the components on this agent 别名设置a1.sources = r1a1.sinks = k1 k2 a1.channels = c1 c2 # 将数据流复制给多个 channela1.sources.r1.selector.type = replicating # Describe/configure the source a1.sources.r1.type = execa1.sources.r1.command = tail -F /tmp/root/hive.loga1.sources.r1.shell = /bin/bash -c # Describe the sink# 分两个端口发送数据 a1.sinks.k1.type = avro a1.sinks.k1.hostname = hsiehchou121 a1.sinks.k1.port = 4141 a1.sinks.k2.type = avro a1.sinks.k2.hostname = hsiehchou121 a1.sinks.k2.port = 4142 # Describe the channel a1.channels.c1.type = memory a1.channels.c1.capacity = 1000a1.channels.c1.transactionCapacity = 100 a1.channels.c2.type = memory a1.channels.c2.capacity = 1000a1.channels.c2.transactionCapacity = 100 # Bind the source and sink to the channel a1.sources.r1.channels = c1 c2 a1.sinks.k1.channel = c1a1.sinks.k2.channel = c2
[root@hsiehchou121 flume]# bin/flume-ng agent –conf conf/log4j.properties –name a1 –conf-file conf/flumejob_1.conf
flumejob_2.conf
# Name the components on this agent a2.sources = r1a2.sinks = k1 a2.channels = c1 # Describe/configure the sourcea2.sources.r1.type = avro # 端口抓取数据a2.sources.r1.bind = hsiehchou121a2.sources.r1.port = 4141 # Describe the sink a2.sinks.k1.type = hdfsa2.sinks.k1.hdfs.path = hdfs://hsiehchou121:9000/flume2/%Y%m%d/%H #上传文件的前缀a2.sinks.k1.hdfs.filePrefix = flume2-#是否按照时间滚动文件夹a2.sinks.k1.hdfs.round = true#多少时间单位创建一个新的文件夹a2.sinks.k1.hdfs.roundValue = 1#重新定义时间单位a2.sinks.k1.hdfs.roundUnit = hour#是否使用本地时间戳a2.sinks.k1.hdfs.useLocalTimeStamp = true#积攒多少个 Event 才 flush 到 HDFS 一次a2.sinks.k1.hdfs.batchSize = 100 #设置文件类型,可支持压缩a2.sinks.k1.hdfs.fileType = DataStream#多久生成一个新的文件a2.sinks.k1.hdfs.rollInterval = 600#设置每个文件的滚动大小大概是 128M a2.sinks.k1.hdfs.rollSize = 134217700#文件的滚动与 Event 数量无关a2.sinks.k1.hdfs.rollCount = 0#最小副本数a2.sinks.k1.hdfs.minBlockReplicas = 1 # Describe the channel a2.channels.c1.type = memory a2.channels.c1.capacity = 1000a2.channels.c1.transactionCapacity = 100 # Bind the source and sink to the channel a2.sources.r1.channels = c1a2.sinks.k1.channel = c1
[root@hsiehchou121 flume]# bin/flume-ng agent –conf conf/log4j.properties –name a2 –conf-file conf/flumejob_1.conf
flumejob_3.conf
# Name the components on this agent a3.sources = r1a3.sinks = k1 a3.channels = c1 # Describe/configure the source a3.sources.r1.type = avroa3.sources.r1.bind = hsiehchou121a3.sources.r1.port = 4142 # Describe the sink a3.sinks.k1.type = file_rolla3.sinks.k1.sink.directory = /root/flume2 # Describe the channel a3.channels.c1.type = memory a3.channels.c1.capacity = 1000a3.channels.c1.transactionCapacity = 100 # Bind the source and sink to the channel a3.sources.r1.channels = c1a3.sinks.k1.channel = c1
[root@hsiehchou121 flume]# bin/flume-ng agent –conf conf/log4j.properties –name a3 –conf-file conf/flumejob_1.conf
Flume
原文地址:https://www.cnblogs.com/hsiehchou/p/10502457.html