Flume的介紹和簡單操作

Flume是什麼

Flume是Cloudera提供的一個高可用的，高可靠的，分佈式的海量日誌採集、聚合和傳輸的系統，Flume支持在日誌系統中定製各類數據發送方，用於收集數據；同時，Flume提供對數據進行簡單處理，並寫到各種數據接受方（可定製）的能力。

Flume的功能

• 支持在日誌系統中定製各類數據發送方，用於收集數據
• 提供對數據簡單處理，並寫到各類數據接收方（可定製）的能力

Flume的組成

• Agent:核心組件
• source 負責數據的產生或蒐集
• channel 是一種短暫的存儲容器,負責數據的存儲持久化
• sink 負責數據的轉發

Flume的工作流示意圖

• 數據流模型

• 
  
• 多Agent模型

• 
  
• 合併模型

• 
  
• 混合模型

Flume的安裝

下載安裝包並解壓

wget http://www.apache.org/dyn/closer.lua/flume/1.8.0/apache-flume-1.8.0-bin.tar.gz
tar -zxvf apache-flume-1.8.0-bin.tar.gz

配置環境變量

vim ~/.bashrc
​
export FLUME_HOME=/usr/local/src/apache-flume-1.8.0-bin
export PATH=$PATH:$FLUME_HOME/bin
​
source ~/.bashrc

Flume簡單操作

• netcat模式
• 進入conf目錄下編寫netcat.conf文件，內容如下：

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink
​
agent.sources.netcatSource.type = netcat
agent.sources.netcatSource.bind = localhost
agent.sources.netcatSource.port = 11111
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink.type = logger 

agent.sinks.loggerSink.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 100
agent.channels.memoryChannel.transactionCapacity = 10

啟動一個實例

(py27) [root@master conf]# pwd
/usr/local/src/apache-flume-1.8.0-bin/conf
(py27) [root@master conf]# flume-ng agent --conf conf --conf-file ./netcat.conf --name agent -Dflume.root.logger=INFO,console

啟動成功

18/10/24 11:26:35 INFO node.PollingPropertiesFileConfigurationProvider: Configuration provider starting
18/10/24 11:26:35 INFO node.PollingPropertiesFileConfigurationProvider: Reloading configuration file:./flume_netcat.conf
18/10/24 11:26:35 INFO conf.FlumeConfiguration: Processing:loggerSink
18/10/24 11:26:35 INFO conf.FlumeConfiguration: Processing:loggerSink
18/10/24 11:26:35 INFO conf.FlumeConfiguration: Added sinks: loggerSink Agent: agent
18/10/24 11:26:35 INFO conf.FlumeConfiguration: Post-validation flume configuration contains configuration for agents: [agent]
18/10/24 11:26:35 INFO node.AbstractConfigurationProvider: Creating channels
18/10/24 11:26:35 INFO channel.DefaultChannelFactory: Creating instance of channel memoryChannel type memory
18/10/24 11:26:35 INFO node.AbstractConfigurationProvider: Created channel memoryChannel
18/10/24 11:26:35 INFO source.DefaultSourceFactory: Creating instance of source netcatSource, type netcat
18/10/24 11:26:35 INFO sink.DefaultSinkFactory: Creating instance of sink: loggerSink, type: logger
18/10/24 11:26:35 INFO node.AbstractConfigurationProvider: Channel memoryChannel connected to [netcatSource, loggerSink]
18/10/24 11:26:35 INFO node.Application: Starting new configuration:{ sourceRunners:{netcatSource=EventDrivenSourceRunner: { source:org.apache.flume.source.NetcatSource{name:netcatSource,state:IDLE} }} sinkRunners:{loggerSink=SinkRunner: { policy:org.apache.flume.sink.DefaultSinkProcessor@262b92ac counterGroup:{ name:null counters:{} } }} channels:{memoryChannel=org.apache.flume.channel.MemoryChannel{name: memoryChannel}} }
18/10/24 11:26:35 INFO node.Application: Starting Channel memoryChannel
18/10/24 11:26:35 INFO node.Application: Waiting for channel: memoryChannel to start. Sleeping for 500 ms
18/10/24 11:26:36 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: CHANNEL, name: memoryChannel: Successfully registered new MBean.
18/10/24 11:26:36 INFO instrumentation.MonitoredCounterGroup: Component type: CHANNEL, name: memoryChannel started
18/10/24 11:26:36 INFO node.Application: Starting Sink loggerSink
18/10/24 11:26:36 INFO node.Application: Starting Source netcatSource
18/10/24 11:26:36 INFO source.NetcatSource: Source starting
18/10/24 11:26:36 INFO source.NetcatSource: Created serverSocket:sun.nio.ch.ServerSocketChannelImpl[/172.16.155.120:11111]
​

然後新開一個終端，發送數據

(py27) [root@master apache-flume-1.8.0-bin]# telnet localhost 11111
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
1
OK 

查看接收數據

18/10/24 11:30:15 INFO sink.LoggerSink: Event: { headers:{} body: 31 0D 1. }

注：如果沒有telnet工具，請先安裝：yum install telnet

• Exec模式
• 編寫配置文件exec.conf

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink
​
agent.sources.netcatSource.type = exec 
agent.sources.netcatSource.command = tail -f /home/master/FlumeTest/test_data/exec.log
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink.type = logger
agent.sinks.loggerSink.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 100
agent.channels.memoryChannel.transactionCapacity = 10

啟動實例

(py27) [root@master conf]# flume-ng agent --conf conf --conf-file ./flume_exec.conf --name agent -Dflume.root.logger=INFO,console

啟動成功後，創建配置文件中的exec.log文件

(py27) [root@master test_data]# ls
exec.log
(py27) [root@master test_data]# pwd
/home/master/FlumeTest/test_data
(py27) [root@master test_data]#

然後通過echo命令模擬日誌的產生

(py27) [root@master test_data]# echo 'Hello World!!!' >> exec.log

查看接收的日誌

18/10/25 09:19:52 INFO sink.LoggerSink: Event: { headers:{} body: 48 65 6C 6C 6F 20 57 6F 72 6C 64 21 21 21 Hello World!!! }

如何將日誌保存到HDFS上

修改配置文件

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink
​
agent.sources.netcatSource.type = exec 
agent.sources.netcatSource.command = tail -f /home/master/FlumeTest/test_data/exec.log
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink.type = hdfs 
agent.sinks.loggerSink.hdfs.path = /flume/%y-%m-%d/%H%M/
agent.sinks.loggerSink.hdfs.filePrefix = exec_hdfs_
agent.sinks.loggerSink.hdfs.round = true
agent.sinks.loggerSink.hdfs.roundValue = 1
agent.sinks.loggerSink.hdfs.roundUnit = minute
agent.sinks.loggerSink.hdfs.rollInterval = 3
agent.sinks.loggerSink.hdfs.rollSize = 20
agent.sinks.loggerSink.hdfs.rollCount = 5
agent.sinks.loggerSink.hdfs.useLocalTimeStamp = true
agent.sinks.loggerSink.hdfs.fileType = DataStream
agent.sinks.loggerSink.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 100
agent.channels.memoryChannel.transactionCapacity = 10

然後啟動實例

(py27) [root@master conf]# flume-ng agent --conf conf --conf-file ./flume_exec_hdfs.conf --name agent -Dflume.root.logger=INFO,console

然後可以看到它把exec.log文件裡的日誌給寫到了HDFS上

18/10/25 09:54:26 INFO hdfs.HDFSDataStream: Serializer = TEXT, UseRawLocalFileSystem = false
18/10/25 09:54:26 INFO hdfs.BucketWriter: Creating /flume/18-10-25/0954//exec_hdfs_.1540475666623.tmp
18/10/25 09:54:32 INFO hdfs.BucketWriter: Closing /flume/18-10-25/0954//exec_hdfs_.1540475666623.tmp
18/10/25 09:54:32 INFO hdfs.BucketWriter: Renaming /flume/18-10-25/0954/exec_hdfs_.1540475666623.tmp to /flume/18-10-25/0954/exec_hdfs_.1540475666623
18/10/25 09:54:32 INFO hdfs.HDFSEventSink: Writer callback called.

我們進入HDFS查看，可以看到log裡的內容

(py27) [root@master sbin]# hadoop fs -ls /flume/18-10-25/0954
Found 1 items
-rw-r--r-- 3 root supergroup 15 2018-10-25 09:54 /flume/18-10-25/0954/exec_hdfs_.1540475666623
(py27) [root@master sbin]# hadoop fs -text /flume/18-10-25/0954/exec_hdfs_.1540475666623
Hello World!!!

然後我們再次寫入寫的log，然後再查看

//寫入新的log
(py27) [root@master test_data]# echo 'test001' >> exec.log 
(py27) [root@master test_data]# echo 'test002' >> exec.log
//進入HDFS目錄查看
(py27) [root@master sbin]# hadoop fs -ls /flume/18-10-25
Found 2 items
drwxr-xr-x - root supergroup 0 2018-10-25 09:54 /flume/18-10-25/0954
drwxr-xr-x - root supergroup 0 2018-10-25 09:56 /flume/18-10-25/0956
(py27) [root@master sbin]# hadoop fs -ls /flume/18-10-25/0956
Found 1 items
-rw-r--r-- 3 root supergroup 16 2018-10-25 09:56 /flume/18-10-25/0956/exec_hdfs_.1540475766338
(py27) [root@master sbin]# hadoop fs -text /flume/18-10-25/0956/exec_hdfs_.1540475766338
test001
test002

• 故障轉移實例
• 首先需要三臺機器，master、slave1、slave2，然後分別配置實例並啟動，master上的agent實例發送日誌，slave1和slave2接收日誌
• master配置

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink1 loggerSink2
​
agent.sinkgroups = group
​
agent.sources.netcatSource.type = exec
agent.sources.netcatSource.command = tail -f /home/master/FlumeTest/test_data/exec.log
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink1.type = avro
agent.sinks.loggerSink1.hostname = slave1
agent.sinks.loggerSink1.port = 52020
agent.sinks.loggerSink1.channel = memoryChannel
​
agent.sinks.loggerSink2.type = avro
agent.sinks.loggerSink2.hostname = slave2
agent.sinks.loggerSink2.port = 52020
agent.sinks.loggerSink2.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 10000
agent.channels.memoryChannel.transactionCapacity = 1000
​
agent.sinkgroups.group.sinks = loggerSink1 loggerSink2
​
​
agent.sinkgroups.group.processor.type = failover
agent.sinkgroups.group.processor.loggerSink1 = 10
agent.sinkgroups.group.processor.loggerSink2 = 1
agent.sinkgroups.group.processor.maxpenalty = 10000

slave1配置

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink
​
agent.sources.netcatSource.type = avro
agent.sources.netcatSource.bind = slave1
agent.sources.netcatSource.port = 52020
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink.type = logger
agent.sinks.loggerSink.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 10000
agent.channels.memoryChannel.transactionCapacity = 1000 

slave2配置

agent.sources = netcatSource
agent.channels = memoryChannel
agent.sinks = loggerSink
​
agent.sources.netcatSource.type = avro
agent.sources.netcatSource.bind = slave2
agent.sources.netcatSource.port = 52020
agent.sources.netcatSource.channels = memoryChannel
​
agent.sinks.loggerSink.type = logger
agent.sinks.loggerSink.channel = memoryChannel
​
agent.channels.memoryChannel.type = memory
agent.channels.memoryChannel.capacity = 10000
agent.channels.memoryChannel.transactionCapacity = 1000

分別啟動master、slave1、slave2的agent，然後在mater上寫入日誌，然後觀察誰收到了

//master
(py27) [root@master test_data]# echo 'hello' >> exec.log 
//slave1
18/10/25 10:53:53 INFO sink.LoggerSink: Event: { headers:{} body: 68 65 6C 6C 6F hello }
//slave2
18/10/25 10:43:00 INFO ipc.NettyServer: [id: 0x8da012e3, /172.16.155.120:39726 => /172.16.155.122:52020] CONNECTED: /172.16.155.120:39726 

發現是slave1收到數據，然後我們把slave1的agent關掉，再次發送日誌

//master
(py27) [root@master test_data]# echo '11111' >> exec.log 
//slave2
18/10/25 10:43:00 INFO ipc.NettyServer: [id: 0x8da012e3, /172.16.155.120:39726 => /172.16.155.122:52020] CONNECTED: /172.16.155.120:39726
18/10/25 10:56:53 INFO sink.LoggerSink: Event: { headers:{} body: 31 31 31 31 31 11111 }
​

然後再次啟動slave1的agent

//master
(py27) [root@master test_data]# echo '22222' >> exec.log  

//slave1
18/10/25 10:58:21 INFO sink.LoggerSink: Event: { headers:{} body: 32 32 32 32 32 22222 }
//slave2
18/10/25 10:43:00 INFO ipc.NettyServer: [id: 0x8da012e3, /172.16.155.120:39726 => /172.16.155.122:52020] CONNECTED: /172.16.155.120:39726
18/10/25 10:56:53 INFO sink.LoggerSink: Event: { headers:{} body: 31 31 31 31 31 11111 }

如有錯誤，還請指正。