Flume原型图.png
Flume事件流.png
Flume数据流.png 
Flume数据流2.png
下载地址：http://flume.apache.org/download.html
解压缩安装包文件
[hadoop@hadoop01 apps]$ tar -zxvf apache-flume-1.8.0-bin.tar.gz
[hadoop@hadoop01 apps]$ cd apache-flume-1.8.0-bin/
[hadoop@hadoop01 apache-flume-1.8.0-bin]$ ll
总用量 148
drwxr-xr-x. 2 hadoop hadoop 62 1月 21 14:31 bin
-rw-r--r--. 1 hadoop hadoop 81264 9月 15 20:26 CHANGELOG
drwxr-xr-x. 2 hadoop hadoop 127 1月 21 14:31 conf
-rw-r--r--. 1 hadoop hadoop 5681 9月 15 20:26 DEVNOTES
-rw-r--r--. 1 hadoop hadoop 2873 9月 15 20:26 doap_Flume.rdf
drwxr-xr-x. 10 hadoop hadoop 4096 9月 15 20:48 docs
drwxr-xr-x. 2 hadoop hadoop 8192 1月 21 14:31 lib
-rw-r--r--. 1 hadoop hadoop 27663 9月 15 20:26 LICENSE
-rw-r--r--. 1 hadoop hadoop 249 9月 15 20:26 NOTICE
-rw-r--r--. 1 hadoop hadoop 2483 9月 15 20:26 README.md
-rw-r--r--. 1 hadoop hadoop 1588 9月 15 20:26 RELEASE-NOTES
drwxr-xr-x. 2 hadoop hadoop 68 1月 21 14:31 tools
[hadoop@hadoop01 apache-flume-1.8.0-bin]$
[root@hadoop01 bin]# ln -s /home/hadoop/apps/apache-flume-1.8.0-bin /usr/local/flume
编辑 /etc/profile文件，增加以下内容：
export FLUME_HOME=/usr/local/flume
export PATH=$PATH:${JAVA_HOME}/bin:${ZOOKEEPER_HOME}/bin:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:${HIVE_HOME}/bin:${FLUME_HOME}/bin
使用example.conf 配置文件启动一个实例
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
a1.sources.r1.channels = c1
a1.channels.c1.type = memory
a1.channels.c1.capacity = 1000
a1.channels.c1.transactionCapacity = 100
a1.sinks.k1.type = logger
a1.sinks.k1.channel = c1
启动命令如下：
[root@hadoop01 conf]# pwd
/home/hadoop/apps/apache-flume-1.8.0-bin/conf
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file example.conf --name a1 -Dflume.root.logger=INFO,console
启动成功后如下图所示：
........略
18/01/27 18:17:25 INFO node.AbstractConfigurationProvider: Channel c1 connected to [r1, k1]
18/01/27 18:17:25 INFO node.Application: Starting new configuration:{ sourceRunners:{r1=EventDrivenSourceRunner: { source:org.apache.flume.source.NetcatSource{name:r1,state:IDLE} }} sinkRunners:{k1=SinkRunner: { policy:org.apache.flume.sink.DefaultSinkProcessor@20470f counterGroup:{ name:null counters:{} } }} channels:{c1=org.apache.flume.channel.MemoryChannel{name: c1}} }
18/01/27 18:17:25 INFO node.Application: Starting Channel c1
18/01/27 18:17:25 INFO node.Application: Waiting for channel: c1 to start. Sleeping for 500 ms
18/01/27 18:17:25 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: CHANNEL, name: c1: Successfully registered new MBean.
18/01/27 18:17:25 INFO instrumentation.MonitoredCounterGroup: Component type: CHANNEL, name: c1 started
18/01/27 18:17:26 INFO node.Application: Starting Sink k1
18/01/27 18:17:26 INFO node.Application: Starting Source r1
18/01/27 18:17:26 INFO source.NetcatSource: Source starting
18/01/27 18:17:26 INFO source.NetcatSource: Created serverSocket:sun.nio.ch.ServerSocketChannelImpl[/127.0.0.1:44444]
使用telnet发送数据
[root@hadoop01 apps]# telnet localhost 44444
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is ‘^]‘.
Are you OK ?
OK
控制台打印如下：
Impl[/127.0.0.1:44444]
18/01/27 18:21:00 INFO sink.LoggerSink: Event: { headers:{} body: 41 72 65 20 79 6F 75 20 4F 4B 20 3F 0D Are you OK ?. }
如无法使用telnet，请先安装telnet工具
[root@hadoop01 apps]# yum -y install telnet
| Source类型 | Type | 用途 |
|---|---|---|
| Avro Source | avro | 启动一个Avro Server，可与上一级Agent连接 |
| HTTP Source | http | 启动一个HttpServer |
| Exec Source | exec | 执行unix command，获取标准输出，如tail -f |
| Taildir Source | TAILDIR | 监听目录或文件 |
| Spooling Directory Source | spooldir | 监听目录下的新增文件 |
| Kafka Source | org.apache.flume.sourc e.kafka.KafkaSource | 读取Kafka数据 |
| JMS Source | jms | 从JMS源读取数据 |
//avrosource.conf
avroagent.sources = r1
avroagent.channels = c1
avroagent.sinks = k1
avroagent.sources.r1.type = avro
avroagent.sources.r1.bind = 192.168.43.20
avroagent.sources.r1.port = 8888
avroagent.sources.r1.threads= 3
avroagent.sources.r1.channels = c1
avroagent.channels.c1.type = memory
avroagent.channels.c1.capacity = 10000
avroagent.channels.c1.transactionCapacity = 1000
avroagent.sinks.k1.type = logger
avroagent.sinks.k1.channel = c1
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file avrosource.conf --name avroagent -Dflume.root.logger=INFO,console
启动成功入下图所示：
...略
18/01/27 18:46:36 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: CHANNEL, name: c1: Successfully registered new MBean.
18/01/27 18:46:36 INFO instrumentation.MonitoredCounterGroup: Component type: CHANNEL, name: c1 started
18/01/27 18:46:36 INFO node.Application: Starting Sink k1
18/01/27 18:46:36 INFO node.Application: Starting Source r1
18/01/27 18:46:36 INFO source.AvroSource: Starting Avro source r1: { bindAddress: 192.168.43.20, port: 8888 }...
18/01/27 18:46:37 INFO instrumentation.MonitoredCounterGroup: Monitored counter group for type: SOURCE, name: r1: Successfully registered new MBean.
18/01/27 18:46:37 INFO instrumentation.MonitoredCounterGroup: Component type: SOURCE, name: r1 started
18/01/27 18:46:37 INFO source.AvroSource: Avro source r1 started
execagent.sources = r1
execagent.channels = c1
execagent.sinks = k1
execagent.sources.r1.type = exec
execagent.sources.r1.command = tail -F /home/hadoop/apps/flume/execsource/exectest.log
execagent.sources.r1.channels = c1
execagent.channels.c1.type = memory
execagent.channels.c1.capacity = 10000
execagent.channels.c1.transactionCapacity = 1000
execagent.sinks.k1.type = avro
execagent.sinks.k1.channel = c1
execagent.sinks.k1.hostname = 192.168.43.20
execagent.sinks.k1.port = 8888
启动 execAgent
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file execsource.conf --name execagent
启动成功如下下图所示：
18/01/27 18:58:43 INFO instrumentation.MonitoredCounterGroup: Component type: SINK, name: k1 started
18/01/27 18:58:43 INFO sink.AbstractRpcSink: Rpc sink k1: Building RpcClient with hostname: 192.168.43.20, port: 8888
18/01/27 18:58:43 INFO sink.AvroSink: Attempting to create Avro Rpc client.
18/01/27 18:58:43 WARN api.NettyAvroRpcClient: Using default maxIOWorkers
18/01/27 18:58:44 INFO sink.AbstractRpcSink: Rpc sink k1 started.
在execAgent监控的文件下写入内容，观察sourceagent是否接收到变化内容
[root@hadoop01 execsource]# echo 222 > exectest.log
[root@hadoop01 execsource]# echo 5555 >> exectest.log
[root@hadoop01 execsource]# cat exectest.log
222
5555
在sourceagent控制打印台下查看监控消息如下：
18/01/27 18:58:50 INFO sink.LoggerSink: Event: { headers:{} body: 31 32 33 123 }
18/01/27 18:59:55 INFO sink.LoggerSink: Event: { headers:{} body: 35 35 35 35 5555 }
则说明2个串联agent传递信息成功。
说明:
avroagent 配置文件配置项起始名称需要与服务启动 -name 名称相一致。
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = spooldir
a1.sources.r1.channels = c1
a1.sources.r1.spoolDir = /home/hadoop/apps/flume/spoolDir
a1.sources.r1.fileHeader = true
a1.channels.c1.type = memory
a1.channels.c1.capacity = 10000
a1.channels.c1.transactionCapacity = 1000
a1.sinks.k1.type = logger
a1.sinks.k1.channel = c1
/home/hadoop/apps/flume/spoolDir 必须已经创建且具有用户读写权限。
启动 SpoolDirsourceAgent
[hadoop@hadoop01 conf]$ flume-ng agent --conf conf --conf-file spooldirsource.conf --name a1 -Dflume.root.logger=INFO,console
在spoolDir文件夹下创建文件并写入文件内容，观察控制台消息：
18/01/28 17:06:54 INFO avro.ReliableSpoolingFileEventReader: Preparing to move file /home/hadoop/apps/flume/spoolDir/test to /home/hadoop/apps/flume/spoolDir/test.COMPLETED
18/01/28 17:06:55 INFO sink.LoggerSink: Event: { headers:{file=/home/hadoop/apps/flume/spoolDir/test} body: 32 32 32 222 }
此时监测到SpoolDirSourceAgent 可以监控到文件变化。
值得说明的是：Spooling Directory Source Agent 并不能监听子级文件夹的文件变化,也不支持已存在的文件更新数据变化.
kafkasourceagent.sources = r1
kafkasourceagent.channels = c1
kafkasourceagent.sinks = k1
kafkasourceagent.sources.r1.type = org.apache.flume.source.kafka.KafkaSource
kafkasourceagent.sources.r1.channels = c1
kafkasourceagent.sources.r1.batchSize = 100
kafkasourceagent.sources.r1.batchDurationMillis = 1000
kafkasourceagent.sources.r1.kafka.bootstrap.servers = 192.168.43.22:9092,192.168.43.23:9092,192.168.43.24:9092
kafkasourceagent.sources.r1.kafka.topics = flumetopictest1
kafkasourceagent.sources.r1.kafka.consumer.group.id = flumekafkagroupid
kafkasourceagent.channels.c1.type = memory
kafkasourceagent.channels.c1.capacity = 10000
kafkasourceagent.channels.c1.transactionCapacity = 1000
kafkasourceagent.sinks.k1.type = logger
kafkasourceagent.sinks.k1.channel = c1
首先启动3个节点的kafka节点服务，在每个kafka节点执行，以后台方式运行
[root@hadoop03 bin]# ./kafka-server-start.sh -daemon ../config/server.properties
在kafka节点上创建一个配置好的Topic flumetoptest1,命令如下：
[root@hadoop03 bin]# ./kafka-topics.sh --create --zookeeper 192.168.43.20:2181 --replication-factor 1 --partitions 3 --topic flumetopictest1
Created topic "flumetopictest1".
创建成功后，启动一个kafka Source Agent，命令如下：
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file kafkasource.conf --name kafkasourceagent -Dflume.root.logger=INFO,console
创建一个Kafka 生产者，进行消息发送
root@hadoop03 bin]# ./kafka-console-producer.sh --broker-list 192.168.43.22:9092,192.168.43.23:9092 --topic flumetopictest1
发送消息，此时kafka 就可以接收到消息：
18/02/03 20:36:57 INFO sink.LoggerSink: Event: { headers:{topic=flumetopictest1, partition=2, timestamp=1517661413068} body: 31 32 33 31 33 32 32 31 12313221 }
18/02/03 20:37:09 INFO sink.LoggerSink: Event: { headers:{topic=flumetopictest1, partition=1, timestamp=1517661428930} body: 77 69 20 61 69 79 6F 75 08 08 08 wi aiyou... }
监听一个文件夹或者文件，通过正则表达式匹配需要监听的 数据源文件，Taildir Source通过将监听的文件位置写入到文件中来实现断点续传，并且能够保证没有重复数据的读取.
taildiragent.sources=r1
taildiragent.channels=c1
taildiragent.sinks=k1
taildiragent.sources.r1.type=TAILDIR
taildiragent.sources.r1.positionFile=/home/hadoop/apps/flume/taildir/position/taildir_position.json
taildiragent.sources.r1.filegroups=f1 f2
taildiragent.sources.r1.filegroups.f1=/home/hadoop/apps/flume/taildir/test1/test.log
taildiragent.sources.r1.filegroups.f2=/home/hadoop/apps/flume/taildir/test2/.*log.*
taildiragent.sources.r1.channels=c1
taildiragent.channels.c1.type=memory
taildiragent.channels.c1.transcationCapacity=1000
taildiragent.sinks.k1.type=logger
taildiragent.sinks.k1.channel=c1
启动一个taildirSource agent ,代码如下：
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file taildirsource.conf --name taildiragent -Dflume.root.logger=INFO,console
开始在test1和test2文件夹写入文件，观察agent消息接收。
type ：channel类型memory
capacity ：channel中存储的最大event数，默认值100
transactionCapacity ：一次事务中写入和读取的event最大数，默认值100。
keep-alive：在Channel中写入或读取event等待完成的超时时间，默认值3秒
byteCapacityBufferPercentage：缓冲空间占Channel容量（byteCapacity）的百分比，为event中的头信息保留了空间，默认值20（单位百分比）
byteCapacity ：Channel占用内存的最大容量，默认值为Flume堆内存的80%
type：channel类型为file
checkpointDir：检查点目录，默认在启动flume用户目录下创建，建 议单独配置磁盘路径
useDualCheckpoints：是否开启备份检查点，默认false，建议设置为true开启备份检查点，备份检查点的作用是当Agent意外出错导致写 入检查点文件异常，在重新启动File Channel时通过备份检查点将数据回放到内存中，如果不开启备份检查点，在数据回放的过程中发现检查点文件异常会对所数据进行全回放，全回放的过程相当耗时
backupCheckpointDir：备份检查点目录，最好不要和检查点目录在同 一块磁盘上
checkpointInterval：每次写检查点的时间间隔，默认值30000毫秒
dataDirs：数据文件磁盘存储路径，建议配置多块盘的多个路径，通过磁盘的并行写入来提高file channel性能，多个磁盘路径用逗号隔开
transactionCapacity：一次事务中写入和读取的event最大数，默认值 10000
maxFileSize：每个数据文件的最大大小，默认值：2146435071字节
minimumRequiredSpace：磁盘路径最小剩余空间，如果磁盘剩余空 间小于设置值，则不再写入数据
capacity：file channel可容纳的最大event数
keep-alive：在Channel中写入或读取event等待完成的超时时间，默认值3秒
配置一个FileChannel,filechannel.conf 的配置内容如下：
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
a1.sources.r1.channels = c1
a1.channels.c1.type = file
a1.channels.c1.dataDirs = /home/hadoop/apps/flume/filechannel/data
a1.channels.c1.checkpointDir = /home/hadoop/apps/flume/filechannel/checkpoint
a1.channels.c1.useDualCheckpoints = true
a1.channels.c1.backupCheckpointDir = /home/hadoop/apps/flume/filechannel/backup
a1.sinks.k1.type = logger
a1.sinks.k1.channel = c1
启动一个FileChannel,启动命令如下：
[root@hadoop01 bin]# flume-ng agent --conf conf --conf-file filechannle.conf --name a1 -Dflume.root.logger=INFO,console
向配置文件端口44444发送数据，观察Channel记录情况
telnet localhost asdfasd
此时可以观察到控制台打印监控结果
18/02/04 21:15:44 INFO sink.LoggerSink: Event: { headers:{} body: 61 64 66 61 64 66 61 64 66 61 73 66 0D adfadfadfasf. }
18/02/04 21:15:48 INFO file.EventQueueBackingStoreFile: Start checkpoint for /home/hadoop/apps/flume/filechannel/checkpoint/checkpoint, elements to sync = 1
18/02/04 21:15:48 INFO file.EventQueueBackingStoreFile: Updating checkpoint metadata: logWriteOrderID: 1517749968978, queueSize: 0, queueHead: 0
18/02/04 21:15:48 INFO file.EventQueueBackingStoreFile: Attempting to back up checkpoint.
18/02/04 21:15:48 INFO file.Serialization: Skipping in_use.lock because it is in excludes set
18/02/04 21:15:48 INFO file.Serialization: Deleted the following files: , checkpoint, checkpoint.meta, inflightputs, inflighttakes.
18/02/04 21:15:48 INFO file.Log: Updated checkpoint for file: /home/hadoop/apps/flume/filechannel/data/log-2 position: 170 logWriteOrderID: 1517749968978
18/02/04 21:15:49 INFO file.EventQueueBackingStoreFile: Checkpoint backup completed.
Kafka Channel：将分布式消息队列kafka作为channel相对于Memory Channel和File Channel存储容量更大、 容错能力更强，弥补了其他两种Channel的短板，如果合理利用Kafka的性能，能够达到事半功倍的效果。
关键参数如下：
type：Kafka Channel类型org.apache.flume.channel.kafka.KafkaChannel
kafka.bootstrap.servers：Kafka broker列表，格式为ip1:port1, ip2:port2…，建 议配置多个值提高容错能力，多个值之间用逗号隔开
kafka.topic：topic名称，默认值“flume-channel”
kafka.consumer.group.id：Consumer Group Id，全局唯一
parseAsFlumeEvent：是否以Avro FlumeEvent模式写入到Kafka Channel中， 默认值true，event的header信息与event body都写入到kafka中
pollTimeout：轮询超时时间，默认值500毫秒
kafka.consumer.auto.offset.reset：earliest表示从最早的偏移量开始拉取，latest表示从最新的偏移量开始拉取，none表示如果没有发现该Consumer组之前拉 取的偏移量则抛异常
配置一个KafakChannel， kafkachannel.conf 配置内容如下：
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
a1.sources.r1.channels = c1
a1.channels.c1.type = org.apache.flume.channel.kafka.KafkaChannel
a1.channels.c1.kafka.bootstrap.servers = 192.168.43.22:9092,192.168.43.23:9092
a1.channels.c1.kafka.topic = flumechannel2
a1.channels.c1.kafka.consumer.group.id = flumecgtest1
a1.sinks.k1.type = logger
a1.sinks.k1.channel = c1
启动kafak服务，创建一个kafka主题，命令如下：
[root@hadoop03 bin]# ./kafka-server-start.sh -daemon ../config/server.properties
[root@hadoop03 bin]# ./kafka-topics.sh --create --zookeeper 192.168.43.20:2181 --replication-factor 1 --partitions 3 --topic flumechannel2
查看创建的主题信息
[root@hadoop03 bin]# ./kafka-topics.sh --list --zookeeper 192.168.43.20:2181
__consumer_offsets
flumechannel2
topicnewtest1
启动kafka agent,使用telnet发送数据
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file kafkachannel.conf --name a1 -Dflume.root.logger=INFO,console
[root@hadoop01 flume]# clear
[root@hadoop01 flume]# telnet localhost 44444
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is ‘^]‘.
abc
OK
监听信息如下：
18/02/04 21:39:33 INFO sink.LoggerSink: Event: { headers:{} body: 61 62 63 0D abc. }
关键参数说明
type：Sink类型为avro。
hostname：绑定的目标Avro Souce主机名称或者IP
port：绑定的目标Avro Souce端口号
batch-size：批量发送Event数，默认值100
compression-type：是否使用压缩，如果使用压缩设则值为
“deflate”， Avro Sink设置了压缩那么Avro Source也应设置相同的 压缩格式，目前支持zlib压缩，默认值none
compression-level：压缩级别，0表示不压缩，从1到9数字越大压缩
效果越好，默认值6
关键参数信息说明如下：
type：Sink类型为hdfs。
hdfs.path：HDFS存储路径，支持按日期时间分区。
hdfs.filePrefix：Event输出到HDFS的文件名前缀，默认前缀FlumeData
hdfs.fileSuffix：Event输出到HDFS的文件名后缀
hdfs.inUsePrefix：临时文件名前缀
hdfs.inUseSuffix：临时文件名后缀，默认值.tmp
hdfs.rollInterval：HDFS文件滚动生成时间间隔，默认值30秒，该值设置 为0表示文件不根据时间滚动生成
配置一个hdfsink.conf文件，配置内容如下：
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
a1.sources.r1.interceptors = i1
a1.sources.r1.interceptors.i1.type = timestamp
a1.sources.r1.interceptors.i1.preserveExisting = false
a1.sources.r1.channels = c1
a1.channels.c1.type = memory
a1.channels.c1.capacity = 10000
a1.channels.c1.transactionCapacity = 1000
a1.sinks.k1.type = hdfs
a1.sinks.k1.channel = c1
a1.sinks.k1.hdfs.path = /data/flume/%Y%m%d
a1.sinks.k1.hdfs.filePrefix = hdfssink
a1.sinks.k1.hdfs.fileType = DataStream
a1.sinks.k1.hdfs.writeFormat = Text
a1.sinks.k1.hdfs.round = true
a1.sinks.k1.hdfs.roundValue = 1
a1.sinks.k1.hdfs.roundUnit = minute
a1.sinks.k1.hdfs.callTimeout = 60000
启动一个hdfssink agent，命令如下：
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file hdfssink.conf --name a1 -Dflume.root.logger=INFO,console
使用telnet 向44444发送数据，观察数据写入结果
[hadoop@hadoop01 root]$ telnet localhost 44444
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is ‘^]‘.
abc
OK
2323444
OK
此时控制台打印，在HDFS文件系统生成一个临时文件
8/02/04 22:41:52 INFO hdfs.HDFSDataStream: Serializer = TEXT, UseRawLocalFileSystem = false
18/02/04 22:41:52 INFO hdfs.BucketWriter: Creating /data/flume/20180204/hdfssink.1517755312242.tmp
18/02/04 22:42:24 INFO hdfs.BucketWriter: Closing /data/flume/20180204/hdfssink.1517755312242.tmp
18/02/04 22:42:24 INFO hdfs.BucketWriter: Renaming /data/flume/20180204/hdfssink.1517755312242.tmp to /data/flume/20180204/hdfssink.1517755312242
18/02/04 22:42:24 INFO hdfs.HDFSEventSink: Writer callback called.
值得注意的是：请使用hadoop用户来执行agent的创建和消息的发送，避免因权限导致HDFS文件无法写入
Flume通过KafkaSink将Event写入到Kafka指定的主题中
主要参数说明如下：
type：Sink类型，值为KafkaSink类路径 org.apache.flume.sink.kafka.KafkaSink。
kafka.bootstrap.servers：Broker列表，定义格式host:port，多个Broker之间用逗号隔开，可以配置一个也可以配置多个，用于Producer发现集群中的Broker，建议配置多个，防止当个Broker出现问题连接 失败。
kafka.topic：Kafka中Topic主题名称，默认值flume-topic。
flumeBatchSize：Producer端单次批量发送的消息条数，该值应该根据实际环境适当调整，增大批量发送消息的条数能够在一定程度上提高性能，但是同时也增加了延迟和Producer端数据丢失的风险。 默认值100。
kafka.producer.acks：设置Producer端发送消息到Borker是否等待接收Broker返回成功送达信号。0表示Producer发送消息到Broker之后不需要等待Broker返回成功送达的信号，这种方式吞吐量高，但是存 在数据丢失的风险。1表示Broker接收到消息成功写入本地log文件后向Producer返回成功接收的信号，不需要等待所有的Follower全部同步完消息后再做回应，这种方式在数据丢失风险和吞吐量之间做了平衡。all（或者-1）表示Broker接收到Producer的消息成功写入本 地log并且等待所有的Follower成功写入本地log后向Producer返回成功接收的信号，这种方式能够保证消息不丢失，但是性能最差。默 认值1。
useFlumeEventFormat：默认值false，Kafka Sink只会将Event body内 容发送到Kafka Topic中。如果设置为true，Producer发送到KafkaTopic中的Event将能够保留Producer端头信息
配置一个kafkasink.conf,具体配置内容如下：
a1.sources = r1
a1.channels = c1
a1.sinks = k1
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
a1.sources.r1.channels = c1
a1.channels.c1.type = memory
a1.channels.c1.capacity = 10000
a1.channels.c1.transactionCapacity = 1000
a1.sinks.k1.type = org.apache.flume.sink.kafka.KafkaSink
a1.sinks.k1.channel = c1
a1.sinks.k1.kafka.topic = FlumeKafkaSinkTopic1
a1.sinks.k1.kafka.bootstrap.servers = 192.168.43.22:9092,192.168.43.23:9092
a1.sinks.k1.kafka.flumeBatchSize = 100
a1.sinks.k1.kafka.producer.acks = 1
启动kafka Broker节点22和Broker节点23
[root@hadoop03 bin]# ./kafka-server-start.sh -daemon ../config/server.properties
按配置文件创建主题信息
[root@hadoop03 bin]# ./kafka-topics.sh --create --zookeeper 192.168.43.20:2181 --replication-factor 1 --partitions 3 --topic FlumeKafkaSinkTopic1
Created topic "FlumeKafkaSinkTopic1".
启动一个kafkasink agent，启动命令如下：
[root@hadoop01 conf]# flume-ng agent --conf conf --conf-file kafkasink.conf --name a1 >/dev/null 2>&1 &
一个source将一个event拷贝到多个channel，通过不同的sink消费不同的channel，将相同的event输出到不同的地方
配置文件：replicating_selector.conf
a1.sources = r1
a1.channels = c1 c2
a1.sinks = k1 k2
#定义source
a1.sources.r1.type = netcat
a1.sources.r1.bind = localhost
a1.sources.r1.port = 44444
#设置复制选择器
a1.sources.r1.selector.type = replicating
#设置required channel
a1.sources.r1.channels = c1 c2
#设置channel c1
a1.channels.c1.type = memory
a1.channels.c1.capacity = 1000
a1.channels.c1.transactionCapacity = 1000
#设置channel c2
a1.channels.c2.type = memory
a1.channels.c2.capacity = 1000
a1.channels.c2.transactionCapacity = 1000
#设置kafka sink
a1.sinks.k1.channel = c1
a1.sinks.k1.type = org.apache.flume.sink.kafka.KafkaSink
a1.sinks.k1.kafka.topic = FlumeSelectorTopic1
a1.sinks.k1.kafka.bootstrap.servers = 192.168.43.22:9092,192.168.23.103:9092
a1.sinks.k1.kafka.flumeBatchSize = 5
a1.sinks.k1.kafka.producer.acks = 1
#设置file sink
a1.sinks.k2.channel = c2
a1.sinks.k2.type = file_roll
a1.sinks.k2.sink.directory = /home/hadoop/apps/flume/selector
a1.sinks.k2.sink.rollInterval = 60
分别写入到kafka和文件中
创建主题FlumeKafkaSinkTopic1
bin/kafka-topics.sh --create --zookeeper 192.168.183.100:2181 --replication-factor 1 --partitions 3 --topic FlumeSelectorTopic1
启动flume agent
bin/flume-ng agent --conf conf --conf-file conf/replicating_selector.conf --name a1
使用telnet发送数据
telnet localhost 44444
查看/home/hadoop/apps/flume/selector路径下的数据
查看kafka FlumeSelectorTopic1主题数据
bin/kafka-console-consumer.sh --zookeeper 192.168.183.100:2181 --from-beginning --topic FlumeSelectorTopic1
-Multiplexing Channel Selector多路复用选择器根据event的头信息中不
同键值数据来判断Event应该被写入到哪个Channel中
selector.type：Channel选择器类型为multiplexing
selector.header：设置头信息中用于检测的headerName
selector.default：默认写入的Channel列表
selector.mapping.*：headerName对应的不同值映射的不同Channel列表
selector.optional：可选写入的Channel列表
配置文件multiplexing_selector.conf、avro_sink1.conf、avro_sink2.conf、avro_sink3.conf
向不同的avro_sink对应的配置文件的agent发送数据，不同的avro_sink配置文件通过static interceptor在event头信息中写入不同的静态数据
multiplexing_selector根据event头信息中不同的静态数据类型分别发送到不同的目的地
multiplexing_selector.conf
a3.sources = r1
a3.channels = c1 c2 c3
a3.sinks = k1 k2 k3
a3.sources.r1.type = avro
a3.sources.r1.bind = 192.168.183.100
a3.sources.r1.port = 8888
a3.sources.r1.threads= 3
#设置multiplexing selector
a3.sources.r1.selector.type = multiplexing
a3.sources.r1.selector.header = logtype
#通过header中logtype键对应的值来选择不同的sink
a3.sources.r1.selector.mapping.ad = c1
a3.sources.r1.selector.mapping.search = c2
a3.sources.r1.selector.default = c3
a3.sources.r1.channels = c1 c2 c3
a3.channels.c1.type = memory
a3.channels.c1.capacity = 10000
a3.channels.c1.transactionCapacity = 1000
a3.channels.c2.type = memory
a3.channels.c2.capacity = 10000
a3.channels.c2.transactionCapacity = 1000
a3.channels.c3.type = memory
a3.channels.c3.capacity = 10000
a3.channels.c3.transactionCapacity = 1000
#分别设置三个sink的不同输出
a3.sinks.k1.type = file_roll
a3.sinks.k1.channel = c1
a3.sinks.k1.sink.directory = /home/hadoop/apps/flume/multiplexing/k11
a3.sinks.k1.sink.rollInterval = 60
a3.sinks.k2.channel = c2
a3.sinks.k2.type = file_roll
a3.sinks.k2.sink.directory = /home/hadoop/apps/flume/multiplexing/k12
a3.sinks.k2.sink.rollInterval = 60
a3.sinks.k3.channel = c3
a3.sinks.k3.type = file_roll
a3.sinks.k3.sink.directory = /home/hadoop/apps/flume/multiplexing/k13
a3.sinks.k3.sink.rollInterval = 60
avro_sink1.conf
agent1.sources = r1
agent1.channels = c1
agent1.sinks = k1
agent1.sources.r1.type = netcat
agent1.sources.r1.bind = localhost
agent1.sources.r1.port = 44444
agent1.sources.r1.interceptors = i1
agent1.sources.r1.interceptors.i1.type = static
agent1.sources.r1.interceptors.i1.key = logtype
agent1.sources.r1.interceptors.i1.value = ad
agent1.sources.r1.interceptors.i1.preserveExisting = false
agent1.sources.r1.channels = c1
agent1.channels.c1.type = memory
agent1.channels.c1.capacity = 10000
agent1.channels.c1.transactionCapacity = 1000
agent1.sinks.k1.type = avro
agent1.sinks.k1.channel = c1
agent1.sinks.k1.hostname = 192.168.183.100
agent1.sinks.k1.port = 8888
avro_sink2.conf
agent2.sources = r1
agent2.channels = c1
agent2.sinks = k1
agent2.sources.r1.type = netcat
agent2.sources.r1.bind = localhost
agent2.sources.r1.port = 44445
agent2.sources.r1.interceptors = i1
agent2.sources.r1.interceptors.i1.type = static
agent2.sources.r1.interceptors.i1.key = logtype
agent2.sources.r1.interceptors.i1.value = search
agent2.sources.r1.interceptors.i1.preserveExisting = false
agent2.sources.r1.channels = c1
agent2.channels.c1.type = memory
agent2.channels.c1.capacity = 10000
agent2.channels.c1.transactionCapacity = 1000
agent2.sinks.k1.type = avro
agent2.sinks.k1.channel = c1
agent2.sinks.k1.hostname = 192.168.183.100
agent2.sinks.k1.port = 8888
avro_sink3.conf
agent3.sources = r1
agent3.channels = c1
agent3.sinks = k1
agent3.sources.r1.type = netcat
agent3.sources.r1.bind = localhost
agent3.sources.r1.port = 44446
agent3.sources.r1.interceptors = i1
agent3.sources.r1.interceptors.i1.type = static
agent3.sources.r1.interceptors.i1.key = logtype
agent3.sources.r1.interceptors.i1.value = other
agent3.sources.r1.interceptors.i1.preserveExisting = false
agent3.sources.r1.channels = c1
agent3.channels.c1.type = memory
agent3.channels.c1.capacity = 10000
agent3.channels.c1.transactionCapacity = 1000
agent3.sinks.k1.type = avro
agent3.sinks.k1.channel = c1
agent3.sinks.k1.hostname = 192.168.183.100
agent3.sinks.k1.port = 8888
在/home/hadoop/apps/flume/multiplexing目录下分别创建看k1 k2 k3目录
bin/flume-ng agent --conf conf --conf-file conf/multiplexing_selector.conf --name a3 -Dflume.root.logger=INFO,console
bin/flume-ng agent --conf conf --conf-file conf/avro_sink1.conf --name agent1 >/dev/null 2>&1 &
bin/flume-ng agent --conf conf --conf-file conf/avro_sink2.conf --name agent2 >/dev/null 2>&1 &
bin/flume-ng agent --conf conf --conf-file conf/avro_sink3.conf --name agent3 >/dev/null 2>&1 &
使用telnet发送数据
telnet localhost 44444
关键参数说明：
sinks：sink组内的子Sink，多个子sink之间用空格隔开
processor.type：设置负载均衡类型load_balance
processor.backoff：设置为true时，如果在系统运行过程中执行的Sink失败，会将失败的Sink放进一个冷却池中。默认值false
processor.selector.maxTimeOut：失败sink在冷却池中最大驻留时间，默认值30000ms
processor.selector：负载均衡选择算法，可以使用轮询“round_robin”、随机“random”或者是继承AbstractSinkSelector类的自定义负载均衡实现类
示例
关键参数说明：
sinks：sink组内的子Sink，多个子sink之间用空格隔开
processor.type：设置故障转移类型“failover”
processor.priority.<sinkName>：指定Sink组内各子Sink的优先级别，优先级从高到低，数值越大优先级越高
processor.maxpenalty：等待失败的Sink恢复的最长时间，默认值30000毫秒
示例
多个agent收集不同机器上相同类型的日志数据，为了保障高可用，采用分层部署，日志收集层Collector部署两个甚至多个，Agent通过Failover SinkProcessor实现其中任何一个collector挂掉不影响系统的日志收集服务
示例
总结
原文：https://www.cnblogs.com/jeasonit/p/11580031.html