业务系统需要收集监控系统日志,想到了hadoop的flume。经过试验,虽说功能不算足够强大,但基本上能够满足功能需求。Flume 是一个分布式、可靠和高可用的服务日志收集工具,能够和hadoop,hive等配置完成日志收集,存储,分析处理等工作,更详细的介绍可以参见apache网站。下面介绍下简单的安装配置方法
1,网上下载flume-ng安装包,分别部署在收集和接收日志文件的服务器上,服务器上需安装jdk 1.6以上,
http://flume.apache.org/download.html
tar -zxvf apache-flume-1.3.0-bin.tar.gz
2, 日志文件接收端端新建conf/flume-conf.properties server端的具体配置如下
如果flume和hadoop不是同一用户,需要注意相关权限问题
3,日志收集端的conf/flume-conf.properties server文件配置,这里收集二个日志文件到收集端
4,后台运行
nohup bin/flume-ng agent -n agent -c conf -f conf/flume-conf.properties >1.log &
查看日志vi flume.log
端口连接情况 netstat -an|grep 44444
[flume@dtydb6 flume-1.4]$ netstat -an|grep 44444
tcp 0 0 ::ffff:172.16.251.1:44444 :::* LISTEN
http://flume.apache.org/FlumeUserGuide.html
1,网上下载flume-ng安装包,分别部署在收集和接收日志文件的服务器上,服务器上需安装jdk 1.6以上,
http://flume.apache.org/download.html
tar -zxvf apache-flume-1.3.0-bin.tar.gz
2, 日志文件接收端端新建conf/flume-conf.properties server端的具体配置如下
从avro source端接收数据,然后写入到HDFS文件系统中
- [flume@ conf]$ cat flume-conf.properties
- agent.sources = avrosrc
- agent.channels = memoryChanne3
- agent.sinks = hdfsSink
- # For each one of the sources, the type is defined
- agent.sources.avrosrc.type = avro
- agent.sources.avrosrc.bind = 172.16.251.1
- agent.sources.avrosrc.port = 44444
- # The channel can be defined as follows.
- agent.sources.avrosrc.channels = memoryChanne3
- # Each channel's type is defined.
- agent.channels.memoryChanne3.type = memory
- agent.channels.memoryChanne3.keep-alive = 10
- agent.channels.memoryChanne3.capacity = 100000
- agent.channels.memoryChanne3.transactionCapacity =100000
- # Each sink's type must be defined
- agent.sinks.hdfsSink.type = hdfs
- agent.sinks.hdfsSink.channel = memoryChanne3
- agent.sinks.hdfsSink.hdfs.path = /logdata/%{hostname}_linux/%Y%m%d_date
- agent.sinks.hdfsSink.hdfs.filePrefix = %{datacenter}_
- agent.sinks.hdfsSink.hdfs.rollInterval = 0
- agent.sinks.hdfsSink.hdfs.rollSize = 4000000
- agent.sinks.hdfsSink.hdfs.rollCount = 0
- agent.sinks.hdfsSink.hdfs.writeFormat = Text
- agent.sinks.hdfsSink.hdfs.fileType = DataStream
- agent.sinks.hdfsSink.hdfs.batchSize = 10
如果flume和hadoop不是同一用户,需要注意相关权限问题
3,日志收集端的conf/flume-conf.properties server文件配置,这里收集二个日志文件到收集端
- agent.sources = tailsource-1 tailsource-2
- agent.channels = memoryChannel-1 memoryChannel-2
- agent.sinks = remotesink remotesink-2
- agent.sources.tailsource-1.type = exec
- agent.sources.tailsource-1.command = tail -F /tmp/linux2.log
- agent.sources.tailsource-1.channels = memoryChannel-1
- agent.sources.tailsource-2.type = exec
- agent.sources.tailsource-2.command = tail -F /tmp/linux2_2.log
- agent.sources.tailsource-2.channels = memoryChannel-2
- agent.sources.tailsource-1.interceptors = host_int timestamp_int inter1
- agent.sources.tailsource-1.interceptors.host_int.type = host
- agent.sources.tailsource-1.interceptors.host_int.hostHeader = hostname
- agent.sources.tailsource-1.interceptors.timestamp_int.type = org.apache.flume.interceptor.TimestampInterceptor$Builder
- #agent.sources.tailsource-1.interceptors = inter1
- agent.sources.tailsource-1.interceptors.inter1.type = static
- agent.sources.tailsource-1.interceptors.inter1.key = datacenter
- agent.sources.tailsource-1.interceptors.inter1.value = BEIJING
- agent.sources.tailsource-2.interceptors = host_int timestamp_int inter1
- agent.sources.tailsource-2.interceptors.host_int.type = host
- agent.sources.tailsource-2.interceptors.host_int.hostHeader = hostname
- agent.sources.tailsource-2.interceptors.timestamp_int.type = org.apache.flume.interceptor.TimestampInterceptor$Builder
- #agent.sources.tailsource-1.interceptors = inter1
- agent.sources.tailsource-2.interceptors.inter1.type = static
- agent.sources.tailsource-2.interceptors.inter1.key = datacenter
- agent.sources.tailsource-2.interceptors.inter1.value = linux2_2
- agent.channels.memoryChannel-1.type = memory
- agent.channels.memoryChannel-1.keep-alive = 10
- agent.channels.memoryChannel-1.capacity = 100000
- agent.channels.memoryChannel-1.transactionCapacity =100000
- agent.channels.memoryChannel-2.type = memory
- agent.channels.memoryChannel-2.keep-alive = 10
- agent.channels.memoryChannel-2.capacity = 100000
- agent.channels.memoryChannel-2.transactionCapacity =100000
- agent.sinks.remotesink.type = avro
- agent.sinks.remotesink.hostname = 172.16.251.1
- agent.sinks.remotesink.port = 44444
- agent.sinks.remotesink.channel = memoryChannel-1
- agent.sinks.remotesink-2.type = avro
- agent.sinks.remotesink-2.hostname = 172.16.251.1
- agent.sinks.remotesink-2.port = 44444
- agent.sinks.remotesink-2.channel = memoryChannel-2
4,后台运行
nohup bin/flume-ng agent -n agent -c conf -f conf/flume-conf.properties >1.log &
查看日志vi flume.log
端口连接情况 netstat -an|grep 44444
[flume@dtydb6 flume-1.4]$ netstat -an|grep 44444
tcp 0 0 ::ffff:172.16.251.1:44444 :::* LISTEN
5,测试方法
可以使用如下类似的脚本,定期向日志文件写入来进行测试
for i in {1..1000000}; do echo "LINUX2 PRESS ************* Flume log rotation $i" >> /tmp/linux3.log; sleep 0.0001; done
flume-ng-1.4.0-cdh4.4.0.tar.gz参考资料:
http://flume.apache.org/FlumeUserGuide.html