Flume-1.4和Hbase-0.96整合

最近由于项目需要把Flume收集到的日志信息插入到Hbase中，由于第一次接触这些，在整合的过程中，我遇到了许多问题，我相信很多人也应该会遇到这些问题的，于是我把整个整合的过程写出来，希望给那些同样遇到这样问题的朋友帮助。

1、本程序一共用了三台集群搭建集群，这三台机器的Hostname分别为master、node1、node2；master机器是Hadoop以及Hbase集群的master。三台机器上分别启动的进程如下：

[wyp@master ~]$ jps2973 HRegionServer4083 Jps2145 DataNode3496 HMaster2275 NodeManager1740 NameNode2790 QuorumPeerMain1895 ResourceManager[wyp@node1 ~]$ jps7801 QuorumPeerMain11669 DataNode29419 Jps11782 NodeManager29092 HRegionServer[wyp@node2 ~]$ jps2310 DataNode2726 HRegionServer2622 QuorumPeerMain3104 Jps2437 NodeManager

2、以master机器作为flume数据的源、并将数据发送给node1机器上的flume，最后node1机器上的flume将数据插入到Hbase中。master机器上的flume和node1机器上的flume中分别做如下的配置：
在master的$FLUME_HOME/conf/目录下创建以下文件（文件名随便取），并做如下配置，这是数据的发送端：

[wyp@master conf]$ vim example.confagent.sources = baksrcagent.channels = memoryChannelagent.sinks = remotesinkagent.sources.baksrc.type = execagent.sources.baksrc.command = tail -F /home/wyp/Documents/data/data.txtagent.sources.baksrc.checkperiodic = 1000agent.channels.memoryChannel.type = memoryagent.channels.memoryChannel.keep-alive = 30agent.channels.memoryChannel.capacity = 10000agent.channels.memoryChannel.transactionCapacity = 10000agent.sinks.remotesink.type = avroagent.sinks.remotesink.hostname = node1agent.sinks.remotesink.port = 23004agent.sinks.remotesink.channel = memoryChannel

在node1的$FLUME_HOME/conf/目录下创建以下文件（文件名随便取），并做如下配置，这是数据的接收端：

[wyp@node1 conf]$ vim example.confagent.sources = avrosrcagent.channels = memoryChannelagent.sinks = fileSinkagent.sources.avrosrc.type = avroagent.sources.avrosrc.bind = node1agent.sources.avrosrc.port = 23004agent.sources.avrosrc.channels = memoryChannelagent.channels.memoryChannel.type = memoryagent.channels.memoryChannel.keep-alive = 30agent.channels.memoryChannel.capacity = 10000agent.channels.memoryChannel.transactionCapacity =10000agent.sinks.fileSink.type = hbaseagent.sinks.fileSink.table = wypagent.sinks.fileSink.columnFamily = cfagent.sinks.fileSink.column = chargesagent.sinks.fileSink.serializer =                   org.apache.flume.sink.hbase.RegexHbaseEventSerializeragent.sinks.fileSink.channel = memoryChannel

3、在master机器和node1机器上分别启动flume服务进程：

[wyp@master apache-flume-1.4.0-bin]$ bin/flume-ng agent                                   --conf conf                                   --conf-file conf/example.conf                                   --name agent                                   -Dflume.root.logger=INFO,console[wyp@node1 apache-flume-1.4.0-bin]$ bin/flume-ng agent                                   --conf conf                                   --conf-file conf/example.conf                                   --name agent                                   -Dflume.root.logger=INFO,console

当分别在node1和master机器上启动上面的进程之后，在node1机器上将会输出以下的信息：

2014-01-20 22:41:56,179 (pool-3-thread-1) [INFO - org.apache.avro.ipc.NettyServer$NettyServerAvroHandler.                                      handleUpstream(NettyServer.java:171)] [id: 0x16c775c5, /192.168.142.161:42201 => /192.168.142.162:23004] OPEN2014-01-20 22:41:56,182 (pool-4-thread-1) [INFO - org.apache.avro.ipc.NettyServer$NettyServerAvroHandler.                                      handleUpstream(NettyServer.java:171)][id: 0x16c775c5, /192.168.142.161:42201 => /192.168.142.162:23004]                                       BOUND: /192.168.142.162:230042014-01-20 22:41:56,182 (pool-4-thread-1) [INFO - org.apache.avro.ipc.NettyServer$NettyServerAvroHandler.                                      handleUpstream(NettyServer.java:171)] [id: 0x16c775c5, /192.168.142.161:42201 => /192.168.142.162:23004]                                       CONNECTED: /192.168.142.161:42201

在master机器上将会输出以下的信息：

2014-01-20 22:42:16,625 (lifecycleSupervisor-1-0) [INFO - org.apache.flume.sink.AbstractRpcSink.createConnection(AbstractRpcSink.java:205)] Rpc sink remotesink: Building RpcClient with hostname: node1, port: 230042014-01-20 22:42:16,625 (lifecycleSupervisor-1-0) [INFO - org.apache.flume.sink.AvroSink.initializeRpcClient(AvroSink.java:126)] Attempting to create Avro Rpc client.2014-01-20 22:42:19,639 (lifecycleSupervisor-1-0) [INFO - org.apache.flume.sink.AbstractRpcSink.start(AbstractRpcSink.java:300)] Rpc sink remotesink started.

这样暗示node1上的flume和master上的flume已经连接成功了。
　　4、如何测试？可以写一个脚本往/home/wyp/Documents/data/data.txt（见上面master机器上flume上面的配置）文件中追加东西：

for i in {1..1000000}; do     echo "test flume to Hbase $i" >>            /home/wyp/Documents/data/data.txt;     sleep 0.1; done

运行上面的脚本，这样将每隔0.1秒往/home/wyp/Documents/data/data.txt文件中添加内容，这样master上的flume将会接收到/home/wyp/Documents/data/data.txt文件内容的变化，并变化的内容发送到node1机器上的flume，node1机器上的flume把接收到的内容插入到Hbase的wyp表中的cf:charges列中（见上面的配置）。