flume伪分布模式实践

1）启动flume master：bin/flume master ,这时你可以到master的web界面http://localhost:35871/查看flume集群信息，web界面如下：

从截图中可以看出，当前flume集群中没有node节点。

2）启动flume node节点：flume node_nowatch，你可以到http://localhost:35862/查看该节点的信息。

      附注：启动flume节点有两种方式：a）flume node：以守护进程的方式运行node，无法通过console和用户进行交互；b）flume node_nowatch：使用户能够通过console和flume node进程进行交互

3）在成功启动flume node节点之后，刷新flume master的web界面：

可以看到flume 集群中已经新添加了一个节点。

4）通过master的web界面配置node节点：

从configure node中选择所要配置的节点，然后配置节点的source和sink，点击提交查询：

提交成功后，返回到主页，可以看到相关栏目已经更新：

5）测试配置是否成功，到运行flume node_nowatch的窗口输出，配置成功：

this is testhadoop07 [INFO Mon Oct 31 07:34:21 CST 2011] this is test

6）修改配置，将hadoop07的source修改为：tail("/etc/services")，提交查询，等到十几秒之后运行flume node_nowatch的窗口输出了/etc/services当前所有的内容。再次修改source：tail("/var/log/messages")，结果node节点报错，原因是运行flume的用户无法读取文件，将文件权限改为777后正常。将sink修改为：dfs("file:///tmp/var_log_messages.copy")，dfs sink的输出为序列文件；将sink修改为：text("/tmp/var_log_messages.copy")，flume节点出现错误信息：

2011-10-31 07:55:16,721 [Check config] ERROR agent.LogicalNode: Forcing driver to exit uncleanly2011-10-31 07:55:16,723 [logicalNode hadoop07-30] WARN text.TailSource: next unexpectedly interrupted :nulljava.lang.InterruptedException        at java.util.concurrent.SynchronousQueue.poll(SynchronousQueue.java:877)        at com.cloudera.flume.handlers.text.TailSource.next(TailSource.java:271)        at com.cloudera.flume.core.connector.DirectDriver$PumperThread.run(DirectDriver.java:105)2011-10-31 07:55:16,723 [logicalNode hadoop07-30] ERROR connector.DirectDriver: Closing down due to exception during append callsjava.lang.InterruptedException        at java.util.concurrent.SynchronousQueue.poll(SynchronousQueue.java:877)        at com.cloudera.flume.handlers.text.TailSource.next(TailSource.java:271)        at com.cloudera.flume.core.connector.DirectDriver$PumperThread.run(DirectDriver.java:105)2011-10-31 07:55:16,723 [logicalNode hadoop07-30] INFO connector.DirectDriver: Connector logicalNode hadoop07-30 exited with error: nulljava.lang.InterruptedException        at java.util.concurrent.SynchronousQueue.poll(SynchronousQueue.java:877)        at com.cloudera.flume.handlers.text.TailSource.next(TailSource.java:271)        at com.cloudera.flume.core.connector.DirectDriver$PumperThread.run(DirectDriver.java:105)2011-10-31 07:55:16,725 [Check config] INFO agent.LogicalNode: Node config successfully set to com.cloudera.flume.conf.FlumeConfigData@1a5af9f2011-10-31 07:55:16,825 [TailThread-3] ERROR text.TailSource: Tail thread nterrupted: sleep interruptedjava.lang.InterruptedException: sleep interrupted        at java.lang.Thread.sleep(Native Method)        at com.cloudera.util.Clock$DefaultClock.doSleep(Clock.java:62)        at com.cloudera.util.Clock.sleep(Clock.java:88)        at com.cloudera.flume.handlers.text.TailSource$TailThread.run(TailSource.java:197)2011-10-31 07:55:16,826 [TailThread-3] INFO text.TailSource: TailThread has exited2011-10-31 07:55:16,826 [logicalNode hadoop07-30] INFO hdfs.DFSEventSink: Closing file:///tmp/var_log_messages.copy2011-10-31 07:55:16,828 [logicalNode hadoop07-30] ERROR connector.DirectDriver: Exiting driver logicalNode hadoop07-30 in error state TailSource | DFSEventSink because null

但是数据已经正确的写入到文件：/tmp/var_log_messages.copy，但文件格式不是原始的文件格式。将sink修改为：text( "/tmp/var_log_messages1.copy", "raw" )，警告：

conf.FlumeBuilder: Deprecated syntax: Expected a format spec but instead had a (String) raw

查看数据文件：/tmp/var_log_messages1.copy，文件已经和原始格式一致。

7）启动collector，注意：flume node_nowatch 会绑定35862端口，flume node_nowatch -n collector会绑定35863端口，所以查看节点信息时的端口依据于最先启动的程序。

flume node_nowatch -n collector

8）使用Configure multiple nodes进行配置，配置代码如下：

hadoop07 : console | agentSink("localhost",35853) ;collector : collectorSource(35853) | console ;

配置失败，错误信息如下：

com.cloudera.flume.conf.FlumeSpecException: Parser error: unexpected 'collector' at position 0 line 2: 'hadoop07:console|agentSink("localhost",35853); collector:collectorSource(35853)|console;'

使用Configure a single node进行配置，最终配置信息如下：

测试配置，成功。

9）最终的master web界面：