Hadoop实战2:MapReduce编程-WordCount实例-streaming-python环境

　　这是搭建hadoop环境后的第一个MapReduce程序；

　　基于hadoop streaming的python的脚本；

　　1 map.py文件，把文本的内容划分成单词：

#!/usr/bin/pythonimport sys

for line in sys.stdin:    line = line.strip()    words = line.split()    for word in words:        print('%s\t%s' % (word, 1))

　　

　　2 reduce文件，把统计单词出现的次数；

#!/usr/bin/pythonimport sys

last_key = Nonerunning_total = 0

for input_line in sys.stdin:    input_line = input_line.strip()    this_key, value = input_line.split("\t", 1)    value = int(value)

    if last_key == this_key:        running_total += value    else:        if last_key:            print ("%s\t%d" % (last_key, running_total))        running_total = value        last_key = this_keyif last_key == this_key:    print( "%s\t%d" % (last_key, running_total) )
        

　　

　　3 本地测试下python脚本，结果是否正确：

cat in.txt | python map.py | python reduce.py

　　4 Hadoop调用脚本：指定输出目录OUTPUT；

　　调用支持多语言的streaming的编程环境，参数-input是输入的log文件，为了用mapreduce模式统计这个文件每个单词出现的次数；-output是输出路径；-mapper是mapper编译 此处是python语言；-reducer是reduce编译语法；-file是mapper文件路径和reduce文件路径；-numReduceTaskers 是使用的子tasker数目，这里是3，代表分成了3了tasker分布式的处理计数任务；

#!/bin/bash

OUTPUT=/home/apm3/outdir
hadoop fs -rmr $OUTPUT
hadoop jar /usr/local/hadoop/share/hadoop/tools/lib/hadoop-streaming-.jar \
-input /opt/mapr/logs/warden.log \
-output $OUTPUT \
-mapper "python map.py" \
-reducer "python reduce.py" \
-file map.py \
-file reduce.py \
-numReduceTasks
 

　　bash -x start.sh 会在输出路径中生成三个输出文件，及三分ReduceTasks 输出的结果；（MapReduce 模式主要做了shuffle和sort任务，shuffle是按照hashkey分配单词到子tasker中，而sort是排序的功能。）

　　5 MapR里执行程序，run.sh:

hadoop fs -rm -r /user/rongyu/output

hadoop jar hadoop-streaming-2.7.0-mapr-1602.jar \-input "/user/input/*" \-output "/user/rongyu/output" \-file "/home/mapr/Develop/rongyu/mapreduce/map.py"-mapper "python map.py" \-file "/home/mapr/Develop/rongyu/mapreduce/reduce.py"-reducer "python reduce.py" \-numReduceTasks 3

　　6 查看结果

　　查看输出目录： 命令 $ hadoop fs -ls /user/rongyu/output/

Found  items
-rwxr-xr-x    mapr mapr           -- : /user/rongyu/output/_SUCCESS
-rwxr-xr-x    mapr mapr     -- : /user/rongyu/output/part-
-rwxr-xr-x    mapr mapr     -- : /user/rongyu/output/part-
-rwxr-xr-x    mapr mapr     -- : /user/rongyu/output/part-

　　输出三个输出文件之一part-00000：命令 $ hadoop fs -cat /user/rongyu/output/part-00000 | less

/nodes/apm1/services/nfs        17/opt/mapr/conf/cldb.conf        12/opt/mapr/hostid        6/services/cldb/master.  4/services/fileserver.   2/services/fileserver/master     1/services/hbmaster/apm2.        1/services/hbregionserver/apm4.  207/services/hbregionserver/master 1/services/historyserver/master  1/services/hoststats/apm2.       2/services/kvstore/apm3. 2/services/nfs.  22/services/nfs/master.   53/services_config/kvstore.       2/services_config/nodemanager.   3/services_config/nodemanager/apm4.      2600:00:00,3402   100:00:00,4710   100:00:01,6710   100:00:01,7916   100:00:01,9725   1

　　7异常：

// :: INFO mapreduce.Job: Task Id : attempt_1469682745105_0016_m_000001_2, Status : FAILED
Error: java.lang.RuntimeException: PipeMapRed.waitOutputThreads(): subprocess failed with code
    at org.apache.hadoop.streaming.PipeMapRed.waitOutputThreads(PipeMapRed.java:)
    at org.apache.hadoop.streaming.PipeMapRed.mapRedFinished(PipeMapRed.java:)
    at org.apache.hadoop.streaming.PipeMapper.close(PipeMapper.java:)
    at org.apache.hadoop.mapred.MapRunner.run(MapRunner.java:)
    at org.apache.hadoop.streaming.PipeMapRunner.run(PipeMapRunner.java:)
    at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:)
    at org.apache.hadoop.mapred.YarnChild$.run(YarnChild.java:)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:)
    at org.apache.hadoop.mapred.YarnChild.main(YarnChild.java:)

　　解决方案：在python脚本头部增加 #!/usr/bin/python  并且注意run.sh的-reducer -mapper等参数设置

　　代码下载： https://github.com/rongyux/Hadoop_WordCount