Hadoop Streaming原理及实践

Hadoop Streaming提供了一个便于进行MapReduce编程的工具包，使用它可以基于一些可执行命令、脚本语言或其他编程语言来实现Mapper和 Reducer，从而充分利用Hadoop并行计算框架的优势和能力，来处理大数据。需要注意的是，Streaming方式是基于Unix系统的标准输入输出来进行MapReduce Job的运行，它区别与Pipes的地方主要是通信协议，Pipes使用的是Socket通信，是对使用C++语言来实现MapReduce Job并通过Socket通信来与Hadopp平台通信，完成Job的执行。任何支持标准输入输出特性的编程语言都可以使用Streaming方式来实现MapReduce Job，基本原理就是输入从Unix系统标准输入，输出使用Unix系统的标准输出。
Hadoop是使用Java语言编写的，所以最直接的方式的就是使用Java语言来实现Mapper和Reducer，然后配置MapReduce Job，提交到集群计算环境来完成计算。但是很多开发者可能对Java并不熟悉，而是对一些具有脚本特性的语言，如C++、Shell、Python、 Ruby、PHP、Perl有实际开发经验，Hadoop Streaming为这一类开发者提供了使用Hadoop集群来进行处理数据的工具，即工具包hadoop-streaming-.jar。
Hadoop Streaming使用了Unix的标准输入输出作为Hadoop和其他编程语言的开发接口，因此在其他的编程语言所写的程序中，只需要将标准输入作为程 序的输入，将标准输出作为程序的输出就可以了。在标准的输入输出中，Key和Value是以Tab作为分隔符，并且在Reducer的标准输入中，Hadoop框架保证了输入的数据是经过了按Key排序的。
如何使用Hadoop Streaming工具呢？我们可以查看该工具的使用方法，通过命令行来获取，如下所示：

01	xiaoxiang@ubuntu3:~/hadoop$ bin/hadoop jar ./contrib/streaming/hadoop-streaming-1.0.3.jar -info

02	Usage: $HADOOP_HOME/bin/hadoop jar $HADOOP_HOME/hadoop-streaming.jar [options]

03

Options:

04	-input <path> DFS input file(s) for the Map step

05	-output <path> DFS output directory for the Reduce step

06	-mapper <cmd|JavaClassName> The streaming command to run

07	-combiner <cmd|JavaClassName> The streaming command to run

08	-reducer <cmd|JavaClassName> The streaming command to run

09	-file <file> File/dir to be shipped in the Job jar file

10	-inputformat TextInputFormat(default)|SequenceFileAsTextInputFormat|JavaClassName Optional.

11	-outputformat TextOutputFormat(default)|JavaClassName Optional.

12	-partitioner JavaClassName Optional.

13	-numReduceTasks <num> Optional.

14	-inputreader <spec> Optional.

15	-cmdenv <n>=<v> Optional. Pass env.var to streaming commands

16	-mapdebug <path> Optional. To run this script when a map task fails

17	-reducedebug <path> Optional. To run this script when a reduce task fails

18	-io <identifier> Optional.

19

-verbose

20

21	Generic options supported are

22	-conf <configuration file> specify an application configuration file

23	-D <property=value> use value for given property

24	-fs <local|namenode:port> specify a namenode

25	-jt <local|jobtracker:port> specify a job tracker

26	-files <comma separated list of files> specify comma separated files to be copied to the map reduce cluster

27	-libjars <comma separated list of jars> specify comma separated jar files to include in the classpath.

28	-archives <comma separated list of archives> specify comma separated archives to be unarchived on the compute machines.

29

30	The general command line syntax is

31	bin/hadoop command [genericOptions] [commandOptions]

32

33

34	In -input: globbing on <path> is supported and can have multiple -input

35	Default Map input format: a line is a record in UTF-8

36	the key part ends at first TAB, the rest of the line is the value

37	Custom input format: -inputformat package.MyInputFormat

38	Map output format, reduce input/output format:

39	Format defined by what the mapper command outputs. Line-oriented

40

41	The files named in the -file argument[s] end up in the

42	working directory when the mapper and reducer are run.

43	The location of this working directory is unspecified.

44

45	To set the number of reduce tasks (num. of output files):

46	-D mapred.reduce.tasks=10

47	To skip the sort/combine/shuffle/sort/reduce step:

48	Use -numReduceTasks 0

49	A Task's Map output then becomes a 'side-effect output' rather than a reduce input

50	This speeds up processing, This also feels more like "in-place" processing

51	because the input filename and the map input order are preserved

52	This equivalent -reducer NONE

53

54	To speed up the last maps:

55	-D mapred.map.tasks.speculative.execution=true

56	To speed up the last reduces:

57	-D mapred.reduce.tasks.speculative.execution=true

58	To name the job (appears in the JobTracker Web UI):

59	-D mapred.job.name='My Job'

60	To change the local temp directory:

61	-D dfs.data.dir=/tmp/dfs

62	-D stream.tmpdir=/tmp/streaming

63	Additional local temp directories with -cluster local:

64	-D mapred.local.dir=/tmp/local

65	-D mapred.system.dir=/tmp/system

66	-D mapred.temp.dir=/tmp/temp

67	To treat tasks with non-zero exit status as SUCCEDED:

68	-D stream.non.zero.exit.is.failure=false

69	Use a custom hadoopStreaming build along a standard hadoop install:

70	$HADOOP_HOME/bin/hadoop jar /path/my-hadoop-streaming.jar [...]\

71	[...] -D stream.shipped.hadoopstreaming=/path/my-hadoop-streaming.jar

72	For more details about jobconf parameters see:

73

74	http://wiki.apache.org/hadoop/JobConfFile

75

76	To set an environement variable in a streaming command:

77	-cmdenv EXAMPLE_DIR=/home/example/dictionaries/

78

79

Shortcut:

80	setenv HSTREAMING "$HADOOP_HOME/bin/hadoop jar $HADOOP_HOME/hadoop-streaming.jar"

81

82	Example: $HSTREAMING -mapper "/usr/local/bin/perl5 filter.pl"

83	-file /local/filter.pl -input "/logs/0604*/*" [...]

84	Ships a script, invokes the non-shipped perl interpreter

85	Shipped files go to the working directory so filter.pl is found by perl

86	Input files are all the daily logs for days in month 2006-04

面，我们分别选择几个可以使用Hadoop Streaming工具来进行计算的例子，比如对单词词频进行统计计算，即WordCount功能。
首先，我们准备测试使用的数据集，如下所示：

1	xiaoxiang@ubuntu3:~/hadoop$ bin/hadoop fs -lsr /user/xiaoxiang/dataset/join/

2	-rw-r--r-- 3 xiaoxiang supergroup 391103029 2013-03-26 12:19 /user/xiaoxiang/dataset/join/irc_basic_info.ds

3	-rw-r--r-- 3 xiaoxiang supergroup 11577164 2013-03-26 12:20 /user/xiaoxiang/dataset/join/irc_net_block.ds

4	-rw-r--r-- 3 xiaoxiang supergroup 8335235 2013-03-26 12:20 /user/xiaoxiang/dataset/join/irc_org_info.ds

一共有3个数据文件，大概将近400M大小。
下面，选择Python语言来实现MapReduce Job的运行。
使用Python实现Mapper，代码文件为word_count_mapper.py，代码如下所示：

1	#!/usr/bin/env python

2

3	import sys

4

5	for line in sys.stdin:

6	line = line.strip()

7	words = filter(lambda word: word, line.split())

8	for word in words:

9	print '%s\t%s' % (word, 1)

使用Python实现Reducer，代码文件为word_count_reducer.py，代码如下所示：

01	#!/usr/bin/env python

02

03	import sys

04	from operator import itemgetter

05

06	wc_dict = {}

07

08	for line in sys.stdin:

09	line = line.strip()

10	word, count = line.split()

11

try:

12	count = int(count)

13	wc_dict[word] = wc_dict.get(word, 0) + count

14	except ValueError:

15

pass

16

17	sorted_dict = sorted(wc_dict.items(), key=itemgetter(0))

18	for word, count in sorted_dict:

19	print '%s\t%s' % (word, count)

运行Python实现的MapReduce程序，如下所示：

01	xiaoxiang@ubuntu3:~/hadoop$ bin/hadoop jar ./contrib/streaming/hadoop-streaming-1.0.3.jar -input /user/xiaoxiang/dataset/join/ -output /user/xiaoxiang/output/streaming/python -mapper word_count_mapper.py -reducer word_count_reducer.py -numReduceTasks 2 -file *.py

02	packageJobJar: [word_count_mapper.py, word_count_reducer.py, /opt/stone/cloud/storage/tmp/hadoop-xiaoxiang/hadoop-unjar4066863202997744310/] [] /tmp/streamjob2336302975421423718.jar tmpDir=null

03	13/04/18 17:50:17 INFO util.NativeCodeLoader: Loaded the native-hadoop library

04	13/04/18 17:50:17 WARN snappy.LoadSnappy: Snappy native library not loaded

05	13/04/18 17:50:17 INFO mapred.FileInputFormat: Total input paths to process : 3

06	13/04/18 17:50:17 INFO streaming.StreamJob: getLocalDirs(): [/opt/stone/cloud/storage/mapred/local]

07	13/04/18 17:50:17 INFO streaming.StreamJob: Running job: job_201303302227_0047

08	13/04/18 17:50:17 INFO streaming.StreamJob: To kill this job, run:

09	13/04/18 17:50:17 INFO streaming.StreamJob: /opt/stone/cloud/hadoop-1.0.3/libexec/../bin/hadoop job -Dmapred.job.tracker=hdfs://ubuntu3:9001/ -killjob_201303302227_0047

10	13/04/18 17:50:17 INFO streaming.StreamJob: Tracking URL:http://ubuntu3:50030/jobdetails.jsp?jobid=job_201303302227_0047

11	13/04/18 17:50:18 INFO streaming.StreamJob: map 0% reduce 0%

12	13/04/18 17:50:33 INFO streaming.StreamJob: map 7% reduce 0%

13	13/04/18 17:50:36 INFO streaming.StreamJob: map 11% reduce 0%

14	13/04/18 17:50:39 INFO streaming.StreamJob: map 15% reduce 0%

15	13/04/18 17:50:42 INFO streaming.StreamJob: map 19% reduce 0%

16	13/04/18 17:50:45 INFO streaming.StreamJob: map 23% reduce 0%

17	13/04/18 17:50:48 INFO streaming.StreamJob: map 25% reduce 0%

18	13/04/18 17:51:09 INFO streaming.StreamJob: map 32% reduce 2%

19	13/04/18 17:51:12 INFO streaming.StreamJob: map 36% reduce 4%

20	13/04/18 17:51:15 INFO streaming.StreamJob: map 40% reduce 8%

21	13/04/18 17:51:18 INFO streaming.StreamJob: map 44% reduce 8%

22	13/04/18 17:51:21 INFO streaming.StreamJob: map 47% reduce 8%

23	13/04/18 17:51:24 INFO streaming.StreamJob: map 50% reduce 8%

24	13/04/18 17:51:45 INFO streaming.StreamJob: map 54% reduce 10%

25	13/04/18 17:51:48 INFO streaming.StreamJob: map 60% reduce 15%

26	13/04/18 17:51:51 INFO streaming.StreamJob: map 65% reduce 17%

27	13/04/18 17:51:55 INFO streaming.StreamJob: map 66% reduce 17%

28	13/04/18 17:51:58 INFO streaming.StreamJob: map 68% reduce 17%

29	13/04/18 17:52:01 INFO streaming.StreamJob: map 72% reduce 17%

30	13/04/18 17:52:04 INFO streaming.StreamJob: map 75% reduce 17%

31	13/04/18 17:52:19 INFO streaming.StreamJob: map 75% reduce 19%

32	13/04/18 17:52:22 INFO streaming.StreamJob: map 87% reduce 21%

33	13/04/18 17:52:25 INFO streaming.StreamJob: map 100% reduce 23%

34	13/04/18 17:52:28 INFO streaming.StreamJob: map 100% reduce 27%

35	13/04/18 17:52:31 INFO streaming.StreamJob: map 100% reduce 29%

36	13/04/18 17:52:37 INFO streaming.StreamJob: map 100% reduce 49%

37	13/04/18 17:52:40 INFO streaming.StreamJob: map 100% reduce 69%

38	13/04/18 17:52:43 INFO streaming.StreamJob: map 100% reduce 72%

39	13/04/18 17:52:46 INFO streaming.StreamJob: map 100% reduce 74%

40	13/04/18 17:52:49 INFO streaming.StreamJob: map 100% reduce 76%

41	13/04/18 17:52:52 INFO streaming.StreamJob: map 100% reduce 78%

42	13/04/18 17:52:55 INFO streaming.StreamJob: map 100% reduce 79%

43	13/04/18 17:52:58 INFO streaming.StreamJob: map 100% reduce 81%

44	13/04/18 17:53:01 INFO streaming.StreamJob: map 100% reduce 84%

45	13/04/18 17:53:04 INFO streaming.StreamJob: map 100% reduce 87%

46	13/04/18 17:53:07 INFO streaming.StreamJob: map 100% reduce 90%

47	13/04/18 17:53:10 INFO streaming.StreamJob: map 100% reduce 93%

48	13/04/18 17:53:13 INFO streaming.StreamJob: map 100% reduce 96%

49	13/04/18 17:53:16 INFO streaming.StreamJob: map 100% reduce 98%

50	13/04/18 17:53:19 INFO streaming.StreamJob: map 100% reduce 99%

51	13/04/18 17:53:22 INFO streaming.StreamJob: map 100% reduce 100%

52	13/04/18 17:54:10 INFO streaming.StreamJob: Job complete: job_201303302227_0047

53	13/04/18 17:54:10 INFO streaming.StreamJob: Output: /user/xiaoxiang/output/streaming/python

验证结果，如下所示：

01	xiaoxiang@ubuntu3:~/hadoop$ bin/hadoop fs -lsr /user/xiaoxiang/output/streaming/python

02	-rw-r--r-- 3 xiaoxiang supergroup 0 2013-04-18 17:54 /user/xiaoxiang/output/streaming/python/_SUCCESS

03	drwxr-xr-x - xiaoxiang supergroup 0 2013-04-18 17:50 /user/xiaoxiang/output/streaming/python/_logs

04	drwxr-xr-x - xiaoxiang supergroup 0 2013-04-18 17:50 /user/xiaoxiang/output/streaming/python/_logs/history

05	-rw-r--r-- 3 xiaoxiang supergroup 37646 2013-04-18 17:50 /user/xiaoxiang/output/streaming/python/_logs/history/job_201303302227_0047_1366278617842_xiaoxiang_streamjob2336302975421423718.jar

06	-rw-r--r-- 3 xiaoxiang supergroup 21656 2013-04-18 17:50 /user/xiaoxiang/output/streaming/python/_logs/history/job_201303302227_0047_conf.xml

07	-rw-r--r-- 3 xiaoxiang supergroup 91367389 2013-04-18 17:52 /user/xiaoxiang/output/streaming/python/part-00000

08	-rw-r--r-- 3 xiaoxiang supergroup 91268074 2013-04-18 17:52 /user/xiaoxiang/output/streaming/python/part-00001

09	xiaoxiang@ubuntu3:~/hadoop$ bin/hadoop fs -cat/user/xiaoxiang/output/streaming/python/part-00000 | head -5

10

! 2

11

# 36

12

#039 1

13

#1059) 1

14

#1098 1

相关问题

在使用Python实现MapReduce时，总是执行失败？
可以查看TaskTracker结点运行日志，可以看到，总是找不到对应的Python脚本文件，错误示例如下：

01	xiaoxiang@ubuntu1:/opt/stone/cloud/storage/logs/hadoop/userlogs/job_201303302227_0045/attempt_201303302227_0045_m_000001_0$cat stderr

02	java.io.IOException: Cannot run program"/user/xiaoxiang/streaming/python/word_count_mapper.py": java.io.IOException: error=2, No such file or directory

03	at java.lang.ProcessBuilder.start(ProcessBuilder.java:460)

04	at org.apache.hadoop.streaming.PipeMapRed.configure(PipeMapRed.java:214)

05	at org.apache.hadoop.streaming.PipeMapper.configure(PipeMapper.java:66)

06	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)

07	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)

08	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)

09	at java.lang.reflect.Method.invoke(Method.java:597)

10	at org.apache.hadoop.util.ReflectionUtils.setJobConf(ReflectionUtils.java:88)

11	at org.apache.hadoop.util.ReflectionUtils.setConf(ReflectionUtils.java:64)

12	at org.apache.hadoop.util.ReflectionUtils.newInstance(ReflectionUtils.java:117)

13	at org.apache.hadoop.mapred.MapRunner.configure(MapRunner.java:34)

14	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)

15	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)

16	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)

17	at java.lang.reflect.Method.invoke(Method.java:597)

18	at org.apache.hadoop.util.ReflectionUtils.setJobConf(ReflectionUtils.java:88)

19	at org.apache.hadoop.util.ReflectionUtils.setConf(ReflectionUtils.java:64)

20	at org.apache.hadoop.util.ReflectionUtils.newInstance(ReflectionUtils.java:117)

21	at org.apache.hadoop.mapred.MapTask.runOldMapper(MapTask.java:432)

22	at org.apache.hadoop.mapred.MapTask.run(MapTask.java:372)

23	at org.apache.hadoop.mapred.Child$4.run(Child.java:255)

24	at java.security.AccessController.doPrivileged(Native Method)

25	at javax.security.auth.Subject.doAs(Subject.java:396)

26	at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1121)

27	at org.apache.hadoop.mapred.Child.main(Child.java:249)

28	Caused by: java.io.IOException: java.io.IOException: error=2, No such file or directory

29	at java.lang.UNIXProcess.<init>(UNIXProcess.java:148)

30	at java.lang.ProcessImpl.start(ProcessImpl.java:65)

31	at java.lang.ProcessBuilder.start(ProcessBuilder.java:453)

32	... 24 more

可以使用Streaming的-file选项指定脚本文件加入到Job的Jar文件中，即使用上面运行的命令行中指定的“-file *.py”， 而实现的2个Python脚本文件就在当前运行Job的结点当前目录下。