[[PageOutline]]

◢ <[wiki:III141129/Lab4 實作四]> | <[wiki:III141129 回課程大綱]> ▲ | <[wiki:III141129/Lab6 實作六]> ◣

= 實作五 Lab 5 =

{{{
#!html
<div style="text-align: center;"><big style="font-weight: bold;"><big>在單機模式執行 MapReduce 基本運算<br/>Running MapReduce in local mode by Examples</big></big></div>
}}}

{{{
#!text
以下練習，請在本機的 Hadoop4Win 環境操作。
}}}

== 範例一『字數統計(WordCount)』 ==

 * STEP 1 : 練習 MapReduce 丟 Job 指令: 『__'''hadoop jar <local jar file> <class name> <parameters>'''__』
{{{
Jazz@human ~
$ cd /opt/hadoop/

Jazz@human /opt/hadoop
$ hadoop jar hadoop-*-examples.jar wordcount input output
11/10/21 14:08:58 INFO input.FileInputFormat: Total input paths to process : 12
11/10/21 14:09:00 INFO mapred.JobClient: Running job: job_201110211130_0001
11/10/21 14:09:01 INFO mapred.JobClient:  map 0% reduce 0%
11/10/21 14:09:31 INFO mapred.JobClient:  map 16% reduce 0%
11/10/21 14:10:29 INFO mapred.JobClient:  map 100% reduce 27%
11/10/21 14:10:33 INFO mapred.JobClient:  map 100% reduce 100%
11/10/21 14:10:35 INFO mapred.JobClient: Job complete: job_201110211130_0001
11/10/21 14:10:35 INFO mapred.JobClient: Counters: 17
11/10/21 14:10:35 INFO mapred.JobClient:   Job Counters
11/10/21 14:10:35 INFO mapred.JobClient:     Launched reduce tasks=1
11/10/21 14:10:35 INFO mapred.JobClient:     Launched map tasks=12
11/10/21 14:10:35 INFO mapred.JobClient:     Data-local map tasks=12
11/10/21 14:10:35 INFO mapred.JobClient:   FileSystemCounters
11/10/21 14:10:35 INFO mapred.JobClient:     FILE_BYTES_READ=16578
11/10/21 14:10:35 INFO mapred.JobClient:     HDFS_BYTES_READ=18312
11/10/21 14:10:35 INFO mapred.JobClient:     FILE_BYTES_WRITTEN=32636
11/10/21 14:10:35 INFO mapred.JobClient:     HDFS_BYTES_WRITTEN=10922
11/10/21 14:10:35 INFO mapred.JobClient:   Map-Reduce Framework
11/10/21 14:10:35 INFO mapred.JobClient:     Reduce input groups=592
11/10/21 14:10:35 INFO mapred.JobClient:     Combine output records=750
11/10/21 14:10:35 INFO mapred.JobClient:     Map input records=553
11/10/21 14:10:35 INFO mapred.JobClient:     Reduce shuffle bytes=15674
11/10/21 14:10:35 INFO mapred.JobClient:     Reduce output records=592
11/10/21 14:10:35 INFO mapred.JobClient:     Spilled Records=1500
11/10/21 14:10:35 INFO mapred.JobClient:     Map output bytes=24438
11/10/21 14:10:35 INFO mapred.JobClient:     Combine input records=1755
11/10/21 14:10:35 INFO mapred.JobClient:     Map output records=1755
11/10/21 14:10:35 INFO mapred.JobClient:     Reduce input records=750
}}}
   * [[BR]][[Image(Hadoop4Win:hadoop4win_14.jpg,width=600)]]

 * STEP 2 : 練習從 http://localhost:50030 查看目前 MapReduce Job 的運作情形
   * [[BR]][[Image(Hadoop4Win:hadoop4win_15.jpg,width=600)]]

 * STEP 3 : 使用 HDFS 指令: 『__'''hadoop fs -get <HDFS file/dir> <local file/dir>'''__』，並了解輸出檔案檔名均為 part-r-*****，且執行參數會紀錄於 <HOSTNAME>_<TIME>_job_<JOBID>_0001_conf.xml，不妨可以觀察 xml 內容與 hadoop config 檔的參數關聯。
{{{
Jazz@human /opt/hadoop
$ hadoop fs -get output my_output

Jazz@human /opt/hadoop
$ ls -alR my_output
my_output:
total 12
drwxr-xr-x+  3 Jazz None     0 Oct 21 14:12 .
drwxr-xr-x+ 15 Jazz None     0 Oct 21 14:12 ..
drwxr-xr-x+  3 Jazz None     0 Oct 21 14:12 _logs
-rwxr-xr-x   1 Jazz None 10922 Oct 21 14:12 part-r-00000

my_output/_logs:
total 0
drwxr-xr-x+ 3 Jazz None 0 Oct 21 14:12 .
drwxr-xr-x+ 3 Jazz None 0 Oct 21 14:12 ..
drwxr-xr-x+ 2 Jazz None 0 Oct 21 14:12 history

my_output/_logs/history:
total 48
drwxr-xr-x+ 2 Jazz None     0 Oct 21 14:12 .
drwxr-xr-x+ 3 Jazz None     0 Oct 21 14:12 ..
-rwxr-xr-x  1 Jazz None 26004 Oct 21 14:12 localhost_1319167815125_job_201110211130_0001_Jazz_word+count
-rwxr-xr-x  1 Jazz None 16984 Oct 21 14:12 localhost_1319167815125_job_201110211130_0001_conf.xml
}}}
   * [[BR]][[Image(Hadoop4Win:hadoop4win_22.jpg,width=600)]]

 * 比較熟悉 Windows 的,也可以用 cgystart my_output 用檔案總管開啟下載的內容
{{{
Jazz@human /opt/hadoop
$ cgystart my_output
}}}

 * 也可以透過 NameNode 網頁介面觀看結果
   * http://localhost:50075/browseDirectory.jsp?dir=/user&namenodeInfoPort=50070

== 範例二『用標準表示法過濾內容 grep』 ==

 * grep 這個命令是擷取文件裡面特定的字元，在 Hadoop example 中此指令可以擷取文件中有此指定文字的字串，並作計數統計[[BR]]grep is a command to extract specific characters in documents. In hadoop examples, you can use this command to extract strings match the regular expression and count for matched strings.
{{{
Jazz@human /opt/hadoop
$ hadoop jar hadoop-*-examples.jar  grep input lab5_out1 'dfs[a-z.]+'
}}}
 * 運作的畫面如下：[[BR]]You should see procedure like this: 
{{{
Jazz@human /opt/hadoop
$ hadoop jar hadoop-*-examples.jar  grep input lab5_out1 'dfs[a-z.]+'
11/10/21 14:17:39 INFO mapred.FileInputFormat: Total input paths to process : 12

11/10/21 14:17:39 INFO mapred.JobClient: Running job: job_201110211130_0002
11/10/21 14:17:40 INFO mapred.JobClient:  map 0% reduce 0%
11/10/21 14:17:54 INFO mapred.JobClient:  map 8% reduce 0%
11/10/21 14:17:57 INFO mapred.JobClient:  map 16% reduce 0%
11/10/21 14:18:03 INFO mapred.JobClient:  map 33% reduce 0%
11/10/21 14:18:13 INFO mapred.JobClient:  map 41% reduce 0%
11/10/21 14:18:16 INFO mapred.JobClient:  map 50% reduce 11%
11/10/21 14:18:19 INFO mapred.JobClient:  map 58% reduce 11%
11/10/21 14:18:23 INFO mapred.JobClient:  map 66% reduce 11%
11/10/21 14:18:30 INFO mapred.JobClient:  map 83% reduce 16%
11/10/21 14:18:33 INFO mapred.JobClient:  map 83% reduce 22%
11/10/21 14:18:36 INFO mapred.JobClient:  map 91% reduce 22%
11/10/21 14:18:39 INFO mapred.JobClient:  map 100% reduce 22%
11/10/21 14:18:42 INFO mapred.JobClient:  map 100% reduce 27%
11/10/21 14:18:48 INFO mapred.JobClient:  map 100% reduce 30%
11/10/21 14:18:54 INFO mapred.JobClient:  map 100% reduce 100%
11/10/21 14:18:56 INFO mapred.JobClient: Job complete: job_201110211130_0002
11/10/21 14:18:56 INFO mapred.JobClient: Counters: 18
11/10/21 14:18:56 INFO mapred.JobClient:   Job Counters
11/10/21 14:18:56 INFO mapred.JobClient:     Launched reduce tasks=1
11/10/21 14:18:56 INFO mapred.JobClient:     Launched map tasks=12
11/10/21 14:18:56 INFO mapred.JobClient:     Data-local map tasks=12
11/10/21 14:18:56 INFO mapred.JobClient:   FileSystemCounters
11/10/21 14:18:56 INFO mapred.JobClient:     FILE_BYTES_READ=888
11/10/21 14:18:56 INFO mapred.JobClient:     HDFS_BYTES_READ=18312
11/10/21 14:18:56 INFO mapred.JobClient:     FILE_BYTES_WRITTEN=1496
11/10/21 14:18:56 INFO mapred.JobClient:     HDFS_BYTES_WRITTEN=280
11/10/21 14:18:56 INFO mapred.JobClient:   Map-Reduce Framework
11/10/21 14:18:56 INFO mapred.JobClient:     Reduce input groups=7
11/10/21 14:18:56 INFO mapred.JobClient:     Combine output records=7
11/10/21 14:18:56 INFO mapred.JobClient:     Map input records=553
11/10/21 14:18:56 INFO mapred.JobClient:     Reduce shuffle bytes=224
11/10/21 14:18:56 INFO mapred.JobClient:     Reduce output records=7
11/10/21 14:18:56 INFO mapred.JobClient:     Spilled Records=14
11/10/21 14:18:56 INFO mapred.JobClient:     Map output bytes=193
11/10/21 14:18:56 INFO mapred.JobClient:     Map input bytes=18312
11/10/21 14:18:56 INFO mapred.JobClient:     Combine input records=10
11/10/21 14:18:56 INFO mapred.JobClient:     Map output records=10
11/10/21 14:18:56 INFO mapred.JobClient:     Reduce input records=7
11/10/21 14:18:56 WARN mapred.JobClient: Use GenericOptionsParser for parsing th
e arguments. Applications should implement Tool for the same.
11/10/21 14:18:57 INFO mapred.FileInputFormat: Total input paths to process : 1
11/10/21 14:18:57 INFO mapred.JobClient: Running job: job_201110211130_0003
( ... skip ... )
}}}
 * 接著查看結果[[BR]]Let's check the computed result of '''grep''' from HDFS : 
 * 這個例子是要從 input 目錄中的所有檔案中找出符合 dfs 後面接著 a-z 字母一個以上的字串
{{{
Jazz@human /opt/hadoop
$ hadoop fs -ls lab5_out1
Found 2 items
drwxr-xr-x   - Jazz supergroup          0 2011-10-21 14:18 /user/Jazz/lab5_out1/_logs
-rw-r--r--   1 Jazz supergroup         96 2011-10-21 14:19 /user/Jazz/lab5_out1/part-00000

Jazz@human /opt/hadoop
$ hadoop fs -cat lab5_out1/part-00000
3       dfs.class
2       dfs.period
1       dfs.file
1       dfs.replication
1       dfs.servers
1       dfsadmin
1       dfsmetrics.log
}}}