Hadoop 2.0.0-alpha尝鲜安装和hello world

仅供测试学习的文章，不推荐在生产环境使用2.0，因为2.0采用YARN，hive，hbase，mahout等需要map/reduceV1的可能无法使用hadoop 2.0或者会出现意外情况。

5月23日，apache发布了hadoop 2.0的测试版。正好跟家呆着没事干，小小的体会了一下map/reduce V2。

环境，virtual box虚拟机ubuntu server 12.04，openjdk-7。

简单介绍一下，2.0.0是从hadoop 0.23.x发展出来的。取消了jobtracker和tasktracker，或者说，是把这两个封装到了container里面。使用YARN替代了原来的map/reduce。

YARN号称是第二代map/reduce，速度比一代更快，且支持集群服务器数量更大。hadoop 0.20.x和由其发展过来的1.0.x支持集群数量建议在3000台左右，最大支持到4000台。而hadoop 2.0和YARN宣称支持6000-10000台，CPU核心数支持200000颗。从集群数量和运算能力上说，似乎还是提高了不少的。并且加入了namenode的HA，也就是高可用。我说似乎，因为没有在实际生产环境测试速度。而namenode的HA，因为是虚拟机测试，也就没有测试。只是简单的看了一下。

2.0的文件结构相比1.0有所变化，更加清晰明了了。可执行文件在bin/下，server启动放到了sbin/下，map/red,streaming,pipes的jar包放到了share/下。很容易找到。

安装包解压缩后，先进入etc/hadoop/目录下，按照单机版方式配置几个配置文件。有core-site.xml，hdfs-site.xml，但是没有了mapred-site.xml，取而代之的是yarn-site.xml

假设已经按照单机配置配好了，那么进入$HADOOP_HOME/bin/目录下

执行如下

./hadoop namenode -format
#先格式化
cd ../sbin/
#进入sbin目录，这里放的都是server启动脚本
./hadoop-daemon.sh start namenode
./hadoop-daemon.sh start datanode
./hadoop-daemon.sh start secondarynamenode
#备份服起不起都无所谓，不影响使用，不过可以用来试试HA功能
#下面较重要，2.0取消了jobtracker和tasktracker，以YARN来代替，所以如果运行start jobtracker一类的，会报错。
#且hadoop,hdfs,map/reduce功能都分离出了单独脚本，所以不能用hadoop-daemon.sh启动所有了。
./yarn-daemon.sh start resourcemanager
#这个就相当于原来的jobtracker，用作运算资源分配的进程，跟namenode可放在一起。
./yarn-daemon.sh start nodemanager
#这个相当于原来的tasktracker，每台datanode或者叫slave的服务器上都要启动。

ps aux一下，如果看到4个java进程，就算启动成功了，访问http://localhost:50070看看hdfs情况。且由于取消了jobtracker，所以也就没有50030端口来查看任务情况了，这个以后再说吧。

然后来试试编写第一个map/reduce V2的程序。其实从程序的编写方式来说跟V1没有任何区别，只是最后调用方式变化了一下。hadoop 2.0为了保证兼容性，用户接口方面对于用户来说，还是跟原来是一样的。

这样一段数据

20120503 04 2012-05-03 04:49:22 222.139.35.72 Log_ASF ProductVer="5.12.0425.2111"

20120503 04 2012-05-03 04:49:21 113.232.38.239 Log_ASF ProductVer="5.09.0119.1112"

假设就2条不一样的吧，一共20条。

还是用python来写map/red脚本

#!/usr/bin/python
#-*- encoding:UTF-8 -*-
#map.py
import sys

debug = True
if debug:
                lzo = 0
else:
                lzo = 1

count='0'
for line in sys.stdin:
                try:
                                flags = line[:-1].split('\t')
                                if len(flags) == 0:
                                                break
                                if len(flags) != 5+lzo:
                                                continue

                                stat_date = flags[2+lzo].split(' ')[0]
                                version = flags[5+lzo].split('"')[1]

                                str = stat_date+','+version+'\t'+count
                                print str

                except Exception,e:
                                print e

------------------------------------------------------------------

#!/usr/bin/python
#-*- encoding:UTF-8 -*-
#reduce.py
import sys

import string

res = {}
#声明字典

for line in sys.stdin:
                try:
                                flags = line[:-1].split('\t')
                                if len(flags) != 2:
                                                continue
                                field_key = flags[0]
                                if res.has_key(field_key) == False:
                                                res[field_key] = 0
                                res[field_key] += 1
                except Exception,e:
                                pass

for key in res.keys():
                print key+','+'%s' % (res[key])

然后把范例数据复制到hdfs上面用

./hadoop fs -mkdir /tmp
./hadoop fs -copyFromLocal /root/asf /tmp/asf

测试一下，还跟以前hadoop一样。不过两种streaming的方式都可以

./hadoop jar /opt/hadoop/share/hadoop/tools/lib/hadoop-streaming-2.0.0-alpha.jar -mapper /opt/hadoop/mrs/map.py -reducer /opt/hadoop/mrs/red.py -input /tmp/asf -output /asf

或者

./yarn jar /opt/hadoop/share/hadoop/tools/lib/hadoop-streaming-2.0.0-alpha.jar -mapper /opt/hadoop/mrs/map.py -reducer /opt/hadoop/mrs/red.py -input /tmp/asf -output /asf

然后

./hadoop fs -cat /asf/part-00000文件

2012-05-03,5.09.0119.1112,2

2012-05-03,5.12.0425.2111,18

结果正确。

附map/reduce V2执行日志：

root@localhost:/opt/hadoop/bin# ./yarn jar /opt/hadoop/share/hadoop/tools/lib/hadoop-streaming-2.0.0-alpha.jar -mapper /opt/hadoop/mrs/map.py -reducer /opt/hadoop/mrs/red.py -input /tmp/asf -output /asf
12/06/01 23:26:40 WARN util.KerberosName: Kerberos krb5 configuration not found, setting default realm to empty
12/06/01 23:26:41 WARN conf.Configuration: session.id is deprecated. Instead, use dfs.metrics.session-id
12/06/01 23:26:41 INFO jvm.JvmMetrics: Initializing JVM Metrics with processName=JobTracker, sessionId=
12/06/01 23:26:41 INFO jvm.JvmMetrics: Cannot initialize JVM Metrics with processName=JobTracker, sessionId= - already initialized
12/06/01 23:26:41 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
12/06/01 23:26:42 WARN snappy.LoadSnappy: Snappy native library not loaded
12/06/01 23:26:42 INFO mapred.FileInputFormat: Total input paths to process : 1
12/06/01 23:26:42 INFO mapreduce.JobSubmitter: number of splits:1
12/06/01 23:26:42 WARN conf.Configuration: mapred.jar is deprecated. Instead, use mapreduce.job.jar
12/06/01 23:26:42 WARN conf.Configuration: mapred.create.symlink is deprecated. Instead, use mapreduce.job.cache.symlink.create
12/06/01 23:26:42 WARN conf.Configuration: mapred.job.name is deprecated. Instead, use mapreduce.job.name
12/06/01 23:26:42 WARN conf.Configuration: mapred.input.dir is deprecated. Instead, use mapreduce.input.fileinputformat.inputdir
12/06/01 23:26:42 WARN conf.Configuration: mapred.output.dir is deprecated. Instead, use mapreduce.output.fileoutputformat.outputdir
12/06/01 23:26:42 WARN conf.Configuration: mapred.map.tasks is deprecated. Instead, use mapreduce.job.maps
12/06/01 23:26:42 WARN conf.Configuration: mapred.output.value.class is deprecated. Instead, use mapreduce.job.output.value.class
12/06/01 23:26:42 WARN conf.Configuration: mapred.output.key.class is deprecated. Instead, use mapreduce.job.output.key.class
12/06/01 23:26:42 WARN conf.Configuration: mapred.mapoutput.value.class is deprecated. Instead, use mapreduce.map.output.value.class
12/06/01 23:26:42 WARN conf.Configuration: mapred.mapoutput.key.class is deprecated. Instead, use mapreduce.map.output.key.class
12/06/01 23:26:42 WARN conf.Configuration: mapred.working.dir is deprecated. Instead, use mapreduce.job.working.dir
12/06/01 23:26:42 WARN mapred.LocalDistributedCacheManager: LocalJobRunner does not support symlinking into current working dir.
12/06/01 23:26:42 INFO mapreduce.Job: The url to track the job: http://localhost:8080/
12/06/01 23:26:42 INFO mapreduce.Job: Running job: job_local_0001
12/06/01 23:26:42 INFO mapred.LocalJobRunner: OutputCommitter set in config null
12/06/01 23:26:42 INFO mapred.LocalJobRunner: OutputCommitter is org.apache.hadoop.mapred.FileOutputCommitter
12/06/01 23:26:42 INFO mapred.LocalJobRunner: Waiting for map tasks
12/06/01 23:26:42 INFO mapred.LocalJobRunner: Starting task: attempt_local_0001_m_000000_0
12/06/01 23:26:42 INFO mapred.Task:    Using ResourceCalculatorPlugin : org.apache.hadoop.yarn.util.LinuxResourceCalculatorPlugin@52b5ef94
12/06/01 23:26:42 INFO mapred.MapTask: numReduceTasks: 1
12/06/01 23:26:42 INFO mapred.MapTask: (EQUATOR) 0 kvi 26214396(104857584)
12/06/01 23:26:42 INFO mapred.MapTask: mapreduce.task.io.sort.mb: 100
12/06/01 23:26:42 INFO mapred.MapTask: soft limit at 83886080
12/06/01 23:26:42 INFO mapred.MapTask: bufstart = 0; bufvoid = 104857600
12/06/01 23:26:42 INFO mapred.MapTask: kvstart = 26214396; length = 6553600
12/06/01 23:26:42 INFO streaming.PipeMapRed: PipeMapRed exec [/opt/hadoop/mrs/map.py]
12/06/01 23:26:42 WARN conf.Configuration: mapred.task.id is deprecated. Instead, use mapreduce.task.attempt.id
12/06/01 23:26:42 WARN conf.Configuration: user.name is deprecated. Instead, use mapreduce.job.user.name
12/06/01 23:26:42 WARN conf.Configuration: map.input.start is deprecated. Instead, use mapreduce.map.input.start
12/06/01 23:26:42 WARN conf.Configuration: mapred.task.is.map is deprecated. Instead, use mapreduce.task.ismap
12/06/01 23:26:42 WARN conf.Configuration: mapred.tip.id is deprecated. Instead, use mapreduce.task.id
12/06/01 23:26:42 WARN conf.Configuration: mapred.skip.on is deprecated. Instead, use mapreduce.job.skiprecords
12/06/01 23:26:42 WARN conf.Configuration: mapred.task.partition is deprecated. Instead, use mapreduce.task.partition
12/06/01 23:26:42 WARN conf.Configuration: map.input.length is deprecated. Instead, use mapreduce.map.input.length
12/06/01 23:26:42 WARN conf.Configuration: mapred.local.dir is deprecated. Instead, use mapreduce.cluster.local.dir
12/06/01 23:26:42 WARN conf.Configuration: mapred.work.output.dir is deprecated. Instead, use mapreduce.task.output.dir
12/06/01 23:26:42 WARN conf.Configuration: map.input.file is deprecated. Instead, use mapreduce.map.input.file
12/06/01 23:26:42 WARN conf.Configuration: mapred.job.id is deprecated. Instead, use mapreduce.job.id
12/06/01 23:26:43 INFO streaming.PipeMapRed: R/W/S=1/0/0 in:NA [rec/s] out:NA [rec/s]
12/06/01 23:26:43 INFO streaming.PipeMapRed: R/W/S=10/0/0 in:NA [rec/s] out:NA [rec/s]
12/06/01 23:26:43 INFO streaming.PipeMapRed: MRErrorThread done
12/06/01 23:26:43 INFO streaming.PipeMapRed: Records R/W=20/1
12/06/01 23:26:43 INFO streaming.PipeMapRed: mapRedFinished
12/06/01 23:26:43 INFO mapred.LocalJobRunner:
12/06/01 23:26:43 INFO mapred.MapTask: Starting flush of map output
12/06/01 23:26:43 INFO mapred.MapTask: Spilling map output
12/06/01 23:26:43 INFO mapred.MapTask: bufstart = 0; bufend = 560; bufvoid = 104857600
12/06/01 23:26:43 INFO mapred.MapTask: kvstart = 26214396(104857584); kvend = 26214320(104857280); length = 77/6553600
12/06/01 23:26:43 INFO mapred.MapTask: Finished spill 0
12/06/01 23:26:43 INFO mapred.Task: Task:attempt_local_0001_m_000000_0 is done. And is in the process of committing
12/06/01 23:26:43 INFO mapred.LocalJobRunner: Records R/W=20/1
12/06/01 23:26:43 INFO mapred.Task: Task 'attempt_local_0001_m_000000_0' done.
12/06/01 23:26:43 INFO mapred.LocalJobRunner: Finishing task: attempt_local_0001_m_000000_0
12/06/01 23:26:43 INFO mapred.LocalJobRunner: Map task executor complete.
12/06/01 23:26:43 INFO mapred.Task:    Using ResourceCalculatorPlugin : org.apache.hadoop.yarn.util.LinuxResourceCalculatorPlugin@25d71236
12/06/01 23:26:43 INFO mapred.Merger: Merging 1 sorted segments
12/06/01 23:26:43 INFO mapred.Merger: Down to the last merge-pass, with 1 segments left of total size: 574 bytes
12/06/01 23:26:43 INFO mapred.LocalJobRunner:
12/06/01 23:26:43 INFO streaming.PipeMapRed: PipeMapRed exec [/opt/hadoop/mrs/red.py]
12/06/01 23:26:43 WARN conf.Configuration: mapred.job.tracker is deprecated. Instead, use mapreduce.jobtracker.address
12/06/01 23:26:43 WARN conf.Configuration: mapred.map.tasks is deprecated. Instead, use mapreduce.job.maps
12/06/01 23:26:43 INFO streaming.PipeMapRed: R/W/S=1/0/0 in:NA [rec/s] out:NA [rec/s]
12/06/01 23:26:43 INFO streaming.PipeMapRed: R/W/S=10/0/0 in:NA [rec/s] out:NA [rec/s]
12/06/01 23:26:43 INFO streaming.PipeMapRed: Records R/W=20/1
12/06/01 23:26:43 INFO streaming.PipeMapRed: MRErrorThread done
12/06/01 23:26:43 INFO streaming.PipeMapRed: mapRedFinished
12/06/01 23:26:43 INFO mapred.Task: Task:attempt_local_0001_r_000000_0 is done. And is in the process of committing
12/06/01 23:26:43 INFO mapred.LocalJobRunner:
12/06/01 23:26:43 INFO mapred.Task: Task attempt_local_0001_r_000000_0 is allowed to commit now
12/06/01 23:26:43 INFO output.FileOutputCommitter: Saved output of task 'attempt_local_0001_r_000000_0' to hdfs://localhost:9000/asf/_temporary/0/task_local_0001_r_000000
12/06/01 23:26:43 INFO mapred.LocalJobRunner: Records R/W=20/1 > reduce
12/06/01 23:26:43 INFO mapred.Task: Task 'attempt_local_0001_r_000000_0' done.
12/06/01 23:26:43 INFO mapreduce.Job: Job job_local_0001 running in uber mode : false
12/06/01 23:26:43 INFO mapreduce.Job:    map 100% reduce 100%
12/06/01 23:26:43 INFO mapreduce.Job: Job job_local_0001 completed successfully
12/06/01 23:26:43 INFO mapreduce.Job: Counters: 32
                File System Counters
                                FILE: Number of bytes read=205938
                                FILE: Number of bytes written=452840
                                FILE: Number of read operations=0
                                FILE: Number of large read operations=0
                                FILE: Number of write operations=0
                                HDFS: Number of bytes read=252230
                                HDFS: Number of bytes written=59
                                HDFS: Number of read operations=13
                                HDFS: Number of large read operations=0
                                HDFS: Number of write operations=4
                Map-Reduce Framework
                                Map input records=20
                                Map output records=20
                                Map output bytes=560
                                Map output materialized bytes=606
                                Input split bytes=81
                                Combine input records=0
                                Combine output records=0
                                Reduce input groups=2
                                Reduce shuffle bytes=0
                                Reduce input records=20
                                Reduce output records=2
                                Spilled Records=40
                                Shuffled Maps =0
                                Failed Shuffles=0
                                Merged Map outputs=0
                                GC time elapsed (ms)=12
                                CPU time spent (ms)=0
                                Physical memory (bytes) snapshot=0
                                Virtual memory (bytes) snapshot=0
                                Total committed heap usage (bytes)=396361728
                File Input Format Counters
                                Bytes Read=126115
                File Output Format Counters
                                Bytes Written=59
12/06/01 23:26:43 INFO streaming.StreamJob: Output directory: /asf

当然map/reduce V2的功能还不止这些，还需要深入的研究一下。因为2.0虽然是0.23发展过来，但是跟0.23还有些不同，比如0.23中有ApplicationManager，2.0里好像没有在外面露出来了。也许也封装到container里面了。另外，那些xml的配置选项好像跟0.20.x也有很大不同了，具体还没细看。HA功能是支持多个namenode，且多个namenode分管不同的datanode。可以支持手工从某台namenode切换到另外一台namenode。这样做到高可用，据说未来会支持自动检测切换。