CDH5: 使用parcels配置lzo

问题导读

1.Parcel 部署有哪些步骤？
2.如何使用parcels配置lzo，需要修改哪些配置？







一、Parcel 部署步骤   

1 下载: 首先需要下载 Parcel。下载完成后，Parcel 将驻留在 Cloudera Manager 主机的本地目录中。
    2 分配: Parcel 下载后，将分配到群集中的所有主机上并解压缩。
    3 激活: 分配后，激活 Parcel 为群集重启后使用做准备。激活前可能还需要升级。


二、lzo parcels本地化

    1、到http://archive-primary.cloudera.com/gplextras/parcels/latest/下载最新lzo parcels包，根据安装hadoop集群的服务器操作系统版本下载，我使用的是rhel6.2, 所以下载的是HADOOP_LZO-0.4.15-1.gplextras.p0.64-el6.parcel

    2、同时下载manifest.json，并根据manifest.json文件中的hash值创建sha文件（注意：sha文件的名称与parcels包名一样）

    3、命令行进入Apache（如果没有安装，则需要安装）的网站根目录下，默认是/var/www/html,在此目录下创建lzo，并将这三个文件放在lzo目录中

    4、启动httpd服务，在浏览器查看，如http://ip/lzo,则结果如下：

        

    5、将发布的local parcels发布地址配置到远程 Parcel 存储库 URL地址中，见下图

         

6、在cloud manager的parcel页面的可下载parcel中，就可以看到lzo parcels, 点击并进行下载

7、根据parcels的部署步骤，进行分配、激活。结果如下图

三、修改配置

    修改hdfs的配置

    将io.compression.codecs属性值中追加,org.apache.hadoop.io.compress.Lz4Codec,com.hadoop.compression.lzo.LzopCodec

    修改yarn配置

    将mapreduce.application.classpath的属性值修改为：

$HADOOP_MAPRED_HOME/*,$HADOOP_MAPRED_HOME/lib/*,$MR2_CLASSPATH,/opt/cloudera/parcels/HADOOP_LZO/lib/hadoop/lib/*
复制代码

    将mapreduce.admin.user.env的属性值修改为：

LD_LIBRARY_PATH=$HADOOP_COMMON_HOME/lib/native:$JAVA_LIBRARY_PATH:/opt/cloudera/parcels/HADOOP_LZO/lib/hadoop/lib/native
复制代码

四、验证

    create external table lzo(id int,name string)  row format delimited fields terminated by '#' STORED AS INPUTFORMAT 'com.hadoop.mapred.DeprecatedLzoTextInputFormat' OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat' location '/test';

    创建一个data.txt,内容如下：  

1#tianhe
2#gz
3#sz
4#sz
5#bx
复制代码

?


    然后使用lzop命令对此文件压缩，然后上传到hdfs的/test目录下

    启动hive,建表并进行数据查询，结果如下：

hive> create external table lzo(id int,name string)  row format delimited fields terminated by '#' STORED AS INPUTFORMAT 'com.hadoop.mapred.DeprecatedLzoTextInputFormat' OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat' location '/test';
OK
Time taken: 0.108 seconds
hive> select * from lzo where id>2;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_1404206497656_0002, Tracking URL = http://hadoop01.kt:8088/proxy/application_1404206497656_0002/
Kill Command = /opt/cloudera/parcels/CDH-5.0.1-1.cdh5.0.1.p0.47/lib/hadoop/bin/hadoop job  -kill job_1404206497656_0002
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 0
2014-07-01 17:30:27,547 Stage-1 map = 0%,  reduce = 0%
2014-07-01 17:30:37,403 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 2.84 sec
2014-07-01 17:30:38,469 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 2.84 sec
2014-07-01 17:30:39,527 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 2.84 sec
MapReduce Total cumulative CPU time: 2 seconds 840 msec
Ended Job = job_1404206497656_0002
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 2.84 sec   HDFS Read: 295 HDFS Write: 15 SUCCESS
Total MapReduce CPU Time Spent: 2 seconds 840 msec
OK
3       sz
4       sz
5       bx
Time taken: 32.803 seconds, Fetched: 3 row(s)

hive> SET hive.exec.compress.output=true;
hive> SET mapred.output.compression.codec=com.hadoop.compression.lzo.LzopCodec;
hive> create external table lzo2(id int,name string)  row format delimited fields terminated by '#' STORED AS INPUTFORMAT 'com.hadoop.mapred.DeprecatedLzoTextInputFormat' OUTPUTFORMAT 'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat' location '/test';
OK
Time taken: 0.092 seconds
hive> insert into table lzo2 select * from lzo;
Total MapReduce jobs = 3
Launching Job 1 out of 3
Number of reduce tasks is set to 0 since there's no reduce operator
Starting Job = job_1404206497656_0003, Tracking URL = http://hadoop01.kt:8088/proxy/application_1404206497656_0003/
Kill Command = /opt/cloudera/parcels/CDH-5.0.1-1.cdh5.0.1.p0.47/lib/hadoop/bin/hadoop job  -kill job_1404206497656_0003
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 0
2014-07-01 17:33:47,351 Stage-1 map = 0%,  reduce = 0%
2014-07-01 17:33:57,114 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 1.96 sec
2014-07-01 17:33:58,170 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 1.96 sec
MapReduce Total cumulative CPU time: 1 seconds 960 msec
Ended Job = job_1404206497656_0003
Stage-4 is selected by condition resolver.
Stage-3 is filtered out by condition resolver.
Stage-5 is filtered out by condition resolver.
Moving data to: hdfs://hadoop01.kt:8020/tmp/hive-hdfs/hive_2014-07-01_17-33-22_504_966970548620625440-1/-ext-10000
Loading data to table default.lzo2
Table default.lzo2 stats: [num_partitions: 0, num_files: 2, num_rows: 0, total_size: 171, raw_data_size: 0]
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 1.96 sec   HDFS Read: 295 HDFS Write: 79 SUCCESS
Total MapReduce CPU Time Spent: 1 seconds 960 msec
OK
Time taken: 36.625 seconds
复制代码

请问版主有自己做过吗？
我的CDH版本是CDH5.3.5，按照上述的方法集成出现了点问题，请有时间帮忙看一下吧。

15/10/22 15:17:38 INFO lzo.GPLNativeCodeLoader: Loaded native gpl library
15/10/22 15:17:38 WARN lzo.LzoCompressor: java.lang.UnsatisfiedLinkError: Cannot load liblzo2.so.2 (liblzo2.so.2: cannot open shared object file: No such file or directory)!
15/10/22 15:17:38 ERROR lzo.LzoCodec: Failed to load/initialize native-lzo library

谢谢了！

我用的是5.11，cloudera的那个远程的parcel源要求是https连接吗？我不管是http还是https都找不到，麻烦问下知道为什么吗？总是显示错误：备注: 未在已配置的存储库中找到任何 parcel。尝试在更多选项下添加一个自定义存储库。否则，您可能只能继续使用包。