?

1安裝Hive

1.1下載解壓

wget http://mirrors.cnnic.cn/apache/hive/stable/hive-0.10.0.tar.gz

tar xzvfhive-0.10.0.tar.gz

1.2配置環(huán)境變量

exportHIVE_HOME=/usr/local/src/hive-0.10.0

export PATH=$HIVE_HOME/bin:$PATH

1.3建立Hive倉庫目錄

hadoop fs -mkdir/tmp

hadoop fs -mkdir/user/hive/warehouse

hadoop fs -chmodg+w /tmp

hadoop fs -chmodg+w /user/hive/warehouse

1.4啟動命令行

通過hive命令進入命令行，操作與MySQL的命令行類似：

2安裝Sqoop

2.1下載解壓

下載適合Hadoop 0.20版本的Sqoop：

wget http://mirrors.cnnic.cn/apache/sqoop/1.4.3/sqoop-1.4.3.bin__hadoop-0.20.tar.gz

tar -xvf sqoop-1.4.3.bin__hadoop-0.20.tar.gz

2.2配置環(huán)境變量

export SQOOP_HOME=/usr/local/src/sqoop-1.4.3.bin__hadoop-0.20

export PATH=$SQOOP_HOME/bin:$PATH

export HADOOP_COMMON_HOME=/home/admin/hadoop-0.20.2

export HADOOP_MAPRED_HOME=/home/admin/hadoop-0.20.2

3用Sqoop導(dǎo)入數(shù)據(jù)到HIVE

3.1導(dǎo)入HDFS

我們從MySQL數(shù)據(jù)庫中導(dǎo)入一張表的數(shù)據(jù)來測試一下Sqoop是否配置成功。首先上傳mysql-connector-java-5.1.23.jar到sqoop的lib文件夾下，然后在sqoop/bin下執(zhí)行下列命令：

sqoop import--connect jdbc:mysql://ip/database --table tb1 --username user -P

===============================================================================

Warning: /usr/lib/hbase does not exist!HBase imports will fail.

Please set $HBASE_HOME to the root of yourHBase installation.

Enter password:

13/06/07 16:51:46 INFOmanager.MySQLManager: Preparing to use a MySQL streaming resultset.

13/06/07 16:51:46 INFO tool.CodeGenTool: Beginning codegeneration

13/06/07 16:51:48 INFO manager.SqlManager:Executing SQL statement: SELECT t.* FROM `tb1` AS t LIMIT 1

13/06/07 16:51:48 INFO manager.SqlManager:Executing SQL statement: SELECT t.* FROM `tb1` AS t LIMIT 1

13/06/07 16:51:48 INFOorm.CompilationManager: HADOOP_MAPRED_HOME is /home/admin/hadoop-0.20.2

13/06/07 16:51:48 INFOorm.CompilationManager: Found hadoop core jar at:/home/admin/hadoop-0.20.2/hadoop-0.20.2-core.jar

Note:/tmp/sqoop-root/compile/44c4b6c5ac57de04b487eb90633ac33e/tb1.java uses oroverrides a deprecated API.

Note: Recompile with -Xlint:deprecation fordetails.

13/06/07 16:51:54 INFO orm.CompilationManager:Writing jar file:/tmp/sqoop-root/compile/44c4b6c5ac57de04b487eb90633ac33e/tb1.jar

13/06/07 16:51:54 WARNmanager.MySQLManager: It looks like you are importing from mysql.

13/06/07 16:51:54 WARNmanager.MySQLManager: This transfer can be faster! Use the --direct

13/06/07 16:51:54 WARNmanager.MySQLManager: option to exercise a MySQL-specific fast path.

13/06/07 16:51:54 INFOmanager.MySQLManager: Setting zero DATETIME behavior to convertToNull (mysql)

13/06/07 16:51:54 INFO mapreduce.ImportJobBase:Beginning import of tb1

13/06/07 16:51:57 INFOdb.DataDrivenDBInputFormat: BoundingValsQuery: SELECT MIN(`id`), MAX(`id`) FROM`tb1`

13/06/07 16:51:59 INFO mapred.JobClient:Running job: job_201306071651_0001

13/06/07 16:52:00 INFOmapred.JobClient:? map 0% reduce 0%

13/06/07 16:52:38 INFOmapred.JobClient:? map 50% reduce 0%

13/06/07 16:52:44 INFOmapred.JobClient:? map 100% reduce 0%

13/06/07 16:52:46 INFO mapred.JobClient:Job complete: job_201306071651_0001

13/06/07 16:52:46 INFO mapred.JobClient:Counters: 5

13/06/07 16:52:46 INFOmapred.JobClient:?? Job Counters

13/06/07 16:52:46 INFOmapred.JobClient:???? Launched map tasks=2

13/06/07 16:52:46 INFOmapred.JobClient:?? FileSystemCounters

13/06/07 16:52:46 INFO mapred.JobClient:???? HDFS_BYTES_WRITTEN=212

13/06/07 16:52:46 INFOmapred.JobClient:?? Map-Reduce Framework

13/06/07 16:52:46 INFOmapred.JobClient:???? Map input records=2

13/06/07 16:52:46 INFOmapred.JobClient:???? Spilled Records=0

13/06/07 16:52:46 INFO mapred.JobClient:???? Map output records=2

13/06/07 16:52:46 INFOmapreduce.ImportJobBase: Transferred 212 bytes in 51.383 seconds (4.1259bytes/sec)

13/06/07 16:52:46 INFOmapreduce.ImportJobBase: Retrieved 2 records.

===============================================================================

?

數(shù)據(jù)文件默認(rèn)被導(dǎo)入到當(dāng)前用戶文件夾下表名對應(yīng)的文件夾了：

?

Sqoop默認(rèn)會同時啟動四個Map任務(wù)來加速數(shù)據(jù)導(dǎo)入，可以通過-m 1命令來強制只啟動一個map任務(wù)，這樣就只會在HDFS中生成一個數(shù)據(jù)文件了。因為tb1表目前就兩條數(shù)據(jù)，所以一共產(chǎn)生兩個文件，查看下生成的文件內(nèi)容：

3.2創(chuàng)建Hive表

首先在hive命令行中創(chuàng)建tb1表。注意hive支持的數(shù)據(jù)類型有限，并且 一定要設(shè)置表的分隔符為逗號 ，否則Hive默認(rèn)分隔符為Ctrl+A。

CREATE TABLE tb1(

? id int,

?......

) row format delimited fields terminated by ‘,’ ;

?

也可以通過下面的命令讓Sqoop根據(jù)MySQL表結(jié)構(gòu)自動創(chuàng)建出Hive表：

sqoop create-hive-table --connect jdbc:mysql://ip/database --table tb1 --hive-table tb1 --username user -P

3.3導(dǎo)入Hive

現(xiàn)在導(dǎo)入HDFS中的文件到Hive，注意Hive從HDFS導(dǎo)入數(shù)據(jù)后，會將HDFS中的文件/user/root/tb1移動到/user/hive/tb1：

???????? LOADDATA INPATH '/user/root/tb1/part-m-*' OVERWRITE INTO TABLE tb1

3.4一條強大的命令

上面的從MySQL導(dǎo)出數(shù)據(jù)到HDFS、創(chuàng)建Hive表格、導(dǎo)入數(shù)據(jù)到Hive三步，可以直接用一條Sqoop命令完成：

sqoop import--connect jdbc:mysql://ip/database --table tb1 --username user -P ?--hive-import

?

4用HiveQL做分析

待續(xù)......?

參考資料

Hive安裝

https://cwiki.apache.org/confluence/display/Hive/GettingStarted

?

http://sqoop.apache.org/docs/1.99.1/Installation.html

?

?

Hadoop入門（三）：Sqoop和Hive的使用