用sqoop实现mysql和hive数据互导

本文主要是介绍用sqoop实现mysql和hive数据互导，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

在此测试的是伪分布式

hadoop1.2.1

sqoop-1.4.4.bin__hadoop-1.0.0

hive-0.12.0

1、数据准备

1.1建立测试用户sqoop
grant all privileges on *.* to 'sqoop'@'%' identified by 'sqoop' with grant option;
注意：这一步不是必须，也可以在mysql的root用户下进行

1.2在sqoop用户下，创建sqoop数据库，在sqoop库中建表employee
create database sqoop;use sqoop;create table employee(employee_id int not null primary key,
employee_name varchar(30));
insert into employee values(101,'zhangsan');
insert into employee values(102,'lisi');
insert into employee values(103,'wangwu');
此时，在sqoop库中的employee表中已经存在三条记录，可以select查询验证。

2、将Mysql中的sqoop用户下的sqoop库中的employee表中的数据导入hive中

2.1mysql中的employee表导入hive

sqoop import --connect jdbc:mysql://localhost:3306/sqoop --username sqoop --password sqoop --table employee --hive-import -m 1

在HDFS的/user/hive/warehouse目录下查看导入的结果

cwjy1202@cwjy1202-Lenovo-IdeaPad-Y471A:~$ hadoop dfs -ls /user/hive/warehouse/employee/
Warning: $HADOOP_HOME is deprecated.Found 2 items
-rw-r--r--   1 cwjy1202 supergroup          0 2014-06-04 14:38 /user/hive/warehouse/employee/_SUCCESS
-rw-r--r--   1 cwjy1202 supergroup         33 2014-06-04 14:38 /user/hive/warehouse/employee/part-m-00000

cwjy1202@cwjy1202-Lenovo-IdeaPad-Y471A:~$ hadoop dfs -cat /user/hive/warehouse/employee/part-m-00000
Warning: $HADOOP_HOME is deprecated.101zhangsan
102lisi
103wangwu

2.2去掉导入语句的参数 -m 1

sqoop import --connect jdbc:mysql://localhost:3306/sqoop --username sqoop --password sqoop --table employee --hive-import

显示的结果如下：

Found 4 items
-rw-r--r--   1 cwjy1202 supergroup          0 2014-06-04 14:10 /user/hive/warehouse/employee/_SUCCESS
-rw-r--r--   1 cwjy1202 supergroup         13 2014-06-04 14:10 /user/hive/warehouse/employee/part-m-00000
-rw-r--r--   1 cwjy1202 supergroup          9 2014-06-04 14:10 /user/hive/warehouse/employee/part-m-00001
-rw-r--r--   1 cwjy1202 supergroup         11 2014-06-04 14:10 /user/hive/warehouse/employee/part-m-00002

三条记录分别存在part-m-00000、part-m-00001和part-m-00002中

3、hive的数据导入mysql

3.1把mysql中的employee表清空

在linux终端输入：
mysql -usqoop -p回车
密码truncate table sqoop.employee;
mysql> select * from sqoop.employee;
Empty set (0.00 sec)
删除成功！！！

3.2从hive到mysql

在linux终端输入：
sqoop export --connect jdbc:mysql://localhost:3306/sqoop -username sqoop -password sqoop --table employee 
--export-dir /user/hive/warehouse/employee --input-fields-terminated-by '\001'
说明：jdbc:mysql://localhost:3306/sqoop 3306是mysql的端口；localhost是mysql的安装ip地址，运行时根据自己的情况而定；sqoop是mysql中的数据库。

hive中字段的默认分割符为'\001'

运行过程的最后显示Export 3 records如下：
14/06/04 14:49:45 INFO mapreduce.ExportJobBase: Transferred 760 bytes in 11.8562 seconds (64.1017 bytes/sec)
14/06/04 14:49:45 INFO mapreduce.ExportJobBase: Exported 3 records.
查询mysql中的employee验证数据是否存在：
mysql> select * from sqoop.employee;
+-------------+---------------+
| employee_id | employee_name |
+-------------+---------------+
| 101 | zhangsan |
| 102 | lisi |
| 103 | wangwu |
+-------------+---------------+
3 rows in set (0.00 sec)
导入mysql成功！！！

额外说明：

因为mysql中的employee表的结构已经确定，employee_id为primary key，所以多次从hive导入相同的数据，只会第一成功。

操作如下，再一次在linux终端输入：
sqoop export --connect jdbc:mysql://localhost:3306/sqoop -username sqoop -password sqoop --table employee 
--export-dir /user/hive/warehouse/employee --input-fields-terminated-by '\001'
运行过程显示：
14/06/04 15:14:34 INFO mapred.JobClient: Task Id : attempt_201406041140_0012_m_000001_1, Status : FAILED
Task attempt_201406041140_0012_m_000001_1 failed to report status for 600 seconds. Killing!
14/06/04 15:14:36 INFO mapred.JobClient: map 25% reduce 0%
14/06/04 15:14:37 INFO mapred.JobClient: Task Id : attempt_201406041140_0012_m_000002_1, Status : FAILED
Task attempt_201406041140_0012_m_000002_1 failed to report status for 600 seconds. Killing!