完成sqoop的安装后,可以这样测试是否可以连接到mysql(注意:mysql的jar包要放到 SQOOP_HOME/lib 下):
sqoop list-databases --connect jdbc:mysql://192.168.1.109:3306/ --username root --password 19891231
结果如下
即说明sqoop已经可以正常使用了。
下面,要将mysql中的数据导入到hadoop中。
我准备的是一个300万条数据的身份证数据表:
先启动hive(使用命令行:hive 即可启动)
然后使用sqoop导入数据到hive:
sqoop import --connect jdbc:mysql://192.168.1.109:3306/hadoop --username root --password 19891231 --table test_sfz --hive-import
sqoop 会启动job来完成导入工作。
完成导入用了2分20秒,还是不错的。
在hive中可以看到刚刚导入的数据表:
我们来一句sql测试一下数据:
select * from test_sfz where id < 10;
可以看到,hive完成这个任务用了将近25秒,确实是挺慢的(在mysql中几乎是不费时间),但是要考虑到hive是创建了job在hadoop中跑,时间当然多。
接下来,我们会对这些数据进行复杂查询的测试:
我机子的配置如下:
hadoop 是运行在虚拟机上的伪分布式,虚拟机OS是ubuntu12.04 64位,配置如下:
TEST 1 计算平均年龄
测试数据:300.8 W
1. 计算广东的平均年龄
mysql:select (sum(year(NOW()) - SUBSTRING(borth,1,4))/count(*)) as ageAvge from test_sfz where address like '广东%';
用时: 0.877s
hive:select (sum(year('2014-10-01') - SUBSTRING(borth,1,4))/count(*)) as ageAvge from test_sfz where address like '广东%';
用时:25.012s
2. 对每个城市的的平均年龄进行从高到低的排序
mysql:select
address, (sum(year(NOW()) - SUBSTRING(borth,1,4))/count(*)) as ageAvge
from test_sfz GROUP BY address order by ageAvge desc;
用时:2.949s
hive:select
address, (sum(year('2014-10-01') - SUBSTRING(borth,1,4))/count(*)) as
ageAvge from test_sfz GROUP BY address order by ageAvge desc;
用时:51.29s
可以看到,在耗时上面,hive的增长速度较mysql慢。
TEST 2
测试数据:1200W
mysql 引擎: MyISAM(为了加快查询速度)
导入到hive:
1. 计算广东的平均年龄
mysql:select (sum(year(NOW()) - SUBSTRING(borth,1,4))/count(*)) as ageAvge from test_sfz2 where address like '广东%';
用时: 5.642s
hive:select (sum(year('2014-10-01') - SUBSTRING(borth,1,4))/count(*)) as ageAvge from test_sfz2 where address like '广东%';
用时:168.259s
2. 对每个城市的的平均年龄进行从高到低的排序
mysql:select
address, (sum(year(NOW()) - SUBSTRING(borth,1,4))/count(*)) as ageAvge
from test_sfz2 GROUP BY address order by ageAvge desc;
用时:11.964s
hive:select
address, (sum(year('2014-10-01') - SUBSTRING(borth,1,4))/count(*)) as
ageAvge from test_sfz2 GROUP BY address order by ageAvge desc;
用时:311.714s
用sqoop
sqoop job --create myjob -- import --connect jdbc:mysql://192.168.80.1:3306/hive --username root --password root --table tbls --fields-terminated-by '\t' --null-string '' --m 1 --check-column 'TBL_ID' --incremental append
可以导入,太大,就不要用记事本或者其它编辑器打开了,
查询分析器--连接数据库--文件--打开--.sql文件--F5执行
不过你把数据导出为SQL文件,而且200多M,真的是。。。。
为何不压缩后,直接拷库文件,或者直接备份导出呢。