二,Hive运行机制与使用
hive介绍
hive是基于Hadoop的一个数据仓库工具,可以将结构化的数据文件映射为一张数据库表,并提供简单的sql查询功能,可以将sql语句转换为MapReduce任务进行运行。其优点是学习成本低,可以通过类SQL语句快速实现简单的MapReduce统计,不必开发专门的MapReduce应用,十分适合数据仓库的统计分析。
hive的运行机制
图示
假设我在hive命令行客户端使用创建了一个数据库(database)myhive,接着又在该数据库中创建了一张表emp。
create database myhive;
use myhive;
create table emp(id int,name string);
那么hive会将元数据存储在数据库中。Hive 中的元数据包括表的名字,表的列和分区及其属性,表的属性(是否为外部表等),表的数据所在目录等。
hive是基于hadoop的,所以数据库和表均表现在hdfs上的目录,数据信息当然也是存储在hdfs上。
对于上面的库和表来说,会在hdfs上创建/user/hive/warehouse/myhive.db这样的目录结构,而表的信息则可以自己上传个文件比如图中的emp.data到/user/hive/warehouse/myhive.db目录下。那么就可以写sql进行查询了(注:写查询语句写的是myhive这张表不删emp.data,如select * from myhive,但是查询到的是emp.data中的信息,两者结合可以理解为传统数据库的某张表),而这些元数据信息都会存储到外部的数据库中(如mysql,当然也可以使用内嵌的derby,不推荐使用derby毕竟是内嵌的不能共享信息)。
然后我再写个查询语句
select id,name from emp where id>2 order by id desc;
那么是怎么执行的呢?查询语句交给hive,hive利用解析器、优化器等(图中表示Compiler),调用mapreduce模板,形成计划,生成的查询计划存储在 HDFS 中,随后由Mapreduce程序调用,提交给job放在Yarn上运行。
hive与mapreduce关系
imagehive的数据存储
1、Hive中所有的数据都存储在 HDFS 中,没有专门的数据存储格式(可支持Text,SequenceFile,ParquetFile,RCFILE等)
2、只需要在创建表的时候告诉 Hive 数据中的列分隔符和行分隔符,Hive 就可以解析数据。
3、Hive 中包含以下数据模型:DB、Table,External Table,Partition,Bucket。
db:在hdfs中表现为${hive.metastore.warehouse.dir}目录下一个文件夹
table:在hdfs中表现所属db目录下一个文件夹
external table:外部表, 与table类似,不过其数据存放位置可以在任意指定路径
普通表: 删除表后, hdfs上的文件都删了
External外部表删除后, hdfs上的文件没有删除, 只是把文件删除了
partition:在hdfs中表现为table目录下的子目录
bucket:桶, 在hdfs中表现为同一个表目录下根据hash散列之后的多个文件, 会根据不同的文件把数据放到不同的文件中
理论总让人头昏,下面介绍hive的初步使用上面的自然就明白了。
hive的使用
虽然可以使用hive与shell交互的方式启动hive
[root@mini1 ~]# cd apps/hive/bin
[root@mini1 bin]# ll
总用量 32
-rwxr-xr-x. 1 root root 1031 4月 30 2015 beeline
drwxr-xr-x. 3 root root 4096 10月 17 12:38 ext
-rwxr-xr-x. 1 root root 7844 5月 8 2015 hive
-rwxr-xr-x. 1 root root 1900 4月 30 2015 hive-config.sh
-rwxr-xr-x. 1 root root 885 4月 30 2015 hiveserver2
-rwxr-xr-x. 1 root root 832 4月 30 2015 metatool
-rwxr-xr-x. 1 root root 884 4月 30 2015 schematool
[root@mini1 bin]# ./hive
hive>
但是界面并不好看,而hive也可以发布为服务(Hive thrift服务),然后可以使用hive自带的beeline去连接。如下
窗口1,开启服务
[root@mini1 ~]# cd apps/hive/bin
[root@mini1 bin]# ll
总用量 32
-rwxr-xr-x. 1 root root 1031 4月 30 2015 beeline
drwxr-xr-x. 3 root root 4096 10月 17 12:38 ext
-rwxr-xr-x. 1 root root 7844 5月 8 2015 hive
-rwxr-xr-x. 1 root root 1900 4月 30 2015 hive-config.sh
-rwxr-xr-x. 1 root root 885 4月 30 2015 hiveserver2
-rwxr-xr-x. 1 root root 832 4月 30 2015 metatool
-rwxr-xr-x. 1 root root 884 4月 30 2015 schematool
[root@mini1 bin]# ./hiveserver2
窗口2,作为客户端连接
[root@mini1 bin]# ./beeline
Beeline version 1.2.1 by Apache Hive
beeline> [root@mini1 bin]#
[root@mini1 bin]# ./beeline
Beeline version 1.2.1 by Apache Hive
beeline> !connect jdbc:hive2://localhost:10000
Connecting to jdbc:hive2://localhost:10000
Enter username for jdbc:hive2://localhost:10000: root
Enter password for jdbc:hive2://localhost:10000: ******
Connected to: Apache Hive (version 1.2.1)
Driver: Hive JDBC (version 1.2.1)
Transaction isolation: TRANSACTION_REPEATABLE_READ
0: jdbc:hive2://localhost:10000>
可能出现错误
Error: Failed to open new session: java.lang.RuntimeException: java.lang.RuntimeException: org.apache.hadoop.security.AccessControlException: Permission denied: user=root, access=EXECUTE, inode="/tmp":hadoop3:supergroup:drwx------
./hadoop dfs -chmod -R 777 /tmp
下面进行简单使用,感觉下使用sql的舒适吧
1、查看数据库
0: jdbc:hive2://localhost:10000> show databases;
+----------------+--+
| database_name |
+----------------+--+
| default |
+----------------+--+
1 row selected (1.456 seconds)
2、创建并使用数据库,查看表
0: jdbc:hive2://localhost:10000> create database myhive;
No rows affected (0.576 seconds)
0: jdbc:hive2://localhost:10000> show databases;
+----------------+--+
| database_name |
+----------------+--+
| default |
| myhive |
+----------------+--+
0: jdbc:hive2://localhost:10000> use myhive;
No rows affected (0.265 seconds)
0: jdbc:hive2://localhost:10000> show tables;
+-----------+--+
| tab_name |
+-----------+--+
+-----------+--+
3、创建表
0: jdbc:hive2://localhost:10000> create table emp(id int,name string);
No rows affected (0.29 seconds)
0: jdbc:hive2://localhost:10000> show tables;
+-----------+--+
| tab_name |
+-----------+--+
| emp |
+-----------+--+
1 row selected (0.261 seconds)
上传数据到该目录下,从页面看的话是个目录,如下
image里面没有文件当然没有数据,那么我们需要上传个文件到该目录下。
[root@mini1 ~]# cat sz.data
1,zhangsan
2,lisi
3,wangwu
4,furong
5,fengjie
[root@mini1 ~]# hadoop fs -put sz.data /user/hive/warehouse/myhive.db/emp
再查看
image4、查看表信息
0: jdbc:hive2://localhost:10000> select * from emp;
+---------+-----------+--+
| emp.id | emp.name |
+---------+-----------+--+
| NULL | NULL |
| NULL | NULL |
| NULL | NULL |
| NULL | NULL |
| NULL | NULL |
+---------+-----------+--+
结果肯定都是null,因为创建表的时候根本没指定根据”,”来切分,而文件中的字段分隔用了逗号。那么删除该表,重新上传文件,重新建表语句如下
0: jdbc:hive2://localhost:10000> drop table emp;
No rows affected (1.122 seconds)
0: jdbc:hive2://localhost:10000> show tables;
+-----------+--+
| tab_name |
+-----------+--+
+-----------+--+
0: jdbc:hive2://localhost:10000> create table emp(id int,name string)
0: jdbc:hive2://localhost:10000> row format delimited
0: jdbc:hive2://localhost:10000> fields terminated by ',';
No rows affected (0.265 seconds)
0: jdbc:hive2://localhost:10000>
[root@mini1 ~]# hadoop fs -put sz.data /user/hive/warehouse/myhive.db/emp
0: jdbc:hive2://localhost:10000> select * from emp;
+---------+-----------+--+
| emp.id | emp.name |
+---------+-----------+--+
| 1 | zhangsan |
| 2 | lisi |
| 3 | wangwu |
| 4 | furong |
| 5 | fengjie |
+---------+-----------+--+
6、条件查询
0: jdbc:hive2://localhost:10000> select id,name from emp where id>2 order by id desc;
INFO : Number of reduce tasks determined at compile time: 1
INFO : In order to change the average load for a reducer (in bytes):
INFO : set hive.exec.reducers.bytes.per.reducer=<number>
INFO : In order to limit the maximum number of reducers:
INFO : set hive.exec.reducers.max=<number>
INFO : In order to set a constant number of reducers:
INFO : set mapreduce.job.reduces=<number>
INFO : number of splits:1
INFO : Submitting tokens for job: job_1508216103995_0004
INFO : The url to track the job: http://mini1:8088/proxy/application_1508216103995_0004/
INFO : Starting Job = job_1508216103995_0004, Tracking URL = http://mini1:8088/proxy/application_1508216103995_0004/
INFO : Kill Command = /root/apps/hadoop-2.6.4/bin/hadoop job -kill job_1508216103995_0004
INFO : Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 1
INFO : 2017-10-18 00:35:39,865 Stage-1 map = 0%, reduce = 0%
INFO : 2017-10-18 00:35:46,275 Stage-1 map = 100%, reduce = 0%, Cumulative CPU 1.33 sec
INFO : 2017-10-18 00:35:51,487 Stage-1 map = 100%, reduce = 100%, Cumulative CPU 2.34 sec
INFO : MapReduce Total cumulative CPU time: 2 seconds 340 msec
INFO : Ended Job = job_1508216103995_0004
+-----+----------+--+
| id | name |
+-----+----------+--+
| 5 | fengjie |
| 4 | furong |
| 3 | wangwu |
+-----+----------+--+
3 rows selected (18.96 seconds)
看到这就能明白了,写的sql最后是被解析为了mapreduce程序放到yarn上来跑的,hive其实是提供了众多的mapreduce模板。
7、创建外部表
0: jdbc:hive2://localhost:10000> create external table emp2(id int,name string)
0: jdbc:hive2://localhost:10000> row format delimited fields terminated by ','//指定逗号分割
0: jdbc:hive2://localhost:10000> stored as textfile//文本存储方式
0: jdbc:hive2://localhost:10000> location '/company';
No rows affected (0.101 seconds)//存储在/company目录下
0: jdbc:hive2://localhost:10000> dfs -ls /;
+----------------------------------------------------------------------------------------+--+
| DFS Output |
+----------------------------------------------------------------------------------------+--+
| Found 16 items |
| -rw-r--r-- 2 angelababy mygirls 7 2017-10-01 20:22 /canglaoshi_wuma.avi |
| -rw-r--r-- 2 root supergroup 22 2017-10-03 21:12 /cangmumayi.avi |
| drwxr-xr-x - root supergroup 0 2017-10-18 00:55 /company |
| drwxr-xr-x - root supergroup 0 2017-10-13 04:44 /flowcount |
| drwxr-xr-x - root supergroup 0 2017-10-17 03:44 /friends |
| drwxr-xr-x - root supergroup 0 2017-10-17 06:19 /gc |
| drwxr-xr-x - root supergroup 0 2017-10-07 07:28 /liushishi.log |
| -rw-r--r-- 3 12706 supergroup 60 2017-10-04 21:58 /liushishi.love |
| drwxr-xr-x - root supergroup 0 2017-10-17 07:32 /logenhance |
| -rw-r--r-- 2 root supergroup 26 2017-10-16 20:49 /mapjoin |
| drwxr-xr-x - root supergroup 0 2017-10-16 21:16 /mapjoincache |
| drwxr-xr-x - root supergroup 0 2017-10-13 13:15 /mrjoin |
| drwxr-xr-x - root supergroup 0 2017-10-16 23:35 /reverse |
| drwx------ - root supergroup 0 2017-10-17 13:10 /tmp |
| drwxr-xr-x - root supergroup 0 2017-10-17 13:13 /user |
| drwxr-xr-x - root supergroup 0 2017-10-14 01:33 /wordcount |
+----------------------------------------------------------------------------------------+--+
0: jdbc:hive2://localhost:10000> create external table t_sz_ext(id int,name string)
0: jdbc:hive2://localhost:10000> row format delimited fields terminated by '\t'
0: jdbc:hive2://localhost:10000> stored as textfile
0: jdbc:hive2://localhost:10000> location '/company';
No rows affected (0.135 seconds)
0: jdbc:hive2://localhost:10000> show tables;
+-----------+--+
| tab_name |
+-----------+--+
| emp |
| emp2 |
| t_sz_ext |
+-----------+--+
能发现多了目录/company和两张表,不过这个时候/company下是没任何东西的。
8、加载文件信息到表中
前面使用了hadoop命令将文件上传到了表对应的目录下,但是也可以在命令行下直接导入文件信息
0: jdbc:hive2://localhost:10000> load data local inpath '/root/sz.data' into table emp2;(也可以用hadoo直接上传)
INFO : Loading data to table myhive.emp2 from file:/root/sz.data
INFO : Table myhive.emp2 stats: [numFiles=0, totalSize=0]
No rows affected (0.414 seconds)
0: jdbc:hive2://localhost:10000> select * from emp2;
+----------+------------+--+
| emp2.id | emp2.name |
+----------+------------+--+
| 1 | zhangsan |
| 2 | lisi |
| 3 | wangwu |
| 4 | furong |
| 5 | fengjie |
+----------+------------+--+
9、表分区,分区字段为school,导入数据到2个不同的分区中
0: jdbc:hive2://localhost:10000> create table stu(id int,name string)
0: jdbc:hive2://localhost:10000> partitioned by(school string)
0: jdbc:hive2://localhost:10000> row format delimited fields terminated by ',';
No rows affected (0.319 seconds)
0: jdbc:hive2://localhost:10000> show tables;
+-----------+--+
| tab_name |
+-----------+--+
| emp |
| emp2 |
| stu |
| t_sz_ext |
+-----------+--+
0: jdbc:hive2://localhost:10000> load data local inpath '/root/sz.data' into table stu partition(school='scu');
INFO : Loading data to table myhive.stu partition (school=scu) from file:/root/sz.data
INFO : Partition myhive.stu{school=scu} stats: [numFiles=1, numRows=0, totalSize=46, rawDataSize=0]
No rows affected (0.607 seconds)
0: jdbc:hive2://localhost:10000> select * from stu;
+---------+-----------+-------------+--+
| stu.id | stu.name | stu.school |
+---------+-----------+-------------+--+
| 1 | zhangsan | scu |
| 2 | lisi | scu |
| 3 | wangwu | scu |
| 4 | furong | scu |
| 5 | fengjie | scu |
+---------+-----------+-------------+--+
5 rows selected (0.286 seconds)
0: jdbc:hive2://localhost:10000> load data local inpath '/root/sz2.data' into table stu partition(school='hfut');
INFO : Loading data to table myhive.stu partition (school=hfut) from file:/root/sz2.data
INFO : Partition myhive.stu{school=hfut} stats: [numFiles=1, numRows=0, totalSize=46, rawDataSize=0]
No rows affected (0.671 seconds)
0: jdbc:hive2://localhost:10000> select * from stu;
+---------+-----------+-------------+--+
| stu.id | stu.name | stu.school |
+---------+-----------+-------------+--+
| 1 | Tom | hfut |
| 2 | Jack | hfut |
| 3 | Lucy | hfut |
| 4 | Kitty | hfut |
| 5 | Lucene | hfut |
| 6 | Sakura | hfut |
| 1 | zhangsan | scu |
| 2 | lisi | scu |
| 3 | wangwu | scu |
| 4 | furong | scu |
| 5 | fengjie | scu |
+---------+-----------+-------------+--+
注:hive是不遵循三范式的,别去考虑主键了。
10、添加分区
0: jdbc:hive2://localhost:10000> alter table stu add partition (school='Tokyo');
为了更直观,去页面查看
image image imagehive元数据表说明
https://blog.csdn.net/haozhugogo/article/details/73274832
谢谢