hive与hbase整合过程

实现目标apache

  1. Hive能够实时查询Hbase中的数据.
  2. hive中的表插入数据会同步更新到hbase对应的表中.
  3. 能够将hbase中不一样的表中的列经过 left 或 inner join 方式映射到hive 中的一个视图中.

Hive map hbaseapp

1,启动hive hbaseoop

在hive hbase服务启动的状况下, $HIVE_HOME/bin/hive --auxpath $HIVE_HOME/lib/hive-hbase-handler-1.1.0-cdh5.7.1.jar,$HIVE_HOME/lib/hbase-common-1.2.0-cdh5.7.1.jar,$HIVE_HOME/lib/zookeeper-3.4.5-cdh5.7.1.jar,$HIVE_HOME/lib/guava-14.0.1.jar --hiveconfhbase.master=dwrj5123:60000      (可能不须要这个过程)  .spa

2, 查询hbase中表的结构hadoop

(1)查询jinan:SI3U_AC06_TEMPci

describe 'jinan:SI3U_AC06_TEMP'同步

Table jinan:SI3U_AC06_TEMP is ENABLED                                                                                                                   string

jinan:SI3U_AC06_TEMP                                                                                                                                     io

COLUMN FAMILIES DESCRIPTION                                                                                                                              table

{NAME => 'AC06_TEMP', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE', DATA_BLOCK_ENCODING => 'NONE', TTL => '

FOREVER', COMPRESSION => 'NONE', MIN_VERSIONS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE => '0'}                             

1 row(s) in 0.1960 seconds

其中jinan:SI3U_AC06_TEMP为表名, 列族为AC06_TEMP,经过查询表中的数据得知有不限于如下列:

AC06_TEMP:AAC001,

AC06_TEMP:AAE140,

AC06_TEMP:AAE149,

AC06_TEMP:BAA044,

AC06_TEMP:BAA035,

AC06_TEMP:BAA036,

AC06_TEMP:AAE034

(2) 查询jinan:SI3U_AC01

hbase(main):003:0> describe 'jinan:SI3U_AC01'

Table jinan:SI3U_AC01 is ENABLED                                                                          

jinan:SI3U_AC01                                                                                            

COLUMN FAMILIES DESCRIPTION                                                                               

{NAME => 'AC01', BLOOMFILTER => 'ROW', VERSIONS => '1', IN_MEMORY => 'false', KEEP_DELETED_CELLS => 'FALSE'

MPRESSION => 'NONE', MIN_VERSIONS => '0', BLOCKCACHE => 'true', BLOCKSIZE => '65536', REPLICATION_SCOPE =>

1 row(s) in 0.1850 seconds

jinan:SI3U_AC01 经过查询表中的数据获得不限于如下列:

AC01:AAC001,

AC01:AAC003,

AC01:AAA109

3, 建立hive 表到hbase映射

(1) 建立hive 表jinan_SI3U_AC01 到hbase 表"jinan:SI3U_AC01的映射:

解释: jinan_SI3U_AC01为hive 中表名, jinan:SI3U_AC01为须要映射的hbase表名.

":key,AC01:AAC001,AC01:AAC003,AC01:AAA109": 为须要映射的列, AC01为列族,多列以逗号隔开.

 

 

(2) 建立hive 表jinan_SI3U_AC06_TEMP 到hbase表jinan:SI3U_AC06_TEMP的映射:

CREATE EXTERNAL  TABLE jinan_SI3U_AC06_TEMP(key string,AAC001 string,AAE140 string,AAE149 string,BAA044 string,BAA035 decimal(19,4),BAA036 decimal(19,4),AAE034 TIMESTAMP ) 

STORED BY 'org.apache.hadoop.hive.hbase.HBaseStorageHandler' 

WITH SERDEPROPERTIES ("hbase.columns.mapping" = ":key,AC06_TEMP:AAC001,AC06_TEMP:AAE140,AC06_TEMP:AAE149,AC06_TEMP:BAA044,AC06_TEMP:BAA035,AC06_TEMP:BAA036,AC06_TEMP:AAE034")  TBLPROPERTIES ("hbase.table.name" = "jinan:SI3U_AC06_TEMP");

 

Hive建立视图

hive 中能够经过建立视图的方式将多个表的数据整合成一个视图,方便查询和使用. 这里以上面映射的jinan_SI3U_AC01jinan_SI3U_AC06_TEMP两个表为例.

create view fact_view (AAC001,AAC003,AAA109,AAE140,AAE149,BAA044,BAA035,BAA036, AAE034 ) as  SELECT a.AAC001, a.AAC003,a.AAA109, b.AAE140,b.AAE149, b.BAA044,b.BAA035, b.BAA036,b.AAE034 FROM  jinan_SI3U_AC01 a RIGHT JOIN jinan_SI3U_AC06_TEMP b ON a.aac001 =b.aac001;

表名

列名

 

视图名

jinan_SI3U_AC01

AAC001

AAC003

AAA109

 

fact_view

jinan_SI3U_AC06_TEMP

AAE140

AAE149

BAA044

BAA035

BAA036

AAE034

 

经过SELECT * FROM fact_view; 能够查询到有效数据.

 

KYLIN hive 视图的使用

kylin 支持hive 视图构创建方体, 过程与使用hive 表相同. 构创建方体完成以后,执行查询,

SELECT SUM(BAA035) FROM FACT_VIEW  left inner join DATE_VIEW ON FACT_VIEW.aae034=DATE_VIEW.start_date where(DATE_VIEW.start_date>'2014-05-01' and DATE_VIEW.start_date<'2015-01-01');

总结

Hive 与hbase实现了经过一次映射, 能够实时查询hbase中的数据, 也能够从hive表中插入数据到hbase. 经过构建视图的方式能够将多个hive 表的数据整合到一个视图中, 方便数据的使用, 经过以上方式对hbase 中数据的利用不占用数据储存空间.缺点,以上过程无数据清洗过程,可能会存在数据冲突的问题.

参考

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

相关文章
相关标签/搜索