PostgreSQL 创建表分区

在pg里表分区是通过表继承来实现的，一般都是建立一个主表，里面是空，然后每个分区都去继承它。

创建表分区步骤如下： 

1. 创建主表 

CREATE TABLE users ( uid int not null primary key, name varchar(20)); 

2. 创建分区表(必须继承上面的主表) 

CREATE TABLE users_0 ( check (uid >= 0 and uid< 100) ) INHERITS (users); 

CREATE TABLE users_1 ( check (uid >= 100)) INHERITS (users); 

3. 在分区表上建立索引，其实这步可以省略的哦 

CREATE INDEX users_0_uidindex on users_0(uid); 

CREATE INDEX users_1_uidindex on users_1(uid); 

4. 创建规则RULE 

CREATE RULE users_insert_0 AS 

ON INSERT TO users WHERE 

(uid >= 0 and uid < 100) 

DO INSTEAD 

INSERT INTO users_0 VALUES (NEW.uid,NEW.name); 

CREATE RULE users_insert_1 AS 

ON INSERT TO users WHERE 

(uid >= 100) 

DO INSTEAD 

INSERT INTO users_1 VALUES (NEW.uid,NEW.name); 

下面就可以测试写入数据啦： 

postgres=# INSERT INTO users VALUES (100,'smallfish'); 

INSERT 0 0 

postgres=# INSERT INTO users VALUES (20,'aaaaa'); 

INSERT 0 0 

postgres=# select * from users; 

uid | name 

-----+----------- 

20 | aaaaa 

100 | smallfish 

(2 笔资料列) 

postgres=# select * from users_0; 

uid | name 

-----+------- 

20 | aaaaa 

(1 笔资料列) 

postgres=# select * from users_1; 

uid | name 

-----+----------- 

100 | smallfish 

(1 笔资料列) 

到这里表分区已经可以算完了，不过还有个地方需要修改下，先看count查询把。 

postgres=# EXPLAIN SELECT count(*) FROM users where uid<100; 

QUERY PLAN 

--------------------------------------------------------------------------------------------- 

Aggregate (cost=62.75..62.76 rows=1 width=0) 

-> Append (cost=6.52..60.55 rows=879 width=0) 

-> Bitmap Heap Scan on users (cost=6.52..20.18 rows=293 width=0) 

Recheck Cond: (uid < 100) 

-> Bitmap Index Scan on users_pkey (cost=0.00..6.45 rows=293 width=0) 

Index Cond: (uid < 100) 

-> Bitmap Heap Scan on users_0 users (cost=6.52..20.18 rows=293 width=0) 

Recheck Cond: (uid < 100) 

-> Bitmap Index Scan on users_0_uidindex (cost=0.00..6.45 rows=293 width=0) 

Index Cond: (uid < 100) 

-> Bitmap Heap Scan on users_1 users (cost=6.52..20.18 rows=293 width=0) 

Recheck Cond: (uid < 100) 

-> Bitmap Index Scan on users_1_uidindex (cost=0.00..6.45 rows=293 width=0) 

Index Cond: (uid < 100) 

(14 笔资料列) 

按照本来想法，uid小于100，理论上应该只是查询users_0表，通过EXPLAIN可以看到其他他扫描了所有分区的表。 

postgres=# SET constraint_exclusion = on; 

SET 

postgres=# EXPLAIN SELECT count(*) FROM users where uid<100; 

QUERY PLAN 

--------------------------------------------------------------------------------------------- 

Aggregate (cost=41.83..41.84 rows=1 width=0) 

-> Append (cost=6.52..40.37 rows=586 width=0) 

-> Bitmap Heap Scan on users (cost=6.52..20.18 rows=293 width=0) 

Recheck Cond: (uid < 100) 

-> Bitmap Index Scan on users_pkey (cost=0.00..6.45 rows=293 width=0) 

Index Cond: (uid < 100) 

-> Bitmap Heap Scan on users_0 users (cost=6.52..20.18 rows=293 width=0) 

Recheck Cond: (uid < 100) 

-> Bitmap Index Scan on users_0_uidindex (cost=0.00..6.45 rows=293 width=0) 

Index Cond: (uid < 100) 

(10 笔资料列) 

到这里整个过程都OK啦！