文章目录
PostgreSQL 10 引入了声明式分区(declarative partitioning)特性,但是实现的功能有限。PostgreSQL 11 为此带来了许多关于分区的增强功能。首先就是增加了 HASH 分区。哈希分区基于分区键的哈希值计算数据所在的分区。
CREATE TABLE htable (c1 bigint, c2 VARCHAR(10)) PARTITION BY HASH(c1);
为哈希分区表创建分区时,使用 FOR VALUES WITH 子句指定分区的计算方法,其中的 MODULUS 子句用于指定除数,REMAINDER 子句用于指定哈希值被除后的余数。
CREATE TABLE htable_p0 PARTITION OF htable FOR VALUES WITH (MODULUS 4, REMAINDER 0);CREATE TABLE htable_p1 PARTITION OF htable FOR VALUES WITH (MODULUS 4, REMAINDER 1);CREATE TABLE htable_p2 PARTITION OF htable FOR VALUES WITH (MODULUS 4, REMAINDER 2);CREATE TABLE htable_p3 PARTITION OF htable FOR VALUES WITH (MODULUS 4, REMAINDER 3);
REMAINDER 子句需要指定一个小于 MODULUS 子句的值。如果指定的分区数量少于 MODULES 子句的值,将会导致无法插入某些数据,因为没有用于存储这些值的分区。
\d+ htableTable "public.htable"Column | Type| Collation | Nullable | Default | Storage | Stats target | Description --------+-----------------------+-----------+----------+---------+----------+--------------+-------------c1| bigint| || | plain | | c2| character varying(10) | || | extended | | Partition key: HASH (c1)Partitions: htable_p0 FOR VALUES WITH (modulus 4, remainder 0),htable_p1 FOR VALUES WITH (modulus 4, remainder 1),htable_p2 FOR VALUES WITH (modulus 4, remainder 2),htable_p3 FOR VALUES WITH (modulus 4, remainder 3)
为表 htable 生成一些数据,查看数据在各个分区的分布是否均匀:
INSERT INTO htable SELECT val,'val:'||val FROM generate_series(1,100000) val;INSERT 0 100000SELECT COUNT(*) FROM htable_p0;count -------25126(1 row)SELECT COUNT(*) FROM htable_p1;count -------24978(1 row)SELECT COUNT(*) FROM htable_p2;count -------24971(1 row)SELECT COUNT(*) FROM htable_p3;count -------24925(1 row)
每个分区大概包含四分之一(25000)的数据。
对于哈希分区,同样支持分区裁剪(Partition Pruning):
show enable_partition_pruning;enable_partition_pruning --------------------------onEXPLAIN SELECT * FROM htable where c1 = 200;QUERY PLAN ------------------------------------------------------------------Append (cost=0.00..470.57 rows=1 width=17)-> Seq Scan on htable_p3 (cost=0.00..470.56 rows=1 width=17)Filter: (c1 = 200)(3 rows)
官方文档:Table Partitioning