[原创]PostgreSQL Plus Advanced Server批量创建分区表写入亿级别数据实例

当前情况：大表的数据量已接近2亿条
我的解决思路：为它创建n*100个分区表，将各个分区表放在不同的tablespace上
这样做的优点：
1、首先是对这个级别的数据表的性能会有所提升
2、数据管理更科学
3、生产运维过程故障排查便捷

实验环境
VM Ware ESXi
Redhat 6.3 64bit
PostgreSQL 9.2/9.3
内存：2G

实现方法（同样适用于Oracle）

1、创建测试表

/**
创建300个分区表，每张表写入5万数据
*/
-- -------------------- step1 : CREATE TABLE--------------------
DROP TABLE parttest;
CREATE TABLE parttest (some_test TEXT,some_tstimestamptz TEXT,random_value INT4, p_condition DATE)
PARTITION BY RANGE(p_condition)
(
    PARTITION p_2013_09_02_and_early VALUES LESS THAN ('2013-09-03')
);

 

2、已时间为分区条件，创建300个分区表的过程

-- -------------------- step2 : CREATE PARTITION PROCEDUCE --------------------
CREATE OR REPLACE PROCEDURE create_partition_withindex() AS

DECLARE
   tblname varchar := 'parttest';
   tblname_prefix varchar := 'p_condition';   
   part_tbl_name varchar;
   current_part_name varchar;
   sql_addpartition_cmd varchar;
   
BEGIN
    -- POPULATE SQL COMMAND
    -- PARTITION TABLE'S NAME IS LIKE "TABLENAME_p_2019_12_31"
    FOR j IN 1..300 LOOP
    current_part_name := to_char(current_date + j,'YYYY_MM_DD');
    part_tbl_name := tblname_prefix || '_p_' || current_part_name;

    -- POPULATE SQL COMMAND STRING
        sql_addpartition_cmd = 'ALTER TABLE ' || tblname || ' ADD PARTITION p_' || current_part_name || ' VALUES LESS THAN (''' || (current_date + j + 1) || ''');';
    DBMS_OUTPUT.PUT_LINE(sql_addpartition_cmd);
        EXECUTE IMMEDIATE sql_addpartition_cmd;

    IF ((j % 30) = 0) THEN
        COMMIT;
    END IF;
        
    END LOOP;

    EXECUTE IMMEDIATE 'ALTER TABLE ' || tblname || ' ADD PARTITION p_' || to_char(current_date + 300 + 1,'YYYY_MM_DD') || '_and_later VALUES LESS THAN (''' || (current_date + 300 + 2) || ''');';
END;

3、创建分区表

-- -------------------- step3 CREATE PARTITION --------------------
EXEC create_partition_withindex;

DEMO测试
1、创建写入数据的存储过程

-- -------------------- step4 CREATE TEST DATA PROCEDUCE --------------------

CREATE OR REPLACE PROCEDURE create_partition_testdata() AS

DECLARE

BEGIN
    FOR i IN 0..300 LOOP
    INSERT INTO parttest (some_test, some_tstimestamptz, random_value,p_condition)
    SELECT
        'user #' || cast(floor(random() * 10) as int4),    
    now() - '1 year'::INTERVAL * random(),
    cast(random() * 100000000 as INT4),
    (current_date + i)
    FROM
    generate_series(1,500000);
    COMMIT;
    END LOOP;
END;

2、执行写入数据的存储过程

-- -------------------- step5 CREATE TEST DATA --------------------
EXEC create_partition_testdata;

3、检查写入数据的数据量

SELECT COUNT(*) FROM parttest;

如果有必要，在写入数据时可以记录一下创建分区表及写入数据的时间数据。