白天自动分区 - PostgreSQL

Question

我想做每日分区。我知道oracle就是这样的。

CREATE TABLE "PUBLIC"."TEST"     
(   
"ID" NUMBER(38,0) NOT NULL ENABLE, 
"SOME_FIELD" VARCHAR2(20 BYTE) NOT NULL ENABLE,     
"ANOTHER_FIELD" VARCHAR2(36 BYTE) NOT NULL ENABLE,    
TABLESPACE "PUBLIC"."TEST_DATA" 
PARTITION BY RANGE ("TEST_DATE") INTERVAL (NUMTODSINTERVAL(1,'DAY'))  
(PARTITION "TEST_P1"  
VALUES LESS THAN (TIMESTAMP' 2019-01-01 00:00:00')   TABLESPACE "TEST_DATA" );

那么PostgreSQL呢？

新编辑：示例脚本：

将在一个表中保存前15天数据的脚本在另一个分区中说“p1”和剩余天数据。

1-创建自动分区取决于insert命令的日期范围

2-在脚本中我还提到了我们如何在所需列上添加索引。

3-日期范围从1到14的数据将添加到分区“p1”中，剩余的数据将添加到分区“p2”中。

示例脚本：

    CREATE TABLE measurement (
        city_id         int not null,
        logdate         date not null,
        peaktemp        int,
        unitsales       int
    );

    CREATE OR REPLACE FUNCTION new_partition_creator() RETURNS trigger AS
  $BODY$
    DECLARE
      partition_date TEXT;
      partition TEXT;
      partition_day int;
      startdate date;
      enddate date;
    BEGIN
      partition_day := to_char(NEW.logdate,'DD');
      partition_date := to_char(NEW.logdate,'YYYY_MM');

         IF partition_day < 15 THEN
      partition := TG_RELNAME || '_' || partition_date || '_p1';
      startdate := to_char(NEW.logdate,'YYYY-MM-01');
      enddate := date_trunc('MONTH', NEW.logdate) + INTERVAL '1 MONTH - 1 day';
      ELSE 
      partition := TG_RELNAME || '_' || partition_date || '_p2';
      startdate := to_char(NEW.logdate,'YYYY-MM-15');
      enddate := date_trunc('MONTH', NEW.logdate) + INTERVAL '1 MONTH - 1 day';
      END IF;

      IF NOT EXISTS(SELECT relname FROM pg_class WHERE relname=partition) THEN
        RAISE NOTICE 'A partition has been created %',partition;
        EXECUTE 'CREATE TABLE ' || partition || ' ( CHECK ( logdate >= DATE ''' || startdate || '''  AND logdate <=  DATE ''' ||  enddate || ''' )) INHERITS (' || TG_RELNAME || ');';
        EXECUTE 'CREATE INDEX ' || partition || '_logdate ON '  || partition || '(logdate)';
        EXECUTE 'ALTER TABLE ' || partition || ' add primary key(city_id);';       
        END IF;
        EXECUTE 'INSERT INTO ' || partition || ' SELECT(' || TG_RELNAME || ' ' || quote_literal(NEW) || ').* RETURNING city_id;';
      RETURN NULL;
    END;
  $BODY$
LANGUAGE plpgsql VOLATILE
COST 100; 


    CREATE TRIGGER testing_partition_insert_trigger BEFORE INSERT ON measurement FOR EACH ROW EXECUTE PROCEDURE new_partition_creator();


    postgres=# insert into measurement values(1,'2017-10-11',10,10);
    NOTICE:  A partition has been created measurement_2017_10_p1
    INSERT 0 0

Answer 1

Postgres确实支持对值进行分区。但是，它不会是自动的，因为您需要在创建基表后手动创建分区，从Postgres 10开始，它们不会自动生成。

请参阅以下链接：https://www.postgresql.org/docs/10/ddl-partitioning.html

看看这个例子是否有意义：

CREATE TABLE PartTest
(
    idx INTEGER,
    partMe Date
) PARTITION BY LIST (partMe);

CREATE TABLE PartTest_2019_04_11 PARTITION OF PartTest
    FOR VALUES IN ('2019-04-11');

Answer 2

您可以使用扩展名pg_partman来创建自动分区。 https://github.com/pgpartman/pg_partman

或者您甚至可以使用scheduler pg_agent，您将在每天18:00:00执行一个过程来创建下一天的分区。