Change the large running table to a partitioned table

The code of this article is limited to monthly horizontal partition based on data time. Other requirements can be implemented by modifying the code

1. Create a partition table

The table field of this table is the same as that of the original table, with partition

CREATE TABLE `metric_data_tmp`  (
    id bigint primary key auto_increment,
    metric varchar(128),
    datadt datetime not null unqine,
    value decimal(30, 6)
) ENGINE=InnoDB AUTO_INCREMENT=0 DEFAULT CHARSET=utf8
partition by range (to_days(DATADT)) (
    PARTITION p201811 VALUES LESS THAN (to_days("2018-12-01")),
    PARTITION p201812 VALUES LESS THAN (to_days("2019-01-01")),
    PARTITION p201901 VALUES LESS THAN (to_days("2019-02-01")),
    PARTITION p201902 VALUES LESS THAN (to_days("2019-03-01")),
);

2. Copy the original table data to the temporary table

• Directly through the insert statement

insert into metric_data_tmp select * from metric_data;

• The data volume is very large. You can use select into outfile and load data file to export and import

SELECT * INTO OUTFILE 'data.txt' FIELDS TERMINATED BY ',' FROM metric_data;
LOAD DATA INFILE 'data.txt' INTO TABLE metric_data_tmp FIELDS TERMINATED BY ',';

3. Rename partition table and history table:

rename table metric_data to metric_data_bak;
rename table metric_data_tmp to metric_data;

4. Automatically create the partition of next month through the scheduled task of database

• stored procedure

delimiter $$
use `db_orbit`$$
drop procedure if exists `create_partition_by_month`$$
create procedure `create_partition_by_month`(in_schemaname varchar(64), in_tablename varchar(64))
begin
    # Used to determine whether the table partition to be created already exists
    declare rows_cnt int unsigned;
    # Time to create table partition
    declare target_date timestamp;
    #The name of the partition in the format p201811
    declare partition_name varchar(8);
        
    #Partition to be created next month
    set target_date = date_add(now(), interval 1 month);
    set partition_name = date_format( target_date, 'p%Y%m' );
        
    # Determine whether the partition to be created exists
    select count(1) into rows_cnt from information_schema.partitions t where table_schema = in_schemaname and table_name = in_tablename and ifnull(t.partition_name, '') = partition_name;
    if rows_cnt = 0 then
        set @sql = concat(
            'alter table `', 
            in_schemaname, 
            '`.`', 
            in_tablename, 
            '`',
            ' add partition (partition ', 
            partition_name, 
            " values less than (to_days('",
            date_format(DATE_ADD(target_date, INTERVAL 1 month), '%Y-%m-01'), 
            "')) engine = innodb);" 
        );
        prepare stmt from @sql;
        execute stmt;
        deallocate prepare stmt;
     else
       select concat("partition `", partition_name, "` for table `",in_schemaname, ".", in_tablename, "` already exists") as result;
     end if;
end$$
delimiter ;

• Create scheduled tasks and execute stored procedures regularly to create partitions

DELIMITER $$
#The database name of the table
USE `db_orbit`$$
CREATE EVENT IF NOT EXISTS `generate_partition_for_metric_data`
ON SCHEDULE EVERY 1 MONTH   #Execution cycle, days, months, etc
STARTS '2019-03-15 00:00:00'
ON COMPLETION PRESERVE
ENABLE
COMMENT 'Creating partitions'
DO BEGIN
    #Call the stored procedure you just created. The first parameter is the database name, and the second parameter is the table name
    CALL db_orbit.create_partition_by_month('db_orbit', 'metric_data');
END$$
DELIMITER ;

5. other

• SQL to view table partition

select 
    partition_name part,  
    partition_expression expr, 
    partition_description descr, 
    table_rows  
from information_schema.partitions where table_name='metric_data';