Percona-Toolkit 之 pt-archiver 删除历史数据

pt-archiver是Percona-Toolkit工具集中的一个组件,是一个主要用于对MySQL表数据进行归档和清除工具。它可以将数据归档到另一张表或者是一个文件中。pt-archiver在清除表数据的过程中并不会影响OLTP事务的查询性能。对于数据的归档,它可以归档到另一台服务器上的另一张表,也可归档到一个文件中,文件可以用LOAD DATA INFILE进行数据装载,这个功能其实就类似是表历史数据的增量删除。

对线上历史数据的归档,本人直接用mysqldump后还原,但是归档后删除历史数据,通过pt-archiver感觉比写个shell脚本方便很多。

pt-archiver --help 参数说明主要介绍删除历史数据能用到的。

1、--progress 每多少行打印进度信息

2、--limit  限制select返回的行数

3、--sleep  指定select语句休眠时间

4、--txn-size 指定多少行提交一次事务

5、--bulk-delete 用单个DELETE语句批量删除每个行块。该语句删除块的第一行和最后一行之间的每一行,隐含--commit-each

6、--dry-run 打印查询,不做任何操作后退出

 

新建表employees_ptarc测试:

CREATE TABLE `employees_ptarc` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`v_int` int(11) DEFAULT NULL,
`v_string` varchar(50) DEFAULT NULL,
`s_string` char(20) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

 

通过存储过程造数据:

delimiter $$
CREATE PROCEDURE i_employees_ptarc (IN row_num INT)
BEGIN

DECLARE i INT DEFAULT 0 ;
WHILE i < row_num DO
INSERT INTO employees_ptarc (v_int, v_string, s_string)
VALUES
(
floor(1 + rand() * 1000000),
substring(
MD5(RAND()),
1,
floor(1 + rand() * 20)
),
substring(MD5(RAND()), 1, 20)
) ;
SET i = i + 1 ;
END
WHILE ; END$$

delimiter ;

 

call i_employees_ptarc(400000);

 

# pt-archiver --source h=localhost,P=7777,u=root,p='Asd.123@#',D=yoon,t=employees_ptarc,A=utf8 --purge --charset=utf8 --where "id < 400000" --progress=100  --txn-size=100 --limit=100 --bulk-delete  --statistics  --dry-run

SELECT /*!40001 SQL_NO_CACHE */ `id`,`v_int`,`v_string`,`s_string` FROM `yoon`.`employees_ptarc` FORCE INDEX(`PRIMARY`) WHERE (id < 400000) AND (`id` < '400000') ORDER BY `id` LIMIT 100
SELECT /*!40001 SQL_NO_CACHE */ `id`,`v_int`,`v_string`,`s_string` FROM `yoon`.`employees_ptarc` FORCE INDEX(`PRIMARY`) WHERE (id < 400000) AND (`id` < '400000') AND ((`id` >= ?)) ORDER BY `id` LIMIT 100
DELETE FROM `yoon`.`employees_ptarc` WHERE (((`id` >= ?))) AND (((`id` <= ?))) AND (id < 400000) LIMIT 100

生成的删除数据语句为:DELETE FROM `yoon`.`employees_ptarc` WHERE (((`id` >= ?))) AND (((`id` <= ?))) AND (id < 400000) LIMIT 100   如果不是核心业务历史数据,这样操作也是没问题的,但是对于核心库删除历史数据,更希望能够实现的删除语句为:delete from table where id=?  

 

将pt-archiver的参数bulk-delete去掉,改成txn-size,并加上sleep来控制,这样不会影响DB,如下:

#pt-archiver --source h=localhost,P=7777,u=root,p='Asd.123@#',D=yoon,t=employees_ptarc --purge --charset=utf8 --where "id <= 400000" --progress=200  --limit=200 --sleep=1 --txn-size=200  --statistics  --dry-run

SELECT /*!40001 SQL_NO_CACHE */ `id`,`v_int`,`v_string`,`s_string` FROM `yoon`.`employees_ptarc` FORCE INDEX(`PRIMARY`) WHERE (id <= 600000) AND (`id` < '400000') ORDER BY `id` LIMIT 100
SELECT /*!40001 SQL_NO_CACHE */ `id`,`v_int`,`v_string`,`s_string` FROM `yoon`.`employees_ptarc` FORCE INDEX(`PRIMARY`) WHERE (id <= 600000) AND (`id` < '400000') AND ((`id` >= ?)) ORDER BY `id` LIMIT 100
DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = ?)

查看审计日志:

root,localhost,79,608616,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110367\')',0
root,localhost,79,608617,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110368\')',0
root,localhost,79,608618,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110369\')',0
root,localhost,79,608619,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110370\')',0
root,localhost,79,608620,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110371\')',0
root,localhost,79,608621,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110372\')',0
root,localhost,79,608622,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110373\')',0
root,localhost,79,608623,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110374\')',0
root,localhost,79,608624,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110375\')',0
root,localhost,79,608625,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110376\')',0
root,localhost,79,608626,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110377\')',0
root,localhost,79,608627,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110378\')',0
root,localhost,79,608628,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110379\')',0
root,localhost,79,608629,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110380\')',0
root,localhost,79,608630,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110381\')',0
root,localhost,79,608631,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110382\')',0
root,localhost,79,608632,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110383\')',0
root,localhost,79,608633,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110384\')',0
root,localhost,79,608634,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110385\')',0
root,localhost,79,608635,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110386\')',0
root,localhost,79,608636,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110387\')',0
root,localhost,79,608637,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110388\')',0
root,localhost,79,608638,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110389\')',0
root,localhost,79,608639,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110390\')',0
root,localhost,79,608640,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110391\')',0
root,localhost,79,608641,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110392\')',0
root,localhost,79,608642,QUERY,`yoon`,'DELETE FROM `yoon`.`employees_ptarc` WHERE (`id` = \'110393\')',0

 

posted @ 2019-08-28 19:55  __Yoon  阅读(312)  评论(0编辑  收藏  举报