标签:pt-online-schema-change 添加字段过程解析
对于大表的 DDL操作,我们一般使用 pt-online-schema-change 来进行。
具体的操作步骤如下:
1、创建一张新表_xxx_new ,对其做DDL操作
2、创建3个触发器(delete\update\insert),在复制数据开始之后,将对源数据表继续进行数据修改的操作记录下来,以便在数据复制结束后执行这些操作,保证数据不会丢失
3、复制数据,从源数据表复制数据到新表(分成多个chunk,小事务提交)
4、修改外键相关的子表,根据修改后的数据,修改外键关联的子表
5、将源数据表重命名为old表,将新表更改为源表名
6、删除原表
7、删除触发器
下面看一个例子,来亲自验证下这个过程。
如下是一条DDL测试语句:
ALTER TABLE tb_2 ADD COLUMN content text ;
对应的pt-osc写法如下:
pt-online-schema-change --user=root --password=123456 -h localhost --alter "ADD COLUMN content text" D=db1,t=tb_2--no-check-replication-filters --alter-foreign-keys-method=auto --recursion-method=none --print --charset=utf8 --execute
下面是我在使用pt-osc执行上述DDL时候,generallog里面记录的内容(部分不太重要的地方有删减..)
Connectroot@localhost on db1
set autocommit=1
SELECT @@SQL_MODE
/*!40101 SET NAMES "utf8"*/
SHOW VARIABLES LIKE ‘innodb\_lock_wait_timeout‘
SET SESSION innodb_lock_wait_timeout=1
SHOW VARIABLES LIKE ‘lock\_wait_timeout‘
SET SESSION lock_wait_timeout=60
SHOW VARIABLES LIKE ‘wait\_timeout‘
SET SESSION wait_timeout=10000
SET @@SQL_QUOTE_SHOW_CREATE = 1/*!40101, @@SQL_MODE=‘NO_AUTO_VALUE_ON_ZERO,NO_ENGINE_SUBSTITUTION‘*/
SELECT @@server_id /*!50038 , @@hostname*/
Connectroot@localhost on db1
set autocommit=1
SELECT @@SQL_MODE
/*!40101 SET NAMES "utf8"*/
SHOW VARIABLES LIKE ‘innodb\_lock_wait_timeout‘
SET SESSION innodb_lock_wait_timeout=1
SHOW VARIABLES LIKE ‘lock\_wait_timeout‘
SET SESSION lock_wait_timeout=60
SHOW VARIABLES LIKE ‘wait\_timeout‘
SET SESSION wait_timeout=10000
SET @@SQL_QUOTE_SHOW_CREATE = 1/*!40101, @@SQL_MODE=‘NO_AUTO_VALUE_ON_ZERO,NO_ENGINE_SUBSTITUTION‘*/
SELECT @@server_id /*!50038 , @@hostname*/
SHOW VARIABLES LIKE ‘wsrep_on‘
SHOW VARIABLES LIKE ‘version%‘
SHOW ENGINES
SHOW VARIABLES LIKE ‘innodb_version‘
SHOW GLOBAL STATUS LIKE ‘Threads_running‘
SHOW GLOBAL STATUS LIKE ‘Threads_running‘
SELECT CONCAT(@@hostname, @@port)
SHOW VARIABLES
SHOW TABLES FROM `db1` LIKE ‘tb_2‘
### 查看原表是否已存在触发器
SHOW TRIGGERS FROM `db1` LIKE ‘tb_2‘
/*!40101 SET @OLD_SQL_MODE := @@SQL_MODE, @@SQL_MODE := ‘‘, @OLD_QUOTE := @@SQL_QUOTE_SHOW_CREATE, @@SQL_QUOTE_SHOW_CREATE := 1 */
USE `db1`
SHOW CREATE TABLE `db1`.`tb_2`
/*!40101 SET @@SQL_MODE := @OLD_SQL_MODE, @@SQL_QUOTE_SHOW_CREATE := @OLD_QUOTE */
EXPLAIN SELECT * FROM `db1`.`tb_2` WHERE 1=1
SELECT table_schema, table_name FROM information_schema.key_column_usage WHERE referenced_table_schema=‘db1‘ AND referenced_table_name=‘tb_2‘
SHOW VARIABLES LIKE ‘wsrep_on‘
/*!40101 SET @OLD_SQL_MODE := @@SQL_MODE, @@SQL_MODE := ‘‘, @OLD_QUOTE := @@SQL_QUOTE_SHOW_CREATE, @@SQL_QUOTE_SHOW_CREATE := 1 */
USE `db1`
SHOW CREATE TABLE `db1`.`tb_2`
/*!40101 SET @@SQL_MODE := @OLD_SQL_MODE, @@SQL_QUOTE_SHOW_CREATE := @OLD_QUOTE */
#### 创建新表,并对其做ALTER操作
CREATE TABLE `db1`.`_tb_2_new` (
`id` bigint(20) NOT NULL,
`url` varchar(2048) NOT NULL DEFAULT ‘‘,
`appid` smallint(6) NOT NULL,
`rand_code` int(11) NOT NULL DEFAULT ‘0‘ COMMENT ‘随机码‘,
`create_time` bigint(20) NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8
ALTER TABLE `db1`.`_tb_2_new` ADD COLUMN content text
/*!40101 SET @OLD_SQL_MODE := @@SQL_MODE, @@SQL_MODE := ‘‘, @OLD_QUOTE := @@SQL_QUOTE_SHOW_CREATE, @@SQL_QUOTE_SHOW_CREATE := 1 */
USE `db1`
SHOW CREATE TABLE `db1`.`_tb_2_new`
/*!40101 SET @@SQL_MODE := @OLD_SQL_MODE, @@SQL_QUOTE_SHOW_CREATE := @OLD_QUOTE */
##### 创建3个触发器(delete、update、insert) (在原表上update,新临时表上是replace into整行数据,所以达到有则更新,无则插入。同时配合后面的 insert ignore,保证这条数据不会因为重复而失败)
CREATE TRIGGER `pt_osc_db1_tb_2_del` AFTER DELETE ON `db1`.`tb_2` FOR EACH ROW DELETE IGNORE FROM `db1`.`_tb_2_new` WHERE `db1`.`_tb_2_new`.`id` <=> OLD.`id`
CREATE TRIGGER `pt_osc_db1_tb_2_upd` AFTER UPDATE ON `db1`.`tb_2` FOR EACH ROW REPLACE INTO `db1`.`_tb_2_new` (`id`, `url`, `appid`, `rand_code`, `create_time`) VALUES (NEW.`id`, NEW.`url`, NEW.`appid`, NEW.`rand_code`, NEW.`create_time`)
CREATE TRIGGER `pt_osc_db1_tb_2_ins` AFTER INSERT ON `db1`.`tb_2` FOR EACH ROW REPLACE INTO `db1`.`_tb_2_new` (`id`, `url`, `appid`, `rand_code`, `create_time`) VALUES (NEW.`id`, NEW.`url`, NEW.`appid`, NEW.`rand_code`, NEW.`create_time`)
EXPLAIN SELECT * FROM `db1`.`tb_2` WHERE 1=1
SELECT /*!40001 SQL_NO_CACHE */ `id` FROM `db1`.`tb_2` FORCE INDEX(`PRIMARY`) ORDER BY `id` LIMIT 1 /*first lower boundary*/
SELECT /*!40001 SQL_NO_CACHE */ `id` FROM `db1`.`tb_2` FORCE INDEX (`PRIMARY`) WHERE `id` IS NOT NULL ORDER BY `id` LIMIT 1 /*key_len*/
EXPLAIN SELECT /*!40001 SQL_NO_CACHE */ * FROM `db1`.`tb_2` FORCE INDEX (`PRIMARY`) WHERE `id` >= ‘20000000‘ /*key_len*/
### 分块查询数据,减小后续操作的持锁范围
EXPLAIN SELECT /*!40001 SQL_NO_CACHE */ `id` FROM `db1`.`tb_2` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= ‘20000000‘)) ORDER BY `id` LIMIT 999, 2 /*next chunk boundary*/
SELECT /*!40001 SQL_NO_CACHE */ `id` FROM `db1`.`tb_2` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= ‘20000000‘)) ORDER BY `id` LIMIT 999, 2 /*next chunk boundary*/
EXPLAIN SELECT `id`, `url`, `appid`, `rand_code`, `create_time` FROM `db1`.`tb_2` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= ‘20000000‘)) AND ((`id` <= ‘20000999‘)) LOCK IN SHARE MODE /*explain pt-online-schema-change 12296 copy nibble*/
### 开始灌数据操作
INSERT LOW_PRIORITY IGNORE INTO `db1`.`_tb_2_new` (`id`, `url`, `appid`, `rand_code`, `create_time`) SELECT `id`, `url`, `appid`, `rand_code`, `create_time` FROM `db1`.`tb_2` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= ‘20000000‘)) AND ((`id` <= ‘20000999‘)) LOCK IN SHARE MODE /*pt-online-schema-change 12296 copy nibble*/
SHOW WARNINGS
SHOW GLOBAL STATUS LIKE ‘Threads_running‘
。。。 对于操作期间有数据INSERT、UPDATE写入的话,这里还会出现REPLACE INTO 类型的SQL语句 。。。
#### 重命名新、老表名(这个操作期间是锁表的,时间很短暂)
RENAME TABLE `db1`.`tb_2` TO `db1`.`_tb_2_old`, `db1`.`_tb_2_new` TO `db1`.`tb_2`
#### 删除原表
DROP TABLE IF EXISTS `db1`.`_tb_2_old`
#### 删除触发器
DROP TRIGGER IF EXISTS `db1`.`pt_osc_db1_tb_2_del`
DROP TRIGGER IF EXISTS `db1`.`pt_osc_db1_tb_2_upd`
DROP TRIGGER IF EXISTS `db1`.`pt_osc_db1_tb_2_ins`
SHOW TABLES FROM `db1` LIKE ‘\_tb_2\_new‘
Quit
Quit
本文出自 “一只菜鸡的笔记” 博客,请务必保留此出处http://lee90.blog.51cto.com/10414478/1974950
pt-online-schema-change 添加字段过程解析
标签:pt-online-schema-change 添加字段过程解析
原文地址:http://lee90.blog.51cto.com/10414478/1974950