pt-osc原理
pt-osc原理
1、检查设置环境
测试db是否可连通,并且验证database是否存在
SET SESSION innodb_lock_wait_timeout=1 //InnoDB事务等待行锁的超时时间
SET SESSION lock_wait_timeout=60 //设置获取元数据琐超时为60s
SET SESSION wait_timeout=10000 //交互超时时间
权限验证 show grants for current_user()
2、创建临时表_tablename_new并修改临时表结构
3、在源表创建三个触发器
--创建delete触发器
CREATE TRIGGER `pt_osc_darren_t_user_del` AFTER DELETE ON `darren`.`t_user` FOR EACH ROW DELETE IGNORE FROM `darren`.`_t_user_new`
WHERE `darren`.`_t_user_new`.`id` <=> OLD.`id`
--创建update触发器
CREATE TRIGGER `pt_osc_darren_t_user_upd` AFTER UPDATE ON `darren`.`t_user` FOR EACH ROW
BEGIN
DELETE IGNORE FROM `darren`.`_t_user_new` WHERE !(OLD.`id` <=> NEW.`id`) AND `darren`.`_t_user_new`.`id` <=> OLD.`id`;
REPLACE INTO `darren`.`_t_user_new` (`id`, `name`, `phone`, `gender`, `type`, `birth`, `is_delete`, `c1`, `c2`, `c3`)
VALUES (NEW.`id`, NEW.`name`, NEW.`phone`, NEW.`gender`, NEW.`type`, NEW.`birth`, NEW.`is_delete`, NEW.`c1`, NEW.`c2`, NEW.`c3`);
END;
--创建insert触发器
CREATE TRIGGER `pt_osc_darren_t_user_ins` AFTER INSERT ON `darren`.`t_user`
FOR EACH ROW
REPLACE INTO `darren`.`_t_user_new` (`id`, `name`, `phone`, `gender`, `type`, `birth`, `is_delete`, `c1`, `c2`, `c3`)
VALUES (NEW.`id`, NEW.`name`, NEW.`phone`, NEW.`gender`, NEW.`type`, NEW.`birth`, NEW.`is_delete`, NEW.`c1`, NEW.`c2`, NEW.`c3`)
4、开始迁移数据:
1)估算表的大小,仅仅用于计算进度:EXPLAIN SELECT * FROM `darren`.`t_user` WHERE 1=1;
2)最小值:SELECT `id` FROM `darren`.`t_user` FORCE INDEX(`PRIMARY`) ORDER BY `id` LIMIT 1;
3) 根据chunk-size,计算本次chunk最大值及下一个chunk开始值:SELECT `id` FROM `darren`.`t_user` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= '1'))
ORDER BY `id` LIMIT 2, 2;
4) 插入数据:
INSERT LOW_PRIORITY IGNORE INTO `darren`.`_t_user_new` (`id`, `name`, `phone`, `gender`, `type`, `birth`, `is_delete`, `c1`, `c2`, `c3`)
SELECT `id`, `name`, `phone`, `gender`, `type`, `birth`, `is_delete`, `c1`, `c2`, `c3` FROM `darren`.`t_user` FORCE INDEX(`PRIMARY`)
WHERE ((`id` >= '1')) AND ((`id` <= '3')) LOCK IN SHARE MODE /*pt-online-schema-change 4670 copy nibble*/
5)开始下一轮迭代......
6) 何时copy结束?
根据SELECT `id` FROM `darren`.`t_user` FORCE INDEX(`PRIMARY`) WHERE ((`id` >= '1')) ORDER BY `id` LIMIT 2, 2;
如果获取不到下一个chunk开始值,视为表copy完成,相关源码片段如下:
sub _get_bounds {
my ($self) = @_;
if ( $self->{one_nibble} ) {
if ( $self->{resume} ) {
$self->{no_more_boundaries} = 1;
}
return;
}
my $dbh = $self->{Cxn}->dbh();
$self->{first_lower} = $dbh->selectrow_arrayref($self->{first_lb_sql});
PTDEBUG && _d('First lower boundary:', Dumper($self->{first_lower}));
if ( my $nibble = $self->{resume} ) {
if ( defined $nibble->{lower_boundary}
&& defined $nibble->{upper_boundary} ) {
my $sth = $dbh->prepare($self->{resume_lb_sql});
my @ub = split ',', $nibble->{upper_boundary};
PTDEBUG && _d($sth->{Statement}, 'params:', @ub);
$sth->execute(@ub);
$self->{next_lower} = $sth->fetchrow_arrayref();
$sth->finish();
}
}
else {
$self->{next_lower} = $self->{first_lower};
}
PTDEBUG && _d('Next lower boundary:', Dumper($self->{next_lower}));
if ( !$self->{next_lower} ) {
PTDEBUG && _d('At end of table, or no more boundaries to resume');
$self->{no_more_boundaries} = 1;
$self->{last_upper} = $dbh->selectrow_arrayref($self->{last_ub_sql});
PTDEBUG && _d('Last upper boundary:', Dumper($self->{last_upper}));
}
return;
}
5、rename表
ANALYZE TABLE `darren`.`_t_user_new`;
RENAME TABLE `darren`.`t_user` TO `darren`.`_t_user_old`, `darren`.`_t_user_new` TO `darren`.`t_user`;
6、清理战场
DROP TABLE IF EXISTS `darren`.`_t_user_old`
DROP TRIGGER IF EXISTS `darren`.`pt_osc_darren_t_user_del`
DROP TRIGGER IF EXISTS `darren`.`pt_osc_darren_t_user_upd`
DROP TRIGGER IF EXISTS `darren`.`pt_osc_darren_t_user_ins`
SHOW TABLES FROM `darren` LIKE '_t_user_new'
【copy rows和触发器实现数据迁移和增量写入原理:】
DDL功能 | 映射执行的SQL语句 |
---|---|
insert触发器 | replace into |
update触发器 | delete ignore + replace into |
delete触发器 | delete ignore |
copy rows | insert ignore into |