概述

分享下最近做的一个mysql大表归档方案，仅供参考。

---

整体思路

---

一、明确哪些大表需做归档

1、数据库表概要信息统计

SELECTt1.table_schema,t1.table_name,`ENGINE`,table_rows,CAST( data_length / 1024.0 / 1024.0 AS DECIMAL ( 10, 2 ) ) `data_size(M)`,CAST( index_length / 1024.0 / 1024.0 AS DECIMAL ( 10, 2 ) ) `index_size(M)`,t2.ct col_count,t3.ct idx_count,create_time,table_comment FROMinformation_schema.TABLES t1LEFT JOIN -- 字段总数( SELECT table_name, COUNT( 1 ) ct FROM information_schema.COLUMNS GROUP BY table_name ) t2 ON t1.table_name = t2.table_nameLEFT JOIN -- 索引总数( SELECT table_name, COUNT( DISTINCT index_name ) ct FROM information_schema.STATISTICS GROUP BY table_name ) t3 ON t1.table_name = t3.table_name WHEREt1.table_schema NOT IN ( 'mysql', 'information_schema', 'performance_schema' ) ORDER BYt1.data_length DESC;

2、整理需备份的大表

---

二、备份大表

1、表定义

或者用navicat导出表结构

 show create table 表名;

2、表主键、外键

--查看表主键信息 SELECTt.TABLE_NAME,t.CONSTRAINT_TYPE,c.COLUMN_NAME,c.ORDINAL_POSITION FROMINFORMATION_SCHEMA.TABLE_CONSTRAINTS AS t,INFORMATION_SCHEMA.KEY_COLUMN_USAGE AS c WHEREt.TABLE_NAME = c.TABLE_NAME AND t.CONSTRAINT_TYPE = 'PRIMARY KEY' AND t.TABLE_NAME = '' AND t.TABLE_SCHEMA = '';--外键定义SELECTC.TABLE_SCHEMA,C.REFERENCED_TABLE_NAME,C.REFERENCED_COLUMN_NAME,C.TABLE_NAME,C.COLUMN_NAME,C.CONSTRAINT_NAME,T.TABLE_COMMENT,R.UPDATE_RULE,R.DELETE_RULE FROMINFORMATION_SCHEMA.KEY_COLUMN_USAGE CJOIN INFORMATION_SCHEMA.TABLES T ON T.TABLE_NAME = C.TABLE_NAMEJOIN INFORMATION_SCHEMA.REFERENTIAL_CONSTRAINTS R ON R.TABLE_NAME = C.TABLE_NAME AND R.CONSTRAINT_NAME = C.CONSTRAINT_NAME AND R.REFERENCED_TABLE_NAME = C.REFERENCED_TABLE_NAME WHEREC.REFERENCED_TABLE_NAME = 'TABLE_NAME' or C.TABLE_NAME='TABLE_NAME';       

3、表触发器

SELECT * FROM information_schema.triggers where event_object_table='table_name';show create trigger trigger_schema.trigger_name;

4、表索引

 show index from 表名;

---

三、历史数据归档

1、建立归档库及在归档库中创建归档表

create database pas_arch default character set utf8mb4;create table pas_arch.t_att_work_hour_report_dmp1 as select * from pas_prod.t_att_work_hour_report where 1=2;

2、归档历史数据

将历史数据插入到归档库中的归档表

 insert into pas_arch.t_att_work_hour_report_dmp1 select * from  pas_prod.t_att_work_hour_report  where work_date

3、确认数据一致

 select count(*) from pas_prod.t_att_work_hour_report where work_date

---

四、表切换

1、在原始库创建中间表

create table pas_prod.t_att_work_hour_report_tmp as select * from pas_prod.t_att_work_hour_report where 1=2;

2、分段insert

insert into pas_prod.t_att_work_hour_report_tmp select * from pas_prod.t_att_work_hour_report where  work_date>='2020-01-01 00:00:00' and  work_date='2020-03-13 00:00:00' and  work_date

3、切换表

理论上会影响业务一秒钟，建议在业务空闲时间段进行

alter table pas_prod.t_att_work_hour_report rename to pas_prod.t_att_work_hour_report_arch;alter table pas_prod.t_att_work_hour_report_tmp rename to pas_prod.t_att_work_hour_report;

4、数据补录

 insert into t_att_work_hour_report select * from t_att_work_hour_report_arch where work_date>='2020-07-13 00:00:00';

5、依次添加主外键、触发器、约束、索引及授权

6、业务测试

业务测试反馈正常。

---

上面的添加主外键、触发器、约束、索引及授权步骤可以考虑在切换表之前，理论上影响业务一秒钟。

后面会分享更多devops和DBA方面内容，感兴趣的朋友可以关注下~