玩转SQL Server复制回路の变更数据类型、未分区表转为分区表
复制的应用:
初级应用:读写分离、数据库备份
高级应用:搬迁大型数据库(跨机房)、变更数据类型、未分区表转为分区表
京东的复制专家 菠萝 曾经写过文章、在数据库大会上也做过演讲,但是我相信真正按照菠萝兄的文章自己去做一次实验的人应该不多
京东的复制专家 菠萝 的文章地址:Replication的犄角旮旯(一)--变更订阅端表名的应用场景
为什麽要玩转复制,大家想象一下:变更数据类型、未分区表转为分区表 这些业务场景经常都会发生,特别在数据量特别大的公司
变更数据类型:没有其他特别好的办法,数据量大,锁表时间会比较长
未分区表转为分区表:有时候一张表的数据量已经很多了,比如体积已经达到100G,那么这时候需要做表分区,方法是重建聚集索引或者导数据
上面的方法不多不少都有一些缺陷,对于数据量特别大的情况下,如果超出业务的预期停机时间……菊花残,满地伤,被领导认为办事不力
常见场景:
1、变更其中的自增列主键,int-》bigint ,将表改为表分区
2、100G+的大表
3、单次最长停机时间:为1小时
下面介绍一下,如何在一个实例下,通过三个数据库,建立一个复制回路,完成上面的需求
实验环境:一台电脑,一个SQL Server实例,SQL Server2012, Windows7
复制类型为事务复制
结构图
从上图可以看出,由于都是在同一个实例,同一台机器下,所以机器磁盘需要有足够的磁盘空间!!
因为[testloopbackA]库有一个[testAltertype]表100G,复制到[testloopbackB]库[testAltertype]表100G
复制到[testloopbackC]库[testAltertype]表100G,最后复制回去[testloopbackA]库[testAltertype]表100G
加上生成的快照文件,当然快照文件可能会压缩,但是一定要保证有足够的磁盘空间
下面是具体演示
1、建库脚本
USE [master]GO/****** Object: Database [testloopbackA] Script Date: 2015/6/3 8:21:01 ******/CREATE DATABASE [testloopbackA] CONTAINMENT = NONE ON PRIMARY ( NAME = N'testloopbackA', FILENAME = N'D:\DataBase\testloopbackA.mdf' , SIZE = 30720KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1024KB ), FILEGROUP [FG_testChangepartition_Id_01] ( NAME = N'FG_testChangepartition_Id_01_data', FILENAME = N'D:\DataBase\testloopbackA\FG_testChangepartition_Id_01_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ), FILEGROUP [FG_testChangepartition_Id_02] ( NAME = N'FG_testChangepartition_Id_02_data', FILENAME = N'D:\DataBase\testloopbackA\FG_testChangepartition_Id_02_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ) LOG ON ( NAME = N'testloopbackA_log', FILENAME = N'D:\DataBase\testloopbackA_log.ldf' , SIZE = 2432KB , MAXSIZE = 2048GB , FILEGROWTH = 10%)GOUSE [master]GO/****** Object: Database [testloopbackB] Script Date: 2015/6/3 8:22:11 ******/CREATE DATABASE [testloopbackB] CONTAINMENT = NONE ON PRIMARY ( NAME = N'testloopbackB', FILENAME = N'D:\DataBase\testloopbackB.mdf' , SIZE = 30720KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1024KB ), FILEGROUP [FG_testChangepartition_Id_01] ( NAME = N'FG_testChangepartition_Id_01_data', FILENAME = N'D:\DataBase\testloopbackB\FG_testChangepartition_Id_01_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ), FILEGROUP [FG_testChangepartition_Id_02] ( NAME = N'FG_testChangepartition_Id_02_data', FILENAME = N'D:\DataBase\testloopbackB\FG_testChangepartition_Id_02_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ) LOG ON ( NAME = N'testloopbackB_log', FILENAME = N'D:\DataBase\testloopbackB_log.ldf' , SIZE = 2432KB , MAXSIZE = 2048GB , FILEGROWTH = 10%)GOUSE [master]GO/****** Object: Database [testloopbackC] Script Date: 2015/6/3 8:22:14 ******/CREATE DATABASE [testloopbackC] CONTAINMENT = NONE ON PRIMARY ( NAME = N'testloopbackC', FILENAME = N'D:\DataBase\testloopbackC.mdf' , SIZE = 30720KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1024KB ), FILEGROUP [FG_testChangepartition_Id_01] ( NAME = N'FG_testChangepartition_Id_01_data', FILENAME = N'D:\DataBase\testloopbackC\FG_testChangepartition_Id_01_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ), FILEGROUP [FG_testChangepartition_Id_02] ( NAME = N'FG_testChangepartition_Id_02_data', FILENAME = N'D:\DataBase\testloopbackC\FG_testChangepartition_Id_02_data.ndf' , SIZE = 24576KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1048576KB ) LOG ON ( NAME = N'testloopbackC_log', FILENAME = N'D:\DataBase\testloopbackC_log.ldf' , SIZE = 2432KB , MAXSIZE = 2048GB , FILEGROWTH = 10%)GO
View Code
下面分区方案和分区函数都在三个库上执行
--1.创建分区函数CREATE PARTITION FUNCTIONFun_testChangepartition_Id(INT) ASRANGE LEFTFOR VALUES(2)--2.创建分区方案CREATE PARTITION SCHEME[Sch_testChangepartition_Id] ASPARTITION [Fun_testChangepartition_Id]TO([FG_testChangepartition_Id_01],[FG_testChangepartition_Id_02])
建表脚本
USE [testloopbackA]GO--更改数据类型CREATE TABLE [testAltertype](id INT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100))GO--变分区表CREATE TABLE [testChangepartition](id INT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100))GO--插入测试数据INSERT INTO [dbo].[testAltertype] ( [name] )VALUES ( N'nihao' -- name - nvarchar(100) )INSERT INTO [dbo].[testChangepartition] ( [name] )VALUES ( N'nihao' -- name - nvarchar(100) )SELECT * FROM [testAltertype]SELECT * FROM [testChangepartition]
View Code
2、在[testloopbackB]库先建好2个表
USE [testloopbackB]GO--更改数据类型CREATE TABLE testAltertype_new(id BIGINT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100))GO--变分区表CREATE TABLE testChangepartition_new(id INT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100)) ON [Sch_testChangepartition_Id](id)GO
3、创建[testloopbackA]库到[testloopbackB]库的发布,这一步很关键,因为在发布的时候需要修改项目属性,在发布属性里,还需要选择快照为字符类型
testChangepartition_new表
testAltertype_new表
[testloopbackA]库到[testloopbackB]库的复制
4、建立[pub_testloopbackAtotestloopbackB]发布的订阅
5、在[testloopbackB]库里, 将[testAltertype_new]表和[testChangepartition_new]表里的id列里的不用于复制设置为"是"
[testAltertype_new]表
[testChangepartition_new]表
6、测试
在[testloopbackA]库的[testAltertype]表和[testChangepartition]表各插入一些记录
USE [testloopbackA]GO--插入测试数据INSERT INTO [dbo].[testAltertype] ( [name] )VALUES ( N'nihao2' -- name - nvarchar(100) )INSERT INTO [dbo].[testChangepartition] ( [name] )VALUES ( N'nihao2' -- name - nvarchar(100) )SELECT * FROM [testAltertype]SELECT * FROM [testChangepartition]
在[testloopbackB]库就能看到新插入的记录
USE [testloopbackB]GOSELECT * FROM [dbo].[testAltertype_new]SELECT * FROM [dbo].[testChangepartition_new]
在[testloopbackB]库里执行
USE [testloopbackB]GO--查看分区架构文件组分布SELECT CONVERT(VARCHAR(MAX), ps.name) AS partition_scheme , p.partition_number , CONVERT(VARCHAR(MAX), ds2.name) AS filegroup , CONVERT(VARCHAR(MAX), ISNULL(v.value, ''), 120) AS range_boundary , STR(p.rows, 9) AS rowsFROM sys.indexes i JOIN sys.partition_schemes ps ON i.data_space_id = ps.data_space_id JOIN sys.destination_data_spaces dds ON ps.data_space_id = dds.partition_scheme_id JOIN sys.data_spaces ds2 ON dds.data_space_id = ds2.data_space_id JOIN sys.partitions p ON dds.destination_id = p.partition_number AND p.object_id = i.object_id AND p.index_id = i.index_id JOIN sys.partition_functions pf ON ps.function_id = pf.function_id LEFT JOIN sys.Partition_Range_values v ON pf.function_id = v.function_id AND v.boundary_id = p.partition_number - pf.boundary_value_on_rightWHERE i.object_id = OBJECT_ID('testChangepartition_new') AND i.index_id IN ( 0, 1 )ORDER BY p.partition_number
数据已经入到相应分区
7、继续将[testloopbackB]库的[testAltertype_new]表和[testChangepartition_new]表复制到[testloopbackC]
这一步需要注意:[testAltertype_new]表不需要再跟[testloopbackA]库到[testloopbackB]库的复制那样设置项目属性->XX_new,只需要保持默认就行了
[testChangepartition_new]表跟刚才一样,需要设置项目属性->XX_new
先在[testloopbackC]库建好 [testChangepartition_new]表,[testAltertype_new]表不需要预先建立
USE [testloopbackC]GO--变分区表CREATE TABLE testChangepartition_new(id INT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100)) ON [Sch_testChangepartition_Id](id)GO
8、建[testloopbackB]库到[testloopbackC]库的发布
8、建立订阅[testloopbackC]库
启动快照初始化
然后需要对[testChangepartition_new]表设置不用于复制 为“是”
[testAltertype_new]表不需要设置
9、测试
[testloopbackA]库插入的记录,[testloopbackC]库马上能看到
USE [testloopbackC]go--查看分区架构文件组分布SELECT CONVERT(VARCHAR(MAX), ps.name) AS partition_scheme , p.partition_number , CONVERT(VARCHAR(MAX), ds2.name) AS filegroup , CONVERT(VARCHAR(MAX), ISNULL(v.value, ''), 120) AS range_boundary , STR(p.rows, 9) AS rowsFROM sys.indexes i JOIN sys.partition_schemes ps ON i.data_space_id = ps.data_space_id JOIN sys.destination_data_spaces dds ON ps.data_space_id = dds.partition_scheme_id JOIN sys.data_spaces ds2 ON dds.data_space_id = ds2.data_space_id JOIN sys.partitions p ON dds.destination_id = p.partition_number AND p.object_id = i.object_id AND p.index_id = i.index_id JOIN sys.partition_functions pf ON ps.function_id = pf.function_id LEFT JOIN sys.Partition_Range_values v ON pf.function_id = v.function_id AND v.boundary_id = p.partition_number - pf.boundary_value_on_rightWHERE i.object_id = OBJECT_ID('testChangepartition_new') AND i.index_id IN ( 0, 1 )ORDER BY p.partition_number--分区区间--SELECT * FROM sys.partition_range_values
View Code
数据入到相应分区
10、跟第7步一样,但是这一次是[testloopbackC]库到[testloopbackA]库
先在[testloopbackA]库建好 [testChangepartition_new]表,[testAltertype_new]表不需要预先建立
USE [testloopbackA]GO--变分区表CREATE TABLE testChangepartition_new(id INT IDENTITY(1,1) PRIMARY KEY,name NVARCHAR(100)) ON [Sch_testChangepartition_Id](id)GO
11、建[testloopbackC]库到[testloopbackA]库的发布
12、建立订阅[testloopbackA]库
启动快照初始化
然后需要对[testChangepartition_new]表设置不用于复制 为“是”
[testAltertype_new]表不需要设置
13、测试
[testloopbackA]库插入的记录,[testloopbackA]库马上能看到
USE [testloopbackA]GOSELECT * FROM [dbo].[testAltertype_new]SELECT * FROM [dbo].[testChangepartition_new]
USE [testloopbackA]go--查看分区架构文件组分布SELECT CONVERT(VARCHAR(MAX), ps.name) AS partition_scheme , p.partition_number , CONVERT(VARCHAR(MAX), ds2.name) AS filegroup , CONVERT(VARCHAR(MAX), ISNULL(v.value, ''), 120) AS range_boundary , STR(p.rows, 9) AS rowsFROM sys.indexes i JOIN sys.partition_schemes ps ON i.data_space_id = ps.data_space_id JOIN sys.destination_data_spaces dds ON ps.data_space_id = dds.partition_scheme_id JOIN sys.data_spaces ds2 ON dds.data_space_id = ds2.data_space_id JOIN sys.partitions p ON dds.destination_id = p.partition_number AND p.object_id = i.object_id AND p.index_id = i.index_id JOIN sys.partition_functions pf ON ps.function_id = pf.function_id LEFT JOIN sys.Partition_Range_values v ON pf.function_id = v.function_id AND v.boundary_id = p.partition_number - pf.boundary_value_on_rightWHERE i.object_id = OBJECT_ID('testChangepartition_new') AND i.index_id IN ( 0, 1 )ORDER BY p.partition_number--分区区间--SELECT * FROM sys.partition_range_values
View Code
数据进入到相应分区
接下来就是找个适当的时间,比如凌晨, 停写, 拆复制, 改表名 打完 收工!!
总结
在搭建复制回路的过程当中,本人发现加字段是不行的,比如testloopbackA库testAddcolumn_new表有四个字段
然后预先在testloopbackB库建立testAddcolumn_new表,并增加一个字段,在快照初始化的时候报错
错误消息:进程无法向表“"dbo"."testAddcolumn_new"”进行大容量复制。 (源: MSSQL_REPL,错误号: MSSQL_REPL20037)获取帮助: http://help/MSSQL_REPL20037已达到文件末尾,缺少结束符或字段数据不完整若要获取详细说明初始化订阅表时所遇到的错误的错误文件,请执行在下面显示的 bcp 命令。有关该 bcp 实用工具及其支持的选项的详细信息,请参阅 BOL。 (源: MSSQLServer,错误号: 20253)获取帮助: http://help/20253bcp "testloopbackB"."dbo"."testAddcolumn_new" in "E:\DataBase\ReplData\unc\NAME-PC_TESTLOOPBACKA_PUB_TESTLOOPBACKATOTES1cf75016\20150604115556\testAddcolumn_2.bcp" -e "errorfile" -t"\n<x$3>\n" -r"\n<,@g>\n" -m10000 -SNAME-PC -T -w (源: MSSQLServer,错误号: 20253)获取帮助: http://help/20253
方法一
在这种情况下,我们可以采取跟MySQL主从一样的策略,建立好发布订阅之后,在订阅上添加字段,这时候订阅表会应用发布端发过来的DML和加字段DDL
但是,基本上影响不是很大,当添加字段DDL执行完毕之后,就可以切表了
方法二
可以考虑在停机维护窗口添加或者考虑升级到SQL Server2012 ,SQL Server2012 对加字段已经作了一些修改
对阻塞减少到最低
相关文章:Sql Server 2012新特性 Online添加非空栏位.