在SSIS 2012中使用CDC（数据变更捕获）

/*
-- =============================================
-- 创建测试数据库及数据表，借助AdventureWorksDW2012示例数据库
---Generate By downmoon(邀月),3w@live.cn
-- =============================================
*/
--Create database CDCTest
--GO
--USE [CDCTest]
--GO

--SELECT * INTO DimCustomer_CDC
--FROM [AdventureWorksDW2012].[dbo].[DimCustomer]
--WHERE CustomerKey < 11500;

--select * from DimCustomer_CDC;

/*
-- =============================================
-- 启用数据库级别CDC，只对企业版和开发版有效
---Generate By downmoon(邀月),3w@live.cn
-- =============================================
*/
USE
 [CDCTest]
GO

EXEC sys.sp_cdc_enable_db
GO

-- add a primary key to the DimCustomer_CDC table so we can enable support for net changes
IF NOT EXISTS (SELECT * FROM sys.indexes WHERE object_id = 
OBJECT_ID(N‘[dbo].[DimCustomer_CDC]‘) AND name = N‘PK_DimCustomer_CDC‘)
  ALTER TABLE [dbo].[DimCustomer_CDC] ADD CONSTRAINT 
[PK_DimCustomer_CDC] PRIMARY KEY CLUSTERED 
(
    [CustomerKey] ASC
)
GO

/*
-- =============================================
-- 启用表级别CDC
---Generate By downmoon(邀月),3w@live.cn
-- =============================================
*/
EXEC sys.sp_cdc_enable_table 
@source_schema = N‘dbo‘,
@source_name = N‘DimCustomer_CDC‘,
@role_name = N‘cdc_admin‘,
@supports_net_changes = 1

GO

/*
-- =============================================
-- 创建一个目标表，与源表(Source)有相同的表结构
--注意，在生产环境中，完全可以是不同的实例或服务器，本例为了方便，在同一个数据库实例的同一个数据库中演示
---Generate By downmoon(邀月),3w@live.cn
-- =============================================
*/
SELECT TOP 0 * INTO DimCustomer_Destination
FROM DimCustomer_CDC
--select @@version;
select * from DimCustomer_Destination;

mamicode.com,码迷

第二步：创建初始包

-- =============================================
-- 我们使用两个包来完成示例，一个初始包完成数据的初始加载，一个增量包完成数据的变更捕获
---Generate By downmoon(邀月),3w@live.cn
-- =============================================

初始包包含如下逻辑：
（1）使用CDC Control Task标记初始加载开始LSN(Use the CDC Control Task to mark the initial load start LSN)
（2）转换所有源表数据到目标表（Transfer all of the data from the source table into our destination table）
（3）使用CDC Control Task标记初始加载结束LSN（Use the CDC Control Task to mark the initial load end LSN）

示例：http://code.msdn.microsoft.com/My-First-Integration-fa41c0b1

新建一个SSIS项目，创建一个包“Initial Load”，如下图：

mamicode.com,码迷

新建两个CDC Control Task,分别命名为“CDC Control Task Start”和“CDC Control Task End”，分别对应属性为“Mark initial load start”和""Mark initial load end"

连接管理器均为ADO.NET方式，其他属性如下图：

mamicode.com,码迷

中间加入一个“Data Flow Task”,属性默认。

mamicode.com,码迷

此时，运行包，可见CDC_States有初始标记。

mamicode.com,码迷

第三步：创建增量包

增量包包含如下逻辑：
（1）创建一个源数据库的连接管理器(Create a connection manager for the Source database)
（2）设置CDC运算符以获取处理边界（Set the CDC Control Operation to Get processing range）
（3）创建一个新的CDC状态变量(CDC_state)（Create a new CDC state variable (CDC_state)）
（4）创建一个目标数据库的连接管理器(Create a connection manager for the Destination database)
（5）选择前面初始加载包创建的状态表(Select the state table (this was created by the Initial Load package) – [dbo].[cdc_states])
（6）设置状态名称（必须匹配初始加载包使用过的状态名称，this must match what was used in the Initial Load package (CDC_State)）

在项目中创建一个新包，命名为“Incremental Load”

在包的"Control Flow"视图中，自上而下分别手动6个Task，顺序如下图，除去上面用到的三个Task,其余均为Execute SQL Task

mamicode.com,码迷

注意：CDC Control Task End的CDC运算符为MARK Process Range，CDC Control Task Start的CDC运算符为Get Process Range

其余4个Execute SQL Task的SQL语句如下：

--Create stage Tables
IF NOT EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N‘[dbo].[stg_DimCustomer_UPDATES]‘) AND type in (N‘U‘))
BEGIN
   SELECT TOP 0 * INTO stg_DimCustomer_UPDATES
   FROM DimCustomer_Destination
END

IF NOT EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N‘[dbo].[stg_DimCustomer_DELETES]‘) AND type in (N‘U‘))
BEGIN
   SELECT TOP 0 * INTO stg_DimCustomer_DELETES
   FROM DimCustomer_Destination
END

-- batch update
UPDATE dest
SET 
    dest.FirstName = stg.FirstName, 
    dest.MiddleName = stg.MiddleName,
    dest.LastName = stg.LastName, 
    dest.YearlyIncome = stg.YearlyIncome
FROM 
    [DimCustomer_Destination] dest, 
    [stg_DimCustomer_UPDATES] stg
WHERE 
    stg.[CustomerKey] = dest.[CustomerKey]

-- batch delete
DELETE FROM [DimCustomer_Destination]
  WHERE[CustomerKey] IN 
(
    SELECT [CustomerKey]
    FROM [dbo].[stg_DimCustomer_DELETES]
)

-- truncate table 
truncate table  [dbo].[stg_DimCustomer_DELETES]
truncate table  [dbo].[stg_DimCustomer_UPDATES]

最关键的一步，选中CDC Control Task Start，并切换到Data Flow，自上而下分别拖动CDC Source，CDC Splitter Transformer，三个ADO.NET Destination，如下图：

mamicode.com,码迷

其中三个的目标表分别为：[DimCustomer_Destination],stg_DimCustomer_DELETES,stg_DimCustomer_UPDATES。

mamicode.com,码迷

而CDC Source的连接管理器属性如下图：

mamicode.com,码迷

此时，可运行增量包，但我们不会看到任何运行结果，因为此时我们还没有进行数据的Insert或Update操作。

下来我们提供一个脚本，测试下效果：

-- =============================================
-- 更新一些数据,以显示SSIS 2012中CDC的效果
---Generate By downmoon(邀月),3w@live.cn
-- =============================================

USE [CDCTest]
GO
 
-- Transfer the remaining customer rows
SET IDENTITY_INSERT DimCustomer_CDC ON
 
INSERT INTO DimCustomer_CDC
(
       CustomerKey, GeographyKey, CustomerAlternateKey, Title, FirstName, 
       MiddleName, LastName, NameStyle, BirthDate, MaritalStatus, 
       Suffix, Gender, EmailAddress, YearlyIncome, TotalChildren, 
       NumberChildrenAtHome, EnglishEducation, SpanishEducation,
       FrenchEducation, EnglishOccupation, SpanishOccupation, 
       FrenchOccupation, HouseOwnerFlag, NumberCarsOwned, AddressLine1, 
       AddressLine2, Phone, DateFirstPurchase, CommuteDistance
)
SELECT CustomerKey, GeographyKey, CustomerAlternateKey, Title, FirstName, 
       MiddleName, LastName, NameStyle, BirthDate, MaritalStatus, 
       Suffix, Gender, EmailAddress, YearlyIncome, TotalChildren, 
       NumberChildrenAtHome, EnglishEducation, SpanishEducation,
       FrenchEducation, EnglishOccupation, SpanishOccupation, 
       FrenchOccupation, HouseOwnerFlag, NumberCarsOwned, AddressLine1, 
       AddressLine2, Phone, DateFirstPurchase, CommuteDistance
FROM [AdventureWorksDW2012].[dbo].[DimCustomer]
WHERE CustomerKey =11502
 
SET IDENTITY_INSERT DimCustomer_CDC OFF
GO
 
-- give 10 people a raise
UPDATE DimCustomer_CDC 
SET 
    YearlyIncome = YearlyIncome + 10
WHERE
    CustomerKey >= 11000 AND CustomerKey <= 11010
 
GO

此时，我们可以看到变更捕获的结果：

mamicode.com,码迷