Sql Server Change Data Capture: Preserving history when adding columns?

Question

When a new column is added to table that is configured for change data capture (cdc), the capture instance table will not have the new column until cdc is disabled and re-enabled for the source table. In the process the existing capture instance is dropped.

I thought I could copy existing data out to a temp table and then copy back using the following SQL. However, other CDC meta information, such as the cdc.change_tables.start_lsn, becomes invalid.

How can the capture instance history be preserved, using the same capture instance name, if at all?

Thanks, Rich

/*Change Data Capture Test - Alter table definition test */

/*Enter restricted mode so we don't lose data changes during this process*/
alter database ChangeDataCaptureTest set AUTO_UPDATE_STATISTICS_ASYNC OFF
alter database ChangeDataCaptureTest set RESTRICTED_USER with ROLLBACK IMMEDIATE
go

/*Add a column to the table*/
alter table dbo.Table1 add value3 varchar(20) DEFAULT '' not null

/*Copy the existing change tracking into a temp table*/
select * into cdc.dbo_Table1_temp from cdc.dbo_Table1_CT

/*Add the new column to the temp table so that we don't have to map
all columns when we copy back, note that we use NULL as the default*/
alter table cdc.dbo_Table1_temp add value3 varchar(20) DEFAULT NULL

/*Disable CDC on the source table, this will drop the associated cdc table*/
exec sys.sp_cdc_disable_table 
@source_schema='dbo',
@source_name='Table1', 
@capture_instance='dbo_Table1'

/*Enable CDC for the table which recreates the CDC table*/
EXEC sys.sp_cdc_enable_table
@source_schema = N'dbo',
@source_name   = N'Table1',
@role_name     = NULL,
@supports_net_changes = 1,
@filegroup_name = N'ChangeDataCapture'
GO

/*Insert values from the temp table back into the new CDC Table*/
Insert into cdc.dbo_Table1_CT 
SELECT * 
From cdc.dbo_Table1_temp
go

/*Drop the temp table*/
drop table cdc.dbo_Table1_temp

/*Go back into multi-user mode*/
alter database ChangeDataCaptureTest set AUTO_UPDATE_STATISTICS_ASYNC ON
alter database ChangeDataCaptureTest set MULTI_USER
go

/*Add a new row to the table*/
insert into table1
values(12,'zz','g')

Answer 1

Rich,

The best method to preserve such data is to create a staging persisted table to capture the _CT table data periodically. Knowing that cdc data generally has a short shelf-life prior to be consumed by the endpoint (warehouse/data mart, etc.) you can ensure that any changes are completed within a maintenance window at which time the _CT table data is copied off into staging prior to the changes being implemented.

The one aspect to consider in this is that once the _CT schema has been changed (by adding or removing one or more columns) the process used to pull that data out into the endpoint must also be updated.

To overcome this we implemented a script store that stores the intended schema of the staging table (used between _CT and endpoint) and once the changes are implemented on the client DB, then we move the data from staging into endpoint and update the staging schema.

Hopefully this will provide food for thought.

Answer 2

I think you'd have to write out the lsn records also and then bring them back into the lsntimemapping table.