Question

I need to create a tool that is able to merge clients production databases.
Usually these databases will have the same schema (I’ll do some check’s later on, but for now we’ll assume it is). Filtering of duplicate data is something for later on too.
This needs to be done automaticly(so no script generation via SSMS etc).
I’ve already had to start over again a couple of times because every time I ran into problems for things I didn’t think off, so this time I wanted to ask you guys for advice before I begin all over again.

My current plan of action is:

• Copy schema from database 1(later on I’ll add some checks here for
  when the schema is different
• Loop over all tables and set all foreign key updates to cascade, and
  set the order in which the tabledata needs to be inserted (so the
  tables containing the PK’s first, then the tables holding the FK’s)

• Loop every table in the correct order

  • Check in database 2 table for identity column, if so, retrieve the
    current seed value from the corresponding table in database 1, drop
    identity property on database 2 table and update each ID to newID =
    currentID + seed(to avoid duplicate primary keys later on)
  • Generate insert script(SMO’s Table.EnumScript) for database 1 table
  • Generate insert script(SMO’s Table.EnumScript) for database 2 table
  • Execute every line in database 1 insert script on the new database
  • Execute every line in database 2 insert script(which now has primary
    keys/identity field data that will follow the ones in database 1) on the new database

• Go to next table

Everything was working when testing(disabling the identity property in SSMS, created a T-SQL script to update every row with the given seed,..)
But the problem now is automating this in C#, more specific the disabling of the identity property. There doesn’t seem to be a clean solution for this. Creating a new table and rebuilding every constraint etc seems like the wrong way, because the only reason I need it is to cascade every FK so everything still points to the correct place..

Another way would be to delay the updating of the identity-column-data, and change it after script generation and before insertion in the new database. But then I’d need to know which data points to which other data, while everything is still in strings(insertscript)?

Any suggestions,thoughts or techniques on how to handle this?

I know about Red Gate ‘s SQL compare, and it is indeed wonderfull, but need to program it myself.

Using: SMO, SQL Server 2005 – 2008R2(no developers or enterprise edition on client servers), ADO.NET , C#, .NET framework 2.0, Visual Studio 2008

Answer 1

I am not sure exactly what you are trying to accomplish with your process here, but managing Database versions is something that I have a keen interest in.
Have a look at DBSourceTools ( http://dbsourcetools.codeplex.com ).
It is a utility to script an entire database to disk, including all foreign key constraints and data.
Using Deployment Targets, you will then be able to re-create these databases on another database server (usually local machine).
The tool will handle dependencies and large database tables using Sql Bulk insert – trying to generate a script with 50,000 insert statements will be a nightmare.
Have fun.

Disclaimer: I am involved in the http://dbsourcetools.codeplex.com project.