Automatic Data Collection of Database Schema Changes in MS SQL Server

Introduction

Have you ever faced a situation when you need to make changes in a stored procedure or a view very quickly? I have, very often, especially at the implementation stage. Unfortunately, a version control system cannot help in this case. Still, how could I understand that something has been modified, and when?

This article describes a possible solution for automatic data collection about database schema changes in MS SQL Server. As usual, I will be glad to hear any alternative solutions.

Solution

1. Create two tables: the first one will be for each database, the second one – for all databases:

   USE [DATABASE_NAME]
   GO
   
   SET ANSI_NULLS ON
   GO
   
   SET QUOTED_IDENTIFIER ON
   GO
   
   CREATE TABLE [srv].[ddl_log](
       [DDL_Log_GUID] [uniqueidentifier] NOT NULL,
       [PostTime] [datetime] NOT NULL,
       [DB_Login] [nvarchar](255) NULL,
       [DB_User] [nvarchar](255) NULL,
       [Event] [nvarchar](255) NULL,
       [TSQL] [nvarchar](max) NULL,
    CONSTRAINT [PK_ddl_log] PRIMARY KEY CLUSTERED 
   (
       [DDL_Log_GUID] ASC
   )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
   ) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]
   
   GO
   
   ALTER TABLE [srv].[ddl_log] ADD  CONSTRAINT [DF_ddl_log_DDL_Log_GUID]  DEFAULT (newid()) FOR [DDL_Log_GUID]
   GO
   
   USE [DATABASE_NAME]
   GO
   
   SET ANSI_NULLS ON
   GO
   
   SET QUOTED_IDENTIFIER ON
   GO
   
   CREATE TABLE [srv].[ddl_log_all](
       [DDL_Log_GUID] [uniqueidentifier] NOT NULL,
       [Server_Name] [nvarchar](255) NOT NULL,
       [DB_Name] [nvarchar](255) NOT NULL,
       [PostTime] [datetime] NOT NULL,
       [DB_Login] [nvarchar](255) NULL,
       [DB_User] [nvarchar](255) NULL,
       [Event] [nvarchar](255) NULL,
       [TSQL] [nvarchar](max) NULL,
       [InsertUTCDate] [datetime] NOT NULL,
    CONSTRAINT [PK_ddl_log_all] PRIMARY KEY CLUSTERED 
   (
       [DDL_Log_GUID] ASC
   )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
   ) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]
   
   GO
   
   ALTER TABLE [srv].[ddl_log_all] ADD  CONSTRAINT [DF_ddl_log_all_DDL_Log_GUID]  DEFAULT (newid()) FOR [DDL_Log_GUID]
   GO
   
   ALTER TABLE [srv].[ddl_log_all] ADD  CONSTRAINT [DF_ddl_log_all_InsertUTCDate]  DEFAULT (getutcdate()) FOR [InsertUTCDate]
   GO

2. Create a DDL-trigger for a database which collects scheme changes:

   USE [DATABASE_NAME]
   GO
   
   SET ANSI_NULLS ON
   GO
   
   SET QUOTED_IDENTIFIER ON
   GO
   
   CREATE TRIGGER [SchemaLog] 
   ON DATABASE --ALL SERVER 
   FOR DDL_DATABASE_LEVEL_EVENTS 
   AS
       SET NOCOUNT ON;
       SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED;
       DECLARE @data XML
       begin try
       if(CURRENT_USER<>'NT AUTHORITY\NETWORK SERVICE' and SYSTEM_USER<>'NT AUTHORITY\NETWORK SERVICE')
       begin
           SET @data = EVENTDATA();
           INSERT srv.ddl_log(
                       PostTime,
                       DB_Login,
                       DB_User,
                       Event,
                       TSQL
                     ) 
           select 
                       GETUTCDATE(),
                       CONVERT(nvarchar(255), SYSTEM_USER),
                       CONVERT(nvarchar(255), CURRENT_USER), 
                       @data.value('(/EVENT_INSTANCE/EventType)[1]', 'nvarchar(255)'), 
                       @data.value('(/EVENT_INSTANCE/TSQLCommand)[1]', 'nvarchar(max)')
           where       @data.value('(/EVENT_INSTANCE/EventType)[1]', 'nvarchar(255)') not in('UPDATE_STATISTICS', 'ALTER_INDEX')
                   and @data.value('(/EVENT_INSTANCE/TSQLCommand)[1]', 'nvarchar(max)') not like '%Msmerge%'; 
                       --there is no need in tracking changes of replication objects
       end
       end try
       begin catch
       end catch
   
   GO
   
   SET ANSI_NULLS OFF
   GO
   
   SET QUOTED_IDENTIFIER OFF
   GO
   
   ENABLE TRIGGER [SchemaLog] ON DATABASE
   GO

I recommend adjusting a filter and not making a DDL-trigger for the whole server. It is useless, as you will get a lot of needless information. In this case, it is better to create a trigger for each database.
However, you will have to turn off this trigger during complicated operations, for example, replication. But later, you will be able to turn it on again.

3.  You will need to gather information in a single table. For example, you can do it with a task in the SQL Server Agent one time a week.
4. It is possible to gather everything in one table by another way you prefer.

Additionally, I recommend deleting old data.

Result

In this article, I have analyzed an example of implementing an automatic data collection about changes of databases schemes in MS SQL Server. It allows us to find out what and when has been modified, and, if necessary, to revert them. In general, this solution may be helpful at the implementation stage where there are a lot of mistakes and when we have different versions of databases copies to be analyzed. If you wish to find out a reason for changes, you can do it by retrieving a revision history.

Also read:

Automatic Data Collection about Completed Tasks in MS SQL Server