Move SQL Server Tables to Different Filegroups

By:   |   Comments (2)   |   Related: More > Database Administration


Problem

Sometimes it's necessary to move SQL Server tables between filegroups or create a copy of a table in a different filegroup. Reasons for having a copy of a table in a different filegroup could be for archiving historical data, using a copy of the table for reporting or for testing purposes. In this tip we will look at both scenarios and how it can be done using T-SQL.

Solution

This article will show methods of copying tables to another filegroup in SQL Server 2016/2017 and in older versions of SQL Server. It will also show how to use SELECT…INTO in SQL 2016 ((13.x) SP2) and 2017 to create tables in a different filegroup.

To illustrate the solution, a test environment is needed, so let's start with the creation of a test database.

Creating the test environment

The script below creates the TestDB database with two tables – UserData and UserLog that stores login information. Both of these tables will be place in the default (primary) filegroup.

USE master
GO

 --Database
CREATE DATABASE TestDB
GO

USE TestDB
GO

--Tables
CREATE TABLE UserData
(
  UserID INT NOT NULL,
  LoginName NVARCHAR(50),
  PRIMARY KEY (UserID)
)
GO 

CREATE TABLE UserLog
(
  UserLogID INT NOT NULL IDENTITY(1,1),
  UserID INT NOT NULL,
  LoginDate DATETIME DEFAULT GETDATE(),
  PRIMARY KEY (UserLogID)
)
GO 

--Data
INSERT INTO UserData(UserID,LoginName)
VALUES(1,'[email protected]'),
      (2,'[email protected] '),
      (3,'[email protected]'),
      (4,'[email protected]')

INSERT INTO UserLog(UserID)
VALUES(1),(2)

WAITFOR DELAY '00:00:10'

INSERT INTO UserLog(UserID)
VALUES(1),(3),(4)

WAITFOR DELAY '00:00:10'

INSERT INTO UserLog(UserID)
VALUES(2),(3),(1)

Now, we'll create a new filegroup to store historical data related to user logins. In other words, the UserData table (or some filtered data from that table) should be moved to a new filegroup.

So, let's create a new filegroup.

USE master
GO

ALTER DATABASE TestDB ADD FILEGROUP HISTORY
ALTER DATABASE TestDB
ADD FILE
(
NAME='History_Data',
FILENAME = 'D:\Microsoft SQL Server\MSSQL14.MSSQLSERVER\MSSQL\DATA\TesDB_2.mdf'
)
TO FILEGROUP HISTORY
GOGO

Running the query below, we can see that we have two filegroups for the database.

USE TestDB
GO

SELECT * FROM sys.filegroups
tables and filegroups

However, both tables in our database are in the PRIMARY filegroup:

USE TestDB
GO

SELECT o.[name] AS TableName, i.[name] AS IndexName, fg.[name] AS FileGroupName
FROM sys.indexes i
INNER JOIN sys.filegroups fg ON i.data_space_id = fg.data_space_id
INNER JOIN sys.all_objects o ON i.[object_id] = o.[object_id]
WHERE i.data_space_id = fg.data_space_id AND o.type = 'U'
tables and filegroups

Now, suppose we have a task to move the UserLog table to the HISTORY filegroup.

Moving a SQL Server table with data to a different filegroup

Moving table with a clustered index

One solution to move a table to another filegroup is by dropping the clustered index and using the MOVE TO option as follows. We can see the IndexName in the above screenshot.

USE TestDB
GO

ALTER TABLE UserLog
DROP CONSTRAINT PK__UserLog__7F8B815172CE9EAE WITH (MOVE TO HISTORY)

As one of our readers pointed out, if we want to keep a clustered index on the table, we could use the following method.  This will drop the existing clustered index and create a new clustered index in the specified filegroup without having to create the clustered index in separate step.

USE TestDB
GO

CREATE UNIQUE CLUSTERED INDEX PK__UserLog__7F8B815172CE9EAE ON UserLog (UserLogID)  
    WITH (DROP_EXISTING = ON)  
    ON HISTORY

We can now run this query to see which filegroup is being used.

USE TestDB
GO

SELECT o.[name] AS TableName, i.[name] AS IndexName, fg.[name] AS FileGroupName
FROM sys.indexes i
INNER JOIN sys.filegroups fg ON i.data_space_id = fg.data_space_id
INNER JOIN sys.all_objects o ON i.[object_id] = o.[object_id]
WHERE i.data_space_id = fg.data_space_id AND o.type = 'U'

We can see the UserLog table is now in the HISTORY filegroup. However, the table no longer has a clustered index. If you need a clustered index, you would need to create one for the UserLog table.

tables and filegroups

Moving table without a clustered index

If we do not have a clustered index on the table, we can create a clustered index and specifying which filegroup to use.

For instance, if we want to move the UserLog table that we just moved to the HISTORY filegroup back to the PRIMARY filegroup, we could issue the following command.

USE TestDB
GO

CREATE UNIQUE CLUSTERED INDEX UIX_UserLogID
ON UserLog(UserLogID) ON [PRIMARY]

Run the following again.

USE TestDB
GO

SELECT o.[name] AS TableName, i.[name] AS IndexName, fg.[name] AS FileGroupName
FROM sys.indexes i
INNER JOIN sys.filegroups fg ON i.data_space_id = fg.data_space_id
INNER JOIN sys.all_objects o ON i.[object_id] = o.[object_id]
WHERE i.data_space_id = fg.data_space_id AND o.type = 'U'

We can see the UserLog table is back in the PRIMARY filegroup and has an index.

tables and filegroups

If we do not need the clustered index, we would need to run additional code to drop it as follows.

USE TestDB
GO

DROP INDEX UserLog.UIX_UserLogID

Hence, if we are moving a heap to another filegroup, we first need to create a clustered index in order to move it to another filegroup and then we could drop the clustered index.

Creating a Copy of a SQL Server Table and Data on Different Filegroup

What if we don't want to move the table, but just create a copy in another filegroup.

Prior to SQL Server 2016 SP2

We could do this as follows where we use SELECT INTO, then create a clustered index and specify the filegroup, then drop the clustered index if we don't want the clustered index.

USE TestDB
GO

SELECT * INTO UserLogHistory FROM UserLog

CREATE UNIQUE CLUSTERED INDEX UIX_UserLogID
ON UserLogHistory(UserLogID) ON [HISTORY]

DROP INDEX UserLogHistory.UIX_UserLogID

As a result, we have a new table UserLogHistory in the HISTORY filegroup:

USE TestDB
GO

SELECT o.[name] AS TableName, i.[name] AS IndexName, fg.[name] AS FileGroupName
FROM sys.indexes i
INNER JOIN sys.filegroups fg ON i.data_space_id = fg.data_space_id
INNER JOIN sys.all_objects o ON i.[object_id] = o.[object_id]
WHERE i.data_space_id = fg.data_space_id AND o.type = 'U'

SELECT * FROM UserLogHistory
tables and filegroups

SQL Server 2016 SP2 and later

Starting with SQL Server 2016 SP2, SELECT…INTO allows you to specify a filegroup when creating the new table. 

USE TestDB
GO

-- Create copy of the table and all data in different filegroup
SELECT * INTO UserLogHistory1 ON HISTORY 
FROM UserLog

We can also use a WHERE clause to minimize the amount of data we want to move to the new table.

USE TestDB
GO

-- Create copy of the table and filtered data in different filegroup
SELECT * INTO UserLogHistory2 ON HISTORY 
FROM UserLog 
WHERE LoginDate > '2018-11-15 16:31:33.983'

To create an empty table in another filegroup we can set the WHERE clause in such a way to not return any data.

USE TestDB
GO

--Creating the copy of table without data
SELECT * INTO UserLogHistory3 ON HISTORY 
FROM UserLog 
WHERE 1=4

Now we have 3 new tables in the HISTORY filegroup:

USE TestDB
GO

SELECT o.[name] AS TableName, i.[name] AS IndexName, fg.[name] AS FileGroupName
FROM sys.indexes i 
INNER JOIN sys.filegroups fg ON i.data_space_id = fg.data_space_id
INNER JOIN sys.all_objects o ON i.[object_id] = o.[object_id]
WHERE i.data_space_id = fg.data_space_id AND o.type = 'U'
tables and filegroups

The data in the first table is the same as the source table, the second table is filtered according to the criteria in the WHERE clause and the third is an empty table.

USE TestDB
GO

SELECT * FROM UserLogHistory1
SELECT * FROM UserLogHistory2
SELECT * FROM UserLogHistory3
query results

Conclusion

To sum up, these are different ways to copy tables between filegroups. The new feature, SELECT…INTO in SQL Server 2016 SP2 and later facilitates this process by making it more flexible.

Next Steps

You can find additional information about this topic below:



sql server categories

sql server webinars

subscribe to mssqltips

sql server tutorials

sql server white papers

next tip



About the author
MSSQLTips author Sergey Gigoyan Sergey Gigoyan is a database professional with more than 10 years of experience, with a focus on database design, development, performance tuning, optimization, high availability, BI and DW design.

This author pledges the content of this article is based on professional experience and not AI generated.

View all my tips


Comments For This Article




Monday, December 17, 2018 - 7:07:16 AM - Sergey Gigoyan Back To Top (78509)

Jmoden, thank you for your detailed and useful comment. I agree with you. In addition, using WITH (DROP_EXISTING = ON) is optimal when we have non-clustered indexes on the table because unlike the previous option this approach will not change non-clustered indexes. Therefore, I have already updated the article.

Thanks,

Sergey

Friday, December 14, 2018 - 9:39:01 AM - jmoden Back To Top (78495)

If you lose the Clustered Index in the process of moving the table to a new file group but still want to have the same Clustered Index, then the SELECT/INTO and the ALTER TABLE with the DROP CONSTRAINT methods have a serious problem, especially when moving large tables.  In both cases, the table will be created as a HEAP.  You already knew that.

What a lot of people forget is that when you create a Clustered Index on any HEAP over 128 extents in size (that's only 8MB), the HEAP will remain in place until the Clustered Index is built and committed and then the HEAP will be dropped.

Why is that a problem?  The answer is simple... it will double the reserved size of the new File Group. 

For example, you create your new File Group and File and then you want to move a 500GB table to it.  You use one of the two methods in this article to do it.  It creates a HEAP on the new File Group/File of 500GB.  When you build the Clustered Index, the 500GB HEAP remains in place while the new Clustered Index is being built and growing, on its own to 500GB.  Just before the commit, your File Group/File is now 1TB in size.  After the commit, the HEAP is dropped but the reserved file space does not change size.  What you end up with is the 500GB Clustered Index and 500GB of wasted free space and that doesn't include the space used to do the SORT while the Clustered Index was being built.  Add about 20% of 500GB or an extra 100GB of empty space.

At the end, you end up with your 500GB Clustered Index, 500GB of empty space where the HEAP was, and 100GB of empty space that was used during the SORT of the Clustered Index Build.  Where you once were using "only" 500GB, you are now using ~1.1TB.

To avoid all of that, you can simply do a CREATE INDEX with DROP EXISTING and mention the new FILE GROUP you want it to live on.  What you end up with (provided that you don't use SORT_IN_TEMPDB=ON, and you probably DON"T want use that because it may blow out TEMPDB) is the new File Group/File will only have the 500GB Clustered index in it, 500GB of free space in the original File Group where the old copy of the table used to be, and an extra 100GB for the SORT, which will also be contained in the original File Group.

Of course, no matter what happens, you'll end up with at least 500GB of free space in the original File Group but you will totally avoid unnecessarily doubling the size of the new File Group.














get free sql tips
agree to terms