Share via

Moving indexes from one filegroup to another in SQL Server

  • Article

This is a pretty simple task but I would like to provide a generic method for doing this. It will help if the requirement is one (or all) of the folowing. I have only considered non-clustered indexes along with the Inclued columns and filtered indexes. No point in moving clustered indexes as moving them would move the data as well. This is because the leaf level of a clustered index is nothing but a data page. 

1. Move all the Indexes of a Table/View

2. Move specific Indexes of a Table/View

2. Move all the indexes present in a Database

Here is a SP to achieve this and I haven't added validations as the idea is to present how to do this. Hope this is useful.

I am assuming that you have already created a user-defined FileGroup to which you would like to move the indexes. The required parameters for this SP are Database name, scehma name, FileGroup name and minimum of one table/view name for the @ObjectNameList parameter. If you would like to move indexes from more than one object, specify a comma separated string (Ex:- @ObjectNameList = 'Table1,Table2'). Specify the Index name argument only when you want to move a single index from that table/view. 

CREATE PROC [dbo].[MoveIndexToFileGroup] (       
 @DBName sysname,  
 @SchemaName sysname = 'dbo',      
 @ObjectNameList Varchar(Max),       
 @IndexName sysname = null, 
 @FileGroupName varchar(100) 
)       
WITH RECOMPILE  
AS       
 
BEGIN 
    
SET NOCOUNT ON 
 
DECLARE @IndexSQL NVarchar(Max) 
DECLARE @IndexKeySQL NVarchar(Max) 
DECLARE @IncludeColSQL NVarchar(Max) 
DECLARE @FinalSQL NVarchar(Max) 
 
DECLARE @CurLoopCount Int 
DECLARE @MaxLoopCount Int 
DECLARE @StartPos Int 
DECLARE @EndPos Int 
 
DECLARE @ObjectName sysname 
DECLARE @IndName sysname 
DECLARE @IsUnique Varchar(10) 
DECLARE @Type Varchar(25) 
DECLARE @IsPadded Varchar(5) 
DECLARE @IgnoreDupKey Varchar(5)
DECLARE @AllowRowLocks Varchar(5) 
DECLARE @AllowPageLocks Varchar(5)
DECLARE @FillFactor Int 
DECLARE @ExistingFGName Varchar(Max)
DECLARE @FilterDef NVarchar(Max)
 
DECLARE @ErrorMessage NVARCHAR(4000) 
DECLARE @SQL nvarchar(4000) 
DECLARE @RetVal Bit 
 
DECLARE @ObjectList Table(Id Int Identity(1,1),ObjectName sysname) 
 
DECLARE @WholeIndexData Table 
( 
ObjectName sysname 
,IndexName sysname 
,Is_Unique Bit 
,Type_Desc Varchar(25) 
,Is_Padded Bit 
,Ignore_Dup_Key Bit 
,Allow_Row_Locks Bit 
,Allow_Page_Locks Bit 
,Fill_Factor Int 
,Is_Descending_Key Bit 
,ColumnName sysname 
,Is_Included_Column Bit 
,FileGroupName Varchar(Max)
,Has_Filter Bit
,Filter_Definition NVarchar(Max) 
) 
 
DECLARE @DistinctIndexData Table 
( 
Id Int IDENTITY(1,1) 
,ObjectName sysname 
,IndexName sysname 
,Is_Unique Bit 
,Type_Desc Varchar(25) 
,Is_Padded Bit 
,Ignore_Dup_Key Bit 
,Allow_Row_Locks Bit 
,Allow_Page_Locks Bit 
,Fill_Factor Int 
,FileGroupName Varchar(Max)
,Has_Filter Bit
,Filter_Definition NVarchar(Max) 
) 
 
-------------Validate arguments----------------------  
 
IF(@DBName IS NULL) 
BEGIN 
 SELECT @ErrorMessage = 'Database Name must be supplied.'  
 GOTO ABEND 
END 
 
IF(@ObjectNameList IS NULL) 
BEGIN 
 SELECT @ErrorMessage = 'Table or View Name(s) must be supplied.'  
 GOTO ABEND 
END 
 
IF(@FileGroupName IS NULL) 
BEGIN 
 SELECT @ErrorMessage = 'FileGroup Name must be supplied.'  
 GOTO ABEND 
END 
 
--Check for the existence of the Database 
IF NOT EXISTS(SELECT Name FROM sys.databases where Name = @DBName)
BEGIN
 SET @ErrorMessage = 'The specified Database does not exist' 
 GOTO ABEND
END
 
--Check for the existence of the Schema 
IF(upper(@SchemaName) <> 'DBO') 
BEGIN 
 SET @SQL = 'SELECT @RetVal = COUNT(*) FROM ' + QUOTENAME(@DBName) + '.sys.schemas WHERE name = ''' + @SchemaName + '''' 
 BEGIN TRY 
  EXEC sp_executesql @SQL,N'@RetVal Bit OUTPUT',@RetVal OUTPUT 
 END TRY 
 BEGIN CATCH 
  SELECT @ErrorMessage = ERROR_MESSAGE()  
  GOTO ABEND 
 END CATCH 
 IF(@RetVal = 0) 
 BEGIN 
  SELECT @ErrorMessage = 'No Schema with the name ' + @SchemaName + ' exists in the Database ' + @DBName  
  GOTO ABEND 
 END  
END 
 
--Check for the existence of the FileGroup 
SET @SQL = 'SELECT @RetVal=COUNT(*) FROM ' + QUOTENAME(@DBName) + '.sys.filegroups WHERE name = ''' + @FileGroupName + '''' 
BEGIN TRY 
 EXEC sp_executesql @SQL,N'@RetVal Bit OUTPUT',@RetVal OUTPUT 
END TRY 
BEGIN CATCH 
 SELECT @ErrorMessage = ERROR_MESSAGE()  
 GOTO ABEND 
END CATCH 
 
IF(@RetVal = 0) 
BEGIN 
 SELECT @ErrorMessage = 'No FileGroup with the name ' + @FileGroupName + ' exists in the Database ' + @DBName  
 GOTO ABEND 
END 
 
----------Get the objects from the concatenated list---------------------------------------------------- 
 
SET @StartPos = 0 
SET @EndPos = 0 
 
WHILE(@EndPos >= 0) 
BEGIN 
 
 SELECT @EndPos = CHARINDEX(',',@ObjectNameList,@StartPos) 
 IF(@EndPos = 0) --Means, separator is not found 
 BEGIN 
  INSERT INTO @ObjectList 
  SELECT SUBSTRING(@ObjectNameList,@StartPos,(LEN(@ObjectNameList) - @StartPos)+1)  
    
  BREAK 
 END 
  
 INSERT INTO @ObjectList 
 SELECT SUBSTRING(@ObjectNameList,@StartPos,(@EndPos - @StartPos)) 
   
 SET @StartPos = @EndPos + 1 
  
END 
 
-------------Check for the validity of all the Objects---------------------- 
 
SET @StartPos = 1 
SELECT @EndPos = COUNT(*) FROM @ObjectList 
 
WHILE(@StartPos <= @EndPos) 
BEGIN 
 
 SELECT @ObjectName = ObjectName FROM @ObjectList WHERE Id = @StartPos 
 
 --Check for existence of the object 
 SET @SQL = 'SELECT @RetVal=COUNT(*) FROM ' + QUOTENAME(@DBName) + '.sys.Objects WHERE type IN (''U'',''V'') AND name = ''' + @ObjectName + '''' 
 BEGIN TRY 
  EXEC sp_executesql @SQL,N'@RetVal Int OUTPUT',@RetVal OUTPUT 
 END TRY 
 BEGIN CATCH 
  SELECT @ErrorMessage = ERROR_MESSAGE()  
  GOTO ABEND 
 END CATCH 
 
 IF(@RetVal = 0) 
 BEGIN 
  SELECT @ErrorMessage = 'No Table or View with the name ' + @ObjectName + ' exists in the Database ' + @DBName  
  GOTO ABEND 
 END  
 
 --Check for existence of Index 
 IF(@IndexName IS NOT NULL) 
 BEGIN 
  SET @SQL = 'SELECT @RetVal=COUNT(*) FROM ' + QUOTENAME(@DBName) + '.sys.Indexes si INNER JOIN ' + QUOTENAME(@DBName) + '.sys.Objects so ' 
  SET @SQL = @SQL + ' ON si.Object_Id = so.Object_Id WHERE so.Schema_id = ' + CAST(Schema_Id(@Schemaname) as varchar(25))  
  SET @SQL = @SQL + ' AND so.name = ''' + @ObjectName + ''' AND si.name = ''' + @IndexName + ''''  
 
  BEGIN TRY 
   EXEC sp_executesql @SQL,N'@RetVal Int OUTPUT',@RetVal OUTPUT 
  END TRY 
  BEGIN CATCH 
   SELECT @ErrorMessage = ERROR_MESSAGE()  
   GOTO ABEND 
  END CATCH 
 
  IF(@RetVal = 0) 
  BEGIN 
   SELECT @ErrorMessage = 'No Index with the name ' + @IndexName + ' exists on the Object ' + @ObjectName  
   GOTO ABEND 
  END 
 END 
  
 SET @StartPos = @StartPos + 1 
END 
 
-------------Loop till all the Objects are processed---------------------- 
 
SET @StartPos = 1 
SELECT @EndPos = COUNT(*) FROM @ObjectList 
 
WHILE(@StartPos <= @EndPos) 
BEGIN 
 
 SELECT @ObjectName = ObjectName FROM @ObjectList WHERE Id = @StartPos 
 
 -------------Build the SQL to get the index data based on the inputs provided----------------------  
 
 SET @IndexSQL =  
 'SELECT so.Name as ObjectName, si.Name as IndexName,si.Is_Unique,si.Type_Desc' 
 + ',si.Is_Padded,si.Ignore_Dup_Key,si.Allow_Row_Locks,si.Allow_Page_Locks,si.Fill_Factor,sic.Is_Descending_Key' 
 + ',sc.Name as ColumnName,sic.Is_Included_Column,sf.Name as FileGroupName,si.Has_Filter,si.Filter_Definition FROM '
 + QUOTENAME(@DBName) + '.sys.Objects so INNER JOIN ' + QUOTENAME(@DBName) + '.sys.Indexes si ON so.Object_Id = si.Object_id INNER JOIN ' 
 + QUOTENAME(@DBName) + '.sys.FileGroups sf ON sf.Data_Space_Id = si.Data_Space_Id INNER JOIN '  
 + QUOTENAME(@DBName) + '.sys.Index_columns sic ON si.Object_Id = sic.Object_Id AND si.Index_id = sic.Index_id INNER JOIN ' 
 + QUOTENAME(@DBName) + '.sys.Columns sc ON sic.Column_Id = sc.Column_Id and sc.Object_Id = sic.Object_Id ' 
 + ' WHERE so.Name = ''' + @ObjectName  + '''' 
 + ' AND so.Schema_id = ' + CAST(Schema_Id(@Schemaname) as varchar(25)) + ' AND si.Type_Desc = ''NONCLUSTERED'' ' 
 
 IF(@IndexName IS NOT NULL) 
 BEGIN 
  SET @IndexSQL = @IndexSQL + ' AND si.Name = ''' + @IndexName + '''' 
 END 
 
 SET @IndexSQL = @IndexSQL + ' ORDER BY ObjectName, IndexName' 
 
 --PRINT @IndexSQL 
 
 -------------Insert the Index Data in to a variable----------------------  
 
 BEGIN TRY 
  INSERT INTO @WholeIndexData 
  EXEC sp_executesql @IndexSQL 
 END TRY 
 BEGIN CATCH 
  SELECT @ErrorMessage = ERROR_MESSAGE()  
  GOTO ABEND 
 END CATCH 
 
 --Check if any indexes are there on the object. Otherwise exit 
 IF (SELECT COUNT(*) FROM @WholeIndexData) = 0 
 BEGIN 
  SELECT 'Object does not have any nonclustered indexes to move'  
  GOTO FINAL  
 END 
   
 -------------Get the distinct index rows in to a variable----------------------  
 
 INSERT INTO @DistinctIndexData 
 SELECT DISTINCT  
 ObjectName,IndexName,Is_Unique,Type_Desc,Is_Padded,Ignore_Dup_Key,Allow_Row_Locks,Allow_Page_Locks,Fill_Factor,FileGroupName,Has_Filter,Filter_Definition 
 FROM @WholeIndexData
 WHERE ObjectName = @ObjectName 
 
 SELECT @CurLoopCount = Min(Id), @MaxLoopCount = Max(Id) FROM @DistinctIndexData WHERE ObjectName = @ObjectName
 
 --SELECT @CurLoopCount, @MaxLoopCount 
 
 -------------Loop till all the indexes are processed----------------------  
 
 WHILE(@CurLoopCount <= @MaxLoopCount) 
 BEGIN 
 
  SET @IndexKeySQL = '' 
  SET @IncludeColSQL = '' 
 
  -------------Get the current index row to be processed---------------------- 
  SELECT  
   @IndName   = IndexName 
   ,@Type   = Type_Desc
   ,@ExistingFGName = FileGroupName
   ,@IsUnique  = CASE WHEN Is_Unique = 1 THEN 'UNIQUE ' ELSE '' END 
   ,@IsPadded  = CASE WHEN Is_Padded = 0 THEN 'OFF,' ELSE 'ON,'  END 
   ,@IgnoreDupKey = CASE WHEN Ignore_Dup_Key = 0 THEN 'OFF,' ELSE 'ON,' END 
   ,@AllowRowLocks = CASE WHEN Allow_Row_Locks = 0 THEN 'OFF,' ELSE 'ON,' END
   ,@AllowPageLocks = CASE WHEN Allow_Page_Locks = 0 THEN 'OFF,' ELSE 'ON,' END 
   ,@FillFactor  = CASE WHEN Fill_Factor = 0 THEN 100 ELSE Fill_Factor END 
   ,@FilterDef  = CASE WHEN Has_Filter = 1 THEN (' WHERE ' + Filter_Definition) ELSE '' END 
  FROM @DistinctIndexData  
  WHERE Id = @CurLoopCount 
   
  -------------Check if the index is already not part of that FileGroup---------------------- 
 
  IF(@ExistingFGName = @FileGroupName) 
  BEGIN 
   PRINT 'Index ' +  @IndName + ' is NOT moved as it is already part of the FileGroup ' + @FileGroupName + '.' 
   SET @CurLoopCount = @CurLoopCount + 1 
   CONTINUE 
  END 
 
  ------- Construct the Index key string along with the direction-------------------- 
  SELECT  
   @IndexKeySQL =  
   CASE 
   WHEN @IndexKeySQL = '' THEN (@IndexKeySQL + QUOTENAME(ColumnName) + CASE WHEN Is_Descending_Key = 0 THEN ' ASC' ELSE ' DESC' END)  
   ELSE (@IndexKeySQL + ',' + QUOTENAME(ColumnName) + CASE WHEN Is_Descending_Key = 0 THEN ' ASC' ELSE ' DESC' END)  
   END 
  FROM @WholeIndexData 
  WHERE ObjectName = @ObjectName  
  AND IndexName = @IndName  
  AND Is_Included_Column = 0 
   
  --PRINT @IndexKeySQL  
   
  ------ Construct the Included Column string -------------------------------------- 
  SELECT  
   @IncludeColSQL =  
   CASE 
   WHEN @IncludeColSQL = '' THEN (@IncludeColSQL + QUOTENAME(ColumnName))  
   ELSE (@IncludeColSQL + ',' + QUOTENAME(ColumnName))  
   END  
  FROM @WholeIndexData 
  WHERE ObjectName = @ObjectName  
  AND IndexName = @IndName  
  AND Is_Included_Column = 1  
   
  --PRINT @IncludeColSQL 
 
  -------------Construct the final Create Index statement---------------------- 
  SELECT
  @FinalSQL = 'CREATE ' + @IsUnique + @Type + ' INDEX ' + QUOTENAME(@IndName)
  + ' ON ' + QUOTENAME(@DBName) + '.' + QUOTENAME(@SchemaName) + '.' + QUOTENAME(@ObjectName) 
  + '(' + @IndexKeySQL + ') '  
  + CASE WHEN LEN(@IncludeColSQL) <> 0 THEN  'INCLUDE(' + @IncludeColSQL + ') ' ELSE '' END
  + @FilterDef 
  + ' WITH ('  
  + 'PAD_INDEX = ' + @IsPadded  
  + 'IGNORE_DUP_KEY = ' + @IgnoreDupKey 
  + 'ALLOW_ROW_LOCKS  = ' + @AllowRowLocks  
  + 'ALLOW_PAGE_LOCKS  = ' + @AllowPageLocks  
  + 'SORT_IN_TEMPDB = OFF,'  
  + 'DROP_EXISTING = ON,'  
  + 'ONLINE = OFF,' 
  + 'FILLFACTOR = ' + CAST(@FillFactor AS Varchar(3)) 
  + ') ON ' + QUOTENAME(@FileGroupName) 
 
  --PRINT @FinalSQL 
 
  -------------Execute the Create Index statement to move to the specified filegroup---------------------- 
  BEGIN TRY 
   EXEC sp_executesql @FinalSQL 
  END TRY 
  BEGIN CATCH 
   SELECT @ErrorMessage = ERROR_MESSAGE()  
   GOTO ABEND 
  END CATCH  
  PRINT 'Index ' +  @IndName + ' on Object ' + @ObjectName + ' is moved successfully.'  
   
  SET @CurLoopCount = @CurLoopCount + 1 
 
 END 
  
 SET @StartPos = @StartPos + 1 
END 
 SELECT 'The procedure completed successfully.' 
 RETURN 
 
ABEND: 
 RAISERROR 500001 @ErrorMessage 
FINAL: 
 RETURN   
END

Comments

  • Anonymous
    September 26, 2008
    PingBack from http://informationsfunnywallpaper.cn/?p=7181

  • Anonymous
    April 08, 2009
    Do you have such a script for SQL 2000. I will appreciate if you send it to me. Here is my email: sarfaraz_malik@hotmail.com Regards, Sarfaraz

  • Anonymous
    April 14, 2009
    I updated this script to not include indexes that are already in the provided file group SET @IndexSQL = 'SELECT so.Object_id as ObjectId, so.Name as ObjectName, si.Name as IndexName,si.Is_Unique,si.Type_Desc,si.Is_Padded,si.Ignore_Dup_Key,'

  • 'si.Allow_Row_Locks,si.Allow_Page_Locks,si.Fill_Factor,sic.Is_Descending_Key,sc.Name as ColumnName,Is_Included_Column'
  • ' FROM ' + @DBName + '.sys.Objects so INNER JOIN ' + @DBName + '.sys.Indexes si ON so.Object_Id = si.Object_id '
  • ' INNER JOIN ' + @DBName + '.sys.Index_columns sic ON si.Object_Id = sic.Object_Id AND si.Index_id = sic.Index_id'
  • ' INNER JOIN ' + @DBName + '.sys.Columns sc ON sic.Column_Id = sc.Column_Id and sc.Object_Id = sic.Object_Id  '
  • ' inner join ' + @DBName + '.sys.data_spaces ds on si.data_space_id = ds.data_space_id '
  • '  WHERE so.Schema_id = ' + CAST(Schema_Id(@Schemaname) as varchar(25))
  • '  AND si.Type_Desc IN (''' + @IndexType + ''') '
  • ' and ds.name not in (''' + @FileGroupName + ''')'
  • Anonymous
    June 19, 2009
    The comment has been removed
  • Anonymous
    June 19, 2009
    Hi David, Thanks for your feedback. The index creation shouldn't fail even if the column order is different from the original one. I think something else is causing the issue. So, try running the generated index creation script (i.e output of the procedure -> finalsql) manually and see what happens. You can also try by reversing the columns. In my testing of the above script, I haven't come across the situation of column order changing. I have noticed that the index key columns are stored in sys.index_columns by default. So, though I haven't put any order by clause in the dynamic sql (I am referring to @IndexSql varibale), the columns are always concatenated as per the initial order defined in the original index. You can try one thing. Add the below order by clause at the end of the dynamic sql that is assigned to @IndexSql and print the finalsql.
  • ' ORDER BY sic.Key_Ordinal ASC'   The index key columns should be in order this time.

  • Anonymous
    September 15, 2009
    Ramoji, Great procedure. I had been thinking about developing a way of identifying and moving indexes that fit specified criteria from the Default Filegroup to a separate filegroup for a couple of months, I starting designing my process and you example is just what I needed. I used functions to concatenate that various columns to be indexed and for the included columns. The procedure that I designed takes the database parameter and then searches the DMVs to find the NonClustered indexes that I want to move. A cursor then builds the index statement on the fly, and executes. Thanks Again!

  • Anonymous
    December 08, 2009
    The comment has been removed

  • Anonymous
    January 12, 2010
    Neat. Do you have anything for 2005?

  • Anonymous
    December 04, 2010
    For 2005, this Works on My Machine: Replace this: si.Has_Filter,si.Filter_Definition with: 0 as Has_Filter,N'''' as Filter_Definition

  • Anonymous
    May 22, 2014
    Great one! but doesn't works on SQL 2000 :) Can you send me a copy if you have sumitchhabra0709@hotmail.com

  • Anonymous
    October 20, 2015
    Looks like somebody updated it for 2012, according to a StackOverflow article. (not mine, just figured it was worth mentioning here) gist.github.com/.../6877012