Mount points seem to be the subject of much confusion on various SQL Server forums. In light of this, the following post will seek to detail exactly how mount points are supposed to work and provide a demonstration on how to create mount points within Windows.

You should have already checked whether your planned SQL Server instance will support Mount Points as not all versions provide this. With that in mind, let's briefly look at this now

Validating SQL Server Compatibility

Check whether or not your chosen version of SQL Server will support mount points. Not all editions of SQL Server do support the use of mount points, therefore, before configuring mount points and attempting to install SQL Server you must refer to the following Microsoft KB article for further details: 

http://support.microsoft.com/kb/819546

In brief, Versions of SQL Server 2005 and beyond fully support mount points for stand alone and clustered environments.

What are Mount Points?

Anybody who has ever used a Unix system would have come across the use of mount points before, they are common within this Network Operating System. Anything from a disk partition to a CD drive or floppy drive all require mounting before use, these file systems are usually mounted under a root partition folder.

Within the Windows operating system we have the ability to perform similar actions with multiple Windows disks. We have 26 drive letters initially available for assignment, subtract from this the reserved set of drives (A, B, C) and you are commonly left with 23. This doesn't leave much scope, especially if you plan to use many data locations per instance. You may have multiple SQL Server instances and require multiple disk sets to offload I\O, For each instance you may originally plan to use separate assignments for

• SQL Server data
• SQL Server logs
• SQL Server backups
• SQL Server tempdb

That's 4 drive letters for a single instance. Mount points enable you to use a single drive letter for all of the chosen locations but still implement separate physical drives for each of the individual data locations. How is this possible?


Mount points are simply file system locations (in the form of folders) under which a physical disk partition may be mounted\referenced. You start with a reference root known as the root disk. The root disk can, in fact should be, as small as possible. This will avoid multiple users creating data on the root disk and make administration easier. You must also bear in mind that all I\O destined for the Mount Point occurs on the mounted physical disk and not the root disk

We'll go through actually creating mount points later in the article, for now let's look at a little theory. Our basic example is as follows

Root disk 1 M:      (Windows disk 2)

Multiple mount points are employed in the following configuration for a single instance, remember each of these mount points is actually a separate physical drive from the Windows disk management snapin.

M:\SQLData         (Windows disk 3)

M:\SQLLog           (Windows disk 4)

M:\SQLTmp          (Windows disk 5)

M:\SQLBak           (Windows disk 6)

The instance may be installed using the paths below. To assign permissions we apply at the MSSQL folder level and they then propagate down to the child folders\files. Applying permissions at the root level M: or the mount point folders would not affect the mount point child file structures.

M:\SQLData\MSSQL\Data

M:\SQLLog\MSSQL\Log

M:\SQLTmp\MSSQL\TempDB

M:\SQLBak\MSSQL\Bak

Going further in another example, you could employ multiple root disks and multiple instances of SQL Server in the following manner

Root disk 1 D:           (Windows disk 2)

Root disk 2 E:           (Windows disk 3)

Root disk 3 M:           (Windows disk 4)

You would have multiple mount points, each instance would have 4 separate physical disks mounted under

Instance1 D:\INST1Data, D:\INST1Log, D:\INST1Tempdb, D:\INST1Bak     (Windows disks 6 - 9)

Instance 2 E:\INST2Data, E:\INST2Log, E:\INST2Tempdb, E:\INST2Bak    (Windows disks 10 - 13)

Instance 3 M:\INST3Data, M:\INST3Log, M:\INST3Tempdb, M:\INST3Bak    (Windows disks 14 - 17)

The SQL server instances would be created along the following paths

D:\INST1Data\MSSQL\Data, D:\INST1Log\MSSQL\Log, D:\INST1Tempdb\MSSQL\TempDB, D:\INST1Bak\MSSQL\Backups

E:\INST2Data\MSSQL\Data, E:\INST2Log\MSSQL\Log, E:\INST2Tempdb\MSSQL\TempDB, E:\INST2Bak\MSSQL\Backups

M:\INST3Data\MSSQL\Data, M:\INST3Log\MSSQL\Log, M:\INST3Tempdb\MSSQL\TempDB, M:\INST3Bak\MSSQL\Backups

It's easy to see how the management becomes more complex with multiple root disks and mounted volumes, it can't be stressed enough that documenting your setup as you go is more valuable than anything else, you do have an HLD and an LLD provided by the architect? Concise documentation gives you and anyone who follows you, a reference point for your system setup and can make troubleshooting easier too.


With Mount Points it's important to remember the following pointers

• Do not install software to the root of the mount point. Rather, create and install into a folder under the mount point to allow the correct propagation of permissions.
• Ideally, the root drive should not have any data created upon it, it is merely the root tree for the new volumes.
• Permissions applied to the root drive file system and the mount point root folders do not propagate to the mount point file systems, they are controlled separately.
• Mount points are not supported for all SQL Server 2000 configurations (in case you had a burning desire to use them). This is detailed in the relevant section above.

Windows requires online disks with formatted partitions before you are able to utilise mount points, let's look at those next as they form the basis for mounting any partition\volume.

Setting up the New Partitions

When initialising and before formatting your disks within Windows disk management, you should first decide on the disk type and partition type to be used.

In brief the 2 disk types possible are Basic and Dynamic. Basic disks are the default disk type used by the Windows Operating System. Dynamic disks are rarely used within Windows and are really only useful for creating spanned volumes or software RAID systems. In reality, the Windows disk you are formatting likely already has an underlying RAID system (such as a local controller) or it is a disk presented from a SAN or other network storage device. For this reason you should generally you should leave your disks as Basic type, especially for cluster support.

As well as the disk type, there are also 2 different partition types that may be used. The 2 types available are MBR (master boot record) or GPT (GUID partition type).

• MBR partition types only allow a maximum of 4 partitions per disk (3 primary and 1 extended).
• GPT partition types support very large disk types, a maximum of 128 partitions and are generally much more efficient and resistant to corruption.

However, I would suggest if your disk is under 500GB and has only 1 partition that you leave this as MBR.

The following Microsoft KB article provides further details on Basic and Dynamic disks and also MBR and GPT partition types. You should digest this before creating disk partitions:

 http://msdn.microsoft.com/en-us/library/aa363785(VS.85).aspx

To re cap, the items you will require to create your mount points are;

• A root disk (or disks) with a formatted partition and a drive letter assigned within Windows
• A defined set of mount point folders to be created upon the root disk.
• Disk devices available for use within Windows disk management (be they local or SAN\NAS attached).

Creating a Mount Point Demonstration.

We will now work through a short demonstration showing how to create mount points within Windows. For the demonstration we need a small root disk, the smaller the better (mines 500MB but it could\should be as small as a few MB).

Note: Remember, the root disk is merely a file system place holder for the mounted drives, it will only contain empty folders which are then used to mount the NTFS volumes. The temptation for users to dump files on the root drive is high and something I've seen a lot over the years.


My partitioned and formatted root disk is shown below

On the root drive (M:) I have created the following folders, these folders currently display the typical Windows folder icon.

Open Windows Disk Management as we first need to create the new volumes. Create the first volume by right clicking an unallocated disk and selecting "New simple volume" as shown below.

Click “Next” in the new volume wizard to continue.

Specify the volume size and click “Next”.

It's at this point we specify where we want to mount this new volume, we could have chosen a drive letter but we want to use a mount point. Select the option to “Mount in the following empty NTFS folder:”, browse to the path and click “OK”.

With the mount point specified, click “Next” to continue.

Specify the format options and click “Next” to continue.

Click “Finish” to create the new volume. Repeat these steps for each volume to be created\mounted.

Once you have created the 3 volumes and mounted them to the specific folders, you will see the following in disk management. Your mount points are now ready to use for a stand alone SQL Server installation.

Mount Points for Clustered Instances

For clustered scenarios there are further checks to be made when utilising mount points for your clustered SQL Server instance drives. Those steps are detailed in the following Microsoft KB: 

http://support.microsoft.com/kb/280297

In brief, you now need to add this storage to the cluster via Failover Cluster Manager. Right click “Storage” and select “Add a disk” as shown below.

With the required disks selected, click “OK”.

Cluster disks 6, 8, 9 and 11 added. Note cluster disk 6 is the Root drive (M:).

Now add the storage to our newly created service\application group. My example uses a pre created application titled “Denali Mount Points”.

You must now set the dependencies between the root drive and the new volumes. Right click the first mounted volume and select “Properties” as shown below.

Note: the root drive must come online before the mounted volumes to ensure the full mount path and drive letter are available.

Set the root disk as the resource dependency, in this case cluster disk 6. Click “Apply” and “OK”. Do this for each of the 3 mounted volumes.

During the SQL server instance installation, select the existing Denali service\application group shown with a green tick.

At the cluster disk selection you must select the root drive for installation. The mounted drives are not available, attempting to select one of the mounted volumes causes setup to direct you to select the root tree drive (in my case cluster Disk 6 M: ).

On the database engine "Data Directories" tab, set your paths for the requested disk locations.

Post installation, these are the new SQL server folders under the mount points when viewed through explorer. Notice that the mount points now have a different icon and not the standard Windows folder icon.

This should give you more of an understanding of mount points and how they are set up. They operate just the same as any other folder with the exception that permissions grants are handled by each individual mount point and not by the root directory. Granting permissions at the root drive and pushing child inheritance will not affect the mount points. Still, they are essentially NTFS folders and, as the screenshot above shows, look seem less to the Windows filesystem. As you can see they are also an excellent way to bypass the Windows 26 drive letter limitation.

As always, test this in an offline environment, virtual machines make a great test lab and are extremely easy to set up using my three part guide on this site. Have fun and of course, post back if you have any troubles.

Hello and welcome to my latest post detailing the Always Encrypted feature within SQL Server. Much has happened within SQL Server these last few years and Always Encrypted is another big feature in the SQL Server encryption arena.

But what is it, how can it help my organisation?

What is "Always Encrypted"?

For a full and detailed explanation, check the Microsoft documentation at the following link;

Always Encrypted (Database Engine)

Always Encrypted was initially made available in Enterprise Edition of SQL Server 2016 RTM, it was later introduced into Standard, Web and Express Editions during Service Pack 1 for SQL Server 2016.
Always encrypted uses a series of steps utilising symmetric and asymmetric keys, the steps are dependent upon who will create and manage the keys and the tools used to create them. Different organisations have different levels of support and role separation. We'll look at these later but for now let's look at exactly what it does and why?

What does Always Encrypted Do and Why?

Traditionally, if you used Transparent Data Encryption, you would have a Certificate and Database Master Key (DMK) in the master database and then a Database Encryption Key (DEK) in the user database. As the Certificate and DMK are SQL Server instance level objects, it's pretty much available to anyone who has the key. Obviously, this cannot be used to truly encrypt and prevent SQL Server users (including DBAs) from reading the encrypted data.

This is where the Column Master Key (CMK) and Column Encryption Key (CEK) come into play. Always Encrypted uses 2 different types of key and an external encryptor to achieve the secure data encryption. The external key provisioning options are; 

• Hardware Security Module using CNG/KSP (Cryptographic Next Generation using Key Storage Provider)
• Hardware Security Module using CAPI/CSP (Cryptographic API using Cryptography Service Provider)
• Azure Key Vault
• Windows Certificate Store (Computer)
• Windows Certificate Store (User)

The first 2 options generate a column master key from an external secure source, the 3rd option uses a certificate in an Azure Key Vault.
The last 2 options use the Windows Certificate key store. There are vast differences between the above options and it all depends on your organisations preferred method of key storage.
Of the first 2 options above, CNG\KSP should be preferred over the legacy CAPI/CSP. Both options require the installation of a client side driver to provide interaction between the key store and the Always Encrypted system.
A typical example of a Windows Certificate key store cert would look similar to the screenshot below;

Column Master Key (CMK)

The CMK is an asymmetric key which is stored externally to the SQL Server instance, this is usually in the form of a certificate. If required, DBAs have no access to the CMK at all.
A reference to the CMK is held inside the user database, but it only holds the path (including the thumbprint) to the external key not the key itself, it's created and managed outside of the SQL server system. 

Column Encryption Key (CEK)

The CEK resides inside the user database and is protected and encrypted by the CMK. The CEK key holds only the encrypted CEK value, no plaintext details are stored.

With the Certificate residing outside of the SQL Server system, The CMK certificate is distributed to only those users who are required to decrypt the data, given security and data best practices this list should be limited.

There are a set of DMVs we can use to track the keys and an example is shown below from a test system;

As we can see from the output above, the CMK metadata holds very limited detail, the key path includes the thumbprint which is required to create the CEK. This value is encrypted and then stored in the "encrypted_value" column in the CEK key values DMV. 

Table Schemas

The base column definitions stay the same in that their data types do not change, there's no creating the columns as varbinary as was necessary with traditional column level encryption. However, once encryption has been applied, the table definition does change slightly as shown below;

When scripting out the schema we can see that the CardNumber, ExpMonth and ExpYear columns now have an extra clause applied to their definition. The type of encryption selected for each chosen column during configuration, now directly applies on a per column basis. The 2 types of encryption which may be applied are defined as one of the following; 

• Randomised
• Deterministic 

Randomised

The Randomised encryption type produces a different value each time even when the data encrypted is the same. This is more secure but also prevents the usage of these columns in indexes, grouping and join criteria.

Deterministic

The Deterministic encryption type produces common encryption values. Whilst this aids the use of the columns in join criteria and indexing, etc it could be possible when analysing the encrypted strings to work out patterns in the encrypted values, so exposing the protected data.

Always Encrypted With Secure Enclaves

New in SQL Server 2019 is the feature Always Encrypted with Secure Enclaves. This feature was added to overcome some of the limitations with the way encrypted data is used. Traditionally, the data encryption operations occur on the client side via the client driver. The client modifies the data unencrypted and then sends the encrypted value(s) back to the server. The data and the encryption keys are never exposed over the link. This affects the way the data is used though and one reason why the choice for Randomised or Deterministic encryption should be carefully made, as in the case of column joins and grouping, etc.

With the feature enhancement, some computations of the unencrypted clear text data are available on the server within secure enclaves. They are protected regions of memory assigned to the SQL Server process so that further computation operations are possible over the standard Always Encrypted functionality. Of course, this would involve extra design and configuration tasks and also development routes would need to be changed too, but the feature is there for those who wish to consume it. This link details further;

https://learn.microsoft.com/en-us/sql/relational-databases/security/encryption/always-encrypted-enclaves?view=sql-server-ver16

Always Encrypted Deployment Routes

Whilst it is possible for a DBA to manage the whole process, this may not fit all organisations, especially where role separation is enforced. So the way in which the encryption is applied depends on whether role separation is employed and whether you fully trust the DBA team or are they supplied by a 3rd party and not authorised to view the sensitive data.

Of course you may have Core employee DBAs who are trusted but the Security team handle all encryption. The answer to these questions would straight away drive a particular route, let's look at some scenarios;

Scenario A

The DBA is a direct employee of the organisation and trusted fully with the database systems. They are either authorised to view the data or do not exercise any elevated privilege to view data they are not required to have access to. They would have access to an Azure Key Vault or Windows Certificate Stores or HSM modules.

Scenario B

The DBA is a resource supplied by a 3rd party support vendor to the organisation. The organisation also has a separate security admin team that manage security keys. Neither team are authorised to have access to the encrypted data but are required to implement encryption and provide BAU management of the database system.

The scenarios above map to the following;

Scenario A (Singularly Managed)

The management via SQL Server Management Studio will be suitable for those who entrust their DBAs with the sensitive data held in the database or where the DBA has a wide range of permissions for their role outside of SQL server. For instance, if selecting a Computer Certificate Store the DBA would require local admin permissions on their machine, otherwise when attempting to create a column master key and selecting the local computer certificate store, an error will be raised.

Allowing the DBAs to create and manage the keys and view the data is fairly straight forward and can be achieved via SQL Server Management Studio, using the [Sales].[CreditCard].[CardNumber] column in the AdventureWorks database as an example, a typical deployment would be as follows;

First, right click the table in SSMS and select "Encrypt Columns" as shown below;

Note: The CMKs and CEKs can be created and managed from SQL Server Management Studio and requires v17.0 and above.

Click "Next" through the Introduction screen and you will see the "Column Selection" dialog, select the columns and the encryption type. For this I have selected the CardNumber column and Randomised encryption. A new CEK has been automatically selected since i do not have any existing keys in the database, I also do not have the option of changing the CEK name. Select your options and click "Next";

 Landing at the CMK configuration dialog we see the following, since no keys exist the system will auto generate with default names. Click "Next" to continue;

 At this point you may script the configuration, leave as "Proceed to Finish Now" and click "Next";

Clicking "Finish" will apply the Always Encrypted configuration.

At this point you are left with a configuration that may not meet naming standards, this is one of the downsides of using the Always Encrypted wizard to deploy your configuration.
Ideally, you would deploy your CMK certificate and create your CEK before you invoke the "Encrypt Columns" wizard. The CMK would be created on a secure asset within the organisation. In this scenario I have used a cert store on the same computer as the SQL server instance, this is for demonstration purposes only and would not meet best practice, this would likely breach any security restrictions imposed by your organisation too.

For the purposes of this demonstration, you should have a basic idea of what is required for a single role deployment.

Scenario B (Role Separated)

For scenarios where role separation is employed, the process will need to be broken down and actioned by the appropriate agent. The process employs PowerShell cmdlets in the first steps, which are used by the Security or Infrastructure admins. The DBAs are then passed the relevant thumbprint values and may complete the process via PowerShell or T-SQL.

To use a scenario, this could be where an organisation has a separate security administration team and a team of Database Administrators provided by a 3rd party support vendor.

Neither party are permitted to view encrypted data in the salary column of the employee table. The security operative is to manage the key but no permission to select the column, the DBA is to manage the SQL Server configuration but have no ability to read the table. Without access to the security key itself, selecting the column returns the encrypted data.

The configuration steps for "role separated" organisations are different and are broken down into subsets for each role to perform. The key values provided by the security team to the DBA are the CMK type and path\location and the CEK encrypted value. In this example we will use a user based certificate in the local Windows Certificate store. The Always Encrypted deployment via a role separated route is as follows;

Security Admin

• Generate a user based certificate on a secure asset within the organisation. Typically the PowerShell cmdlet "New-SelfSignedCertificate" could be used.
• Create a CMK setting using the PowerShell cmdlet "New-SqlCertificateStoreColumnMasterKeySettings". This cmdlet connects to the SQL PowerShell provider but performs no actual database connectivity.
• Create a CEK setting using the PowerShell cmdlet "New-SqlColumnEncryptionKeyEncryptedValue". Again, this cmdlet connects to the SQL PowerShell provider but performs no actual database connectivity.
• Provide the DBA with the CMK path\location and the CEK encrypted value, this can be done via secure sharing of a file generated by the previous steps.

 DBA

• Connect to the SQL Server instance and database via PowerShell.
• Create a column master key setting from the settings previously provided by the Security Admin. This uses the PowerShell cmdlet "New-SqlColumnMasterKeySettings". No actual database connectivity is performed at this point.
• Create a CMK in the user database by using the PowerShell cmdlet "New-SqlColumnMasterKey". This performs database connectivity and creates the new CMK. This key would be available during the column encryption wizard steps as previously seen in the single role deployment.
• Create a CEK in the user database by using the PowerShell cmdlet "New-SqlColumnEncryptionKey". This also performs database connectivity and creates the new CEK based on the CMK details. Again, this key would be available during the column encryption wizard steps as previously seen in the single role deployment.
• Complete the column encryption using the newly created keys

After completing the steps above, when first querying the data the encrypted version is returned.

After ensuring access to the certificate in my current user Windows Certificate store and reconfiguring the SSMS connection to allow Always Encrypted as shown below;

Re running the query now shows the unencrypted data

To remove the encryption simply re run the "Encrypt Columns" wizard and set the columns to "PlainText", then complete and Finish the wizard.

It should be understood from the outset that the configuration and implementation of Always Encrypted requires the skills of an experienced Security Admin and an experienced DBA. Where necessary, Proof Of Concept environments can be used to drive the High Level Designs and Low Level Designs. Configurations outside the basics of the Windows Certificate Store will require deep knowledge of all associated subsystems and PowerShell. It's important that the data security be paramount at all times and that all options are fully evaluated to ensure the correct configurations are applied, observing any rules the organisation follows or are regulated to follow.

We've all been used to using the stored procedure "sp_change_users_login" for some time now. This SP is no longer the route to take when fixing orphaned users. Now we have the following TSQL command:

ALTER USER xxxx WITH LOGIN = xxxx

Admittedly this does not resolve the issue of a missing login, in this situation you simply need a check to ensure that the login exists before attempting to alter the users SID, the following can be used:

IF SUSER_SID('someuser') IS NULL

BEGIN

CREATE LOGIN [xxxx] WITH PASSWORD = 'xxxx' --sql

CREATE LOGIN [xxx\xxxx] FROM WINDOWS --Windows

END

ALTER USER xxxx WITH LOGIN = xxxx

In this article I will be discussing the moving of database files within a SQL Server instance. We'll also work through a typical move scenario, looking at the scripts we should use and the meta data available to help us.

Let's first begin with what we do know.

Under SQL Server 2000, altering database paths was all but limited to the TempDB database. You had to detach the user database and then re-attach it, or you could use backup, then restore. Hmm, messy to say the least. 

Under normal operation, once SQL Server has a handle on a database you will not be able to manipulate the files at the OS level. Any attempt to copy the files, etc., will result in the dialog box below. 

To address the first point, thankfully in SQL Server 2005 onwards, this is no longer necessary, and in fact SP_ATTACH_DB has been deprecated and will be removed in a future version of SQL Server. You should now use:

CREATE DATABASE ... FOR ATTACH

To address the second point, in order to release the handle the database engine has on the user database files we merely need to Offline the database. We do not need to stop the SQL Server services.

Let's just re-cap that; we do not need to stop the SQL Server services.

You may issue the following command to Offline the database;

ALTER DATABASE [yourDB] SET OFFLINE

If you have active connections and wish to roll them back and take the database offline you may do so using;

ALTER DATABASE [yourDB] SET OFFLINE WITH ROLLBACK IMMEDIATE

Once the database is Offline you may move and\or rename your database files. Just remember that if you delete or rename the files, the database will not come back online again. When attempting to Online the database you will usually receive an error along the lines of: 

You must first amend the system catalogs to provide the new paths\filenames, this is done using the ALTER DATABASE command passing in the MODIFY FILE parameters as shown in the following query construct:

ALTER DATABASE [yourDB]
  MODIFY FILE ( name=logicalfilename, 
                filename=N'c:\folder1\folder2\adbfile.mdf'
              )

Important Note: When using the T-SQL command above, SQL Server will accept whatever you type and issue in the ALTER DATABASE statement so be careful and check your typing!

For example this would be valid

ALTER DATABASE [yourDB] MODIFY FILE(name=logicalfilename, 

filename=N'c:\MSSQL\DATA\gobbeldygook.dat')

If the path\filename does not exist when the database tries to start you will receive an error!!

Moving the Files

With the above in mind, let's look at how we would achieve the goal of moving a databases disk files to new locations.

Our Scenario

The server drives are filling up quickly and you have been asked by the manager to move the disk files to a new set of drives provided by the Windows administrator. The engineer has created your new file paths for you and retained all NTFS ACLs required for the SQL Server services. The drives\paths supplied are as follows (I am using my C drive but this could easily be G or F or some other drive letter);

C:\Program Files\Microsoft SQL Server\MSSQL10_50.DBA\MSSQL\Data

Before making any changes to the OS files and their locations first check the metadata available to you in the following system catalog master.sys.master_files. The important metadata to collect consists of the Logical Filenames and the Physical Names. You may obtain this information using the following query;

SELECT database_id,
       name, 
       physical_name 
 FROM sys.master_files 
 WHERE database_id = DB_ID('SampleServiceCompany2009')

For my database I have the following: 

I need to amend these paths from the "C:\Program Files (x86)\Microsoft SQL Server\MSSQL.1\MSSQL\Data" directory to match the new location provided to me. The first task is to construct a set of T-SQL move commands. The ALTER DATABASE ... MODIFY FILE command really only needs 2 parameters for the file move: the files logical name and the physical_name. Both of these are retrieved in the query shown above. The move commands are extremely simple and as shown earlier take the following form; 

Issuing these commands now will update the system catalogs, however, the new paths will not take effect until the database has been restarted (or taken Offline then brought back Online). The following message informs us of this: 

I'll now take the database Offline using the command highlighted below. 

With the database Offline, I may now move the files(and even rename them if I really wanted to).

A word of caution here. It is advisable to copy and paste the files to the new locations. Only when the database comes online successfully would I then remove the old files.

Once you have copied the files to the new locations you would then bring the database Online, this is done using:

ALTER DATABASE [yourDB] SET ONLINE

Shown below are the typical screenshots you will see if the database fails to start. From the information dialog below click the message column and you will see details of the issue. 

The error details show below provide an indication to the issue, the files probably do not exist (in fact that's generally exactly what an OS Error 2 means). 

If the database starts successfully you will see the following 

Help, My Database Won't Start

In every scenario I have encountered whereby a database file move has failed, the issue has been down to a mistyped path and\or filename, resulting in the DBA then getting into a vicious loop trying to correct themselves. Should your database fail to start, don't panic. Perform the following tasks;

• Check the script you used to modify the database file locations, have you got the paths and filenames correct?
• Have your admin check the permissions to the new path for the SQL Server Database Engine service account.
• Query the catalog master.sys.master_files, do the paths\filenames here match what you're expecting?
• If you are unable to complete the move successfully, revert back to the original file paths\names. As you left the files in the source directory simply issue the appropriate ALTER DATABASE ... MODIFY FILE statements and bring the database back online.

Querying master.sys.master_files to obtain the current database file paths 

This is a very easy task to complete providing you pay full attention to the task in hand. If you encounter an error, stop and review what you currently have in place both from the system catalog metadata and the physical OS locations\filenames.

As always, enjoy and if you're still stuck post back and I'll help all I can

It's quite possible that at some point you may want to have the use of a corrupted SQL Server database for test or DR practice purposes. In fact, this can also aid in the recovery steps for a corrupted database once the file and object architectures are fully understood.
Don't ever be tempted to just go run the "Repair_allow_data_loss" clause of DBCC CHECKDB, it does what it says on the tin. I was recently involved on a forum post where a user had done exactly that, without understanding the ramifications or understanding the output of DBCC CHECKDB.

To understand more we will create and corrupt our own test database, this is very easy to achieve as I will detail below. 

For this exercise we merely need a Hex editor and the use of a SQL Server instance.

Note: do not use a Production SQL Server instance!

I have chosen XVI32 as this editor is free of charge and requires no installation to take place, simply place the files into a folder and create a shortcut to the program.

The core database will be created using the following simple script. We’ll go through the process in stages with diagrams to see exactly what’s happening. Start with the code below;

Don't forget to modify any drive letters and paths before executing the script ;-)

USE [master]

CREATE DATABASE [Corrupt2K8] ON PRIMARY

( NAME =N'Corrupt2K8', FILENAME=N'C:\Program Files\Microsoft SQL Server\MSSQL10_50.MSSQLSERVER\MSSQL\DATA\Corrupt2K8.mdf',

SIZE = 524288KB , MAXSIZE = UNLIMITED, FILEGROWTH = 1024KB )

LOG ON

( NAME = N'Corrupt2K8_log', FILENAME = N'C:\Program Files\Microsoft SQL Server\MSSQL10_50.MSSQLSERVER\MSSQL\DATA\Corrupt2K8_log.ldf',

SIZE = 262144KB , MAXSIZE = 2048GB , FILEGROWTH = 1024KB)

GO

USE [Corrupt2K8]

GO

CREATE TABLE dbo.NoddyTable(

NoddyID UNIQUEIDENTIFIER NOT NULL DEFAULT NEWID()

, NoddyName VARCHAR(128) NULL

, NoddyInt BIGINT NULL

, NoddyDate DATETIME NULL

) 

GO

INSERT INTO dbo.NoddyTable

SELECT NEWID(), name, ROUND(RAND(object_id)*856542, 0), GETDATE() FROM sys.columns

UNION ALL

SELECT NEWID(), name, ROUND(RAND(object_id)* 1048576, 0), GETDATE() FROM sys.columns

ALTER TABLE dbo.NoddyTable ADD CONSTRAINT PK_NoddyID

PRIMARY KEY CLUSTERED (NoddyID)

WITH (IGNORE_DUP_KEY=OFF)

CREATE NONCLUSTERED INDEX IDX_NoddyName_NoddyDate

ON dbo.NoddyTable(NoddyName, NoddyDate)

WHERE NoddyName IN ('password','length','created','crtype','offset','intprop')

CREATE NONCLUSTERED INDEX IDX_NoddyDate

ON dbo.NoddyTable(NoddyDate)

Once you have the database created, take a full backup followed by a differential and then a transaction log backup, you may then use these in future testing scenarios.

We now want to choose an object as the target of our corruption exercise. I am going to choose a non clustered index on the table 'dbo.NoddyTable', to get a list of indexes on this table use the following query;

SELECT OBJECT_NAME(object_id), name, index_id, type_desc FROM sys.indexes

ORDER BY 1

I will be using the non-clustered index 'IDX_NoddyDate', this has an index id of 3. To find details of the page numbers in use by this index we now need to turn to an undocumented DBCC command 'DBCC IND'. Full details of how to use this command may be found at the links below but basically this is used as follows;

DBCC IND (DatabaseName,'tablename', index_id)

So, I have

DBCC IND (Corrupt2K8,'dbo.NoddyTable', 3)

Below is the output from the command above

I'm going to pick a page of page type 2 (an index page), my chosen page number here is 174.

Next I need to go view a dump of this page just to have a look at the records it contains. This requires the use of another undocumented DBCC command called DBCC PAGE. Again full details of this are in the links below but basically it's used as follows;

DBCC PAGE (DatabaseName, filenumber, pagenumber, printoption)

So, I have the following code

 --switch on client output first

DBCC TRACEON(3604)

--now read the page

DBCC PAGE (Corrupt2K8, 1, 174, 1)

This is the page header;

I'm going to home in on slot7 or record 7. I'll use the Hex editor to modify this record in the page which will then generate an error when DBCC CHECKDB is run. The detail for slot 7 looks as follows;

So, to hack the record at slot 7 on page 174, I first need to work out some figures to find the address locations within the file. Convert the record offset (indicated in the screenshot above) from hex to decimal first and then the address for slot 7 is calculated as follows

page number x num of bytes per page + record offset

This equates to 174 x 8192 + 292 = 1425700

Take the database offline and now open the primary data file using XVI32. From the File menu select open and then browse to the MDF file.

Now, from the File menu select "Address" > "Goto". In the dialog box which appears ensure you select the decimal radio button and enter the address which was calculated above, in my case 1425700. As you can see from the screenshot below, the editor has placed me at the start of my chosen record in page 174.

This record has a length of 28 bytes which was detailed in the page dump we did earlier, now to modify the record. First switch the editor to Text and Overwrite Mode if it isn't already. From the File menu ensure "Tools" > "Text Mode" and "Tools" > "Overwrite" are selected. Now I'll mark the blocks I wish to mangle. To do this, from the File menu select "Edit" > "Block <n> chars". Switching to decimal, I enter the record length of 28. The blocks have now been marked in red as shown below 

Now, to overwrite the record in slot 7 as shown below, in text mode type a simple string

Now click "File" > "Exit" and save the file when prompted to do so. In SQL Server you may now bring the database back online. We'll re run the page dump and check the results which are shown below;

Well, as we can see above the record was modified in the anticipated location, of course the only part hosed here is the non clustered index which is easily fixed by dropping and re creating it. What does DBCC CHECKDB show us?

Now you have a corrupt database which you may use for your DR and script tests. Give this a go in your test systems and by all means post back if you're stuck. 

Credits

Information on these undocumented procedures was digested from Paul Randal's blogs at the following links

http://blogs.msdn.com/b/sqlserverstorageengine/archive/2006/12/13/more-undocumented-fun_3a00_-dbcc-ind_2c00_-dbcc-page_2c00_-and-off_2d00_row-columns.aspx

http://blogs.msdn.com/b/sqlserverstorageengine/archive/2006/08/09/692806.aspx

http://blogs.msdn.com/b/sqlserverstorageengine/archive/2006/06/10/625659.aspx

Warnings

Obviously you should not perform this on your production databases\servers, complete all tests in your offline environments. I will not be held responsible for those who wish to apply this in online environments