Active Directory Backup and Restore

Overview

Active Directory is backed up as part of system state, a collection of system components that depend on each other. You must back up and restore system state components together.
Components that comprise the system state on a domain controller include:

• System Start-up Files (boot files). These are the files required for Windows 2000 Server to start.
  
• System registry.
  
• Class registration database of Component Services. The Component Object Model (COM) is a binary standard for writing component software in a distributed systems environment.
  
• SYSVOL. The system volume provides a default Active Directory location for files that must be shared for common access throughout a domain. The SYSVOL folder on a domain controller contains:
  
  • NETLOGON shared folders. These usually host user logon scripts and Group Policy objects (GPOs) for non-Windows 2000based network clients.
    
  • User logon scripts for Windows 2000 Professionalbased clients and clients that are running Windows 95, Windows 98, or Windows NT 4.0.
    
  • Windows 2000 GPOs.
    
  • File system junctions.
    
  • File Replication service (FRS) staging directories and files that are required to be available and synchronized between domain controllers.
    
• Active Directory. Active Directory includes:
  
  • Ntds.dit: The Active Directory database.
    
  • Edb.chk: The checkpoint file.
    
  • Edb*.log: The transaction logs, each 10 megabytes (MB) in size.
    
  • Res1.log and Res2.log: Reserved transaction logs.
    

Note: If you use Active Directory-integrated DNS, then the zone data is backed up as part of the Active Directory database. If you do not use Active Directory-integrated DNS, you must explicitly back up the zone files. However, if you back up the system disk along with the system state, zone data is backed up as part of the system disk.If you installed Windows Clustering or Certificate Services on your domain controller, they are also backed up as part of system state. Details of these components are not discussed in this guide.

General Guidelines for Backup

The backup tool in Windows 2000 Server supports multiple types of backup: normal, copy, incremental, differential, and daily. However, because Active Directory is backed up as part of system state, the only type of backup available for Active Directory is normal. A normal backup creates a backup of the entire system state while the domain controller is online. In addition, the backup tool marks each file as a backed up file, which clears the archive attribute of the file.
Considerations for ensuring a good backup
To ensure a successful restore from backup, you must know what defines a good backup.
Which domain controllers to back up
At a minimum, back up two domain controllers in each domain, one of which should be an operations master role holder (excluding the relative ID (RID) master, which should not be restored). Note that backup data from a domain controller can only be used to restore that domain controller. You cannot use a backup of one domain controller to restore another.
Contents
A good backup includes at least the system state and the contents of the system disk. Backing up the system disk ensures that all the required system files and folders are present so you can successfully restore the data.
Note: Best performance practice states that the Active Directory's logs and database files should be on separate disks. If you have configured your domain controllers in this manner you will have Active Directory components spread out on multiple drives, such as D:\Winnt\NTDS for your logs and E:\Winnt\NTDS for your database. You do not need to specify these log and database locations in order for them to be backed up; the backup utility will automatically locate and include them when you back up system state.
Age
A backup that is older than the tombstone lifetime set in Active Directory is not a good backup. At a minimum, perform at least two backups within the tombstone lifetime. The default tombstone lifetime is 60 days. Active Directory incorporates the tombstone lifetime into the backup and restore process as a means of protecting itself from inconsistent data.
Deleting an object from Active Directory is a two-step process. When an object is deleted in Active Directory, the object gets converted into a tombstone, which is then replicated to the other domain controllers in the environment to inform them of the deletion. Active Directory purges the tombstone when the tombstone lifetime is reached.
If you restore a domain controller to a state prior to the deletion of an object, and the tombstone for that object is not replicated to the restored domain controller before the tombstone expires, the object remains present only on the restored domain controller, resulting in inconsistent data. Thus, you must restore the domain controller prior to expiration of the tombstone, and allow inbound replication from a domain controller containing the tombstone to complete prior to expiration of the tombstone.
Active Directory protects itself from restoring data older than the tombstone lifetime by disallowing the restore. As a result, the useful life of a backup is equivalent to the tombstone lifetime setting for the enterprise.

General Guidelines for Restore

You can start the restore process by using either the Windows 2000 Server backup utility or another supported utility. You can perform either a non-authoritative restore or an authoritative restore.

How to Select the Appropriate Restore Method

You select the appropriate restore method by considering:

• Circumstances and characteristics of the failure. The two major categories of failure, from an Active Directory perspective, are Active Directory data corruption and hardware failure. Active Directory data corruption occurs when the directory contains corrupt data that has been replicated to all domain controllers or when a large portion of the Active Directory hierarchy has been changed accidentally (such as deletion of an OU) and this change has replicated to other domain controllers.
  
• Roles and functions of the failed server.
  

Non-authoritative restore of Active Directory
A non-authoritative restore returns the domain controller to its state at the time of backup, then allows normal replication to overwrite that state with any changes that have occurred after the backup was taken. After you restore the system state, the domain controller queries its replication partners. The replication partners replicate any changes to the restored domain controller, ensuring that the domain controller has an accurate and updated copy of the Active Directory database.
Non-authoritative restore is the default method for restoring Active Directory, and you will use it in most situations that result from Active Directory data loss or corruption. To perform a non-authoritative restore, you must be able to start the domain controller in Directory Services Restore Mode.
Non-authoritative restore of SYSVOL
When you non-authoritatively restore the SYSVOL, the local copy of SYSVOL on the restored domain controller is compared with that of its replication partners. After the domain controller restarts, it contacts its replication partners, compares SYSVOL information, and replicate the any necessary changes, bringing it up-to-date with the other domain controllers within the domain.
Perform a non-authoritative restore of SYSVOL if at least one other functioning domain controller exists in the domain. This is the default method for restoring SYSVOL and occurs automatically if you perform a non-authoritative restore of the Active Directory.
If no other functioning domain controller exists in the domain, then perform a primary restore of the SYSVOL. A primary restore builds a new File Replication service (FRS) database by loading the data present under SYSVOL on the local domain controller. This method is the same as a non-authoritative restore, except that the SYSVOL is marked primary.
Authoritative restore of Active Directory
An authoritative restore is an extension of the non-authoritative restore process. You must perform the steps of a non-authoritative restore before you can perform an authoritative restore. The main difference is that an authoritative restore has the ability to increment the version number of the attributes of all objects in an entire directory, all objects in a subtree, or an individual object (provided that it is a leaf object) to make it authoritative in the directory. Restore the smallest unit necessary, for example, do not restore the entire directory in order to restore a single subtree.
As with a non-authoritative restore, after a domain controller is back online, it will contact its replication partners to determine any changes since the time of the last backup. However, because the version number of the object attributes that you want to be authoritative will be higher than the existing version numbers of the attribute held on replication partners, the object on the restored domain controller will appear to be more recent and therefore will be replicated out to the rest of the domain controllers within the environment.
Unlike a non-authoritative restore, an authoritative restore requires the use of a separate tool, Ntdsutil.exe. No backup utilities— including the Windows 2000 Server system tools— can perform an authoritative restore.
An authoritative restore will not overwrite new objects that have been created after the backup was taken. You can authoritatively restore only objects from the configuration and domain-naming contexts. Authoritative restores of schema-naming contexts are not supported.
Perform an authoritative restore when human error is involved, such as when an administrator accidentally deletes a number of objects and that change replicates to the other domain controllers and you cannot easily recreate the objects. To perform an authoritative restore, you must start the domain controller in Directory Services Restore Mode.
Authoritative restore of SYSVOL
By authoritatively restoring the SYSVOL, you are specifying that the copy of SYSVOL that is restored from backup is authoritative for the domain. After the necessary configurations have been made, Active Directory marks the local SYSVOL as authoritative and it is replicated to the other domain controllers within the domain.
The authoritative restore of SYSVOL does not occur automatically after an authoritative restore of Active Directory. Additional steps are required.
As with Active Directory authoritative restore, you typically perform an authoritative restore of SYSVOL when human error is involved and the error has replicated to other domain controllers. For example, you might perform an authoritative restore of SYSVOL if an administrator has accidentally deleted an object that resides in SYSVOL, such as a Group Policy object.
Recover a domain controller through reinstallation
To recover a domain controller through reinstallation, you do not restore the system state from backup media; instead, you reinstall Windows, install Active Directory, and allow replication partners to bring the recovered domain controller up to date.
Recovering a domain controller through reinstallation can quickly return the computer to service if the following conditions exist:

• A domain controller has failed and you cannot restart in Directory Services Restore mode. If failure was caused by a hardware failure, you have resolved the hardware problem (for example, by replacing the disk).
  
• There are other domain controllers in the domain, to serve as replication partners.
  
• The computer is functioning only as a domain controller (it does not run other server services such as Exchange), and it does not contain other data that needs to be recovered from a backup.
  

Restore a domain controller through reinstallation and restore from backup
This method involves first reinstalling Windows 2000, to enable you to start in Directory Services Restore Mode. During the Windows 2000 Server setup process, you will obtain more information about the nature of the failure and you can then determine whether you can reinstall Windows 2000 Server into the same partition as it was previously installed or whether you will need to re-partition the drive. After you successfully reinstall Windows 2000, you can start in Directory Services Restore Mode and perform a normal non-authoritative restore from backup media.
Restore a domain controller through reinstallation and restore the system state from backup if the following conditions exist:

• A domain controller has failed and you cannot restart in Directory Services Restore mode. If failure was caused by a hardware failure, you have resolved the hardware problem (for example, by replacing the disk).
  
• You have the following information about the failed domain controller:
  
  • Disk configuration. You need a record of the volumes and sizes of the disks and partitions. You use this information to recreate the disk configuration in the case of a complete disk failure. You must recreate all disk configurations prior to restoring system state. Failure to recreate all disk configurations can cause the restore process to fail and can prevent you from starting the domain controller following the restore.
    
  • Computer name. You need the computer name to restore a domain controller of the same name and avoid changing client configuration settings.
    
  • Domain membership. You must know the domain name because even if the computer name does not change, you might need to re-establish a new computer account.
    
  • Local Administrator password. You must know the local computer's Administrator password that was used when the backup was created. Without it, you will not be able to log on to the computer to establish a domain account for the computer after you restore it. If you are not part of the domain, you will not be able to log on by using a domain account, even if you are a domain administrator. The local Administrator password is also required to restore the system state on a domain controller.
    
• The domain controller is running other server services such as Exchange, or contains other data you must restore from a backup.
  
• You have a good backup, made within the tombstone lifetime.
  

Considerations for restoring operations masters
To restore an operations master role holder, you must perform one of the following procedures:

• Restore the failed operations master from backup.
  
• Seize the role to another domain controller within the environment. Seize the operations master role only if you do not intend to restore the original role holder from backup. For more information about seizing operations master roles, see "Managing Operations Masters" in this guide.
  

Restoring the RID Master can result in Active Directory data corruption, so it is not recommended.
Restoring the Schema Master can result in orphaned objects, so it is not recommended.
Considerations for recovering global catalog servers
To recover the global catalog server you can either:

• Restore the failed global catalog server from backup.
  
• Assign a new global catalog to compensate for the loss of the original.
  

Restoring from backup is the only way that a domain controller that was functioning as a global catalog at the time of backup can automatically be restored to the role of global catalog. Restoring a domain controller by reinstallation does not automatically reinstate the global catalog role. In a multi-domain environment, be aware that restoring a global catalog server from backup requires more time than restoring a domain controller that does not host the global catalog.
As there are no real disadvantages in configuring multiple global catalogs, you might want to create a new global catalog in your environment if you anticipate an extended downtime for the failed global catalog server. Creating a new global catalog server is particularly relevant if users associated with the original global catalog server can no longer access a global catalog server, or if the requirement for the global catalog service is significant in your environment, such as when you are running Exchange 2000.
For more information about creating a new global catalog server, see "Managing Global Catalogs Servers" in this guide.
Note: Configuring multiple global catalogs servers in a forest increases the availability of the system, but also increases replication traffic and database size. If you do restore the failed domain controller and maintain its role as a global catalog server, you might want to remove any additional global catalogs servers that you configured during its absence.
Considerations for restoring onto different hardware
It is possible to restore a domain controller onto different hardware. However, you should consider the following issues:

• Different hardware abstraction layers (HALs). By default, the Hal.dll is not backed up as part of system state, however the Kernel32.dll is. Therefore, if you try to restore a backup onto a computer that requires a different HAL (for example, to support a multiprocessor environment) compatibility issues exist between the new HAL and the original Kernel32.dll. To overcome this incompatibility, manually copy the Hal.dll from the original computer and install it on the new computer. The limitation is that the new computer can use only a single processor.
  
• Incompatible Boot.ini File. If you backup and restore the boot.ini file, you might have some incompatibility with your new hardware configuration, resulting in a failure to start. Before you restore it, ensure that the boot.ini file is correct for your new hardware environment.
  
• Different Network or Video Cards. If your new hardware has a different video adapter or multiple network adapters, then uninstall them before you restore data. When you restart the computer; the normal Plug and Play functionality makes the necessary changes.
  
• Disk Space and Partition Configuration. Partitions on the new computer must match those on the original computer. Specifically, all the drive mappings must be the same and the partition size must be at least equal to that on the original computer.
  

Considerations for authoritative restores
Performing an authoritative restore can affect group membership and passwords for trusts and computer accounts.
Impact on group membership
By performing an authoritative restore, you risk possible loss of group membership information.
Because group membership is a multi-valued attribute, and because of how Active Directory handles links, back links and deletions, an authoritative restore can produce varying results to group membership. These variations are based on which objects replicate first after an authoritative restore: the User object or the Group object.
If the un-deletion of the user replicates first, then the group membership information of both the group (the members it contains) and the user (the groups to which the user belongs) will be represented correctly.
If the un-deletion of the group replicates first, the replication partners will drop the addition of the (locally) deleted user from the group membership. The only exception to this is the user's primary group, which is always represented correctly both from the user and group reference.
You cannot control which object replicates first after you perform an authoritative restore. If your environment is affected by this situation, the only option is to modify the group membership attribute of the affected groups on the domain controller where you performed the authoritative restore.
This issue stems not from the integrity of the restored data, but from the way in which the data is replicated. By looking at this domain controller, administrators can view the way the directory should look and take steps to replicate the accurate directory information to the other domain controllers within the domain.
The best way to do this is to add a fictitious user and then delete that same fictitious user to and from each group that was involved in the authoritative restore.
A group is involved in the restore if it was either authoritatively restored itself or if it had members restored who did not have that group defined as their primary group.
By doing this, you force the correct group membership information to be replicated out from the source domain controller (the domain controller on which you performed the original authoritative restore) and update the group membership information on its replication partners. These updated objects reflect the correct memberships and also correct the information represented in the Member of tab of the restored user objects' properties.
You must ensure that no additions are made to group membership (for the affected groups and users) on any of the other domain controllers within the environment.
If you do not adhere to this process, the accurate version of the directory (held on the domain controller where the restore was performed) can become corrupted by the incorrect membership information. If the accurate version of the directory becomes corrupted, you must either update group membership manually or perform another authoritative restore of the objects by using the verinc option, and perform the process again.
Impact on trusts and computer accounts
In Windows 2000, trust relationships and computer account passwords are negotiated at a specified interval (by default 30 days for trust relationships and computer passwords).
When you perform an authoritative restore, you might restore previously used passwords for the objects in the Active Directory that maintain trust relationships and computer accounts.
In the case of trust relationships, this can impact communication with other domain controllers from other domains, causing permissions errors when users try to access resources in other domain. To rectify this, you must remove and recreate NTLM trust relationships to Windows 2000 or Windows NT 4.0 domains.
In the case of a computer account password, this can impact communications between the member workstation or server and a domain controller of its domain. This effect might cause users on Windows NT or Windows 2000 computers to have authentication difficulty due to an invalid computer account.

Top of page

Backup and Restore Tasks and Procedures

Table 1.8 shows the tasks and procedures for backup and restore.
Table 1.8 Backup and Restore Tasks and Procedures

Tasks	Procedures	Tools	Frequency
Back up Active Directory and associated components.	Back up system state on a domain controller. Back up system state and system disk on a domain controller.	NTBackup.exe	At least twice within the tombstone lifetime
Perform a non-authoritative restore.	Restart the domain controller in Directory Services Restore Mode (locally or remotely). Restore from backup media. Verify Active Directory restore.	NTBackup.exe Ntdsutil.exe Event Viewer Repadmin.exe	As needed
Perform an authoritative restore of a subtree or leaf object.	Restart in Directory Services Restore Mode. Restore from backup media for authoritative restore. Restore system state to an alternate location. Perform authoritative restore of the subtree or leaf object. Restart in normal mode. Restore applicable portion of SYSVOL from alternate location. Verify Active Directory restore.	NTBackup.exe Ntdsutil.exe Event Viewer Repadmin.exe	As needed
Perform an authoritative restore of the entire directory.	Restart in Directory Services Restore Mode. Restore from backup media for authoritative restore. Restore system state to an alternate location. Restore the database. Restart in normal mode. Copy SYSVOL from alternate location. Verify Active Directory restore.	NTBackup.exe Ntdsutil.exe Event Viewer Repadmin.exe	As needed
Recover a domain controller through reinstallation.	Clean up metadata. Install Windows 2000 Server. Install Active Directory.	Ntdsutil.exe Active Directory Sites and Services Active Directory Users and Computers Dcpromo.exe	As needed
Restore a domain controller through reinstallation and subsequent restore from backup.	Install Windows 2000 Server on the same drive letter and partition as before the failure, partitioning the drive if necessary. Restore from backup media (non-authoritative restore). Verify Active Directory restore.	NTBackup.exe	As needed

Backing Up Active Directory and Associated Components

To back up Active Directory and associated components on a domain controller, you can back up only system state or you can back up both system state and the system disk.

Procedures for Backing Up Active Directory and Associated Components

Use one of the following procedures to back up Active Directory and associated components. Procedures are explained in detail in the linked topics.

1. Back up system state.
   
2. Back up system state and the system disk.
   

Top of page

Performing a Non-Authoritative Restore

Non-authoritative restore is the default method for restoring Active Directory, and you use it in most situations that result from Active Directory data loss or corruption. You must be able to start in Directory Services Restore Mode to perform a non-authoritative restore. After you restore the domain controller from backup media, replication partners use the standard replication protocols to update both the Active Directory and FRS on the restored domain controller.

Procedures for Performing a Non-Authoritative Restore

Use the following procedures to perform a non-authoritative restore of a domain controller. Procedures are explained in detail in the linked topics.

1. Restart the domain controller in Directory Services Restore Mode (locally or remotely).
   
2. Restore from backup media.
   
3. Verify Active Directory restore.
   

Top of page

Performing an Authoritative Restore of a Subtree or Leaf Object

An authoritative restore of a subtree or leaf object restores that subtree or leaf and marks it as authoritative for the directory. You begin by restoring from backup media, just as in a non-authoritative restore, but then you perform additional steps to complete an authoritative restore.

Top of page

Procedures for Authoritative Restore of a Subtree or Leaf Object

Use the following procedures to perform an authoritative restore of an Active Directory subtree or leaf object. Procedures are explained in detail in the linked topics.

1. Restart the domain controller in Directory Services Restore Mode (locally or remotely).
   
2. Restore from backup media for authoritative restore.
   
3. Restore system state to an alternate location.
   
4. Perform authoritative restore of the subtree or leaf object.
   
5. Restore applicable portion of SYSVOL from alternate location if necessary.
   
6. Verify Active Directory restore.
   

Top of page

Performing an Authoritative Restore of Entire Directory

Authoritative restore of the entire directory is a major operation. Perform an authoritative restore of the entire directory only after consultation with a Microsoft Support professional. Do not perform an authoritative restore of the entire directory if only one domain controller exists in the domain.

Top of page

Procedures for Authoritative Restore of the Entire Directory

Use the following procedures to perform an authoritative restore of the entire Active Directory. Procedures are explained in detail in the linked topics.

1. Restart the domain controller in Directory Services Restore Mode (locally or remotely).
   
2. Restore from backup media.
   
3. Restore system state to an alternate location.
   
4. Perform authoritative restore of entire directory.
   
5. Restore SYSVOL from alternate location.
   
6. Verify Active Directory restore.
   

Recovering a Domain Controller Through Reinstallation

Recovering through reinstallation is the same process as creating a new domain controller. It does not involve restoring from backup media. This method relies on Active Directory replication to restore a domain controller to a working state, and is only valid if another healthy domain controller exists in the same domain. This option is normally used on computers that function only as a domain controller.

Bandwidth Considerations

The primary consideration when recovering a domain controller through replication is bandwidth. The bandwidth required is directly proportional to the size of the Active Directory database and the time in which the domain controller is required to be at a functioning state. Ideally, the existing functional domain controller is located in the same Active Directory site as the replicating domain controller (new domain controller) in order to reduce network impact and restore duration.

Procedures for Recovering a Domain Controller Through Reinstallation

Use the following procedures to recover a domain controller. Procedures are explained in detail in the linked topics.

1. Clean up metadata.
   
2. Reinstall Windows 2000 Server. (This procedure is not covered in this guide.)
   
3. Install Active Directory. During the installation process, replication occurs, ensuring that the domain controller has an accurate and up to date copy of the Active Directory. For more information about seizing operations master roles, see "Installing Active Directory" in this guide.
   

Restoring a Domain Controller Through Reinstallation and Subsequent Restore from Backup

If you cannot restart a domain controller in Directory Services Restore Mode, you can restore a domain controller through reinstallation and subsequently restore Active Directory from backup. This option is normally used on domain controllers that also run other services, such as Exchange, or have other data you want to recover.

Procedures for Restoring a Domain Controller Through Reinstallation and Subsequent Restore from Backup

To restore a domain controller through reinstallation and subsequently restore Active Directory from backup, you must ensure that you install Windows 2000 Server on the same drive letter and on a partition that is at least as large as the partition used before the failure. You must repartition the drive if necessary. After you reinstall Windows 2000, perform a non-authoritative restore of the system state and the system disk. Procedures are explained in detail in the linked topics.

1. Install Windows 2000 Server on the same drive letter and partition as before the failure. (This procedure is not covered in this guide.)
   
2. Restore from backup media.
   
3. Verify Active Directory restore.
   

Active Directory database basics: Performing an offline defrag

http://searchwindowsserver.techtarget.com/tip/0,289483,sid68_gci1325578_mem1,00.html

For the most part, the Active Directory database just works. It performs automatic online defrag, rarely becomes corrupt and never requires manual database recovery procedures like Exchange Server does. Still, it's important for administrators to understand the basics of an Active Directory database – how it works and certain important maintenance procedures.
Figure 1 shows where the database (NTDS.DIT file) sits in the Active Directory architecture. You can see how protocols like LDAP interface with the Directory Service Agent (DSA). The DSA is responsible for functions such as schema enforcement of updates, access control enforcement, object identification, referrals and functional level definition. It is associated with a GUID -- specifically the objectGUID attribute of NTDSsettings object -- that is used in replication to identify replication partners. This is exposed in the Repadmin/showrepl command and is the GUID that is mapped to the server's FQDN in the DNS Alias record.
Figure 1 (click to enlarge)

There is also a database GUID that is the invocationID attribute of the NTDSsettings object. The database layer is responsible for the creation, deletion and modification of objects, as well as the retrieval of objects, attributes and the schema cache. The schema defines rules for the organization of the database in terms of classes and attributes.
The NTDS.dit -- located in the %systemroot%\ntds directory -- exists on every Windows server installation. It is a basic transactional Jet database just like Exchange Server, and it is recommended to store the database and logs on separate physical disks. The location for these files is determined during Dcpromo, but you can change the location using the NTDSUtil program described later in this article. Figure 2 shows a typical NTDS directory.
Figure 2 (click to enlarge)

NTDS.DIT is the database. During Dcpromo, it is enhanced with data from other domain controllers if it's joined to an existing domain or starts a new domain. As objects and attributes are added, deleted or modified, the database gains "whitespace" (unused space).
Adding users, computers, printers and other objects along with defining various attributes will cause the database to grow, and it can be anywhere from a few MB in size to several GB. Being able to load the database entirely in addressable memory will significantly improve operations such as authentication. Therefore it is normally recommended for DCs with 4 GB or more of physical RAM to use the \3 GB switch in Boot.ini to expand the user mode section of memory, which permits more of the NTDS.DIT to fit in there.
Windows Server 2003 made an important change that significantly reduced the database size. While a security descriptor value is stored for every object in Windows 2000 Server, Windows Server 2003 uses single instance descriptors, allowing multiple objects to use a single descriptor.
Defragging the Active Directory database
It's important to defragment the Active Directory database for best performance. Normally there is an online defragmentation that occurs about twice a day on the database, but this is more of a backup than defrag. It can give admins a false sense of security, thinking that the database has been defragged without any downtime.
The only way to truly defragment the Active Directory database, remove whitespace and decrease its size is with an offline defrag. This requires you to take Active Directory offline by booting a DC into Directory Service Restore Mode (DSRM), which boots up the DC in safe mode without mounting the AD database. Once booted, enter the NTDSUtil program and you will be able to perform a number of actions that are not possible with Active Directory online, including those in the File menu as shown in Figure 3.

Figure 3 (click to enlarge)

Significant functions include:

• Header -- Dumps the Jet database header with information, such as database location, the state (i.e., clean or dirty shutdown, last backup, database size, etc.). This is shown in Figure 4.
  
• Checksum -- Performs a Jet physical integrity check, a good thing to do before defragmenting. See Figure 2.
  
• Integrity -- Performs a logical integrity check of the Jet database. After running this command you will see a message recommending that you run the Semantic Database Analysis function. This option is found in the main NTDSUtil menu. This is a good one to run if you get events in the system event log indicating database inconsistency, corruption or other errors. The Semantic Database Analysis function will be covered in more detail in a future article.
  
• Compact to %S -- This command is the defrag operation that squeezes the whitespace out of the database.

Figure 4 (click to enlarge)

Note that there are other commands available that allow you to move the database and log files, set the online backup directory path and even perform a soft database recovery.
To perform the actual defrag of the Active Directory database, the following steps are required:

1. Create a directory that will temporarily store the compacted (defragmented) database. In our example here, we have created C:\ESE-Backup.
   
2. From NTDSUtil, go to the File menu option. In the File Maintenance prompt, enter: Compact c:\ESE-Backup
   When the compacting operation completes, you will see the following text:

   If compaction was successful, you need to:
         copy "c:\ese-backup\ntds.dit"
         "C:\WINDOWS\NTDS\ntds.dit"
   and delete the old log files:
         del C:\WINDOWS\NTDS\*.log

   
   
3. It follows that the next step is to copy the new compacted ntds.dit in c:\ese-backup to c:\windows\ntds\ntds.dit (overwrite the old file), and then delete the 4 logs. You will only have the edb.chk, temp.edb and ntds.dit files left.
   
4. Restart the DC in normal mode.

If you want to play with this in a test domain, you can record the size of the NTDS.DIT and then create a large number of users -- say 10,000 -- in the domain. Note the increased size of the NTDS.DIT. Then delete the users and follow the procedure just described to compact the database and replace the old one with the whitespace for the 10,000 deleted users. Compare the size of the new, compacted database with the size after you created the users. You could also wait for an online defrag of the database to occur (recorded with an event in the system event log) and see if that changes the NTDS.DIT size. Note that in a production situation, you want to perform database integrity checks to ensure stability.
It is not necessary to perform this offline defrag on a regular basis, but it's good to do it after significant changes have been made -- such as the removal of a large number of users or groups -- to keep the database at an efficient size.