Managing Domain Controllers with Active Directory Directory Services
On This Page
Overview Installing and Removing Active Directory Renaming Domain Controllers Managing Global Catalog Servers Managing Operations Masters Managing the Database Managing SYSVOL Managing Windows Time Service Managing Long-Disconnected Domain Controllers
Overview
While individual domain controllers require little management, your overall operations environment might require change-related tasks, such as adding or removing domain controllers, or reintroducing a domain controller that has been offline for more than one replication cycle. During your day-to-day operations, you might need to do some or all of the following:
Install and remove Active Directory
Rename domain controllers
Manage global catalog servers
Manage operations masters
Manage the database
Manage SYSVOL
Manage Windows Time Service
Manage long-disconnected domain controllers
Installing and Removing Active Directory
Only domain controllers can host Active Directory. All servers that are not domain controllers must access the directory in the same manner as the workstations. They send requests for information to a domain controller, which processes the request and returns the information back to them.
Domain controllers store and maintain portions of the directory. They also have services that allow them to directly store and retrieve information from the directory. These services are referred to as the Active Directory. When you install Active Directory on a Windows 2000based server, it becomes a Windows 2000based domain controller.
The process of removing Active Directory involves steps similar to those for installation. You run many of the same tests before you remove the directory as you run before you install the directory. These tests ensure that the process occurs without any problems. In the event that a domain controller suffers a hardware failure and you plan to never return it to service, you must take additional steps to remove it from the directory.
The Active Directory Installation Wizard
You install Active Directory by running the Active Directory Installation Wizard on a Windows 2000based server. The wizard simplifies the process by automating as much of the installation process as possible. During the installation, the wizard asks for the name of the domain that you want this domain controller to host, and for the location where you want to install required files. To run the Active Directory Installation Wizard, you must be a member of the Domain Admins group.
Active Directory Installation Prerequisites
This guide covers the installation of Active Directory in an environment that is configured according to the best practices described in Best Practice Active Directory Design for Managing Windows Networks and Best Practice Active Directory Deployment for Managing Windows Networks. To download these guides, see the Active Directory link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources. They describe the process of planning your forests and domains and provide recommendations for deploying DNS. They also provide guidelines for estimating the number of domains as well as the number of domain controllers in each domain.
Before you begin your installation, the following conditions must exist in your environment:
Your Active Directory forest must already exist. At least two properly functioning domain controllers must reside in the forest root.
Your Active Directory Domain must already exist. At least two properly functioning domain controllers must reside in the domain.
DNS must be functioning properly.
You must use Active Directoryintegrated DNS zones. You must configure at least one domain controller as a DNS server.
Note: Creating or removing a domain or forest is beyond the scope of this guide. This guide does not cover deploying DNS into an environment that has not previously hosted a DNS infrastructure.
For information about these options, see the Active Directory link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources and the Microsoft Windows 2000 Server Deployment Planning Guide.
Active Directory Installation Preparation
Properly preparing for the installation of Active Directory decreases the chances of problems during the installation process and helps you quickly complete the operation. Preparation includes installing and configuring DNS and gathering information that you need for the installation.
Configure all domain controllers as DNS servers. Install the DNS server service prior to installing Active Directory. Follow the recommendations mentioned earlier so that your domain is already configured, DNS is functioning, and you have Active Directoryintegrated DNS zones. Installing the DNS Server service prior to installing Active Directory allows the DNS Server service to automatically start using the DNS zones that are stored on the directory after you complete the Active Directory installation.
The installation wizard asks for specific configuration information, such as the domain administrator's user name and password, location of the directory database and log files, and the password needed to us Directory Services Restore Mode, before it begins installing Active Directory. Have that information ready before you run the Active Directory Installation Wizard.
Note: For better performance, store the log files and the Ntds.dit file on separate hard disks.
Active Directory Installation
During the installation, the Active Directory Installation Wizard communicates with other domain controllers to obtain configuration information. This information can come from any domain controller in the same domain. The Active Directory Installation Wizard also communicates with the various operations masters so that the new domain controller can properly join the domain and be added to the directory. For this process to succeed, the wizard must be able to communicate with the various domain controllers involved. Test these channels of communication prior to installing Active Directory to help ensure that the process does not encounter problems during the installation.
After successfully testing the communication paths, the Active Directory Installation Wizard installs Active Directory on the server to make it a domain controller. During the installation process, the wizard asks for the information that you gathered during the preparation phase. After the wizard finishes, it restarts the domain controller and the installation completes during the restart process.
Active Directory Post-installation Tasks
After you complete the installation of Active Directory, perform some validation tests to ensure that the domain controller is properly joined to the domain and is functioning as expected. The areas you must test include:
Site placement
DNS configuration
Network connectivity
SYSVOL
Replication
If your tests show that all of these areas are configured and functioning properly, the Active Directory installation is successful.
Active Directory Unattended Installation
You can automate the Active Directory installation process by performing an unattended installation. You can create an answer file to answer the questions that the Active Directory Installation Wizard asks during the installation. The installation does not require user input and proceeds quickly.
For more information about unattended installation options, see "Using the Answer File with the Active Directory Installation Wizard" in the Deployment Planning Guide.
Domain Controller Removal
A domain controller can be removed from a domain in one of two ways: by removing Active Directory or by a system failure that renders the domain controller inoperable so that you cannot restore it to service.
Active Directory removal
Similarly to how you can install Active Directory to turn a Windows 2000based server into a domain controller, you can remove Active Directory and turn a Windows 2000based domain controller back into a server. This process removes most of the references to the domain controller from the directory. You must manually remove the server object that represents the domain controller from the computer container after you remove Active Directory. This method properly removes the domain controller from the directory.
Domain controller failure
A hardware failure on a domain controller can render it inoperable. If the problem is severe enough, you might never be able to return the domain controller to service. In this case, the other domain controllers eventually reconfigure themselves so that they can continue to replicate directory information without the failed domain controller.
When a domain controller is removed from the domain without removing Active Directory, all the information about that domain controller remains in the directory. You must take additional steps to remove this information from the directory.
Active Directory Installation and Removal Management Tasks and Procedures
Table 1.9 shows the tasks and procedures for managing Active Directory installation and removal.
Table 1.9 Active Directory Installation and Removal Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Prepare for Active Directory Installation. |
|
|
As needed. |
Install Active Directory. |
|
|
As needed. |
Perform Active Directory post-installation tasks. |
|
|
As needed. |
Decommission a domain controller. |
|
|
As needed. |
Preparing for Active Directory Installation
Preparation helps the Active Directory installation proceed successfully. To prepare for the installation process, you must have the appropriate domain information and credentials available before you start the Active Directory Installation Wizard. It is recommended that you configure all domain controllers as DNS servers. You must have your DNS server configuration information available for that portion of the installation process.
DNS Service Installation
Domain controllers use DNS to locate other domain controllers that are hosting Active Directory. Configure every domain controller as a DNS server to help ensure that a DNS server is always available. Using Active Directoryintegrated DNS zones simplifies the configuration required because you do not need to create the zone files on each DNS server. Active Directoryintegrated zones are stored in the directory and are replicated to each domain controller along with other Active Directory data. When you start a domain controller that also runs DNS, the DNS Server service detects the zones in the directory and uses them.
Before you install DNS server on a domain controller that you want to host Active Directoryintegrated zones, ensure that you already have other domain controllers functioning in the domain with at least one configured as a DNS server that uses Active Directoryintegrated zones.
For more information about DNS configuration and operations master role placement, see Best Practice Active Directory Design for Managing Windows Networks and Best Practice Active Directory Deployment for Managing Windows Networks. To download these guides, see the Active Directory link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources.
Active Directory Installation Information
Gather the information that you must supply to the Active Directory Installation Wizard before you run the wizard.
Procedures for Preparing for Active Directory Installation
To prepare for the Active Directory installation, install the DNS Server service on the server that you want to make a domain controller and gather the information that you must supply to the Active Directory Installation Wizard.
Gather installation information, including:
The user name, password, and the domain that contains the user account that you intend to use to run the Active Directory Installation Wizard.
The name of the domain that you want the new domain controller to host.
Location for the Active Directory database (Ntds.dit).
Location for the log files.
The server administrator account name and password to use in Directory Services Restore mode.
Installing Active Directory
You install Active Directory by using the Active Directory Installation Wizard (DCPromo.exe). During installation, the wizard contacts other domain controllers for information that it needs to complete the installation. If the wizard cannot communicate with other domain controllers, the installation fails. To help ensure successful installation, test the communication channels prior to running the wizard.
Site Placement
During installation, the Active Directory Installation Wizard attempts to place the new domain controller in the appropriate site. The appropriate site is determined by the domain controller's IP address and subnet mask. The wizard uses the IP information to calculate the subnet address of the domain controller and checks to see if a subnet object exists in the directory for that subnet address. If the subnet object exists, the wizard uses it to place the new server object in the appropriate site. If not, the wizard places the new server object in the same site as the domain controller that is being used as a source to replicate the directory database to the new domain controller. Make sure the subnet object has been created for the desired site prior to running the wizard.
Domain Connectivity
During the installation process, the Active Directory Installation Wizard needs to communicate with other domain controllers in order to join the new domain controller to the domain. The wizard needs to communicate with a member of the domain to receive the initial copy of the directory database for the new domain controller. It needs to communicate with the domain naming master so that the new domain controller can be added to the domain. The wizard also needs to contact the RID master so that the new domain controller can receive its RID pool, and it needs to communicate with another domain controller in order to populate the SYSVOL shared folder on the new domain controller. All of this communication depends on proper DNS installation and configuration. By using Netdiag.exe and Dcdiag.exe, you can test all of these connections prior to starting the Active Directory Installation Wizard.
The Active Directory Installation Wizard
After you have gathered all the information that you need to run the Active Directory Installation Wizard and performed the tests to verify that all of necessary domain controllers are available, you are ready to install Active Directory on your server and turn it into a domain controller.
You need to log on with local administrative credentials to start the wizard. Start the wizard and supply the information you gathered earlier. If the wizard asks for information that you did not gather, such as if you want to install DNS Server service, it is indicating that it cannot locate the DNS servers. The wizard assumes that none exist and asks you if you want to install one. Running the verification tests prior to using the installation wizard helps prevent this kind of situation from happening.
During the installation process, the wizard asks for information that it needs to properly configure the new domain controller. First, it asks is if you want to install a domain controller in a new domain or an additional domain controller in an existing domain. Because this guide pertains to adding domain controllers to domains that already exist, choose Additional domain controller in an existing domain.
During the installation process, the wizard needs to communicate with other domain controllers in order to add this new domain controller to the domain and get the appropriate information into the Active Directory database. To maintain security, you must provide credentials that have administrative access to the directory. Once your credentials are validated, the wizard guides you through the following steps:
The wizard asks for a user name, password, and domain name of the account it uses to add this domain controller to the directory.
The wizard then asks for the name of the domain that you want this new domain controller to host. Enter the fully qualified domain name of the appropriate domain.
Next, the wizard asks where you want to store the Active Directory database and the database log files. For better performance, store these files on separate hard disks.
The wizard then asks for the location where you want to store the shared System Volume (SYSVOL). Ensure that the location has adequate disk space. For more information about ensuring adequate disk space for SYSVOL, see "Managing Sysvol" later in this guide.
The wizard then asks for the password that is assigned to the Directory Services Restore Mode administrator account. This account is not the domain administrator account or the local administrator account on the server, but a special account that can only be used when the domain controller starts in Directory Services Restore Mode.
Before installation begins, the wizard displays a dialog box that summarizes the information that you supplied. Verify that the information is correct before the installation process begins.
Procedures for Installing Active Directory
Verify that an IP address maps to a subnet and determine the site association.
Verify the existence of the operations masters.
Note: If any of the verification tests fail, do not continue until you determine and fix the problems. If these tests fail, the installation is also likely to fail.
Performing Active Directory Post-Installation Tasks
After completing the installation of Active Directory, perform some validation tests to ensure that the domain controller is properly installed into the domain and is functioning as expected. Successfully passing these tests is a good indication that the new domain controller is functioning properly. You might also need to perform additional tasks regarding DNS configuration and hosting the global catalog.
Proper Site Placement
You must ensure that the new domain controller is located in the proper site so that after the installation is complete, the new domain controller can locate replication partners and become part of the replication topology. During Active Directory installation, the wizard creates a server object for the new domain controller in the directory and attempts to place the server object in the proper site. To place the server object, the wizard uses the current IP address and subnet mask of the new domain controller. If the subnet associated with the domain controller's IP address is not defined by an existing subnet object, the wizard places the new server object in the same site as the source domain controller, which is the domain controller from which the new domain controller downloaded a copy of the directory database. If the site is not correct, you can use the Active Directory Sites and Services snap-in to move the server object for the domain controller to the proper site after Active Directory installation is complete.
The last dialog box displayed by the Active Directory Installation Wizard lists the site where the new domain controller is installed. If this is not the proper site, you need must move the server object.
For more information about sites or to create a new site object, see "Managing Site Topology" later in this guide.
Final DNS Configuration
If you installed the DNS server service and made this domain controller a DNS server, you might need to perform some additional configuration of the DNS installation to ensure that it conforms to the recommended practices. The configuration that you must perform depends upon whether this is a new domain controller in the forest root domain or a new domain controller in a child domain. Performing final DNS configuration helps balance the load among your DNS servers and provides redundancy in case a DNS server becomes unavailable.
You might need to add a delegation for the new domain controller. If your forest root domain is a child domain in your corporate DNS domain structure, you must add a delegation for the new domain controller in the forest root's parent DNS domain. If the forest root domain has no parent DNS domain, you do not need to add the delegation.
If the new domain controller is located in a child domain of the forest root domain, you must add a delegation for the new domain controller to the forest root domain.
You also need to configure the DNS client settings on the new domain controller. Configure a domain controller in the forest root domain to refer to another DNS server located nearby as its primary DNS server and refer to itself as the secondary DNS server. If the new domain controller is located in a child domain of the forest root domain, configure the DNS client to use its own IP address as its primary DNS server address, and another local DNS server as the secondary server address.
If the new domain controller is located in a child domain below the forest root, create a secondary zone to make the process of locating domain controllers more reliable.
Whether or not the new domain controller is located in a parent or child domain, you must also configure the DNS server to use either root hints or forwarders for recursive name resolution. Follow the established method on your network.
Domain Connectivity
After the Active Directory Installation Wizard finishes, the domain controller restarts and performs a few tasks before it is ready to assume its role as a domain controller. It registers itself with its DNS server so that other members of the domain know that it is a domain controller and can locate it.
When a new domain controller first joins the network, it receives SYSVOL information from its replication partners. Until it finishes the initial replication of the SYSVOL, it does not create the NETLOGON and SYSVOL shared folders and does not start the Net Logon service, both of which are necessary for it to assume the role of a domain controller. An event number 13516 in the File Replication Service event log indicates that replication is complete and is working properly. At this point, the domain controller starts the Net Logon service and the domain controller becomes available to the domain.
Note: This process can take 15 minutes or longer to complete, depending on the connection speed between the domain controller and its replication partners.
Domain controllers make changes to the directory and replicate these changes among themselves through a series of connections that are established when the domain controller joins the network. The connections can be generated automatically or an administrator might manually create the connections objects. If these connections are not functioning properly, the domain controller cannot replicate changes to the other domain controllers and cannot receive changes from other domain controllers.
To function properly, domain controllers must periodically communicate with various operations masters. The domain controllers send password changes to the PDC emulator. They receive a RID pool from the RID master. As their pools are depleted, the domain controller periodically replenishes their allocations by sending requests to the RID master.
All of these features depend upon communication between the new domain controller and other domain controllers in the domain and forest. When a new domain controller joins the network, perform tests that verify the communication channels used by these features.
Configure Other Roles
After the domain controller is functioning properly and you complete verification tests and final DNS configuration, configure any additional roles, such as global catalog server, on the domain controller. For information about configuring a global catalog server, see "Managing Global Catalog Servers" later in this guide.
Procedures for Performing Active Directory Post-Installation Tasks
To perform this task, the site object must already be defined in Active Directory Sites and Services and you must know the site in which you want to place the server object.
Move a server object to a different site if the domain controller is located in the wrong site.
Perform final DNS configuration for a new domain controller that is located in the forest root domain:
Create a delegation for the new domain controller in the parent domain of the DNS infrastructure if a parent domain exists and a Microsoft DNS server hosts it. If a Microsoft DNS server does not host the parent domain, follow the procedures outlined in the vendor documentation to add the delegation for the new domain controller.
Configure the DNS client settings.
or
Perform final DNS configuration for a new domain controller that is located in a child domain:
Create a delegation for the new domain controller in the forest root domain.
Decommissioning a Domain Controller
Just as you can install Active Directory to make a Windows 2000based server a domain controller, you can also remove Active Directory and to make a Windows 2000based domain controller back into a server.
Removing Active Directory is a similar process to installing it. You use the Active Directory Installation Wizard and it contacts other domain controllers to copy information from the domain controller that you want to decommission. As with installation, if the domain controller cannot contact the other domain controllers during the Active Directory removal, the process is likely to fail. Perform the same connectivity tests prior to decommissioning a domain controller as you perform prior to installing Active Directory.
This guide does not include procedures for decommissioning the last domain controller in a domain. Decommissioning the last domain controller in a domain constitutes the removal of the domain from the forest. For more information about removing domains, see "Removing Active Directory" in the Windows 2000 Server Distributed Systems Guide.
Operations Master Role Transfer
During the decommissioning process, the Active Directory Installation Wizard transfers the operations master roles to other domain controllers without any user interaction. You do not have control over which domain controller receives the roles. The wizard transfers the roles to any available domain controller and does not indicate which domain controller hosts them.
Because of this behavior, transfer any operations master roles prior to running the Active Directory Installation Wizard to decommission a domain controller so you can control operations master role placement. If you need to transfer any roles from a domain controller, understand all the recommendations for role placement before performing the transfer. For more information about transferring operations master roles and role placement, see "Managing Operations Master Roles" later in this guide.
Global Catalog Removal
If you remove Active Directory from a domain controller that hosts the global catalog, the Active Directory Installation Wizard confirms that you want to continue with removing Active Directory. This confirmation ensures that you are aware that you are removing a global catalog from your environment. Do not remove the last global catalog server from your environment because users cannot logon without an available global catalog server. If you are not sure, do not proceed with removing Active Directory until you know at least one other global catalog server is available. For more information about removing and creating global catalog servers, see "Managing Global Catalog Servers" later in the guide.
Domain Connectivity
During the removal of Active Directory, the Active Directory Installation Wizard must communicate with various domain controllers. Any unreplicated changes to the directory must be replicated to another domain controller. The wizard attempts to connect to another domain controller to replicate these changes. The wizard must contact another domain controller so that Active Directory can remove the domain controller from the directory database. If the domain controller hosts any operations master roles that you chose not to transfer, the wizard must contact another domain controller in order to transfer the operations master roles.
If the domain controller cannot contact the other domain controllers during Active Directory removal, the decommissioning operation fails. As with the installation process, test the communication infrastructure prior to running the installation wizard. When you remove Active Directory, use the same connectivity tests that you use during Active Directory installation.
Active Directory Removal
After you transfer operations master roles and verify that all the necessary domain controllers are available, you can use the Active Directory Installation Wizard to remove Active Directory. When you run the wizard on a server that is already a domain controller, it displays the Remove Active Directory options.
The wizard asks whether or not this is the last domain controller in the domain and requests the password that is assigned to the local administrator account on the server after Active Directory is removed. Note that the procedures in this guide do not pertain to removing Active Directory from the last domain controller in the domain, because that action also deletes the domain from the forest.
Server Object Removal
After removing Active Directory from a domain controller, the Active Directory Installation Wizard removes information about that domain controller from the directory. Because it no longer acts as a domain controller, the server is not part of the replication topology and the directory does not maintain connections to it. During the decommissioning process, the Active Directory Installation Wizard removes the server object from the Domain Controller container in Active Directory Users and Computers and removes the connection objects associated with the domain controller from the NTDS Settings object in Active Directory Sites and Services.
The Active Directory Installation Wizard does not delete the server object from the site object during the removal of Active Directory because other services, such as Microsoft Operations Manager 2000 (MOM), use this container to store their own site-specific information. After you remove Active Directory, you can use the Active Directory Sites and Services snap-in to safely remove the server object that represents the decommissioned domain controller in Active Directory Sites and Services if the server object container is empty.
Procedures for Decommissioning Domain Controllers
View the current operations master role holders to see if any roles are assigned to this domain controller.
Transfer the forest-level operations master roles to another domain controller in the forest root domain if this domain controller hosts either the schema master or domain naming master roles.
Transfer the domain-level operations master roles if this domain controller hosts the PDC emulator, infrastructure master, or RID master.
Determine whether a domain controller is a global catalog server to ensure that other domain controllers are configured as global catalog servers before you remove Active Directory.
Verify the existence of the operations masters.
Note: If any of the verification tests fail, do not continue until you determine and fix the problems. If these tests fail, the installation is also likely to fail.
Renaming Domain Controllers
Renaming a domain controller that is running Windows 2000 Server involves the following steps:
Removing Active Directory
Renaming the computer
Reinstalling Active Directory
Restoring the domain controller to its original configuration
When you rename a domain controller, you must reinstall any services that cannot identify the computer name dynamically or that can only operate on a domain controller. You do not need to reinstall any of the services that ship with Windows 2000 Server, such as File and Print sharing or DNS.
It is recommended that you do not rename a domain controller unless it is absolutely necessary. For example, it would be necessary to rename a domain controller if:
You moved the domain controller to another site and the name of the domain controller needs to map to the naming convention of the new site.
The name of the domain controller was chosen in error; such as when the naming convention requires the site name and a derivative of the domain, but the name includes the incorrect site or domain.
Because renaming a domain controller requires that Active Directory be removed and then reinstalled on the computer, the impact on the network of renaming a domain controller is identical to the impact of installing Active Directory to create a new domain controller or global catalog server.
Renaming Domain Controllers Tasks and Procedures
Table 1.10 lists the tasks and procedures for renaming domain controllers.
Table 1.10 Tasks and Procedures for Renaming Domain Controllers
Tasks |
Procedures |
Tools |
Recommended Frequency |
|---|---|---|---|
Identify the current configuration of the domain controller. |
|
|
As needed. |
Rename the domain controller. |
|
|
As needed. |
Restore the original configuration of the domain controller. |
|
|
As needed. |
Identifying the Current Configuration of a Domain Controller
Because renaming a domain controller involves removing and reinstalling Active Directory, you must be able to reestablish the current configuration of the domain controller after you rename it. Before you begin, identify the current configuration of the domain controller so that you can restore it after you reinstall Active Directory. Specifically, determine the status of the following roles and configurations:
Global catalog server. If the domain controller is a global catalog server, the global catalog partial directory partitions are removed when you remove Active directory. Therefore, after you rename the domain controller, you need to reconfigure the domain controller as a global catalog server. For information about configuring a domain controller as a global catalog server, see "Managing Global Catalog Servers" in this guide.
Operations master role holder. If the domain controller holds operations master roles, it is recommended that you transfer the roles to the standby master for the roles prior to removing Active Directory. If you do not transfer the roles, they are transferred automatically, but you have no control over the placement of the roles. By manually transferring the roles prior to removing Active Directory, you control the role placement. For information about transferring operations master roles, see "Managing Operations Masters" in this guide.
DNS server. Removing Active Directory does not remove the DNS Server service if it is installed. However, when you reinstall Active Directory, you need to reconfigure the domain controller to assume authority for the appropriate DNS zones and to contain all appropriate delegations. For information about configuring DNS server after installing Active Directory, see "Managing the Installation and Removal of Active Directory" in this guide.
Initial change notification delay. This server-specific configuration determines how long the domain controller waits before it signals its first replication partner that it has changes. If you change the default initial change notification delay setting on the domain controller, you need to reconfigure the setting when you reinstall Active Directory.
Preferred bridgehead server. This configuration is not recommended for domain controllers running Windows 2000 Server. However, if the domain controller is configured to be a preferred bridgehead server, you must reconfigure the domain controller as a preferred bridgehead server after you reinstall Active Directory. For more information about using preferred bridgehead servers, see "Managing Site Topology" in this guide.
Procedures for Identifying the Current Configuration of a Domain Controller
Use the following procedures to identify the current configuration of the domain controller. You need to reconfigure the current configuration on the renamed domain controller after you reinstall Active Directory.
Determine whether the domain controller is a global catalog server.
View the operations master role holders. If roles are held by this domain controller, transfer the roles to the standby operations master prior to removing Active Directory, as follows:
If the domain controller holds any forest-level roles, transfer forest-level operations master roles.
If the domain controller holds any domain-level roles, transfer domain-level operations master roles.
Determine whether the domain controller is a DNS server. Make a note of the DNS configuration so that you can reproduce it when you reinstall Active Directory.
Determine the initial change notification delay. If this setting has been changed from the default on this domain controller, you need to reconfigure the setting after you rename the server and add Active Directory.
Determine whether the domain controller is a preferred bridgehead server.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Renaming a Domain Controller
Before you rename a domain controller, you must remove Active Directory to return the domain controller to member server status. Prior to performing this procedure, be sure that you have transferred any operations master roles that are held by the domain controller.
After you remove Active Directory, rename the member server and then reinstall Active Directory on the member server to restore it to domain controller status.
Procedures for Renaming a Domain Controller
Use the following procedures to rename a domain controller. You must perform these procedures directly on the domain controller; they cannot be performed remotely.
Remove Active Directory. This procedure results in the domain controller becoming a member server in the domain.
Run the Active Directory Installation Wizard. This procedure installs Active Directory on the member server to restore it to domain controller status.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Restoring the Original Configuration of a Domain Controller
After you have renamed a member server and returned it to domain controller status, you must restore the original configuration of the domain controller.
If you transferred any domain operations master roles to another domain controller in the domain prior to renaming the domain controller, you can now transfer them back to the renamed domain controller.
If the domain controller was originally configured as a DNS server, you must restore the zone and delegation configurations. The following instructions are based upon best practice recommendations for DNS design, as described in "Best Practice Active Directory Design for Managing Windows Networks" and "Best Practice Active Directory Deployment for Managing Windows Networks" at https://www.microsoft.com/windows/default.mspx. Follow the links under Products to Windows 2000 Server, Technical Resources, Planning & Deployment, Deploying the Windows 2000 Server Family. If your deployment uses a different DNS design, you might not use the delegations and secondary zones described below.
If the domain controller is located in a child domain anywhere in the forest, then you must:
Create a delegation for the domain controller in the forest root domain.
Create a secondary zone.
If the domain controller is located in the forest root domain and the forest root domain has a parent domain, then you must:
- Create a delegation for the new domain controller in the parent domain.
For information about how to configure DNS servers after installing Active Directory, see "Completing Active Directory Installation" in this guide.
Procedures for Restoring the Original Configuration of a Domain Controller
Use the following procedures to restore a domain controller to its original configuration.
Configure the domain controller as a global catalog server, if appropriate.
Transfer the domain operations master roles, if appropriate.
Transfer the forest operations master roles, if appropriate.
Create a delegation for the new domain controller, if appropriate. Perform this procedure in the parent domain of the domain of the DNS server, if one exists.
Create a secondary DNS zone, if appropriate. Perform this procedure only if the DNS server is located in a child domain, not in the forest root domain.
Change the delay for initial notification of an intrasite replication partner, if appropriate.
Configure the domain controller as a preferred bridgehead server, if appropriate.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Managing Global Catalog Servers
Designate global catalog servers in sites to accommodate forest-wide directory searching and so that Active Directory can determine universal group membership of native-mode domain clients.
Global Catalog Placement
To improve the speed of logging on and searching, place at least one global catalog server in each site, and at least two global catalog servers if the site has multiple domain controllers. As a best practice, make half of all domain controllers in a site global catalog servers if the site contains more than three domain controllers. If your deployment uses a single global domain, configure all domain controllers as global catalog servers. In a single-domain forest, configuring all domain controllers as global catalog servers requires no additional resources.
When placing global catalog servers, primary concerns are:
Does any site have no global catalog servers?
What domain controllers are designated as global catalog servers in a particular site?
Initial Global Catalog Replication
When you add a global catalog server to a site, the Knowledge Consistency Checker (KCC) updates the replication topology, after which replication of partial domain directory partitions that are available within the site begins. Replication of partial domain directory partitions that are available only from other sites begins at the next scheduled interval.
Adding subsequent global catalog servers within a site requires only intrasite replication and does not affect network performance. Replication of the global catalog potentially affects network performance only when adding the first global catalog server in the site, and the impact varies depending on the following conditions:
The speed and reliability of the wide area network (WAN) link or links to the site.
The size of the forest.
For example, in a forest that has a large hub site, five domains, and thirty small branch sites (some of which are connected by only dial-up connections), global catalog replication to the small sites takes considerably longer than replication of one or two domains to a few well-connected sites.
Global Catalog Readiness
After replication of the partial domain directory partitions, the domain controller advertises as a global catalog server and begins accepting queries on ports 3268 and 3269. The requirements for advertising as a global catalog server differ in Microsoft Windows 2000 Server Service Pack 3 (SP3) and in Windows 2000 Server Service Pack 2 (SP2). The default requirements in Windows 2000 Server SP3 include replication of all domain directory partitions in the forest. The default requirements in Windows 2000 Server SP2 are limited to replication of the domain directory partitions that are local to the site. If the domain controller advertises as a global catalog server before it has complete information from all domains in the forest, it might return false information to applications that begin using the server for forest-wide searches.
For example, Microsoft Exchange 2000 servers use the global catalog exclusively when looking up addresses. A domain controller that advertises as a global catalog server before it contains all partial directory partitions can cause Address Book lookup and mail delivery problems for Exchange clients. To avoid this problem, ensure that the domain controller does not advertise as global catalog server before it contains all partial domain directory partitions.
Premature advertisement of the global catalog is an issue only for global catalog servers that are running Windows 2000 Server SP2, and only when you add the first global catalog server in a site that does not include all domains. If all domains are represented in the site, or if a global catalog server already exists in the site, then the new global catalog server always has all domains prior to advertising as a global catalog server.
Global Catalog Removal
When you remove the global catalog, the domain controller immediately stops advertising as a global catalog server. The KCC gradually removes the read-only replicas from the domain controller.
Global Catalog Server Management Tasks and Procedures
Table 1.11 shows the tasks and procedures for managing global catalog servers.
Table 1.11 Global Catalog Server Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Identify the global catalog servers in a site. |
|
|
Monthly. |
Identify a site that has no global catalog server. |
|
|
Daily to weekly, depending on environment. |
Add the global catalog to a domain controller and verify global catalog readiness. |
Windows 2000 Server SP2:
Windows 2000 Server SP3:
|
|
As needed. |
Remove the global catalog from a domain controller. |
|
|
As needed. |
Identifying Global Catalog Servers in a Site
Maintain a list of those servers that are designated as global catalog servers. Routinely check these servers to ensure that no one has changed the designation. Check other servers to ensure that no one has erroneously designated a global catalog server.
Procedure for Identifying a Global Catalog Server
Use the following procedure to determine whether a domain controller is a global catalog server. The procedure is explained in detail in the linked topic.
- To determine whether a domain controller is a global catalog server, check the properties on the NTDS Settings object of the respective server object.
Identifying a Site That Has No Global Catalog Servers
To quickly identify a site that has no global catalog servers, you can perform one command rather than check each server individually. You can perform this test any time you add a site, or routinely if global catalog servers can potentially be removed inappropriately.
Procedure for Identifying a Site that has No Global Catalog Servers
Use the following procedure to determine whether a site has a global catalog server. The procedure is explained in detail in the linked topic.
- To identify a site that has no global catalog servers, determine whether the site has at least one global catalog server.
Adding the Global Catalog to a Domain Controller and Verifying Readiness
When conditions in a site warrant adding a global catalog server, you can configure a domain controller to be a global catalog server. Selecting the Global catalog setting on the NTDS Settings object prompts the KCC to update the topology. After the topology is updated, then read-only partial domain directory partitions are replicated to the designated domain controller. When replication must occur between sites to create the global catalog, the site link schedule determines when replication can occur.
Minimum hardware requirements for global catalog servers depend upon the numbers of users in the site. Table 1.12 contains guidelines for assessing hardware requirements.
Table 1.12 Global Catalog Hardware Requirements
Users in site |
Domain controller |
|---|---|
<= 100 |
One uniprocessor PIII 500, 512 MB. |
101 500 |
One uniprocessor PIII 500, 512 MB. |
500 1,000 |
One Dual PIII 500, 1 GB. |
1,001 10,000 |
Two Quad PIII XEON, 2 GB. |
> 10,000 users |
One Quad PIII XEON, 2 GB for every 5,000 users. |
When configuring a global catalog server, be sure the machine has adequate hard disk space. Use the information in Table 1.13 to determine how much storage to provide for the Active Directory database.
Table 1.13 Global Catalog Storage Requirements for the Active Directory Database
Server |
Active Directory database storage requirements |
|---|---|
Domain controller |
0.4 GB of storage for each 1,000 users. |
Global catalog server |
|
For example, in a forest with two 10,000-user domains, all domain controllers need 4 GB of storage. All global catalog servers require 6 GB of storage.
These requirements represent conservative estimates. For a more accurate determination of storage requirements, download and run the Active Directory Sizer Tool (ADSizer.exe). You can download the ADSizer.exe tool from the Active Directory Sizer Tool link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources/.
Occupancy Levels and Global Catalog Server Readiness
The occupancy level setting on a domain controller determines the criteria for advertising itself as a global catalog server in DNS. If a global catalog server advertises itself before it has synchronized all read-only directory partition replicas, clients can receive incorrect information.
The requirements of the occupancy levels are as follows (each higher level includes all levels below it):
0: No occupancy requirement.
1: An inbound connection for at least one read-only directory partition in the site of the global catalog server is added to the designated server by the KCC. Event ID 1264 in the Directory Service log signals creation of the inbound connection.
2: At least one read-only directory partition in the site is replicated to the global catalog server.
3: Inbound connections for all read-only directory partitions in the site are added by the KCC, and at least one is replicated to the server.
4: All read-only directory partitions in the site are replicated to the server.
5: Inbound connections for all read-only directory partitions in the forest are added by the KCC, and all directory partitions in the site are replicated to the server.
6: All directory partitions in the forest are replicated to the server.
Default occupancy levels for domain controllers that are running Windows 2000 Server depend on the Windows 2000 Server service pack release that is installed, as follows:
Windows 2000 Server SP2 or earlier: default and maximum occupancy level = 4.
Windows 2000 Server SP3: default and maximum occupancy level = 6.
Exchange 2000 servers use the global catalog exclusively when looking up addresses. Therefore, in addition to causing Active Directory client search problems, the condition of a global catalog server being advertised before it receives all partial replicas can cause Address Book lookup and mail delivery problems for Exchange clients.
The Name Service Provider Interface (NSPI) must be running on a global catalog server to enable MAPI access to Active Directory. To enable NSPI, you must restart the global catalog server after replication of the partial directory partitions is complete.
Verification of Global Catalog Server Readiness
A global catalog is considered ready to serve clients when the following events occur, in this order:
Occupancy level requirements are met by replicating read-only replicas.
The isGlobalCatalogReady rootDSE attribute is set to TRUE.
The Net Logon service on the domain controller has updated DNS with global-catalog-specific SRV resource records.
At this point, the global catalog server is available to respond to requests on ports 3268 and 3269. However, in response to various tests, the local system can indicate that it is a global catalog server as soon as replication requirements are met, but before DNS has been updated. For a global catalog server that is running Windows 2000 Server SP2, you must also consider the replication requirements for the occupancy level. For the first global catalog server in a site, the occupancy level is significant if all domains are not represented in the site.
Global Catalog Readiness in the SP2 Environment
Because the default occupancy level requirement in Windows 2000 Server SP2 is limited to replicating only the domain directory partitions that are available in the local site, a global catalog server in this environment might advertise itself as ready when other domains are not present on the server. For this reason, when adding the first global catalog to a site where all domains in the forest are not represented, you must take steps to ensure that the global catalog server does not advertise itself before all domain directory partitions are present on the server, as follows:
Prior to configuring the domain controller to be a global catalog server, stop the Net Logon service on the domain controller. If the Net Logon service is not running, then the server cannot update DNS prematurely.
Monitor replication until all domain directory partitions are replicated to the server.
Verify successful replication of all domain directory partitions in the forest.
Restart the domain controller to enable NSPI. Restarting will also start the Net Logon service automatically.
Verify DNS updates.
Global Catalog Readiness in the SP3 Environment
Because the default occupancy level requirement in Windows 2000 Server SP3 is level 6, a new global catalog server does not advertise itself until all partial domain directory partitions in the forest are replicated to the server. In this case, you do not have to stop the Net Logon service before configuring the domain controller as a global catalog server. However, you do need to restart the domain controller to enable NSPI.
Procedures for Adding the Global Catalog to a Domain Controller and Verifying Global Catalog Readiness
Use the following procedures to add a global catalog server to a domain controller. The procedures are explained in detail in the linked topics. Some procedures are performed only when you are configuring the first global catalog server in the site or only when Windows 2000 Server SP2 is running on the domain controller that you are configuring.
Stop the Net Logon service on the domain controller (SP2 only, first global catalog server in the site only).
Configure the domain controller as a global catalog server. Setting the Global Catalog check box initiates the process of replicating all domains to the server.
Monitor global catalog replication progress (first global catalog server in the site only).
Verify successful replication to a domain controller on the global catalog server. Check for inbound replication of all partial domain directory partitions in the forest, to ensure that all domain directory partitions have replicated to the global catalog server.
Verify global catalog readiness. This procedure indicates that the replication requirements have been met.
Restart the Net Logon service, if needed. If you are adding the first global catalog server in a site to a domain controller that is running Windows 2000 Server SP2 and you stopped the Net Logon service prior to adding the global catalog, then restart the service now.
Restart the global catalog server and verify global catalog DNS registrationss by checking DNS for global catalog SRV resource records.
Removing the Global Catalog from a Domain Controller
If the user population of a site decreases to the point where multiple global catalog servers are not required, or if a global catalog server is being replaced with a more powerful machine, then you can remove the global catalog from the domain controller.
The procedure to remove the global catalog is simply to clear the Global Catalog check box on the NTDS Settings object properties page. As soon as you perform this step, the domain controller stops advertising itself as a global catalog server (Net Logon de-registers the global catalog-related records in DNS) and immediately stops accepting LDAP requests over ports 3268 and 3269.
When you remove the global catalog from a domain controller, the KCC begins removing the read-only replicas one at a time by means of an asynchronous process that removes objects gradually over time. Each time the KCC runs (every 15 minutes by default), it attempts the removal of the read-only replica until there are no remaining objects. At an estimated rate of 2000 objects per hour, complete removal of the global catalog from the domain controller can take from several hours to days, depending on the size of the directory.
Procedures for Removing the Global Catalog from a Domain Controller
Use the following procedures to remove the global catalog from a domain controller. The procedures are explained in detail in the linked topics.
Managing Operations Masters
Operations masters keep the directory functioning properly by performing specific tasks that no other domain controllers are permitted to perform. Because operations masters are critical to the long-term performance of the directory, they must be available to all domain controllers and desktop clients that require their services. Careful placement of your operations masters becomes more important as you add more domains and sites to build your forest.
Operations Master Roles
Three operations master roles exist in each domain:
The primary domain controller (PDC) emulator. The PDC emulator processes all replication requests from Microsoft Windows NT 4.0 backup domain controllers and processes all password updates for clients that are not running Active Directoryenabled client software.
The relative identifier (RID) master. The RID master allocates RIDs to all domain controllers to ensure that all security principals have a unique identifier.
The infrastructure master. The infrastructure master for a given domain maintains a list of the security principals from other domains that are members of groups within its domain.
In addition to the three domain-level operation master roles, two operations master roles exist in each forest:
The schema master, which governs all changes to the schema.
The domain naming master, which adds and removes domains to and from the forest.
To perform these functions, the domain controllers hosting these operations master roles must be located in areas where network reliability is high and they need to be consistently available.
Reasons to Move an Operations Master Role
Operations master role holders are placed automatically when the first domain controller in a given domain is created. The three domain-level roles are assigned to the first domain controller created in a domain. The two forest-level roles are assigned to the first domain controller created in a forest.
You might need to move a master operations role to a different domain controller if the service level becomes insufficient, if the domain controller holding the operations master role fails or is decommissioned, or if you make incompatible configuration changes.
Insufficient service level
The PDC emulator is the operations master role that most impacts the performance of a domain controller. For clients that do not run Active Directory client software, the PDC emulator processes requests for password changes, replication, and user authentication. While providing support for these clients, the domain controller continues to perform its normal services, such as authenticating Active Directoryenabled clients. As the network grows, the volume of client requests can increase the workload for the domain controller that hosts the PDC emulator role and its performance can suffer. To solve this problem, you can transfer all or some of the master operation roles to another, more powerful domain controller. You may choose to transfer the role to another domain controller, upgrade the hardware on the original domain controller and then transfer the role back again.
Master operations role holder failure
In the event of a failure, you must decide if you need to relocate the master operations roles to another domain controller or wait for the domain controller to be returned to service. Base that determination on the role that the domain controller hosts and the expected down time.
Decommissioning of the domain controller
Before permanently taking a domain controller offline, transfer any operations master roles that the domain controller holds to another domain controller.
Incompatible configuration changes
Configuration changes to domain controllers or the network topology can result in the need to transfer master operations roles. Except for the infrastructure master, you can assign operations master roles to any domain controller regardless of any other tasks that the domain controller performs. Do not host the infrastructure master role on a domain controller that is also acting as a global catalog server, unless all of the domain controllers in the domain are global catalog servers, or unless only one domain is in the forest. If the domain controller hosting the infrastructure master role is configured to be a global catalog server, you must transfer the infrastructure master role to another domain controller. Changes to the network topology can result in the need to transfer operation master roles in order to keep them in a particular site.
Considerations for Moving Operations Master Roles
You can reassign an operations master role by transfer or, as a last resort, by seizure.
Role transfer
Role transfer is the preferred method to move an operations master role from one domain controller to another. During a role transfer, the two domain controllers replicate to ensure that no information is lost. After the transfer completes, the previous role holder reconfigures itself so that it no longer attempts to perform as the operations master while the new domain controller assumes those duties. This prevents the possibility of duplicate operations masters existing on the network at the same time, which can lead to corruption in the directory.
Role seizure
Seize a role only as a last resort to assign a role to a different domain controller. Use this process only when the previous operations master fails and remains out of service for an extended amount of time. During a role seizure, the domain controller does not verify that replication is updated, so recent changes can be lost. Because the previous role holder is unavailable during the role seizure, it cannot know that a new role holder exists. If the previous role holder comes back online it might still assume that it is the operations master. This can result in duplicate operations master roles on the network, which can lead to corruption of data in the directory and ultimately to the failure of the domain or forest.
To transfer a role to a new domain controller, ensure that the destination domain controller is a direct replication partner of the previous role holder and that replication between them is up to date and functioning properly. This minimizes the time required to complete the role transfer. If replication is sufficiently out of date, the transfer can take a while, but it eventually finishes.
Important: If you must seize an operations master role, never reattach the previous role holder to the network without following the procedures in this guide. Incorrectly reattaching the previous role holder to the network can result in invalid data and corruption of data in the directory.
Guidelines for Role Placement
By improperly placing operations master role holders, you might prevent clients from changing their passwords, or be unable to add domains and new objects, such as users and groups. You might also be unable to make changes to the schema. In addition, name changes might not properly appear within group memberships that are displayed in the user interface.
As your environment changes, you must avoid the problems associated with improperly placed operations master role holders. Eventually, you might need to reassign the roles to other domain controllers.
Although you can assign the forest-level and domain-level operations master roles to any domain controller in the forest and domain respectively, improperly placing the infrastructure master role can cause it to not function properly. Other improper configurations can increase administrative overhead.
Requirements for infrastructure master placement
Do not place the infrastructure master on a domain controller that is also a global catalog server.
The infrastructure master updates the names of security principals from other domains that are added to groups in its own domain. For example, if a user from one domain is a member of a group in a second domain and the user's name is changed in the first domain, then the second domain is not notified that the user's name must be updated in the group's membership list. Because domain controllers in one domain do not replicate security principals to domain controllers in another domain, the second domain never becomes aware of the change. The infrastructure master constantly monitors group memberships, looking for security principals from other domains. If it finds one, it checks with the security principal's domain to verify that the information is updated. If the information is out of date the infrastructure master performs the update and then replicates the change to the other domain controllers in its domain.
Two exceptions apply to this rule. First, if all the domain controllers are global catalog servers, the domain controller that hosts the infrastructure master role is insignificant because global catalogs do replicate the updated information regardless of the domain to which they belong. Second, if the forest has only one domain, the domain controller that hosts the infrastructure master role is not needed because security principals from other domains do not exist.
Recommendations for role placement
Although you can assign the operations master roles to any domain controller, follow these guidelines to minimize administrative overhead and ensure the performance of Active Directory. If a domain controller that is hosting operation master roles fails, following these guidelines also simplifies the recovery process. Guidelines for role placement include:
Leave the two forest-level roles on a domain controller in the forest root domain.
Place the two forest-level roles on a global catalog server.
Place the three domain-level roles on the same domain controller.
Do not place the domain-level roles on a global catalog server.
Place the domain-level roles on a higher performance domain controller.
Adjust the workload of the operations master role holder, if necessary.
Choose an additional domain controller as the standby operations master for the forest-level roles and choose an additional domain controller as the standby for the domain-level roles.
Forest-level role placement in the forest root domain
The first domain controller created in the forest is assigned the schema master and domain naming master roles. To ease administration and backup and restore procedures, leave these roles on the original forest root domain controller. Moving the roles to other domain controllers does not improve performance. Separating the roles creates additional administrative overhead when you must identify the standby operations masters and when you implement a backup and restore policy.
Unlike the PDC emulator role, forest-level roles rarely place a significant burden on the domain controller. Keep these roles together to provide easy, predictable management.
Forest-level role placement on a global catalog server
In addition to hosting the schema master and domain naming master roles, the first domain controller created in a forest also hosts the global catalog. In Windows 2000 Server, you must leave the domain naming master on a global catalog server. When the domain naming master creates an object representing a new domain, it uses the global catalog to ensure that no other object has the same name. The domain naming master achieves this consistency by running on a global catalog server, which contains a partial replica of every object in the forest.
Domain-level role placement on the same domain controller
The three domain-level roles are assigned to the first domain controller created in a new domain. Except for the forest root domain, leave the roles at that location. Keep the roles together unless the workload on your operations master justifies the additional management burden of separating the roles.
For the forest root domain, the first domain controller also hosts the two forest-level roles as well as the global catalog. This additional workload requires you to take two precautionary steps to avoid potential problems. First, the domain-level roles must not remain on a global catalog server. In addition, because hosting all five roles on a single domain controller can overload the server and hurt performance, transfer the three domain-level roles to another domain controller.
Because all pre-Active Directory clients submit updates to the PDC emulator, the domain controller holding that role uses a higher number of RIDs. Place the PDC emulator and RID master roles on the same domain controller so these two roles interact more efficiently.
If you must separate the roles, you can still use a single standby operations master for all three roles. However, you must ensure that the standby is a replication partner of all three of the role holders.
Backup and restore procedures also become more complex if you separate the roles. Special care must be taken to restore a domain controller that hosted an operations master role. By hosting the roles on a single computer, you minimize the steps that are required to restore a role holder.
Domain-level role absence on a global catalog server
Do not host the infrastructure master on a domain controller that is acting as a global catalog server. Because it is best to keep the three domain-level roles together, avoid putting any of them on a global catalog server.
Domain-level role placement on a higher performance domain controller
Host the PDC emulator role on a powerful and reliable domain controller to ensure that it is available and capable of handling the workload. Of all the operations master roles, the PDC emulator creates the most overhead on the server that is hosting the role. It has the most intensive daily interaction with other systems on the network. The PDC emulator has the greatest potential to affect daily operations of the directory.
Workload adjustment of the operations master role holder
Domain controllers can become overloaded while attempting to service client requests on the network, manage their own resources, and handle any specialized tasks such as performing the various operations master roles. This is especially true of the domain controller holding the PDC emulator role. Pre-Active Directory clients and domain controllers running Windows NT 4.0 rely more heavily on the PDC emulator than Active Directory clients and Windows 2000 Server domain controllers. If your networking environment has pre-Active Directory clients and domain controllers, you might need to reduce the workload of the PDC emulator.
If a domain controller begins to indicate that it is overloaded and the performance is affected, you can reconfigure the environment so that some tasks are performed by other, less-used domain controllers. By adjusting the domain controller's weight in the DNS environment, you can configure the domain controller to receive fewer client requests than other domain controllers on your network. Optionally, you can adjust the domain controller's priority in the DNS environment so it processes client requests only if other DNS servers are unavailable. With fewer DNS client requests to process, the domain controller can use more resources to perform operations master services for the domain.
Standby operations master
The standby operations master is a domain controller that you identify as the computer that assumes the operations master role if the original computer fails. You do not need to perform any special configuration steps or run any type of setup utilities to make a domain controller a standby operations master. This precautionary planning step helps make your operation more resilient if a problem arises that requires you to reassign a master operations role to a new domain controller.
Ensure that the standby operations master is a direct replication partner of the actual operations master. If the standby operations master domain controller is a direct replication partner of the original operations master, it most likely contains the most recent changes to the domain. This reduces the time required to transfer the role to the standby operations master and, in the case of a failure, reduces the chances of losing information. Even if replication is not totally complete, only few outstanding updates exist. Those outstanding updates can be replicated by a normal replication cycle rather than requiring a full synchronization, which replicates all of the account information in the partition. To guarantee that the two domain controllers are replication partners, you must manually create a connection object between them. Although creating manual connection objects is not generally recommended, in this one case it is necessary because it is so important that these two domain controllers be replication partners.
If you must reassign the domain-level operations master roles to the standby operations master, do not place the infrastructure master role on a global catalog server.
Ramifications of Role Seizure
If a role is seized, the new role holder is configured to host the operations master role with the assumption that you do not intend to return the previous role holder to service. Use role seizure only when the previous role holder is not available and you need the operations master role to keep the directory functioning. Because the previous role holder is not available during a seizure, you cannot reconfigure the previous role holder and inform it that another domain controller is now hosting the operations master role.
If the previous role holder comes back online, its behavior depends on your current service pack level. If you are running Windows 2000 Server Service Pack 2 (SP2) or earlier, the domain controller waits for one replication cycle while it attempts to verify the current role holder. If the previous role holder receives data that indicates that another domain controller is performing the operations master role, it reconfigures itself so that it no longer hosts the operations master role and Active Directory functions properly. If for any reason replication fails, it does not receive any replicated data indicating that a new operations master exists. Whether or not replication actually occurs, after one replication cycle it assumes that the data it has is correct. It leaves itself configured as the current operations master and attempts to resume its duties as the operations master role holder. This results in duplicate operations masters on the network. As shown in Table 1.14, this can cause serious problems in the directory.
If you are running Windows 2000 Server Service Pack 3 (SP3), the previous role holder waits for a full replication cycle to complete successfully before it resumes the role of operations master. By waiting for a full replication cycle, it can see if another operations master exists before it brings itself back online. If the previous role holder detects that another operations master exists, it reconfigures itself so that it no longer hosts the roles in question.
To reduce risk, perform a role seizure only if the missing operations master role unacceptably affects performance of the directory. Calculate the effect by comparing the impact of the missing service provided by the operations master to the amount of work that is needed to bring the previous role holder safely back online after you perform the role seizure.
Active Directory continues to function when the operations master roles are not available. If the role holder is only offline for a short period, you might not need to seize the role to a new domain controller. Remember that returning an operation master to service after the role is seized can have dire consequences if it is not done properly.
Table 1.14 Operations Master Role Functionality Risk Assessment
Operations Master Role |
Consequences if Role is Unavailable |
Risk of Improper Restoration |
Recommendation for Returning to Service After Seizure |
|---|---|---|---|
Schema master |
You cannot make changes to the schema. |
Conflicting changes can be introduced to the schema if both schema masters attempt to modify the schema at the same time. This can result in a fragmented schema. |
Not recommended. Can lead to a corrupted forest and require rebuilding the entire forest. |
Domain naming master |
You cannot add or remove domains from the forest. |
You cannot add or remove domains or clean-up metadata. Domains might appear as though they are still in the forest even though they are not. |
Not recommended. Can require rebuilding domains. |
PDC emulator |
You cannot change passwords on pre-Active Directory clients. No replication to Windows NT 4.0 backup domain controllers. |
Password validation can randomly pass or fail. Password changes take much longer to replicate throughout the domain. |
Allowed. User authentication can be erratic for a time, but no permanent damage occurs. |
Infrastructure master |
Delays displaying updated group membership lists in the user interface when you move users from one group to another. |
Displays incorrect user names in group membership lists in the user interface after you move users from one group to another. |
Allowed. May impact the performance of the domain controller hosting the role, but no damage occurs to the directory. |
RID master |
Eventually, domain controllers cannot create new directory objects as each of their individual RID pools is depleted. |
Duplicate RID pools can be allocated to domain controllers, resulting in data corruption in the directory. This can lead to security risks and unauthorized access. |
Not recommended. Can lead to data corruption that can require rebuilding the domain. |
Operations Master Role Management Tasks and Procedures
Table 1.15 shows the tasks and procedures for managing operations master roles.
Table 1.15 Operations Master Role Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Designate operations master roles. |
|
|
As needed |
Reduce the workload on the PDC emulator. |
|
|
As needed |
Decommission a role holder. |
|
|
As needed |
Seize operations master roles. |
|
|
As needed |
Choose a standby operations master. |
|
|
As needed |
Designating Operations Master Roles
When you create a new domain, the Active Directory Installation Wizard automatically assigns all of the domain-level operations master roles to the first domain controller that is created in that domain. When you create a new forest, the wizard also assigns the two forest-level operations master roles to the first domain controller. After the domain is created and functioning, you might transfer various operations master roles to different domain controllers to optimize performance and simplify administration.
Transferring the forest-level and domain-level operations master roles is performed as needed and governed by the guidelines for placing operations master roles. Before you transfer an operations master role, use Repadmin.exe with the /showreps option to ensure that replication between the current role holder and the domain controller assuming the role is updated.
In addition, you must determine if the domain controller that you intend to assume an operations master role is a global catalog server. The domain naming master, a forest-level role, must also host the global catalog. However, the infrastructure master for each domain must not host the global catalog.
Do not change the global catalog configuration on the domain controller that you intend to assume an operations master role unless your IT management authorizes that change. Changing the global catalog configuration can cause changes that can take days to complete and the domain controller might not be available during that period. Instead, transfer the operations master roles to a different domain controller that is already properly configured.
Procedures for Designating Operations Master Roles
Procedures are explained in detail in the linked topics.
Reducing the Workload on the PDC Emulator
You can configure a domain controller so that DNS sends the majority of client requests to other domain controllers. Reducing the number of client requests helps reduce the workload on a domain controller, giving it more time to function as an operations master, and is especially important for the PDC emulator. Of all the operations master roles, the PDC role has the highest impact on the domain controller hosting that role. You might need to take steps to keep that domain controller from becoming overloaded.
To receive information from the domain, a client uses DNS to locate a domain controller and then sends the request to that domain controller. By default, DNS performs rudimentary load balancing and randomizes the distribution of client requests so they are not always sent to the same domain controller. If too many client requests are sent to a domain controller while it attempts to perform other duties, such as those of the PDC emulator, it can become overloaded, which has a negative impact on performance. To reduce the number of client requests that are processed by the PDC emulator, you can adjust its weight in the DNS environment or you can adjust its priority in the DNS environment.
DNS Weight Registry Setting
Adjusting the weight of a domain controller to less than other domain controllers reduces the number of clients that DNS refers to that domain controller. The default weight for all domain controllers is 100. By reducing this value, DNS refers clients to a domain controller less frequently based on the proportion of this value to the value of other domain controllers. For example, to configure the system so that the domain controller hosting the PDC emulator role receives requests only half as many times as the other domain controllers, configure the weight of the domain controller hosting the PDC emulator role to be 50. DNS determines the weight ratio for that domain controller to be 50/100 (50 for that domain controller and 100 for the other domain controllers). After you reduce this ratio to 1/2, DNS refers clients to the other domain controllers twice as often as it refers to the domain controller with the reduced weight setting. By reducing client referrals, the domain controller receives fewer client requests and has more resources for other tasks, such as performing the role of PDC emulator.
DNS Priority Registry Setting
Adjusting the priority of the domain controller also reduces the number of client referrals. However, rather than reducing it proportionally to the other domain controllers, changing the priority causes DNS to stop referring all clients to this domain controller unless all domain controllers with a lower priority setting are unavailable.
To configure the PDC emulator in this manner, use Regedit.exe to modify the ldapsrvpriority or ldapsrvweight registry entries.
Procedures for Reducing the Number of Client Requests Processed by the PDC Emulator
Procedures are explained in detail in the linked topics.
Decommissioning a Role Holder
When you use the Active Directory Installation Wizard to decommission a domain controller that currently hosts one or more operations master roles, the wizard reassigns the roles to a different domain controller. When the wizard is run, it determines whether the domain controller currently hosts any operations master roles. If it detects any operations master roles, it queries the directory for other eligible domain controllers and transfers the roles to a new domain controller. A domain controller is eligible to host the domain-level roles if it is a member of the same domain. A domain controller is eligible to host a forest-level role if it is a member of the same forest.
You cannot control which domain controller the wizard chooses and the wizard does not indicate which domain controller receives the roles. Because of this behavior, it is best to transfer the roles prior to running the wizard. That way you can control role placement and can transfer the roles according to the recommendations discussed earlier in this guide.
Transfer to the Operations Master Standby
Transfer the operations master roles to the standby operations master. By following the recommendations for operations master role placement, the standby operations master is a direct replication partner and is ready to assume the roles. Remember to designate a new standby for the domain controller that assumes the roles.
Transfer when No Standby Operations Master is Ready
If you do not follow the recommendations for role placement and you have not designated a standby operations master, you must properly prepare a domain controller to which you intend to transfer the operations master roles. Preparing the future role holder is the same process as preparing a standby operations master. You must manually create a connection object to ensure that it is a replication partner with the current role holder and that replication between the two domain controllers is updated. To determine whether the standby domain controller received the latest replicated updates from the current operations master, use Repadmin.exe with the /showreps option.
In addition, you must determine whether the domain controller intended to assume an operations master role is a global catalog server. The domain naming master, a forest-level role, must also host the global catalog. However, the infrastructure master for each domain must not host the global catalog.
Do not change the global catalog configuration on the domain controller that you intend to assume an operations master role unless your IT management authorizes that change. Changing the global catalog configuration can cause changes that can take days to complete and the domain controller might not be available during that period. Instead, transfer the operations master roles to a different domain controller that is already properly configured.
Procedures for Decommissioning a Role Holder
Procedures are explained in detail in the linked topics.
Seizing Operations Master Roles
Seize an operations master role only as a last resort. If at all possible, transfer an operations master role to a new domain controller instead. Seize an operations master role only if the current role owner is offline and is unlikely to return to service.
Role seizure is the act of assigning an operations master role to a new domain controller without the cooperation of the current role holder (usually because it is offline due to a hardware failure). During role seizure, a new domain controller assumes the operations master role without communicating with the current role holder.
Role seizure can create two conditions that can cause problems in the directory. First, the new role holder starts performing its duties based on the data located in its current directory partition. The new role holder might not receive changes that were made to the previous role holder before it went offline if replication did not complete prior to the time when the original role holder went offline. This can cause data loss or data inconsistency into the directory database.
To minimize the risk of losing data to incomplete replication, do not perform a role seizure until enough time has passed to complete at least one complete end-to-end replication cycle across your network. Allowing enough time for complete end-to-end replication ensures that the domain controller that assumes the role is as up-to-date as possible.
Second, the original role holder is not informed that it is no longer the operations master role holder, which is not a problem if the original role holder stays offline. However, if it comes back online (for example, if the hardware is repaired or the server is restored from a backup), it might try to perform the operations master role that it previously owned. This can result in two domain controllers performing the same operations master role simultaneously. Depending on the role in question and whether your environment runs Windows 2000 Server SP2 or Windows 2000 Server SP3, this can disrupt the directory service. For example, a RID master might reallocate a duplicate RID pool resulting in corruption of data in the directory. The severity of duplicate operations master roles varies from no visible effect to the need to rebuild the entire forest. For more information about the risks of returning an operations master to service after the role is seized to another domain controller, see "Ramifications of Role Seizure" earlier in this guide.
If you are seizing a role and you have not designated another domain controller as the standby operations master, you can use Repadmin.exe with the /showreps option to identify a domain controller that has the most recent updates from the current role holder. Seize the operations master role to that domain controller to minimize the impact of the role seizure.
Procedures for Seizing Operations Master Roles
Procedures are explained in detail in the linked topics.
Verify that a complete end-to-end replication cycle has occurred. During the design process, you calculated the maximum end-to-end replication latency. The maximum end-to-end replication latency is the maximum amount of time it should take for replication to take place between the two domain controllers in your enterprise that are farthest from each other based on the topology of your network. If you verify that replication is functioning properly and wait this amount of time without making any additional changes to the directory then you can assume that all changes have been replicated and the domain controller is up to date.
Verify successful replication to a domain controller (the domain controller that will be seizing the role).
Choosing a Standby Operations Master
A single domain controller can act as the standby operations master for all of the operations master roles in a domain, or you can designate a separate standby for each operations master role. Following the recommendations, it is best to select one standby for the forest-level roles and another standby in each domain that can be used to host the three domain-level roles if their host fails.
No utilities or special steps are required to designate a domain controller as a standby operations master. However, the current operations master and the standby should be well connected. This means that the network connection between them must support at least 10 megabit transmission rate and be available at all times. In addition, configure the current role holder and the standby as direct replication partners by manually creating a connection object between them.
Configuring a replication partner can save some time if you must reassign any operations master roles to the standby operations master. Before transferring a role from the current role holder to the standby operations master, ensure that replication between the two computers is functioning properly. Because they are replication partners, the new operations master is as updated as the original operations master, thus reducing the time required for the transfer operation. To determine whether the standby domain controller received the latest replicated updates from the current operations master, use Repadmin.exe with the /showreps option.
During role transfer, the two domain controllers exchange any unreplicated information to ensure that no transactions are lost. If the two domain controllers are not direct replication partners, a substantial amount of information might need to be replicated before the domain controllers completely synchronize with each other. The role transfer requires extra time to replicate the outstanding transactions. If the two domain controllers are direct replication partners, fewer outstanding transactions exist and the role transfer operation completes sooner.
Designating a domain controller as a standby also minimizes the risk of role seizure. By making the operations master and the standby direct replication partners, you reduce the chance of data loss in the event of a role seizure, thereby reducing the chances of introducing corruption into the directory.
When you designate a domain controller as the standby, follow all recommendations that are discussed in "Guidelines for Role Placement" earlier in this guide. To designate a standby for the forest-level roles, choose a global catalog server so it can interact more efficiently with the domain naming master. To designate a standby for the domain-level roles, ensure that the domain controller is not a global catalog server so that the infrastructure master continues to function properly if you must transfer the roles.
Manually create a connection object between the operations master and the designated standby operations master to ensure that replication occurs between the two domain controllers.
Procedures for Choosing a Standby Operations Master
Procedures are explained in detail in the linked topics.
Managing the Database
Active Directory is stored in the Ntds.dit database file. In addition to this file, the directory uses log files, which store transactions prior to committing them to the database file. For best performance, store the log files and the database on separate hard drives.
The directory database is a self-maintained system. Other than regular backup, the directory database requires no daily maintenance during ordinary operation. However, you might need to manage the following conditions:
Low disk space: Monitor free disk space on the partition or partitions that store the directory database and logs. Provide warnings at the following logical-disk-space thresholds:
Ntds.dit partition: The greater of 20 percent of the Ntds.dit file size or 500 megabytes (MB).
Log file partition: The greater of 20 percent of the combined log files size or 500 MB.
Ntds.dit and logs on the same volume: The greater of 1 gigabyte (GB) or 20 percent of the combined Ntds.dit and log files sizes.
Note: If you also set an alert threshold, divide the above warning thresholds in half.
Database size: During ordinary operation, the database removes expired tombstones and defragments (consolidates) white space. This automatic online defragmentation redistributes and retains white space for use by the database. The following conditions might warrant taking steps to regulate database size manually:
Temporary backlog of expired tombstones following bulk deletions: Large-scale deletions can temporarily increase the database file size if tombstones expire in larger numbers than garbage collection can remove in one cycle (5,000 tombstones per cycle). After objects are deleted, their tombstones are stored in the directory for 60 days by default and cannot be removed prior to that time. However, after the tombstone lifetime expires, you can speed removal of the tombstone backlog by temporarily decreasing the default garbage collection period (12 hours).
Increased white space due to large-scale deletions: If data is decreased significantly, such as when the global catalog is removed from a domain controller, white space is not automatically returned to the file system. Although this condition does not affect database operation, it does result in a larger file size. You can use offline defragmentation to decrease the size of the database file by returning white space from the database file to the file system.
Hardware upgrade or failure: If you need to upgrade or replace the disk on which the database or log files are stored, move the files to a different location, either permanently or temporarily.
For information about monitoring the database and log file partitions for low disk space, see "Monitoring Active Directory" earlier in this guide.
General Guidelines for Directory Database Management
For all database management tasks, follow these guidelines:
Prior to performing any procedures that affect the directory database, be sure that you have a current system state backup. For information about performing system state backup, see "Active Directory Backup and Restore" earlier in this guide.
To manage the database file itself, you must take the domain controller offline by restarting in Directory Services Restore Mode, and then use Ntdsutil.exe to manage the file.
To start a domain controller in Directory Services Restore Mode, you must log on to the domain controller as the local Administrator. To remotely manage the database, you can use Terminal Services Client to restart the domain controller in Directory Services Restore Mode.
NTFS Disk Compression is not supported for the database and log files.
Directory Database Management Tasks and Procedures
Table 1.16 shows the tasks and the procedures for managing the database.
Table 1.16 Directory Database Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Relocate directory database files. |
|
|
As needed. |
Return unused disk space from the directory database to the file system. |
|
|
As needed. |
Speed removal of an expired-tombstone backlog. |
|
|
As needed. |
Relocating Directory Database Files
The following conditions require moving database files:
Hardware maintenance: If the physical disk on which the database or log files are stored requires upgrading or maintenance, the database files must be moved, either temporarily or permanently.
Low disk space: When free disk space is low on the logical drive that stores the database file (Ntds.dit), the log files, or both, first verify that no other files are causing the problem. If the database file or log files are the cause of the growth, then provide more disk space by taking one of the following actions:
Expand the partition on the disk that currently stores the database file, the log files, or both. This procedure does not change the path to the files and does not require updating the registry.
Use Ntdsutil.exe to move the database file, the log files, or both to a larger existing partition. Moving files to a different partition changes the path to the files and therefore requires updating the registry. Ntdsutil.exe automatically updates the registry when you use it to move database files.
Path Considerations
If the path to the database file or log files changes as a result of moving the files, be sure that you:
Use Ntdsutil.exe to move the files (rather than copying them) so that the registry is updated with the new path. Even if you are moving the files only temporarily, use Ntdsutil.exe to move files locally so that the registry is always current.
Perform a system state backup as soon as the move is complete so that the restore procedure uses the correct path.
Verify that the correct permissions are applied on the destination folder following the move. Revise permissions to those that are required to protect the database files, if needed.
SYSVOL Considerations
If you replace or reconfigure a drive that stores the SYSVOL folder, you must first move the SYSVOL folder manually. For information about moving SYSVOL manually, see "Managing SYSVOL" later in this guide.
Procedures for Relocating Directory Database Files
Use the following procedures to move or copy the database file, the log files, or both. Procedures are explained in detail in the linked topics.
Determine the location and size of the directory database files. Use the database size to prepare a destination location of the appropriate size. Track the respective file sizes during the move to ensure that you successfully move the correct files. Be sure to use the same method to check file sizes when you compare them. The size is reported differently, depending on whether the domain controller is online or offline, as follows:
Determine the database size and location online. This size is reported in bytes.
Determine the database size and location offline. This size is reported in megabytes (MB). Use this method if the domain controller is already started in Directory Services Restore Mode.
Compare the size of the directory database files to the volume size. Before moving any files in response to low disk space, verify that no other files on the volume are responsible for the condition of low disk space.
Back up system state. System state includes the database file and log files as well as SYSVOL and NETLOGON shared folders, among other things. Always ensure that you have a current backup prior to moving database files.
Restart the domain controller in Directory Services Restore Mode, as follows:
If you are logged on to the domain controller console, locally restart the domain controller in Directory Services Restore Mode.
If you are using Terminal Services for remote administration, modify the Boot.ini file on the remote server so that you can remotely restart the domain controller in Directory Services Restore Mode.
Move the database file, the log files, or both. Move the files to a temporary destination if you need to reformat the original location, or to a permanent location if you have additional disk space. Moving the files can be performed locally by using Ntdsutil.exe or remotely (temporarily) by using a file copy, as follows:
Copy the directory database files to a remote share and back. When copying any database files off the local computer, always copy both the database file and the log files.
If the path to the database or log files has changed, back up system state so that the restore procedure has the correct information.
Returning Unused Disk Space from the Directory Database to the File System
During ordinary operation, the white space in the directory database file becomes fragmented. Each time garbage collection runs (every 12 hours by default), white space is automatically defragmented online to optimize its use within the database file. The unused disk space is thereby maintained for the database; it is not returned to the file system.
Only offline defragmentation can return unused disk space from the directory database to the file system. When database contents have decreased considerably through a bulk deletion (for example, you remove the global catalog from a domain controller), if the size of the database backup is significantly increased due to the white space, use offline defragmentation to reduce the size of the Ntds.dit file.
You can determine how much free disk space is recoverable from the Ntds.dit file by setting the Garbage Collection logging level in the registry. Changing the Garbage Collection logging level from the default value of 0 to a value of 1 results in event ID 1646 being logged in the Directory Service log. This event describes the total amount of disk space used by the database file as well as the amount of free disk space that is recoverable from the Ntds.dit file through offline defragmentation.
At Garbage Collection logging level 0, only critical events and error events are logged in the Directory Service log. At level 1, high-level events are logged as well. Events can include one message for each major task that is performed by the service. At level 1, the following events are logged for garbage collection:
700 and 701: report when online defragmentation begins and ends, respectively.
1646: reports the amount of free space available in the database out of the amount of allocated space.
Caution: Setting the value of entries in the Diagnostics subkey to greater than 3 can degrade server performance and is not recommended.
Following offline defragmentation, perform a database integrity check. The integrity command in Ntdsutil.exe detects binary-level database corruption by reading every byte in the database file. The process ensures that the correct headers exist in the database itself and that all of the tables are functioning and consistent. Therefore, depending upon the size of your Ntds.dit file and the domain controller hardware, the process might take considerable time. In testing environments, the speed of 2 GB per hour is considered to be typical. When you run the command, an online graph displays the percentage completed.
Procedures for Performing Offline Defragmentation
Use the following procedures to perform offline defragmentation. Procedures are explained in detail in the linked topics.
Change the garbage collection logging level to 1. Check the Directory Service event log for event ID 1646, which reports the amount of disk space that you can recover by performing offline defragmentation.
Back up system state. System state includes the database file and database log files as well as SYSVOL, NETLOGON, and the registry, among other things. Always ensure that a current backup exists prior to defragmenting database files.
Take the domain controller offline, as follows:
If you are logged on to the domain controller locally, restart the domain controller in Directory Services Restore Mode.
If you are using Terminal Services for remote administration, you can remotely restart the domain controller in Directory Services Restore Mode after modifying the Boot.ini file on the remote server.
Compact the directory database file (offline defragmentation). As part of the offline defragmentation procedure, check directory database integrity.
If database integrity check fails, perform semantic database analysis with fixup.
Speeding Removal of an Expired-Tombstone Backlog
An object that is deleted from Active Directory is stored as a tombstone, which represents the deleted object in the directory so that the deletion is replicated. Tombstones remain in the directory for a default period of 60 days from the time of deletion, at which point they expire and are permanently removed by garbage collection.
Note: Tombstones cannot be removed prior to expiration of the tombstone lifetime.
Although tombstones use less space than the full object, they can affect the size of the database temporarily following large bulk deletions. A maximum of 5,000 expired tombstones can be deleted at one time. If the number of expired tombstones exceeds 5,000, more than one garbage collection interval is required to clear the backlog. During the backlog, tombstones that are no longer needed are retained, consuming database space.
Increased Rate of Tombstone Removal
The default garbage collection period is 12 hours. Temporarily decreasing the garbage collection period (for example, to 1 hour) can help speed the removal of expired tombstones. However, setting this period too low can also cause slow performance, so be sure to return the value to the original setting as soon as the backlog is cleared. To reduce database size by returning the white space left by the removed tombstones to the file system, perform offline defragmentation after the backlog is cleared.
Logging of Tombstone Removal
The default logging level for garbage collection is 0. At this level, only errors are reported. When garbage collection logging is set to 3, event ID 1006 reports the number of expired tombstones removed during each garbage collection cycle.
If you want to track removal of expired tombstones, increase the logging level to 3 and decrease the garbage collection period until the backlog is cleared, and then return the logging level and the garbage collection period to normal.
Procedures for Regulating Directory Database Growth Caused by Tombstones
Use the following procedures to manage removal of tombstones following bulk deletions.
Change the garbage collection period to a lower interval. Decreasing the interval between garbage collections helps the system eliminate the tombstone backlog more quickly.
Change the garbage collection logging level to 3. Increasing the logging level to 3 causes an event that reports the number of tombstones removed each time garbage collection occurs.
Verify removal of tombstones in the event log. Check the Directory Service event log for NTDS event ID 1006, which reports the number of expired tombstones removed. When this event indicates that the number of tombstones removed is less than 5,000, the backlog has been cleared.
Change the garbage collection period. When the event ID 1006 reports a number of removed tombstones less than 5,000, you can return the interval between garbage collections to the normal level.
Change the garbage collection logging level, if needed. If you no longer want informational events logged for garbage collection, return the logging level to 0.
Compact the directory database file (offline defragmentation), if needed. Clearing the backlog does not remove the white space created by the tombstones. Only offline defragmentation returns unused disk space to the file system.
Managing SYSVOL
The Windows 2000 Server System Volume (SYSVOL) is a collection of folders and reparse points in the file systems that exist on each domain controller in a domain. SYSVOL provides a standard location to store Group Policy objects (GPOs) and scripts so that the File Replication service (FRS) can distribute them to other domain controllers and member computers in a domain.
FRS monitors SYSVOL and if a change occurs to any file stored on SYSVOL, then FRS automatically replicates the changed file to the SYSVOL folders on the other domain controllers in the domain.
Computers that run Windows 2000 Server obtain GPOs, logon, logoff, startup, and shutdown scripts from the SYSVOL shared folder. Windows NT 4.0based domain controllers and Windows-based clients that do not run Active Directory client software obtain GPOs and scripts from the NETLOGON shared folder.
During the installation of Active Directory, the folders and reparse points are automatically created in the %SystemRoot%/SYSVOL folder. FRS automatically replicates any files or GPOs that are written to these folders to the other domain controllers in the domain, to ensure that they are available and ready to be used when a user logs on to the domain.
The day-to-day operation of SYSVOL is an automated process that does not require any human intervention other than watching for alerts from the monitoring system. Occasionally, you might perform some system maintenance as you change your network. The procedures you might perform include:
Relocating SYSVOL
Relocating the Staging Area
Changing the size of the Staging Area
These procedures involve moving SYSVOL or portions of SYSVOL to alternate locations. You might perform these procedures to maintain capacity and performance of SYSVOL, for hardware maintenance, or for data organization.
Capacity
Depending upon the configuration of your network, SYSVOL can require much disk space to function properly. During the initial deployment, SYSVOL might be allocated adequate disk space to function. However, as your network grows, the required capacity can exceed the available disk space.
Note: If you receive indications that disk space is low, determine if the cause is inadequate physical space on the disk, or a registry setting that allocates inadequate disk space to SYSVOL. By modifying a setting in the registry, you can allocate more disk space to SYSVOL rather than relocating SYSVOL or the Staging Area. Increasing the space allocation in the registry is much faster and easier than relocation. For more information about managing disk space, see "Maintaining Sufficient Disk Space" later in this section.
Performance
Any changes made to SYSVOL are automatically replicated to the other domain controllers in the domain. If the files stored in SYSVOL change frequently, the replication increases the input and output for the volume where SYSVOL is located. If the volume is also host to other system files, such as the directory database or the pagefile, then the increased input and output for the volume can impact the performance of the server.
Hardware Maintenance
System maintenance, such as removal of a disk drive, can require you to relocate SYSVOL. Even if the maintenance occurs on a different disk drive, verify that that maintenance does not affect the system volume. Logical drive letters can change after you add and remove disks. FRS locates SYSVOL by using pointers stored in the directory and the registry. If drive letters change after you add or remove disk drives, be aware that these pointers are not automatically updated.
Data Organization
Some organizations prefer to control where specific data is stored for organizational purposes and established backup and restore policies.
Guidelines for Managing SYSVOL
To manage SYSVOL, ensure that FRS properly replicates the SYSVOL data, and provide enough space to store SYSVOL. Implement a monitoring system that can detect low disk space and potential FRS disruptions so that you can address those issues before the system stops replicating. For more information about monitoring SYSVOL, see "Monitoring Active Directory" in this guide.
Disk space maintenance
SYSVOL stores and replicates GPOs, Distributed File System (DFS) information, and scripts. As the network grows, SYSVOL can begin to require substantial storage space. Although you do plan for storage requirements for SYSVOL during the planning stages of deployment, you might need to adjust the storage requirements after you deploy additional domain controllers due to network growth and the way in which FRS replicates files.
FRS replicates files by making a temporary copy of the files in a Staging Area folder and then sending the copies to replication partners. This method avoids problems that locked files can cause while replication occurs. Because FRS replicates copies of the files, the original files remain available for user access during replication. However, this method requires making a copy of every file prior to replication. Based on the size and number of files involved, a substantial amount of disk space might be required for temporary storage.
When the Staging Area folder runs out of disk space, FRS behaves differently depending on the version of Windows 2000 that is running. If Windows 2000 Server Service Pack 2 (SP2) or earlier is running, then the system will stop replicating until space is made available. If Windows 2000 Server Service Pack 3 (SP3) is running, then FRS will detect when it is about to run out of disk space and start removing the least recently used files to provide more space. Although this prevents the system from halting replication, it does increase input and output for the server's disk and can impact performance. For more information about the changes to FRS from Windows 2000 Server SP2 to Windows 2000 Server SP3, see KB article 321557 in the Microsoft Knowledge Base. To view the Microsoft Knowledge Base, see the Microsoft Knowledge Base link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources.
Both FRS and DFS use the Staging Area folder. To maintain sufficient disk space for SYSVOL, estimate the amount of space that DFS uses as well as the space that FRS uses. For more information about DFS, see "Distributed File System" in the Distributed Systems Guide of the Windows 2000 Server Resource Kit.
Because the Staging Area folder holds files from all replication partners, you must consider traffic to and from all partners when you estimate the disk space requirements for the Staging Area folder on each computer.
If replication must occur between domain controllers that are located in different sites, remember that FRS uses the same connection objects as Active Directory. You can configure those connection objects so that replication can occur only during certain times of the day. Each connection object has an associated schedule that dictates what hours of the day the connection is available for replication. Allocate enough time in the schedule for all Active Directory replication and all FRS replication to occur. If FRS does not complete all outstanding replication requests when the schedule makes the connection available, it will hold the remaining unreplicated files until the next time the connection becomes available. Over time, this backlog of unreplicated files can grow to consume an enormous amount of disk space.
Additional SYSVOL recommendations
You can preserve Staging Area and bandwidth usage by following these best practices:
Run Windows 2000 SP2 on all domain controllers that run FRS. Install Windows 2000 SP3 as soon as possible.
Always keep FRS service running, especially when you make bulk changes to FRS-replicated files or files outside the tree on the same drive.
Do not run anti-virus software against FRS-replicated directories.
Do not enable File System Group Policy on any FRS-replicated tree.
Watch for inconsistent directories. Duplicate folders that appear in the FRS replication tree on multiple domain controllers can cause inconsistent directories. Although this is not a critical problem, it can result in unanticipated behavior, such as changes appearing to be lost. If this occurs, examine the files in these directories to determine which directory is the proper version and then delete the duplicated directories from the tree.
Do not leave files open for extended amounts of time. FRS cannot replicate a file while it is open. Avoid using elements in scripts that cause a file to be open for an extended amount of time, such as a script that waits for user input before proceeding. If the user is not present when the script runs, the file can remain open and cannot be replicated until the script terminates.
Do not attempt to relocate SYSVOL or the Staging Area if the FRS environment on your network is unstable and you are having problems with system volumes becoming unsynchronized among replication partners. Troubleshoot the FRS problems and ensure that the environment is stable before attempting any relocation operations. During all relocation operations except authoritative restore, FRS rebuilds the SYSVOL content by replicating data from its replication partners. If FRS is not functioning properly on the partners, their SYSVOL data may be invalid. This can result in invalid SYSVOL data in the new location. The relocation operation can also fail because FRS cannot replicate the necessary data from the domain controller's replication partners.
SYSVOL and Staging Area relocation
Deployment is the best time to determine the location of SYSVOL. Consider performance and disk capacity to determine the best location for the SYSVOL folders. During the Active Directory installation, you must specify the location of the SYSVOL folders. After installation, you might need to relocate SYSVOL or the Staging Area folder.
Relocating only the Staging Area
Although SYSVOL contains many folders, the Staging Area requires the most capacity because it is used for replication. You can leave SYSVOL in its original location and relocate only the Staging Area.
Relocating SYSVOL and the Staging Area
You can relocate the entire SYSVOL folder and its associated subtrees, including the Staging Area.
You can relocate SYSVOL by removing and reinstalling Active Directory on the domain controller or by manually recreating SYSVOL at a new location.
Active Directory removal and reinstallation
To relocate SYSVOL, removing and reinstalling Active Directory is far easier and more reliable than manually recreating SYSVOL at a new location, but it can also be impractical. To relocate SYSVOL by using this method, you use the Active Directory Installation Wizard to remove Active Directory from the domain controller then use it again to reinstall Active Directory on the same domain controller. During the reinstallation, provide the new location for SYSVOL. The replication process populates the folders with the appropriate files from another domain controller. This method might not be practical to use because having a large number of objects in your directory increases the required time for reinstallation and you might need to reinstall and reconfigure other services if the domain controller runs additional services.
Manual SYSVOL relocation
To manually recreate the SYSVOL folder at the new location, copy the data from the existing location to the new location and then reconfigure FRS to point to the new location. Ensure that you properly copy all files to the new location.
Manually relocate SYSVOL only as a last resort, when you cannot remove and reinstall Active Directory on the domain controller. If you must perform this procedure, ensure that the SYSVOL replication between the domain controller and its replication partners is as up-to-date as possible. If the domain controller is not replicating properly with its partners, do not attempt to recreate SYSVOL until you determine why replication is not functioning and make the necessary fixes. For more information about recreating SYSVOL manually, see KB article 304300 in the Microsoft Knowledge Base. To view the Microsoft Knowledge Base, see the Microsoft Knowledge Base link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources.
SYSVOL Management Tasks and Procedures
Table 1.17 shows the tasks and procedures for managing SYSVOL.
Table 1.17 SYSVOL Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Change the space allocated to the Staging Area folder. |
|
|
As needed |
Relocate the Staging Area folder. |
|
|
As needed |
Move SYSVOL by using the Active Directory Installation Wizard. |
|
|
As needed |
Move SYSVOL manually. |
|
|
As needed |
Update the SYSVOL path. |
|
|
As needed |
Restore and rebuild SYSVOL. |
|
|
As needed |
Changing the Space Allocated to the Staging Area
The Staging Area is a folder inside the SYSVOL folder. FRS replicates files by making copies of the files, storing these copies in the Staging Area folder, and then sending them to replication partners. Because FRS replicates a copy of the file, the original file remains available for user access during replication.
The Staging Area stores files prior to being replicated and stores files that it has just received through replication. Although FRS compresses the data and attributes of the replicated files to save space in the Staging Area folder and reduce the time that is needed to replicate the files, this method requires making and storing a copy of every file prior to replication and can require a substantial amount of disk space to store all of the copies.
When you examine the disk space that SYSVOL uses, you need to examine both physical disk space and allocated disk space. Physical disk space refers to the amount of space that is available on the disk drive. To prevent SYSVOL from using all physical disk space available on the drive, an entry in the registry limits the amount of space that SYSVOL can use. This is the allocated disk space.
The default size of the Staging Area folder is 675 megabytes(MB). The minimum size is 10 MB and the maximum size is 2 terabytes. You can adjust the size limit of the Staging Area folder by setting the value in kilobytes (KB) of the Staging Space Limit registry entry in HKEY_Local_Machine\System\CurrentControlSet\Services\NtFrs\Parameters. For more information about setting the Staging Space Limit in the registry, see KB article 221111 in the Microsoft Knowledge Base. To view the Microsoft Knowledge Base, see the Microsoft Knowledge Base link on the Web Resources page at https://www.microsoft.com/windows/reskits/webresources.
When the Staging Area folder runs out of disk space, FRS behaves differently depending on the version of Windows 2000 Server that is running. If Windows 2000 Server Service Pack 2 (SP2) or earlier is running, then FRS fills the Staging Area to the limit defined in the registry and then suspends inbound and outbound replication until disk space is made available. In this situation, you can avoid suspension of replication by generously estimating the amount of disk space that SYSVOL requires.
If Windows 2000 Server Service Pack 3 (SP3) is running, then FRS fills the Staging Area to 90 percent of the limit specified in the registry and then starts removing the least recently used files to make more space available. While this prevents FRS from suspending replication, it can affect the performance of the domain controller. If a large number of files are constantly being updated, then FRS constantly stages, removes, and restages files to maintain available disk space in the Staging Area. In this case, making more space available reduces the amount of work that the domain controller performs in order to keep FRS functioning.
Other Considerations for Estimating Required Disk Space
Both FRS and DFS use the Staging Area folder. The Staging Space Limit in the registry applies to the sum of the space that is used by DFS and FRS. To maintain sufficient disk space for SYSVOL, estimate the amount of space that DFS uses as well as the space that FRS uses.
If a file changes, FRS replicates the entire file and not just the change. If two replication partners have different values set for the Staging Space Limit, the maximum size of a file that FRS can replicate is the lower of the two values.
The Staging Area folder holds files from all replication partners. You must consider traffic to and from all partners when you estimate the disk space requirements for the Staging Area folder in each SYSVOL.
Active Directory replication uses connection objects to establish connections between replication partners. FRS uses the same connections for its own replication. Two factors control the rate that replication can take place over those connections: availability of the connection and transmission speed of the network. Each connection object has an associated schedule that allows administrators to dictate when the connection is available for replication. Network administrators can limit the time that replication can take place so that processes that are more important to the daily operation of the business can use available network bandwidth over a specific connection. This becomes especially important if two replication partners are connected by a slow link (such as a 128 Kbps dial-up connection). The schedule makes it possible to limit replication traffic so that it occurs only at night or during off-peak hours.
FRS stages all replication traffic and waits for the connection to become available. When the connection is available, it begins replication and continues until it replicates all outstanding files, or the connection becomes unavailable. If many files are awaiting replication and the network is busy handling other traffic, then FRS might not get a chance to replicate all outstanding files before the schedule makes the connection unavailable. If this happens, FRS holds the remaining files until the schedule permits replication to continue. While FRS is waiting for the schedule to permit replication, it continues to stage new files for replication. The Staging Area folder needs enough space to store the staged files as well as to handle any backlog of files that might not get replicated due to limited availability of the connection.
Procedures for Changing the Space Allocated to the Staging Area
Use the following procedures to change the amount of space that is allocated to the Staging Area folder. Procedures are explained in detail in the linked topics.
Relocating the Staging Area
The Staging Area folder is likely to use most of the disk space that is allocated to SYSVOL. This is because the Staging Area folder stores all inbound and outbound files, and sometimes multiple copies of those files. As the disk space requirements increase, you can allocate more space until you reach 2 terabytes or the physical limit of the disk drive. The maximum disk space allowed for the Staging Area is 2 terabytes. If you reach the limit of the disk drive and still have not reached the 2 TB limit, consider relocating the Staging Area folder to a different disk that has more space available.
By default, the Active Directory Installation Wizard installs the Staging Area folder within the SYSVOL. The Active Directory Installation Wizard creates two folders, Staging and Staging Area, which FRS uses for the staging process. When you relocate the Staging Area, you can change the folder name. Ensure that you identify the proper folder in case the folder is renamed in your environment.
Two parameters determine the location of the Staging Area. One parameter, fRSStagingPath, is stored in the directory and contains the path to the actual location that FRS uses to stage files. The other parameter is a junction point stored in the Staging Areas folder in SYSVOL that links to the actual location that FRS uses to stage files. When relocating the Staging Area, you must update these two parameters to point to the new location.
Procedures for Relocating the Staging Area Folder
Except where noted, perform these procedures on the domain controller that contains the Staging Area folder that you want to relocate. Procedures are explained in detail in the linked topics.
On the replication partners, check the status of the shared system volume. You do not need to perform the test on every partner, but you need to perform enough tests to be confident that the shared system volumes on the partners are healthy.
Prepare a domain controller for non-authoritative SYSVOL restore.
Moving SYSVOL by Using the Active Directory Installation Wizard
Relocate SYSVOL only as a last resort. The many steps involved present many opportunities to incorrectly configure the system. If you must relocate SYSVOL, use the Active Directory Installation Wizard because it is far easier and more reliable that manually moving SYSVOL. The Active Directory Installation Wizard asks for the new location and then automatically configures the system for you.
Although using the Active Directory Installation Wizard is the preferred method for relocating SYSVOL, it is also the least practical because it involves decommissioning the domain controller. When this process is used, the Active Directory Installation Wizard is run on the domain controller to remove Active Directory. After Active Directory is removed, you run the wizard again to reinstall Active Directory. During the reinstallation, the wizard asks where you want to store SYSVOL. You enter the new location and the wizard configures it for you.
Using the Active Directory Installation Wizard to relocate SYSVOL can be too impractical for two reasons. First, because you are removing Active Directory and then reinstalling it, you also need to reinstall any other services that depend on Active Directory that are running on that domain controller. This can amount to hours of additional work and an unacceptable amount of time for the domain controller to be unavailable. Second, if a large number of objects exist in your directory, it can take hours or even days to complete the reinstallation when the new domain controller joins the network and completes the initial replication of the directory.
If this domain controller is not hosting any additional services that depend on the directory, and your directory does not take an extensive amount of time to complete the initial replication to new domain controllers, then moving SYSVOL with the Active Directory Installation Wizard can save you time and be easier and more reliable than moving SYSVOL manually.
Warning: Do not move SYSVOL with the Active Directory Installation Wizard unless you completely understand the risks and consequences of decommissioning the domain controller in question.
Procedures for Moving SYSVOL with the Active Directory Installation Wizard
Use the following procedures to remove and reinstall Active Directory in order to move SYSVOL. For more information about installing and removing Active Directory, see "Managing Installation and Removal of Active Directory" in this guide. Procedures are explained in detail in the linked topics.
View the current operations master role holders to see if any roles are assigned to this domain controller.
If this domain controller is listed as hosting either the schema master or domain naming master roles, then transfer the forest-level roles to another domain controller in the forest root domain. Any domain controller in the forest is capable of hosting these roles but it is recommended that they remain in the forest root domain. Ensure that you place the domain naming master role on a global catalog server.
If this domain controller is listed as hosting the primary domain controller (PDC) emulator, infrastructure master or relative identifier (RID) master roles, transfer the domain-level roles to another domain controller in the same domain. Do not place the infrastructure master role on a global catalog server unless all of the domain controllers host the global catalog or unless only one domain exists in the forest.
Determine whether a domain controller is a global catalog server and ensure that other domain controllers are configured as global catalog servers before continuing.
Verify the existence of the operations masters on the network.
Note: If any of the verification tests fail, do not continue until you identify and fix the problems. If these tests fail, the decommissioning operation is also likely to fail.
Verify DNS registration and functionality.
Note: If the verification test fails, do not continue until you identify and fix the problems. If the test fails, then installation is also likely to fail.
Install Active Directory. Provide the wizard with the new location for SYSVOL when prompted.
Move a server object to a different site if the domain controller is located in the wrong site.
Perform final DNS configuration for a new domain controller that is located in the forest root domain:
Create a delegation for the new domain controller in the parent domain of the DNS infrastructure if a parent domain exists and a DNS server hosts it. If a DNS server does not host the parent domain, then follow the procedures outlined in the vendor documentation to add the delegation for the new domain controller.
Or
Perform final DNS configuration for a new domain controller that is located in a child domain:
Create a delegation for the new domain controller in the forest root domain.
Moving SYSVOL Manually
If you must move the entire system volume, not just the Staging Area folder, and you have determined that moving the system volume by using the Active Directory Installation Wizard is impractical, then you can relocate the system volume manually. Because no utilities can automate this process, you must carefully ensure that you properly move all folders and maintain the same level of security at the new location.
Regardless of the method used to move SYSVOL, these events occur:
The File Replication service is stopped.
The proper folder structure is created at the new location.
The SYSVOL path information is updated in the directory and in the registry.
Default security settings are set on the new folder structure.
The File Replication service is restarted.
FRS is stopped while the changes are made and then restarted after the changes are completed. During the restart process, FRS reads the new configuration information in the directory and the registry and reconfigures itself to use the new location.
SYSVOL uses an extensive folder structure that must be recreated accurately at the new location. The easiest method is to copy the folder structure by using Windows Explorer. You must ensure that you copy any folders that may have special attributes, such as hidden folders.
The folder structure also includes junction points. Junction points look like folders when they appear in Windows Explorer but they are not really folders. Junction points contain links to other folders. When you open a junction in Windows Explorer, you see the contents of the folder to which the junction is linked. If you open a command prompt and display a directory listing that contains junction points, they are designated as <JUNCTION>, while regular folders are designated with <DIR>. Junction points behave like regular folders. When you are working in the file system, you have no indication whether you are working with a junction or a folder.
The difference between folders and junctions appears when you copy or move a junction to a new location. Because a junction is a link to another location, when you copy a junction to a new location, the link still refers to the original location. SYSVOL contains two junction points that point to folders in the SYSVOL tree. When you move the tree to a new location, you must update the junction points to point to the new location. Otherwise, the junction points continue to point to the original SYSVOL folders.
The registry and Active Directory store path information that FRS uses to locate the SYSVOL and the Staging Area folders. You must update these settings to point to the new locations.
After you create the new folders and update the paths and junctions, ensure that the folders get repopulated with the proper data. You can repopulate the files stored in SYSVOL at the new location is done by replicating the data into the new location from one of the domain controller's replication partners. The BURFLAGS option is set in the registry and when FRS restarts, it replicates the data into the new folders from one of the replication partners. Because this data is restored to the new location by means of replication, be certain that the system volumes on the replication partners are updated and functioning properly to ensure that the data replicated into the new folders is updated and has no errors.
Important: Remember, if the system volumes on your domain controllers are becoming unsynchronized to the point that you need to relocate the system volumes, be sure to troubleshoot the FRS problems and resolve the issues that cause the system volumes to become unsynchronized before you attempt to relocate the system volumes.
Procedures for Moving SYSVOL Manually
Except where noted, perform these steps on the domain controller that contains the system volume that you want to move. Procedures are explained in detail in the linked topics.
Warning: This procedure can alter security settings. After you complete the procedure, the security settings on the new system volume are reset to the default settings that were established when you installed Active Directory. You must reapply any changes to the security settings on the system volume that you made since you installed Active Directory. Failure to do so can result in unauthorized access to Group Policy objects and logon and logoff scripts.
On the replication partners, check the status of the shared system volume. You do not need to perform the test on every partner, but you need to perform enough tests to be confident that the shared system volumes on the partners are healthy.
Set the Staging Area path. If you have moved the Staging Area folder to a different location already, you do not need to do this step.
Prepare a domain controller for non-authoritative SYSVOL restore.
Updating the System Volume Path
Due to system maintenance, you might need to update the system volume path. When you add or remove disk drives, the logical drive letters of the other drives on the system can change. If either your SYSVOL or Staging Area folder is located on one of the drives whose letter changes, FRS cannot locate them. You must update the paths that FRS uses to locate these folders to solve this problem. To change the path for the system volume, make changes to the registry and in the directory. Changing the Staging Area path requires a change in the directory. Both changes require that you update the junction points. After updating the path information, you must restart FRS so it can reinitialize with the new values.
Procedures for Updating the System Volume Path
Use the following procedures to change the amount of space that is allocated to the Staging Area folder. Procedures are explained in detail in the linked topics.
Set the SYSVOL path (if needed).
Set the fRSRootPath (if needed).
Set the Staging Area path (if needed).
Restoring and Rebuilding SYSVOL
In some cases, you must recreate or rebuild the SYSVOL on a single domain controller. Attempt to rebuild SYSVOL on a single domain controller only when all other domain controllers in the domain have a healthy and functioning SYSVOL. Do not attempt to rebuild SYSVOL until you correct any problems that are occurring with FRS in a domain.
Procedure for Restoring and Rebuilding SYSVOL
Use these procedures only if you are working on a domain controller that does not have a functional SYSVOL. Procedures are explained in detail in the linked topics.
Choose a partner and check the status of the SYSVOL on the partner. Because you will be copying the system volume from one of the partners, you need to make sure that the system volume you copy from the partner is up-to-date.
Verify that replication is functioning on the partner.
Restart the domain controller that is being repaired in Directory Services Restore Mode. If you are sitting at the console of the domain controller, locally restart a domain controller in directory services restore mode. If you are accessing the domain controller remotely using Terminal Services, remotely restart a domain controller in directory services restore mode.
Prepare a domain controller for non-authoritative SYSVOL restore.
Managing Windows Time Service
The Windows 2000 time service, W32Time, requires little management and is installed by default on all Windows 2000based computers. W32Time uses coordinated universal time (UTC), which is based on an atomic time scale and is independent of time zone.
On computers that are joined to a domain, time synchronization occurs when the W32Time service starts during system startup. The Net Logon service looks for a domain controller that can authenticate and synchronize time with the client.
Time Configuration on the Forest-Root PDC Emulator
The time service uses a hierarchical relationship that controls authority and ensures common time usage. By default, the PDC emulator in the forest root domain is the authoritative time source for that forest.
Follow these best practices for configuring time on the forest-root PDC emulator, in this order of preference:
Install a hardware clock that uses the Network Time Protocol (NTP) on an internal network, and synchronize the forest-root PDC emulator and the standby PDC emulator to it.
Use IPSec to securely synchronize with another network time server.
Monitor the forest-root PDC emulator closely to ensure that its time is accurate. Do not synchronize the forest-root PDC emulator with another computer.
If none of these options are acceptable in your organization, you can synchronize with an external reliable time source. However, this option is not recommended, as it synchronizes time in an unauthenticated manner, potentially making time packets vulnerable to an attacker.
System Time Maintenance
Do not advance or roll back the system time on Windows 2000based servers under any circumstances, including attempts to:
Test significant time and date transitions such as Year 2000 testing.
Force the deletion of tombstones (objects that have been marked for deletion in the Active Directory).
Make objects on one computer override the objects on another computer.
Extend the useful life of a system backup.
Return a computer to an earlier system state including schema rollback.
Incorporate test environments into production, after you test time and date transitions on lab computers.
Troubleshoot Active Directory or File Replication Service (FRS) issues, by advancing the system time of a computer in an effort to make the content of one computer authoritative over another. Advancing the time can adversely affect the operation of the system, and it is not a useful method of resolving Active Directory or FRS replication problems.
How advancing system time affects FRS
Advancing the system time affects FRS in the following manner:
Active Directory prematurely deletes tombstones for deleted objects, causing incorrect reconciliation later. When an object is deleted, it is not actually removed from the database. It is instead marked for deletion after 60 days by default. This tombstone is replicated to other domain controllers. When the tombstone expires, the object is then permanently deleted. If the tombstone is deleted prematurely, then updates from replication partners are inconsistent.
Local file changes create change orders with event times reflecting the advanced clock time. These change orders are inserted into the outbound log but are not sent because the computer with the advanced clock will not join with the partners that remain at the correct time. Later, when the time on this computer is restored to the correct time and the computer is able to join with its outbound partners, it sends the change orders with the advanced event time. The downstream partner ignores these change orders because the event time is too far into the future. The result is that the files that changed while the time was advanced are not replicated to the other members, but remain on the computer. Furthermore, the advanced event times cause the computer to reject updates to these files that originate from other replication partners.
How advancing system time affects Active Directory
Advancing the system time affects Active Directory in the following manner:
Replication conflicts might be incorrectly resolved. Active Directory uses the time service to resolve replication conflicts. When the same attribute on the same object is changed on two different servers during a latency period, the most recent change is replicated. Thus, if you advance the time on a computer, all changes originating on that computer appear as more recent changes and are replicated despite the fact that they might not be the most recent changes.
Name conflicts might be incorrectly resolved. Active Directory also uses the time service to resolve name conflicts. When two different objects with the same name are created on two servers, Active Directory saves the most recently created object. Advancing the time on a computer might cause Active Directory to save the wrong object simply because it reflects a more recent change.
Restoring from a backup might fail. Backups are only good for the period of the tombstone lifetime. When you back up the system state, Active Directory generates an expiry token. The token is submitted when you restore the system state from the backup and is used to verify that the backup is not too old. Attempting to restore a backup after you advance the system clock might make the backup appear too old and cause the backup to fail. Do not restore a backup that you made from a computer with an advanced time setting.
Link value replication is impaired. Link value replication uses a timestamp to distinguish values. Changing the system clock hinders this mechanism.
Kerberos authentication might fail. Kerberos authentication is based on clock synchronization. Furthermore, the lifetimes of the Kerberos tickets are exceeded if the clock is moved too far ahead.
Windows Time Service Management Tasks and Procedures
Table 1.18 lists the tasks and procedures for managing Windows Time Service.
Table 1.18 Windows Time Service Management Tasks and Procedures
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Configure a time source for the forest. |
|
|
As needed |
Configure a reliable time source on a computer other than the PDC emulator. |
|
|
As needed |
Configure a client to request time from a specific time source. |
|
|
As needed |
Optimize the polling interval. |
|
|
As needed |
Disable the Windows Time Service. |
|
|
As needed |
Configuring a Time Source for the Forest
After initial deployment of your network, you typically only reconfigure the time service on the PDC emulator in two situations:
If you move the PDC emulator role to a different computer. In this case, you must configure the time source on the new role holder.
If you change your time source. For example, if you change from synchronizing with an external source to an internal hardware device.
Procedures for Configuring Time on the Forest-Root PDC Emulator
To configure time service for the forest-root PDC emulator, you might need to remove an external time source that you used previously, or, if you transferred that operations master role, you might only need to configure the time service on the new PDC emulator. To configure time on the forest-root PDC emulator, you can use the following procedures. Procedures are explained in detail in the linked topics.
Configuring a Reliable Time Source on a Computer Other than the PDC Emulator
By default, the PDC emulator in the forest root is the authoritative time source for that forest. However, you might want to configure a different computer in your network to be authoritative for the forest, in the following situations:
If you plan to move the PDC Operations Master role, you can configure a reliable time source on a different computer prior to the move(s) to avoid resets or disruption of the time service. The role of PDC emulator can move between computers, which means that every time the role of PDC emulator moves, the new PDC emulator must be manually configured to point to the external source, and the manual configuration must be removed from the original PDC emulator. To avoid this process, you can set one of the domain controllers in the parent domain as reliable and manually configure just that computer to point to an external source. Then, no matter which computer is the PDC emulator, the root of the time service stays the same and thus remains properly configured.
If you have security reasons for wanting to segregate the authoritative time computer.
When domain controllers look for a time source to synchronize with, they choose a reliable source if one is available. It is important to note that the automatic discovery mechanism in the time service client never chooses a computer that is not a domain controller. Clients must be manually configured to use any server that is not a domain controller.
Note: Setting a computer that is already synchronizing from the domain hierarchy as a reliable time source can create loops in the synchronization tree and cause unpredictable results.
Procedure for Configuring a Reliable Time Source on a Computer Other than the PDC Emulator
Although the PDC emulator in the forest root domain is the authoritative time source for that forest, you can configure a reliable time source on a computer other than the PDC emulator.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Configuring a Client to Request Time from a Specific Time Source
Certain computers do not automatically synchronize their time through the Windows 2000 time service hierarchy, so you might want to configure these clients to request time from a specific source. If you do not specify a source, each computer's internal hardware clock governs its time. The following client computers do not automatically synchronize through the time service:
Client computers that run Windows NT 4.0
Client computers that run UNIX
Computers that are not members of a domain
Note: Manually specified time sources are not authenticated, and therefore can enable an attacker to manipulate the time source and then start Kerberos V5 replay attacks. Also, a computer that does not synchronize with its domain controller can have an unsynchronized time. This causes Kerberos V5 authentication to fail, which in turn causes other actions requiring network authentication, such as printing or file sharing, to fail. When only one computer in the forest root domain is getting time from an external source, all computers within the forest remain synchronized to each other, making replay attacks difficult.
Procedures for Configuring a Client to Request Time from a Specific Time Source
The following procedures allow you to specify a time source for client computers that do not automatically synchronize through the time service. Procedures are explained in detail in the linked topics.
Set a manually configured time source on a selected computer.
Remove a manually configured time source on a selected computer.
Optimizing the Polling Interval
By default, the time service synchronizes once every 45 minutes until three successful synchronizations occur, then once every eight hours. You might want to change this interval in the following situations:
If computers are polling over a paid line, you can increase the polling interval. By polling less often, you will decrease usage of the paid line.
If you have applications or devices that require increased time accuracy, you can decrease the polling interval.
Procedure for Optimizing the Polling Interval
You only need to perform one procedure to disable the Windows Time service.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Disabling the Windows Time Service
If you choose to implement another time synchronization product that uses the NTP protocol, you must disable the W32Time time service because all NTP servers need access to UDP port 123. If W32Time is running on a Windows 2000based computer, port 123 remains occupied.
Procedure for disabling the Windows Time service
You only need to perform one procedure to disable the Windows Time service.
Managing Long-Disconnected Domain Controllers
A disconnected domain controller is a domain controller that is not replicating. Domain controllers can become disconnected deliberately or inadvertently. Short-term disconnections are not problematic because Active Directory replication automatically updates domain controllers with all changes that they have not received. However, if a domain controller must be separated from the replication topology for several weeks, you can take preliminary steps to ensure a smooth reconnection.
For example, when domain controllers must be moved long distances or are pre-staged and possibly stored for a period of time prior to being shipped to a destination, you must prepare them to ensure that no gaps occur in operations master coverage during the disconnection and that SYSVOL is updated when you reconnect the domain controller. If you plan to disconnect a domain controller for longer than a domain controller keeps track of object deletions, you must take additional steps to ensure directory consistency, as described in "Preparing a Domain Controller for a Long Disconnection" later in this section.
By monitoring replication, you can detect disconnections that occur due to network failures, service failures, or configuration errors. For information about implementing monitoring for replication failures, see "Monitoring Active Directory" earlier in this guide.
Operations Master Considerations
If a domain controller holds an operations master role, you must transfer the role prior to disconnecting the domain controller. Normal directory functioning depends on all roles being active, so when you plan to disconnect the domain controller, you must first transfer any operations master roles. Role transfer ensures that no gaps in master operations coverage occur, which can cause directory inconsistencies. For information about transferring operations master roles, see "Managing Operations Masters" earlier in this guide.
Active Directory Replication Considerations
Ensure that the domain controller is updated before you disconnect it. Immediately prior to disconnecting the domain controller, force replication with all replication partners and verify that each directory partition replicates to the domain controller that you are disconnecting. If replication of any directory partition does not succeed, resolve the replication problem prior to disconnecting. By ensuring that replication is up-to-date, you can maximize the possible safe disconnection period, which cannot exceed the tombstone lifetime for the forest. For information about estimating the maximum safe disconnection period, see "Preparing a Domain Controller for a Long Disconnection" later in this guide.
Tombstone Lifetime and Backup Considerations
Active Directory backups are useful for recovering a domain controller for only as long as the tombstone lifetime. When an object is deleted, Active Directory replicates the object as a tombstone, which consists of a small subset of the attributes of the deleted object. The tombstone is retained in Active Directory for 60 days by default, after which it is permanently removed. Because a domain controller that is disconnected for longer than the tombstone lifetime cannot receive deletions that occurred prior to the beginning of the tombstone lifetime, a backup that is older than the tombstone lifetime cannot be used to restore Active Directory.
When conditions beyond your control cause a domain controller to be disconnected longer than the tombstone lifetime, one or more objects that have been deleted from the rest of the directory while the domain controller was offline might remain on the disconnected domain controller. The best practice recommendation for reconciling this condition of inconsistency is to reinstall Windows on the outdated domain controller and then reinstall Active Directory. Otherwise, the outdated domain controller can potentially reintroduce (reanimate) objects into Active Directory that were deleted while the domain controller was disconnected. For information about how objects become reanimated in Active Directory, see "Reconnecting Long-Disconnected Domain Controllers" later in this guide.
If planned domain controller disconnections are consistently lasting longer than 60 days, alert the design team and consider extending the tombstone lifetime for the forest.
SYSVOL Consistency Considerations
SYSVOL is a file system folder that stores files that must be available and synchronized among all domain controllers. SYSVOL contains the NETLOGON share, Group Policy settings, and File Replication service (FRS) staging directories and files. SYSVOL is required for Active Directory to function properly.
SYSVOL is replicated by the File Replication service (FRS). FRS has a fixed tombstone lifetime of 60 days. Because you cannot change this interval, any domain controller that is disconnected for more than 60 days potentially has an outdated SYSVOL. Updating SYSVOL requires performing a non-authoritative restore of SYSVOL.
In addition, SYSVOL replication cannot be synchronized manually. For this reason, ensuring that SYSVOL is updated prior to disconnecting the domain controller is more difficult than simply updating SYSVOL when the domain controller is reconnected. Regardless of the length of the disconnection, to ensure that SYSVOL is synchronized when the domain controller is reconnected, prepare the domain controller to perform a non-authoritative restore of SYSVOL prior to disconnecting it. When it restarts, non-authoritative restore of SYSVOL occurs automatically. For information about performing non-authoritative restore of SYSVOL, see "Restoring and Rebuilding SYSVOL" earlier in this guide.
Windows 2000 Server with SP3
Windows 2000 Server with Service Pack 3 (SP3) provides the ability to force strict replication consistency, which prevents outdated domain controllers from reintroducing objects that no longer exist in Active Directory. When deploying new domain controllers that are running Windows 2000 Server SP3, modify the registry to enforce strict replication consistency. For information about strict replication consistency, see "Removing Lingering Objects from an Outdated Writable Domain Controller" in this guide. For information about installing domain controllers, see "Installing and Removing Active Directory" earlier in this guide.
Best Practice Recommendations for Managing Long Disconnections
If you must disconnect a domain controller for a period of several weeks or months, follow these recommendations:
Prior to disconnecting, determine the maximum length of time that the domain controller will be disconnected and subtract a generous estimate of the end-to-end replication latency. This amount of time is the maximum period for which the domain controller can safely be disconnected.
Prior to disconnecting, determine the value of the tombstone lifetime for the forest. If you estimate the maximum safe time of disconnection to be longer than the tombstone lifetime, contact a supervisor. The design team must determine whether to extend the tombstone lifetime or rebuild the domain controller prior to reconnecting it.
Prior to disconnecting, prepare the registry for automatic non-authoritative restore of SYSVOL when the domain controller restarts.
Immediately prior to disconnecting, ensure that the domain controller replicates successfully with all replication partners.
When you disconnect the domain controller, attach a label to the computer that identifies the date and time of disconnection.
When reconnecting the domain controller, if the site contains no other domain controller that is authoritative for the domain, time the restart of the domain controller to coincide with the beginning of intersite replication to restore SYSVOL as quickly as possible. If the site has one or more other domain controllers that are authoritative for the domain, start the domain controller at any time.
If a domain controller has been disconnected for longer than the maximum safe time of disconnection (tombstone lifetime less end-to-end replication latency), do not allow the domain controller to replicate. Reinstall Windows 2000 Server. This recommendation applies to all such domain controllers, regardless of the version of Windows 2000 Server they are running (SP3, SP2, or earlier).
If you deploy Windows 2000 Server SP3, modify the registry to enforce strict replication behavior at the time the domain controller is installed.
Tasks and Procedures for Managing Long-Disconnected Domain Controllers
Table 1.19 shows the tasks and procedures for managing long disconnected domain controllers, including tasks that address removing lingering objects.
Table 1.19 Tasks and Procedures for Managing Long-Disconnected Domain Controllers
Tasks |
Procedures |
Tools |
Frequency |
|---|---|---|---|
Prepare a domain controller for long disconnection. |
|
|
As needed |
Reconnect a long-disconnected domain controller. |
|
|
As needed |
Remove lingering objects from an outdated writable domain controller. |
Windows 2000 Server with SP2:
Windows 2000 Server with SP3:
Windows 2000 Server with SP2 or SP3, continue as follows:
|
|
As needed |
Remove lingering objects from a global catalog server. |
Windows 2000 Server with SP2:
Windows 2000 Server with SP3:
|
Ldp.exe |
As needed |
Preparing a Domain Controller for a Long Disconnection
When you need to take a domain controller offline for a prolonged period, prepare the domain controller by doing the following:
Establish the maximum safe disconnection period. Determine the tombstone lifetime interval and subtract a generous estimate of the end-to-end replication latency to establish the maximum safe period of disconnection. Otherwise, even when the domain controller is reconnected prior to the end of the tombstone lifetime, a tombstone can potentially expire before reaching the reconnected domain controller.
Verify replication success on the domain controller prior to disconnecting it. If replication is not successful, troubleshoot and fix the problem prior to disconnecting the domain controller.
Modify the registry to prepare the domain controller to perform a non-authoritative restore of SYSVOL when it restarts. SYSVOL inconsistencies are not easily verifiable prior to disconnecting. Therefore, by setting the registry to restore SYSVOL when the domain controller restarts, you can ensure that SYSVOL reinitializes its membership in the replica set and updates its content at the earliest opportunity after reconnecting the domain controller.
When modifying the registry to restore SYSVOL, consider the following:
If SYSVOL is the only replica set that is represented on the domain controller, modify the global BurFlags registry entry.
If other replica sets are represented on the domain controller and you want to update only SYSVOL, modify the replica-set-specific BurFlags registry entry for SYSVOL.
For information about restoring SYSVOL, see "Restoring and Rebuilding SYSVOL" earlier in this guide.
Determine whether the domain controller holds an operations master role. If the domain controller is an operations master, transfer the role prior to disconnecting. For information about transferring operations master roles, see "Managing Operations Masters" earlier in this guide.
If the length of the disconnection is predicted to be longer than the current tombstone lifetime, consult the design team about extending the tombstone lifetime.
Procedures for Preparing a Domain Controller for Long Disconnection
Perform the following procedures prior to disconnecting a domain controller. Procedures are explained in detail in the linked topics.
Determine the anticipated length of the disconnection.
Determine the maximum safe disconnection period by subtracting a generous estimate of the end-to-end replication latency from the tombstone lifetime. Either find the latency estimate in the design documentation for your deployment, or request the information from a member of the design or deployment team.
If the anticipated time of disconnection exceeds the maximum safe disconnection period, do not disconnect the domain controller. Contact a supervisor.
If the estimated time of disconnection does not exceed the maximum safe disconnection time, proceed with disconnection.
View the current operations master role holders to determine whether the domain controller is an operations master role holder.
Transfer a domain-level operations master role, if appropriate.
Transfer a forest-level operations master role, if appropriate.
Prepare the domain controller for non-authoritative SYSVOL restore on the domain controller that you are disconnecting. This process ensures an up-to-date SYSVOL when the domain controller is restarted.
Synchronize replication from all inbound (source) replication partners. Each connection object below the NTDS Settings object for the server you are disconnecting represents an inbound replication partner.
Verify successful replication to the domain controller that you are disconnecting.
Label the domain controller with the date and time of disconnection and the maximum safe disconnection period.
Caution: The registry editor bypasses standard safeguards, allowing settings that can damage your system, or even require you to reinstall Windows. If you must edit the registry, back up system state first. For information about backing up system state, see "Active Directory Backup and Restore" in this guide.
Reconnecting Long-Disconnected Domain Controllers
Assuming that the domain controller has not been disconnected for longer than the maximum safe period of disconnection (tombstone lifetime minus end-to-end replication latency), reconnecting it to the replication topology requires no special procedures. By default, the Knowledge Consistency Checker (KCC) on a domain controller runs 5 minutes after the domain controller starts, automatically incorporating the reconnected domain controller into the replication topology.
If you plan appropriately for disconnecting and reconnecting domain controllers, no domain controller will be disconnected from the replication topology for longer than a tombstone lifetime. However, if unexpected events result in a domain controller becoming outdated, do not reconnect the domain controller. Do not attempt to remove Active Directory because this process requires replication. To ensure directory consistency, reinstall Windows 2000 Server on the outdated domain controller. For information about how to reinstall a domain controller that has not replicated for longer than a tombstone lifetime, see "Recovering a Domain Controller Through Reinstallation."
By monitoring replication, you avoid unexpected lengthy disconnections of domain controllers. For information about monitoring replication, see "Monitoring Active Directory" in this guide.
Long Disconnections and Tombstone Lifetime
If a domain controller remains disconnected for longer than a tombstone lifetime, an object that has been deleted from the directory can remain on the disconnected domain controller. For this reason, such objects are called "lingering objects."
Lingering objects can occur in the following circumstances:
A domain controller goes offline immediately prior to the deletion of an object on another domain controller and remains offline for:
A period that exceeds the tombstone lifetime.
A period that is less than the tombstone lifetime, but replication latency exceeds the remaining duration of the tombstone lifetime.
A domain controller goes offline following the deletion of an object on another domain controller but prior to receiving replication of the tombstone, and remains offline for a period that exceeds the tombstone lifetime.
A domain controller goes offline, an object is deleted on that domain controller, and the object tombstone is removed by garbage collection on that domain controller prior to the domain controller being reconnected to replication.
In the latter case, an object exists on all domain controllers in the domain (for a domain-specific object) or forest (for a configuration or schema object) except the reconnected domain controller. In this case, the remedy is simply to delete the object on any writable domain controller.
However, in the first two cases, if the domain controller is then reconnected to the replication topology, objects that exist nowhere else in the forest remain on the domain controller and potentially can be reintroduced into the directory.
If lingering objects are security principals, reintroducing them can have serious consequences. For more information about how lingering objects are reintroduced into the directory and how to remove them, see "Removing Lingering Objects from an Outdated Writable Domain Controller."
Best Practice Recommendations for Avoiding Lingering Objects
Take the following precautions to ensure that lingering objects do not occur:
Monitor the KCC topology and replication to ensure that long disconnections are detected. For information about monitoring the KCC and replication, see "Monitoring Active Directory" earlier in this guide.
Ensure that the tombstone lifetime is not lowered below the default of 60 days.
If you know that a domain controller will be offline for longer than the tombstone lifetime, consult the design team about increasing the tombstone lifetime to a period that safely encompasses the offline duration plus a generous period of replication latency.
Install Windows 2000 Server SP3 as soon as possible and enable strict replication consistency to ensure that lingering objects cannot replicate.
Long Disconnections and SYSVOL
If the tombstone lifetime has been extended to longer than 60 days, SYSVOL will be outdated when you reconnect the domain controller. The recommended practice is to prepare a domain controller for a long disconnection by modifying the registry so that SYSVOL is restored automatically when the domain controller is restarted. To update SYSVOL as soon as possible after reconnecting, plan the time that you restart the domain controller to optimize the replication schedule, as follows:
If the closest replication partner for the domain is in a different site, view site link properties to determine the schedule and then restart the domain controller as soon as possible after the schedule opens.
If a replication partner for the domain is available within the site, verify replication success on that partner prior to restarting the domain controller.
Important: Do not use file copy utilities such as xcopy or robocopy to update an outdated SYSVOL.
In the event that a domain controller has been disconnected for a tombstone lifetime or longer but has already replicated, follow the instructions for detecting and removing lingering objects in ".Removing Lingering Objects from an Outdated Writable Domain Controller."
Procedures for Reconnecting a Long-Disconnected Domain Controller
Follow these procedures to reconnect the domain controller. Procedures are explained in detail in the linked topics.
Determine whether the maximum safe disconnection time has been exceeded, and proceed accordingly:
If the domain controller has been disconnected for a period that exceeds the maximum safe disconnection period, do not reconnect the domain controller. Contact a supervisor about reinstalling the domain controller.
If the maximum safe time has not been exceeded, proceed with reconnecting.
If the site in which you are reconnecting the domain controller has one or more other domain controllers that are authoritative for the domain, start the domain controller at any time.
If the site in which you are reconnecting the domain controller has no other domain controllers that are authoritative for the domain, proceed as follows:
Determine when the next intersite replication cycle is scheduled to begin by viewing the replication properties on the site link that connects this site to the next closest site that includes domain controllers for this domain.
As soon as possible after the next replication cycle begins, start the domain controller.
After replication is complete, verify successful replication to the domain controller (the reconnected domain controller) of the domain, configuration, and schema directory partitions. If the domain controller is a global catalog server, check for successful replication of all domain directory partitions.
In the event that a domain controller has been disconnected for a tombstone lifetime or longer but has already replicated, follow the instructions for detecting and removing lingering objects in "Removing Lingering Objects from an Outdated Writable Domain Controller."
Removing Lingering Objects from an Outdated Writable Domain Controller
If a domain controller does not replicate for a period that is longer than the tombstone lifetime and the domain controller is then reintroduced into the replication topology, objects that have been deleted from Active Directory while the domain controller was offline can remain on the domain controller as lingering objects.
Causes for Lingering Objects
Lingering objects can occur whenever a domain controller does not replicate for a period that exceeds the tombstone lifetime. Unexpectedly long disconnections can be caused by the following conditions:
A domain controller is left in a storage room and forgotten, or shipment of the pre-staged domain controller to its remote location takes longer than a tombstone lifetime.
Replication fails and monitoring is not in place. For example, if a bridgehead server is overloaded, replication can become backlogged indefinitely.
WAN connections are unavailable for long periods. For example, a domain controller on board a cruise ship might be unable to replicate because the ship is at sea for longer than the tombstone lifetime.
Garbage collection tampering. For example:
Someone changes the time on a domain controller to force garbage collection.
Someone changes the tombstone lifetime to force garbage collection.
Indications that a Domain Controller has Lingering Objects
An outdated domain controller can store lingering objects with no noticeable effect as long as no one updates the lingering object or tries to create an object with the same name in the domain or the same user principal name in the forest. However, the existence of lingering objects can cause problems, especially if the object is a security principal.
The following conditions indicate that a domain controller has lingering objects:
A deleted user or group account does not disappear from the Global Address List on Exchange servers. Therefore, although the account name appears in the list, attempts to send mail result in errors.
E-mail messages are not delivered to a user whose user object was moved between domains. After an outdated domain controller or global catalog server becomes reconnected, both instances of the user object appear in the global catalog. Both objects have the same e-mail address, so e-mail messages cannot be delivered.
A universal group that no longer exists still appears in a user's access token. Although the group no longer exists, if a user account still has the group in its security token, the user might have access to a resource that you intended to be unavailable to that user.
A new object or Exchange mailbox cannot be created when the samAccountName attribute value of the new object is the same as a lingering object. An error reports that the object already exists.
Replication succeeds with "no such object" error (event ID 1388) when "loose replication consistency" is in effect. This error indicates that the source domain controller revived a lingering object in the directory.
Replication fails with a "no such object" error (event ID 1084) when "strict replication consistency" is in effect. This error indicates that the source domain controller tried to replicate a lingering object.
Replication of Lingering Objects
If a user updates a lingering object on the outdated domain controller, the destination domain controller that receives the request for the update cannot update the object because the object does not exist. The destination domain controller logs an NTDS Replication error in the Directory Service log in Event Viewer. The error that is reported depends on the type of replication consistency that is in effect on the domain controller.
The replication response differs on domain controllers that use loose replication consistency and domain controllers that use strict replication consistency. On domain controllers that use loose replication consistency (the default behavior with Windows 2000 Server SP2), the destination domain controller requests a full copy of the object from the replication source. If the object is being modified, the destination requests the full object and the object is revived in the directory. If the object is being deleted, the destination replicates the tombstone. In either case, the NTDS Replication event ID 1388 is logged in the Directory Service log by the destination. The error reports that the object being updated does not exist and the domain controller does not have enough information to create it, and so it will request a complete copy. This error alerts you to the fact that you have at least one lingering object and gives you the information that you need in order to locate that object and delete it if it has been revived. Deleting the revived object on a writable domain controller removes it from the directory
Domain controllers on which strict replication consistency is enabled (configurable behavior with Windows 2000 Server SP3) refuse replication from the outdated replication source. This action stops replication from the outdated source and logs NTDS Replication event ID 1084 in the Directory Service log. The error reports that the object cannot be updated and replication will not be accepted from the source until the issue is resolved. The information in the error includes the name, GUID, and source of the lingering object so that you can delete the object and determine whether additional lingering objects exist on the source. For this error to be logged, however, you must have modified the registry to implement strict replication consistency.
In both cases, you can delete the identified lingering object and then take steps to identify and remove all additional lingering objects from the outdated domain controller.
Sequence for Removing Lingering Objects
The process for removing lingering objects from an outdated writable domain controller involves several procedures that must be performed in sequence. After an error indicates the existence of a lingering object, use the following general sequence to remove the lingering object and determine whether there are other lingering objects on the source domain controller:
Identify the domain controller that replicated the update to a lingering object. Use the information in event ID 1388 (Windows 2000 Server with SP2) or event ID 1084 (Windows 2000 Server with SP3) to identify the source domain controller.
Disable outbound replication on the source domain controller.
Delete the lingering object from the source domain controller.
Compare the database contents of the outdated source domain controller and an up-to-date replication partner to determine whether the outdated source domain controller contains objects that do not exist on its replication partner.
Identify the distinguished names of the objects that exist on the outdated domain controller but not on the replication partner.
Examine metadata of the object to determine when it was created.
Delete the objects that were created prior to disconnecting the domain controller.
Restart outbound replication on the source domain controller.
Deletions of the lingering objects replicate to the other domain controllers. Any domain controller that is running Windows 2000 Server with SP2, and that does not have the object, logs event ID 1388. In this case, the missing object is revived as a tombstone, and replicates as such. The errors on domain controllers that do not have the object can be ignored; they will cease after the second replication cycle.
If you have domain controllers that are running Windows 2000 Server with SP3, you can set the registry to enforce strict replication consistency, which ensures that lingering objects do not replicate. For this reason, attempted replication of the deletions will not be accepted. You must delete lingering objects from only the outdated domain controller. For information about setting strict replication consistency for domain controllers that are running Windows 2000 Server with SP3, see "Managing Active Directory Installation and Removal" in this guide.
Procedures for Removing Lingering Objects from an Outdated Writable Domain Controller
Use the following process to identify and remove lingering objects after you have discovered an outdated domain controller. The initial step in the process varies according to the version of Windows 2000 Server that you are using. Procedures are explained in detail in the linked topics.
Identify and delete the initial occurrence of a lingering object, as follows:
For Windows 2000 Server with SP2:
Identify a revived lingering object and its replication source on a writable domain controller. Event ID 1388 provides the distinguished name of an object that has been updated on an outdated domain controller. The message also provides the GUID of the domain controller from which the update was replicated. Use the GUID to discover the name of the source domain controller. Repeat this process on each source domain controller until you identify a source domain controller that does not have the error. This domain controller is the outdated source domain controller.
Disable outbound replication on the outdated source domain controller.
Delete the object from the outdated source domain controller.
For Windows 2000 Server with SP3:
- Identify and delete a known non-replicated lingering object on an outdated domain controller, as identified in event ID 1084. The object and source domain controller are named in the error message.
Identify unknown lingering objects on an outdated domain controller. This procedure requires the following series of subprocedures to be performed sequentially:
Compare the directory databases of the outdated domain controller and the domain controller that received the initial replication error.
Identify the distinguished names of the objects that exist on the outdated domain controller but not on the partner domain controller.
Note: The results of this procedure identify only objects where the numbers of objects did not agree between domain controllers. If numbers match but an object of a class was added on one domain controller and a different object of the same class was deleted on the other, and these changes did not replicate, this test cannot identify these inconsistent objects.
On the outdated domain controller, view the replication metadata of objects that you identified in the previous procedure to determine whether they were created prior to the time the domain controller was disconnected or were created during the time that the domain controller was offline. If the newest date in the Org.Time/Date column is older than the date on which the domain controller was disconnected, the object is a lingering object.
On the outdated domain controller, delete the objects that were created prior to the date and time that the domain controller was disconnected.
Restart disabled outbound replication on the outdated domain controller (SP2 only).
Synchronize replication from the outdated domain controller to the partner domain controller to replicate the deletions. Use the connection object on the replication partner that shows the name of the outdated domain controller in the From Server column. This procedure results in error messages on domain controllers that do not have the objects, but these messages can be ignored and will cease by the second replication cycle.
Removing Lingering Objects from a Global Catalog Server
If you delete a lingering object on a writable domain controller, the object deletion replicates to all writable domain controllers in the domain as well as to all global catalog servers. However, if a global catalog server becomes outdated, lingering objects can potentially exist in a read-only replica on the global catalog server and nowhere else, in which case you cannot delete the object by the normal method. The recommended solution to this problem depends on the version of Windows 2000 Server that is running on the outdated global catalog server:
Windows 2000 Server with SP2: Contact Microsoft Product Support Services.
Windows 2000 Server with SP3: Use Ldp.exe to identify and delete the object from all global catalog servers that retain the object.
Causes for Lingering Objects on Global Catalog Servers
Excessively high replication load on a global catalog server, in combination with a short intersite replication interval, can result in updates not being replicated. Global catalog servers replicate read-only replicas of all domain directory partitions in the forest. The replication of read-only replicas has a lower priority than the replication of writable replicas. In addition, global catalog servers are often bridgehead servers, which adds to the replication load. If the replication load on global catalog servers acting as bridgehead servers is too high due to an extremely short replication interval, excessive numbers of concurrent outbound replication partners, or a combination of both, the replication queue can become backlogged. If the condition persists, read-only replicas can remain in the queue indefinitely. These conditions can result in lingering objects on a global catalog server.
If replication of a read-only replica is stalled or the domain controller is disconnected for longer than a tombstone lifetime, the deletion of an object from the corresponding writable directory partition can potentially expire without ever reaching the global catalog server. In this case, the only location of this object is in the read-only replica on the global catalog server.
As with writable domain controllers, a global catalog server that is not monitored for replication can potentially become outdated. When appropriate monitoring is in place and sensible intersite replication schedules are configured, global catalog servers are not susceptible to becoming outdated. For information about monitoring replication, see "Monitoring Active Directory" in this document. For information about scheduling replication, see "Managing Sites" in this document.
Indications that Lingering Objects Exist on a Global Catalog Server
The following events indicate that a lingering objects exists on a global catalog server:
A deleted user or group account does not disappear from the Global Address List on Exchange servers.
E-mail messages are not deliverable to a user whose Active Directory account appears to be current.
A new user account or Exchange mailbox cannot be created because the object already exists, but you do not see the object in Active Directory.
Searches that use attributes of an existing object find an object of the same name that has been deleted from the domain but remains in an isolated global catalog server.
Sequence for Removing Lingering Objects from a Global Catalog Server
To remove a lingering object from a global catalog server, you need an attribute value to use for the search to identify the object in the global catalog. For example, when you are trying to create a mailbox, user account, or other object in Active Directory, and error messages indicate that the object already exists, use the name of the object that you are trying to create. If you know that a deleted group or user name appears in the Global Address List, use that name.
Use the following general sequence of tasks to locate and remove a lingering object from a global catalog server:
Use an LDAP search to establish the distinguished name and GUID of the duplicate (lingering) object.
Use the distinguished name to identify the domain of the object.
Identify a writable domain controller for that domain.
Identify the GUID of the writable domain controller.
Delete the object from the global catalog server. This procedure requires the preceding information.
Repeat the previous steps for every object and global catalog server that is outdated.
When deleting an object that has child objects, you must delete the child object first, then delete the parent. You can tell from the distinguished name whether the object has parent objects.
Procedures for Removing a Lingering Object from a Global Catalog Server
Use the following procedures to identify and remove a read-only lingering object from a global catalog server that is running Windows 2000 Server with SP3. Procedures are explained in detail in the linked topics.
Establish the distinguished name and GUID of the object by searching the global catalog on an attribute that can uniquely identify the object. From the distinguished name, you can identify the domain by the DC= components.
Identify the GUID of a domain controller that has a writable replica of the domain of the lingering object.
Delete the lingering object from the global catalog server. In this procedure, use the GUID of the object and the GUID of the writable domain controller that you identify in procedures 1 and 2.