Perform initial recovery

Applies To: Windows Server 2016, Windows Server 2012 and 2012 R2, Windows Server 2008 and 2008 R2

This section includes the following steps:

Restore the first writeable domain controller in each domain

Beginning with a writeable DC in the forest root domain, complete the steps in this section in order to restore the first DC. The forest root domain is important because it stores the Schema Admins and Enterprise Admins groups. It also helps maintain the trust hierarchy in the forest. In addition, the forest root domain usually holds the DNS root server for the forest’s DNS namespace. Consequently, the Active Directory–integrated DNS zone for that domain contains the alias (CNAME) resource records for all other DCs in the forest (which are required for replication) and the global catalog DNS resource records.

After you recover the forest root domain, repeat the same steps to recover the remaining domains in the forest. You can recover more than one domain simultaneously; however, always recover a parent domain before recovering a child to prevent any break in the trust hierarchy or DNS name resolution.

For each domain that you recover, restore only one writeable DC from backup. This is the most important part of the recovery because the DC must have a database that has not been influenced by whatever caused the forest to fail. It is important to have a trusted backup that is thoroughly tested before it is introduced into the production environment.

Then perform the following steps. Procedures for performing certain steps are in AD Forest Recovery - Procedures.

  1. If you plan to restore a physical server, ensure that the network cable of the target DC is not attached and therefore is not connected to the production network. For a virtual machine, you can remove the network adapter or use a network adapter that is attached to another network where you can test the recovery process while isolated from the production network.

  2. Because this is the first writeable DC in the domain, you must perform a nonauthoritative restore of AD DS and an authoritative restore of SYSVOL. The restore operation must be completed by using an Active Directory-aware backup and restore application, such as Windows Server Backup (that is, you should not restore the DC by using unsupported methods such as restoring a VM snapshot).

    An authoritative restore of SYSVOL is required because replication of the SYSVOL replicated folder must be started after you recover from a disaster. All subsequent DCs that are added in the domain must resynchronize their SYSVOL folder with a copy of the folder that has been selected to be authoritative before the folder can be advertised.

    Caution

    Perform an authoritative (or primary) restore operation of SYSVOL only for the first DC to be restored in the forest root domain. Incorrectly performing primary restore operations of the SYSVOL on other DCs leads to replication conflicts of SYSVOL data.

    There are two options perform a nonauthoritative restore of AD DS and an authoritative restore of SYSVOL:

  3. After you restore and restart the writeable DC, verify that the failure did not affect the data on the DC. If the DC data is damaged, then repeat step 2 with a different backup.

    If the restored domain controller hosts an operations master role, you may need to add the following registry entry to avoid AD DS being unavailable until it has completed replication of a writeable directory partition:

    HKLM\System\CurrentControlSet\Services\NTDS\Parameters\Repl Perform Initial Synchronizations

    Create the entry with the data type REG_DWORD and a value of 0. After the forest is recovered completely, you can reset the value of this entry to 1, which requires a domain controller that restarts and holds operations master roles to have successful AD DS inbound and outbound replication with its known replica partners before it advertises itself as domain controller and starts providing services to clients. For more information about initial synchronization requirements, see KB article 305476.

    Continue to the next steps only after you restore and verify the data and before you join this computer to the production network.

  4. If you suspect that the forest-wide failure was related to network intrusion or malicious attack, reset the account passwords for all administrative accounts, including members of the Enterprise Admins, Domain Admins, Schema Admins, Server Operators, Account Operators groups, and so on. The reset of administrative account passwords should be completed before additional domain controllers are installed during the next phase of the forest recovery.

  5. On the first restored DC in the forest root domain, seize all domain-wide and forest-wide operations master roles. Enterprise Admins and Schema Admins credentials are needed to seize forest-wide operations master roles.

    In each child domain, seize domain-wide operations master roles. Although you might retain the operations master roles on the restored DC only temporarily, seizing these roles assures you regarding which DC hosts them at this point in the forest recovery process. As part of your post-recovery process, you can redistribute the operations master roles as needed. For more information about seizing operations master roles, see Seizing an operations master role. For recommendations about where to place operations master roles, see What Are Operations Masters?.

  6. Clean up metadata of all other writeable DCs in the forest root domain that you are not restoring from backup (all writeable DCs in the domain except for this first DC). If you use the version of Active Directory Users and Computers or Active Directory Sites and Services that is included with Windows Server 2008 or later or RSAT for Windows Vista or later, metadata cleanup is performed automatically when you delete a DC object. In addition, the server object and computer object for the deleted DC are also deleted automatically. For more information, see Cleaning metadata of removed writable DCs.

    Cleaning up metadata prevents possible duplication of NTDS-settings objects if AD DS is installed on a DC in a different site. Potentially, this could also save the Knowledge Consistency Checker (KCC) the process of creating replication links when the DCs themselves might not be present. Moreover, as part of metadata cleanup, DC Locator DNS resource records for all other DCs in the domain will be deleted from DNS.

    Until the metadata of all other DCs in the domain is removed, this DC, if it were a RID master before recovery, will not assume the RID master role and therefore will not be able to issue new RIDs. You might see event ID 16650 in the System log in Event Viewer indicating this failure, but you should see event ID 16648 indicating success a little while after you have cleaned the metadata.

  7. If you have DNS zones that are stored in AD DS, ensure that the local DNS Server service is installed and running on the DC that you have restored. If this DC was not a DNS server before the forest failure, you must install and configure the DNS server.

    Note

    If the restored DC runs Windows Server 2008, you need to install the hotfix in KB article 975654 or connect the server to an isolated network temporarily in order to install DNS server. The hotfix is not required for any other versions of Windows Server.

    In the forest root domain, configure the restored DC with its own IP address (or a loopback address, such as 127.0.0.1) as its preferred DNS server. You can configure this setting in the TCP/IP properties of the local area network (LAN) adapter. This is the first DNS server in the forest. For more information, see Configure TCP/IP to use DNS.

    In each child domain, configure the restored DC with the IP address of the first DNS server in the forest root domain as its preferred DNS server. You can configure this setting in the TCP/IP properties of the LAN adapter. For more information, see Configure TCP/IP to use DNS.

    In the _msdcs and domain DNS zones, delete NS records of DCs that no longer exist after metadata cleanup. Check if the SRV records of the cleaned up DCs have been removed. To help speed up DNS SRV record removal, run:

    nltest.exe /dsderegdns:server.domain.tld  
    
  8. Raise the value of the available RID pool by 100,000. For more information, see Raising the value of available RID pools. If you have reason to believe that raising the RID Pool by 100,000 is insufficient for your particular situation, you should determine the lowest increase that is still safe to use. RIDs are a finite resource that should not be used up needlessly.

    If new security principals were created in the domain after the time of the backup that you use for the restore, these security principals might have access rights on certain objects. These security principals no longer exist after recovery because the recovery has reverted to the backup; however, their access rights might still exist. If the available RID pool is not raised after a restore, new user objects that are created after the forest recovery might obtain identical security IDs (SIDs) and could have access to those objects, which was not originally intended.

    To illustrate, consider the example of the new employee named Amy that was mentioned in the introduction. The user object for Amy no longer exists after the restore operation because it was created after the backup that was used to restore the domain. However, any access rights that were assigned to that user object might persist after the restore operation. If the SID for that user object is reassigned to a new object after the restore operation, the new object would obtain those access rights.

  9. Invalidate the current RID pool. The current RID pool is invalidated after a system state restore. But if a system state restore was not performed, the current RID pool needs to be invalidated to prevent the restored DC from re-issuing RIDs from the RID pool that was assigned at the time the backup was created. For more information, see Invalidating the current RID pool.

    Note

    The first time that you attempt to create an object with a SID after you invalidate the RID pool you will receive an error. The attempt to create an object triggers a request for a new RID pool. Retry of the operation succeeds because the new RID pool will be allocated.

  10. Reset the computer account password of this DC twice. For more information, see Resetting the computer account password of the domain controller.

  11. Reset the krbtgt password twice. For more information, see Resetting the krbtgt password.

    Because the krbtgt password history is two passwords, reset passwords twice to remove the original (prefailure) password from password history.

    Note

    If the forest recovery is in response to a security breach, you may also reset the trust passwords. For more information, see Resetting a trust password on one side of the trust.

  12. If the forest has multiple domains and the restored DC was a global catalog server before the failure, clear the Global catalog check box in the NTDS Settings properties to remove the global catalog from the DC. The exception to this rule is the common case of a forest with just one domain. In this case, it is not required to remove the global catalog. For more information, see Removing the global catalog.

    By restoring a global catalog from a backup that is more recent than other backups that are used to restore DCs in other domains, you might introduce lingering objects. Consider the following example. In domain A, DC1 is restored from a backup that was taken at time T1. In domain B, DC2 is restored from a global catalog backup that was taken at time T2. Suppose T2 is more recent than T1, and some objects were created between T1 and T2. After these DCs are restored, DC2, which is a global catalog, holds newer data for domain A's partial replica than domain A holds itself. DC2, in this case, holds lingering objects because these objects are not present on DC1.

    The presence of lingering objects can lead to problems. For instance, e-mail messages might not be delivered to a user whose user object was moved between domains. After you bring the outdated DC or global catalog server back online, both instances of the user object appear in the global catalog. Both objects have the same e-mail address; therefore, e-mail messages cannot be delivered.

    A second problem is that a user account that no longer exists might still appear in the global address list. A third problem is that a universal group that no longer exists might still appear in a user's access token.

    If you did restore a DC that was a global catalog—either inadvertently or because that was the solitary backup that you trusted—we recommend that you prevent the occurrence of lingering objects by disabling the global catalog soon after the restore operation is complete. Disabling the global catalog flag will result in the computer losing all its partial replicas (partitions) and relegating itself to regular DC status.

  13. Configure Windows Time Service. In the forest root domain, configure the PDC emulator to synchronize time from an external time source. For more information, see Configure the Windows Time service on the PDC emulator in the Forest Root Domain.

Reconnect each restored writeable domain controller to a common network

At this stage you should have one DC restored (and recovery steps performed) in the forest root domain and in each of the remaining domains. Join these DCs to a common network that is isolated from the rest of the environment and complete the following steps in order to validate forest health and replication.

Note

When you join the physical DCs to an isolated network, you may need to change their IP addresses. As a result, the IP addresses of DNS records will be wrong. Because a global catalog server is not available, secure dynamic updates for DNS will fail. Virtual DCs are more advantageous in this case because they can be joined to a new virtual network without changing their IP addresses. This is one reason why virtual DCs are recommended as the first domain controllers to be restored during forest recovery.

After validation, Join the DCs to the production network and complete the steps to verify forest replication health.

  • To fix name resolution, create DNS delegation records and configure DNS forwarding and root hints as needed. Run repadmin /replsum to check replication between DCs.

  • If the restored DC’s are not direct replication partners, replication recovery will be much faster by creating temporary connection objects between them.

  • To validate metadata cleanup, run Repadmin /viewlist \* for a list of all DCs in the forest. Run Nltest /DCList: <domain> for a list of all DCs in the domain.

  • To check DC and DNS health, run DCDiag /v to report errors on all DCs in the forest.

Add the global catalog to a domain controller in the forest root domain

A global catalog is required for these and other reasons:

  • To enable logons for users.

  • To enable the Net Logon service running on the DCs in each child domain to register and remove records on the DNS server in the root domain.

    Although it is preferred that the forest root DC become a global catalog, it is possible to elect any of the restored DCs to become a global catalog.

Note

A DC will not be advertised as a global catalog server until it has completed a full synchronization of all directory partitions in the forest. Therefore, the DC should be forced to replicate with each of the restored DCs in the forest.

Monitor the Directory Service event log in Event Viewer for event ID 1119, which indicates that this DC is a global catalog server, or verify the following registry key has a value of 1:

HKLM\System\CurrentControlSet\Services\NTDS\Parameters\Global Catalog Promotion Complete

For more information, see Adding the global catalog.

At this stage you should have a stable forest, with one DC for each domain and one global catalog in the forest. You should make a new backup of each of the DCs that you have just restored. You can now begin to redeploy other DCs in the forest by installing AD DS.

Next Steps