The list in the following table provides the definitions of the names associated with registered SRV records. It also describes the lookup criteria supported by each record and the checks performed by Net Logon as each record is registered.

In the descriptions of registered SRV records, DnsDomainName refers to the DNS domain name that is used during creation of the domain controller when the domain tree is joined or created (that is, when the computer runs the Active Directory Installation Wizard). DnsForestName refers to the DNS domain name of the forest root domain.

An owner name is the name of the DNS node to which a resource record pertains. These resource records are used by domain controller Locator. The following table lists the owner names of the SRV records that are registered by Net Logon.

SRV Records That Are Registered by Net Logon

In addition to the SRV records listed in the table above, Net Logon also registers a DNS alias (CNAME) record for use by Active Directory replication, in the format: DsaGuid._msdcs.DnsForestName. The Locator does not use this record. This record enables a client to locate any domain controller in the forest by looking up an A record. The only information that is known about the domain controller is the GUID of the directory system agent (DSA) object for the domain controller and the name of the forest in which the domain controller is located. This record is used to facilitate renaming a domain controller.

If multiple domain controllers have the same criteria, multiple records exist with the same owner name. A client that is looking for a domain controller with specific criteria would receive all the applicable records from the DNS server. The client would pick one of the returned records to select a domain controller, as described in RFC 2782.

The following information is also included in an SRV record.

The algorithm by which clients interpret and select among SRV resource records is defined in RFC 2782, “A DNS RR for specifying the location of services (DNS SRV).”

Net Logon registers the following DNS A records for the use of LDAP clients that do not support DNS SRV records (that is, clients that are non-SRV-aware). Locator does not use these records.

Host (A) Records Registered by Net Logon

The following example illustrates the combined information that is contained in A resource records and SRV resource records.

A domain controller named Phoenix in the domain contoso.com has an IP address of 157.55.81.157. It registers the following A records and SRV records with DNS:

phoenix.contoso.com   A   157.55.81.157
_ldap._tcp.contoso.com    SRV  0 0 389 phoenix.contoso.com
_kerberos._tcp.contoso.com   SRV  0 0 88 phoenix.contoso.com
_ldap._tcp.dc._msdcs.contoso.com  SRV  0 0 389 phoenix.contoso.com
_kerberos._tcp.dc._msdcs.contoso.com  SRV  0 0 88 phoenix.contoso.com.

When the appropriate SRV records and A records are in place, a DNS query for _ldap._tcp.dc._msdcs.contoso.com returns the names and addresses of all domain controllers in the domain.

For more information about A records, SRV records, DNS, and dynamic updates, see “DNS Protocol” and “DNS Physical Structure” in How DNS Works.

Active Directory supports dynamic registration of domain controller addresses in DNS. After Active Directory has been installed during domain controller creation, the Net Logon service dynamically creates records in the DNS database that are used to locate the server. Dynamic update (described in RFC 2136) is a recent addition to the DNS standard; this addition to the standard defines a protocol for dynamically updating a DNS server with new or changed resource record values. Before the advent of this new protocol, administrators had to manually create the records that are stored on DNS servers. The implementation of DNS server that is included with Windows Server 2003 supports dynamic updates, as does the Berkeley Internet Name Domain (BIND) version 8.x implementation of DNS.

Note

• By default, the Windows Server 2003 DNS Server service running on a domain controller is configured to accept secure dynamic update only.
  

Every Windows Server 2003–based domain controller dynamically registers SRV records in DNS. The SRV records enable servers to be located by service type (for example, LDAP) and protocol (for example, TCP). Because domain controllers are LDAP servers that communicate over TCP, SRV records can be used to find the DNS computer names of domain controllers. In addition to registering LDAP-specific SRV records, Net Logon also registers Kerberos v5 authentication protocol–specific SRV records to enable locating servers that run the Kerberos Key Distribution Center (KDC) service.

Every Windows Server 2003–based domain controller also dynamically registers a single host resource record (an A resource record) that contains the name of the domain (DnsDomainName) where the domain controller is and the IP address of the domain controller. The A resource record makes it possible for clients that do not recognize SRV records to locate a domain controller by means of a generic host lookup.

A domain controller registers its NetBIOS name (DomainName[1C]) by broadcasting or directing a NetBIOS name registration request to a NetBIOS name server, such as a WINS server. Registering the NetBIOS name makes it possible for Windows-based clients that are not DNS-enabled to find the domain controllers that are running Windows Server 2003, Windows 2000, Windows NT 4.0, or Windows NT 3.51. In this case, the client finds the domain controller by sending a Net Logon mailslot request that is based on the NetBIOS domain name.

Note

• NetBIOS recognizes domain controllers by the [1C] registration.
  

Each Windows Server 2003–based domain controller registers its DNS domain name on the DNS server and registers its NetBIOS name by using a transport-specific mechanism (for example, WINS). Thus, a DNS client locates a domain controller by querying DNS, and a NetBIOS client locates a domain controller by querying the appropriate transport-specific name service. Because the code for the DNS-compatible Locator and the Windows NT 4.0–compatible Locator is shared, both DNS clients and NetBIOS clients are supported.

The process that the Locator follows can be summarized as follows:

1. On the client (the computer that is locating the domain controller), the Locator is initiated as a remote procedure call (RPC) to the local Net Logon service. The Locator API (DsGetDcName) is implemented by the Net Logon service.
   
2. The client collects the information that is needed to select a domain controller and passes the information to the Net Logon service by using the DsGetDcName API.
   
3. The Net Logon service on the client uses the collected information to look up a domain controller for the specified domain in one of two ways:
   
   • For a DNS name, Net Logon queries DNS by using the DNS-compatible Locator — that is, DsGetDcName calls DnsQuery to read the SRV records and A records from DNS after it appends an appropriate string to the front of the domain name that specifies the SRV record.
     
   • For a single label name, Net Logon performs domain controller discovery by using the Windows NT 4.0–compatible Locator — that is, by using the transport-specific mechanism (for example, WINS).
     Note
     

In Windows NT 4.0 and earlier, “discovery” is a process for locating a domain controllerfor authentication in either the primary domain or a trusted domain.

1. The Net Logon service sends a datagram to the discovered domain controllers that register the name. For NetBIOS domain names, the datagram is implemented as a mailslot message. For DNS domain names, the datagram is implemented as an LDAP UDP search.
   
2. Each available domain controller responds to the datagram to indicate that it is currently operational and then returns the information to DsGetDcName.
   
3. The Net Logon service returns the information to the client from the domain controller that responds first.
   
4. The Net Logon service caches the domain controller information so that it is not necessary to repeat the discovery process for subsequent requests. Caching this information encourages the consistent use of the same domain controller and, thus, a consistent view of Active Directory.
   

During a search for a domain controller, the Locator attempts to find a domain controller in the site closest to the client. When the domain that is being sought is a Windows Server 2003 domain, the domain controller uses the information stored in Active Directory to determine the closest site. When the domain being sought is a Windows NT 4.0 domain, domain controller discovery occurs when the client starts and uses the first domain controller that it finds.

Each Windows Server 2003–based domain controller registers DNS records that indicate the site where the domain controller is located. The site name (the relative distinguished name of the site object in Active Directory) is registered in several records so that the various roles the domain controller might perform (for example, global catalog server or Kerberos server) can be associated with the domain controller’s site. When DNS is used, the Locator searches first for a site-specific DNS record before it begins to search for a DNS record that is not site-specific (thereby preferentially locating a domain controller in that site).

A client computer stores its own site information in the registry, but the computer is not necessarily located physically in the site associated with its IP address. For example, a portable computer that was moved to a new location could contact a domain controller in its home site, which is not the site to which the computer is currently connected. In this situation, the domain controller looks up the client site on the basis of the client IP address by comparing the address to the sites that are identified in Active Directory, and then returns the name of the site that is closest to the client. The client then updates the information in the registry.

The domain controller stores site information for the entire forest in the Configuration container. The domain controller uses the site information to check the IP address of the client computer against the list of subnets in the forest. In this way, the domain controller ascertains the name of the site in which the client is assumed to be located, or the site that is the closest match, and returns this information to the client.

A site is a collection of subnets that have high-speed connections. In Active Directory, a site is defined by a site object in the cn=Sites,cn=Configuration,dc=ForestRootDomain container. A subnet is an addressed segment within a site and is represented by an object in the cn=Subnets,cn=Sites,cn=Configuration,dc=ForestRootDomain container.

The site in which a domain controller is located is identified in the Configuration container by the domain controller object that is located within the cn=Servers container beneath the site object for a particular site. A domain controller can identify the site of a client by using the subnet object in the Sites container. Each subnet object has a siteObject property (“attribute”) that links it to a site object; the value of the siteObject property is the distinguished name of the site object. This link enables a domain controller to identify clients that have an IP address in the specified subnet as being in the specified site.

Subnet names in Active Directory take the form “network/bits masked” (for example, the subnet object 172.16.72.0/22 has a subnet of 172.16.72.0 and a 22-bit subnet mask). If this subnet had a siteObject property value that contained the distinguished name of the Seattle site object, all IP addresses in the 172.16.72.0/22 subnet would be considered to be in the Seattle site. The siteObject property is a single value, which implies that a single subnet maps to a single site. However, multiple subnet objects can be linked to the same site object. The directory administrator manually creates subnet objects and, hence, the siteObject property value.

The Configuration container (including all of the site and subnet objects in it) is replicated to all domain controllers in the forest. Therefore, any domain controller in the forest can identify the site in which a client is located, compare it to the site in which the domain controller is located, and indicate to the client whether that domain controller’s site is the closest site to the client.

For more information about networks, subnets, and subnet masks, see “How TCP/IP Works.”

During Net Logon startup, the Net Logon service on each domain controller enumerates the site objects in the Configuration container. Net Logon on each domain controller is also notified of any changes made to the site objects. Net Logon uses the site information to build an in-memory structure that is used to map IP addresses to site names.

When a client that is searching for a domain controller receives the list of domain controller IP addresses from DNS, the client begins querying the domain controllers in turn to find out which domain controller is available and appropriate. Active Directory intercepts the query, which contains the IP address of the client, and passes it to Net Logon on the domain controller. Net Logon looks up the client IP address in its subnet-to-site mapping table by finding the subnet object that most closely matches the client IP address and then returns the following information:

• The name of the site in which the client is located, or the site that most closely matches the client IP address.
  
• The name of the site in which the current domain controller is located.
  
• A bit that indicates whether the found domain controller is located (bit is set) or not located (bit is not set) in the site closest to the client.
  

The domain controller returns the information to the client. The response also contains various other pieces of information that describe the domain controller.

The client inspects the information to determine whether to try to find a better domain controller. The decision is made as follows:

• If the returned domain controller is in the closest site (the returned bit is set), the client uses that domain controller.
  
• If the client has already tried to find a domain controller in the site in which the domain controller claims the client is located, the client uses that domain controller.
  
• If the domain controller is not in the closest site, the client updates its site information and sends a new DNS query to find a new domain controller in the site. If the second query is successful, the client uses the new domain controller. If the second query fails, the client uses the original domain controller.
  
• If the domain that is being queried by a computer is the same as the domain to which the computer is joined, the site in which the computer resides (as reported by a domain controller) is stored in the computer registry. The client stores this site name in the DynamicSiteName registry entry in HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\ Netlogon\Parameters. Therefore, the DsGetSiteName API returns the site in which the computer is located.
  

You can override the dynamically determined value using the registry, but you should never change dynamically determined values. To override the dynamic site name, add the SiteName entry with the REG_SZ data type in HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Netlogon\Parameters. When a value is present for the SiteName entry, the DynamicSiteName entry is not used.

Note

• Editing the registry directly can have serious, unexpected consequences that can prevent the system from starting and require that you reinstall Windows Server 2003. There are programs available in Control Panel or Microsoft Management Console (MMC) for performing most administrative tasks. These programs provide safeguards that prevent you from entering conflicting settings or settings that are likely to degrade performance or damage your system. Registry editors bypass the standard safeguards that are provided by these administrative tools. Modifying the registry is recommended only when no administrative tool is available. Before you make changes to the registry, it is recommended that you back up any valuable data on the computer. For instructions about how to edit registry entries, see Help for the registry editor that you are using.
  

If the domain being located is the same as the domain to which the computer is joined and the computer has not physically moved to a different site since the last query, the dynamically determined site name in the registry is the actual site in which the computer is located. Thus, the client finds a domain controller in the correct site without having to retry the operation. On the other hand, if the site name in the registry is not the current site of the computer (for example, if the computer is portable), the domain controller location process serves to update the site information in the registry.

There is not necessarily a domain controller in every site. For various reasons, it is possible that no domain controller exists for a particular domain at the local site. By default, each domain controller checks all sites in the forest and then checks the replication cost matrix. A domain controller advertises itself (registers a site-related SRV record in DNS) in any site that does not have a domain controller for that domain and for which its site has the lowest-cost connections. This process ensures that every site has a domain controller that is defined by default for every domain in the forest, even if a site does not contain a domain controller for that domain. The domain controllers that are published in DNS are those from the closest site (as defined by the replication topology).

For example, given one domain and three sites, a domain controller for that domain might be located in two of the sites, but there might be no domain controller for the domain in the third site. Replication to the domain that does not have a domain controller in the third site might be too expensive in terms of cost or replication latency. To ensure that a domain controller can be located in the site closest to a client computer, if not the same site, Windows Server 2003 automatically attempts to register a domain controller in every site. The algorithm that is used to accomplish automatic site coverage determines how one site can cover another site when no domain controller exists in the second site.

Site coverage is determined according to site-link costs, and domain controllers register themselves in sites accordingly. For example, given one domain and sites A, B, and C, site A has no domain controllers for the domain. If a client in site A attempts to locate a domain controller, which domain controller should be returned? The answer depends on which site covers site A for the domain.

In the example, a site link exists between site A and both of the other sites — that is, the connections between domain controllers in site A, site B, and site C are configured for replication over site links in Active Directory Sites and Services. Costs are associated with site links based on the expense of transferring data over the connections. The administrator uses the speed of the connection between sites to assign a cost to the communication link, and replication uses the cost to establish the least expensive route for replication traffic.

Site A and site B are connected by site link AB. Site A and site C are connected by site link AC, with the following costs:

Site link AB cost = 50.

Site link AC cost = 100.

The link between site A and site C has a much higher cost than the link between site A and site C. The administrator configured this cost based on the expensive Integrated Services Digital Network (ISDN) line that connects site A and site C, and the administrator would prefer that resources in site B be used when possible. The site coverage algorithm (described in the next sub-section) ensures that a domain controller in site B registers itself as a domain controller for site A. In this way, clients in Site A that are looking for a domain controller find one from site B, instead of possibly finding one from site C.

During registration of SRV records in DNS, the following algorithm determines if the domain controllers should register site SRV records to designate themselves as preferred domain controllers in Active Directory sites where no domain controller exists for a particular domain.

For every domain controller in the forest, this procedure is followed:

• Build a list of target sites — sites that have no domain controllers for this domain (the domain of the current domain controller).
  
• Build a list of candidate sites — sites that have domain controllers for this domain.
  
• For every target site, follow these steps:
  
  1. Build a list of candidate sites of which this domain is a member. (If none, do nothing.)
     
  2. Of these, build a list of sites that have the lowest site link cost to the target site. (If none, do nothing.)
     
  3. If more than one, break ties (reduce this list to one candidate site) by choosing the site with the largest number of domain controllers.
     
  4. If more than one, break ties by choosing the site that is first alphabetically.
     
  5. Register target-site-specific SRV records for the domain controllers for this domain in the selected site.
     

If a domain member computer requests a domain controller while all domain controllers in its site are offline, the Locator necessarily returns a domain controller in a different site. The location of this domain controller is stored in the client cache. The cache lifetime is controlled by the CloseSiteTimeout entry in the registry.

In addition, the domain controller performs authentication, and a secure channel is set up. On subsequent location attempts, the lifetime of the cache and the lifetime of the secure channel are secondary to the location of a domain controller in the closest site.

If the domain controller that is stored in the client cache is not in a site that is close to the client, Net Logon attempts to find a close domain controller when either of the following events occurs:

• An interactive logon process uses pass-through authentication on the secure channel.
  
• The value in the CloseSiteTimeout registry entry has elapsed since the last attempt, and any other attempt is made to use the secure channel (for example, pass-through authentication of network logons).
  

Thus, Net Logon attempts to find a close domain controller only on demand. The default value of the CloseSiteTimeout period is 15 minutes; the maximum value is 49 days; the minimum value is 60 seconds. The implications of these settings are:

• If the time-out value is too large, a client never tries to find a close domain controller if there is not one available at startup.
  
• If the time-out value is too small, secure channel traffic is unnecessarily slowed down by discovery attempts.
  

Sometimes the client pings a domain controller and the client IP address cannot be found in the subnet-to-site mapping table. In this case, the domain controller returns a NULL site name, and the client uses the returned domain controller.

On the basis of parameters passed to Net Logon in the DsGetDcName API, the process of locating a domain controller proceeds in one of two ways:

• The DNS-compatible Locator is used if the domain name passed to DsGetDcName is a DNS-compatible name. The Net Logon service on the client looks up the name in DNS (by calling DnsQuery) after it appends an appropriate string to the front of the domain name. The DNS service supports a query for determining the set of domain controllers. If the client site name is known, the client DNS query specifies the site. DNS returns the IP addresses of domain controllers that match the DNS query. The client Net Logon service sends an LDAP UDP message to one or more of the domain controllers that have been returned by DNS in order to determine whether any of the specified domain controllers are running and support the specified domain.
  
• The Windows NT 4.0–compatible Locator is used if the domain name passed to DsGetDcName is a NetBIOS name. The Net Logon service on the client sends a transport-specific logon request query to locate a domain controller in a particular domain and then sends a mailslot message to one or more of the domain controllers to determine whether any of the domain controllers it found are running and support the specified domain.