All About Windows 2000 DNS, Sites, and Replication.

By Russ Bishop

http://www.boneville.net/lanadmin/

 

 

 

 

 

 

            DNS is the core name resolution system for Windows 2000, which is a refreshing change from the WINS days of yore. However, this means that DNS must be setup and working properly for your Windows 2000 Pro clients to login, replication to occur, and the sites to work properly (that is, each client logs into a DC in its site. This depends on sites being correctly specified and setup in AD and DNS.) First I’ll describe how to setup a basic DNS system, then deal with some specific issues.

 

 

            Every Domain Controller in your network ought to run DNS. There is absolutely no good reason to do otherwise. Besides giving you good load-balancing capabilities, you also get fault tolerance. DHCP can issue up to three DNS servers to each client, so take advantage of it. Have DHCP running at a specific location issue the Primary DNS to be the DC at that location. If there are two DCs, have it point the Secondary to another DC. Have the third server point to a DC over your WAN, if available. This helps ensure you don’t have a single point of failure on the client side.

 

            Setting up DNS on a DC isn’t very difficult. Just install the service, and set it up to be “Active-Directory Integrated.” In the DNS Server properties enable scavenging of stale records for every seven days. Also setup a forwarder to the Unix DNS server. This ensures that any DNS names that we cannot resolve (which will be anything outside of msad.company.net) gets forwarded to the Unix boxes, which know about the outside world. The properties here are on a per-server basis, so you might want to setup these settings on each server that runs DNS.

 

            Remember: In the TCP/IP Properties for the Domain Controller, setup the Primary DNS server to point to the DC itself. This step is VERY IMPORTANT. Let me say that again, this step is VERY IMPORTANT. Do not EVER use a 127.0.0.1 address; Always use the machine’s external address. You can fill in the secondary address if you wish, but it isn’t required. I recommend setting the secondary address to the Unix DNS Server IP if anything.

 

            Now create the “msad.company.net” forward lookup zone. You will want to set it to allow Dynamic Updates (not “only secured.” There are specific issues related to all that which would take 10 pages to explain, and that don’t apply to our situation.) Go to the Zone Transfers and select “only to servers on the Name Servers tab”. Then, go to the name servers tab. If you have any other DCs running DNS, they should appear here. Add the UNIX DNS server to the list. This should be all that is required to get a basic DNS setup going. Most other issues, like reverse lookup zones, etc are only conveniences, and are not required to allow clients to login or to perform replication. The settings here are replicated to every DC running DNS, so you can set it once and forget about it.

 

Note: if you see another forward lookup zone, such as  .”, delete it. The only forward lookup zone that should exist is the one holding the domain in question.

 

 

            The next step is setting up Sites, in order to facilitate replication and intelligent client login. Open up the Sites & Services control panel. It is worthy to note that the site you see called “Default First Site” is the site that catches all computers which have an IP address in a range that isn’t assigned to any other site. It is a catch-all for anything and everything, at least at first. Later, after multiple sites are setup, clients are liable to randomly pick a site if their subnet doesn’t fall into any specific assigned range. It would be a very good idea to ask the router guy or check your DHCP configuration and find out ALL the subnets in use, including static ones and ones that are assigned to servers, printers, and such, then enter them into the proper sites.

 

            The first thing to do is rename this default site to a location identifier, typically the main location. Go into the subnets and add a subnet that corresponds to this site. You must then assign the subnet to that site. Repeat this process for each of your locations that are connected at less than 10MBps. “Sites” linked with 10Mbps or more do not need to be defined in AD; they can be part of the site they are connected to. For example, if Site A is the main location, Site B is connected to it with a T1, and Site C has a 11Mbps wireless lan to Site B, there is no need to define Site C in AD. Just make the subnets in Site C part of Site B.

 

Once this is complete, you must start creating Site Links (Intersite-transports\IP).  These tell AD how expensive a trip across the wire is, as well as what sites have wires to other sites. A suggested rating might be a cost of 100 for a T1, 500 for 512k-256k, and 750 for 128k. These costs just tell AD how to rate a line if multiple ones are available. For example, if a site has two links, one with a cost of 100 and the other with a cost of 200, AD will usually send most of the traffic down the 100 link and just a small amount down the 200 link. If a site only has one link to it, then the cost doesn’t really apply, other than to tell it how much processor time it should waste trying to compress the data before sending. You can also select a replication frequency. Since only changes are sent, this data shouldn’t be very large. I’d recommend 15 minutes for all sites as a good replication time.

 

            Another important thing to note about sites is that you can specify a bridgehead server for the site. This means that all traffic coming in and out of that site will go through that server, rather than letting servers in one site randomly talk to servers in another site. If the bridgehead server is down, AD will randomly pick another machine in the site to act in its place temporarily. For a very large site, it might be best to pick a specific powerful machine to be the bridgehead. You will almost always want to use IP as the intersite transport; SMTP can only replicate the Configuration schemas, and not the actual data partition.

 

            Lastly, don’t forget to actually move the DCs to their specific sites when they are first promoted. If a DC’s subnet doesn’t match the subnet for an existing site, the machine gets dropped into the Default First Site usually.

 

 

           

 

            Once sites and DNS are configured, you need to verify that Replication is working. You should give the servers at least an hour to replicate amongst themselves and reconfigure their information to adjust to the new site information you have provided.

 

There is a nifty utility in the Support Tools folder on the Server CD-ROM, called “Replication Monitor.” Install the Support Tools on every DC. That previous sentence was a command and not an option, and one day you will be glad you did. Once they are installed, run “replmon”. Click the Edit menu, and add a server. Repeat for each of your DCs. Now look at the status for each server and see if any are not replicating. If a server shows that it isn’t replicating with one of its partners, there are several issues to address:

 

A.     Check to see that the servers can ping each other.

B.     Make sure that both servers’ DNS entries for each other point to the proper IP addresses

C.    If server A says it replicated fine, but server B says it couldn’t contact Server A, check the DNS setup on Server B. Chances are it has a record for Server A pointing to the wrong place.

D.    Run Netdiag and see if it reports any errors or problems.

 

Once replication shows up as working properly, you should have a basic AD framework setup, and you can move on to other specific issues.

 

 

Procedures for changing a Server’s IP Address

 

Once DNS and replication are setup, it is generally a bad idea to change a servers IP address (at least according to Microsoft). Just be sure that is what you really want to do before starting the process. It is a bit kin to changing the Internal IPX number of A Novell server, but it can be done.

 

1.      Change the Server’s IP address

2.      Stop the NETLOGON service.

3.      Rename or delete SYSTEM32\CONFIG\NETLOGON.DNS and NETLOGON.DNB

4.      Restart the NETLOGON service and run “IPconfig /registerDNS

5.      Go to one of the other DCs and verify that its DNS is now pointing to the new IP address of the server. If not, change the records manually and give it 15 minutes to replicate the DNS changes out.

6.      Run REPLMON and make sure that replication is working now. You may have to wait a little while for things to straighten out. Give it an hour or two if necessary.