…How To…  

 

Within Windows 2003 server you have the option of Network Load Balancing (NLB). This can be used for load balancing your servers and to a certain extent failover. In this instance we are going to use it for three servers running Citrix Web Interface. The idea being if one of the servers goes down then the end users will get redirected to one of the other two.

There are some disadvantages to this ie it’s not true failover, if users are connected when one of the servers go down they will be kicked out of their session, they can reconnect straight away however and will be directed to one of the other servers.

First of all we need to establish how your servers networking is setup. You can use NLB on a single network card but as most servers I come across nowadays have two this document will describe how to set it up for two.

Commonly nowadays you find that if there are two network cards in your server they are in a “Team” using the manufacturers software. You can use NLB with teamed network cards but also, it can cause problems. If you wish to setup keeping your teamed cards please see Microsoft Article 278431 for further information.

For this article I have made the following assumptions:

Your network cards are installed correctly.

They both have a static ip address.

DNS is configured correctly on your network.

You have an available IP address to assign to your cluster.

The first thing to do is to open up Network load balancing Manager. This can be found under administrative tools. As long as the admin tools are installed this can be done from your PC, it doesn’t have to be on a server.

When you have located and opened NLB Manager click Cluster on the menu at the top and then New. You will be presented with the “Cluster Parameters” window as below. The ip address to fill in here is the free one you are going to assign to the cluster. In this example we are using 192.168.1.11 with the standard class C subnet mask. The full internet name is just a name so it doesn’t really matter what you put here but you might as well call it what you want the DNS name is going to be at the end. We will leave in Unicast mode and not allow remote control.
 

We only want one cluster so on the next screen we leave the Additional cluster ip addresses and click next.

The next window refers to port rules  and again we can leave as default in this instance.

The following screen asks us to connect to just one of the hosts that will be part of this new cluster. You can fill in either the server name or one of the ip addresses on this server. When you click connect both interfaces available on that server should show in the “Interfaces available for configuring a new cluster” box. Select the address you have dedicated to being your “Virtual address” and click next. Check the final window as is and click Finish.

Your cluster is now created and will be displayed using the name you specified during setup. If you right click this name you can now select add host to cluster where you will see a familiar window. Add the name or ip address of one of the two remaining servers you want in the cluster and select the correct interface as before. When you have done this repeat the procedure for the final interface and your cluster is setup and ready to go.

Finally you need to add a DNS record to resolve to the cluster ip address. For example we will add one for johncluster.local to resolve to 192.168.1.11.

Testing the cluster   

Open up two command prompts. In the first one type “ping clusterip –t” and in the second one type “ping server1 –t” and hit return. The command prompts should both be continuously replying. Shutdown and restart server1 and the cluster ping should reply all the time whereas the server1 ping will timeout several times before replying again when it comes back up. When the server has come back up try shutting down and restarting the other two at the same time and again the clusterip ping should reply constantly even though only one of the servers is up.