The Windows NT Load Balancing Service (WLBS) and Windows 2000 Network Load Balancing (NLB) does not provide failover at the program, service, or port level, because WLBS and NLB are TCP/IP load balancers. They are NOT designed to detect that a service or program is not working properly.

If, for example, IIS fails on a cluster node, but WLBS or NLB is still running on the node, port 80 or web packets are still routed to the WLBS or NLB node. The IIS service does NOT failover due to design limitations.

When a program or service fails, the cause is usually more complex than the service is not responding on a port.

If you could monitor the health of your programs and/or services and detect 'failure', you could then send a command to stop WLBS or NLB on the 'failed' node. This will prevent any further traffic until the node rejoins the cluster:

1. You can use NETSVC.EXE from the Resource kits, Windows NT 4.0 or Windows 2000, to monitor the service and/or start/Stop/pause a service, either locally or remotely. See tip 0770.

2. You can use Httpmon.exe on a Web server to monitor the status of the Web service. See How to Configure HTTPMon to Monitor NLB or WLBS Web sites.

3. You can use Microsoft Internet Information Server (IIS) Exception Monitor to monitor an IIS 3.0 and IIS 4.0 server.