WLBS problem on C4506 switch

[Newport_s]

User has a Windows Load Balancing Server with 2 nodes setup in a server farm. We recently changed out the 5500 with a 4506 with multiple vlans. The 5500 was running vlan 1 only (spanning tree was enabled) and the setup worked. His pings from one node to another take over a minute to start responding. The WLBS mgt software also doesn't see the other node from both sides during the ICMP timeouts. Once the ICMP starts working between nodes the cluster is manageable from both nodes. User traffic is not affected, the problem only exists within the Balancing cluster and mgt software. All admin and public NICs on the WLBS are on the same vlan (so is the switch management). The ports he connects through are set to portfast but are running spanning-tree in the vlan. I don't see anything strange on the switch.

I've seen this symptom discussed while browsing MS admin websites, but don't find a solution. Any help someone can provide will be appreciated.

Thanks,

[bwilmoth]

The following document addresses a problem where Windows Load Balancing Server (WLBS) causes slow traffic through switches.

http://www.cisco.com/warp/public/473/ms_wlbs.html

[Newport_s]

The issue isn't slowness, but that the NLB management software isn't seeing the both nodes. This is affecting WNLB management functionality.

[dave.keith]

There are a lot of, um, 'issues' with MS NLB, and they change depending on what OS and what implimentation (unicast vs multicast for example). Have you looked at MS for resolutions or discussions of your problems ? If not, have a look at

http://support.microsoft.com/kb/898867

and search for others.

Dave

[Newport_s]

Yup, we've already been down this road. The administrator has set the MaskSourceMAC to "1" since day one per Microsoft "how to" documentation. We've been using their support area due to the lack of information on Cisco's website concerning the issue. The server administrator is going to setup a separate network for his management NICs which should solve the problem. I'll update this post once we get resolution.

Thanks for your response Dave.