There are 4 3120 blade switches on 2 HP c7000 enclosures uplinked to a 3COM switch. These switces are just Layer 2 switches whose L3 vlan resides on the 3com switches.
There are a pair of blade servers being loadbalanced using Microsoft NLB. These servers have blades in two different enclosures whose blade switches are uplinked.
The uplinks are unusually access ports instead of trunk ports.
For Microsoft NLB, the technology being used is Multicast VIP, for which we have
configured a static ARP entry for the VIP and also created a static MAC address static entry and bound it to the uplink ports connecting to the blade enclosures.
The configuration done on the 3com switch is similar to the one we see in the following Cisco link for MS NLB:
The requirement is that if i disable the NIC of one of the blade servers in the NLB cluster, the other blade server
should be accessible through the VIP address automatically. But thats not happening automatically. Its happening manually though if i clear the arp on the 3com switch or if i shutdown both the NIC's of both the blade servers in the NLB cluster and re-enable the NIC on the blade server that i want to bring up.
Thats for sure not a desirable solution and i am seeking for an automatic way of doing that. If i say that the Cisco Blade switch is a pure L2 and nothing else, that does not appear a valid answer here as the customer has Microsoft NLB working correctly if he connects 2 standalone servers to the 3COM switch directly.
Does anybody think that we can really do something here on the Cisco Blade switches to send something to the effect of a gratitutous arp to update the mac address table of the 3com switch so that it sends data through the working uplink.
Note: The HP blade enclosures are not connected to each other directly.
Thanks a lot