Microsoft NLB - Multicast static arp refused by IPSO ?!

[nooon]

One will say it's Microsoft that does not follow RFC's
One will say *BSD is just way to much strict.

However, I need to "static arp" for a Microsoft multicast NLB cluster.

I tried from Voyager - The entry is added to the web page, however an
arp -a shows no entry...

So I decided to add it manually and then :
firewall[admin]# arp -s 10.20.13.73 3:bf:84:ab:d:49
writing to routing socket: Invalid argument
firewall[admin]# arp -s 10.20.13.73 3:bf:84:ab:d:49 temp
writing to routing socket: Invalid argument
Jan 25 18:05:26 firewall [LOG_ERR] kernel: arp: ether address is broadcast or multicast for IP address 10.20.13.73!

Since I'm running a pretty old IPSO 3.6 there, I wanted to know if some of you running more recent 3.[7-9] or 4.[0-1], manages to add a multicast mac@ to the static entries, for ex:
arp -s you.r.sub.net 3:bf:84:ab:d:49

and/or if there is any way to override this..... or should we revert to MS Unicast NLB ?

cheers

PS : I know that Checkpoint or Cisco CSS loadbalancing is better than NLB. off-topic here please :))

[northlandboy]

Try going config -> ARP -> Accept multicast ARP replies "On"

That should do the trick I think

(Oh and I just checked that option is on IPSO 3.6)

[nooon]

hey North !

I found the same in the same time. Works fine !

However, all members from the microsoft cluster responds to ping at the same time, creating a [Dup !] on the ping reply.
Windows technotes says this is a normal behavior.

I think multicast nlb implementation stinks.

[northlandboy]

Great minds, and all that.

Multicast is one way of solving a problem. Nokia uses multicast for one version of clustering. I think it's more elegant than forwarding mode, but the problem is that multicast support is still pretty patchy.