Background
VMware have recommended for quite some time that we stick to multicast when configuring NLB (MS’s Network Load Balancing) where possible:
VMware recommends that you use multicast mode, because unicast mode forces the physical switches on the LAN to broadcast all Network Load Balancing traffic to every machine on the LAN.
If you need to use unicast, then to prevent port flooding you should change the Port Group’s “Notify Switches” policy to No – the default being Yes.
Windows 2008 R2 Failover Clustering
According to this white paper from Microsoft,
Multicast functionality has been discontinued in Windows Server 2008 failover clustering, and cluster communications now use User Datagram Protocol (UDP) unicast.
So Microsoft clustering gurus, does this mean for Window 2008 R2 Failover Clusters we should also change the “Notify Switches” policy off? Is the recommended setting for MS NLB clustering now applicable to MS’s latest version of MSCS?
“Notify Switches”=off is only needed for services that want to hide a common MAC address from the physical switches. Only unicast-NLB does that. It won’t “prevent port flooding”, it will _allow_ port flooding, which is exactly what unicast-NLB wants, in order to receive the same ethernet traffic on all nodes.
Regular failover clustering uses nodes with different MAC addresses, that do not need to hide their MAC address from the switches. Therefore, notify switches can remain on.
NB “Notify Switches”=off will increase the vMotion gap for regular VMs. It should only be set on a port group where only the unicast-NLB machine is connected.
Thanks Bert, yes you’re quiet right that unicast needs to see the traffic, not prevent the port flooding. I had worded it incorrectly. I’ve updated the post to reflect this.
Great to know that we can leave this turned on for non-NLB situations.
Thanks!