[c-nsp] Problem with Microsoft NLB on Server 2008 running in Multicast mode

Tim Durack tdurack at gmail.com
Thu Mar 25 21:53:56 EDT 2010


On Thu, Mar 25, 2010 at 6:20 PM, Asbjorn Hojmark - Lists
<lists at hojmark.org> wrote:
> On Thu, 25 Mar 2010 22:52:07 +0100, you wrote:
>
>> http://www.cisco.com/en/US/customer/products/hw/switches/ps708/products_configuration_example09186a0080a07203.shtml
>
> Yeah, that's the one I was thinking off.
>
>> Together with the static ARP it tends to make the whole system (cluster
>> + network) more fragile IMO. Too much manual configuration, too many
>> normally independent parts of the solution that have to match.
>
> I agree, it isn't pretty.
>
> One could use the 'unicast' model instead, but that lives on unicast
> flooding, so it requires very small and very controlled VLANs, and it
> breaks with vSwitches in there.
>
> Or one could use a hardware loadbalancer...

We've suffered through several MS-NLB clusters for the last couple of
years. Nothing but trouble.

Static arp plus disable dhcp-snooping is required. We've broken it
several times, with vrf migrations and what not.

I would strongly suggest staying away from MS-NLB.

-- 
Tim:>


More information about the cisco-nsp mailing list