[c-nsp] Cisco 3012 IBM Blade Switch Configurations?

[David Hughes]

On 17/02/2009, at 5:21 AM, ChrisSerafin wrote:

> That brings up a good point about STP. If I have 1 etherchannel  
> going from each switch blade to each upstream switch, will the  
> switches detect the loops or do I need to manually configure this?  
> Thanks for your comments!

The blade switches operate as separate switches and can be viewed as  
if they were your normal 29xx or 35xx access switch depending on the  
CIGESM you buy (well, with a couple of caveats about their management  
interfaces).  If the switches are connected to your network via a  
single etherchannel then there's no possibility of a loop - other than  
a mis-configured etherchannel.  You don't get loops but you don't get  
any redundancy at the network layer.  In this setup I assume you are  
using some form of NIC teaming and failover on the blades themselves  
for redundancy.  That's not going to pick up a failed uplink on your  
primary switching path as most teaming drivers only look at link state  
on the NIC.

If you are following a "normal" dual attached model where each access  
switch (be it a top of rack switch or a blade switch module) is  
uplinked to 2 different switches then you get link redundancy but  
naturally need STP.  If you are running dense virtualisation or vm  
server farms then the STP you should be running is MST.


Thanks

David
...