02-26-2010 01:24 PM
Hi there,
I hope somebody can help us. We recently bought some HP c7000 Blade Enclosures with 2 Cisco 3120x switches inside. On the blades we run Windows Server 2003 and 2008 and use teaming.
When a switch needs to be replaced after an hardware failure the new switch will be unconfigured placed back into the c7000 enclosure. The ports of the 3120x are not shut so the Windows servers will start transmit traffic through that ports. The problem is that the switches are not yet properly configured.
How can we solve this problem? Or do we need a test enclosure where we can shut the ports before putting back in the porduction enclosure?
How do you handle hardware failure's?
Thanks for the help!
02-26-2010 02:02 PM
Hi,
In order for your 3120 switches to work with you servers you need to do NIC teaming on the server side and also build an etherchanned on the switch side with your 2 switches stacked together so when you lose one switch the other one take over the function and start forwarding packets.
HTH
Reza
02-26-2010 02:10 PM
Thanks for the answer but we dont want to stack the switches, is there a possibility without stacking the switches?
02-26-2010 02:15 PM
Unfortunately not. If you don't stack them then teaming is not going to do any good. When you stack them you have NIC redundancy and also switch redundancy. We use 3110x with IBM chassis and it works well. Why don't want to stack them?
Reza
02-26-2010 02:21 PM
We use the teaming option from HP in Windows Server 2003 / 2008, we are not using etherchannels, so we use both switches as individual switches.
Somebody already figured out if we can stack but that was not working with our design.
So we need a solutions for the replacement.
02-27-2010 03:50 AM
Hi,
I've a doubt and a suggestion, but I'm not an expert in DC, so bear with me.
First, teaming and etherchannel are different names for the same, ethernet inverse multiplexing,
standardized in 802.3ad (LACP)
What I see as a problem with defining it in only one side is that your MACs will be moving
from one switch to the other, but nothing serious. You will not have link balancing or HA controlled from the switch side,
but as a response to whatever the host sends. So what is the problem with your design that prevents stacking and
multi chassis ethernet bonding is something I'd like to know.
Now, with your reinsertion trouble, I guess admin down all the ports on the host side before reinsertion
is one option.
As a more risky option, you can actually insert the switch and prevent it from booting by breaking into rommon.
The switch supports xmodem, so you can upload the configuration while the switch is non functional.
Good luck.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide