General discussion

Locked

Dual Nics in a cisco switch environment

By mwcombs ·
Scenerio:
Does anyone know the proper configuration/design when using dual nics in a cisco switch environment. For example, if you have two nics in an hp server configured for either fault tolerance or load balancing. There are a few ways you can go about doing this on the cisco side. You can have both network cables plug into the same switch and I can create an etherchannel where both ports are bound together for true aggregate bandwidth. The other preferred option is to have one nic plug into one switch and the other nic into a second switch. This way if either switch or nic fails you still maintain connectivity.

Issue/question:
If you go with the 2nd option then you need to be careful because when you use fault tolerance most of the times you have to use a virtual mac address. This virtual mac address will show up on both switch ports. As I'm sure you all know switches forward frames according to their layer 2 address. Is there any special configurations that you need to take into consideration when using dual nics across different switches. I think this is causing a big issue with spanning tree seeing that the same mac is on two different switch ports. This also becomes an even bigger issue if portfast is turned on at both ports. Any design help would be greatly appreciated.

- Mark

This conversation is currently closed to new comments.

5 total posts (Page 1 of 1)  
| Thread display: Collapse - | Expand +

All Comments

Collapse -

not sure I get it

by CG IT In reply to Dual Nics in a cisco swit ...

Not sure the actual MTBF on network cards but 50,000 hrs or 5.7 years tends to be the norm. Cisco equipment MTBF often runs the same if not double that. Their stuff just doesn't break down often.

for fault tolerance or load balancing with servers [which would tend to fail first than a NIC or Cisco switch], might consider server redundancy on subnets.

If a switch goes down on a subnet, then all hosts on the subnet are down as well. So until you get the switch back up, those hosts on that subnet [or VLAN] are offline. Having the server span subnets with dual nics only provides redundancy for the subnet uneffected by the disaster.

If the server NIC on a subnet does down, then hosts are left without a server on their subnet. you can provide redundancy for that subnet when the server goes down by specifing other servers on other connected subnets in which hosts would go to if the prefered server is down.

Collapse -

Not looking at the whole picture?

by apaxson In reply to not sure I get it

I agree, that Cisco hardware rarely fails. But I don't think you guys are looking at the entire picture.

You build redundancy for many reasons. Only one, of which, is direct hardware failure. What if you need to take the hardware out of service? Possibly upgrade the IOS? What if an intermediary uplink fails to the core?

It's alot cheaper to purchase a 2nd switch, than to build a fully redundant server with hot-swap capability (active/standby). The application may not even be redundant capable.

"....If a switch goes down on a subnet, then all hosts on the subnet are down as well..."

You are confusing Layer2 with Layer3. If a switch goes down, the entire SUBNET is NOT down. Only those hosts connected to the switch.

Collapse -

Nic Teaming

by mwcombs In reply to Dual Nics in a cisco swit ...

I believe that improper NIC teaming is causing network issues. Primarly due to the fact that when you use teaming most of the time if not all you have to use a vitual mac address. This mac address cannot show up on two different Cisco switch ports without causing layer 2 issues. Is there a guide or a does someone have any best practice experience when using teaming in a Cisco siwtch envornment?

Collapse -

Best practice Dont do it

by CG IT In reply to Nic Teaming

Cisco equipment usually doesn't fail and if it did, every host on the switch will lose connectivity so having a server "span" 2 switches for redundancy is just bad design and planning and not knowing your infrastructure hardware very well. If there is only 1 critical server everyone must get to therefore redundancy is required should it fail, best practice would be to not have a single point of failure. Get another server for redundancy.

Collapse -

Best practice

by mwcombs In reply to Best practice Dont do i ...

I agree with you that having teamed nics that are going to two different switches is a bit overkill. I'm not sure if the server team is doing it for redundancy for the server NICs or what. I have been managing cisco devices well over 10 years and I probably might have seen one or two nics/ports go bad. All I know is that spanning tree doesn't like it and for me to prove my case I need some type of best practice white paper so I can show to management. Unfortunately, the management staff does not listen very well. Thanks for the response.

Back to Networks Forum
5 total posts (Page 1 of 1)  

Related Discussions

Related Forums