cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
6217
Views
8
Helpful
8
Replies

NIC Teaming Question

visitor68
Level 4
Level 4

hello:

I have 3 separate questions,.

(1)

(a)  Imagine I have a server with 2 NICs and I want to team them. If NIC 1  is connected to access switch 1 and NIC 2 to access switch 2, the only  NIC teaming methodology that can be used is one in which one NIC is active and the other is in standby. The reason is that you would have a  loop if both NICs were active (assuming L2 adjacency between switches). Is that part about the loop correct? If so, the way to get away from the loop is to have access switches, like  the Nexus 5000, that does vPC. Correct? SEE THE ATTACHED DIAGRAM.

     (b) If there is indeed a loop, which mechanism causes one NIC uplink to be active and the other to go standby. Its not STP. The NICs are connecte to  access ports.


(2) What about 2 separate NICs, as in  the first scenario, BUT the two access switches are stacked, like two  3750s?


(3) What  about the case where you have one dual-port NIC with  the same uplink connections to the two switches and no vPC or stacking?  Still be a loop?

May I please have 3 separate answers?

Thanks

8 Replies 8

Leo Laohoo
Hall of Fame
Hall of Fame

It would work if your NICs are connected to one VIRTUAL switch such as the 2960S, 2975, 3750/3750E/3750X switches.  This would also work if you have 4000/4500 and 6000/6500 switches.

I am a bit confused because I posed 3 separate scenarios..

Can you kindly respond to each scenario separately? Please note that question 1 has 2 parts.

Thanks

Hitesh Vinzoda
Level 4
Level 4

Hi

1)

(a)  Imagine I have a server with 2 NICs and I want to team them. If NIC 1  is connected to access switch 1 and NIC 2 to access switch 2, the only  NIC teaming methodology that can be used is one in which one NIC is active and the other is in standby. The reason is that you would have a  loop if both NICs were active (assuming L2 adjacency between switches). Is that part about the loop correct? If so, the way to get away from the loop is to have access switches, like  the Nexus 5000, that does vPC. Correct? SEE THE ATTACHED DIAGRAM.

No you will not have a loop as server doesnot work like switches.

     (b) If there is indeed a loop, which mechanism causes one NIC uplink to be active and the other to go standby. Its not STP. The NICs are connecte to  access ports.

I ll explain the working of Teaming as below for example Red is client and Blue(1.1.1.2) is server with Teaming

a) Red transmits a broadcast ARP request asking for Blue’s MAC address.
A user on Red issues the command ping 1.1.1.2 to initiate a PING to Blue. First, Red determines whether or
not Blue is on the same Layer 2 network. After Red has determined that Blue is on the same Layer 2 network, Red
must find out what Blue’s MAC address is. First, Red checks its own ARP cache for a MAC address entry
matching the IP address of 1.1.1.2. If Red does not have a static entry or an entry cached from a previous
conversation with Blue, then it must broadcast an ARP request frame on the network asking Blue to respond and
provide its MAC address. Red must broadcast this ARP request because without knowing Blue’s unique MAC
address, it has no way of sending a frame directly (unicast) to Blue.


b) Blue transmits a unicast ARP reply to Red, providing its MAC address.
Blue sees the ARP request (the frame is received on both the primary and non-primary teamed ports) because the
frame is broadcast on the network. However, the team discards all non-heartbeat frames incoming on nonprimary
ports, and responds with a unicast ARP reply to Red. The ARP reply is transmitted by the primary port (B).
In Blue’s ARP reply, Blue provides the MAC address of its teaming driver, which is the same as the current
primary port’s MAC address (B) (see “4-2-3 HP teaming and Layer 2 versus Layer 3 addresses”). Blue also notes
Red’s MAC address (A) and IP address (1.1.1.1) and enters them into its ARP cache. Red receives the reply and
enters the MAC address (B) and the IP address of Blue (1.1.1.2) into its own ARP cache.


c) Red transmits a unicast PING request to Blue using Blue’s destination MAC address.
Red can now create a PING request frame using Blue’s MAC address (B). Red sends the PING request to Blue.
Blue receives the frame on its primary port (B) and notices that a station with an IP address of 1.1.1.1 is asking
for it to respond.

d) Blue transmits a broadcast ARP request asking for Red’s MAC address.
Note: The following step may not occur if Blue’s ARP table still contains an entry for Red as a result of steps (a)
and (b).


Blue checks its ARP cache for a MAC address entry that matches 1.1.1.1. If Blue does not find one, then Blue
broadcasts an ARP request asking for Red’s MAC address.

e) Red transmits a unicast ARP reply to Blue providing its MAC address.
Note: The following step will not occur if step (d) does not take place.
Red sees the ARP request and transmits a unicast ARP reply directly to Blue providing its MAC address (A). Blue
receives the ARP reply and puts Red’s MAC address (A) and IP address (1.1.1.1) in its ARP cache.

f) Blue transmits a unicast PING reply to Red using Red’s destination MAC address.
Blue then transmits a unicast PING reply to Red using Red’s MAC address (A) and the user sees the PING reply
message printed on the screen. This completes the entire conversation.


(2) What about 2 separate NICs, as in  the first scenario, BUT the two access switches are stacked, like two  3750s?

The Team will work across stacks, different access switches

(3) What  about the case where you have one dual-port NIC with  the same uplink connections to the two switches and no vPC or stacking?  Still be a loop?

There would not be a loop as the mac address would be different

There is a good document on this link which explains the working of Team..

http://h20000.www2.hp.com/bc/docs/support/SupportManual/c01415139/c01415139.pdf

Hope this helps

Hitesh Vinzoda

Pls rate useful posts

Hitesh, thank you very much for that thoughtful post. The document is excellent, although it focuses only on HP. I am actually working with Broadcom NICs in Dell servers. But its OK, the document you gave me was very well written and presented. It allowed me to understand general concepts.

I would like you to elaborate on why connecting 2 separate server NICs to 2 different access switches will not cause a loop. If you look at the drawing, you will see why I am asking this question.

they will not create loop because they are NIC not a switch, they will not forward frames from one NIC to other NIC as the switches do.

HTH

Hitesh Vinzoda

Hitesh, I think the answer as to wther there would be a loop or not has to do with whether each NIC port has its own MAC address. That was the discerning factor that was slipping through my mental  cracks...and it makes perfect sense...

The Dell Power Edge servers use the  Broadcom NIC teaming utility called BACS. So, for example, with the PE  blade servers and a blade switch in both the A and A' switch fabric  slots, both ports on the same NIC --  but with 2 different MACs  -- can run active/ative to each switch. If the MACs were the same for both ports they  would only be able to run active standby, otherwise a loop would exist.

Ganesh Hariharan
VIP Alumni
VIP Alumni

hello:

I have 3 separate questions,.

(1)

(a) Imagine I have a server with 2 NICs and I want to team them. If NIC 1 is connected to access switch 1 and NIC 2 to access switch 2, the only NIC teaming methodology that can be used is one in which one NIC is active and the other is in standby. The reason is that you would have a  loop if both NICs were active (assuming L2 adjacency between switches). Is that part about the loop correct? If so, the way to get away from the loop is to have access switches, like  the Nexus 5000, that does vPC. Correct? SEE THE ATTACHED DIAGRAM.

    (b) If there is indeed a loop, which mechanism causes one NIC uplink to be active and the other to go standby. Its not STP. The NICs are connecte to  access ports.


(2) What about 2 separate NICs, as in  the first scenario, BUT the two access switches are stacked, like two  3750s?


(3) What  about the case where you have one dual-port NIC with  the same uplink connections to the two switches and no vPC or stacking?  Still be a loop?

May I please have 3 separate answers?

Thanks

                    Attachments:

Hi,

Check out the below link for genral recommedation on NIC teaming.

http://technet.microsoft.com/en-us/library/cc778400(WS.10).aspx

Hope to Help !!

Ganesh.H

Ganesh, I am not talking about Microsoft NLB and the link you gave me is about that. You also didnt answer any of my questions.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco