G
G
GSnick2015-03-05 22:03:23
Cisco
GSnick, 2015-03-05 22:03:23

How to configure 2 NICs in active/backup mode if they are connected to 2 different switches?

There is the following scheme: It is
ca933e0ae526424d99607366f8d89d02.png
necessary that when the link Port-channel 1 falls, everything goes to Port-channel 2 . Now active / backup bonding is configured on the blade itself from arp-target to the gateway, i.e. 192.168.9.9. But for some reason, arp requests from the eth0 side receive responses, but there are no responses from the eth1 side . Although if you manually send the eth0 interface down, then everything works correctly through eth1 . This still needs to be dealt with, but therefore the question arose, is it possible to configure this with the help of blade switches?
The IBM BladeCenter H blade itself, with two Cisco WS-CBS3012 switches. The description for the switches says:
The Trunk Failover feature allows rapid failover to the redundant switch in the blade enclosure if all uplinks from the primary switch fail. When the uplinks fail, the switch shuts down the ports connected to the blade servers and lets network interface card (NIC) teaming software direct traffic to the redundant switch. This feature is also known as Link State Tracking.

I also heard about Teaming active/backup, but I can't figure out if that's what it is? Can anyone point me in which direction to look?

Answer the question

In order to leave comments, you need to log in

1 answer(s)
A
Armenian Radio, 2015-03-05
@GSnick

Teaming (Etherchannel, link aggregation, the HP Gbe2Ec switch generally called it a trunk, crack me up with a marketer!) Works with only one switch (for Cisco stacks - with one device from the stack) - they took a bunch of wires and connected two points strictly to each other. You can't do it the way you want - so that half of the etherchannel goes to one device, and the other half goes to the other.
What is written in the manual actually means that if all the uplinks from one switch fell, it puts ports that look inside the box with blades. Having discovered such a trick, the OS inside the blade should act on its own.
That is, you must start a channel inside the blade by combining eth0 and eth1, then eth0 goes to the left switch, eth1 to the right, then the left and right switches are connected to one ofrouters - either left or right. In fact, through the box switches, etherchannel will transit, it will begin inside the blade, and end on one of the routers.
To do exactly what you want, you will have to use at least four interfaces, repeating the above scheme for eth2 and eth3.
There is still an option with STP, but this use of STP is a crutch.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question