L2 etherchannel and STP

[eduardonpinto]

Hi all,

I have a couple of questions concerning L2 etherchannels. Can you please help?

1- Is there a way to increase drastically the STP port cost of an unbundled etherchannel interface, forcing the traffic to be switched to another interface?

As an example, imagine you have a direct etherchannel link at 2x1Gbps (STP cost 3) and one of the links fails (STP cost increases to 4). I would like to increase this cost to 10 when the failure happens...is this possible?

2- Is there some sort of dampening system (BGP-like) for etherchannels?

Many thanks in advance,

Eduardo

[Giuseppe Larosa]

Hello Eduardo,

1)

unfortunately as far as I know, there is no command for this

spanning-tree cost accepts only one value in manual configuration

2)

I would recommend to use LACP if possible to have bundles that can detect neighbor state changes on each member link

Hope to help

Giuseppe

[eduardonpinto]

Hi Giuseppe,

thank you for your reply.

Concerning 2) I thought PAgP had the same feature: it can also detect an individual link failure, no? I'm not sure I understood your observation...

Regards

Eduardo

[Giuseppe Larosa]

Hello Eduardo,

I'm suggesting LACP because it is standard based.

PAGP is similar but Cisco proprietary.

We often have to connect servers with multiple NICs and we are fine with LACP.

You can use PAGP if you like with similar results as you suggest.

Hope to help

Giuseppe

[eduardonpinto]

Giuseppe,

I will use PAgP since I'm only dealing with cisco boxes.

It seems to me strange that no one ever requested the feature 1) to Cisco...imagine you have 1,2 Gbps of traffic running on a 2x1Gbps etherchannel directly between 2 switches. Imagine you also have an alternate path also with 2x1Gbps etherchannels but with an extra hop in between. It seems natural to me that you would like your traffic to cross the link with more hops but with the required bandwidth then having 200Mbps of traffic dropped on the etherchannel with the failure....

Am I missing something here??

Regards,

Eduardo

[eduardonpinto]

I was indeed missing something: the min-links feature in LACP.

[Giuseppe Larosa]

Hello Eduardo,

this is the key point you can force an LACP bundle to be not used if there is not a minimal number of links up.

forcing min-links = 2 would cause the bundle to go down when one member link fails.

This may be near to what you want.

I would say LACP provides more features.

I had thought of this feature but I hadn't checked documentation so I didn't write about it.

Hope to help

Giuseppe