Cross-stack etherchannel - one port always getting suspended

[inaiate87]

Hi,

I have configured a cross-stack etherchannel in a Cisco 3850; ports get added to the port-channel but one port always get suspended. If I try to add one when i add the other it gets suspended; if i change the order in which I add the ports to the Port-Channel the last port that I add always get suspended. I check the configurations and all parameters are the same, speed, duplex, everything. 

The stack is comprised of two switches.

Same is appearing to happening in a VSS cluster of two 6880 Catalyst switches.

Can anyone help?

[Giuseppe Larosa]

Hello inaiate87,

are you using LACP on the bundle?

I mean mode active/active or active/passive.

Can you check in the log for LACP related messages ?

show log | inc LACP

or

shw log | inc SUSPEND

You may need to shut both member links and re-enable them at the same time.

Hope to help

Giuseppe

[inaiate87]

Sorry but what is the logging level to see LACP messages? I've configured "logging buffered warnings" but that doesn't show LACP messages.
I'm using LACP; i've tried to use ON mode, but what is happening is that traffic is only being forwarded out of one interface of the port-channel.

[Reza Sharifi]

Hi,

What are these interfaces connected to? Are these connected to servers or other switches? If to servers, are the servers configured with NIC teaming?

One thing that works well sometimes is this:

delete the po interface completely

delete the po number under the physical interfaces. 

Once all deleted try this:

under the physical interface, add the interface to the po group (this will create the po interface as well)

do the same for the other interface

now, go under the po interface and un-shut it. This should bring up both the logical and physical interfaces. This is if everything else on the opposite side is configured correctly.

HTH

[inaiate87]

Please check the attached drawing. The goal is on the Cisco 3850 bundle the interfaces Gi1/1/1 and Gi2/1/1. And on the 6880 bundle the Gi1/5/1 and Gi2/5/1.

I've already tried your suggestion, but with the same behavior.