Applying QoS service policy to etherchannel physical interface problem.

Unanswered Question
Jun 21st, 2012
User Badges:

I have an access-layer 4510-E IOS XE 3.2.4.SG with Sup7-E.  It is connected to the distribution layer switch with 2 10gig interfaces in a portchannel.  When I tried to apply an output QoS service policy to int Te5/1 - one of the two phyical interfaces - the interface was removed from the portchannel with the error msg something like "EC-5-cannot bundle: Te5/1 is not compativle with Te6/1 and will be suspended (Te5/1 has a different output QoS policy than Te6/1)".  This caused a flapping in EIGRP and a disruption of traffic.


When I applied the output policy to T6/1 it brought T5/1 back into the bundle and EIGRP stopped flapping.


Even if the two physical interfaces are next to each other, like Te5/1 & Te5/2, and I use the interface range command to apply the QoS policy, both interfaces fall out of the port channel with a complete lose of EIGRP connectivity to the distribution switch. Eventually the physical interfaces rejoin the portchannel and EIGRP neighbors come up.


How can one apply QoS to multiple interfaces in an etherchannel and not cause the physical ports to fall out of the channel?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
jkeeffe Fri, 06/22/2012 - 07:08
User Badges:

That's a great reference and I've taken a look at it before, which helped me to configure the egress policy. What I'm beginning to understand, though, is there is no way to get around the fact that the physical interfaces will be temporarily pulled out of the logical port channel as the service policy is applied - since until I apply the service policy to the second phyical interface of the port channel, the two will be different. And the rule for a port channel to form is that all physical interfaces must be exactly the same.

Paolo Bevilacqua Sat, 06/23/2012 - 08:29
User Badges:
  • Super Gold, 25000 points or more
  • Hall of Fame,

    Founding Member

Which Qos? You don;t need QoS on such fast interfaces

paul.jerome1 Mon, 03/20/2017 - 06:22
User Badges:

The issue is caused by the difference in config between the physical interfaces. The L2 config o f the interfaces needs to be identical for them to be able to bundle.


The way I found around it was to admin shut one interface so there is only one active interface in the channel group, apply the QoS policy to that. Because there  are no other active interfaces it doesn't cause and conflict. Then you apply the Policy to the shut down interface so it matches the active one, and once the config matches you un-shut that interface and it can then join the channel group as normal.


I'd also maybe move off that IOS version.  I've had big issues with port flapping on L2 Etherchannels between 4500X and 2960X as soon as output QoS policies re applied. I think there are some LACP related bugs in that code.  These switches also ran IOS XE 3.4.2SG.

Actions

This Discussion

Related Content