Etherchannel/PagP & same/different VLANs on trunk

Unanswered Question
Jul 9th, 2009
User Badges:

Hi All,

1) If I have a working etherchannel trunked then use "switchport trunk allow VLAN n remove" to remove some of the VLANs from port-channel interface, will this

1) break the etherchannel

b) break connectivity for the remaining VLANs?

(I know you'e supposed to have same VLANs both ends of etherchannel)

2) If I then do the same command at the other end of the etherchannel will the etherchannel reform & connectivity for remaining VLANs be restored?

If I do this 'remove' at both ends of the etherchannel within say 5 seconds is this fast enough to not affect the etherchannel bundle?

Thanks for your help.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 2 (1 ratings)
Giuseppe Larosa Fri, 07/10/2009 - 03:42
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Mark,

once a Layer2 trunk etherchannel is up and established all changes have to be done on the logical interface port-channel:

doing it directly on physical members caused a bridging loop in our customer networks (twice it was actually an additoon of a new vlan).

Working on both ends of the channel on logical interfaces should be safe.

Hope to help


gnijs Fri, 07/10/2009 - 04:07
User Badges:
  • Bronze, 100 points or more

Adding or removing a VLAN shouldn't bring down or re-initialise an Etherchannel. I have tested this with LACP, however not with Pagp. Also, LACP won't prevent you from running with different allowed VLANs on both ends, so i assume there is no "verification" method inside LACP to check if the same vlans are configured on both sides (not that i know of anyway)

mheuzenroeder Fri, 07/10/2009 - 20:50
User Badges:

Thanks gnis,

I should add I'm using PagP (the default) and mode 'desirable' at both ends.

Cisco doco don't seem to differentiate between "having same VLANs allowed at both ends is best practice" and "having different VLANs allows at both ends will make PagP tear the bundle apart.

Is it correct that worst-case, if the bundle is torn apart by different allowed VLANs each end STP will block some physical ports so that others are still forwarding and connectivity is not completely lost?

Thanks for your help, MH


This Discussion