need clarification on trunking and setting up port-channel

config: interface GigabitEthernet0/1 switchport trunk allowed vlan 2-4094 switchport mode trunk channel-group 1 mode desirable

interface GigabitEthernet0/2 switchport trunk allowed vlan 2-4094 switchport mode trunk channel-group 1 mode desirable

a few points:

• config must match exactly on the two ports

• allowed vlans at either end must have the exact same range

• channel-group number (port channel number) must match all on interfaces and be unique

you should either configure port channel or spanning tree, you might as well double your bandwidth and use a port channel.

traffic will still get across one link if the other goes down so you aren't losing out on diversity.

Flamer.

Flamer,

Thank you for the provided information. Please allow me to ask a few more follow up questions.

1. I understand that the configuration on each port (within a switch) must match, but what about the other switch? Do I need to configure port channel on the other switch as well? If so, what are the requirements for it then?
2. Is it required to use the "switchport trunk allowed vlan 2-4094" statement? I think, that skipping it will allow vlans.
3. What does the "mode desirable" stand for? I googled it, but could not find anything that I could easily comprehend.
4. If I understood it correctly, the "Port-channel1" interface will be created automatically when I add the first physical interface to the "channel-group 1"? Do I need to tweak the settings of the "Port-channel1" interface once it is created?
5. Without creating the "channel-group" (etherchannel), and having 2 wires connecting the two switches the throughput between the two switches will be equal to the speed of a single interface and only one of them will be active at a time? What happens when this link fails? Is there a downtime till the link is established over the other wire?

Thank you in advance for additional clarification.

Regards, AP

Thanks, AP

And yes, you need to configure the channel on both sides/switches. If you don't, it works as you say, and yes there will be downtime as spanning-tree runs since one of those links should be blocking. With the channel, it's only one logical link, so there really isn't any disadvantages to the channel in a small network (other than your switches may not be able to push the throughput).

And yes, you need to configure the channel on both sides/switches. If you don't, it works as you say, and yes there will be downtime as spanning-tree runs since one of those links should be blocking. With the channel, it's only one logical link, so there really isn't any disadvantages to the channel in a small network (other than your switches may not be able to push the throughput).

Thank you for taking time and explaining this. Regards, AP

Hi Folks,

While all this is true, I did notice on a failed 2948G-L3 that the physical ports were grouped in units of 4 per chipset (when fa0/1failed, then fa0/2 - fa0/4 also failed), and while you lost 4 ports in an row, the other ports on the switch kept going. So there may be some way to retain Port-Channel operation by spreading the port aggregation out a bit in such circumstances, but overal it does mean a dead (or dying) switch.

This was also apparent when using both ISL or Dot1q trunking on the

2948G-L3, you had to set a comon form of trunking method for 4 adjacent ports.

Just my 2c worth..

Cheers.............pk.

Hmm. I don't think that is the case. I can't test for the present but I have done a fair bit of work with EtherChannels in the past I am sure I would have noticed this. Maybe some platforms do this, but I would call it a bug.

This is always a good plan.

You can use RSTP to get fast convergence on reasonably recent kit. The one real disadvantage of using a channel for resilience is that it is more difficult to make changes to the port confiuration without risking an outage. This is due to the strict requirements regarding similarity of configuration of ports participating in a channel.

Of course if you are well organised and practise change control that will never be necessary;-)