Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations biv343 on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

EtherChannel Configuration 1

Status
Not open for further replies.

Dektaki

IS-IT--Management
Jan 24, 2007
9
US
I have 3 Cisco 2960's all 10/100/1000. I use Ports 45 - 48 in a Trunk mode and have created EtherChannels to attempt to double my link speed between the switches. I'm not sure that I'm getting that at this point because when I use the "sho int etherchannel" command it shows the secondary port in each Channel-Group as being "Down Not-in-Bndl" instead of Up. Should I be specifying the etherchannel type rather than on?

2960-1 - Ports 46,47 in Channel-Group 1 Mode On (Uplink to Switch 2)Trunk Ports for VTP purposes

2960-2 - Ports 45,46 in Channel-Group 1 Mode On (Uplink to Switch 1)Trunk Ports for VTP purposes
2960-2 - Ports 47,48 in Channel-Group 2 Mode On (Uplink to Switch 3)Trunk Ports for VTP Purposes

2960-3 - Ports 45,46 in Channel-Group 2 Mode On (Uplink to Switch 2)Trunk Ports for VTP purposes.

Above is a very basic idea of the switched environment. My ultimate goal is to have each of the channel groups aggregate so as to achieve a higher switching fabric between the individual uplinks. My understanding is that with 2 Gigabite Ethernet ports in a single Channel-Group I should be able to see a total of 4GB. 2GB Up and 2GB Down.

Any thoughts?
 
Let the switches negotiate the etherchannel , alot of people like to try and force the channels on and when they don't work they wonder why . On each of your interfaces use
"channel-group # mode desirable non-silent and this will bring your channels up . I can almost guarantee this will work so try it .
 
I am not sure how you have those tied together . where do the etherchannels go to ?
 
The ehterchannels go from switch to switch. My hope was to bond two Gig interfaces to make for a higher uplink between all 3 switches.
 
Viper,

It was in fact exactly as you stated in your first response to me. I now do the "show int etherchannel" command and both ports come up as "UP Mstr In-Bndl" now!

So let me throw in a monkey wrench into the equation. Since PAgP is Cisco proprietary, what will the outcome of me attaching an HP Procurve 2848 to the end of this stack and trunking it off of the last switch? I'm assuming the other two Channel-Groups will not be affected since they are off of Cisco switches. Am I correct to assume that when I create the trunk between the last Cisco switch in the stack and the HP 2848 Managed switch that I will have to use LAcP instead?

Thanks again for your help!
 
If you have connections in like a triangle configuration then spanning tree is going to block the path somewhere on those connections as you have a built in loop so traffic will flow all one way .
 
I didn't setup the switches in a full mesh as I didn't want to tie up the other ports to do so. So really it's just:

Switch 1 --> Switch 2 --> Switch 3

I had counted on STP stopping the secondary links in the event that the etherchannel was not setup properly. And in fact, that's exactly what was happening. PVST was stopping the second link in the bundle causing it to come down. I understand all of that, but what I don't get is why the command "channel-group # mode on" didn't work...Is it because I'm trying to force PGaP?
 
Not sure would have to see the old configs that didn't work . PAGP is not even used if you force it on . If you need to have a channel to the HP then you would indeed have to use LACP for that . Spanning tree should not block anything over a single etherchannel as the link as far as spanning tree is concerned is a single logical link . If you have 2 etherchannels running the same vlans across it then you would get some ports blocking because it would have a built in loop . It is not a problem mixing the 2 etherchannel protocols or change it to al LACP if you want to keep it consistent but I always let the switches negotiate the link at least between cisco devices , I haven't had the pleasure of trying to link between manufacturers . A little info on lacp . Also a link

CAT2955# configure terminal
Enter configuration commands, one per line. End with CNTL/Z.
CAT2955(config)# interface fastethernet 0/9
CAT2955(config-if)# channel-group 1 mode active
Assign the interface to a channel group, and specify the LACP mode.

CAT2955(config-if)#
6d08h: %LINEPROTO-5-UPDOWN: Line protocol on Interface FastEthernet0/9,changed state to down
6d08h: %LINEPROTO-5-UPDOWN: Line protocol on Interface FastEthernet0/9,changed state to up
6d08h: %LINK-3-UPDOWN: Interface Port-channel1, changed state to up
6d08h: %LINEPROTO-5-UPDOWN: Line protocol on Interface Port-channel1,changed state to up
CAT2955(config-if)# interface fastethernet 0/10
CAT2955(config-if)# channel-group 1 mode active
CAT2955(config-if)#
6d08h: %LINEPROTO-5-UPDOWN: Line protocol on Interface FastEthernet0/10,changed state to down
6d08h: %LINEPROTO-5-UPDOWN: Line protocol on Interface FastEthernet0/10,changed state to up
CAT2955(config-if)# interface port-channel 1
 
Why even use LACP or PAGP?

What are the advantages to using the protocols?? Why not just manually set the ports to on and not use the protocols.

-E
 
Why even use LACP or PAGP?

You have to use one protocol, there isn't an option for just 'on', the default is PaGP.

With regards to configuring EtherChannels as Automatic or Desirable instead of nailed-up, Cisco's (current?) best practise is to hard-code them as it decreases the convergence time slightly (milliseconds):

interface FastEthernet0/1
channel-group 1 mode on
!
interface FastEthernet0/2
channel-group 1 mode on

The same goes for VLAN Trunk interfaces - hard-code them and disable negotiation, as well as limiting the VLANs:

interface FastEthernet0/1
switchport
switchport access vlan 4000
switchport trunk native vlan 4000
switchport trunk allowed vlan 10,20
switchport trunk encapsulation dot1q
switchport mode trunk
switchport trunk encapsulation dot1q
switchport nonegotiate
!


HTH

Andy
 
I thought so too till I reviewed the material! As usual, vipergg is right. Ports in the on or off mode do not exchange PaGP packets. Because it is disabled, it does not negotiate. That's why it is slightly quicker.


Why do we bother to use it? Because it's neat. If the new guy comes along and plugs his laptop into the etherchannel port and it is set to desirable, his laptop will still work because it won't receive PaGP or LACP packets.
 
Really depends on what you want to use , this is one cisco doc that says use pagp or lacp.

Recommendation

Cisco recommends that PAgP be enabled on all switch-to-switch channel connections, avoiding on mode. The preferred method is to set desirable mode at both ends of a link. The additional recommendation is to leave the silent/non-silent keyword at default - silent on Catalyst 6500/6000 and 4500/4000 switches, non-silent on Catalyst 5500/5000 fiber ports.

 
Well, my implementation went well. I ended up ultimately as my original thread stated going with "channel-group # mode desirable non-silent" on all 4 of my 2960's, and everything is working great. I ended up ditching that HP gig switch because I didn't want to deal with mixing equipment in our server room...So now I'm 2GB uplinked between all 4 switches and using PaGP. I see a little bit of latency, but nothing huge as compared to when I had hard coded the mode to "on" instead of desirable. Thanks again Viper!
 
If you are seeing certain links busier than others you can play with the etherchannel load balancing commands to see if you can even it out a little .
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top