Re: 3750 cross stack etherchannels

cwilson · ‎12-02-2006

I have a 3750 stack connected to another 3750 stack using 2 - 100mb links. I wanted to use an etherchannel to combine the bandwidth to be 200mb/s as well as for redundancy in case one link went down or one of the switches went down. I have a port channel defined trunking the vlans I want and have the 2 ports on each switch in the port channel. Because they are cross stack etherchannels, I have to use channel-group # mode ON so I can't use LACP or PAGP. When I remove one link, it doesn't automatically failover to the remaining link. I have read a ton of info on the website, but don't know what I am missing. Do I have to go back and have both ports on one switch instead of having it cross stack?

wgoulart · ‎12-02-2006

Hello,

Can you please post the configuration of the switch i.e., sh run so we can better diagnose your issue to help you.

Thanks

glen.grant · ‎12-02-2006

It should failover , an etherchannel is considered one link and traffic just gets shared across both channels , if one goes away then it should just flow across the remaining one . Did you check that both ports were channeling before pulling the one link with the show etherchannel summary and show etherchannel detail commands ? . You do have the stack cables in the back connected correct ? Just looking for the easy stuff first . As the other poster said , the config would help for the etherchannel and the connecting ports themselves on both boxes and both sides . Also maybe post a show switch detail command . Also verify the trunk is working correctly with the show int trunk command and verify what is being allowed across the trunk is the same on both sides of the link .

cwilson · ‎12-03-2006

My config is as TAC and the website examples show:

Switch 1 running 3750-IPBASEK9-M, Version 12.2(25)SEE2

hostname CMT-ASE

!

switch 1 provision ws-c3750g-24ps

switch 2 provision ws-c3750g-24ps

switch 3 provision ws-c3750g-24ps

switch 4 provision ws-c3750g-24ps

spanning-tree mode rapid-pvst

no spanning-tree optimize bpdu transmission

spanning-tree extend system-id

!

vlan internal allocation policy ascending

!

interface Port-channel1

description ASE trunk

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

!

interface GigabitEthernet1/0/1

description 2230-1

switchport access vlan 22

switchport mode access

switchport voice vlan 701

mls qos trust cos

spanning-tree portfast

spanning-tree bpduguard enable

!

interface GigabitEthernet1/0/24

description ASE-22 floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

speed 100

duplex full

no mdix auto

channel-group 1 mode on

!

interface GigabitEthernet2/0/24

description ASE-22 floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

speed 100

duplex full

no mdix auto

channel-group 1 mode on

!

end

and the switch on the other end of the 2-100mb links

C3750-IPBASEK9-M, Version 12.2(25)SEB

version 12.2

hostname 34-35

switch 1 provision ws-c3750-24p

switch 2 provision ws-c3750-24p

switch 3 provision ws-c3750-24p

switch 4 provision ws-c3750g-24ps

switch 5 provision ws-c3750-24p

!

spanning-tree mode rapid-pvst

no spanning-tree optimize bpdu transmission

spanning-tree extend system-id

!

vlan internal allocation policy ascending

!

interface Port-channel6

description ASE - 22floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

duplex full

speed 100

!

interface FastEthernet3/0/23

description ASE - 22floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

duplex full

speed 100

no mdix auto

channel-group 6 mode on

!

interface GigabitEthernet4/0/23

description ASE - 22floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

duplex full

speed 100

no mdix auto

channel-group 6 mode on

!

interface Vlan1

ip address 172.16.255.6 255.255.255.0

!

ip classless

no ip http server

ip http secure-server

!

logging 10.0.1.101

access-list 102 remark SSH Access

34-35#

politiezuidhollandzuid · ‎12-04-2006

You should take the settings "channel-group 6 mode active".

From Cisco.com "PAgP cannot be enabled on cross-stack EtherChannels while LACP is supported on cross-stack EtherChannels from Cisco IOS Software Release 12.2(25)SEC and later. "

LACP = mode active

PAGP = mode on

rvv · ‎12-04-2006

Don't thnik so.

mode on - not mean that pagp will take place.

as you say: "PAgP cannot be enabled on cross-stack EtherChannels"

In on mode, a usable EtherChannel exists only when both connected port

groups are in the on mode.

I have similar configuration (mode on used) and all work fine.

But i'am use ISL encapsulation on trunk. Maybe there is problem.

rvv · ‎12-04-2006

Try to not create PortChannel first.

When you issue "channel-group" command portchannel will be created automaticaly. Try to make ALL configuaration on GigabitEthernet and with last command isuue "channel-group".

And you will be shure that all config from GigabitEthernet interface transfered to PortChannel interface.

cwilson · ‎12-04-2006

I upgraded the switch on one side so they were both at 12.2.25 SEE2, so now the etherchannel comes up with LACP (channel-group mode active). Still, when I unplug one connection on the etherchannel, I lose connection and the traffic isn't immediately forwarded to the member that is still up.

glen.grant · ‎12-04-2006

How are you testing to see if the connection is working or not when you pull one side ???

cwilson · ‎12-04-2006

I have a laptop over at the remote location pinging a server in the datacenter across the link. With both interfaces in the etherchannel connected, pings go through. I then pull the cable or shut one connection and my pings keep going so that means that through the source mac load balancing, I am using the other interface that is up. Then I put the connection back, wait a couple minutes and then pull the other connection. My pings to the other side then stop, after a minute or so I put the connection back and my pings resume. From other etherchannels I have done, it should failover with maybe 1 or 2 pings dropped.

yizhou_david · ‎12-10-2006

I have exactly the same issue.

My IOS version is c3750-ipservices-mz.122-25.SED1.bin.

At the beginning I set both sides as 'mode on', when one link fails traffic fails to switch to the other link.

Then I configured both sides to use LACP 'mode active', it did work. But the failover takes around one minute.(customer is not happy about this one min outage)

In my setup, my both 100M links are metro ethernet links instead of direct inhouse cables. When one side fails, Ethernet

provider doesn't pass the signal over to the other side. So the other side remains up without knowing far end is down. Hence

Etherchannel still keep sending packets to both interfaces according to its hashing algorithm. After I configured LACP, LACP detects

the failure in about one minute and brings down the interface on the other end.

Just curious to know if you also get Metro Ethernet connections in between.

I can't see any option available for me to tune down the timer in LACP neigotiation altough some command output does show fast/slow LACP. Would be happy to know if anyone from Cisco is able to answer this question.

alanwright1 · ‎12-11-2006

Hi David,

Try using UDLD. http://www.cisco.com/en/US/products/hw/switches/ps5023/products_configuration_guide_chapter09186a00805a6510.html

Its not really meant to detect this, but it may well do as it is based on a Mcast frame on the wire to the other end.

politiezuidhollandzuid · ‎12-14-2006

It confused me.. So I tried it myself in a lab. I use mode active in a cross stack environment. It works perfectly. When I take a look at your config I see a speed setting at the portchannel interface. I tried it and now I have the same issue. Remove it to see if that is your solution. And LACP is the way to go. The other way is manual and that one is prone to errors. (forgive my english)

cwilson · ‎12-14-2006

Thanks for trying it in a lab, I don't have extra switches. I found one problem was with one of my 100mb links. Anyways, the last time I tried it, this was my config for the portchannel:

interface Port-channel6

description ASE-22 floor

switchport trunk encapsulation dot1q

switchport trunk allowed vlan 1,10,22,701,997,1002-1005

switchport mode trunk

switchport nonegotiate

end

It didn't seem to make much difference for me. I still have faster failover if I just use 2 trunks with STP blocking one and don't use the portchannel-whether it is mode active or mode on.