Another Etherchannel causing STP loops problem

[axa-wongjeff]

Just encountered this last night.

New configuration - Etherchannel 2 ports from Cisco switch modules installed in IBM Bladecenter to Catalyst6509 switch.

Received the following message sequence on the Bladecenter switch modules:

%PM-4-ERR_DISABLE: channel-misconfig error detected on Gi0/17, putting Gi0/17 in err-disable state

%PM-4-ERR_DISABLE: channel-misconfig error detected on Gi0/18, putting Gi0/18 in err-disable state

%LINK-3-UPDOWN: Interface GigabitEthernet0/17, changed state to down

%LINK-3-UPDOWN: Interface GigabitEthernet0/18, changed state to down

%LINK-5-CHANGED: Interface GigabitEthernet0/17, changed state to administratively down

%LINK-5-CHANGED: Interface GigabitEthernet0/18, changed state to administratively down

========================================

My etherchannel configs:

Bladecenter

int range gig0/17 -18

switchport mode access

switchport access vlan 102

duplex full

speed 100

channel-group 1 mode on

interface port-channel 1

switchport access vlan 102

switchport mode access

Cat6509

set vlan 102 6/8-9

set port speed 6/8-9 100

set port duplex 6/8-9 full

set trunk 6/8-9 off

set port channel 6/8-9 mode on

========================================

I checked my Etherchannel configurations on both switches and I should've seen a channel established. Then all of a sudden, Bladecenters on the same VLAN began to err-disable their switch ports. I started seeing 'Duplicate Address' messages and EIGRP neighbor change messages in syslog. I knew I had some sort of STP loop occurring, if that was what was happening at that time.

My questions:

1. What could the switches have seen to say that the channel was misconfigured? Where would I find this information?

2. If the Bladecenter switch ports were admin down via software, then how come I still saw STP loop symptoms? I figure the shutdown state would have stopped or halted any STP calculation issues.

[glen.grant]

I would check the links for any errors , also check to see where the spanning tree root is for the vlans on that channel . One of the dangers of forcing on the channels is that you can get stuff like this , I always set the channels up dynamic non-silent and let pagp do its thing .Have never had a etherchannel problem and we have the same setup with a 6500 going to many bladecenters with cisco igesm switch cards. There is a command to see why it err disabled , show interfaces status err-disabled" .

[Edison Ortiz]

Jeffrey,

When configuring etherchanneling in CATOS, you need to bundle the ports and turn the etherchannel mode on different commands.

You have one command there with set port channel 6/8-9 mode on but you also need this command 'set port channel 6/8-9'

Try that and let us know. To verify etherchanneling in CATOS type the command 'show port channel'.

___

Please rate helpful posts.

Thanks

[axa-wongjeff]

I entered just 'set port channel 6/8-9'. The only change was the AdminGroup number. I looked over the configuration and saw no change in the original 'set port channel' statement.

The 'show port channel' output:

----- ---------- -------------------- ----- -----

6/3 notconnect on 148 802

6/4 notconnect on 148 802

----- ---------- -------------------- ----- -----

6/9 notconnect on 151 805

6/10 notconnect on 151 805

----- ---------- -------------------- ----- -----

6/15 notconnect on 149 808

6/16 notconnect on 149 808

----- ---------- -------------------- ----- -----

[bella]

Remove VLAN membership 102 on either of both sides.

[axa-wongjeff]

> Remove VLAN membership 102 on either of both sides.

What would removing it from the VLAN prove? I need to assign it to a VLAN. I'm not trunking.

[Edison Ortiz]

Correct,

The Admin group number bundled the ports together. Now go to the IBM switch and issue a shut then 'no shut' command to remove the err-disabled state.

[Francois Tallet]

As far as STP is concerned, the message means that the channel keeps receiving BPDUs from two different source mac addresses. I'm not a big fan of this feature, but here, it means that the remote port is probably not bundled correctly: the channel is broken up in at least two different interfaces, each sending its own BPDU.

Regards,

Francois

PS: STP is not the problem here, it is just detecting an error.

[kirkster]

I have just had the same problem myself.

I seem to recall there was a spanning tree feature that helped prevent this exact problem. Can't remember it off the top of my head. Will look into it

[ge4sat]

I seen this exact issue and error on 2 of my 6500's runnning IOS. I was missing the "Interface port-channel 1 on one side. What is its equivelint on CATOS? Maybe that answer will reolve the issue.

[axa-wongjeff]

To configure etherchannel on CatOS, the command would be 'set port channel '. That's it.