MST and Etherchannel Question

criitdept · ‎11-22-2010

Hi

I hope someone can help me with this.

I have 3 Cisco Catalyst 3560 Switches connected as per the diagram.

I have just configured these with MST in Region 1 and Instance 1

When I have the etherchannel unplugged between A + B and I check spanning-tree summary, it shows the connection between A + C as Root FWD. this is the same on B + C.

When i plug the etherchannel connection between A + B, the Span Summary shows A as Desg FWD and B as Altn BLK.

Is this correct?

I get the feeling that both sides should be either BLK or FWD.

When this etherchannel is plugged in, i get some connection issues when connecting to servers that are plugged into these switches. When the etherchannel is unplugged, the issues go away.

What do i need to check or where do i start to troubleshoot this?

Thanks for your help

James

Peter Paluch · ‎11-22-2010

Hello,

When i plug the etherchannel connection between A + B, the Span Summary shows A as Desg FWD and B as Altn BLK.

Is this correct?

I get the feeling that both sides should be either BLK or FWD.

The behavior you have observed is correct. If a link is redundant, only one of its ends is blocked, not both ends. This is consistent with all versions of STP. The rationale behind this is that the link may provide a redundant connection between switches but it may also host some end stations. For example, a link between your switches A and B can be in fact created by another unmanaged switch or a hub plugged between A and B, and this switch or hub could provide connectivity to a number of end stations. Because this unmanaged switch/hub does not speak STP, switches A and B do not know of its presence. If both A and B declared this link as redundant and blocked it from both sides then the clients on this link would be completely cut off the network. Of course, today, such a design would be a gross violation of best practices but there are still networks in which this generalized approach is necessary.

That being said, I believe you should tweak your STP settings to actually make use of the EtherChannel you have configured. According to your description, the EtherChannel link bundle is considered as redundant link via MSTP and it is being blocked, i.e. the EtherChannel is unused - instead of providing your network with additional inter-switch data-carrying capacity.

When this etherchannel is plugged in, i get some connection issues when 
connecting to servers that are plugged into these switches. When the 
etherchannel is unplugged, the issues go away.

Please try to be more descriptive of the "issues", and try to be as precise as possible.

Best regards,

Peter

tverhell · ‎11-22-2010

Hello James,

Yes, this is correct STP(MST)behaviour:

As C is here you root bridge (both port from A and B towards C, showing as root port),

to avoid a loop we need to block the link between A and B (here the etherchannel) either on the A-side or the B-side.

Doing this on 1 side is enough to guarantuee a loop-free environment.

Maybe this page on troubleshooting STP will be helpfull for you:

http://www.cisco.com/en/US/tech/tk389/tk621/technologies_tech_note09186a0080136673.shtml

However regarding your problem,

the exercise will be to get a more detailed problem description and find some pattern:

- what sort of connectivity issue are you seeing? (when, how frequent, hong long problem last, resolving itself?, ...)

I guess both end-users and servers are located in that vlan 30?

- On which ports are those servers/end-users connected that are facing these issues?

- Do certain users see the problem, and others not,...?

- doing 'sh mac address-table' to look for those end-users/servers on each switch would get you also more insight,

how packets would flow (basicly we should not see them on the etherchannel, as that is a blocked link).

- 'sh spanning vlan 30 detail' from each switch will also learn if BPDUs are sent correctly, and if STP is stable (no frequent TCNs)

Cheers,

Tom Verhellen