Spaning tree vs LACP vs multi-chassis LAG

[azi123]

Hi, everyone!

I have a simple ring topology with 10 nodes ( switches). I have two questions:

1-I think I can use multi-chassis LAG( link aggregation)  instead of any spanning tree protocol, but I'm confused whether I'm able to use LACP instead of spanning tree or not?

( To make things clear about LACP case: Well, each switch has two ports connecting to other switches and then I configure both ports in one channel group and disable spanning tree protocol, So at most one of these two ports forwards a specific packet and by this, we could prevent loop in layer 2. Obviously, this configuration will be applied to all switches.) 

2- Am I true about the previous question's assumption? Can multi-chassis LAG  be a substitute for spanning tree protocol?  

Best, Amir

[Georg Pauwen]

Hello,

it is not recommended to turn off STP in anything layer 2 (including LAGs). However, if you are absolutely certain that there are no loops, and there is only one LAG, you could turn it off. 

LACP is just the protocol LAGs use, so it is not a substitute for STP.

[Francesco Molino]

Hi

When you talk about multi chassis lacp, do you mean all your switches are attached to a distribution/core switch connected using a protocol to combine 2 physical switches in 1 logical switch like (stack, vss, vpc, virtual stack)?
If not, you can't do multi chassis lacp just by connecting 2 uplinks to 2 different switches and disabled stp.

Stp is very important in such topology and disabling it could cause a lot more issues.

[azi123]

All I have is simple ring topology and there is no distribution/core switch. ( see attachment #1 please). And all the switches support multi-chassis LAG feature. Then I configure switch number 1, in a way that both switches number 0 and 2  become one chassis from switch number one's point of view. It means that switch number one sees both switch number 0 and 2 as one switch( attachment #2).

and I repeat this configuration for all five switches( attachment #3)

I'm not sure this way of using MC-LAG could be an alternative for STP, but I hope I could make my idea more clear.

Thank you all for your responses.

[Francesco Molino]

What type of switches are you using?
Anyway, if you're connecting every single switches as mlag in front of another mlag, this should be spanning-tree loop free design. But what happens if someone creates a loop in your network? That's why you shouldn't disable spanning-tree.

I'm kind of skeptical with your attachment #3. If i understand correctly, switches 1 and 3 form a logical switch facing 0 and 4 which are also another logical switch?
Maybe with the model of switches it will clarify a bit. Can you share some part of your config?

[azi123]

Maybe I didn't understand MC-LAG well, but let answer your question: switches 1 and 3 make a logical switch, but not facing switches 0 and 4, they face switch 2!

Suppose your configuring switch 2. You configure it in a way that both switch 1 and 3 become one chassis. A simple set up for MC-LAG configuration (the figure 1 in the link below). the difference between this configuration and figure 1 is that there is no intra-chassis link between switch 1 and 3.

Like the previous step, I repeat for other switches: I configure switch 5 in a way so both switch 0 and 4 become on chassis. I configure switch 4 in a way that both switch 3 and 5 become one chassis.  I configure switch 3 in a way that both switch 2 and 4 become one chassis and so on.

and I don't have any switches that support MC-LAG right now, but I'm considering buying one of 3750 series.

the link:

https://www.cisco.com/c/en/us/td/docs/ios/cether/configuration/guide/ce_mlacp.html

[Francesco Molino · Accepted solution]

To do MC-LAG, on 3750, you need to build a stack using the stack ports on the rear of the switch.

If you associate switch 1 and 3, them become 1 logical switch. Now if you associate switch 3 and 5 (as you said), this means you'll add switch 5 to the stack already existing between switch 1 and 3. At the end, you'll have a stack of 3 switches (switch 1, 3 and 5).

Now let's assume your stack has 3 switches and faces 1 switch. The LACP has been built using switch 1 and 3 from the stack to the peer which is in standalone. If the stack fails, switch 1 and 3 will act as 2 different switches and here you'll have a STP loop. If you deactivate STP, you will have issue.