Setting STP properly or disabling at all in MLAG setup

[Robert_Zdzieblo]

Hello Guys,

I got a customer with MLAG’ed stacks as peers (X670-G2 and X460-G2 in each stack), where the edge switches are X440-G2 stacks, LAGged to both peers. So the whole environment is EXOS.

When I upgraded from 22.3 to 22.5 I got LAG uplink ports (master) on edge stacks set to blocked.

Command “debug vlan show vpif default 1:52” run on edge stack showed that STP was registered on the port.

Before the upgrade to 22.5 I didn’t setup STP in any way, although it was enabled by defalut AFAIK.

After the upgrade I tried to disable stpd on edge stacks, but it didn’t help.

As we had a downtime in production network,  there was no time for further troubleshooting and we have to revert to 22.3, which resulted in getting network up and working again.

I haven’t disabled STP on MLAG ports at MLAG peers when running 22.5, as I have no time.

I know that RSTP support for MLAG was introduced in ExtremeXOS 22.5 and that changed the way STP works in my setup.

Do you think disabling STP on MLAG ports at peers would help in this case ?

BEST REGARDS

Robert

[Robert_Zdzieblo]

OK, guys - I disabled STP completely on both MLAG peers.

Now network is stable on EXOS 22.5.

Thanks,

Robert

[Robert_Zdzieblo]

Fredrik,

To be honest I don’t need STP at all. 

I had never took it into consideration at network design stage, although not all network elements are under my control (like routing, for example).

It’s probably just the way EXOS sets STP up, when it’s enabled by default.

I don’t either think that this STP port blocking could be caused by some network loop, as we’ll have it regardless the EXOS is 22.3 or 22.5.

Thanks for all your thoughs, guys !

Robert

[FredrikB]

That’s about the only proper way to configure spanning tree these days. What do you actually need STP for, really? I only ever recommend using it for access port loop detection and blocking but never on network links. There are so much better ways to build networks these days than to use STP. STP was invented 35 years ago!!!

/Fredrik

[Gabriel_G]

Robert,

That is correct. STP should be automatically disabled on the MLAG ports on the 2 peers when MLAG is enabled but double check and be sure to disable STP on the ISC link between the peers as well. I would advise against disabling STP entirely as an unknown loop in your network can cause a lot of issues so let’s isolate everything to this MLAG partition.

The ISC blocking filter will prevent a loop on the 3 node MLAG topology but if you need additional loop protection for the rest of you network you can use ELRP, RSTP for MLAG, or simply keep STP enabled on the rest of the network. Without knowing about the rest of your topology, I cannot give a more detailed suggestion.

Once STP is disabled on the interfaces involved in MLAG, check to see that the peer relation is up and that the LAG ports on the access switch up and not blocking. A problem at this point would likely be due to a bad MLAG config. As a quick reference, you can check this article for troubleshooting steps, or this article for configuration steps.

Also, here is an article for configuring RSTP for MLAG in case that fits your situation.

Thanks,
Gabriel