[External] [c-nsp] N9K traffic lost when redundant link comes up

[Hunter Fuller]

If you are just adding a port to a Port-channel, the box should not see any
MAC "moves" as such. The MAC is learned on the Port-channel interface no
matter which bundle member it uses.

This is going to be something more esoteric, or a bug.

--
Hunter Fuller (they)
Router Jockey
VBH Annex B-5
+1 256 824 5331

Office of Information Technology
The University of Alabama in Huntsville
Network Engineering


On Tue, Mar 9, 2021 at 11:10 AM <cnsp at marenda.net> wrote:

>
> Hi,
>
> i have a pair of N9K-C93180YC-EX running nxos.9.3.1.bin connected with a
> LACP port-channel (pair of 100G Links).
> I got a pair of N9K-C9348GC-FXP running nxos.9.3.5.bin connect with a
> (single-100G Link) LACP post-channel to only one of the above switches.
>
> I finally got more transceivers to create the missing redundant link(s) to
> the other one of the first switches ,
> In a second LACP port-channel with just one single-100G Link.
>
> No Multi-Chassis LACP here, each device works stand-alone,
> spanning tree mode is MST, everywhere identically configured.
>
> Expected behaviour is:
> New link gets active, and
> if spanning tree finds this new link as "lower" it would block it.
> if spanning tree finds it "better" it should start to use it and block
> somewhere else.
>
> But monitoring was crying, and I found in the loggin:
> 16:30:19 dsw2 %L2FM-2-L2FM_MAC_FLAP_DISABLE_LEARN: Disabling learning in
> vlan XXX for 120s due to too many mac moves
> 16:32:19 dsw2 %L2FM-2-L2FM_MAC_FLAP_RE_ENABLE_LEARN: Re-enabling learning
> in vlan XXX
>
> Yes, that was also the duration of the "outage", adding a redundant link
> leads two two minutes outage ☹
>
> Cisco's error-messages finder tells me that there is nothing to do  ?!?
>
> Case opened, infos submitted, but two days (plus weekend) silence.
>
> Any idea what is happening and how I can avoid that (the fourth link wants
> to be plugged in).
>
> Will that happen when a link fails, STP unblocks an other link and
> therefor the switch relearns too much mac-addresses too fast
>
> so I get again 2 minutes "down" instead of just 2..3 seconds ?
>
> The ancient C4900M did not show that behaviour...
>
> Any suggestions?
>
> Thank you for your patience,
>
> Jürgen.
>
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>