Re: Nexus 5k / 7k Loop

In case anyone else has this issue - TAC has identified that this can actually
happen within the 77xx platforms and is a rare but known bug. B They have seen
this before, but are unsure of a root cause. B Either the L2FM or the port
index manager was not properly adding links to the vPC port-channel, and a L2
loop formed. B The fix can be as simple as a system reload, or in some cases
just flapping the links.

________________________________
 From: david bloom
<dbsg13_at_yahoo.com>
To: Cisco Certification <ccielab_at_groupstudy.com>
Sent:
Tuesday, August 19, 2014 10:58 AM
Subject: Re: Nexus 5k / 7k Loop
 

As an
update - we just completed the 2nd uplink from the 5k to the 7k, created
port-channels and back to back vPC, however the loop still exists...
________________________________
 From: david bloom <dbsg13_at_yahoo.com>
To:
Cisco Certification <ccielab_at_groupstudy.com>
Sent: Tuesday, August 19, 2014
9:38 AM
Subject: Nexus 5k / 7k Loop
 

Hello Everyone,

We have an odd
situation that I canbt quite understand - a L2 loop is happening.

We have a
production 5k environment, and a new 7k lab
environment.B When I connect the
two
together, the 5kbs immediately report a loop and disables learning.
B While this happens, traffic in the production vlan is intermittent, i.e.
some pings go though, others dropped.

Both the 5kbs and the 7kbs are in a
back to back vPC
configuration.B There is a single link
between 5k04 and
7k02.B At first, we
set up that single link in a port-channel and vPC on both
the 7k and 5k.B We then removed the port-channel and vPC
config for the link
on both sides, and set the link as a standard trunk.B B Both scenarios
produced the loop b with the
detection showing on the 5k that the same MAC
was learned coming from the 7k,
and on the port-channel going to the upstream
5k.

My guess is that this is an unsupported config.B We do plan do add a
second link from the 7k
ot the 5k, this way we can run back to back vPC
between the 7k and 5k.B Ibm thinking this will solve the issue.

Question
is why would the single link cause a loop?

Mac address flooding coming in
from the upstream 5k goes through the link to the single 7k02. B The flood
goes out the port-channel (to both downstream 7k03 AND 7k04) and across the
vPC peer link, correct? B From there, each switch (other 7k01, and both
downstream 7k03, 7k04) receive the flood on a vPC port-channel, so it should
not flood back, either over the same port-channel, or through the downstream
vPC peer link?

5k01 B ---- B 5k02
B B | B B B B \ / B B B |
5k03
B ---- B 5k04
B B B B B B B B B B B B |
B B B B B B B B B
B B B |
7k01 B ---- 7k02
B B | B B B \ B / B B B |
7k03 B --- 7k04
Each horizontal set is a vPC peer, and all cross links are
port-channels in a
vPC.B The single
connection is a standard trunk.B The loop
errors occur on
the 5k04 that is connected to the 7k02. B 5k03 also reports the errors - the
duplicate mac is seen on the vPC peer-link from 5k04, and the port-channel to
the upstream 5ks.

David

Blogs and organic groups at http://www.ccie.net
Received on Thu Aug 21 2014 - 18:18:19 ART