[c-nsp] Cat 3750E w/ MAC learning disabled shutting port?

Sascha E. Pollok sp at iphh.net
Sat Apr 25 02:52:52 EDT 2015


Hi Lukas, hi Blake, et al,

me again. I noticed I have not been completely honest with you guys :)

>>>> I am looking for an explanation for a strange port flap that I
>>>> experienced this afternoon and out of desperation (and because I can not
>>>> find an answer on Cisco, Google, you name it and I am not yet desperate
>>>> enough to open a TAC case) I am posting here.
>>>>
>>>> Here is the situation:
>>>> Catalyst 3750E has several VLANs. One VLAN has mac-learning disabled.
>>>> The VLAN is tagged on one TenG interface and untagged on a GigE.
>>>> The TenG interface went down due to a transmission problem. The GigE
>>>> interface went Lineproto-down too. Thats weird.
>>>
>>> I would guess that the box connected to the GigE interface has some
>>> kind of heartbeat running towards the other end of the network and
>>> resets itself (or the interface) in case the communication breaks. Like a
>>> "suicide-is-better-than-a-split-brain" protection?
>>>
>>> Seems strange for a Catalyst to drop the link in this case. Is the box
>>> in a stack?
> 
> The 3750E is a 2x3750E stack. The two interfaces are on the same stack
> member (Te1/0/1 is going down, Gig1/0/33 went down too). And yes,
> Gig1/0/33 was the only member left in this VLAN but there is no SVI.
> Plain layer 2 with MAC-learning disabled.
> 
> To make it a little more strange: the link down of Gig1/0/33 was
> actually not logged on the 3750E itself. But the switch connected to
> Gig1/0/33 (a 6509 SUP720) saw its link going to up/down when Te1/0/1 of
> 3750E went down. Do I have to make a drawing? :)
> 
> I will try to reproduce it but as these are production boxes I can not
> just do as I wish.

What I forgot to tell you that the interface Gig1/0/33 (that went down on
the connected device) has a l2 protocol tunnel configured! :-O

 switchport access vlan 82
 switchport mode dot1q-tunnel
 logging event bundle-status
 logging event status
 load-interval 30
 speed 1000
 duplex full
 l2protocol-tunnel cdp
 l2protocol-tunnel stp
 l2protocol-tunnel vtp
 l2protocol-tunnel point-to-point pagp
 l2protocol-tunnel point-to-point lacp
 l2protocol-tunnel point-to-point udld

Could the protocol tunneling have an influence on any link state
forwarding from the uplink to the access port?

Thanks
Sascha



More information about the cisco-nsp mailing list