Nexus 1000v Uplink Ports Broadcast Behaviour

[dumlutimuralp]

Hi,

For Cisco UCS Fabric Interconnect, there are very good explanations in terms of packet forwarding behaviour at Brad Hedlund' s site.

However for Nexus 1000v , in the deployment guide, about upstream switch connectivity it is being said that,

"Defining two standard uplinks to carry the same VLAN involves the risk of creating loops within the environment and is an unsupported configuration. Cisco NX-OS will post warnings when such a condition occurs. "

First of all I have recently deployed Nexus 1000v. No such warnings have been posted. Opened a TAC case and been informed that it is gonna be corrected. I mean when a user accidentally forgets to create a port-channel from two uplinks however configure the same VLAN on them, there will be an actual warning on Nexus CLI.

What I am curious about is, we all know that the forwarding behaviour of Nexus 1000v is, if it receives a packet destined to MAC A, it checks whether if MAC A is a locally known MAC address (meaning that MAC A belongs to a VM on a vEth port), if it is not , the packet simply get dropped.

What if we receive a packet destined to FFFF.FFFF.FFFF ingress on Uplink#1 on NExus 1000v, meaning that it is a broadcast packet and we accidentally configured the same VLAN on two different Uplinks on NExus 1000v without creating a port-channel. For sure this packet should be forwarded to all VMs on vEth ports.

However if this packet is being recevied on Uplink #1 , it should not be forwarded back through uplink #2 ?

In real world I tested this scenario and it didnt create any loops. But I am surprised that this scenario is not well explained anywhere ?

thanks for comments.

Dumlu

[freost]

Hi. Below is a link that discusses N1k and loop prevention. as N1k doesnt run spanning tree.

https://supportforums.cisco.com/message/3057658

Youre right that broadcasts could be received on both non port channelled trunk interfaces but nothing is done except a possible log entry. If standby VSM ctrl VLAN receives the broadcast it should report an error such as 'Duplicate VSM addr received' or something like that. This is because we only want 1 designated receiver of broadcasts so a log will be written but since N1k doesnt run STP a loop wont occur(see explanation in link above).

Multiple connected trunks without a port channel in the system uplink profile can results in unstable behavior but youre correct that the packets wont be forwarded back through Nk1 interfaces. 

HTH.

[dumlutimuralp]

Hi,

Actually the url you referenced is the exact explanation from Nexus 1000v Official Deployment Guide version 2.However that explanation does not cover the scenario which I raised this question about. I am aware that Nexus1000v does not STP and why it does not need to run STP.

The thing is, saying "one of the uplink ports is selected as a broadcaset receiver link" or saying "the packets wich are received on an uplink are not forwarded through another uplink" makes a huge difference and I still can not see such sentences in any of the NExus1000v documents.

Cheers.