VPC question

[jeremytetart]

Hello Guys,

I have one question for you about vPC.

As shown in attachment, you can see two schemas with three diagrams per schema.

Each schemas are separate.

- Schemas 1, 1-1 & 1-2 are together

- Schemas 2, 2-1 & 2-2 are together


The goal is to understand where the flow pass when Host A ping Host B in vPC topology (without & with failure)

I know :

- schemas 1 & 2 are correct. Hash algorithm determine where the flow pass, either NX5K01 or NX5K02.

- schemas 1-1 & 2-1 are correct. With port link down (schema 1-1) the flow pass by vPC peer-link and without link failure (schema 2-1) the flow pass directly to NX5K02.


However, I don't understand why the network flow pass by the vPC peer-link in case of one vPC port failure (schema 1-1) ?

Is there a mechanism to detect vPC port failure on the upstream switch in order to forward traffic by the other port of the downstream switch ? (exemple in diagram 1-2 & 2-2)

Thank in advance for you clarification

Regards.

[Jon Marshall]

Are the 1-2 and 2-2 examples official documentation or are they your examples ?

I ask because as far as I know an downstream switch cannot detect a vPC port failure for a vPC it is not part of.

Jon

[jeremytetart]

1-2 & 2-2 are my examples.

I ask to me and to the forum if there is a mecanism like spanning-tree backbonefast to detect undirect link failure.

[Jon Marshall]

There is no mechanism that I am aware of.

So the downstream switch has no way of knowing a member port of a vPC of which it is not part has failed.

Jon

[jeremytetart]

I'm thought the same thing ;-)

Jon, do you know how the hash algorithm work to loadbalance the network flow between a Nexus 2K & two Nexus 5K ?

[jeremytetart]

Up

[Jon Marshall]

See this link for details -

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus5000/sw/layer2/513_n1_1/b_Cisco_n5k_layer2_config_gd_rel_513_N1_1/b_Cisco_n5k_layer2_config_gd_rel_513_N1_1_chapter_01000.html#con_1207090

is this what you were asking ?

Jon

[jeremytetart]

Hi Jon,

OK I thought that the loadbalancing was different on NX-OS but not. This is the same as etherchannel on IOS.

https://supportforums.cisco.com/document/12458136/load-balancing-methods-port-channels-nexus#Loadbalancing_Methods