Nexus vPC | Orphan Ports

[zekebashi]

Hello,

Topology is as follows:

vPC Domain 10
NX5k_A peer member is connected to NX5K_B peer member via peer-link port
HostZ is connected to both vPC peers using a vPC114/port-channel114)
HostZ is using vPC VLANs (VLANs are members of the peer-link)

HostZ is an ESXi.

Here's the question:
- The output from the lacp neighbor command shows that HostZ is not establishing lacp peering with neither NX5K_A nor NX5K_B; however, the output from the show vpc command shows that the vpc is up/up.

Note: Command used to verify LACP establishment " sh lacp neighbor interface port-channel 114"  -- -See attached.

My thought is that from both vpc peers, the vpc is setup properly; however, HostZ isn't configured for LAG/Port-Channeling!

Is HostZ considered a single-attached host and the vPC member ports are considered orphan ports?

Thanks in adavance,
~zK

[MHM Cisco World]

HostZ what represent ?

[zekebashi]

an ESXi.

[MHM Cisco World]

this ESXi have 1000v?

[zekebashi]

No, It's a standard ESXi. 

[MHM Cisco World]

It depend on design, 
it seem same BUT 
with vPC the ESX is connect to double-side vPC 
why they recommend not use vPC?

[zekebashi]

There is a significant difference between the two topologies you posted:

 - The one on the left is a double-sided vPC design. Our discussion pertains to this design.

 - The one on the left (MAC Pinning) is the traditional design where the top layer switches are using an interlink to     

    provide redundancy for the lower-layer switches(downstream devices). Not relevant in our discussion.

-  The reason why; I am not quite sure why the system team opted to use individual-links instead of port-channeling.

HTH

[paul driver]

Hello
Does the esxi host support or is it enabled for LACP?

Have you tired using a static vpc instead?

[zekebashi]

Hello,

I believe the esxi host supports LACP; however, the system team opted not to use it.

When saying static vPC. you mean port-channel in ON mode?

Thanks in advance.

~zK

[Reza Sharifi]

What is the output of "show port-channel summary"?

As Paul also said, you may want to try the static Portchannel which is mode "on" as sometimes ESXi hosts don't play well with Cisco

Portchannel. If none of these works, just try using the 2 physical interfaces (no Portchannle) with the ESXi host and test redundancy.

HTH

[Reza Sharifi]

When saying static vPC. you mean port-channel in ON mode?

That is correct.

HTH

[zekebashi]

Here's the output of the show port-channel summary:

NX5K_A:

--------------------------------------------------------------------------------
Group   Port-       Type   Protocol  Member Ports
            Channel
--------------------------------------------------------------------------------

114      Po114(SU) Eth   NONE       Eth10/1/1(P)

NX5K_B:

--------------------------------------------------------------------------------
Group   Port-       Type   Protocol  Member Ports
            Channel
--------------------------------------------------------------------------------

114      Po114(SU) Eth   NONE       Eth10/1/1(P)

=======

Config of port-channel on NX5K_A and NX5K_B:

interface port-channel114
description ESXi_HostZ
switchport mode trunk
no lacp suspend-individual
spanning-tree port type edge trunk
speed 10000
vpc 114

----

==============================

Output of show lacp command:

NX5K_A# sh lacp neighbor interface port-channel 114
NX5K_A#

NX5K_B# sh lacp neighbor interface port-channel 114
NX5K_B#

================================

"Portchannel. If none of these works, just try using the 2 physical interfaces (no Portchannle) with the ESXi host and test redundancy."

The system team already chose this method; however, in this case the ESXi will be operating in Active/Standby, right?

[Reza Sharifi]

The outputs look good. 

The system team already chose this method; however, in this case the ESXi will be operating in Active/Standby, right?

I think so. So, basically, one interface is active until it fails and then the other one takes over. This usually works well with ESXi hosts unless the servers are Cisco UCS, which work with Portchannels fine.

HTH

[zekebashi]

So, we can conclude that the following:

- HostZ(ESXi host) is attached to each vPC peer device, each of which has a vpc member port configured in a vPC/port-channel

- HostZ(ESXi host) carries "vPC VLANs" = vPC attached

- HostZ(ESXi host) is not configured in any type of LAG (static nor LACP); therefore, it is operating in active/standby mode

- HostZ(ESXi host) is considered single-attached to each vPC peer and each port is considered an "Orphan Port"

~zK

[Reza Sharifi]

Yes, That is correct.  Without any Portcahnnel configuration, you are just using the physical interfaces in trunk mode. Since the server is connected to both switches, they are not considered "Orphan Port".

HTH