Link addition between 7k and 9k on working port-channel

[Garry Kabbay]

Hi Experts

I need your guidance regarding link upgradation in the DC. Below is the Topology. If I will add more links for example 1 or 2 (40 Gig) links in existing topology between 7K and 9 K switches in the port-channel 100. I have following queries :

1. Will it effect existing connectivity like flap or down the existing port-channel?

2. shall i get down time for this activity?

3. what worse can happen or it is simple and can be done without impact ?

Topology:

Regards

Gurbinder

[MHM Cisco World]

We can not predict' but you want to be careful.

When add new port member to PO the STP become crazy.

during time port member internally add to PO the stp detect new path and start elect port' here some traffic disrupt.

To make it safe do this in window.

MHM

[paul driver]

Hello

---

@MHM Cisco World wrote:

We can not predict' but you want to be careful.

When add new port member to PO the STP become crazy.

during time port member internally add to PO the stp detect new path and start elect port' here some traffic disrupt.

---

FYI - this is incorrect - spanning-tree will see the PC as a single logical link, its the pc that will LB the traffic of those links not spanning -tree

[MHM Cisco World]

you need to refresh your ACK 
I mention that many times

MHM

[Joshqun Ismayilov]

can you please share :
show etherchannel summary 
show interfaces trunk
and which port do you want to add te existing PO group 

Thanks,
Joshqun Ismayilov

[paul driver]

Hello
1- no 
2- yes ( always use change control its your friend)
3- It shouldn't cause an issue as traffic wont traverse the new port members in vPC  until they are up plus any selected primary port in the vPC should not preempt to the new port members being added, but I would suggest to refer to 2 and this under change control

[M02@rt37]

Hello @Garry Kabbay 

When adding 1 or 2 additional 40G links to an existing Port-Channel between 7K and 9K switches, it should not disrupt existing connectivity, as Port-Channels are designed to handle the addition of links without causing flapping or taking down the existing Port-Channel.

As long as the new links are configured consistently with the existing ones (same speed, duplex, MTU, etc.), there should be minimal to no downtime.

However, if there are configuration mismatches or issues with LACP negotiation, there could be temporary disruption. The process is typically straightforward and can be done without significant impact, but it’s important to verify compatibility, ensure correct LACP configuration, and monitor performance during and after the changes to avoid potential issues like misload balancing or resource limitations.

While it's true that adding a new port member to a Port-Channel can trigger a temporary STP recalculation, it's not a guarantee that STP will behave unpredictably. However, there is a risk of STP convergence during the time the new link is added to the Port-Channel, as STP detects the new path and may temporarily block or re-elect ports to prevent loops. This can cause a brief disruption in traffic as STP recalculates, as @MHM Cisco World underlines.

[paul driver]

Hello M02@rt37 

---

M02@rt37 wrote:

 as STP detects the new path and may temporarily block or re-elect ports to prevent loops

Apologies I tend to disagree-  stp is a control plane function correct?
As such stp should really only see the logical PC interface and not it underlying physical interfaces, should a member link fail or be added the change shouldn't be seen or reflected to the logical interface that stp is running over., Granted there is always a risk when performing such change but the effects should minimal to either action and especially put against the effects or adding or removing a physical standalone port.

[M02@rt37]

Hello @paul driver 

Thanks for your input, but STP operates at L2 and is designed to prevent loops by considering the physical topology of the network. While it’s true that STP is a control plane function, it still needs to interact with physical interfaces to determine the best path. When a physical link changes state (goes down or up), STP recalculates and may temporarily block or re-elect ports to maintain a loop-free topology. Even if the logical interface remains unchanged, the underlying physical interfaces directly impact STP’s decision-making process.

You’re right that there’s a risk when adding or removing physical links, but the effects can be more pronounced than expected depending on the network’s design and the traffic flowing at that moment

[paul driver]

Hello M02@rt37 @MHM Cisco World 
Apologies for the late reply
Id thought i would preform some POC as we may end up going around in circles on this one.

I performed 2 tests ( normal Port channel between two switch's and also a vPC member switch between a nxos vPC cluster

As you will from these tests no stp event was incurred on either test as/when a new port member was added to an existing PC/vPC along with no stp running on any physical port and obviously no renegotiation of root port/roles etc..

The only continually stp event i logged  before ,during and after the change was a " selected peer-switch RP"
from the secondary vPC peer i envisage this had to do with the peer-switch feature and running the same stp priority's on the vPC cluster (primary & secondary)

Obviously this was ran on simulation eve-ng software however i seems to suggest what i was explaining in the first instance that should a member link fail or be added the change shouldn't be seen or reflected to the logical interface that stp is running over.

[M02@rt37]

Hello and Thanks @paul driver 

That makes perfect sense. Since STP operates at the logical Port-Channel level, adding or removing a member link shouldn't trigger a topology change or cause STP to reconverge...ok.

The only STP-related log you saw—"selected peer-switch RP"— is expected when using the peer-switch feature in vPC, as both vPC peer share the same STP priority and act as a single logical root bridge.

Even though this was done in eve-ng, the behavior should be the same on real hardware...

[MHM Cisco World]

show spanning-tree internal event-history tree 1 interface 
ethernet x/x brief

show spanning-tree internal interactions 

Share output of above 

MHM

[Garry Kabbay]

Thanks For all your valuable suggestion.

I have below config on 9K and 7K side on the physical ports and port channel 100, So when I will add more links then to avoid any disruption : what command I should use for new ports, so these can easily add to current Po without any breakage.

9K (C93600CD-GX) Config:

Interface Eth1/1

ttag

ttag strip

switchport

switchport mode trunk

switchport trunk native vlan <>

switchport trunk allowed vlan <>

mtu 9216

speed 40000

no negotiate auto

channel-group 100 mode active

no shut

Port-channel100 :

switchport

switchport mode trunk

switchport trunk native vlan <>

switchport trunk allowed vlan <>

spanning-tree bpduguard disable

mtu 9216

speed 40000

no negotiate auto

VPC 100

7K (C7710) Config:

Interface Eth1/1

switchport

switchport mode trunk

switchport trunk native vlan <>

switchport trunk allowed vlan <>

speed 40000

mtu 9216

channel-group 100 mode active

no shut

Port-channel 100:

switchport

switchport mode trunk

switchport trunk native vlan <>

switchport trunk allowed vlan <>

speed 40000

mtu 9216

VPC 100

[paul driver]

Hello
Before adding the physical ports the PC, shut the new ports down
Apply the existing cfg taken from a physical interface that is active in same PC to the new port then enable the new physical port.

[Garry Kabbay]

Thanks @paul  for your suggestion.

However, if I will only add "channel-group 100 mode active" to new ports in shut state. then enable these ports. then, will it not be taking the config form Po100 on both 9k and 7 K  ? means complete config and it will work without any impact?

or there is any risk of doing this ?

Regards