Etherchannel from 1 stack to 2 non cisco devices

[joris.jeannelle]

Hi,

Here is my issue. We have 2 sites connected via a VPLS from a provider. We have L3 switches with a vlan SVI in the shared VLAN to route between sites. The VLAN shared between both site is a interconnection network (/30).

The client requested a redundant link from the VPLS provider. The provide will bring a new fiber and end device that will give access to the shared VLAN.

The question is how can I handle the redundancy on my L3 switch (9300). Ideally I would create an etherchannel to avoid having to deal with STP.

Here is the diagram (L2 and L3). I don't have the hand on the provider edge device, it is NON Cisco and the provider is not open to change it's configuration

[balaji.bandi]

As per the diagram both the side are different network, they are not Layer 2 extend  network, (considering you do not need Layer 2 extention herre)

you going to have new link added both the side,if no Layer 2 requirement, i take advatage of Layer 3 with OSPF ECMP LB.

is this make sense ? or am i miss understood your requirmeent ?

[joris.jeannelle]

Indeed I Wouldn't need a L2 VPLS between both sites and use an L3 link (MPLS or other) but that is what is in place here and I can't change that.

[balaji.bandi]

Then IGP is good solution here, but in case you want to change the desing @Reza Sharifi  suggest make it Layer 3 Lag if possible,

if this is too much change, then go with orginal plan with layer 3 / IGP.

[Reza Sharifi]

Hi,

The client requested a redundant link from the VPLS provider. The provide will bring a new fiber and end device that will give access to the shared VLAN.

VPLS is a point to multipoint technology and it is designed for this type of scenarios but in order to avoid a loop on the site with 2 connections from the PE routers, you need to ask them to configure LAG on their site and you do the same on your site on the 9300 switches. The providers use standard protocols for LAG and so you would need to run LACP with the provider. If the devices are not Cisco, they are most likely Juniper switches. Ask them to configure LAG on both connections if they don't want to change their site then you have to make sure you run STP to block one of the ports.

HTH 

[joris.jeannelle]

It's not even Juniper it is MetroNID GT. I am not sure it supports LAG and even then the provider does not want to change to configuration of their device. They use a template to configure the device and we are not a big enough client to make a change from their template.

Let's assume I have 2 ports on my 9300 in the shared VLAN with Rapid PVST. What convergence time I am looking at in case of a cut ?

The other thing is I know that I will work with spanning tree but for me Spanning is a loop prevention Mechanism and not a redundancy protocol. I don't like using it that way.

[Reza Sharifi]

It's not even Juniper it is MetroNID GT. I am not sure it supports LAG 

It may be that they use MetroNID GT just as a layer-1 device to connect you to their MPLS/VPLS network on their site where the routers and switches (PE devices) are located. So, the LAG would take place between your 9300 and a device located inside the VPLS cloud in your diagram. So, think of MetroNID GT as just a media converter and nothing more and should be transparent for the LAG.

What convergence time I am looking at in case of a cut ?

If they don't want to make changes on their site, deploy RSTP, the convergence time is about 2 seconds. You should be able to test it in a maintenance window once the services are up and running to make sure things failover as expected when you need it.

HTH

[Georg Pauwen]

Hello,

I assume the redundant link does not use the same VPLS network as the existing connection ?

Since both are L3 switches, the easiest would probably be to create a L3 link between both switches, and then use e.g. an IP SLA or policy routing for the failover. That way, you would not have to deal with spanning tree at all.

[joris.jeannelle]

The secondary link is from the same provdier as the primary one. So I assume it will be the same network.

What do you call a L3 link ?

[Georg Pauwen]

What I meant was a simple /30 network between either two routed interfaces, or two SVIs, each on one switch.

However, since you say that the same VPLS network is being used as for the existing connection, that probably won't work anyway. I don't see how you can resolve this with provider intervention, since you will need to connect the new switch to one of the PE routers no matter what.