Nexus 5548UP Pair - Layer 3 configuration

[Jason Whitehead]

I am in need of some guidance,

I have a pair of Nexus 5548UP switches with layer 3 daughter cards. I have setup vPC, and have a layer 2 vPC operational to a 3750 switch.

I now need to configure the 5548 Pair to route to layer 3 3750 switch, How do I set this up so they route to eachother while using both of the full 10 Gig links?

Thank you.

[Jon Marshall]

Jason

Just to clarify. You say you have a vPC to a 3750 switch. Is that switch A in your diagram ie. the L3 3750 is a different switch ?

If so if you want to route between the Nexus switches and the 3750 L3 switch then i would use 2 L3 P2P links between the 3750 and each Nexus switch and run a routing protocol between the 3 switches.

If i have misunderstood please clarify.

Jon

[Jason Whitehead]

Hello Jon,

Yes I have configured a vPC to switch A in the diagram, the L3 3750 is a different switch.

Should I make the ports on the Nexus switches layer 3 ports and run a routing protocol like EIGRP between them and the L3 3750?

Thank you for your fast response

[Jon Marshall · Accepted solution]

Jason

That is what i would do yes ie. "no switchport" then use 2 x /30 subnets, one for each link then use a dynamic routing protocol.

Jon

[Jason Whitehead]

Okay, for the nexus side, I shall create a layer 3 port on each switch in different subnets.

for the 3750 side I will create two layer 3 ports matching the nexus subnets and configure EIGRP on both ends.

I am rusty on my routing, these will be equal path cost links, will load balancing happen automagically?

thanks!

Nexus1

int e1/1

no switchport

ip add    10.10.1.1 /30

Nexus2

int 1/1/

no switchport

ip add 10.10.2.1/30

3750

int Te2/1/1

no switchport

ip add 10.10.1.2 255.255.255.0

int Te2/1/2

no switchport

ip add 10.10.2.2 255.255.255.0

[Jerry Ye]

You need one more L3 link between the N5K's. The reason is HSRP in the Nexus will forward traffic even though it is HSRP standby. In this case, if a uplink, N5K1 <-> 3750, is down, any traffic hits that HSRP interface will not be able to forward any traffic (black hole). With this L3 link, it will forward traffic to N5K2 and out. It will not be an issue for return traffic.

HTH,

jerry

[Marwan ALshawi]

Just to Add to Jerry and Jon posts

in your topology you have the L2 side has been setup with vPC that ok

- from the N5K up to the 3750 it is better to have two L3 links as Jon suggested above and you need to setup your routing to run with equal cost multipathing ECMP so that you can get the benefit of using all your links in the vPC and up to the 3750 Stack

- as Jerry said you will need an additional link ( not the vPC peer link ) to be used i would say you can use a L3 interface with and included in the routing configuration

- also you need to be aware if you are using HSRP in the N5K no to use HSRP tracking while using L3 links to the core or the 3750 switch with ECMP

- Should a peer link failure occur, the SVIs on the operational secondary will be shut down, which enforces only one possible active topology: that is, the SVI on the primary Cisco Nexus 7000 Series Switch is going to be the active HSRP interface. If you want to remove an SVI from this behavior you should use the command

dual-active exclude interface-vlan > in the vPC domain configuration

hope this help

[Jason Whitehead]

Hello Marwanshawi,

Thank you for your response, just to clarify:

Will the links between the N5Ks and the 3750 stack will be seperate layer 3 links on different subnets or will they be apart of a single layer 3 vPC?

Along with my 2 10G connects between Nexus1 and Nexus2 peer links, I will have an additional connection between the switches that is layer 3 and I will include it's network in the advertised routes.

While using HSRP between the N5K pair to not track any interfaces.

Let me know if I have missed anything,

Thank you

[Jerry Ye]

The links between the N5Ks and 3750 should be on its own network (/30 P2P). The L3 cross link between the N5K should be a different network also (/30 P2P). All these can be advertised via the routing protocol.

HTH,

jerry

[Marwan ALshawi]

Hi Jason,

You need the interfaces to be L3 interfaces routed interfaces in it's own subnet as Jerry mentioned above

And the interswitch link too and all being part ofnthe routing and make sure the two uplink l3 to be configured in the routing with equal cost/metric to achieve ECMP

For HSRP yes no tracking to a avoid vPC blocking in case of one uplink down and the traffic will go over the vPC peer link and get blocked ( I mean server to server traffic )

Hope this help

Plz rate the helpful posts

[Jason Whitehead]

Hello,

I have the vPC's and EIGRP running great,  thank you for all the help and fast responses.