Cisco Nexus 3172TQ-XL vPC

[Jean-Francois Gagnon]

Hi!

We have 2 new Cisco Nexus 3172TQ-XL switches. 

We plan to use those as core switches to uplink to other switches. Also, they will route private VLAN.

I was wondering if we could connect ESXi servers with Teamed nic with IP Hash. (Each NIC to a different Nexus switch on a static Etherchannel)

Thanks!

[Peter Paluch]

Hi Jean-Francois,

I was wondering if we could connect ESXi servers with Teamed nic with IP Hash. (Each NIC to a different Nexus switch on a static Etherchannel)

That should not be a problem. Once you configure vPC peering between the pair of N3172 switches, and configure the physical interfaces going to the same server to be located in the same virtual port-channel, you should be okay.

Please note that is almost always better to use LACP instead of static port-channels. Static port-channels are prone to miscabling and misconfiguration, leading to traffic blackholing, or in worse cases, traffic loops. LACP is a very useful detector of misconfiguration, miscabling, or a brain death of the link partner. Unless you have a very good reason to avoid LACP, I strongly recommend using it.

Feel welcome to ask further!

Best regards,
Peter

[Jean-Francois Gagnon]

Thanks, but static is my only option cause we don't have the licence to support vDS and LACP.

HSRP will also be used for redundancy of gateways.

Any other advice?

[Peter Paluch · Accepted solution]

Hi Jean-Francois,

Oh, I see. Okay.

One thing to watch out for: I do not know how exactly the NIC teaming under ESXi works. If the teamed NICs continue using their own MAC addresses instead of a single shared MAC address for the whole team, the ESXi then must make sure that every NIC is ready and willing to accept frames destined to the MAC addresses of other NICs in the team. This is because with vPC, the switches will mutually exchange and synchronize their MAC tables, and they will both learn the MAC addresses on the respective vPC. Each switch will then believe it can reach the same set of MAC addresses through any of the physical links in the same vPC, meaning that for a destination MAC X, the switch may freely choose a physical link in the vPC that happens to be connected to the NIC with MAC Y. Ideally, all NICs in the same team should share a single MAC address - but this is up to the ESXi to accomplish.

HSRP for redundancy is perfect. Please note that with vPC, both Active and Standby switches perform the routing for the vMAC of the HSRP standby group - so even though you have an Active/Standby state on the control plane, you always have an Active/Active state on the data plane.

Without knowing more about your intended network design, no further suggestions :)

Best regards,
Peter