VMware Networking Community
willster07
Contributor
Contributor

VMWare VxRail Hosts connecting to Cisco Nexus 5K

All of our VMware hosts connect to our Nexus 5K vPC pair.  There's a lot on the internet to digest regarding this setup.  And I have the same questions posted to Cisco forums as well.

Our older HP Blades have a built in Cisco switch and the vPC pairs to those hosts from the vPC pair (primary and secondary) appear to work as one would expect.

The VxRail nodes are connected with dual trunk ports to each of the vPC pairs.  Well.  We lost the secondary and all of the VxRails became completely inaccessible.  No redundancy.

My immediate thought was well they weren't setup as vPC and if you have a Nexus vPC pair that's what you need.  TAC confirmed this.  The fix on the CIsco side is pretty straightforward to create the needed vPC.  The config on the Vmware side, no so much.

Anyone have experience setting up VxRail nodes to a Nexus vPC and configure them to use vPC and very specifically tested losing a Nexus in the pair and having full redundancy?

And please if there's a better discussion board to choose let me know, this is the closest I could find and we're not using NSX (hahaha maybe that's the issue?)

0 Kudos
2 Replies
Sreec
VMware Employee
VMware Employee

Since it will not offer any benefit, you should ideally remove VXRAIL and all VMware port-group connectivity from the VPC connections. These nodes will undoubtedly satisfy all technical specifications and complete all validation stages when connected to redundant 5K using the vSphere teaming policy. Static/LACP configurations with VPC switches are required if you wish to continue using VPC design.

Cheers,
Sree | VCIX-5X| VCAP-5X| VExpert 7x|Cisco Certified Specialist
Please KUDO helpful posts and mark the thread as solved if answered
0 Kudos
willster07
Contributor
Contributor

Well the problem presented itself when the secondary nexus crashed and they were simply trunk links. There were no vPC set up to these hosts. All connectivity to those hosts were lost.

Where we didn’t see an outage was with the vPC to other devices. The fix would appear to be setting up vPC connections to those hosts so we can tolerate the loss of a nexus. 

That’s where the technical challenge resides. Getting the actual VMware side configured. The nexus config is straightforward. 

Looks like on the VMware end I’ll need to create a new dvswitch at a minimum to support the LACP config then move my hosts over to that. 

0 Kudos