UCS vNIC configuration best practices

[dganta]

Hi Team,

We are using a Hyper-V environment in UCS B-series Chassis wth FI 6200, We would like to know what is the best practic for the vNIC should we enable hardware failover at the UCS level or just create 2 vNICs one for A and one for B and then enable port-channel in active-active mode. Also we are using an Internal port-channel between the IOM module and the FI 6200.

Thanks,

Dinesh

[Kirk J]

Greetings.

Take a look at http://www.cisco.com/c/dam/en/us/solutions/collateral/borderless-networks/advanced-services/Whitepaper_C11-730725.pdf

It appears it is still Fabric Failover that is preferred, although there will be use cases for teaming.

Please keep in mind that the FIs do not share a dataplane so you cannot create host nic teams utilizing LACP/port-channels (the FIs do not do VPC)

Thanks,

Kirk...

[dganta]

But we can do switch independent Nic teaming or port-channel in Hyper-V.

In Vmware with the same environment without enabling fabric failover we have created active active adapter using DVS and standard switch. Is this is a standard port-channel in that case or it is not a port-channel and just 2 vnics one on A and one on B doing loadbalancing

[Kirk J]

It needs to be independent.  Having two links from the host (each going to a separate switch/FI) will not establish a port-channel.

You need to mimic what you setup in VMware as a team with 'originating port-id' setting.

You are effectively in an active-active for outbound, but inbound is always through one port (unless link failure/failover occurs)

If these were rack servers connecting to a pair of N5K switches in a VPC, then you could use a switch dependant/Port-channel.

In order for a device to form a port-channel with two separate switches, those switches need to a share control plane, and behave as one logical device, in order for that to work.  Examples would would be Stackwise setup on some Catalyst switches, VSS, or VPC on Nexus switches.

Thanks,

Kirk...

[dganta]

Thanks a lot for that Clarification Kirk.

[joehoughes]

Even though these articles are written specifically to VMware, the underlying principles relate to network uplinks, layout and failover for any virtual environment.  These may help you to form your decision:

Trade-off Factor – Cisco UCS Fabric Failover or OS based NIC Teaming using Dual Fabric

Example Architectural Decision – vMotion configuration for Cisco UCS

Why You Should Pin vMotion Port Groups In Converged Environments