Cisco UCS fabric extender networking

[samtech]

Hey Guys..

I am new to Cisco UCS and support an environment with 5108 chassis with B200 M3 blades connected with redundant fabric interconnects. These blades are configured as ESXi servers with 2 Distributed switches and 2 uplinks on each switch.

My query is:

A. Does it matter which Fabric Interconnect is primary? How to determine primary??

B. Two connections go to Fabric interconnect from each Fabric Extender. I assume this is for redundancy? Then, there are 4 Network Adapters acting as Uplinks for Hypervisor. How do the ESXi gets 4 adapters when there are only 2 connections on the fabric interconnect?? 

C. Is each port on the fabric extender capable of provisioning\pinning additional uplink NIC's for Hypervisor?

D. There are two connections from Fabric Interconnects to the core switch. Is there going to be an impact on the vSphere environment in the event one of the connection to the core is unavailable?

Your responses will be highly appreciated. Cheers,

[balaji.bandi]

My query is:

A. Does it matter which Fabric Interconnect is primary? How to determine primary??

This configuration you can view in UCS Gui.

B. Two connections go to Fabric interconnect from each Fabric Extender. I assume this is for redundancy? Then, there are 4 Network Adapters acting as Uplinks for Hypervisor. How do the ESXi gets 4 adapters when there are only 2 connections on the fabric interconnect?? 

As per the diagram you have 4 physical adapter connected. (depends on how the configuration done) - the picture not good enough see and advise  how these are configured.

C. Is each port on the fabric extender capable of provisioning\pinning additional uplink NIC's for Hypervisor?

Yes you can configure dedicated interface - make sure you  configure all in change window. it may require some reboots.

D. There are two connections from Fabric Interconnects to the core switch. Is there going to be an impact on the vSphere environment in the event one of the connection to the core is unavailable?

As per diagram it shows only 1 core switch (you need to confirm that), in generally Cisco best practice there will be 2 Cores switches dual homed to FEX.

[samtech]

Thanks BB for your response.Attached diag for your reference.Hopefully this is clear. The basic idea is to extend the uplinks available on the core switch by placing a Nexus 3524x in between FI 6248 and core.

So, what I have understood is that it does not matter which Fabric interconnect is primary\secondary as both have active workloads. I'm thinking the way to do this is below (pls. confirm if looks good)

A. Stop the workflows on one of the fabric interconnect.

B. Disconnect the connection going from core to this fabric interconnect. Connect it to the port configured on Cisco Nexus       3524x and from there to the Fabric interconnect.

C. Start workflows to this fabric interconnect and monitor vSphere connectivity etc...
D. Then repeat this procedure for the other fabric interconnect.

Is there still a need to make one of the FI's as lead?

[balaji.bandi]

Networking point of view looks good.