UCS Environment reaction without CORE Switches

[McDungus]

Hi all,

Forgive me, as I am going to try to explain this the best I can. 

Background Information:

We have a major downtime coming up soon and we have come to a situation where we are not sure what could happen. Soon, we will be replacing our CORE Switches (Nexus 7ks). These switches have been troublesome and have not been set up correctly from day one. One Core Switch resides at the Datacenter and the other sits across the street on the Business side.  As you can tell, there is no redundancy and the CORE switches have been scheduled to be replaced with two cores on each side. Our contractor is turning off the 7ks before bringing the cables over to the new redundant core switches. (there is more behind this like out of support cards and mismatching firmware, but that's not important for the question)

The situation:

The UCS hosts sit in blade chassis and are connected to two fabric interconnects, which are connected to two Cisco MDS, which are connected to the 7ks. All of our storage is connected to the MDS. All of this resides on the Datacenter side. We know that connectivity to the Business side will be down and that accessing the USC environment will unavailable until the new COREs are in place. 

Questions:

When the 7ks are turned off, what will happen from a UCS standpoint?

How do the servers/hosts react? 

What would be the best course of action? 

If more detail is needed, I can definitely expand upon it. 

Thank you,

McDungus

[Kirk J · Accepted solution]

Greetings.

So your FIs are connected directly to the MDS for native FC storage traffic?

If the N7Ks are providing just networking traffic, then you have a couple of options.

When the FI links that carry your various Ethernet traffic VLANs go down, the default behavior for your vnics that are pinned to these uplinks will be to go down.  While it sounds like you are expecting a complete ethernet traffic outage, there maybe some things you can do to mitigate certain traffic.

If you have server to server traffic that is L2 adjacent (i.e. same subnet & vlan), then make sure the uplinks are all pinned to the same FI side (i.e. side A) by tweeking your ESXi vswitch teaming uplinks (i.e. disable side B uplinks).  Additionally you would need to adjust the default network control policy that is applied to your vnics/vnic templates.  You would want to change the "Action on Uplink Fail" from 'link down' to Warning, during this maintenance window.

Again, this would only mitigate L2 adjacent traffic between blades or Rack servers connected to the UCSM domain.

Traffic that has to go to L3 gateway/SVI will get black holed as it can't reach anything outside the FIs during this period.

Your Native FC uplinks going to MDS will not be impacted, so at least your guestVMs storage shouldn't be impacted.

Kirk...

[McDungus]

This is exactly what I was needing! Thank you so much. This will save us from the need to turning off all the VMs as our contractor suggested.

Thank you again!

McDungus