BladeSystem Virtual Connect
cancel
Showing results for 
Search instead for 
Did you mean: 

Best practices for configuring VC Shared Uplink Sets (SUS) in stacked enclosures

chuckk281
Trusted Contributor

Best practices for configuring VC Shared Uplink Sets (SUS) in stacked enclosures

Alexandar was helping a customer with 2 c7000 enclosures:

 

****************

 

We are deploying two stacked enclosures and they will host two RAC clusters that will have their nodes spread between the two chassis. We will use the internal vNets for hearbeat traffic and rest of the ports on the Flex-10 modules for the production traffic.

 

 

  1. What is the best practice for configuring the Shared uplink sets in such cases? We have two options identified:

2x SUS (each sus combines the two Flex-10 in the enclosure in Active-Passive mode)

4x SUS (each SUS encompass only one Flex-10, all is Active-Active)

 

  1. The enclosure are stacked in dual ring topology (ports x5 and x6). How the private vNet will use the stacking link for the heartbeat communication?

**************

 

Input from Vincent:

 

***************

 

Why are you considering a 2-enclosure domain ? Is it “just” for the heartbeat network to stay local and not go to the upstream switches ? You can do that even without joining the 2 enclosures in a domain. Check the paragraph titled “Directly connecting VC Domains” in the VC setup and install guide.

 

The reason I’m asking is that some customers want to spread cluster members over different enclosures to increase the resilience of the solution and that makes sense. But having the 2 enclosures in a single VC domain somewhat negates that benefit: if you happen to “lose” the primary enclosure in the domain, you no longer can manage VC in the remaining enclosure. Servers would still be connected but you can’t change anything in the config, you can’t even look at port statistics, you basically have no VCM until the primary enclosure is restored.

 

I recommend multi-enclosure stacking in 2 cases: if you’re short on 10Gb ports in the upstream switches (doesn’t seem to be the case since you’re considering uplinks on all modules) or if you want to consolidate management consoles and be able to move server profiles between enclosures without VCEM.

 

****************

 

Comments or questions?