MDS Core/Edge Config

[cfries]

I have an established Cisco SAN core with a couple of MDS 9513's.  Right now all our hosts have home runs back to the core.  As we go through and rebuild our datacenter, we are starting to move towards a core/edge type environment and putting MDS 9124's in each rack. 

Would someone be willing to send me the steps to configure these 9124's as edge devices back to our 9513's?  I'm know there are steps on both sides (core and edge) but I am unsure of what these are other than establishing an EISL.  The 9513's are in production and I don't want to run the risk of bring those down because of a mistake I made.  Or as an alternative, can anyone point me in the direction where I might find some documentation on how to accomplish this? 

Thank you in advance.  

Carl

[dynamoxxx]

Carl,

are there going to be a lot of these 9124s ? Take a look at NPV functionality that could be enabled on 9124s and turn them into simpler "pass-through" like switches. Simpler fabric to manage, easier to support.

If you still decided to connect these in regular mode then on 9513 you will need to look at domain configuration and document domain id for each VSAN.

sh fcdomain domain-list

1) make sure 9124s are running the same code level (ideally)

2) create VSANs on 9124 and make sure they have the same number but unique domain id.

3) make VSAN domain id static ( fcdomain domain 22 static vsan 130)

4) if you are going to connect multiple cables from 9124 to 9513, bind multiple links into portchannel.

5) on portchannel only trunk VSANs that need to be on the other switch. For example if you have a "tape" vsan on your core switch and your blade servers do not need connectivity to that vsan ..there is no reason to trunk over to 9124s.

[cfries]

Dynamoxxx,

I really appreciate the response.  We have a small environment here and I don't anticipating needing that many edge switches right now.  I did look into N Port Virtualization earlier and felt it wasn't something necessary at this time but thanks for pointing that out. 

At the moment we have 6 VSANs for each one of our application environments.  For instance, one VSAN for MS SQL, one for ESX, and so on and of course each VSAN has one or more storage systems zoned to them. 

Our Cisco SAN has two 9513s with one being for HBA0 and the other switch being for HBA1.  When deploying these edge switches, how do I make sure we have two paths setup for each host?  Do I need to deploy 2 edge switches in each rack for this to happen? 

The 9124s and 9513s are running the same NX-OS of 5.0(1a). 

We are using DPVM, does this matter when setting the edge switches up? 

Thanks again for the response. 

Carl

[dynamoxxx]

I have not used DPVM so i can't comment if there are any specific nuances. As far as you connectivity options, yes you want to maintain Fabric A and Fabric B topology. So your fabric may look like this:

Fabric A:

9513A ----- 9124A

Fabirc B:

9513B ---- 9124B

Each server HBA0 connect to 9124A and HBA1 to 9124B. 

Do you create VSANs for each major application or for some functional separation ? Unless you are using IVR you are probably "burning" a lot of array ports to be able to provide storage in each VSAN ?

[cfries]

We create VSANs based on application.  For instance, there is a VSAN for Production MS SQL, one for Non-Production SQL, and one for ESX, and so on.  The reason for this is our storage arrays are classfied as Tier1 to Tier3.  We may zone Tier1 to the production SQL VSAN and Tier2 storage to non-prod SQL.  Then we may zone all three to the ESX VSAN.  Does this make sense? 

We are not using Inter-VSAN Routing.  Currently we have two core switches for fabric a and b.  Then we have 4 storage arrays with one of them virtualized behind our Tier1 device.  We are not really hurting for ports right now and don't anticipate this happening for some time. 

I hope that answers your questions.  What you wrote about validating my thinking about having 2 edge switches per rack helps. 

Thanks,

Carl

[santkumd528]

Carl Fries wrote:
  Then we have 4 storage arrays with one of them virtualized behind our Tier1 device.  

what is the technology used behind? from which vendor?

[cfries]

Technologies are Hitachi USP/VM, (2x) AMS2500, and Pillar Axiom 500.