vCloud Director 5.1 running hosts with difference hardware in the same cluster?

[SgRddY]

Hello Everyone,

I am running into a bit of a problem and just wondering if anyone has came across it and have a possible solution.

Here is my setup:

Cloud Director 5.1

8 x R710 hosts with Intel Xeon x5690 @3.47Ghz

4 x R720 hosts with Intel Xeon E5-2690v2 @3GHz (these are new hosts to be added)

One cluster for all 12 hosts

The problem I am seeing is that when a user deploy and new vApp and has no customization setting.  They are force to reboot because the Windows is detecting new hardware since the last time it was deployed.  This is because the new hosts have different CPU, RAM, and a lot of other things.  This is causing a lot of problems to our automation process.

Is there a way to mask the VM's so that they can't see the changes in the hardware? 

What are my options?

Thanks,

Ken

[IamTHEvilONE]

You should not blend hardware, especially between two very different generations of Intel hardware.  you can try to enable EVC (enhanced vmotion compatibility) to mask out the extra features of the newer processors.  I'm not sure how effective that will be.

However, we can't help if Windows detects new hardware since it's a different core architecture and may want to change accordingly.

[SgRddY]

I have EVC turned on and it does help with vMotion but not the problem I am seeing.  I know ideally, you would have the same hardware in your cluster but when you are adding capacity, there isn't much that you can do because you can not buy old hardware any more.

Is it possible to have 2 different clusters and one resource pool for vCloud?

Thanks,

[IamTHEvilONE]

you can have an elastic provider, assuming you are using Pay-Go or Elastic Allocation Pools (enable it in general settings).

this means we'll deploy to one cluster until it's full, then overflow to the new cluster.  you can also ask VMs to be moved from one resource pool to another.

To be able to do an elastic Provider, I think you need to use VXLAN to span the two clusters together.

more information on the requirements here

[SgRddY]

I don't have VCD Elastic turn on at the moment.  I would imagine it's not just something that I can enable and add the new cluster in and not making major changes to the Provider VCD and causing an outage. 

I have VXLAN setup and that takes care of that part.  Having 2 clusters will reduce the inconsistency of where and when the VM's will land causing new hardware detection reboots.

I don't think our automation process will be able to detect which cluster to VM's are landing in and be able to deal with the reboots.

This solution worth considering... Thanks for the suggestion.

[IamTHEvilONE]

If you already have VXLAN enabled, you can split the 4 new hosts out to a new cluster ... verify VXLAN works, then add the additional resource pool to the provider.  that's done for the elastic part.

when in elastic mode you'll have a 'primary' and multiple 'secondary' pools.  from a programming perspective, it doesn't change ... since you target an Org vDC (not the pools/clusters which back them).

[SgRddY]

Great... Thanks for the information.