I’ll have to re-read that section to be sure. I would have meant that a license pack applies to two cores. You cannot buy an individual license for a single core.

According to your ebook, for Windows Server 2016: “A single license is applied to two CPU cores in the same physical host”.

It confuses me because it contradicts with what Microsoft said in their licensing guide and example (WindowsServer2016-Licensing-Guide.pdf): “the number of core licenses required equals the number of physical cores on the licensed server”, i.e. one license one core (not two cores as you said).

This guide can be downloaded from Microsoft site. I hope you could spare some time to check it.

Many thanks for the guides on Hyper-V. They’re really, really great.

Nat,

I will trust that you have validated your license count with a credentialed specialist. That math looks strange to me, but I’m not the one who will be slapped with a fine.
Still not making any sense to me why there are so many LUNs. Also not understanding why a company would drop the money for that many hosts and a SAN but then balk at a few thousand dollars for licensing.
Replica is for disaster recovery. If all of your hosts live in the same site and use the same SAN, then the list of disasters that you are protected from is very short. It’s a lot of effort and overhead with next to zero payout.
If you can’t cluster the VMs over a few hundred dollars, then you’re better to just leave the third host off and connect it to the necessary LUNs if a host fails. The 90 day rule is waived in the event of hardware failure. But it’s still better to cluster.

But this whole thing is ridiculously over-architected. Yeah, it will work, but whatever they think they’re saving in licensing fees will be lost in administrative effort. I’m already at the point where I would need to start charging my own consultancy rates even to think about this build anymore. And it’s only 6 VMs.

Hi Eric,

I understand this is over-architected. The company where I’m setting up this config already has 4 physical servers. One of them is going to be replaced for a new one (HV1). First my thought was to setup a cluster with the new host and one of the older servers. But then I miscounted the double license costs (each host needs enough licenses to run all VM’s), and also the amount of RAM necessary on the old server should be enough to run all VM’s on it.

With the four licenses I meant 32*9EM-00301 in total 64 licenses, enough for three 1CPU servers with on one server 2 licenses stacked, so it can hosts 4 VM’s.

That’s the reason to use both Hosts not clustered (for now. When the 2nd server will be replaced HV1 and HV2 will be setup in a clustered environment). That’s also the reason for me to use the third host as backup in case of a failure of one of the primary hosts. The third host is available, has licenses to hosts 2 VM’s and has enough ressources to run 2 VM’s.

From your response I understand not to use the storagepool function of Server 2016, because the SAN is dealing with that. What isn’t clear to me is why it’s pointless using 3 hosts and one SAN and replica in the same site. Is it because the third host can startup the VMs of HV1 and HV2?
Microsoft told me that I couldn’t run VM1 through VM6 on HV3 in case of HV1 or HV2 should have a failure. Only once in 90 days it would be in compliance to do that.
That’s the reason I was thinking to replicate the VM’s to HV3, but it isn’t my first choice.
I just can’t figure it out if it would be legible to map the LUNS from the SAN also to HV3 and start VM 3, 4, 5 or 6 from HV3 (only in case of failure or perhaps in case of reconfiguration of HV1 or HV2).

With import I meant to import the VM in the Hyper-V manager from the VMCX / VMRS file which defines the VMs, created on HV1 or HV2.

I hope this makes it a bit more clear and although it isn’t the most suitable solution I would like to know if it will be legible and if it will be working…

That’s excessively over-architected. Why use storage pools if you have a SAN? And I don’t really understand your angle with backup hosts. I’ve seen builds of 250 VMs that are simpler than what you have. That looks like a management nightmare.
I hope that you’ve talked to your reseller about 2016 licensing. Two 2016 licenses don’t get you anywhere at all. Were they SA from 2012 licenses?
If you’ve got three hosts, six VMs, and one SAN, all in one site, then replica is pointless. And what’s this about import/export? I don’t understand what you’re going to accomplish with all of that.
A couple of LUNs and a failover cluster is what this calls for. Backup that can be transported offsite is what you need, not replica. If you want replica, take one of those hosts offsite, find a storage unit for it, and replicate that way.

Thank you very much for your articles and all replies!

I’m currently setting up this environment:

HV1 = host1 on 1 CPU server with two license of 2016 Std. (for 4 VM’s).
HV2 = host2 on 1 CPU server with it’s own license of 2016 Std. (for 2 VM’s).
HV3 = host3 on 1 CPU server with it’s own license of 2016 Std. (for 2 VM’s).

(On all three hosts only the Hyper-V role is installed. And all four licenses have SA.)

One SAN attached to all three hosts with MPIO.

LUN1 -> VM1 on SAN
LUN2 -> VM2 on SAN
LUN3 -> VM3 on SAN
LUN4 -> VM4 on SAN
LUN5 -> Data disk for VM3 (2TB) on SAN
LUN6 -> Data disk for VM4 (1,5 TB) on SAN
LUN7 -> VM5 on SAN
LUN8 -> VM6 on SAN

HV1 has discovered LUN 1-4 (with on it installed it’s own VM) and LUN5 and 6 for data.
HV2 has discovered LUN7 and 8 (with on it installed it’s own VM).
HV3 has discovered LUN 1-8, but all disks are offline.

HV1 it the main Host for running VM1 through VM4, so I bring six disks online to be able to import VM1 – 4 in the Hyper-V manager. All four VM’s are running, and VM3 also is using LUN5 as data disk, and VM4 is using LUN6 as data disk.
HV2 is the main Host for running VM5 and VM6, both VM’s are running.

HV3 is a backup host. In the Hyper-V manager I’ve imported VM1 through VM6. After importing the VM’s they are not running and all discovered disks (LUNS) are set offline.

In case of HV2 shuts down, I want to bring LUN7 and LUN8 online on HV3 and startup VM5 and VM6 on HV3.
In case of HV1 shuts down, I want to bring up LUN3 through LUN6 online (2 VM disks and 2 data disks) and startup VM3 and VM4.
VM1 and VM2 are less important in case of a temporary fail of HV1.

I’m confused if this is legible. I read in your post (first hit on ctrl f replica) when I use replication (and create VM7 as replica for VM1, VM8 as replica for VM2 and so on) HV3 even don’t need it’s own license for two VM’s, because it’s covered by the SA on the licenses for HV1 and HV2.

I hope you can tell me what to do:

1a.) use it as described: HV3 has 6 VM’s imported but all turned off. in case of failure of HV2, VM5&6 are turned on, in case of failure of HV1, VM3&4 are turned on.
1b.) use it as described: but only the LUNS are discovered by HV3 all six VM’s are not imported yet on HV3. in case of failure of HV2, I set disk7 and disk8 online and import VM5&6 into the Hyper-V manager and then they are turned on. After repairing HV1 I remove VM5 and VM6 from the Hyper-V manager, bring disk 7 and 8 offline. In case of failure of HV1, disks 3 through 6 are set online and VM3&4 are imported and turned on.

2.) use Hyper-V replica and create VM7 through VM12 on HV3.
3.) use Live Migration to move VM5&6 from HV1 to HV3 or move VM3&4 from HV2 to HV3. (in case of failure live migration isn’t an option…)

4.) If the 1st option is for me the one to choose, will there be any trouble if HV1 shutdown unexpectedly and I want to start VM3 (with LUN5 for the data vhdx) and VM4 (with LUN6 for the data vhdx)?
5.) If the 2nd option is for me the one to choose, can I replicate VM3 to HV3 (without LUN5 for saving storage space on HV3) and replicate VM4 to HV3 (without LUN6 for saving storage space on HV3)? Will I be able to attach the vhdx file from LUN5 to VM7 (as it is the replica of VM3) and attach the vhdx file from LUN6 to VM8 (as it is the replica of VM4)?

I’m looking forward to your response.

Kind regards,
Leander van Gorsel

no

They say, “prior to the workload running on it”, I say, “needs to have sufficient licenses to run the maximum number of virtual machines it can be realistically expected to ever run”. They convey roughly the same meaning, don’t they? Mine is wordier and doesn’t quite leave the door open to buying licenses later if you are going to add VMs, but theirs does. In that regard, their wording is more accurate. At the moment that you go into production, every host in a cluster needs enough licenses to run the maximum number of Windows Server VMs that it might ever host. To use an extreme example, you wouldn’t need to license each host in a 64-node cluster to run all 1,000 VMs separately if it’s not possible for any single host to run 1,000 VMs. You would calculate the maximum that a node could possibly run and license for that number. Of course, Datacenter licensing usually makes sense long before you get to that distinction. Most sites will license each node to run all VMs or choose Datacenter for each node.