Q&A from the Windows Server 2016 Preview JumpStart – Part III – Upgrades & Operational Efficiencies

Here we are in part three of the Q&A series, where I’m working through the multitude of questions that were asked during the recent ‘What’s new in Windows Server 2016 Preview’ JumpStart course, which, if you haven’t heard, is now available on-demand, so check it out! In this post, I’ll be walking through the questions focused around upgrading to Windows Server 2016, specifically using the new rolling upgrade capabilities, then wrapping up with some questions under the banner of ‘Operational Efficiencies’, which, between you and I, are a load of cool and handy features that just make life better for the Hyper-V admin.  Here we go…

Rolling Upgrades

Can you upgrade from Windows Server 2012 to Windows Server 2016 or only from Windows Server 2012 R2?

No, as detailed on TechNet, this feature only works for clusters transitioning from Windows Server 2012 R2, to Windows Server 2016. Any earlier editions of Windows Server would have to be upgraded to Windows Server 2012 R2, before you could take advantage of the rolling upgrade.  There are no plans to port this feature back to previous versions of Windows Server.

Does it matter if the VMs running on the cluster, are Generation 1 or 2, when upgrading the cluster?

No, it doesn’t matter, both Generation 1 and Generation 2 VMs are supported running on a Windows Server 2012 R2, and Windows Server 2016 cluster.

Does this work with the free Hyper-V Server 2012 R2 and 2016?

Yes, the process would be no different, however you would need to manage the whole process remotely through a UI, with PowerShell, or with System Center Virtual Machine Manager.

Is there downtime for upgrading the virtual machine configuration version?

Yes. Users of Hyper-V virtual machines may choose to upgrade virtual machines by scheduling a brief maintenance window, turning off virtual machines, and running the Update-VMConfigurationVersion PowerShell cmdlet. This will update the virtual machine version, and enable new Hyper-V features, eliminating the need for future Hyper-V Integration Component (IC) updates.

Production Checkpoints

Does a production checkpoint shut down the virtual machine in order to capture the checkpoint?

No, the VM is 100% online during the checkpoint process. Production checkpoints allow you to easily create “point in time” images of a virtual machine, which can be restored later on in a way that is completely supported for all production workloads. This is achieved by using backup technology inside the guest to create the checkpoint, instead of using saved state technology. For production checkpoints, the Volume Snapshot Service (VSS) is used inside Windows virtual machines. Upon restoration from a production checkpoint, you’ll notice that the VM starts from the off position, which is different from a standard checkpoint, which will start from exactly the same point in time as when the checkpoint was taken.

Are production checkpoints supported with Linux and FreeBSD virtual machines?

Yes. Upon triggering a production checkpoint, Linux virtual machines flush their file system buffers to create a file system consistent checkpoint. The mechanism to flush buffers to create file system consistency was originally implemented for the Hyper-V Live Backup feature, but the same mechanism also used for production checkpoints. FreeBSD does not implement Live Backup, so it also does not implement production checkpoints.

Do you need to merge production checkpoints before enabling Hyper-V Replica for a particular VM?

No, not according to Ben Armstrong, who I checked with for the answer for this question!

Will production checkpoints be supported by the respected SQL Server and Exchange support teams?

Yes. A production checkpoint is using VSS technologies within the virtual machine – the same technology used by many backup vendors.

PowerShell Direct

Does PowerShell Direct work through the integration components?

PowerShell Direct provides a powerful scripting and automation experience with the simplicity of VMConnect. PowerShell Direct runs between the host and the virtual machine, over VMBus, for virtual machines with the integration services installed. You don’t need a network connection or to enable remote management. You do need guest credentials to log into the virtual machine.

Will PowerShell Direct work with a Windows Server 2008 virtual machine?

No, as detailed on this TechNet page, you must be connected to a Windows 10 or Windows Server 2016 host with virtual machines that run Windows 10 or Windows Server 2016 as the guest operating system.

Can PowerShell Direct be enabled and disabled?

No, according to the engineering team, only removing the Hyper-V PowerShell Module will disable PowerShell Direct, however MVP Charbel Nemnom, has detailed lots of useful info on PowerShell Direct, including a couple of ways it can be disabled.

Will PowerShell Direct be available from System Center Virtual Machine Manager, or only directly from a Hyper-V host?

Only directly from a Hyper-V host, however, you could create a PS Session from the Virtual Machine Manager host, to the Hyper-V host, and then jump into the VM from there.

Which credentials are required to use PowerShell Direct into a VM?

Firstly, you’ll need to be logged in with Hyper-V Administrator credentials on the host, and from there, you’ll establish a session into the VM using guest credentials.  This could be a local account, or a domain account.

You mentioned that PowerShell Direct would be great for a hosted/multi-tenant scenario, but don’t you need host access to use PowerShell Direct?

In a multi-tenant/hosted scenario, you would expect the service provider to have administrative access to the fabric, i.e. the Hyper-V hosts upon which the tenant VM’s run. With that in mind, one of the requirements is met. The next requirement that will become important, is whether or not the service provider has an account within the tenant guest operating system. There are multiple scenarios here, where the service provider may have an account. It may be, that the service provider has a account that could be used to add roles, features, or handle other configuration tasks, during the actual creation of the virtual machine. In this respect, having an account to access inside the guest would be beneficial, and could in fact be temporary, with the account cleaned up and removed after creation has completed. Alternatively, if the service provider is offering managed services, it may be the case that they have a designated account within the tenant environment, again, for performing certain administrative tasks.  One of our MVP’s, Kristian Nese, has a good write-up on PS Direct, along with a couple of scenarios where he feels it’s relevant.


If ReFS is now recommended for Hyper-V volumes, is it supported with Cluster Shared Volumes and Failover Clustering?

Yes, and you can see some of the interesting results in VHDX creation times, using ReFS and Cluster Shared Volumes, over on Didier’s blog.

Is Deduplication supported with ReFS?

No. If you feel this is an important feature to have in the base platform, raise the feedback on the Windows Server Uservoice website, in the appropriate section.

Are there scenarios where one would choose NTFS over ReFS? Is ReFS mostly for Hyper-V storage, or are there other suitable scenarios?

Traditionally, ReFS has been aimed at archival type scenarios, with it’s excellent data resiliency capabilities. There’s a great write-up on ReFS, and some of the key goals on this Windows blog. Sure, it’s an old blog post, but it highlights some of the important reasons why ReFS is a future direction, and that the aim isn’t just to build a newer, better NTFS. In Windows Server 2016, the new scenarios that ReFS lights up for Hyper-V, such as the rapid creation of fixed virtual disks, and the rapid merging of checkpoints are just two of the scenarios available today, with more coming in the future I’m sure. Why would you choose NTFS? Well, for Hyper-V in Windows Server 2016, I don’t see many compelling reasons as to why you’d stick with NTFS, and don’t forget, you can always run NTFS inside guests that are stored on ReFS, but for other workloads and applications, there may be dependencies on certain components or features of NTFS, and thus, it would be the right choice over ReFS. Certain apps and workloads, like Exchange 2016, have already declared that they will be supporting ReFS within their deployment topologies, as denoted on TechNet, and I’m sure there will be more in the future.

How to you create a new ReFS drive on a new server?

Firstly, ReFS can’t be the OS drive, that still has to be NTFS, so we’ll focus on the other drives in your system. In essence, there’s not a huge difference in the process of creating an ReFS volume versus creating an NTFS volume. Using Disk Manager as an example, you’ll simply select a drive, create a new volume, and instead of selecting NTFS, you’ll select ReFS. You’ll have many other familiar selections, such as drive letter, allocation unit size etc. You can see a more detailed view of how to create an ReFS volume on this blog post.

If all ReFS does is update pointers/references, will you have more fragmentation issues? If NTFS moves bits around to perform merges, potentially that can reduce fragmentation. Correct? Or am I mistaken?

We’ve not disclosed all the technical details on ReFS in Windows Server 2016 at this time, and, as we’re still in the preview phase, there are still lots of moving parts which will be clarified in detail, in the not-too-distant future.  You’ll be able to learn all about it on TechNet.

Can we format a virtual disk as ReFS or only physical disks?

Any disk, physical or virtual, can have a volume formatted with ReFS, just as long as it’s not the OS volume.

Is there any documentation on ReFS?

The formal documentation for ReFS, specifically on Windows Server 2016, isn’t available just yet, but there are some legacy resources that can provide some useful context and background information. Resources include guidance on TechNet, and useful context on why we developed ReFS along with some of the goals. There’s also, for those of you who want to get that bit deeper, a presentation delivered at the Storage Developers Conference on File System Structures.

That’s all for now!


Say what?