Wednesday, August 19, 2015

Getting started with Nano Server for Compute and Cluster

I assume you have heard the news that Windows Server and System Center 2016 TP3 is publicly available by now.

This means you can download and play around with the bits in order to get some early hands-on experience on the available scenarios and features.

One of the key scenarios that’s available in this preview is the following:

·         Nano Server (enhanced – and covered in this blog post)
·         Windows Container (new – and very well explained by Aidan Finn at www.aidanfinn.com )
·         Storage Spaces Direct (enhanced – and covered here very soon)
·         Network Controller (new – and covered here in detail very very soon J )

So, let us start to talk about Nano Server.

During Ignite earlier this year, Nano Server was introduced by the legend himself, Mr. Snover.
Let us be very clear: Nano Server is not even comparable to Server Core, that Microsoft has been pushing since the release of it, where you run a full Windows Server without any graphical user interface. However, some of the concepts are the same and applicable when it comes to Nano.

Some of drivers for Nano Server was based on customer feedback, and you might be familiar with the following statements:

-          Reboots impact my business
Think about Windows Server in general, not just Hyper-V in a cluster context – which more or less deals with reboots.
Very often you would find yourself in a situation where you had to reboot a server due to an update – of a component you in fact wasn’t using, nor aware of was installed on the server (that’s a different topic, but you get the point).

-          What’s up with the server image? It’s way too big!
From a WAP standpoint, using VMM as the VM Cloud Provider, you have been doing plenty of VM deployments. You normally have to sit and wait for several minutes just for the data transfer to complete. Then there’s the VM customization if it’s a VM Role, and so on and so forth. Although thing has been improving over the last years with Fast-File-Copy and support of ODX, the image size is very big. And don’t forget - this affects backup, restore and DR scenarios too, in addition to the extra cost on our networking fabric infrastructure.

-          Infrastructure requires too many resources
I am running and operating a large datacenter today, where I have effectively been able to standardize on the server roles and features I only need. However, the cost per server is too high when it comes to utilization, and really make an impact on the VM density.
Higher VM density lower my costs and increases my efficiency & margins.

I just want the components I need….and nothing more… please

So speaking of which components we really need.

Nano Server is designed for the Cloud, which means it’s effective and goes along with a “Zero-footprint” model. Server Roles and optional features live outside of the Nano Server itself, and we have stand-alone packages that we adds to the image by using DISM. More about that later.
Nano Server is a “headless”, 64-bit only, deployment option for Windows Server that according to Microsoft marketing is refactored to focus on “Cloud OS Infrastructure” and “Born-in-the-cloud applications”.

The key roles and features we have today is the following:

-          Hyper-V
Yes, this is (If you ask me) the key – and the flagship when it comes to Nano Server. You might remember the stand-alone Hyper-V server that was based on the Windows Kernel but only ran the Hyper-V Role? Well, the Nano Server is much smaller and only is based on Hyper-V, sharing the exact same architecture as the Hypervisor we know from the GUI based Windows Server edition.

-          Storage (SOFS)
As you probably know already, compute without storage is quite useless, given the fact that Virtual Machines is nothing but a set of files on a disk J
With a package for storage, we are able to instantiate several Nano Servers with the storage role to act as storage nodes based on Storage Spaces Direct (shared-nothing storage). This is very cool and will of course qualify for its own blog post in the near future.

-          Clustering
Both Hyper-V and Storage (SOFS) relies (in many situations) on the Windows Failover Cluster feature. Luckily, the cluster feature servers as its own package for Nano Server and we can effectively enable critical infra roles in a HA configuration using clustering.

-          Windows Container
This is new in TP3 – and I suggest you read Aidan’s blog about the topic. However, you won’t be able to test/verify this package on Nano Server in this TP, as it is missing several of its key requirements and dependencies.

-          Guest Package
Did you think that you had to run Nano Server on your physical servers only? Remember that Nano is designed for the “born-in-the-cloud applications” too, so you can of course run them as virtual machines. However, you would have to add the Guest Package to make them aware that they are running on top of Hyper-V.

In addition, we have packages for OEM Drivers (package of all drivers in Server Core), OneCore ReverseForwarders and Defender.

Remote Management

Nano Server is all about being effective, leverage the cloud computing attributes, being effective, scalable and achieve more. In order to do so, we must understand that Nano Server is all about remote management.
With a subset of Win32 support, PowerShell Core, ASP.NET5, we aren’t able to use Nano Server for everything. But that is also the point here.

Although Nano is refactored to run on CoreCLR, we have full PowerShell language compatibility and remoting. Examples here are Invoke-Command, New-PSSession, Enter-PSSession etc.

Getting started with Nano Server for Compute

Alright, so let us get over to some practical examples on how to get started with Nano Server for Compute, and how to actually do the configuration.

Originally, this blog post was a bit longer than it is now, since Microsoft just published some new content over at TechNet. Here you will find a good guidance on how to deploy Nano: https://technet.microsoft.com/en-us/library/mt126167.aspx

I must admit, that the experience of installing and configuring Nano wasn’t state of the art in TP2.
Now, in TP3, you can see that we have the required scripts and files located on the media itself, which simplifies the process.



1.       Mount the media and dot-source the ‘convert-windowsimage.ps1’ and ‘new-nanoserverimage.ps1’ script in a PowerShell ISE session
2.       Next, see the following example on how to create a new image for your Nano server (this will create a VHD that you could either upload to a WDS if you want to deploy it on a physical server, or mount it to a virtual machine



3.       By running the cmdlet, you should have a new image

In our example, we uploaded the vhd to our WDS (Thanks Flemming Riis for facilitating this).

If you pay close attention to the paramhash table, you can see the following:

$paramHash = @{
MediaPath = 'G:\'
BasePath = 'C:\nano\new'
TargetPath = 'C:\Nano\compute'
AdministratorPassword = $pass
ComputerName = 'nanohosttp3'
Compute = $true
Clustering = $true
DriversPath = "c:\drivers"
EnableIPDisplayOnBoot = $True
EnableRemoteManagementPort = $True
Language = 'en-us'
DomainName = 'drinking.azurestack.coffee'
}

Compute = $true and Clustering = $true.
This means that both the compute and the clustering package will be added to the image. In addition, since we are deploying this on a physical server, we learned the hard way (thanks again Flemming) that we needed some HP drivers for networks and storage controller. We are therefore pointing to the location (DriversPath = “c:\drivers” ) where we extracted the drivers so they get added to the image.
Through this process, we are also pre-creating the computer name object in Active Directory as we want to domain join the box to “drinking.azurestack.coffee”.
If you pay attention to the guide at Technet, you can see how you can set a static IP address on your Nano Server. We have simplified the deployment process in our fabric as we are rapidly deploying and decommissioning compute on the fly, so all servers get their IP config from a DHCP server.

Once the servers were deployed (this took literally under 4 minutes!), we could move forward and very that everything was as we desired.

1)      Nano Servers were joined to domain
2)      We had remote access to the nano servers



Since Nano Server is all about remote management, we used the following PowerShell cmdlets in order to configure the compute nodes, create the cluster etc.

# Preparing your mgmt server

Install-WindowsFeature -Name RSAT-Hyper-V-Tools, Hyper-V-Tools, Hyper-V-PowerShell, RSAT-Clustering, RSAT-Clustering-MGMT, RSAT-AD-PowerShell -Verbose

# Creating Nano Compute Cluster

$clustername = "nanocltp3"
$nodes = "hvtp301, hvtp302"
$ip = "10.0.0.50"

New-Cluster -Name $clustername -Node $nodes -StaticAddress $ip -NoStorage -Verbose

# Connecting to storage server and create SMB share with proper permissions

$storage = "nanostor"

Enter-PSSession -ComputerName nanostor

MD D:\VMS
ICACLS.EXE D:\VMS --% /Grant drinking\knadm:(CI)(OI)F
ICACLS.EXE D:\VMS --% /Grant drinking\hvtp301$:(CI)(OI)F
ICACLS.EXE D:\VMS --% /Grant NTDEV\benarm-zeta$:(CI)(OI)F
ICACLS.EXE D:\VMS /Inheritance:R
New-SmbShare -Name VMS -Path D:\VMS –FullAccess drinking\knadm, drinking\hvtp301$, drinking\hvtp302$

# Configuring Constrained Delegation

Enable-SmbDelegation –SmbServer $storage –SmbClient hvtp301 -Verbose
Enable-SmbDelegation –SmbServer $storage -SmbClient hvtp302 -Verbose

# Configure Hyper-V settings for Cluster usage

$vmhosts =@("hvtp301", "hvtp302")
$vhdpath = "\\nanostor\vms\"
$vmconfigpath = "\\nanostor\vms\"
$lmsettings = "5"

foreach ($vmhost in $vmhosts)
    {
        Set-VMHost -ComputerName $vmhost -MaximumVirtualMachineMigrations $lmsettings -VirtualHardDiskPath $vhdpath -VirtualMachinePath $vmconfigpath -VirtualMachineMigrationAuthenticationType Kerberos -Verbose
    }

# Create VM based on Nano Image

$vm = "nanovm1"
$nanohost = "hvtp301"

New-VM -ComputerName $nanohost -Name $vm -MemoryStartupBytes 512mb -VHDPath \\nanostor\vms\blank1.vhd -SwitchName VMSwitch -Generation 1 -Verbose

# Make the VM highly available

Add-ClusterVirtualMachineRole -VMName $vm -Cluster $clustername -Verbose

# Start the VM

Start-VM -ComputerName hvtp301 -Name $vm -Verbose

As you can see, we are also creating a virtual machine here that is obviously based on a vhd with the guest drivers installed. We tested on how to do this manually by using DISM on an empty image.

The following example can be used in order to service your Nano vhd.

# Nano servicing

# Create a mountpoint

md mountpoint

# Mount the image into the mountpoint you just created

dism /Mount-Image /ImageFile:.\blank.vhd /Index:1 /MountDir:.\mountpoint

# Add your package. In this example, we will add packages for Storage, Cluster and Virtual Guest Services

dism /Add-Package /PackagePath:G:\NanoServer\Packages\Microsoft-NanoServer-Guest-Package.cab /Image:.\mountpoint

dism /Add-Package /PackagaPath:G:\NanoServer\Packages\Microsoft-NanoServer-FailoverCluster-Package.cab /Image:.\mountpoint

dism /Add-Package /PackagePath:G:\NanoServer\Packages\Microsoft-NanoServer-Storage-Package.cab /Image:.\mountpoint

# Commit and dismount

dism /Unmount-Image /MountDir:.\mountpoint /commit

# Copy the vhd over to the smb share for the compute cluster

Copy-Item -Path .\blank.vhd -Destination \\nanostor\vms -Verbose

The following screen shot shows the Nano Cluster that is running a virtual machine with Nano Server installed:



NB: I am aware that my PowerShell cmdlets didn’t configure any VMswitch as part of the process. In fact, I have reported that as a bug as it is not possible to do so using the Hyper-V module. The VM switch was created successfully using the Hyper-V Manager console.

Happy Nano’ing, and I will cover more later.

(I also hope that I will see you during our SCU session on this topic next week)


No comments: