Provisioning Windows Server on a rented machine comes down to one decision made before the first logon: which image you install. A provider template gets you to a desktop in five minutes and arrives with drivers, a licence channel and a pre-set administrator password. A custom ISO takes longer and gives you a clean build with your own edition and partitioning. Both routes have traps, and the traps are different.
This walkthrough covers the template path, the ISO path with its driver and virtual disk pitfalls, and the first-boot hardening that should happen before the machine is exposed to anything. If you have not yet picked a host, our Windows VPS options list which Windows versions are available as templates versus ISO on each plan.
Template or ISO: what you actually get
A provider template is a pre-sysprepped image restored onto your virtual disk. You get an administrator account, often with a password mailed to you or set in the control panel, RDP enabled, and network settings from DHCP. The image typically has generic virtio or SATA drivers already injected, which is the main practical advantage: it will boot on the provider’s hypervisor without you knowing anything about the storage controller.
A custom ISO mounts a Microsoft evaluation or volume-licensed image through the provider’s virtual media feature and runs the normal setup wizard. You choose the edition, set the partition layout, and supply the product key or evaluation period. The advantage is control — no unknown local accounts, no vendor agents, no pre-installed tools you did not ask for.
Template Custom ISO
--------------------------------- ---------------------------------
5-10 minute provision 15-40 minute install
Unknown local admin account You set the credentials
Provider agents may be present Clean OS only
Pre-injected storage drivers You must supply drivers if needed
Licence handled by provider Key or 180-day evaluation
Snapshot/backup integration ready You configure it yourself
The licence question is the one people get wrong. A template usually carries the provider’s SPLA or datacentre licensing; an ISO built from a retail key is your responsibility to activate and your responsibility to keep compliant. If the plan’s price includes licensing, do not reinstall from your own media and then claim the difference — that is a compliance problem and it is easy to avoid by asking before provisioning.
Installing from a template: what to change immediately
First logon should never be on the shipped administrator password. Change it, and create a named administrative account while you are at it, so that audit logs distinguish you from a shared default. Then confirm what is actually running on the box before you trust it.
Get-LocalUser | Select-Object Name, Enabled, LastLogon
Get-LocalGroupMember -Group Administrators
Get-Service | Where-Object { $_.Status -eq 'Running' -and $_.StartType -eq 'Auto' } |
Select-Object Name, DisplayName
Get-CimInstance Win32_LogicalDisk | Select-Object DeviceID, Size, FreeSpace
Get-Volume | Select-Object DriveLetter, FileSystemLabel, FileSystem, SizeRemaining
Look for services whose names you do not recognise, and for a second local account beyond the default ones. Provider monitoring agents are legitimate and usually documented; anything undocumented deserves a search before you disable it. Check the disk layout too — templates frequently provision a single 40 GB system volume and leave the rest of the allocation unformatted.
Installing from a custom ISO
Mount the ISO through the provider’s virtual media interface, set the boot order to CD first, and reboot into setup. The wizard is the familiar one: language, edition, custom install, partition. Two points deserve attention.
First, Drivers. If setup reports no disks, the image lacks the storage controller driver. On KVM this is usually the VirtIO SCSI driver, which is not on the Microsoft ISO; you must load it from a separate virtual floppy or ISO supplied by the provider. This is the single most common failure on a custom ISO install, and it is not a sign the disk is missing.
Second, Editions. Pick the edition that matches your licence and your intention. Server Core has no GUI and a much smaller patch footprint; Desktop Experience gives you the familiar shell. You can add the GUI later with Install-WindowsFeature Server-Gui-Mgmt-Infra,Server-Gui-Shell and a reboot, but removing it is more awkward, so decide at install time.
REM after first boot, from an elevated prompt
DISM /Online /Get-CurrentEdition
DISM /Online /Get-TargetEditions
DISM /Online /Set-Edition:ServerStandard /ProductKey:XXXXX-XXXXX-XXXXX-XXXXX-XXXXX /AcceptEula
Evaluation media rearm up to a limited number of times; track the deadline rather than discovering it when the machine starts shutting down hourly. Convert to a licensed edition early if the host is going into production, not at the last minute.
Partitioning and second disks
A 40 GB system volume is tight but workable for a role server with no data; it is not workable for one that stores logs, databases or user profiles. During setup, if the virtual disk is larger than the default partition, extend the system partition to use the whole disk rather than leaving unallocated space behind — it is easier at install time than afterwards.
Get-Partition | Select-Object DiskNumber, PartitionNumber, DriveLetter, Size, Type
Get-Disk | Select-Object Number, FriendlyName, Size, PartitionStyle, OperationalStatus
Resize-Partition -DriveLetter C -Size (Get-PartitionSupportedSize -DriveLetter C).SizeMax
If the provider attached a second disk, initialise and format it before anything writes to it. Keep in mind that a disk that shows as RAW is uninitialised, not broken.
Initialize-Disk -Number 1 -PartitionStyle GPT
New-Partition -DiskNumber 1 -UseMaximumSize -DriveLetter D |
Format-Volume -FileSystem NTFS -NewFileSystemLabel "Data" -Confirm:$false
Putting data on D: from day one pays off later: the system volume stays small enough to image, and a reinstall does not require moving terabytes across the network. It also makes disk-level cleanup far simpler, as covered in our guide to reclaiming space on a Windows VPS system drive.
First boot: drivers, network and RDP
On a custom ISO build, network adapters may be missing drivers as well. Check Device Manager through Get-PnpDevice for unknown devices, then install the provider’s virtio network package if required. Until the NIC works, the machine is unreachable over RDP and you are working through the provider console.
Get-NetAdapter | Select-Object Name, InterfaceDescription, Status, LinkSpeed
Get-NetIPConfiguration
Get-PnpDevice -Status Error, Unknown | Select-Object Class, FriendlyName, Status
Enable-NetFirewallRule -DisplayGroup "Remote Desktop"
Set-ItemProperty 'HKLM:\System\CurrentControlSet\Control\Terminal Server' -Name fDenyTSConnections -Value 0
Then confirm the firewall profile. A fresh install on an unknown network classifies the NIC as Public, where the RDP rule group is closed. Set the network category explicitly if the machine is behind the provider’s firewall and RDP is intended.
Get-NetConnectionProfile | Select-Object Name, NetworkCategory, IPv4Connectivity
Set-NetConnectionProfile -InterfaceAlias "Ethernet" -NetworkCategory Private
Hardening before you expose it
The gap between “RDP works” and “the machine is safe to leave running” is where most compromises happen. An internet-facing RDP port with a default name and no lockout policy is scanned within hours. Make these changes before the host is reachable, ideally from the provider console while the public firewall still blocks 3389.
net accounts /lockoutthreshold:5 /lockoutduration:15 /lockoutwindow:15
Set-ItemProperty 'HKLM:\System\CurrentControlSet\Control\Lsa' -Name LimitBlankPasswordUse -Value 1
Set-ItemProperty 'HKLM:\System\CurrentControlSet\Control\Lsa' -Name LmCompatibilityLevel -Value 5
New-NetFirewallRule -DisplayName "RDP restrict" -Direction Inbound -Protocol TCP `
-LocalPort 3389 -RemoteAddress 203.0.113.10 -Action Allow
Enable-WindowsOptionalFeature -Online -FeatureName MicrosoftWindowsPowerShellV2Root
Better still, do not expose 3389 at all. Terminate RDP behind the provider’s firewall with an allowlist, or use a VPN or RD Gateway so the port is never publicly reachable. Whichever you choose, verify from outside afterwards — a rule you believe is blocking but is not is worse than no rule, because it gives false confidence.
Install updates, set the time zone correctly, and enable the backup mechanism you intend to rely on. A host that has been online for a month with no patch baseline and no backup is not finished being installed, no matter how long ago the install wizard completed.
Post-install checklist
The list below is what separates “it boots” from “it is ready”. Work through it in one sitting and record the results, so the next person who touches the host — possibly you in six months — knows what state it was left in.
[ ] Administrator password rotated, named admin account created
[ ] Windows Update run to current, reboot loop completed
[ ] Time zone and clock verified (w32tm /query /status)
[ ] Storage layout confirmed, data volume formatted and in use
[ ] Unknown services and local accounts reviewed
[ ] RDP reachable only from allowed addresses
[ ] Account lockout policy applied
[ ] Activation status confirmed (slmgr /xpr)
[ ] Backup or snapshot schedule configured and tested once
[ ] Hostname and static IP or DHCP reservation final
Run slmgr /dli and slmgr /xpr to confirm activation state rather than assuming it from the install path. Templates are often activated against the provider’s KMS at first boot, which can fail silently if the network is not yet up when the activation task runs. A quick re-arm at the end of setup costs nothing and prevents a surprise in three months.
Provisioning is the cheapest place to get things right. Choosing the correct edition, a sensible disk layout and a locked-down RDP surface takes an extra twenty minutes on install day and saves a rebuild later. Start from the plan that gives you enough disk and a template you can audit, then do the hardening pass above before anything touches the internet.



