Creating and managing Windows Server instances

Windows VMs

Compute Engine provides public images with Windows Server that you can use to create instances. For instructions on how to create a Windows Server instance with SQL Server preinstalled, see Creating SQL Server instances.

For more general information about Windows Server instances and Windows applications that you can run on Compute Engine, see Windows on Compute Engine.

Before you begin

If you want to use the command-line examples in this guide:
1. Install or update to the latest version of the gcloud command-line tool.
2. Set a default region and zone.
If you want to use the API examples in this guide, set up API access.

Creating a Windows Server instance

To create an instance with Windows Server, specify the image family for the specific version of Windows that you need. Compute Engine offers several versions of Windows Server, most of which are available as Shielded VM images. Shielded VM images offer security features like UEFI-compliant firmware, Secure Boot, and vTPM-protected Measured Boot. For a list of the available image families, see public images.

Creating a Windows Server instance that uses an external IP to activate

This section describes how to create a Windows Server instance that has an external IP address. Your VPC network must be configured to allow access to kms.windows.googlecloud.com.

Creating the Windows Server instance

Console

To create a basic Windows instance:

In the Cloud Console, go to the VM instances page.
Go to the VM instances page
Click Create instance.
In the Boot disk section, click Change to begin configuring your boot disk.
On the Public images tab, choose a Windows image.
Click Select.
Click Create to create the instance.

To create a Shielded VM Windows instance:

Go to the VM instances page.
Go to the VM instances page
Click Create instance.
In the Boot disk section, click Change to begin configuring your boot disk.
In the Public images tab, choose Windows Server operating system version.
To see only the Shielded VM images, select Show images with Shielded VM features.
Click Save to save your boot disk configuration.
Optionally, change the instance's Shielded VM settings:
1. Click on the Security tab in the Management Security Disks Networking section.
2. If you want to disable Secure Boot, uncheck Turn on Secure Boot. Secure Boot helps protect your VM instances against boot-level and kernel-level malware and rootkits. For more information, see Secure Boot.
3. If you want to disable the virtual trusted platform module (vTPM), uncheck Turn on vTPM. The vTPM enables Measured Boot, which validates the VM pre-boot and boot integrity. For more information, see Virtual Trusted Platform Module (vTPM).
  
  Important: Disabling the vTPM also disables integrity monitoring, because integrity monitoring relies on data gathered by Measured Boot.
4. If you want to disable integrity monitoring, uncheck Turn on Integrity Monitoring. Integrity monitoring lets you monitor the boot integrity of your Shielded VM instances using Cloud Monitoring. For more information, see Integrity monitoring.
Click Create to create the instance.

gcloud

Use the compute images list command to see a list of available Windows Server images:

gcloud compute images list --project windows-cloud --no-standard-images

To determine whether an image supports Shielded VM features, run the following command, and check for UEFI_COMPATIBLE in the output:

gcloud compute images describe [IMAGE_NAME] --project windows-cloud

where [IMAGE_NAME] is the name of the image to check for support of Shielded VM features.

Use the compute instances create command to create a new instance and specify the image family for one of the Windows Server public images.

gcloud compute instances create [INSTANCE_NAME] \
    --image-project windows-cloud \
    --image-family [IMAGE_FAMILY] \
    --machine-type [MACHINE_TYPE] \
    --boot-disk-size [BOOT_DISK_SIZE] \
    --boot-disk-type [BOOT_DISK_TYPE]

where:

[INSTANCE_NAME] is the name for the new instance.
[IMAGE_FAMILY] is one of the public image families for Windows Server images.
[MACHINE_TYPE] is one of the available machine types.
[BOOT_DISK_SIZE] is the size of the boot disk in GB. Larger persistent disks have higher throughput.
[BOOT_DISK_TYPE] is the type of the boot disk for your instance. For example, pd-ssd.

If you chose an image that supports Shielded VM, you can optionally change the instance's Shielded VM settings using one of the following flags:

--no-shielded-secure-boot: Disable Secure Boot. Secure Boot helps protect your VM instances against boot-level and kernel-level malware and rootkits. For more information, see Secure Boot.
--no-shielded-vtpm: Disable the virtual trusted platform module (vTPM). The vTPM enables Measured Boot, which validates the VM pre-boot and boot integrity. For more information, see Virtual Trusted Platform Module (vTPM).

Important: Disabling the vTPM also disables integrity monitoring, because integrity monitoring relies on data gathered by Measured Boot.
--no-shielded-integrity-monitoring: Disable integrity monitoring. Integrity monitoring lets you monitor the boot integrity of your Shielded VM instances using Cloud Monitoring. For more information, see Integrity monitoring.

The following example creates a Windows 2012 Shielded VM instance with Secure Boot disabled:

gcloud compute instances create my-instance \
--image-family windows-2012-r2 --image-project windows-cloud \
--no-shielded-secure-boot

API

To create an instance with the API, include the initializeParams property in your instance creation request and specify a Windows image. For example, your request body might look like the following:

instance = {
  "name": "[INSTANCE_NAME]",
  "machineType": "zones/[ZONE]/machineTypes/[MACHINE_TYPE]",
  "disks": [{
      "boot": "true",
      "type": "PERSISTENT",
      "initializeParams": {
         "diskName": "[DISK_NAME]",
         "sourceImage": "https://www.googleapis.com/compute/v1/projects/windows-cloud/global/images/family/[IMAGE_FAMILY]",
         "diskSizeGb": "[BOOT_DISK_SIZE]",
         "diskType": "[BOOT_DISK_TYPE]",
       }
    }],
  "networkInterfaces": [{
    "accessConfigs": [{
      "type": "ONE_TO_ONE_NAT",
      "name": "External NAT"
     }],
    "network": "global/networks/default"
  }],
  "serviceAccounts": [{
       "email": DEFAULT_SERVICE_EMAIL,
       "scopes": DEFAULT_SCOPES
  }]
}

where:

[INSTANCE_NAME] is the name for the new instance.
[IMAGE_FAMILY] is one of the public image families for Windows Server or SQL Server images.
[ZONE] is the zone for this instance.
[MACHINE_TYPE] is one of the available machine types.
[BOOT_DISK_SIZE] is the size of the boot disk in GB. Larger persistent disks have higher throughput.
[BOOT_DISK_TYPE] is the type of the boot disk for your instance. For example, pd-ssd.

If you chose an image that supports Shielded VM, you can optionally change the instance's Shielded VM settings by using the following boolean request body items:

enableSecureBoot: Enable or disable Secure Boot. Secure Boot helps protect your VM instances against boot-level and kernel-level malware and rootkits. For more information, see Secure Boot.
enableVtpm: Enable or disable the virtual trusted platform module (vTPM). The vTPM enables Measured Boot, which validates the VM pre-boot and boot integrity. For more information, see Virtual Trusted Platform Module (vTPM).

Important: Disabling the vTPM also disables integrity monitoring, because integrity monitoring relies on data gathered by Measured Boot.
enableIntegrityMonitoring: Enable or disable integrity monitoring. Integrity monitoring lets you monitor and verify the runtime boot integrity of your Shielded VM instances using Cloud Monitoring reports. For more information, see Integrity monitoring.

For more information about creating an instance, read the instances.insert() documentation.

After you create your Windows or SQL Server instance, set the initial password for the instance so that you can connect to the instance through RDP.

Creating a Windows Server instance that uses an internal IP address to activate

Before you can create a Windows Server instance that has only an internal IP address, you must verify or configure routes and firewall rules in your VPC network to allow access to kms.windows.googlecloud.com. Additionally, you must enable Private Google Access for subnets in your VPC network that contain Windows instances with only internal IP addresses.

When you create a new instance by using the gcloud command line, you can use the --no-address flag to ensure that it is not assigned an external IP address:

gcloud compute instances create [INSTANCE_NAME] --network [NETWORK_NAME] \
   --subnet [SUBNET_NAME] \
   --no-address \
   --zone [ZONE] \
   --image-project windows-cloud \
   --image-family [IMAGE_FAMILY] \
   --machine-type [MACHINE_TYPE] \
   --boot-disk-size [BOOT_DISK_SIZE] \
   --boot-disk-type [BOOT_DISK_TYPE]

Replace the following placeholders with valid values:

[INSTANCE_NAME] is the name for the new instance.
[SUBNET_NAME] is the name of the subnet in the VPC network that the instance will use. The subnet must be in the same region as the zone you choose for the instance.
[IMAGE_FAMILY] is one of the public image families for Windows Server images.
[MACHINE_TYPE] is one of the available machine types.
[BOOT_DISK_SIZE] is the size of the boot disk in GB. Larger persistent disks have higher throughput.
[BOOT_DISK_TYPE] is the type of the boot disk for your instance. For example, pd-ssd.

Because this instance does not have an external IP address, you cannot connect to it directly over the Internet. You can connect from another network connected to your VPC network by using Cloud Interconnect or Cloud VPN, or you can first connect to a bastion instance over RDP and then connect to the instance that has only an internal IP address.

Configuring access to kms.windows.googlecloud.com

For Windows activation and renewal, your VPC network must meet the following routing and firewall rule requirements.

Routing requirements

Your Windows instances must be able to reach kms.windows.googlecloud.com (35.190.247.13) through a route whose next hop is the default Internet gateway. You cannot activate Windows instances using an instance based NAT gateway or Cloud NAT because kms.windows.googlecloud.com rejects activation requests from IP addresses that are not confirmed to be Compute Engine instances.

You can use the default route in your VPC network to route traffic directly to kms.windows.googlecloud.com. If you remove this route, or if you plan to do so in the future, create a custom static route with destination 35.190.247.13 and next hop set to default Internet gateway:

gcloud compute routes create [ROUTE_NAME] \
    --destination-range=35.190.247.13/32 \
    --network=[NETWORK] \
    --next-hop-gateway=default-internet-gateway

Replace [ROUTE_NAME] with a name for this route and [NETWORK] with the name of your VPC network.

Either the default route or a custom static route as described above will permit instances with external IP addresses to reach kms.windows.googlecloud.com. If you have Windows instances without external IP addresses, you must also enable Private Google Access so that instances with only internal IP addresses can send traffic to the external IP address for kms.windows.googlecloud.com. That IP address, 35.190.247.13, is included in the list of IP addresses for Google APIs and services.

Firewall rule requirements

The implied allow egress firewall rule allows instances to make requests and receive established responses. Unless you have created custom firewall rules that deny egress, your Windows instances can communicate with kms.windows.googlecloud.com.

If you customize firewall rules, it's a good practice to create a high priority egress allow rule that explicitly permits communication with 35.190.247.13. This way, as you modify your firewall rules, you won't accidentally disable Windows activation.

The following gcloud examples creates the recommended allow egress rule with the highest priority:

gcloud compute firewall-rules create [RULE_NAME] \
    --direction=EGRESS \
    --network=[NETWORK] \
    --action=ALLOW \
    --rules=tcp:1688 \
    --destination-ranges=35.190.247.13/32 \
    --priority=0

Replace [RULE_NAME] with a name for this firewall rule and [NETWORK] with the name of your VPC network.

Verifying that an instance has successfully started

Windows instances experience a longer startup time because of the sysprep process. The Cloud Console might show that the instance is running even if the sysprep process is not yet complete. To check if your instance has successfully started and is ready to be used, check the serial port output with the following command:

gcloud compute instances get-serial-port-output [INSTANCE_NAME]

where [INSTANCE_NAME] is the name of the instance you want to verify.

...[snip]...
Running schtasks with arguments /run /tn GCEStartup
-->  SUCCESS: Attempted to run the scheduled task "GCEStartup".
-------------------------------------------------------------
Instance setup finished. [INSTANCE_NAME] is ready to use.
-------------------------------------------------------------

Disabling Google-provided component updates

If you have Windows instances with image versions v20170509 and later or with agent version 4.1.0 and later, Google-provided components such as the agent, metadata, and sysprep scripts are updated automatically using a scheduled task. The scheduled task is set up by the google-compute-engine-auto-updater package.

If you want to manage updates manually or manage updates using an alternative system you can disable these automatic component updates by removing the google-compute-engine-auto-updater package:

On the Windows Server instance, open a PowerShell terminal as an administrator.

Run the googet remove command to remove the package:

PS C:\\> googet remove google-compute-engine-auto-updater

Optionally, you can reinstall the package to enable automatic component updates:

On the Windows Server instance, open a PowerShell terminal as an administrator.

Run the googet install command to install the package:

PS C:\\> googet install google-compute-engine-auto-updater

Alternatively, you can disable the updates by setting the disable-agent-updates value to true in project or instance custom metadata. The metadata value disables updates without removing the package or the task.

Enabling and disabling Windows instance features

If you have Windows instances with image versions v20170509 and later or with agent version 4.1.0 and later, you can set instance configuration in a config file or in project or instance custom metadata. The config file is in INI format, and is located at the following path:

C:\Program Files\Google\Compute Engine\instance_configs.cfg

The system overrides configuration settings in the following order of priority from the highest priority to the lowest priority:

Configuration parameters that you set in the config file
Configuration parameters set in instance-level custom metadata
Configuration parameters set in project-level custom metadata

For example, if you can enable the accountManager feature in a config file, your instance ignores parameters that you set in custom metadata to disable that feature.

One benefit of setting these parameters in the config file is that those settings persist when you create a custom image for a Windows Server instance. Instance-level custom metadata does not persist beyond the life of the instance.

You can disable different Windows instance features using the following examples:

Disable the account manager, which also disables resetting passwords with the gcloud command-line tool or the Console:

Config file:
```
[accountManager]
disable=true
```
In custom metadata, set disable-account-manager to true in metadata.

Disable the address manager:

Config file entry:
```
[addressManager]
disable=true
```
In custom metadata, set disable-address-manager to true in metadata.

Windows Server Failover Clustering

Enable the Windows Server Failover Clustering agent:

Config file entry:
```
[wsfc]
enable=true
```
In custom metadata, set enable-wsfc to true in metadata.

Using multiple internal load balancers

Specify the IP address of the internal load balancing instance for failover clustering. This is an advanced configuration that you don't need to set for a dedicated failover cluster.

Normally you use an instance of internal load balancing to direct network traffic to one VM instance at a time. If you add a second instance of internal load balancing that uses the failover clustering VM instances as part of a load-balanced website backend, you would have two internal load balancing IP addresses. If failover clustering uses 10.0.0.10 and the website's load balancer uses 10.0.0.11, you must specify the IP address of the load balancer that you use for failover clustering. This disambiguates which address is in use for the cluster.

Config file entry:
```
[wsfc]
addresses=10.0.0.10
```
In custom metadata, set wsfc-addrs to a 10.0.0.10.

Changing the clustering agent port

Set the failover clustering agent port. The default port is 59998. You need to specify a port only when you want to use a different port:

Config file entry:
```
[wsfc]
port=12345
```
In custom metadata, set wsfc-agent-port to the port number.

Image version notes

Older images do not use a config file and only have a subset of features. Image versions between version v20160112 and version v20170509, or Windows agent version between 3.2.1.0 and 4.0.0 require you to use the following custom metadata values:

Set disable-account-manager to true in instance metadata to disable the account manager.
Set disable-address-manager to true in instance metadata to disable the address manager.