cloud-foundation-fabric/modules/compute-vm

Google Compute Engine VM module

This module can operate in two distinct modes:

• instance creation, with optional unmanaged group
• instance template creation

In both modes, an optional service account can be created and assigned to either instances or template. If you need a managed instance group when using the module in template mode, refer to the compute-mig module.

Examples

• Examples
  • Instance using defaults
  • Service account management
    • Compute default service account
    • Custom service account
    • Custom service account, auto created
    • No service account
  • Disk management
    • Disk sources
    • Disk types and options
    • Boot disk as an independent resource
  • Network interfaces
    • Internal and external IPs
    • Using Alias IPs
    • Using gVNIC
    • PSC interfaces
  • Metadata
  • IAM
  • Spot VM
  • Confidential compute
  • Disk encryption with Cloud KMS
  • Instance template
  • Instance group
  • Instance Schedule
  • Snapshot Schedules
  • Resource Manager Tags (non-firewall)
  • Resource Manager Tags (firewall)
  • Sole Tenancy
• Variables
• Outputs

Instance using defaults

The simplest example leverages defaults for the boot disk image and size, and uses a service account created by the module. Multiple instances can be managed via the instance_count variable.

module "simple-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
}
# tftest modules=1 resources=1 inventory=defaults.yaml

Service account management

VM service accounts can be managed in four different ways:

• in its default configuration, the module uses the Compute default service account with a basic set of scopes (devstorage.read_only, logging.write, monitoring.write)
• a custom service account can be used by passing its email in the service_account.email variable
• a custom service account can be created by the module and used by setting the service_account.auto_create variable to true
• the instance can be created with no service account by setting the service_account variable to null

Scopes for custom service accounts are set by default to cloud-platform and userinfo.email, and can be further customized regardless of which service account is used by directly setting the service_account.scopes variable.

Compute default service account

module "vm-managed-sa-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test1"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
}
# tftest modules=1 resources=1 inventory=sa-default.yaml

Custom service account

module "vm-managed-sa-example2" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test2"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  service_account = {
    email = "sa-0@myproj.iam.gserviceaccount.com"
  }
}
# tftest modules=1 resources=1 inventory=sa-custom.yaml

Custom service account, auto created

module "vm-managed-sa-example2" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test2"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=2 inventory=sa-managed.yaml

No service account

module "vm-managed-sa-example2" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test2"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  service_account = null
}
# tftest modules=1 resources=1 inventory=sa-none.yaml

Disk management

Disk sources

Attached disks can be created and optionally initialized from a pre-existing source, or attached to VMs when pre-existing. The source and source_type attributes of the attached_disks variable allows several modes of operation:

• source_type = "image" can be used with zonal disks in instances and templates, set source to the image name or self link
• source_type = "snapshot" can be used with instances only, set source to the snapshot name or self link
• source_type = "attach" can be used for both instances and templates to attach an existing disk, set source to the name (for zonal disks) or self link (for regional disks) of the existing disk to attach; no disk will be created
• source_type = null can be used where an empty disk is needed, source becomes irrelevant and can be left null

This is an example of attaching a pre-existing regional PD to a new instance:

module "vm-disks-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "${var.region}-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  attached_disks = [{
    name        = "repd-1"
    size        = 10
    source_type = "attach"
    source      = "regions/${var.region}/disks/repd-test-1"
    options = {
      replica_zone = "${var.region}-c"
    }
  }]
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=2

And the same example for an instance template (where not using the full self link of the disk triggers recreation of the template)

module "vm-disks-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "${var.region}-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  attached_disks = [{
    name        = "repd"
    size        = 10
    source_type = "attach"
    source      = "https://www.googleapis.com/compute/v1/projects/${var.project_id}/regions/${var.region}/disks/repd-test-1"
    options = {
      replica_zone = "${var.region}-c"
    }
  }]
  service_account = {
    auto_create = true
  }
  create_template = true
}
# tftest modules=1 resources=2

Disk types and options

The attached_disks variable exposes an option attribute that can be used to fine tune the configuration of each disk. The following example shows a VM with multiple disks

module "vm-disk-options-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  attached_disks = [
    {
      name        = "data1"
      size        = "10"
      source_type = "image"
      source      = "image-1"
      options = {
        auto_delete  = false
        replica_zone = "europe-west1-c"
      }
    },
    {
      name        = "data2"
      size        = "20"
      source_type = "snapshot"
      source      = "snapshot-2"
      options = {
        type = "pd-ssd"
        mode = "READ_ONLY"
      }
    }
  ]
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=4 inventory=disk-options.yaml

Boot disk as an independent resource

To create the boot disk as an independent resources instead of as part of the instance creation flow, set boot_disk.use_independent_disk to true and optionally configure boot_disk.initialize_params.

This will create the boot disk as its own resource and attach it to the instance, allowing to recreate the instance from Terraform while preserving the boot.

module "simple-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  boot_disk = {
    initialize_params    = {}
    use_independent_disk = true
  }
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=3 inventory=independent-boot-disk.yaml

Network interfaces

Internal and external IPs

By default VNs are create with an automatically assigned IP addresses, but you can change it through the addresses and nat attributes of the network_interfaces variable:

module "vm-internal-ip" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "vm-internal-ip"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
    addresses  = { internal = "10.0.0.2" }
  }]
}

module "vm-external-ip" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "vm-external-ip"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
    nat        = true
    addresses  = { external = "8.8.8.8" }
  }]
}
# tftest modules=2 resources=2 inventory=ips.yaml

Using Alias IPs

This example shows how to add additional Alias IPs to your VM.

module "vm-with-alias-ips" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
    alias_ips = {
      alias1 = "10.16.0.10/32"
    }
  }]
}
# tftest modules=1 resources=1 inventory=alias-ips.yaml

Using gVNIC

This example shows how to enable gVNIC on your VM by customizing a cos image. Given that gVNIC needs to be enabled as an instance configuration and as a guest os configuration, you'll need to supply a bootable disk with guest_os_features=GVNIC. SEV_CAPABLE, UEFI_COMPATIBLE and VIRTIO_SCSI_MULTIQUEUE are enabled implicitly in the cos, rhel, centos and other images.

resource "google_compute_image" "cos-gvnic" {
  project      = "my-project"
  name         = "my-image"
  source_image = "https://www.googleapis.com/compute/v1/projects/cos-cloud/global/images/cos-89-16108-534-18"

  guest_os_features {
    type = "GVNIC"
  }
  guest_os_features {
    type = "SEV_CAPABLE"
  }
  guest_os_features {
    type = "UEFI_COMPATIBLE"
  }
  guest_os_features {
    type = "VIRTIO_SCSI_MULTIQUEUE"
  }
}

module "vm-with-gvnic" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "test"
  boot_disk = {
    initialize_params = {
      image = google_compute_image.cos-gvnic.self_link
      type  = "pd-ssd"
    }
  }
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
    nic_type   = "GVNIC"
  }]
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=3 inventory=gvnic.yaml

PSC interfaces

Private Service Connect interfaces can be configured via the network_attached_interfaces variable, which is a simple list of network attachment ids, one per interface. PSC interfaces will be defined after regular interfaces.

# create the network attachment from a service project
module "net-attachment" {
  source     = "./fabric/modules/net-address"
  project_id = "prj-svc"
  network_attachments = {
    svc-0 = {
      subnet_self_link      = "projects/prj-host/regions/europe-west8/subnetworks/gce"
      producer_accept_lists = ["my-vm-project"]
    }
  }
}

module "vm-psc-interface" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-vm-project"
  zone       = "europe-west8-b"
  name       = "vm-internal-ip"
  network_interfaces = [{
    network    = "internal"
    subnetwork = "internal"
  }]
  network_attached_interfaces = [
    module.net-attachment.network_attachment_ids["svc-0"]
  ]
}
# tftest modules=2 resources=2

Metadata

You can define labels and custom metadata values. Metadata can be leveraged, for example, to define a custom startup script.

module "vm-metadata-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "nginx-server"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  labels = {
    env    = "dev"
    system = "crm"
  }
  metadata = {
    startup-script = <<-EOF
      #! /bin/bash
      apt-get update
      apt-get install -y nginx
    EOF
  }
  service_account = {
    auto_create = true
  }
}
# tftest modules=1 resources=2 inventory=metadata.yaml

IAM

Like most modules, you can assign IAM roles to the instance using the iam variable.

module "vm-iam-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "webserver"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  iam = {
    "roles/compute.instanceAdmin" = [
      "group:webserver@example.com",
      "group:admin@example.com"
    ]
  }
}
# tftest modules=1 resources=2 inventory=iam.yaml

Spot VM

Spot VMs are ephemeral compute instances suitable for batch jobs and fault-tolerant workloads. Spot VMs provide new features that preemptible instances do not support, such as the absence of a maximum runtime.

module "spot-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  options = {
    spot               = true
    termination_action = "STOP"
  }
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
}
# tftest modules=1 resources=1 inventory=spot.yaml

Confidential compute

You can enable confidential compute with the confidential_compute variable, which can be used for standalone instances or for instance templates.

module "vm-confidential-example" {
  source               = "./fabric/modules/compute-vm"
  project_id           = var.project_id
  zone                 = "europe-west1-b"
  name                 = "confidential-vm"
  confidential_compute = true
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]

}

module "template-confidential-example" {
  source               = "./fabric/modules/compute-vm"
  project_id           = var.project_id
  zone                 = "europe-west1-b"
  name                 = "confidential-template"
  confidential_compute = true
  create_template      = true
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
}

# tftest modules=2 resources=2 inventory=confidential.yaml

Disk encryption with Cloud KMS

This example shows how to control disk encryption via the the encryption variable, in this case the self link to a KMS CryptoKey that will be used to encrypt boot and attached disk. Managing the key with the ../kms module is of course possible, but is not shown here.

module "kms-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "kms-test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  attached_disks = [{
    name = "attached-disk"
    size = 10
  }]
  service_account = {
    auto_create = true
  }
  encryption = {
    encrypt_boot      = true
    kms_key_self_link = var.kms_key.id
  }
}
# tftest modules=1 resources=3 inventory=cmek.yaml

Instance template

This example shows how to use the module to manage an instance template that defines an additional attached disk for each instance, and overrides defaults for the boot disk image and service account.

module "cos-test" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  boot_disk = {
    initialize_params = {
      image = "projects/cos-cloud/global/images/family/cos-stable"
    }
  }
  attached_disks = [
    {
      name = "disk-1"
      size = 10
    }
  ]
  service_account = {
    email = "vm-default@my-project.iam.gserviceaccount.com"
  }
  create_template = true
}
# tftest modules=1 resources=1 inventory=template.yaml

Instance group

If an instance group is needed when operating in instance mode, simply set the group variable to a non null map. The map can contain named port declarations, or be empty if named ports are not needed.

locals {
  cloud_config = "my cloud config"
}

module "instance-group" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "ilb-test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  boot_disk = {
    image = "projects/cos-cloud/global/images/family/cos-stable"
  }
  service_account = {
    email  = var.service_account.email
    scopes = ["https://www.googleapis.com/auth/cloud-platform"]
  }
  metadata = {
    user-data = local.cloud_config
  }
  group = { named_ports = {} }
}
# tftest modules=1 resources=2 inventory=group.yaml

Instance Schedule

Instance start and stop schedules can be defined via an existing or auto-created resource policy.

To use an existing policy pass its id to the instance_schedule variable:

module "instance" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "schedule-test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  boot_disk = {
    image = "projects/cos-cloud/global/images/family/cos-stable"
  }
  instance_schedule = {
    resource_policy_id = "projects/my-project/regions/europe-west1/resourcePolicies/test"
  }
}
# tftest modules=1 resources=1 inventory=instance-schedule-id.yaml

To create a new policy set its configuration in the instance_schedule variable. When removing the policy follow a two-step process by first setting active = false in the schedule configuration, which will unattach the policy, then removing the variable so the policy is destroyed.

module "instance" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "schedule-test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  boot_disk = {
    image = "projects/cos-cloud/global/images/family/cos-stable"
  }
  instance_schedule = {
    create_config = {
      vm_start = "0 8 * * *"
      vm_stop  = "0 17 * * *"
    }
  }
}
# tftest modules=1 resources=2 inventory=instance-schedule-create.yaml

Snapshot Schedules

Snapshot policies can be attached to disks with optional creation managed by the module.

module "instance" {
  source     = "./fabric/modules/compute-vm"
  project_id = "my-project"
  zone       = "europe-west1-b"
  name       = "schedule-test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  boot_disk = {
    image             = "projects/cos-cloud/global/images/family/cos-stable"
    snapshot_schedule = "boot"
  }
  attached_disks = [
    {
      name              = "disk-1"
      size              = 10
      snapshot_schedule = "generic-vm"
    }
  ]
  snapshot_schedules = {
    boot = {
      schedule = {
        daily = {
          days_in_cycle = 1
          start_time    = "03:00"
        }
      }
    }
  }
}
# tftest modules=1 resources=5 inventory=snapshot-schedule-create.yaml

Resource Manager Tags (non-firewall)

Resource manager tags bindings for use in IAM or org policy conditions are supported via the tag_bindings variable with the following limitations:

• tag bindings are not created for attached disks
• tag bindings will not be created for the boot disk if the use_independent_disk flag is true
• tag bindings are ignored for instance templates

The current provider implementation is sub-optimal and forces

• recreation of the instance on tag changes
• specifying both the key and value where only the value is actually needed

This is an example of setting tag bindings:

module "simple-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  tag_bindings = {
    "tagKeys/1234567890" = "tagValues/7890123456"
  }
}
# tftest modules=1 resources=1 inventory=tag-bindings.yaml

Resource Manager Tags (firewall)

Network-scoped resource manager tags (or "secure tags") bindings for use in firewall rules are supported with similar limitations as in the section above, via a separate tag_bindings_firewall variable that only applies bindings to the instance and not the boot disk.

This is an example of setting both types of tag bindings:

module "simple-vm-example" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  tag_bindings = {
    "tagKeys/1234567890" = "tagValues/7890123456"
  }
  # tags here need to be scoped to a VPC
  tag_bindings_firewall = {
    "tagKeys/5678901234" = "tagValues/3456789012"
  }
}
# tftest modules=1 resources=1 inventory=tag-bindings.yaml

Sole Tenancy

You can add node affinities (and anti-affinity) configurations to allocate the VM on sole tenant nodes.

module "sole-tenancy" {
  source     = "./fabric/modules/compute-vm"
  project_id = var.project_id
  zone       = "europe-west1-b"
  name       = "test"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }]
  options = {
    node_affinities = {
      workload = {
        values = ["frontend"]
      }
      cpu = {
        in     = false
        values = ["c3"]
      }
    }
  }
}
# tftest modules=1 resources=1 inventory=sole-tenancy.yaml

Variables

name	description	type	required	default
name	Instance name.	string	✓
network_interfaces	Network interfaces configuration. Use self links for Shared VPC, set addresses to null if not needed.	list(object({…}))	✓
project_id	Project id.	string	✓
zone	Compute zone.	string	✓
attached_disk_defaults	Defaults for attached disks options.	object({…})		{…}
attached_disks	Additional disks, if options is null defaults will be used in its place. Source type is one of 'image' (zonal disks in vms and template), 'snapshot' (vm), 'existing', and null.	list(object({…}))		[]
boot_disk	Boot disk properties.	object({…})		{…}
can_ip_forward	Enable IP forwarding.	bool		false
confidential_compute	Enable Confidential Compute for these instances.	bool		false
create_template	Create instance template instead of instances.	bool		false
description	Description of a Compute Instance.	string		"Managed by the compute-vm Terraform module."
enable_display	Enable virtual display on the instances.	bool		false
encryption	Encryption options. Only one of kms_key_self_link and disk_encryption_key_raw may be set. If needed, you can specify to encrypt or not the boot disk.	object({…})		null
group	Define this variable to create an instance group for instances. Disabled for template use.	object({…})		null
hostname	Instance FQDN name.	string		null
iam	IAM bindings in {ROLE => [MEMBERS]} format.	map(list(string))		{}
instance_schedule	Assign or create and assign an instance schedule policy. Either resource policy id or create_config must be specified if not null. Set active to null to dtach a policy from vm before destroying.	object({…})		null
instance_type	Instance type.	string		"f1-micro"
labels	Instance labels.	map(string)		{}
metadata	Instance metadata.	map(string)		{}
min_cpu_platform	Minimum CPU platform.	string		null
network_attached_interfaces	Network interfaces using network attachments.	list(string)		[]
options	Instance options.	object({…})		{…}
scratch_disks	Scratch disks configuration.	object({…})		{…}
service_account	Service account email and scopes. If email is null, the default Compute service account will be used unless auto_create is true, in which case a service account will be created. Set the variable to null to avoid attaching a service account.	object({…})		{}
shielded_config	Shielded VM configuration of the instances.	object({…})		null
snapshot_schedules	Snapshot schedule resource policies that can be attached to disks.	map(object({…}))		{}
tag_bindings	Resource manager tag bindings for this instance, in tag key => tag value format.	map(string)		null
tag_bindings_firewall	Firewall (network scoped) tag bindings for this instance, in tag key => tag value format.	map(string)		null
tags	Instance network tags for firewall rule targets.	list(string)		[]

Outputs

name	description	sensitive
external_ip	Instance main interface external IP addresses.
group	Instance group resource.
id	Fully qualified instance id.
instance	Instance resource.	✓
internal_ip	Instance main interface internal IP address.
internal_ips	Instance interfaces internal IP addresses.
self_link	Instance self links.
service_account	Service account resource.
service_account_email	Service account email.
service_account_iam_email	Service account email.
template	Template resource.
template_name	Template name.