Terraform is an open source tool from HashiCorp that can be used to create and manage application infrastructure. Terraform can also provide software for example with Ansible or Chef.


  • Create a modules.tf file and define the OpenStack provider (https://www.terraform.io/docs/providers/openstack/index.html)

    provider "openstack" {
      user_name   = "USERNAME"
      password    = "PASSWORD"
      auth_url    = "https://api-1.betacloud.de:5000/v3"
      region      = "betacloud-1"
    module "sample" {
      source = "modules/sample"
  • Or create a clouds.yaml and secure.yaml file (see Authentication chapter)

    provider "openstack" {
      cloud = "sample"
    module "sample" {
      source = "modules/sample"
  • Create a modules directory

    $ mkdir -p modules/sample


  • Create a variables.tf file below the sample module directory

    variable "network" {
      description = "The default name for the new network"
      default     = "sample"
    variable "pool" {
      description = "The default pool where the floating IPs are taken from"
      default     = "public"
    variable "flavor" {
      description = "The default flavor of the new instance"
      default     = "1C-1GB-10GB"
    variable "image" {
      description = "The default image used to create the new instance"
      default     = "Ubuntu 18.04"


Start the deployment

First we initialize the working directory with the terraform init command after writing a new Terraform configuration. This will ensure that Terraform has all the necessary components to build the template in OpenStack.

If the working directory is initialized, we create an execution plan with terraform plan. In this step, the required resources are compared with the state information stored by Terraform.

After checking the plan, the configuration can be carried out with terraform apply. The instance can easily be deleted using the terraform destroy command.

Adjust the defaults

In the variables.tf file you defined variables, all of which have a default value. You can override these defaults by creating a terraform.tfvars file, setting one or more of the variables to a different value:

# use a larger flavor
flavor = "2C-2GB-20GB"

Terraform will automatically use your new value and create a larger instance. If you remove the terraform.tfvars file again and create another instance, it will again use the default set in variables.tf.

Adding output for the user

To have terraform return some information that the user might need (e.g. floating IPs, hostnames), create outputs in a file called e.g. outputs.tf:

output "instance_names" {
  value = "${openstack_compute_instance_v2.sample.*.name}"
output "floating_ips" {
  value = "${openstack_networking_floatingip_v2.sample.*.address}"

This example will return the names and floating IPs for all of the instances you create.

Regarding the tfstate files

After you have successfully created your resources, you will notice a terraform.tfstate file (and some others of the same kind) in your working directory. Those files are where terraform keeps track of which resources you actually have. This file will be refreshed at each start of a terraform run, but nevertheless should never be deleted.

Example: Instance with additional disks

If you need to have two additional disks in your instances, try something like this:

  • in variables.tf

    variable volume_size" {
      description = "Size of the additional block devices (in GB)"
      size        = 1
  • in main.tf

    # Create the instances
    resource "openstack_compute_instance_v2" "my_instances" {
      count           = "${var.instance_count}"
      name            = "my_instance_0${count.index + 1}"
      flavor_name     = "${var.flavor}"
      image_name      = "${var.image}"
      network {
        uuid = "<enter your network id here>"
    # Create two additional volumes for each instance
    # count is ${var.instance_count * 2 }
    resource "openstack_blockstorage_volume_v2" "my_volumes" {
      count = "${var.instance_count * 2 }"
      size  = "${var.volume_size}"
    # Associate the first volume with the instances
    # use count.index*2 to get the first of each pair of volumes
    resource "openstack_compute_volume_attach_v2" "first_volume_association" {
      count = "${var.instance_count}"
      instance_id = "${openstack_compute_instance_v2.my_instances[count.index].id}"
      volume_id = "${openstack_blockstorage_volume_v2.my_volumes[count.index*2].id}"
    # Associate the second volume with the instances
    # use count.index*2+1 to get the second of each pair of volumes
    resource "openstack_compute_volume_attach_v2" "second_volume_association" {
      count = "${var.instance_count}"
      instance_id = "${openstack_compute_instance_v2.my_instances[count.index].id}"
      volume_id = "${openstack_blockstorage_volume_v2.my_volumes[count.index*2+1].id}"

Example: Shared VIPs for clustered services

In case you want to create shared VIPs for use with e.g. pacemaker (the Neutron chapter explains this from an OpenStack perspective), have a look at the following example:

  • in main.tf

    # reserve an IP
    resource "openstack_networking_port_v2" "reserved_port" {
      name           = "my-reserved-ip"
      network_id     = ...
      admin_state_up = "true"
    resource "openstack_networking_port_v2" "my_instance_port" {
      name               = "my-instance"
      network_id         = ...
      admin_state_up     = true
      security_group_ids = ...
      allowed_address_pairs {
        # allow the virtual IP address created above
        ip_address = openstack_networking_port_v2.reserved_port.all_fixed_ips.0
    resource "openstack_compute_instance_v2" "suse-ses-mon-instance" {
      name            = "my-instance"
      flavor_name     = ...
      image_name      = ...
      network {
        port = openstack_networking_port_v2.my_instance_port.id