Configurators

Configurators is a software plugin that implements an operation and applies changes to instances. There are built-in configurators for shell scripts, Ansible playbooks, Terraform configurations, and Kubernetes resources or you can include your own as part of your blueprint.

To use a configurator, set it in the implementation field of an Operation and set its inputs as documented below. Configurator names are case-sensitive; if a configurator name isn’t found it is treated as an external command.

If you set an external command line directly as the implementation, Unfurl will choose the appropriate one to use. If operation_host is local it will use the Shell configurator, if it is remote, it will use the Ansible configurator and generate a playbook that invokes it on the remote machine:

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    topology_template:
      node_templates:
        test_remote:
          type: tosca:Root
          interfaces:
            Standard:
              configure: echo "abbreviated configuration"

PYTHON

@operation(name="configure")
def test_remote_configure(self, **kw):
    return unfurl.configurators.shell.ShellConfigurator(
        command='echo "abbreviated configuration"',
    )

test_remote = tosca.nodes.Root()
test_remote.set_operation(test_remote_configure)

Configurators are fairly low-level. You can use “Installer” nodes that. Artifacts

Available configurators include:

Ansible

The Ansible configurator executes the given playbook. You can access the same Unfurl filters and queries available in the Ensemble manifest from inside a playbook.

Example

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    topology_template:
      node_templates:
        test_remote:
          type: tosca:Root
          interfaces:
            Standard:
              configure:
                implementation: Ansible
                inputs:
                  playbook:
                    # quote this yaml so its templates are not evaluated before we pass it to Ansible
                    q:
                      - set_fact:
                          fact1: "{{ '.name' | eval }}"
                      - name: Hello
                        command: echo "{{ fact1 }}"
                outputs:
                  fact1:

PYTHON

@operation(name="configure", outputs={"fact1": None})
def test_remote_configure(self, **kw):
    return unfurl.configurators.ansible.AnsibleConfigurator(
        playbook="""
        - set_fact:
            fact1: "{{ '.name' | eval }}"
        - name: Hello
          command: echo "{{ fact1 }}"
        """
    )

test_remote = tosca.nodes.Root()
test_remote.set_operation(test_remote_configure)

Inputs

playbook

(required) If string, treat as a file path to the Ansible playbook to run, otherwise treat as an inline playbook

inventory

If string, treat as a file path to an Ansible inventory file or directory, otherwise treat as in inline YAML inventory. If omitted, the inventory will be generated (see below)

arguments

A dictionary of variables that will be passed to the playbook as Ansible facts. See Arguments

playbookArgs

A list of strings that will be passed to ansible-playbook as command-line arguments

resultTemplate

Same behavior as defined for Shell but will also include outputs as a variable.

Outputs

Keys declared as outputs are used as the names of the Ansible facts to be extracted after the playbook completes.

implementation key notes

operation_host

If set, names the Ansible host.

environment

If set, environment directives will processed and passed to the playbook’s environment

Playbook processing

The playbook input can be set to a full playbook or a list of tasks. If inventory is auto-generated and the “hosts” keyword is empty or missing from the playbook, “hosts” will be set to the host found in the auto-generated inventory, as described below.

Inventory

If an inventory file isn’t specified in inputs, Unfurl will generate an Ansible inventory for the target host. The target host will be selected by searching for a node in the following order:

• The operation_host if explicitly set.

• The current target if it looks like a host (i.e. has an Ansible or SSH endpoint or is a Compute resource)

• Search the current target’s hostedOn relationship for a node that looks like a host.

• Fallback to “localhost” with a local ansible connection.

The inventory facts for the selected host is built from the following sources:

• If host has an endpoint of type unfurl.capabilities.Endpoint.SSH or unfurl.capabilities.Endpoint.Ansible use that capability’s host, port, connection, user, and hostvars properties.

• If there is a relationship template or connection of type unfurl.relationships.ConnectsTo.Ansible that targets the endpoint, uses its credential and hostvars properties. (These can be set in the environment’s Connections section.)

• If the host is declared as a member of group of type unfurl.groups.AnsibleInventoryGroup in the service template, the group’s name will be added as an ansible group along with the contents of the group’s hostvars property.

• If ansible_host wasn’t previously set, ansible_host will be set to the host’s public_ip or private_ip in that order if present, otherwise set it to localhost.

• If the host is a Google compute instance the host name will be set to INSTANCE_NAME.ZONE.PROJECT e.g. instance-1.us-central1-a.purple-sanctum-25912. This is for compatibility with the gcloud compute config-ssh command to enable Unfurl to use those credentials.

Execution environment

Unfurl runs Ansible in an environment isolated from your machine’s Ansible installation and will not load the ansible configuration files in the standard locations. If you want to load an Ansible configuration file set the ANSIBLE_CONFIG environment variable. If you want Ansible to search standard locations set to an empty or invalid value like ANSIBLE_CONFIG=. (See also the Ansible Configurations Documentation)

Note: Because Ansible is initialized at the beginning of execution, if the --no-runtime command option is used or if no runtime is available ANSIBLE_CONFIG will only be applied in the environment that executes Unfurl. It will not be applied if set via environment declaration.

Cmd

The Cmd configurator executes a shell command using either the Shell configurator or the Ansible configurator for remote execution if the operation_host is set to a remote node. As described above, Cmd is the default configurator if none is specified.

Example

In this example, operation_host is set to a remote instance so the command is executed remotely using Ansible.

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    topology_template:
      node_templates:
        test_remote:
          type: tosca:Root
          interfaces:
            Standard:
              configure:
                implementation:
                  primary: Cmd
                  operation_host: staging.example.com
                inputs:
                  cmd: echo "test"

PYTHON

import unfurl.configurators
import tosca
@tosca.operation(name="configure", operation_host="staging.example.com")
def test_remote_configure(self, **kw):
    return unfurl.configurators.CmdConfigurator(
        cmd='echo "test"',
    )


test_remote = tosca.nodes.Root()
test_remote.set_operation(test_remote_configure)

Delegate

The delegate configurator will delegate the current operation to the specified one.

Inputs

operation

(required) The operation to delegate to, e.g. Standard.configure

target

The name of the instance to delegate to. If omitted the current target will be used.

inputs

Inputs to pass to the operation. If omitted the current inputs will be used.

when

If set, only perform the delegated operation if its value evaluates to true.

Shell

The Shell configurator executes a shell command.

Inline shell script example

This example executes an inline shell script and uses the cwd and shell input options.

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    topology_template:
      node_templates:
        shellscript-example:
          type: tosca:Root
          interfaces:
            Standard:
              configure:
                implementation: |
                  if ! [ -x "$(command -v testvars)" ]; then
                    source testvars.sh
                  fi
                inputs:
                    cwd: '{{ "project" | get_dir }}'
                    keeplines: true
                    # our script requires bash
                    shell: '{{ "bash" | which }}'

PYTHON

@operation(name="configure")
def shellscript_example_configure(self, **kw):
    return unfurl.configurators.shell.ShellConfigurator(
        command='if ! [ -x "$(command -v testvars)" ]; then\n  source testvars.sh\nfi\n',
        cwd=Eval('{{ "project" | get_dir }}'),
        keeplines=True,
        shell=Eval('{{ "bash" | which }}'),
    )

shellscript_example = tosca.nodes.Root()
shellscript_example.set_operation(shellscript_example_configure)

Example with artifact

Declaring an artifact of a type that is associated with the shell configurator ensures Unfurl will install the artifact if necessary, before it runs the command.

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    imports:
    - repository: unfurl
      file: tosca_plugins/artifacts.yaml
    node_types:
      artifact_example:
          derived_from: tosca:Root
          artifacts:
            ripgrep:
              type: artifact.AsdfTool
              file: ripgrep
              properties:
                version: 13.0.0
          interfaces:
            Standard:
              configure:
                implementation: 
                  primary: ripgrep
                inputs:
                  cmd: rg search

PYTHON

from unfurl.tosca_plugins.artifacts import artifact_AsdfTool
import tosca

class artifact_example(tosca.nodes.Root):
    ripgrep: artifact_AsdfTool = artifact_AsdfTool(
        "ripgrep",
        version="13.0.0",
        file="ripgrep",
    )

    def configure(self, **kw):
        return self.ripgrep.execute(
            cmd="rg search",
        )

Inputs

command

(required) The command to execute It can be either a string or a list of command arguments.

arguments

A map of arguments to pass to the command.

cwd

Set the current working directory to execute the command in.

dryrun

During a during a dryrun job this will be either appended to the command line or replace the string %dryrun% if it appears in the command. (%dryrun% is stripped out when running regular jobs.) If not set, the task will not be executed at all during a dry run job.

shell

If a string, the executable of the shell to execute the command in (e.g. /usr/bin/bash). A boolean indicates whether the command if invoked through the default shell or not. If omitted, it will be set to true if command is a string or false if it is a list.

echo

A boolean that indicates whether or not should be standard output (and stderr) should be echoed to Unfurl’s stdout while the command is being run. If omitted, true unless running with --quiet. (Doesn’t affect the capture of stdout and stderr.)

input

Optional string to pass as stdin.

keeplines

(Default: false) If true, preserve line breaks in the given command.

done

As as done defined by the template configurator.

outputsTemplate

A Jinja2 template or runtime expression that is processed after shell command completes, with same variables as resultTemplate. The template should evaluate to a map to be used as the operation’s outputs or null to skip.

resultTemplate

A Jinja2 template or runtime expression that is processed after shell command completes, it will have the following template variables:

Result template variables

All values will be either string or null unless otherwise noted.

success

true unless an error occurred or the returncode wasn’t 0

cmd

(string) The command line that was executed

stdout

stderr

returncode

Integer (Null if the process didn’t complete)

error

Set if an exception was raised

timeout

(Null unless a timeout occurred)

The processing of the resultTemplate is equivalent to passing its resulting YAML to update_instances.

Outputs

No outputs are set unless outputsTemplate is present.

Template

The template configurator lets you implement an operation entirely within the template.

Inputs

run

Sets the result of this task.

dryrun

During a --dryrun job used instead of run.

done

If set, a map whose values are passed as arguments to unfurl.configurator.TaskView.done(). Embedded runtime expressions can access the previous value of those arguments as variables.

resultTemplate

A Jinja2 template or runtime expression that is processed with results of run as its variables.

Outputs

Operation outputs are set from the outputs key on the done input if present.

Terraform

The Terraform configurator will be invoked on any Node Template with the type unfurl.nodes.Installer.Terraform. It can also be used to implement any operation regardless of the node type by setting the implementation to Terraform. It will invoke the appropriate terraform command (e.g “apply” or “destroy”) based on the job’s workflow.

Unless you set the stateLocation input parameter to “remote”, the Terraform configurator manages the Terraform state file itself and commits it to the ensemble’s repository so you don’t use Terraform’s remote state – it will be self-contained and sharable like the rest of the Ensemble. Any sensitive state will be encrypted using Ansible Vault.

During a --dryrun job the configurator will validate and generate the Terraform plan but not execute it. You can override this behavior with the dryrun_mode input parameter and you can specify dummy outputs to use with the dryrun_outputs input parameter.

You can use the unfurl.nodes.Installer.Terraform node type with your node template to the avoid boilerplate and set the needed inputs.

Example

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    imports:
    - repository: unfurl
      file: tosca_plugins/artifacts.yaml
    topology_template:
      node_templates:

        terraform-example:
          type: unfurl.nodes.Installer.Terraform
          interfaces:
            defaults:
              inputs:
                tfvars:
                  tag: test
                main: |
                  variable "tag" {
                    type        = string
                  }

                  output "name" {
                    value = var.tag
                  }
            Standard:
              operations:
                configure:

PYTHON

import tosca
import unfurl
from unfurl.tosca_plugins.artifacts import unfurl_nodes_Installer_Terraform
import unfurl.configurators.terraform


@tosca.operation(
    name="default", apply_to=["Install.check", "Standard.configure", "Standard.delete"]
)
def terraform_example_default(self, **kw):
    return unfurl.configurators.terraform.TerraformConfigurator(
        tfvars={"tag": "test"},
        main="""
      variable "tag" {
        type  = string
      }
      output "name" {
        value = var.tag
      }""",
    )

terraform_example = unfurl_nodes_Installer_Terraform()
terraform_example.set_operation(terraform_example_default)

Inputs

main

The contents of the root Terraform module or a path to a directory containing the Terraform configuration. If it is a directory path, the configurator will treat it as a local Terraform module. Otherwise, if main is a string it will be treated as HCL and if it is a map, it will be written out as JSON. (See the note below about HCL in YAML.) If omitted, the configurator will look in get_dir("spec.home") for the Terraform configuration.

tfvars

A map of Terraform variables to passed to the main Terraform module or a string equivalent to “.tfvars” file.

stateLocation

If set to “secrets” (the default) the Terraform state file will be encrypted and saved into the instance’s “secrets” folder. If set to “artifacts”, it will be saved in the instance’s “artifacts” folder with only sensitive values encrypted inline. If set to “remote”, Unfurl will not manage the Terraform state at all.

command

Path to the terraform executable. Default: “terraform”

dryrun_mode

How to run during a dry run job. If set to “plan” just generate the Terraform plan. If set to “real”, run the task without any dry run logic. Default: “plan”

dryrun_outputs

During a dry run job, this map of outputs will be used simulate the task’s outputs (otherwise outputs will be empty).

resultTemplate

A Jinja2 template or runtime expression that is processed with the Terraform state JSON file as its variables as well as the Result template variables variables documented above for Shell. See the Terraform providers’ schema documentation for details but top-level keys will include “resources” and “outputs”.

Outputs

Specifies which outputs defined by the Terraform module that will be set as the operation’s outputs. If omitted and the Terraform configuration is specified inline, all of the Terraform outputs will be included. But if a Terraform configuration directory was specified instead, its outputs need to be declared here to be exposed.

tfvar and tfoutput Metadata

You can automatically map properties and attributes to a Terraform variables and outputs by setting tfvar and tfoutput keys in the property and attribute metadata, respectively. For example:

YAML

node_types:
  ExampleTerraformManagedResource:
    derived_from: tosca.nodes.Root
    description: A type of resource that is managed (create/update/delete) by a terraform
      resource
    properties:
      example_terraform_var:
        type: string
        description: A property the UI will render for user input
        metadata:
          # TOSCA properties are conceptually similar to Terraform variables.
          # This declares that this property will set as a terraform variable with the same name:
          tfvar: true
    attributes:
      example_terraform_output:
        type: string
        metadata:
          # TOSCA attributes are conceptually similar to Terraform outputs.
          # This declares that this attribute will be set to the terraform output with the same name:
          tfoutput: true
    interfaces:      
      Install:
        operations:
          check:
      Standard:
        operations:
          configure:
          delete:
      defaults:
        # set this operation for the basic CRUD operations in the TOSCA deploy workflow
        implementation:
          className: unfurl.configurators.terraform.TerraformConfigurator
        inputs:
          main: terraform

PYTHON

from unfurl.configurators.terraform import (
    TerraformConfigurator,
    TerraformInputs,
)
from unfurl.tosca_plugins.expr import tfvar, tfoutput
import tosca

class ExampleTerraformManagedResource(tosca.nodes.Root):
    """A type of resource that is managed (create/update/delete) by a terraform resource"""

    # TOSCA properties are conceptually similar to Terraform variables.
    # Set the tfvar option to indicate the property should set a Terraform variable with the same name.
    example_terraform_var: str = tosca.Property(options=tfvar)
    """A property the UI will render for user input"""

    # TOSCA attributes are conceptually similar to Terraform outputs.
    # Set the tfoutput option to indicate the attribute should be set to the terraform output with the same name.
    example_terraform_output: str = tosca.Attribute(options=tfoutput)

    # call this function for the basic CRUD operations in the TOSCA deploy workflow
    @tosca.operation(apply_to=["Install.check", "Standard.configure", "Standard.delete"])
    def default(self, **kw) -> TerraformConfigurator:
        # Implement these operations using the Terraform module found in the directory named "terraform"
        return TerraformConfigurator(TerraformInputs(main="terraform"))

Environment Variables

If the TF_DATA_DIR environment variable is not defined it will be set to .terraform relative to the current working directory.

Note on HCL in YAML

The json representation of the Terraform’s HashiCorp Configuration Language (HCL) is quite readable when serialized as YAML:

Example 1: variable declaration

variable "example" {
  default = "hello"
}

Becomes:

variable:
  example:
    default: hello

Example 2: Resource declaration

resource "aws_instance" "example" {
  instance_type = "t2.micro"
  ami           = "ami-abc123"
}

becomes:

resource:
  aws_instance:
   example:
    instance_type: t2.micro
    ami:           ami-abc123

Example 3: Resource with multiple provisioners

resource "aws_instance" "example" {
  provisioner "local-exec" {
    command = "echo 'Hello World' >example.txt"
  }
  provisioner "file" {
    source      = "example.txt"
    destination = "/tmp/example.txt"
  }
  provisioner "remote-exec" {
    inline = [
      "sudo install-something -f /tmp/example.txt",
    ]
  }
}

Multiple provisioners become a list:

resource:
  aws_instance:
    example:
      provisioner:
        - local-exec
            command: "echo 'Hello World' >example.txt"
        - file:
            source: example.txt
            destination: /tmp/example.txt
        - remote-exec:
            inline: ["sudo install-something -f /tmp/example.txt"]

You can convert HCL to JSON and YAML using tools like hcl2json and yq, for example:

hcl2json main.tf | yq -P -oyaml

Expressing terraform modules as YAML or JSON instead of HCL exposes the terraform in a structured way, making it easier to provide extensibility. For example, a derived node template or artifact could add or update terraform resources defined on the base type: In the example below, a derived type customizes its base type’s main property without have to replace its entire definition.

main: "{{ '.super::main' | eval | combine(SELF.custom_changes, recursive=True, list_merge='append_rp') }}"

Here we merge the base class’s property using .super with a property called custom_changes using Ansible Jinja2’s combine filter which lets you recursively merge maps and lists.

Installers

Installer node types already have operations defined. You just need to import the service template containing the TOSCA type definitions and declare node templates with the needed properties and operation inputs.

Docker

Required TOSCA import: configurators/templates/docker.yaml (in the unfurl repository)

unfurl.nodes.Container.Application.Docker

TOSCA node type that represents a Docker container.

artifacts

image

(required) An artifact of type tosca.artifacts.Deployment.Image.Container.Docker

By default, the configurator will assume the image is in https://registry.hub.docker.com. If the image is in a different registry you can declare it as a repository and have the image artifact reference that repository.

Inputs

configuration

A map that will included as parameters to Ansible’s Docker container module They are enumerated here

Example

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    imports:
    - repository: unfurl
      file: configurators/templates/docker.yaml
    topology_template:
      node_templates:
        hello-world-container:
          type: unfurl.nodes.Container.Application.Docker
          artifacts:
            image:
              type: tosca.artifacts.Deployment.Image.Container.Docker
              file: busybox
          interfaces:
            Standard:
              inputs:
                configuration:
                  command: ["echo", "hello world"]
                  detach: no
                  output_logs: yes

PYTHON

from unfurl.configurators.templates.docker import (
    unfurl_nodes_Container_Application_Docker,
)
import tosca

hello_world_container = unfurl_nodes_Container_Application_Docker(
    "hello-world-container",
    image=tosca.artifacts.DeploymentImageContainerDocker(
        "image",
        file="busybox",
    ),
)
hello_world_container._Standard_default_inputs = dict(
    configuration=dict(command=["echo", "hello world"], detach=False, output_logs=True)
)

DNS

The DNS installer support nearly all major DNS providers using OctoDNS.

Required TOSCA import: configurators/templates/dns.yaml (in the unfurl repository)

unfurl.nodes.DNSZone

TOSCA node type that represents a DNS zone.

Properties

name

(required) DNS hostname of the zone (should end with “.”).

provider

(required) A map containing the OctoDNS provider configuration

records

A map of DNS records to add to the zone (default: an empty map)

exclusive

Set to true if the zone is exclusively managed by this instance (removes unrecognized records) (default: false)

Attributes

zone

A map containing the records found in the live zone

managed_records

A map containing the current records that are managed by this instance

unfurl.relationships.DNSRecords

TOSCA relationship type to connect a DNS record to a DNS zone. The DNS records specified here will be added, updated or removed from the zone when the relationship is established, changed or removed.

Properties

records

(required) A map containing the DNS records to add to the zone.

Example

YAML

apiVersion: unfurl/v1.0.0
kind: Ensemble
spec:
  service_template:
    imports:
    - repository: unfurl
      file: configurators/templates/dns.yaml
    topology_template:
      node_templates:
        example_com_zone:
          type: unfurl.nodes.DNSZone
          properties:
            name: example.com.
            provider:
              # Amazon Route53 (Note: this provider requires that the zone already exists.)
              class: octodns.provider.route53.Route53Provider

        test_app:
          type: tosca.nodes.WebServer
          requirements:
            - host: compute
            - dns:
                node: example_com_zone
                relationship:
                  type: unfurl.relationships.DNSRecords
                  properties:
                    records:
                      www:
                        type: A
                        value:
                          # get the ip address of the Compute instance that this is hosted on
                          eval: .source::.requirements::[.name=host]::.target::public_address

PYTHON

import unfurl
import tosca
from tosca import Eval

from unfurl.configurators.templates.dns import unfurl_nodes_DNSZone, unfurl_relationships_DNSRecords

example_com_zone = unfurl_nodes_DNSZone(
    name="example.com.",
    provider={"class": "octodns.provider.route53.Route53Provider"},
)

test_app = tosca.nodes.WebServer(
    host=[tosca.find_node("compute")],
)
test_app.dns = unfurl_relationships_DNSRecords(
    records=Eval(
        {
            "www": {
                "type": "A",
                "value": {
                    "eval": ".source::.requirements::[.name=host]::.target::public_address"
                },
            }
        }
    ),
)[example_com_zone]

Helm

Requires Helm 3, which will be installed automatically if missing.

Required TOSCA import: configurators/templates/helm.yaml (in the unfurl repository)

unfurl.nodes.HelmRelease

TOSCA type that represents a Helm release. Deploying or discovering a Helm release will add to the ensemble any Kubernetes resources managed by that release.

Requirements

host

A node template of type unfurl.nodes.K8sNamespace

repository

A node template of type unfurl.nodes.HelmRepository

Properties

release_name

(required) The name of the helm release

chart

The name of the chart (default: the instance name)

chart_values

A map of chart values

Inputs

All operations can be passed the following input parameters:

flags

A list of flags to pass to the helm command

unfurl.nodes.HelmRepository

TOSCA node type that represents a Helm repository.

Properties

name

The name of the repository (default: the instance name)

url

(required) The URL of the repository

Kubernetes

Use these types to manage Kubernetes resources.

unfurl.nodes.K8sCluster

TOSCA type that represents a Kubernetes cluster. Its attributes are set by introspecting the current Kubernetes connection (unfurl.relationships.ConnectsTo.K8sCluster). There are no default implementations defined for creating or destroying a cluster.

Attributes

apiServer

The url used to connect to the cluster’s api server.

unfurl.nodes.K8sNamespace

Represents a Kubernetes namespace. Destroying a namespace deletes any resources in it. Derived from unfurl.nodes.K8sRawResource.

Requirements

host

A node template of type unfurl.nodes.K8sCluster

Properties

name

The name of the namespace.

unfurl.nodes.K8sResource

Requirements

host

A node template of type unfurl.nodes.K8sNamespace

Properties

definition

(map or string) The YAML definition for the Kubernetes resource.

Attributes

apiResource

(map) The YAML representation for the resource as retrieved from the Kubernetes cluster.

name

(string) The Kubernetes name of the resource.

unfurl.nodes.K8sSecretResource

Represents a Kubernetes secret. Derived from unfurl.nodes.K8sResource.

Requirements

host

A node template of type unfurl.nodes.K8sNamespace

Properties

data

(map) Name/value pairs that define the secret. Values will be marked as sensitive.

Attributes

apiResource

(map) The YAML representation for the resource as retrieved from the Kubernetes cluster. Data values will be marked as sensitive.

name

(string) The Kubernetes name of the resource.

unfurl.nodes.K8sRawResource

A Kubernetes resource that isn’t part of a namespace.

Requirements

host

A node template of type unfurl.nodes.K8sCluster

Properties

definition

(map or string) The YAML definition for the Kubernetes resource.

Attributes

apiResource

(map) The YAML representation for the resource as retrieved from the Kubernetes cluster.

name

(string) The Kubernetes name of the resource.

Supervisor

Supervisor is a light-weight process manager that is useful when you want to run local development instances of server applications.

Required TOSCA import: configurators/templates/supervisor.yaml (in the unfurl repository)

unfurl.nodes.Supervisor

TOSCA type that represents an instance of Supervisor process manager. Derived from tosca.nodes.SoftwareComponent.

properties

homeDir

(string) The location the Supervisor configuration directory (default: {get_dir: local})

confFile

(string) Name of the confiration file to create (default: supervisord.conf)

conf

(string) The supervisord configuration. A default one will be generated if omitted.

unfurl.nodes.ProcessController.Supervisor

TOSCA type that represents a process (“program” in supervisord terminology) that is managed by a Supervisor instance. Derived from unfurl.nodes.ProcessController.

requirements

host

A node template of type unfurl.nodes.Supervisor.

properties

name

(string) The name of this program.

program

(map) A map of settings for this program.

Artifacts

Instead of setting an operation’s Implementation to a configurator, you can set it to an artifact. Using an artifact allows you to reuse an implementation with more than one operation. For example, you can create artifacts for specific Terraform modules, Ansible playbooks, or executables.

You define an execute operation on an artifact’s type or template definition to specify the inputs and outputs that can be passed to the artifact’s configurator. How the inputs and outputs are used depends on the artifact’s type. For example, with a Terraform module artifact, its inputs will be used as the Terraform module’s variables and its outputs the Terraform module’s outputs. Or with a shell executable artifact, the inputs specify the command line arguments passed to the executable.

The example below declares an artifact that represents a shell script and shows how an operation can invoke the artifact and pass values to it.

PYTHON

import unfurl
import tosca
from tosca import ToscaOutputs, Attribute, Eval


class MyArtifact(unfurl.artifacts.ShellExecutable):
    file: str = "myscript.sh"

    # evaluates the script's output
    outputsTemplate = Eval("{{ stdout | from_json }}")

    class Outputs(ToscaOutputs):
        output1: str = Attribute()
        output2: int = Attribute()

    # no implementation is needed for execute() because:
    # - execute() arguments are passed as command line arguments by default
    # - MyArtifact.Outputs is constructed from the json map returned by outputsTemplate
    def execute(self, arg1: str, arg2: int) -> Outputs:
        return MyArtifact.Outputs()


class MyNode(tosca.nodes.Root):
    prop: int

    def configure(self) -> MyArtifact.Outputs:
        return MyArtifact(input="y").execute("hello", self.prop)


my_node = MyNode(prop=1)

YAML

tosca_definitions_version: tosca_simple_unfurl_1_0_0
topology_template:
  node_templates:
    my_node:
      type: MyNode
      properties:
        prop1: foo
        prop2: 1
      metadata:
        module: docs.examples.artifact2
artifact_types:
  MyArtifact:
    derived_from: unfurl.artifacts.ShellExecutable
    properties:
      outputsTemplate:
        type: any
        required: false
        default: '{{ stdout | from_json }}'
    interfaces:
      Executable:
        type: unfurl.interfaces.Executable
        operations:
          execute:
            metadata:
              output_match:
              - Outputs
            inputs:
              arg1:
                type: string
              arg2:
                type: integer
            outputs:
              output1:
                type: string
              output2:
                type: integer
node_types:
  MyNode:
    derived_from: tosca.nodes.Root
    properties:
      prop1:
        type: string
      prop2:
        type: integer
    interfaces:
      Standard:
        operations:
          configure:
            metadata:
              output_match:
              - Outputs
              arguments:
              - arg1
              - arg2
            inputs:
              arg1: hello
              arg2:
                eval: .::prop2
            outputs:
              output1:
                type: string
              output2:
                type: integer
            implementation:
              primary:
                type: MyArtifact
                properties:
                  input: y
                file: myscript.sh

Like other TOSCA operations, when generating TOSCA YAML, Unfurl looks at the execute method’s signature to determine its TOSCA inputs and outputs. So in simple cases the method can be a no-op. In more complex cases, you can use the method to validate or transform the inputs before they are passed to the configurator. In that case, you can a call set_inputs in the method body, which will force the method to be invoked at runtime when the operation’s task is created in a job’s planning stage. Or you could implement completely custom behavior by having the execute method return a run method using the pattern shown in this example.

Arguments

When a node operation invokes its implementation artifact’s execute operation, Unfurl looks for a operation input named arguments to pass as the execute operation’s inputs. This should be dictionary whose keys and values corresponds to the execute operation’s input specification.

If an arguments input isn’t explicitly declared, it will be synthesized from the following sources (listed here from lowest to highest merge order):

• Default input values defined for the execution operation.

• Properties on the node template and on the implementation artifact if they have matching input_match metadata keys in their definitions (see Shared Properties).

• The operation’s inputs whose names are listed in the operation’s arguments metadata key, if set. The Python DSL generates this based on the operation method’s call to the execute method, as shown in the example above.

• If the arguments metadata key is missing, operation inputs whose name matches execute a operation’s input name or one of the above matching property names.

Shared Properties

In the Python DSL, TOSCA types can inherit from tosca.ToscaInputs and tosca.ToscaOutputs classes using multiple inheritance and their inherited fields will be will be treated as TOSCA properties and attributes respectively. If an execute operation declares ToscaInputs subclass as a parameter in its method signature, any node or artifact that inherited that ToscaInputs class will have those properties automatically passed as arguments. This way implementation definitions stay in sync with the nodes that use them.

In YAML, you can do the equivalent by adding a input_match metadata key to those properties to indicate they should be treated as arguments to operations. When invoking an operation, any property on the node or on the implementation artifact has that set will be added arguments. Its value can be a boolean or the name of an artifact to indicate that it should only be passed as arguments to operations that use that artifact. You can also control with properties are passed as arguments by adding an input_match metadata key to the artifact execute interface’s metadata – if set, only properties with matching input_match values will be set. Similarly, you can set the output_match metadata key to map operation outputs to node attributes, as an alternative to declaring an attribute mapping on an operation. The YAML generated by the Python DSL uses that mechanism as the example below shows:

PYTHON

import unfurl
import tosca

class KubernetesClusterInputs(tosca.ToscaInputs):
    do_region: str = "nyc3" 

class KubernetesClusterOutputs(tosca.ToscaOutputs):
    do_id: str = tosca.Attribute()

class ClusterTerraformModule(unfurl.artifacts.TerraformModule):
    file: str = "my_module/main.tf"
    def execute(self, inputs: KubernetesClusterInputs) -> KubernetesClusterOutputs:
        return KubernetesClusterOutputs()

# inherited ToscaInputs and ToscaOutputs are treated as properties and attributes
class Cluster(tosca.nodes.Root, KubernetesClusterInputs, KubernetesClusterOutputs):
    cluster_config: "ClusterTerraformModule" = ClusterTerraformModule()

    my_property: str = "default"

    def configure(self) -> KubernetesClusterOutputs:
        return self.cluster_config.execute(self)

YAML

tosca_definitions_version: tosca_simple_unfurl_1_0_0
topology_template: {}
artifact_types:
  ClusterTerraformModule:
    derived_from: unfurl.artifacts.TerraformModule
    interfaces:
      Executable:
        type: unfurl.interfaces.Executable
        operations:
          execute:
            metadata:
              output_match:
              - KubernetesClusterOutputs
              input_match:
              - KubernetesClusterInputs
            inputs:
              do_region:
                type: string
                default: nyc3
            outputs:
              do_id:
                type: string
node_types:
  Cluster:
    derived_from: tosca.nodes.Root
    properties:
      do_region:
        type: string
        default: nyc3
        metadata:
          input_match: KubernetesClusterInputs
      my_property:
        type: string
        default: default
    attributes:
      do_id:
        type: string
        metadata:
          output_match: KubernetesClusterOutputs
    artifacts:
      cluster_config:
        type: ClusterTerraformModule
        file: my_module/main.tf
    interfaces:
      Standard:
        operations:
          configure:
            metadata:
              output_match:
              - KubernetesClusterOutputs
              arguments:
              - do_region
            inputs:
              do_region:
                eval: .::do_region
            outputs:
              do_id:
                type: string
            implementation:
              primary: cluster_config

Abstract artifacts

You can define abstract artifact types that just define the inputs and outputs it expects by defining an artifact type with an execute operation that doesn’t have an implementation declared. Artifacts can implement that by, for example, by using multiple inheritance to inherit both the abstract artifact type and a concrete artifact type like unfurl.artifacts.TerraformModule.

This way a node type can declare operations with abstract artifacts and node templates or a node subclass can set a concrete artifact without having to reimplement the operations that use it – with the assurance that the static type checker will check that operation signatures are compatible.

The example below defines a node type specifies the abstract artifact type its configuration operation will use and a node template that uses a concrete artifact that implements the abstract artifact type.

PYTHON

import unfurl
import tosca

class ClusterOp(unfurl.artifacts.Executable):
    file: str = "kubernetes"

    def execute(self, prop1: str, prop2: int):
        pass

class DOCluster(tosca.nodes.Root):
    cluster_config: "ClusterOp"

    my_property: str = "default"

    def configure(self, **kw):
        return self.cluster_config.execute(prop1=self.my_property, prop2=0)

class ClusterTerraform(unfurl.artifacts.TerraformModule, ClusterOp):
    file: str = "main.tf"

mycluster = DOCluster(cluster_config=ClusterTerraform())

YAML

tosca_definitions_version: tosca_simple_unfurl_1_0_0
topology_template:
  node_templates:
    mycluster:
      type: DOCluster
      artifacts:
        cluster_config:
          type: ClusterTerraform
          file: main.tf
      metadata:
        module: docs.examples.artifact3
artifact_types:
  ClusterOp:
    derived_from: unfurl.artifacts.Executable
    interfaces:
      Executable:
        type: unfurl.interfaces.Executable
        operations:
          execute:
            inputs:
              prop1:
                type: string
              prop2:
                type: integer
  ClusterTerraform:
    derived_from:
    - unfurl.artifacts.TerraformModule
    - ClusterOp
node_types:
  DOCluster:
    derived_from: tosca.nodes.Root
    artifacts:
      cluster_config:
        type: ClusterOp
    properties:
      my_property:
        type: string
        default: default
    interfaces:
      Standard:
        operations:
          configure:
            inputs:
              prop1:
                eval: .::my_property
              prop2: 0
            metadata:
              arguments:
              - prop1
              - prop2
            implementation:
              primary: cluster_config