Intrinsic Functions

Intrinsic functions are functions that can be used within blueprints. Depending on the function, evaluation occurs on deployment creation or in runtime. During deployment creation, the “runtime only evaluation” flag can be set, which makes all functions be evaluated on-demand in runtime.

For example, the get_input intrinsic function is used to retrieve an input defined within the blueprint. Intrinsic functions make blueprints dynamic, enabling the retrieval and setting of data structures in different parts of the blueprint.

get_secret

get_secret is used for referencing secrets described in the secrets API. get_secret can be used in node properties, outputs, node/relationship operation inputs, and runtime-properties of node instances. The function is evaluated at runtime.

Example:

node_templates:
  host:
    type: cloudify.nodes.Compute
    properties:
      ip: { get_secret: ip }
      cloudify_agent:
        key: { get_secret: agent_key }
        user: { get_secret: user }
    interfaces:
      test_interface:
        test_operation:
          implementation: central_deployment_agent
          inputs:
            operation_input: { get_secret: operation_input }

outputs:

  webserver_url:
    description: Web server url
    value: { concat: ['http://', { get_secret: ip }, ':', { get_secret: webserver_port }] }

In this example, get_secret is used for completing several of the host node’s properties, as well as an operation input. In addition, it is used twice in the concatenated webserver_url output.

get_input

get_input is used for referencing inputs described in the inputs section of the blueprint. get_input can be used in node properties, outputs, and node/relationship operation inputs. The function is evaluated on deployment creation by default (unless the “runtime only evaluation” flag is set).

Example:

inputs:

  webserver_port:
    description: The HTTP web server port
    default: 8080
    
  vm_info:
    description: The HTTP web server port
    default:
      key_name: 'my-openstack-key-name'

node_templates:
  ...

  vm:
    type: cloudify.openstack.nodes.Server
    properties:
      server:
        image_name: { get_input: image_name }
        key_name: { get_input: [ vm_info, key_name ] }

  ...

  http_web_server:
    type: cloudify.nodes.WebServer
    properties:
      port: { get_input: webserver_port }
    relationships:
      - type: cloudify.relationships.contained_in
        target: vm

outputs:

  webserver_port:
    description: Web server port
    value: { get_input: webserver_port }

  vm_key_name:
    description: Web server port
    value: { get_input: [ vm_info, key_name ] }

In the example, get_input is used for supplying the http_web_server node’s port property and the vm node’s key_name property. If on deployment creation the webserver_port input is not specified, get_input returns the default value of the webserver_port input. Similarly, if the vm_info input is not specified, get_input returns the default value of the vm_info input.

get_capability

get_capability is used for referencing capabilities defined in other deployments, as described in the capabilities section of the blueprint. get_capability can be used in node properties, outputs, and node/relationship operation inputs. The function is evaluated at runtime. This means that the results of the evaluation may differ according to their original values in the defining deployment.

Example:

First, we need to create a deployment that defines capabilities:

inputs:
  some_input: some_value

node_types:
  test_type:
    derived_from: cloudify.nodes.Root
    properties:
      key:
        default: default_value
  
  dummy_type:
    derived_from: cloudify.nodes.Root
    properties:
      input_property: { get_input: some_input }

node_templates:
  node1:
    type: test_type
  node2:
    type: test_type
    properties:
      key: override_value
  dummy_node:
    type: dummy_type

capabilities:
  node_1_key:
    value: { get_attribute: [ node1, key ]}
  node_2_key:
    value: { get_attribute: [ node2, key ]}
  complex_capability:
    value:
      level_1:
        level_2:
          level_3: [ value_1, value_2 ]
          key_1: value_3
        key_2: value_4
  input_capability:
    value: { get_attribute: [ dummy_node, input_property ]

We should note several things here:

Let’s assume now that a deployment with the ID shared was created from the above blueprint. Let’s now create a second deployment to utilize the get_capability intrinsic function:

node_types:
  test_type:
    derived_from: cloudify.nodes.Root
    properties:
      key:
        default: { get_capability: [ shared, node_1_key ] }

node_templates:
  node1:
    type: test_type
  node2:
    type: test_type
    properties:
      key: { get_capability: [ shared, node_2_key ] }

outputs:
  complex_output:
    value: { get_capability: [ shared, complex_capability ] }

  nested_complex_output:
    value: { get_capability: [ shared, complex_capability, level_1, level_2, level_3, 0 ] }
    

Here we can see how get_capability is used - the input to the function is a list with at least 2 values: the ID of the deployment, the name of the capability and optionally nested attributed/list indices, as defined in the shared blueprint. Note that both the deployment ID and the capability name can be provided using other intrinsic functions (e.g. get_input or get_secret). So, in case the deployment ID is not known in advance, we could do something like this:

outputs:
  some_output:
    value: { get_capability: [ { get_secret: shared_deployment_id }, complex_capability ] }

get_property

get_property is used for referencing node properties within a blueprint. get_property can be used in node properties, outputs, and node/relationship operation inputs. The function is evaluated on deployment creation by default (unless the “runtime only evaluation” flag is set).

Usage and Examples

get_property in node properties and interface operation inputs

node_templates:
  security_group:
    type: cloudify.openstack.nodes.SecurityGroup
    properties:
      rules:
        - remote_ip_prefix: 0.0.0.0/0
          port: { get_property: [web_server, port] }

  web_server:
    type: cloudify.nodes.WebServer
    properties:
      port: 80
    interfaces:
      cloudify.interfaces.lifecycle
        create:
          ...
        configure:
          implementation: some_plugin.tasks.configure
          inputs:
            port: { get_property: [SELF, port] }

In the example, get_property is used for specifying a security group’s rule port as the web_server node’s port. In addition, get_property is used for passing the web_server’s port property as an input to the configure operation. The keyword SELF is used for specifying that the referenced property belongs to the current node. In this case, using web_server instead of SELF provides the same outcome.

node_templates:
  db_server:
    type: cloudify.nodes.DBMS
    properties:
      endpoint: 10.0.0.1:3376

  web_server:
    type: cloudify.nodes.WebServer
    properties:
      port: 8080
    relationships:
      - target: db_server
        type: cloudify.relationships.connected_to
        source_interfaces:
          cloudify.interfaces.relationship_lifecycle:
            preconfigure:
              implementation: some_plugin.tasks.my_preconfigure
              inputs:
                db_endpoint: { get_property: [TARGET, endpoint] }
                webserver_port: { get_property: [SOURCE, port] }

In this example, get_property is used to reference the source and target nodes’ properties. The SOURCE and TARGET keywords can only be used in a relationship interface.

node_templates:
  web_server
    type: cloudify.nodes.WebServer
    properties:
      port: 80

outputs:
  web_server_id:
    description: Web server port
    value: { get_property: [web_server, port] }

get_property Nested Properties and Complex Structures

It is possible to reference nested properties within dictionaries/hashes and lists in any nesting level. To access a property within a list, the index of the item must be specified. To access values in a dictionary/hash, a key must be specified.

node_templates:
  vm:
    type: cloudify.nodes.Compute
    properties:
      ip_addresses:
        - 192.168.0.7
        - 15.67.45.29

  web_server:
    type: cloudify.nodes.WebServer
    properties:
      endpoint:
        type: http
        port: 80
    relationships:
      - target: vm
        type: cloudify.relationships.contained_in
        source_interfaces:
          cloudify.interfaces.relationship_lifecycle:
            preconfigure:
              implementation: some_plugin.tasks.my_preconfigure
              inputs:
                public_ip: { get_property: [TARGET, ip_addresses, 1] }
                endpoint_type: { get_property: [SOURCE, endpoint, type] }

get_attribute

get_attribute is used to reference runtime-properties of different node-instances from within a blueprint.

Usage and Examples

get_attribute in outputs

For this example, assume a webserver_id runtime property has been set on the web_server instance.

node_templates:
  web_server
    type: cloudify.nodes.WebServer

outputs:
  web_server_id:
    description: Web server ID
    value: { get_attribute: [web_server, webserver_id] }

In the example, the web_server_id deployment output is configured to reference the web_server runtime property webserver_id. Each time the deployment outputs are evaluated, the reference is replaced with its current value.

For this example, assume a connection_url runtime property has been set on the db_server instance and a requested_version runtime property has been set on the web_server instance.

node_templates:
  db_server:
    type: cloudify.nodes.DBMS
  web_server:
    type: cloudify.nodes.WebServer
    interfaces:
      cloudify.interfaces.lifecycle
        create:
          ...
        configure:
          implementation: some_plugin.tasks.configure
          inputs:
            db_connection_url: { get_attribute: [db_server, connection_url] }
            webserver_version: { get_attribute: [SELF, requested_version] }

In the example, each time the configure operation of web_server instances is invoked, the inputs db_connection_url and webserver_version are evaluated. The db_connection_url input evaluates to the db_server runtime property connection_url, and the webserver_version evaluates to the web_server runtime property requested_version. SELF is used to reference run-time properties of the current node instance in webserver_version.

For this example, assume a connection_url runtime property has been set on the db_server instance and a requested_version runtime property has been set on the web_server instance.

node_templates:
  db_server:
    type: cloudify.nodes.DBMS
  web_server:
    type: cloudify.nodes.WebServer
    relationships:
      - target: db_server
        type: cloudify.relationships.connected_to
        source_interfaces:
          cloudify.interfaces.relationship_lifecycle:
            preconfigure:
              implementation: some_plugin.tasks.my_preconfigure
              inputs:
                db_connection_url: { get_attribute: [TARGET, connection_url] }
                webserver_version: { get_attribute: [SOURCE, requested_version] }

In the example, each time the preconfigure relationship operation is invoked, the inputs db_connection_url and webserver_version are evaluated. The db_connection_url input evaluates to the db_server runtime property connection_url. The webserver_version evaluates to the web_server runtime property requested_version. SOURCE and TARGET are used to reference the relationship source and target node instances respectively.

get_attribute Nested Properties and Complex Structures

Attribute access may be nested and is not restricted to top-level properties. In this example, assume a webserver_spec runtime property has been set on the web_server instance with this value:

{
  "requested_version": "11.2",
  "alternative_versions": ["11.3", "12.0"],
  "endpoints": {
    "endpoint_1": {
      "description": "An endpoint of the web server",
      "url": "/endpoint1"
    },
    "endpoint_2": {
      "description": "Another endpoint of the web server",
      "url": "/endpoint2"
    }
  }
}

With this value in place, nested properties can be accessed as follows:

outputs:
  alt_version1:
    # evaluates to "12.0"
    value: { get_attribute: [web_server, webserver_spec, alternative_version, 1] }
  enpoint_2_url:
    # evaluates to "/endpoint2"
    value: { get_attribute: [web_server, webserver_spec, endpoints, endpoint_2, url] }
  partial_spec:
    value:
      version: { get_attribute: [web_server, webserver_spec, requested_version] }
      alt_versions:
        version1: { get_attribute: [web_server, webserver_spec, alternative_versions, 0] }
        version2: { get_attribute: [web_server, webserver_spec, alternative_versions, 1] }

Notice that nested properties can be either a key name in the case of a map, or an index in case of a list. Also note in partial_spec that get_attribute can be used in complex data structures and not only in a flat key/value manner.

get_attribute to get Node Instance ID

Use get_attribute to retrieve the ID of a node instance:

{ get_attribute: [node, node_instance_id] }

node can be SELF, SOURCE, TARGET or node name according to the context in the blueprint.

get_attribute Between Members of Shared Scaling Groups

In general, get_attribute cannot be used with an explicit reference (i.e. specifying a node name directly) when more than one node instance matching the specified node exists.

If however, the referenced node shares a scaling group with the referencing node, the ambiguity may be resolved.

Resolving the ambiguity for get_attribute usages in the blueprint outputs is not supported.

Following is a more detailed explanation, followed by an example.

The term referencing node depends on where in the blueprint, get_attribute is used. If it is used in a node operation’s inputs (e.g. cloudify.interfaces.lifecycle.start), referencing node is the node template under which the operation is defined. If get_attribute is used in a relationship operation’s inputs (e.g. cloudify.interfaces.relationship_lifecycle.establish), referencing node is actually referencing nodes, which are the source and target nodes in the relationship operation. Both can be used as a referencing node, and the first to resolve the ambiguity is used.

Consider the case in which A is the referencing node and B the referenced node. If A and B belong to a scaling group, and that scaling group’s instances contain only one instance of B, get_attribute resolves to using that B's instance when evaluating the get_attribute.

For example:

node_templates:
  db_server:
    type: cloudify.nodes.DBMS
  web_server:
    type: cloudify.nodes.WebServer
    interfaces:
      cloudify.interfaces.lifecycle
        configure:
          implementation: some_plugin.tasks.configure
          inputs:
            # Here, the referencing node is the web_server and the referenced
            # node is the db_server
            db_connection_url: { get_attribute: [db_server, connection_url] }
    relationships:
      - target: db_server
        type: cloudify.relationships.connected_to
        source_interfaces:
          cloudify.interfaces.relationship_lifecycle:
            preconfigure:
              implementation: some_plugin.tasks.my_preconfigure
              inputs:
                # Here, the referencing nodes are web_server and db_server and the
                # referenced node is db_server (a node can reference itself)
                db_connection_url: { get_attribute: [db_server, connection_url] }

groups:
  db_and_webserver:
    members: [db_server, web_server]

policies:
  scaling_policy1:
    type: cloudify.policies.scaling
    properties:
      default_instances: 2
    targets: [db_and_webserver]

The blueprint example defines an application with one scaling group db_and_webserver that initially has two instances. Each group instance contains one db_server node instance and one web_server node instance. Both usages of get_attribute will correctly resolve to the node instance that is together with the referencing node instance in the same scaling group instance.

<div class="panel panel-default">
<div class="panel-heading">Tip</div>
<div class="panel-body">

If a node template is contained in another node template (for example, a Webserver contained in a VM), and the containing node template is a member in a scaling group, the contained node instance is implicilty a member of the same scaling group.

Using this, you can define a scaling group containing one node (for example, a compute node).

All nodes transitively contained in that compute node can reference each other using an explicit get_attribute (reference by node name), even if the compute node has several instances (if the compute node is scaled using its scaling group and not directly). This is possible because they all implicitly belong to the same scaling group instance (that of the compute node instance containing them).

Notes, Restrictions and Limitations

When using get_attribute with an explicit reference, that is, a node’s name { get_attribute: [ web_server, webserver_spec ] } and not an implicit reference such as { get_attribute: [ SELF, webserver_spec ] } the following limitation exists.

If, at the time of evaluation, more than one node instance with that name exists and the ambiguity cannot be resolved as described in the previous section, an error is returned.

This limitation has significant implications when using get_attribute in node/relationship operation inputs, because it means the operation cannot be executed.

concat

concat is used for concatenating strings in different sections of a blueprint. concat can be used in node properties, outputs, and node/relationship operation inputs. The function is evaluated once on deployment creation, which replaces get_input and get_property usages. It is also evaluated on every operation execution and outputs evaluation, to replace usages of get_attribute.

Example

node_templates:
  ...

  http_web_server:
    type: cloudify.nodes.WebServer
    properties:
      port: 8080
      # This will evaluate to 'http://localhost:8080' during deployment creation
      local_endpoint: { concat: ['http://localhost:', { get_property: [SELF, port] }] }
    interfaces:
      cloudify.interfaces.lifecycle:
        configure: scripts/configure.sh
        start:
          implementation: scripts/start.sh
          inputs:
            process:
              env:
                port: { get_input: webserver_port }
                # This will evaluate to 'http://192.168.12.12:8080' before the 'start'
                # operation execution, assuming `the_vm` private ip address is 192.168.12.12
                internal_endpoint: { concat: ['http://', { get_attribute: [the_vm, ip] },
                                              ':', { get_property: [SELF, port] }] }
        stop: scripts/stop.sh

outputs:
  external_endpoint:
    description: Web server external endpoint
    # This will evaluate to 'http://15.16.17.18:8080' every time outputs are evaluated
    # assuming `the_floating_ip` address is 15.16.17.18
    value: { concat: ['http://', { get_attribute: [the_foating_ip, floating_ip_address] },
                      ':', { get_property: [http_web_server, port] }] }

Intrinsic Functions as arguments of other Intrinsic Functions

Intrinsic Functions can be passed as arguments of other Intrinsic Functions. For example, you may write this in your blueprint:

node_templates:
  ...

  http_web_server:
    type: cloudify.nodes.Compute
    properties:
      # The parser will first evaluate the result of the inner `get_input` and 
      # then using it's result, will evaluate the outter get_input.
      # If the input available_ports = [8000, 8080] and web_server_port_no = 0, 
      # then http_web_server.port = 8000.
      port: { get_input: [ available_ports, { get_input: web_server_port_no } ] }
  
  ...

<div class="panel panel-primary panel-warning">
<div class="panel-heading">Limitation</div>
<div class="panel-body">

The arguments of a static function (e.g. get_input, get_property) may not contain any runtime function (e.g. get_attribute, get_secret). You may also note that function concat for example is neither static or runtime, therefore it may be passed as an argument to a static function as long as it doesn’t contain runtime functions. For example this is OK:

outputs: some_output: value: { get_input: { concat: [ { get_property: [ server1, protocol_prefix ] }, { get_property: [ server1, preset_ip_addr ] } ] }} …

This is not OK (a runtime function get_attribute within a static function get_input):

outputs: some_output: value: { get_input: { concat: [ { get_property: [ server1, protocol_prefix ] }, { get_attribute: [ server1, ip_addr ] } ] }} …