Built-in Workflows

Overview

Cloudify comes with a number of built-in workflows, covering:

Built-in workflows are declared and mapped in types.yaml, which is usually imported either directly or indirectly via other imports.

# Snippet from types.yaml

workflows:
  install: default_workflows.cloudify.plugins.workflows.install
  uninstall: default_workflows.cloudify.plugins.workflows.uninstall
  execute_operation:
    mapping: default_workflows.cloudify.plugins.workflows.execute_operation
    parameters:
      operation: {}
      operation_kwargs:
        default: {}
      run_by_dependency_order:
        default: false
      type_names:
        default: []
      node_ids:
        default: []
      node_instance_ids:
        default: []

The implementations for these workflows can be found at cloudify-plugins-common.

Built-in workflows are not special in any way - they use the same API and framework as any custom workflow is able to use, and one may replace them with different workflows with the same names.

The Install Workflow

Workflow name: install

Workflow description: Workflow for installing applications.

Workflow high-level pseudo-code:

For each node, for each node instance (in parallel):

  1. Wait for node instance relationships to be started. (Only start processing this node instance when the node instances it depends on are started).
  2. Execute cloudify.interfaces.lifecycle.create operation. 1
  3. Execute cloudify.interfaces.relationship_lifecycle.preconfigure relationship operations.2
  4. Execute cloudify.interfaces.lifecycle.configure operation.1
  5. Execute cloudify.interfaces.relationship_lifecycle.postconfigure relationship operations.2
  6. Execute cloudify.interfaces.lifecycle.start operation.1
  7. If the node instance is a host node (its type is a subtype of cloudify.nodes.Compute):
    • Install agent workers and required plugins on this host.
    • Execute cloudify.interfaces.monitoring_agent interface install and start operations. 1
  8. Execute cloudify.interfaces.monitoring.start operation. 1
  9. Execute cloudify.interfaces.relationship_lifecycle.establish relationship operations.2

The Uninstall Workflow

Workflow name: uninstall

Workflow description: Workflow for uninstalling applications.

Workflow parameters:

Workflow high-level pseudo-code:

For each node, for each node instance (in parallel):

  1. Wait for dependent node instances to be deleted. (Only start processing this node instance when the node instances dependent on it are deleted).
  2. Execute cloudify.interfaces.monitoring.stop operation. 1
  3. If node instance is host node (its type is a subtype of cloudify.nodes.Compute):
    • Execute cloudify.interfaces.monitoring_agent interface stop and uninstall operations. 1
    • Stop and uninstall agent workers.
  4. Execute cloudify.interfaces.lifecycle.stop operation.1
  5. Execute cloudify.interfaces.relationship_lifecycle.unlink relationship operations.2
  6. Execute cloudify.interfaces.lifecycle.delete operation.1

The Execute Operation Workflow

Workflow name: execute_operation

Workflow description: Generic workflow for executing arbitrary operations on nodes.

Workflow parameters:

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

Executing an operation on a node which has that interface operation but has not mapped it to any concrete implementation will simply do nothing. However, attempting to execute an operation on a node which doesn’t have the relevant interface operation will result in a workflow execution error. Use the filtering fields to ensure the operation is only executed on nodes which the operation might be relevant to.

Workflow high-level psuedo-code:

For each node, for each node instance:

  1. Filter out instances whose node is not in the node_ids list (unless its empty).
  2. Filter out instances whose id is not in the node_instance_ids list (unless its empty).
  3. Filter out instances whose node type is not in or a descendant of a type which is in the type_names list (unless its empty).

If run_by_dependency_order is set to true: create a task dependency between the following section’s (1) task for a given instance and the (3) task for all instances it depends on.1

For each of the remaining node instances:

  1. Send a node instance event about starting the execution operation.
  2. Execute the operation operation for the instance, with the operation_kwargs passed to the operation invocation.
  3. Send a node instance event about completing the execution of the operation.

The Start Workflow

Workflow name: start

Workflow description: Can be used to start all, or a subset of, node templates.

This workflow is a wrapper for the execute_operation workflow, allowing the user to easily start the topology (or a subset thereof). Calling the start workflow is equivalent to calling execute_operation while passing cloudify.interfaces.lifecycle.start as the operation name.

Workflow parameters:

The Stop Workflow

Workflow name: stop

Workflow description: Can be used to stop all, or a subset of, node templates.

This workflow is a wrapper for the execute_operation workflow, allowing the user to easily stop the topology (or a subset thereof). Calling the stop workflow is equivalent to calling execute_operation while passing cloudify.interfaces.lifecycle.stop as the operation name.

Workflow parameters:

See workflow parameters for the start workflow above.

The Restart Workflow

Workflow name: restart

Workflow description: Can be used to restart all, or a subset of, node templates.

This workflow simply calls the stop workflow, followed by start.

Workflow parameters:

NOTE: The restart workflow performs all stop operations first, and then performs all start operations.

The Heal Workflow

Workflow name: heal

Workflow description: Reinstalls the whole subgraph of the system topology by applying the uninstall and install workflows’ logic respectively. The subgraph consists of all the node instances that are contained in the compute node instance which contains the failing node instance and/or the compute node instance itself. Additionally, this workflow handles unlinking and establishing all affected relationships in an appropriate order.

Workflow parameters:

Workflow high-level pseudo-code:

  1. Retrieve the compute node instance of the failed node instance.

  2. Construct a compute sub-graph (see note below).

  3. Uninstall the sub-graph:

    • Execute uninstall lifecycle operations (stop, delete) on the compute node instance and all it’s contained node instances. (1)
    • Execute uninstall relationship lifecycle operations (unlink) for all affected relationships.
  4. Install the sub-graph:

    • Execute install lifecycle operations (create, configure, start) on the compute node instance and all it’s contained nodes instances.
    • Execute install relationship lifecycle operations (preconfigure, postconfigure, establish) for all affected relationships.

A compute sub-graph can be thought of as a blueprint that defines only nodes that are contained inside a compute node. For example, if the full blueprint looks something like this:

...

node_templates:

  webserver_host:
    type: cloudify.nodes.Compute
    relationships:
      - target: floating_ip
        type: cloudify.relationships.connected_to

  webserver:
    type: cloudify.nodes.WebServer
    relationships:
      - target: webserver_host
        type: cloudify.relationships.contained_in

  war:
    type: cloudify.nodes.ApplicationModule
    relationships:
      - target: webserver
        type: cloudify.relationships.contained_in
      - target: database
        type: cloudify.relationships.connected_to

  database_host:
    type: cloudify.nodes.Compute

  database:
    type: cloudify.nodes.Database
    relationships:
      - target: database_host
        type: cloudify.relationships.contained_in

  floating_ip:
    type: cloudify.nodes.VirtualIP

...

Then the corresponding graph will look like so:

Blueprint as Graph

And a compute sub-graph for the webserver_host will look like:

Blueprint as Graph

This sub-graph determines the operations that will be executed during the workflow execution. In this example:

The Scale Workflow

Workflow name: scale

Workflow description:

Scales out/in the node subgraph of the system topology applying the install/uninstall workflows’ logic respectively.

If the entity denoted by scalable_entity_name is a node template that is contained in a compute node (or is a compute node itself) and scale_compute is true, the node graph will consist of all nodes that are contained in the compute node which contains scalable_entity_name and the compute node itself. Otherwise, the subgraph will consist of all nodes that are contained in the node/scaling group denoted by scalable_entity_name.

In addition, nodes that are connected to nodes that are part of the contained subgraph will have their establish relationship operations executed during scale out and their unlink relationship operations executed during scale in.

Workflow parameters:

Workflow high-level pseudo-code:

  1. Retrieve the scaled node/scaling group, based on scalable_entity_name and scale_compute parameters.
  2. Start deployment modification, adding or removing node instances and relationship instances.
  3. If delta > 0:
    • Execute install lifecycle operations (create, configure, start) on added node instances.
    • Execute the establish relationship lifecycle operation for all affected relationships.
  4. If delta < 0:
    • Execute the unlink relationship lifecycle operation for all affected relationships.
    • Execute uninstall lifecycle operations (stop, delete) on removed node instances.

The Install New Agents Workflow

Workflow name: install_new_agents

Workflow description:

Installs agents on all VMs related to a particular deployment and connects them to the Cloudify Manager’s RabbitMQ instance. Please note that the old Manager has to be running during the execution of this workflow. What is worth mentioning as well is that the old agents don’t get uninstalled. This workflow’s common use case is executing it after having successfully restored a snapshot on a new Manager in order for the Manager to gain control over applications that have been orchestrated by the previous Manager.

Workflow parameters: