compute-daisy

Reusing Workflows

Instead of requiring you to copy-paste large sections of workflows to perform similar tasks, Daisy provides two different step types for workflow reuse: IncludeWorkflow and SubWorkflow. Both step types will cause Daisy to run the workflow they specify, but they differ slightly in how they share (or do not share) resources.

IncludeWorkflow

The parent workflow of a workflow included using IncludeWorkflow has access to the child workflow’s resources, and vice versa. Sources and named resources (e.g. disks, instances) present in either workflow can be referenced by name from either workflow. Two children with a common parent can also access each other’s resources. It is up to you to ensure there are no naming collisions amongst included workflows and their parents.

IncludeWorkflow will run the steps of the included workflow in parallel with the parent workflow’s steps, according to the dependency rules of both the parent and included workflows. A workflow included with IncludeWorkflow will behave largely as if the included workflow had been copy-pasted into the parent workflow.

See also the documentation for IncludeWorkflow.

SubWorkflow

SubWorkflow runs the subworkflow specified as one step. Subworkflows do not have access to their parent’s resources. This means that any workflow can be used as a subworkflow without fear of resource name collisions. Additionally, the subworkflow inherits its parent’s Project, Zone, GCSPath, and OAuthPath fields. Even if these fields are set in the subworkflow, they will be overwritten with the parent’s values.

See also the documentation for SubWorkflow.

Using Vars

To allow parent workflows to control the behavior of their children, they can set the Vars of a child workflow. Vars are represented by simple JSON objects with three fields: Value, Description, and Required. Both IncludeWorkflow and SubWorkflow tasks provide their own Vars field, which allows the parent workflow to set the Vars of the child workflow. If a child has a Var marked Required that is not set by the parent, an error will occur.

The typical use case for parent/child Vars is to set the Required and Description of the child’s Vars (but not the Value), and then pass values from the parent. Here is an example of the child:

{
  "Name": "child",
  "Vars": {
    "var1": { "Required": true, "Description": "passed from the parent" },
    "var2": { "Required": true, "Description": "also important" }
  },
  "Steps:" { ...  },
  "Dependencies": { ... }
}

And the corresponding parent:

{
  "Name": "parent",
  ...
  "Steps": {
    "include-step": {
      "IncludeWorkflow": {
        "Path": "path-to-child.wf.json",
        "Vars": {
          "var1": "value1",
          "var2": "value2"
        }
      }
    },
    ...
  },
  "Dependencies": { ... }
}