Upbound Docs - Authoring Compositions

Authoring Compositions

After you author an XRD, up composition generate allows you to create a composition based on the parameters of your XRD.

Scaffold the composition from the XRD

In the root folder of your control plane project, run the up composition generate command.

up composition generate apis/xbuckets/definition.yaml

This generates a new composition for you in apis/xbuckets/composition.yaml. Open the file in your editor to review the minimal file created.

apiVersion: apiextensions.crossplane.io/v1
kind: Composition
metadata:
  creationTimestamp: null
  name: xbuckets.devexdemo.upbound.io
spec:
  compositeTypeRef:
    apiVersion: devexdemo.upbound.io/v1alpha1
    kind: XBucket
  mode: Pipeline
  pipeline:
  - functionRef:
      name: crossplane-contrib-function-auto-ready
    step: crossplane-contrib-function-auto-ready

Generate an embedded function

Functions allow you to define the logic of your composition. Composition functions build, package, and manage deployment logic as part of your configuration. You can write functions in familiar programming languages rather than using the built-in patch-and-transform YAML workflow.

To generate a function based on your composition, run the following command:

up function generate --language=python test-function apis/xbuckets/composition.yaml

This command generates an embedded Python function called test-function and creates a new file in your project under functions/test-function/main.py. The up function generate command also creates schema models to help with your authoring experience.

To generate a function based on your composition, run the following command:

up function generate --language=kcl test-function apis/xbuckets/composition.yaml

This command generates an embedded KCL function called test-function and creates a new file in your project under functions/test-function/main.k. The up function generate command also creates schema models to help with your authoring experience.

To generate a function based on your composition, run the following command:

up function generate --language=go test-function apis/xbuckets/composition.yaml

This command generates an embedded Go function called test-function in a new directory, functions/test-function. The up function generate command also generates a Go module containing types you can use when authoring your function.

To generate a function based on your composition, run the following command:

up function generate --language=go-templating test-function apis/xbuckets/composition.yaml

This command generates an embedded Go Templating function called test-function and scaffolds it in a new directory in your project, functions/test-function. The up function generate command also creates schema models to help with your authoring experience.

Authoring the composition function

For this example, you need Python and the Python Visual Studio Code extension. Refer to the Visual Studio Code Extensions documentation to learn how to install them.

Open the main.py function file in Visual Studio Code.

from crossplane.function import resource
from crossplane.function.proto.v1 import run_function_pb2 as fnv1

from .model.io.upbound.aws.s3.bucket import v1beta1

def compose(req: fnv1.RunFunctionRequest, rsp: fnv1.RunFunctionResponse):
    # Specify an S3 Bucket, using a generated model.
    bucket = v1beta1.Bucket(
        apiVersion="s3.aws.upbound.io/v1beta1",
        kind="Bucket",
        spec=bucketv1beta1.Spec(
            forProvider=bucketv1beta1.ForProvider(
                # Derive the bucket's region from the XR's region.
                region=req.observed.composite.resource["spec"]["region"],
            ),
        ),
    )

    # Update the function's desired composed resources to include the bucket.
    resource.update(rsp.desired.resources["bucket"], bucket)

Use import statements to load Crossplane’s Python SDK and Upbound’s generated models into your function.

Define your function’s logic in the compose function. Crossplane calls this function. It passes it a RunFunctionRequest and a RunFunctionResponse.

Specify your desired composed resources by passing them to resource.update. You can pass resource.update a model object, or a Python dictionary.

You can also use resource.update to update the desired XR:

from crossplane.function import resource
from crossplane.function.proto.v1 import run_function_pb2 as fnv1

def compose(req: fnv1.RunFunctionRequest, rsp: fnv1.RunFunctionResponse):
    # Update the function's desired XR's status using a dictionary.
    resource.update(rsp.desired.composite, {"status": {"widgets": 42}})

With the Visual Studio Code Python extension you get autocompletion, linting, type errors, and more.

For more Python best practices, please refer to the documentation.

For this example, make sure you have KCL and the KCL language server installed:

curl -fsSL "https://kcl-lang.io/script/install-cli.sh" | /bin/bash
curl -fsSL "https://kcl-lang.io/script/install-kcl-lsp.sh" | /bin/bash

Next, install and enable the KCL Visual Studio Code Extension.

Open the main.k function file in Visual Studio Code. The schema scaffold here builds your composition logic and contains placeholders for your desired inputs.

import models.v1beta1 as v1beta1
import models.v1beta2 as v1beta2
import models.k8s.apimachinery.pkg.apis.meta.v1 as metav1

oxr = option("params").oxr # observed composite resource
_ocds = option("params").ocds # observed composed resources
_dxr = option("params").dxr # desired composite resource
dcds = option("params").dcds # desired composed resources

_metadata = lambda name: str -> any {
    { annotations = { "krm.kcl.dev/composition-resource-name" = name }}
}

_items = [

]
items = _items

First, the function uses an import statement to load KCL language server logic into your composition.

import models.v1beta1 as v1beta1
import models.v1beta2 as v1beta2
import models.k8s.apimachinery.pkg.apis.meta.v1 as metav1

Next, the variable statements capture your desired resources and observed resources.

oxr = option("params").oxr # observed composite resource
_ocds = option("params").ocds # observed composed resources
_dxr = option("params").dxr # desired composite resource
dcds = option("params").dcds # desired composed resources

The _items stanza is empty and expects a resource object. Your function uses this assignment to pass your desired infrastructure to the control plane for management.

_items = [
  v1beta1.Bucket{
    metadata.name = oxr.metadata.name
    spec.forProvider: {
      objectLockEnabled = True
      forceDestroy = False
    }
  }

  v1beta1.BucketVersioning {
    spec.forProvider: {
        bucketRef.name = oxr.metadata.name
    }
   }   if oxr.spec.versioning and oxr.status.conditions == 'True' else {}

   v1beta1.BucketServerSideEncryptionConfiguration {
    spec.forProvider = {
        bucketRef.name = oxr.metadata.name
        rule: [
          {
            applyServerSideEncryptionByDefault: [
                {
                    sseAlgorithm = "AES256"
                }
            ]
            bucketKeyEnabled = True
          }
        ]
    }
   } if oxr.spec.versioning and oxr.status.conditions == 'True' else {}
]

The statement in the _items variable is a fully functional KCL function. With the Visual Studio Code KCL extension and KCL language server, you get autocompletion, linting, type errors, and more.

In this example, the oxr assignment captures the composite resources and the function adds server side encryption to the buckets your deployment creates.

For more KCL best practices, please refer to the documentation.

For this example, you need Go and the Go Visual Studio Code extension. Refer to the Visual Studio Code Extensions documentation to learn how to install them.

Open the fn.go function file in Visual Studio Code.

package main

import (
	"context"
	"encoding/json"

	"dev.upbound.io/models/com/example/platform/v1alpha1"
	"dev.upbound.io/models/io/upbound/aws/s3/v1beta1"
	"k8s.io/utils/ptr"

	"github.com/crossplane/function-sdk-go/errors"
	"github.com/crossplane/function-sdk-go/logging"
	fnv1 "github.com/crossplane/function-sdk-go/proto/v1"
	"github.com/crossplane/function-sdk-go/request"
	"github.com/crossplane/function-sdk-go/resource"
	"github.com/crossplane/function-sdk-go/resource/composed"
	"github.com/crossplane/function-sdk-go/response"
)

// Function is your composition function.
type Function struct {
	fnv1.UnimplementedFunctionRunnerServiceServer

	log logging.Logger
}

// RunFunction runs the Function.
func (f *Function) RunFunction(_ context.Context, req *fnv1.RunFunctionRequest) (*fnv1.RunFunctionResponse, error) {
	f.log.Info("Running function", "tag", req.GetMeta().GetTag())
	rsp := response.To(req, response.DefaultTTL)

	observedComposite, err := request.GetObservedCompositeResource(req)
	if err != nil {
		response.Fatal(rsp, errors.Wrap(err, "cannot get xr"))
		return rsp, nil
	}

	var xr v1alpha1.XStorageBucket
	if err := convertViaJSON(&xr, observedComposite.Resource); err != nil {
		response.Fatal(rsp, errors.Wrap(err, "cannot convert xr"))
		return rsp, nil
	}

	params := xr.Spec.Parameters
	if params.Region == nil || *params.Region == "" {
		response.Fatal(rsp, errors.Wrap(err, "missing region"))
		return rsp, nil
	}

	bucket := &v1beta1.Bucket{
		APIVersion: ptr.To("s3.aws.upbound.io/v1beta1"),
		Kind:       ptr.To("Bucket"),
		Spec: &v1beta1.BucketSpec{
			ForProvider: &v1beta1.BucketSpecForProvider{
				Region: params.Region,
			},
		},
	}

	composedBucket := composed.New()
	if err := convertViaJSON(composedBucket, bucket); err != nil {
		response.Fatal(rsp, errors.Wrap(err, "cannot convert bucket to unstructured"))
		return rsp, nil
	}
	desiredComposedResources, err := request.GetDesiredComposedResources(req)
	if err != nil {
		response.Fatal(rsp, errors.Wrap(err, "cannot get desired resources"))
		return rsp, nil
	}
	desiredComposedResources["bucket"] = &resource.DesiredComposed{Resource: composedBucket}
	if err := response.SetDesiredComposedResources(rsp, desiredComposedResources); err != nil {
		response.Fatal(rsp, errors.Wrap(err, "cannot set desired resources"))
		return rsp, nil
	}

	return rsp, nil
}

func convertViaJSON(to, from any) error {
	bs, err := json.Marshal(from)
	if err != nil {
		return err
	}
	return json.Unmarshal(bs, to)
}

Use import statements to load Crossplane’s Go SDK and Upbound’s generated models into your function.

Define your function’s logic in the RunFunction function. Crossplane calls this function via gRPC, passing it a RunFunctionRequest. It returns a RunFunctionResponse that includes resources for Crossplane to create and update.

Convert your desired composed resources to the necessary SDK types and add them to the response with response.SetDesiredComposedResources. Crossplane creates or updates the desired composed resources returned in the response. You can also use response.SetDesiredCompositeResource to update the status of the XR.

With the Visual Studio Code Go extension you get autocompletion, linting, type errors, and more.

For more Go best practices, please refer to the documentation.

For this example, you need the YAML Visual Studio Code extension and optionally the Modelines extension. Refer to the Visual Studio Code Extensions documentation to learn how to install them.

A Go templating function is a set of Go templates that produce YAML when executed. The resulting YAML may contain desired composed resources for Crossplane to create, or updates for Crossplane to apply to the composite resource’s status. See the documentation for Crossplane’s function-go-templating for full details on the features available.

Open the 01-compose.yaml.gotmpl file in Visual Studio Code.

# code: language=yaml
# yaml-language-server: $schema=../../.up/json/models/index.schema.json

---
apiVersion: s3.aws.upbound.io/v1beta1
kind: Bucket
metadata:
  annotations:
    {{ setResourceNameAnnotation "bucket" }}
spec:
  forProvider:
    region: "{{ $xr.spec.parameters.region }}"

The generated boilerplate code in 00-prelude.yaml.gotmpl has already loaded the observed composite resource into the $xr Go templating variable. The function runtime executes template files in lexical order, so subsequent templates (including 01-compose.yaml.gotmpl) can use $xr to access the observed XR. In this case, the template copies the region from the XR into the Bucket managed resource.

Use the setResourceNameAnnotation Go templating function to set a resource name on each desired composed resource. Crossplane will generate an appropriate name, allowing you to update the same resource on subsequent function invocations.

You can also update the status of the desired XR by including it in your template:

---
apiVersion: devexdemo.upbound.io/v1alpha1
kind: XBucket
status:
  widgets: 42

With the Visual Studio Code YAML extension you get autocompletion, linting, type errors, and more.

For more Go templating best practices, please refer to the documentation.

In the next guide, you’ll run and test your composition.

Next steps

After you author your compositions and embedded functions, you can build and test your control plane projects in Upbound. To learn more about testing, review the testing guide.