Skip to content

jonashackt/crossplane-aws-azure

Repository files navigation

crossplane-aws-azure

provision-aws provision-azure crossplane-version provider-aws-s3 provider-azure-storage License renovateenabled

Example project showing how to get started with Crossplane, connect it to multiple providers like AWS, Azure - and provision some resources like a S3Bucket or a StorageAccount through K8s CRDs

This repo is accompanied by this blog post https://blog.codecentric.de/en/2022/07/crossplane/

Also it is used as an example in these three iX Magazin issues (simply click on the magazine covers to access the full articles):

Crossplane https://crossplane.io/ claims to be the "The cloud native control plane framework". It introduces a new way how to manage any cloud resource (beeing it Kubernetes-native or not). It's an alternative Infrastructure-as-Code tooling to Terraform, AWS CDK/Bicep or Pulumi and introduces a higher level of abstraction - based on Kubernetes CRDs.

Literally the best intro post to Crossplane for me was https://blog.crossplane.io/crossplane-vs-cloud-infrastructure-addons/ - here the real key benefits especially compared to other tools are described. Without marketing blabla. If you love deep dives, I can also recommend Nate Reid's blog https://vrelevant.net/ who works as Staff Solutions Engineer at Upbound.

Crossplane basic concepts

https://docs.crossplane.io/latest/concepts/

  • Managed Resourced (MR): Kubernetes custom resources (CRDs) that represent infrastructure primitives (mostly in cloud providers). All Crossplane Managed Resources could be found via https://doc.crds.dev/
  • Composite Resources (XR): compose Managed Resources into higher level infrastructure units (especially interesting for platform teams). They are defined by:
    • a CompositeResourceDefinition (XRD) (which defines an OpenAPI schema the Composition needs to be conform to)
    • (optional) CompositeResourceClaims (XRC) (which is an abstraction of the XR for the application team to consume) - but is fantastic to hold the exact configuration parameters for the concrete resources you want to provision
    • a Composition that describes the actual infrastructure primitives aka Managed Resources used to build the Composite Resource. One XRD could have multiple Compositions - e.g. to one for every environment like development, stating and production
    • and configured by a Configuration
  • Packages: OCI container images to handle distribution, version updates, dependency management & permissions for Providers & Configurations. Packages were formerly named Stacks.
    • Providers: are Packages that bundle a set of Managed Resources & a Controller to provision infrastructure resources - all providers can be found on GitHub, e.g. provider-aws or on docs.crds.dev. A list of all available Providers can also be found on GitHub.
    • Configuration: define your own Composite Resources (XRs) & package them via kubectl crossplane build configuration (now they are a Package) - and push them to an OCI registry via kubectl crossplane push configuration. With this Configurations can also be easily installed into other Crossplane clusters.

composition-how-it-works

Getting started with Crossplane

In order to use Crossplane we'll need any kind of Kubernetes cluster to let it operate in. This management cluster with Crossplane installed will then provision the defined infrastructure. Using any managed Kubernetes cluster like EKS, AKS and so on is possible - or even a local Minikube, kind or k3d.

Install prerequisites & fire up a K8s cluster with kind

https://crossplane.io/docs/v1.8/getting-started/install-configure.html

Be sure to have kind, the package manager Helm and kubectl installed:

brew install kind helm kubectl

https://docs.crossplane.io/latest/cli/

Also we should install the crossplane CLI

curl -sL "https://raw.githubusercontent.com/crossplane/crossplane/master/install.sh" | sh
sudo mv crossplane /usr/local/bin

Now the kubectl crossplane --help command should be ready to use.

Now spin up a local kind cluster

kind create cluster --image kindest/node:v1.29.2 --wait 5m

Install Crossplane with Helm

Plain via Helm

The Crossplane docs tell us to use Helm for installation:

helm repo add crossplane-stable https://charts.crossplane.io/stable
helm repo update

helm upgrade --install crossplane --namespace crossplane-system --create-namespace crossplane-stable/crossplane

Using the appended --create-namespace, we don't need to explicitely create the namespace before running helm upgrade.

Renovate-powered installation via local Helm Chart

As an Renovate-powered alternative we can [create our own simple Chart.yaml to enable automatic updates](https://stackoverflow.com/a/71765472/4964553) of our installation if new crossplane versions get released:

apiVersion: v2
type: application
name: crossplane
version: 0.0.0 # unused
appVersion: 0.0.0 # unused
dependencies:
  - name: crossplane
    repository: https://charts.crossplane.io/stable
    version: 1.15.1

To install Crossplane using our own Chart.yaml simply run:

helm dependency update crossplane-install
helm upgrade --install crossplane --namespace crossplane-system crossplane-install --create-namespace

Be sure to exclude charts and Chart.lock files via .gitignore.

# Exclude Helm charts lock and packages
**/**/charts
**/**/Chart.lock

Check Crossplane installation

Check Crossplane version installed with helm list -n crossplane-system :

$ helm list -n crossplane-system
NAME      	NAMESPACE        	REVISION	UPDATED                              	STATUS  	CHART           	APP VERSION
crossplane	crossplane-system	1       	2022-06-21 09:28:21.178357 +0200 CEST	deployed	crossplane-1.8.1	1.8.1

Before we can actually apply a Provider we have to make sure that crossplane is actually healthy and running. Therefore we can use the kubectl wait command like this:

kubectl wait --for=condition=ready pod -l app=crossplane --namespace crossplane-system --timeout=120s

Otherwise we will run into errors like this when applying a Provider:

error: resource mapping not found for name: "provider-aws" namespace: "" from "provider-aws.yaml": no matches for kind "Provider" in version "pkg.crossplane.io/v1"
ensure CRDs are installed first

Finally check crossplane status with kubectl get all -n crossplane-system:

$ kubectl get all -n crossplane-system
NAME                                           READY   STATUS    RESTARTS   AGE
pod/crossplane-7c88c45998-d26wl                1/1     Running   0          69s
pod/crossplane-rbac-manager-8466dfb7fc-db9rb   1/1     Running   0          69s

NAME                                     TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
service/crossplane-webhooks              ClusterIP   10.96.160.156   <none>        9443/TCP   4d

NAME                                      READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/crossplane                1/1     1            1           69s
deployment.apps/crossplane-rbac-manager   1/1     1            1           69s

NAME                                                 DESIRED   CURRENT   READY   AGE
replicaset.apps/crossplane-7c88c45998                1         1         1       69s
replicaset.apps/crossplane-rbac-manager-8466dfb7fc   1         1         1       69s

"The base Crossplane installation consists of two pods, the crossplane pod and the crossplane-rbac-manager pod. Before starting the core Crossplane container an init container runs. The init container installs the core Crossplane Custom Resource Definitions (CRDs), configures Crossplane webhooks and installs any supplied Providers or Configurations."" See https://docs.crossplane.io/latest/concepts/pods/

Now we should be able to find some new Kubernetes API objects:

$ kubectl api-resources  | grep crossplane
compositeresourcedefinitions               xrd,xrds     apiextensions.crossplane.io/v1         false        CompositeResourceDefinition
compositionrevisions                       comprev      apiextensions.crossplane.io/v1         false        CompositionRevision
compositions                               comp         apiextensions.crossplane.io/v1         false        Composition
environmentconfigs                         envcfg       apiextensions.crossplane.io/v1alpha1   false        EnvironmentConfig
usages                                                  apiextensions.crossplane.io/v1alpha1   false        Usage
configurationrevisions                                  pkg.crossplane.io/v1                   false        ConfigurationRevision
configurations                                          pkg.crossplane.io/v1                   false        Configuration
controllerconfigs                                       pkg.crossplane.io/v1alpha1             false        ControllerConfig
deploymentruntimeconfigs                                pkg.crossplane.io/v1beta1              false        DeploymentRuntimeConfig
functionrevisions                                       pkg.crossplane.io/v1beta1              false        FunctionRevision
functions                                               pkg.crossplane.io/v1beta1              false        Function
locks                                                   pkg.crossplane.io/v1beta1              false        Lock
providerrevisions                                       pkg.crossplane.io/v1                   false        ProviderRevision
providers                                               pkg.crossplane.io/v1                   false        Provider
storeconfigs                                            secrets.crossplane.io/v1alpha1         false        StoreConfig

Configure Crossplane to access AWS

https://docs.crossplane.io/latest/getting-started/provider-aws/

Choosing a Crossplane AWS Provider

Currently there are two crossplane providers for AWS:

The first is written "by hand" and supports around 200 Managed Resources and the other is generated using Upjet supporting the full AWS API with 924 Managed Resources.

The second provider, which is also called the Upbound "official" provider has been donated to the OpenSource community in September 2023 - so there are now 2 OpenSource providers for AWS.

Which one should you choose? This issue clarifies it:

"Upbound (the company behind crossplane) has moved to its own Terraform-based set of providers. This means that https://github.com/crossplane-contrib/provider-aws is now only maintained by community volunteers since there is a number of people out there who are still using it. It is kind of legacy but it will receive further updates as long as the community is willing to contribute to it.

Until June 2023 the Upbound providers had a problem. As Managed Resources are CRDs, the Upbound provider introduced a problem to most control planes, since Kubernetes wasn't designed for that amount of CRDs. I gave a talk about crossplane at our local DevOps Meetup and already struggled with the Upbound AWS provider.

Therefore in June 2023 the Upjet generated official Upbound provider has been split into Provider Families: https://blog.crossplane.io/crd-scaling-provider-families/ These split up Upbound providers are the way to go if you start using crossplane today:

The migration guide also states that the monolithic provider is already deprecated:

"Warning: The monolithic AWS provider (upbound/provider-aws) has been deprecated in favor of the AWS provider family. You can read more about the provider families in our blog post and the official documentation for the provider families is here. We will continue support for the monolithic AWS provider until June 12, 2024. https://marketplace.upbound.io/providers/upbound/provider-aws/"

So head over to the Upbound marketplace and search for your provider familiy. As we want to use AWS S3 for example, here we went with https://marketplace.upbound.io/providers/upbound/provider-aws-s3.

Using the official Upbound (Upjet generated) AWS Provider

https://marketplace.upbound.io/providers/upbound/provider-aws-s3

Now we can install the Upbound AWS Provider. Therefore the upbound/provider-aws-s3/config/provider-aws-s3.yaml looks only slightly different compared to the classic AWS Provider - only the spec.package changed:

apiVersion: pkg.crossplane.io/v1
kind: Provider
metadata:
  name: upbound-provider-aws-s3
spec:
  package: xpkg.upbound.io/upbound/provider-aws-s3:v0.46.0
  packagePullPolicy: Always
  revisionActivationPolicy: Automatic
  revisionHistoryLimit: 1

Install it via kubectl:

kubectl apply -f upbound/provider-aws-s3/config/provider-aws-s3.yaml

The package version in combination with the packagePullPolicy configuration here is crucial, since we can configure an update strategy for the Provider here. I'am not sure, if the Crossplane team will provide an installation method where we can use tools like Renovate to keep our Crossplane providers up to date (now Renovate supports Crossplane, see paragraph Provider & Configuration Package Upgrades with Renovate). A full table of all possible fields can be found in the docs: https://crossplane.io/docs/v1.8/concepts/packages.html#specpackagepullpolicy We can also let crossplane itself manage new versions for us. But we also leave the GitOps way here!

Multiple Provider Package revisions can also be installed at the same time (especially when using packagePullPolicy: Always). If you installed multiple package versions, you'll see them as providerrevision.pkg.x when running kubectl get crossplane:

$ kubectl get crossplane
...
NAME                                                                             HEALTHY   REVISION   IMAGE                                                   STATE      DEP-FOUND   DEP-INSTALLED   AGE
providerrevision.pkg.crossplane.io/upbound-provider-aws-s3-4c95b368de88          True      1          xpkg.upbound.io/upbound/provider-aws-s3:v1.2.1          Active     1           1               3d23h
providerrevision.pkg.crossplane.io/upbound-provider-aws-s3-d6a6663caff3          True      1          xpkg.upbound.io/upbound/provider-aws-s3:v1.1.0          Inactive   1           1               5d22h
...

As our first Crossplane Provider has been installed. We need to wait for the Provider to become healthy:

kubectl wait "providers.pkg.crossplane.io/upbound-provider-aws-s3" --for=condition=Healthy --timeout=180s

Obtain the status via kubectl get provider.pkg.crossplane.io:

$ kubectl get provider.pkg.crossplane.io
NAME                          INSTALLED   HEALTHY   PACKAGE                                               AGE
upbound-provider-aws-s3       True        True      xpkg.upbound.io/upbound/provider-aws-s3:v0.46.0       113s
upbound-provider-family-aws   True        True      xpkg.upbound.io/upbound/provider-family-aws:v0.46.0   108s

Otherwise we may run into errors like this when applying the ProviderConfig right after the Provider.

Create aws-creds.conf file

I assume here that you have aws CLI installed and configured. So that the command aws configure should work on your system.

With this prepared we can create an aws-creds.conf file:

echo "[default]
aws_access_key_id = $(aws configure get aws_access_key_id)
aws_secret_access_key = $(aws configure get aws_secret_access_key)
" > aws-creds.conf

Don't ever check this file into source control - it holds your AWS credentials! Add it to your .gitignore file right now:

# Exclude credential configuration files like aws-creds.conf
*-creds.conf

If you're using a CI system like GitHub Actions (as this repository is based on), you need to have both AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY configured as Repository Secrets:

github-actions-secrets

Also make sure to have your Default region configured locally - or as a env: variable in your CI system. All three needed variables in GitHub Actions for example look like this:

env:
  AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
  AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
  AWS_DEFAULT_REGION: 'eu-central-1'

Create AWS Provider secret

Now we need to use the aws-creds.conf file to create the Crossplane AWS Provider secret:

kubectl create secret generic aws-creds -n crossplane-system --from-file=creds=./aws-creds.conf

If everything went well there should be a new aws-creds Secret ready:

provider-aws-secret

Create ProviderConfig to consume the Secret containing AWS credentials

To get our Provider finally working we also need to create a ProviderConfig accordingly that will tell the Provider where to find it's AWS credentials. Therefore we create a upbound/provider-aws-s3/config/provider-config-aws.yaml:

apiVersion: aws.upbound.io/v1beta1
kind: ProviderConfig
metadata:
  name: default
spec:
  credentials:
    source: Secret
    secretRef:
      namespace: crossplane-system
      name: aws-creds
      key: creds

Crossplane resources use the ProviderConfig named default if no specific ProviderConfig is specified, so this ProviderConfig will be the default for all AWS resources.

The secretRef.name and secretRef.key has to match the fields of the already created Secret.

Apply it via kubectl:

kubectl apply -f upbound/provider-aws-s3/config/provider-config-aws.yaml

Now we should have everything in place to use the Upbound AWS Provider! We can double check via kubectl get crossplane:

$ kubectl get crossplane
NAME                                    AGE
providerconfig.aws.upbound.io/default   55s

NAME                                                                             HEALTHY   REVISION   IMAGE                                                   STATE      DEP-FOUND   DEP-INSTALLED   AGE
providerrevision.pkg.crossplane.io/upbound-provider-aws-s3-4c95b368de88          True      1          xpkg.upbound.io/upbound/provider-aws-s3:v1.2.1          Active     1           1               3d23h

NAME                                                        INSTALLED   HEALTHY   PACKAGE                                                 AGE
provider.pkg.crossplane.io/upbound-provider-aws-s3          True        True      xpkg.upbound.io/upbound/provider-aws-s3:v1.2.1          3d23h

NAME                                        AGE   TYPE         DEFAULT-SCOPE
storeconfig.secrets.crossplane.io/default   4d    Kubernetes   crossplane-system

Now we should have some more Kubernetes API resources available:

$ kubectl api-resources | grep aws
providerconfigs                                         aws.upbound.io/v1beta1                 false        ProviderConfig
providerconfigusages                                    aws.upbound.io/v1beta1                 false        ProviderConfigUsage
storeconfigs                                            aws.upbound.io/v1alpha1                false        StoreConfig
bucketaccelerateconfigurations                          s3.aws.upbound.io/v1beta1              false        BucketAccelerateConfiguration
bucketacls                                              s3.aws.upbound.io/v1beta1              false        BucketACL
bucketanalyticsconfigurations                           s3.aws.upbound.io/v1beta1              false        BucketAnalyticsConfiguration
bucketcorsconfigurations                                s3.aws.upbound.io/v1beta1              false        BucketCorsConfiguration
bucketintelligenttieringconfigurations                  s3.aws.upbound.io/v1beta1              false        BucketIntelligentTieringConfiguration
bucketinventories                                       s3.aws.upbound.io/v1beta1              false        BucketInventory
bucketlifecycleconfigurations                           s3.aws.upbound.io/v1beta1              false        BucketLifecycleConfiguration
bucketloggings                                          s3.aws.upbound.io/v1beta1              false        BucketLogging
bucketmetrics                                           s3.aws.upbound.io/v1beta1              false        BucketMetric
bucketnotifications                                     s3.aws.upbound.io/v1beta1              false        BucketNotification
bucketobjectlockconfigurations                          s3.aws.upbound.io/v1beta1              false        BucketObjectLockConfiguration
bucketobjects                                           s3.aws.upbound.io/v1beta1              false        BucketObject
bucketownershipcontrols                                 s3.aws.upbound.io/v1beta1              false        BucketOwnershipControls
bucketpolicies                                          s3.aws.upbound.io/v1beta1              false        BucketPolicy
bucketpublicaccessblocks                                s3.aws.upbound.io/v1beta1              false        BucketPublicAccessBlock
bucketreplicationconfigurations                         s3.aws.upbound.io/v1beta1              false        BucketReplicationConfiguration
bucketrequestpaymentconfigurations                      s3.aws.upbound.io/v1beta1              false        BucketRequestPaymentConfiguration
buckets                                                 s3.aws.upbound.io/v1beta1              false        Bucket
bucketserversideencryptionconfigurations                s3.aws.upbound.io/v1beta1              false        BucketServerSideEncryptionConfiguration
bucketversionings                                       s3.aws.upbound.io/v1beta1              false        BucketVersioning
bucketwebsiteconfigurations                             s3.aws.upbound.io/v1beta1              false        BucketWebsiteConfiguration
objectcopies                                            s3.aws.upbound.io/v1beta1              false        ObjectCopy
objects                                                 s3.aws.upbound.io/v1beta1              false        Object

Provision a S3 Bucket with Crossplane

The crossplane core Controller and the Provider AWS Controller should now be ready to provision any infrastructure component in AWS!

You heard right: we don't create a Kubernetes based infrastructure component - but we start with a simple S3 Bucket.

Therefore we can have a look into the Crossplane AWS provider API docs:https://doc.crds.dev/github.com/upbound/provider-aws/s3.aws.upbound.io/Bucket/v1beta1@v0.45.0

CompositeResourceDefinitions (or XRDs)

https://docs.crossplane.io/latest/concepts/composite-resource-definitions/

A CompositeResourceDefinition (or XRD) defines the type and schema of your XR. It lets Crossplane know that you want a particular kind of XR to exist, and what fields that XR should have.

Since defining your own CompositeResourceDefinitions and Compositions is the main work todo with Crossplane, it's always good to know the full Reference documentation which can be found here https://docs.crossplane.io/latest/concepts/compositions/

One of the things to know is that Crossplane automatically injects some common 'machinery' into the manifests of the XRDs and Compositions: https://docs.crossplane.io/latest/concepts/composite-resources/

Defining a CompositeResourceDefinition (XRD) for our S3 Bucket

All possible fields an XRD can have are documented in the docs.

The field spec.versions.schema must contain a OpenAPI schema, which is similar to the ones used by any Kubernetes CRDs. They determine what fields the XR (and Claim) will have. The full CRD documentation and a guide on how to write OpenAPI schemas could be found in the Kubernetes docs.

Note that Crossplane will be automatically extended this section. Therefore the following fields are used by Crossplane and will be ignored if they're found in the schema:

spec.resourceRef
spec.resourceRefs
spec.claimRef
spec.writeConnectionSecretToRef
status.conditions
status.connectionDetails

So our Composite Resource Definition (XRD) for our S3 Bucket could look like crossplane-contrib/provider-aws/s3/definition.yaml:

apiVersion: apiextensions.crossplane.io/v1
kind: CompositeResourceDefinition
metadata:
  # XRDs must be named 'x<plural>.<group>'
  name: xobjectstorages.crossplane.jonashackt.io
spec:
  # This XRD defines an XR in the 'crossplane.jonashackt.io' API group.
  # The XR or Claim must use this group together with the spec.versions[0].name as it's apiVersion, like this:
  # 'crossplane.jonashackt.io/v1alpha1'
  group: crossplane.jonashackt.io
  
  # XR names should always be prefixed with an 'X'
  names:
    kind: XObjectStorage
    plural: xobjectstorages
  # This type of XR offers a claim, which should have the same name without the 'X' prefix
  claimNames:
    kind: ObjectStorage
    plural: objectstorages
  
  # default Composition when none is specified (must match metadata.name of a provided Composition)
  # e.g. in composition.yaml
  defaultCompositionRef:
    name: objectstorage-composition

  versions:
  - name: v1alpha1
    served: true
    referenceable: true
    # OpenAPI schema (like the one used by Kubernetes CRDs). Determines what fields
    # the XR (and claim) will have. Will be automatically extended by crossplane.
    # See https://kubernetes.io/docs/tasks/extend-kubernetes/custom-resources/custom-resource-definitions/
    # for full CRD documentation and guide on how to write OpenAPI schemas
    schema:
      openAPIV3Schema:
        type: object
        properties:
          spec:
            type: object
            # We define 2 needed parameters here one has to provide as XR or Claim spec.parameters
            properties:
              parameters:
                type: object
                properties:
                  bucketName:
                    type: string
                  region:
                    type: string
                required:
                  - bucketName
                  - region

Install the XRD into our cluster with:

kubectl apply -f crossplane-contrib/provider-aws/s3/definition.yaml

We can double check the CRDs beeing created with kubectl get crds and filter them using grep to our group name crossplane.jonashackt.io:

$ kubectl get crds | grep crossplane.jonashackt.io
objectstorages.crossplane.jonashackt.io                         2022-06-27T09:54:18Z
xobjectstorages.crossplane.jonashackt.io                        2022-06-27T09:54:18Z

Craft a Composition to manage our needed cloud resources

The main work in Crossplane has to be done crafting the Compositions. This is because they interact with the infrastructure primitives the cloud provider APIs provide.

Detailled docs to many of the possible manifest configurations can be found here https://docs.crossplane.io/latest/concepts/compositions/

aws/aws-cdk#25288 (comment)

https://doc.crds.dev/github.com/crossplane/provider-aws/s3.aws.crossplane.io/Bucket/v1beta1@v0.39.0

https://docs.aws.amazon.com/AmazonS3/latest/userguide/access-policy-language-overview.html

https://stackoverflow.com/questions/76097031/aws-s3-bucket-cannot-have-acls-set-with-objectownerships-bucketownerenforced-s

https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/BucketPublicAccessBlock/v1beta1

According to hashicorp/terraform-provider-aws#28353 and https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/s3_bucket_acl we need to separate Bucket creation from BucketPublicAccessBlock, BucketOwnershipControls and BucketACL - which should now be available finally leveraging the Upbound AWS Provider:

apiVersion: apiextensions.crossplane.io/v1
kind: Composition
metadata:
  name: objectstorage-composition
  labels:
    # An optional convention is to include a label of the XRD. This allows
    # easy discovery of compatible Compositions.
    crossplane.io/xrd: xobjectstorages.crossplane.jonashackt.io
    # The following label marks this Composition for AWS. This label can 
    # be used in 'compositionSelector' in an XR or Claim.
    provider: aws
spec:
  # Each Composition must declare that it is compatible with a particular type
  # of Composite Resource using its 'compositeTypeRef' field. The referenced
  # version must be marked 'referenceable' in the XRD that defines the XR.
  compositeTypeRef:
    apiVersion: crossplane.jonashackt.io/v1alpha1
    kind: XObjectStorage
  
  # When an XR is created in response to a claim Crossplane needs to know where
  # it should create the XR's connection secret. This is configured using the
  # 'writeConnectionSecretsToNamespace' field.
  writeConnectionSecretsToNamespace: crossplane-system
  
  # Each Composition must specify at least one composed resource template.
  resources:
    # Providing a unique name for each entry is good practice.
    # Only identifies the resources entry within the Composition. Required in future crossplane API versions.
    - name: bucket
      base:
        # see https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/Bucket/v1beta1
        apiVersion: s3.aws.upbound.io/v1beta1
        kind: Bucket
        metadata: {}
        spec:
          deletionPolicy: Delete
      
      patches:
        - fromFieldPath: "spec.parameters.bucketName"
          toFieldPath: "metadata.name"
        - fromFieldPath: "spec.parameters.region"
          toFieldPath: "spec.forProvider.region"

    - name: bucketpublicaccessblock
      base:
        # see https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/BucketPublicAccessBlock/v1beta1
        apiVersion: s3.aws.upbound.io/v1beta1
        kind: BucketPublicAccessBlock
        spec:
          forProvider:
            blockPublicAcls: false
            blockPublicPolicy: false
            ignorePublicAcls: false
            restrictPublicBuckets: false

      patches:
        - fromFieldPath: "spec.parameters.bucketPABName"
          toFieldPath: "metadata.name"
        - fromFieldPath: "spec.parameters.bucketName"
          toFieldPath: "spec.forProvider.bucketRef.name"
        - fromFieldPath: "spec.parameters.region"
          toFieldPath: "spec.forProvider.region"

    - name: bucketownershipcontrols
      base:
        # see https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/BucketOwnershipControls/v1beta1#doc:spec-forProvider-rule-objectOwnership
        apiVersion: s3.aws.upbound.io/v1beta1
        kind: BucketOwnershipControls
        spec:
          forProvider:
            rule:
              - objectOwnership: ObjectWriter

      patches:
        - fromFieldPath: "spec.parameters.bucketOSCName"
          toFieldPath: "metadata.name"
        - fromFieldPath: "spec.parameters.bucketName"
          toFieldPath: "spec.forProvider.bucketRef.name"
        - fromFieldPath: "spec.parameters.region"
          toFieldPath: "spec.forProvider.region"

    - name: bucketacl
      base:
        # see https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/BucketACL/v1beta1
        apiVersion: s3.aws.upbound.io/v1beta1
        kind: BucketACL
        spec:
          forProvider:
            acl: "public-read"

      patches:
        - fromFieldPath: "spec.parameters.bucketAclName"
          toFieldPath: "metadata.name"
        - fromFieldPath: "spec.parameters.bucketName"
          toFieldPath: "spec.forProvider.bucketRef.name"
        - fromFieldPath: "spec.parameters.region"
          toFieldPath: "spec.forProvider.region"
  
    - name: bucketwebsiteconfiguration
      base:
        # see https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/resources/s3.aws.upbound.io/BucketWebsiteConfiguration/v1beta1
        apiVersion: s3.aws.upbound.io/v1beta1
        kind: BucketWebsiteConfiguration
        spec:
          forProvider:
            indexDocument:
              - suffix: index.html

      patches:
        - fromFieldPath: "spec.parameters.bucketWebConfName"
          toFieldPath: "metadata.name"
        - fromFieldPath: "spec.parameters.bucketName"
          toFieldPath: "spec.forProvider.bucketRef.name"
        - fromFieldPath: "spec.parameters.region"
          toFieldPath: "spec.forProvider.region"

  # If you find yourself repeating patches a lot you can group them as a named
  # 'patch set' then use a PatchSet type patch to reference them.
  #patchSets:

Let's create Composition via:

kubectl apply -f upbound/provider-aws-s3/composition.yaml

Craft a Composite Resource (XR) or Claim (XRC)

Crossplane could look quite intimidating when having a first look. There are few guides around to show how to approach a setup when using Crossplane the first time. You can choose between writing an XR OR XRC! You don't need both, since the XR will be generated from the XRC, if you choose to craft a XRC.

https://docs.crossplane.io/v1.14/concepts/composite-resources/

Since we want to create a S3 Bucket, here's an suggestion for an claim.yaml:

# Use the spec.group/spec.versions[0].name defined in the XRD
apiVersion: crossplane.jonashackt.io/v1alpha1
kind: ObjectStorage
metadata:
  # Only claims are namespaced, unlike XRs.
  namespace: default
  name: managed-s3
spec:
  # The compositionRef specifies which Composition this XR will use to compose
  # resources when it is created, updated, or deleted.
  compositionRef:
    name: objectstorage-composition
  
  # Parameters for the Composition to provide the Managed Resources (MR) with
  # to create the actual infrastructure components
  parameters:
    bucketName: microservice-ui-nuxt-js-static-bucket2
    region: eu-central-1

Testdrive with:

kubectl apply -f crossplane-contrib/provider-aws/s3/claim.yaml

When somthing goes wrong with the validation, this could look like this:

$ kubectl apply -f claim.yaml
error: error validating "claim.yaml": error validating data: [ValidationError(S3Bucket.metadata): unknown field "crossplane.io/external-name" in io.k8s.apimachinery.pkg.apis.meta.v1.ObjectMeta_v2, ValidationError(S3Bucket.spec): unknown field "parameters" in io.jonashackt.crossplane.v1alpha1.S3Bucket.spec, ValidationError(S3Bucket.spec.writeConnectionSecretToRef): missing required field "namespace" in io.jonashackt.crossplane.v1alpha1.S3Bucket.spec.writeConnectionSecretToRef, ValidationError(S3Bucket.spec): missing required field "bucketName" in io.jonashackt.crossplane.v1alpha1.S3Bucket.spec, ValidationError(S3Bucket.spec): missing required field "region" in io.jonashackt.crossplane.v1alpha1.S3Bucket.spec]; if you choose to ignore these errors, turn validation off with --validate=false

The Crossplane validation is a great way to debug your yaml configuration - it hints you to the actual problems that are still present.

Troubleshooting your crossplane configuration

https://docs.crossplane.io/knowledge-base/guides/troubleshoot/

Per Kubernetes convention, Crossplane keeps errors close to the place they happen. This means that if your claim is not becoming ready due to an issue with your Composition or with a composed resource you’ll need to “follow the references” to find out why. Your claim will only tell you that the XR is not yet ready.

The docs also tell us what they mean by "follow the references":

  • Find your XR by running kubectl describe <claim-kind> <claim-metadata.name> and look for its “Resource Ref” (aka spec.resourceRef).
  • Run kubectl describe on your XR. This is where you’ll find out about issues with the Composition you’re using, if any.
  • If there are no issues but your XR doesn’t seem to be becoming ready, take a look for the “Resource Refs” (or spec.resourceRefs) to find your composed resources.
  • Run kubectl describe on each referenced composed resource to determine whether it is ready and what issues, if any, it is encountering.

Waiting for resources to become ready

There are some possible things to check while your resources (may) get deployed after running a kubectl apply -f claim.yaml (see https://crossplane.io/docs/v1.8/getting-started/provision-infrastructure.html#claim-your-infrastructure).

The best overview gives a kubectl get crossplane which will simply list all the crossplane resources:

$ kubectl get crossplane
Warning: Please use v1beta1 version of this resource that has identical schema.
NAME                                                                                          ESTABLISHED   OFFERED   AGE
compositeresourcedefinition.apiextensions.crossplane.io/xs3buckets.crossplane.jonashackt.io   True          True      23m

NAME                                               AGE
composition.apiextensions.crossplane.io/s3bucket   2d17h

NAME                                      INSTALLED   HEALTHY   PACKAGE                           AGE
provider.pkg.crossplane.io/provider-aws   True        True      crossplanecontrib/provider-aws:v0.34.0   4d21h

NAME                                                           HEALTHY   REVISION   IMAGE                             STATE    DEP-FOUND   DEP-INSTALLED   AGE
providerrevision.pkg.crossplane.io/provider-aws-2189bc61e0bd   True      1          crossplanecontrib/provider-aws:v0.34.0   Active                               4d21h

NAME                                        AGE     TYPE         DEFAULT-SCOPE
storeconfig.secrets.crossplane.io/default   5d23h   Kubernetes   crossplane-system
  • kubectl get claim: get all resources of all claim kinds, like PostgreSQLInstance.
  • kubectl get composite: get all resources that are of composite kind, like XPostgreSQLInstance.
  • kubectl get managed: get all resources that represent a unit of external infrastructure.
  • kubectl get <name-of-provider>: get all resources related to .
  • kubectl get crossplane: get all resources related to Crossplane.

We can also check our claim with kubectl get <claim-kind> <claim-metadata.name> like this:

$ kubectl get ObjectStorage managed-s3
NAME         READY   CONNECTION-SECRET               AGE
managed-s3           managed-s3-connection-details   5s

To watch the provisioned resources become ready we can run kubectl get crossplane -l crossplane.io/claim-name=<claim-metadata.name>:

kubectl get crossplane -l crossplane.io/claim-name=managed-s3

Check if the S3 Bucket has been created successfully via aws CLI with aws s3 ls | grep microservice-ui-nuxt-js-static-bucket2.

$ aws s3 ls | grep microservice-ui-nuxt-js-static-bucket2
2022-06-27 11:56:26 microservice-ui-nuxt-js-static-bucket2

Our bucket should be there! We can also double check in the AWS console:

aws-console-s3-bucket-created

Deploy our app

Let's deploy our app (a simple index.html) to our S3 Bucket using the aws CLI like this:

aws s3 sync static s3://devopsthde-bucket --acl public-read

Now we can open up http://microservice-ui-nuxt-js-static-bucket2.s3-website.eu-central-1.amazonaws.com/ in our Browser and should see our app beeing deployed:

s3-static-webseite-deployed

Before removing the claim, we should remove our index.html - otherwise we'll run into errors like this:

  Warning  CannotDeleteExternalResource  37s (x16 over 57s)  managed/bucket.s3.aws.crossplane.io  (combined from similar events): operation error S3: DeleteBucket, https response error StatusCode: 409, RequestID: 0WHR906YZRF0YDSH, HostID: x7cz2iYF/8Ag2wKtKRZUy1j3hPk67tBUOTFeR//+grrD7plqQ5Zo6EecO70KOOgHKbY7hUyp9vU=, api error BucketNotEmpty: The bucket you tried to delete is not empty

So first delete the index.html:

aws s3 rm s3://devopsthde-bucket/index.html

Finally remove our S3 Bucket again with

kubectl delete -f upbound/provider-aws-s3/claim.yaml

Now also the S3 Bucket should be removed by crossplane.

Configure Crossplane to access Azure

https://docs.crossplane.io/v1.14/getting-started/provider-azure/

https://marketplace.upbound.io/providers/upbound/provider-azure-storage/v0.39.0

Create crossplane-azure-provider-key.json

https://docs.crossplane.io/latest/getting-started/provider-azure/#create-a-kubernetes-secret-for-azure

I assume here that you have azure CLI installed and you're logged in to your Azure subscription via az login. So that the command az account show should work on your system.

The current Crossplane docs on Azure propose a az ad sp create-for-rbac command that isn't working: ERROR: Usage error: To create role assignments, specify both --role and --scopes. When using --role we must use --scopes also. The documentation about the latter isn't easy to grasp https://docs.microsoft.com/en-us/cli/azure/ad/sp?view=azure-cli-latest#az-ad-sp-create-for-rbac at first sight. As there are examples like --scopes /subscriptions/{subscriptionId}/resourceGroups/{resourceGroup1} I thought I need to create a resourceGroup before even connecting to Azure to create a resourceGroup with Crossplane (that doesn't make any sense). BUT: It's possible to use --scopes /subscriptions/{subscriptionId} here without any resourceGroups, which will create a scope for your whole subscription.

So prepare the SUBSCRIPTION_ID variable like this:

export SUBSCRIPTION_ID=$(az account show --query id --output tsv)

With this prepared we can create a service principal in Azure AD using the following command:

az ad sp create-for-rbac --sdk-auth --role Owner --scopes /subscriptions/$SUBSCRIPTION_ID --name servicePrincipalCrossplaneGHActions > crossplane-azure-provider-key.json

Although this produces a waring message like WARNING: Option '--sdk-auth' has been deprecated and will be removed in a future release. we definitely need to use the --sdk-auth parameter. Otherwise our Managed Resources will run into errors like connect failed: cannot get authorizer from client credentials config: failed to get SPT from client credentials: parameter 'activeDirectoryEndpoint' cannot be empty because there are configuration entries missing in the file like "activeDirectoryEndpointUrl": "https://login.microsoftonline.com",.

This produces a crossplane-azure-provider-key.json file you should never ever check into version control! For this repository I added it to the .gitignore file.

If you're using a CI system like GitHub Actions (as this repository is based on), define 3 repository secrets from the output. Choose the appId as the ARM_CLIENT_ID, the password as the ARM_CLIENT_SECRET and the tenant as the ARM_TENANT_ID. Additionally we need to define the ARM_SUBSCRIPTION_ID secret. Therefore run a az account show (after you logged your CLI into your Azure subscription via azure login) and use the value of "id":.

github-actions-secrets-azure

On GitHub Actions we need to

echo "{
  \"clientId\": \"$ARM_CLIENT_ID\",
  \"clientSecret\": \"$ARM_CLIENT_SECRET\",
  \"subscriptionId\": \"$ARM_SUBSCRIPTION_ID\",
  \"tenantId\": \"$ARM_TENANT_ID\",
  \"activeDirectoryEndpointUrl\": \"https://login.microsoftonline.com\",
  \"resourceManagerEndpointUrl\": \"https://management.azure.com/\",
  \"activeDirectoryGraphResourceId\": \"https://graph.windows.net/\",
  \"sqlManagementEndpointUrl\": \"https://management.core.windows.net:8443/\",
  \"galleryEndpointUrl\": \"https://gallery.azure.com/\",
  \"managementEndpointUrl\": \"https://management.core.windows.net/\"
}" > crossplane-azure-provider-key.json

All three needed variables in GitHub Actions for example look like this:

env:
  ARM_CLIENT_ID: ${{ secrets.ARM_CLIENT_ID }}
  ARM_CLIENT_SECRET: ${{ secrets.ARM_CLIENT_SECRET }}
  ARM_SUBSCRIPTION_ID: ${{ secrets.ARM_SUBSCRIPTION_ID }}
  ARM_TENANT_ID: ${{ secrets.ARM_TENANT_ID }}

Create Azure Provider secret

Now we need to use the crossplane-azure-provider-key.json file to create the Crossplane Azure Provider secret:

kubectl create secret generic azure-account-creds -n crossplane-system --from-file=creds=./crossplane-azure-provider-key.json

If everything went well there should be a new azure-account-creds Secret ready.

Install the Crossplane Azure Provider

https://docs.crossplane.io/latest/getting-started/provider-azure/#install-the-azure-provider

Let's create a provider-azure-storage.yaml file like this:

apiVersion: pkg.crossplane.io/v1
kind: Provider
metadata:
  name: provider-azure-storage
spec:
  package: xpkg.upbound.io/upbound/provider-azure-storage:v0.39.0
  packagePullPolicy: Always
  revisionActivationPolicy: Automatic
  revisionHistoryLimit: 1

Install the Azure provider:

kubectl apply -f upbound/provider-azure-storage/config/provider-azure-storage.yaml

Now our first Crossplane Provider has been installed. You may check it with kubectl get provider:

$ kubectl get provider
NAME             INSTALLED   HEALTHY   PACKAGE                            AGE
provider-azure   True        Unknown   crossplane/provider-azure:v0.19.0  13s

Before we can actually apply a ProviderConfig to our Azure provider we have to make sure that it's actually healthy and running. Therefore we can use the kubectl wait command like this:

kubectl wait --for=condition=healthy --timeout=120s provider/provider-azure

Otherwise we may run into errors like this when applying the ProviderConfig right after the Provider.

Create ProviderConfig to consume the Secret containing Azure credentials

https://docs.crossplane.io/latest/getting-started/provider-azure/#create-a-providerconfig

Now we need to create ProviderConfig object that will tell the AWS Provider where to find it's AWS credentials. Therefore we create a provider-config-azure.yaml:

apiVersion: azure.upbound.io/v1beta1
kind: ProviderConfig
metadata:
  name: default
spec:
  credentials:
    source: Secret
    secretRef:
      namespace: crossplane-system
      name: azure-account-creds
      key: creds

Crossplane resources use the ProviderConfig named default if no specific ProviderConfig is specified, so this ProviderConfig will be the default for all Azure resources.

The secretRef.name and secretRef.key has to match the fields of the already created Secret.

Apply it with:

kubectl apply -f upbound/provider-azure-storage/config/provider-config-azure.yaml

The crossplane core Controller and the Provider Azure Controller should now be ready to provision any infrastructure component in Azure!

Provision a StorageAccount in Azure with Crossplane

https://doc.crds.dev/github.com/upbound/provider-azure/storage.azure.upbound.io/Account/v1beta1@v0.38.2

https://doc.crds.dev/github.com/upbound/provider-azure/azure.upbound.io/ResourceGroup/v1beta1@v0.38.2

Defining a CompositeResourceDefinition (XRD) for our Storage Account

So our Composite Resource Definition (XRD) for our Storage Account could look like upbound/provider-azure-storage/definition.yaml:

---
apiVersion: apiextensions.crossplane.io/v1
kind: CompositeResourceDefinition
metadata:
  name: xstoragesazure.crossplane.jonashackt.io
spec:
  group: crossplane.jonashackt.io
  names:
    kind: XStorageAzure
    plural: xstoragesazure
  claimNames:
    kind: StorageAzure
    plural: storagesazure
  
  defaultCompositionRef:
    name: storageazure-composition

  versions:
  - name: v1alpha1
    served: true
    referenceable: true
    # See https://kubernetes.io/docs/tasks/extend-kubernetes/custom-resources/custom-resource-definitions/
    schema:
      openAPIV3Schema:
        type: object
        properties:
          spec:
            type: object
            properties:
              parameters:
                type: object
                properties:
                  location:
                    type: string
                  resourceGroupName:
                    type: string
                  storageAccountName:
                    type: string
                required:
                  - location
                  - resourceGroupName
                  - storageAccountName

Install the XRD into our cluster with:

kubectl apply -f upbound/provider-azure-storage/definition.yaml

Let's wait for the XRD to become Offered:

kubectl wait --for=condition=Offered --timeout=120s xrd xstoragesazure.crossplane.jonashackt.io  

Craft a Composition to manage our needed cloud resources

A Composite to manage an Storage Account in Azure with public access for static website hosting could for example look like this upbound/provider-azure-storage/composition.yaml:

---
apiVersion: apiextensions.crossplane.io/v1
kind: Composition
metadata:
  name: storageazure-composition
  labels:
    crossplane.io/xrd: xstoragesazure.crossplane.jonashackt.io
    provider: azure
spec:
  compositeTypeRef:
    apiVersion: crossplane.jonashackt.io/v1alpha1
    kind: XStorageAzure
  
  writeConnectionSecretsToNamespace: crossplane-system
  
  resources:
    - name: storageaccount
      base:
        # see https://doc.crds.dev/github.com/crossplane/provider-azure/storage.azure.crossplane.io/Account/v1alpha3@v0.19.0
        apiVersion: storage.azure.upbound.io/v1beta1
        kind: Account
        metadata: {}
        spec:
          forProvider:
            accountKind: StorageV2
            accountTier: Standard
            accountReplicationType: LRS
      patches:
        - fromFieldPath: spec.parameters.storageAccountName
          toFieldPath: metadata.name
        - fromFieldPath: spec.parameters.resourceGroupName
          toFieldPath: spec.forProvider.resourceGroupName
        - fromFieldPath: spec.parameters.location
          toFieldPath: spec.forProvider.location
          
    - name: resourcegroup
      base:
        # see https://doc.crds.dev/github.com/crossplane/provider-azure/azure.crossplane.io/ResourceGroup/v1alpha3@v0.19.0
        apiVersion: azure.upbound.io/v1beta1
        kind: ResourceGroup
        metadata: {}
      patches:
        - fromFieldPath: spec.parameters.resourceGroupName
          toFieldPath: metadata.name
        - fromFieldPath: spec.parameters.location
          toFieldPath: spec.forProvider.location

Install our Composition with

kubectl apply -f upbound/provider-azure-storage/composition.yaml

Craft a Composite Resource (XR) or Claim (XRC)

Since we want to create a Storage Account, here's an suggestion for an claim.yaml:

---
apiVersion: crossplane.jonashackt.io/v1alpha1
kind: StorageAzure
metadata:
  namespace: default
  name: managed-storage-account
spec:
  compositionRef:
    name: storageazure-composition
  parameters:
    location: West Europe
    resourceGroupName: rg-crossplane
    storageAccountName: account4c8672f

Testdrive with

kubectl apply -f upbound/provider-azure-storage/claim.yaml

Wait for the Azure ResourceGroup & Storage Account to become ready

We can use the new trace command of the crossplane CLI introduced in 1.14 to have a look, what's going on with our Azure resources:

$ crossplane beta trace storageazure.crossplane.jonashackt.io/managed-storage-account
NAME                                SYNCED   READY   STATUS                                                                                
StorageAzure/account (default)      True     False   Waiting: ...resource claim is waiting for composite resource to become Ready          
└─ XStorageAzure/account-g97s8      False    -       ReconcileError: ... "object": spec.forProvider.accountTier is a required parameter]   
   ├─ Account/account4c8672f        -        -       Error: accounts.storage.azure.upbound.io "account4c8672f" not found                   
   └─ ResourceGroup/rg-crossplane   -        -       Error: resourcegroups.azure.upbound.io "rg-crossplane" not found    

Ahh, so our Composition isn't crafted correctly.

To fix this, we need to have a look into the API docs at https://doc.crds.dev/github.com/upbound/provider-azure/storage.azure.upbound.io/Account/v1beta1@v0.38.2. After fixing our Composition and reapplying it together with the Claim, we can have a look into the Azure Portal. Our Resource Group should show up:

azure-console-resourcegroup

And also our Storage account should be visible inside the group:

azure-console-storageaccount

TODO:

Publish Crossplane Configuration Package into GitHub Container Registry

Let's package our Composite Resource definitions as a Configuration. Therefore we need to use the Crossplane CLI via kubectl crossplane build configuration (now they are a Package) - and push them to an OCI registry via kubectl crossplane push configuration. With this Configurations can also be easily installed into other Crossplane clusters.

Therefore we need a crossplane.yaml file as described in https://crossplane.io/docs/v1.8/getting-started/create-configuration.html#build-and-push-the-configuration

See also https://crossplane.io/docs/v1.8/concepts/packages.html#configuration-packages

Our crossplane-contrib/provider-aws/s3/crossplane.yaml is of kind: Configuration and defines the minimum crossplane version needed alongside the crossplane AWS provider:

apiVersion: meta.pkg.crossplane.io/v1
kind: Configuration
metadata:
  name: s3-bucket-composition
spec:
  crossplane:
    version: ">=v1.8"
  dependsOn:
    - provider: crossplanecontrib/provider-aws
      version: ">=v0.33.0"

Having this crossplane.yaml in place we can build our Configuration at last. On a command line go into the directory where the crossplane.yaml resides and run the kubectl crossplane build command:

$ cd crossplane-s3
$ kubectl crossplane build configuration

Really strange, getting

kubectl crossplane build configuration
kubectl crossplane: error: failed to build package: failed to parse package: {path:/Users/jonashecht/dev/kubernetes/crossplane-awws-azure/crossplane-contrib/provider-aws/s3/composition.yaml position:0}: no kind "S3Bucket" is registered for version "crossplane.jonashackt.io/v1alpha1" in scheme "/home/runner/work/crossplane/crossplane/internal/xpkg/scheme.go:47"

Provision an EKS cluster with crossplane

The crossplane core Controller and the Provider AWS Controller should now be ready to provision any infrastructure component in AWS!

So in order to maximise the Inception let's provision an EKS based Kubernetes cluster in AWS with our crossplane Kubernetes cluster :)

A EKS cluster is a complex AWS construct leveraging different basic AWS services. That's exactly what a crossplane Composite Resource (XR) is all about. So let's create our own Composite Resource.

Higher level abstractions - or why it is so hard to write your own Compositions

Looking only at the Crossplane docs and blog I thought something is missing: A curated library of higher level abstractions (like Pulumi Crosswalk). First initiatives from cloud vendors like AWS Blueprints for Crossplane: https://aws.amazon.com/blogs/opensource/introducing-aws-blueprints-for-crossplane/

But then I found the Upbound blog - and finally stumbled upon the Upbound Platform Reference Architectures https://blog.upbound.io/azure-reference-platform/ which are intended as

foundations to help accelerate understanding and application of Crossplane within your organization

Use Upbound Platform Reference Architecture AWS to provision a EKS cluster

What I didn't wanted to do is to duplicate some code into my Composition from all those sources I found (mainly this https://www.kloia.com/blog/production-ready-eks-cluster-with-crossplane, this https://aws.amazon.com/blogs/containers/gitops-model-for-provisioning-and-bootstrapping-amazon-eks-clusters-using-crossplane-and-argo-cd/ and https://aws.amazon.com/blogs/opensource/introducing-aws-blueprints-for-crossplane/) about provisioning a EKS cluster with crossplane.

Install platform-ref-aws

https://github.com/upbound/platform-ref-aws#install-the-platform-configuration

The problem with Open Source Crossplane: Provider Coverage Problem

See this blog post https://blog.upbound.io/first-official-providers/ and the paragraph Provider Coverage Problem:

The Crossplane community has been adding CustomResourceDefinitions (CRDs) to cover the API surface, but the pace is unable to catch up with users who are blocked from getting into production with Crossplane due to the lack of coverage. To give you an idea, AWS has ~1000 resources and the classic AWS provider has yet to cover ~200 as of writing.

Now with the classic AWS provider https://marketplace.upbound.io/providers/crossplane-contrib/provider-aws/v0.39.0/crds you have around 180 CRDs available. If you only need the CRDs that are provided by this classic provider, everything is okay. But I recently stumbled upon a new security default implementation of AWS S3: #12

This project stopped to work, simply since AWS doesn't support Access Control Lists (ACLs) for all new S3 Buckets anymore. See hashicorp/terraform-provider-aws#28353

Starting in April 2023, Amazon S3 will introduce two new default bucket security settings by automatically enabling S3 Block Public Access and disabling S3 access control lists (ACLs) for all new S3 buckets.

This in fact means, that we not only need the Bucket CRD, but we additionally need BucketACL, BucketPublicAccessBlock and BucketOwnershipControls, which are only available in the so called official Upbound providers with their 901 CRDs: https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/crds

Up until now I didn't really know the difference between the classic providers and the official providers. But the official providers are based on the tool Upjet https://github.com/upbound/upjet/ which is the successor of the former Terrajet https://github.com/crossplane/terrajet, where Terraform modules are used to generate Crossplane CRDs. With this approach, every API Terraform provides is also available in Crossplane. Upjet is currently used to generate all CRDs for the official AWS, Azure and GCP providers.

And with Upjet it's possible to generate really any Crossplane provider from Terraform, meaning one can rely on every Provider Terraform already has: https://registry.terraform.io/browse/providers

Why not use Terraform then you may ask. Theres a good post here https://blog.crossplane.io/crossplane-vs-terraform/ which makes the point, that Terraform can only be used from a bash script - but Crossplane delivers a control plane based approach, which continously reconciles all infrastructure configured inside of it.

So TLDR: To fix the error #12

S3 creation error: CannotCreateExternalResource managed/bucket.s3.aws.crossplane.io failed to create the Bucket: InvalidBucketAclWithObjectOwnership: Bucket cannot have ACLs set with ObjectOwnership's BucketOwnerEnforced setting

we need to switch over to the official AWS provider based on Upjet. Do we need to use the UXP (Universal Control Plane) flavour of Crossplane for that? No, the docs https://marketplace.upbound.io/providers/upbound/provider-aws/v0.34.0/docs/configuration state:

The Upbound AWS official provider may also be used with upstream Crossplane.

So let's use the provider inside our project!

IDE plugins

https://blog.upbound.io/moving-crossplane-package-authoring-from-plain-yaml-to-ide-aided-development/

Upbound VS Code plugin https://blog.upbound.io/crossplane-vscode-plugin-announcement/

upbound-vscode-extension

Links

https://crossplane.io/

Intro post https://medium.com/@khelge/crossplane-7197ad200013

Crossplane is based on OAM (3 roles: Developer, application operator & infrastructure operator) https://github.com/oam-dev/spec

https://www.kloia.com/blog/production-ready-eks-cluster-with-crossplane

https://www.forbes.com/sites/janakirammsv/2021/09/15/how-crossplane-transforms-kubernetes-into-a-universal-control-plane/

Crossplane essentially transforms Kubernetes into a universal control plane that can orchestrate the lifecycle of public cloud-based services such as virtual machines, database instances, Big Data clusters, machine learning jobs, and other managed services offered by the hyperscalers.

Crossplane takes the concept of Infrastructure as Code (IaC) to the next level through its tight integration with Kubernetes.

Intro Slides: https://docs.google.com/presentation/d/1PxZweRpB6HElxd9qGK1McboGZ1kluCDCS5qxgYnX5f0/edit#slide=id.g8801599ecb_2_169

Promoted from sandbox to incubation by the CNCF: https://blog.crossplane.io/crossplane-cncf-incubation/

Crossplane vs. Terraform: https://blog.crossplane.io/crossplane-vs-terraform/

Crossplane 101 https://vrelevant.net/crossplane-bringing-the-basics-together/

https://www.techtarget.com/searchitoperations/news/252508176/Crossplane-project-could-disrupt-infrastructure-as-code

Argo + crossplane https://morningspace.medium.com/using-crossplane-in-gitops-what-to-check-in-git-76c08a5ff0c4

Infrastructure-as-Apps https://codefresh.io/blog/infrastructure-as-apps-the-gitops-future-of-infra-as-code/

New Charter from Sep 2023 on: crossplane/crossplane#4643 Developer Experience and Providers are added to the scope of crossplane

Even Upjet and the major Upbound developed (aka Upjet generated) Providers are donated to the OS project & CNCF: https://blog.crossplane.io/charter-expansion-upjet-donation/

See also crossplane-contrib/provider-aws#1954

Crossplane + AWS

https://aws.amazon.com/blogs/containers/gitops-model-for-provisioning-and-bootstrapping-amazon-eks-clusters-using-crossplane-and-argo-cd/

Crossplane’s infrastructure provider for AWS relies on code generated by the AWS Go Code Generator, which is also used by AWS Controllers for Kubernetes (ACK).

https://www.kloia.com/blog/production-ready-eks-cluster-with-crossplane

AWS Blueprints for Crossplane:

https://aws.amazon.com/blogs/opensource/introducing-aws-blueprints-for-crossplane/

Writing your first Composition and debugging can be intimidating.

And it's quite a doublicate effort IMHO compared to other tools like Pulumi that provide higher level abstractions including well-architectured best practices (like Pulumi https://www.pulumi.com/docs/guides/crosswalk/aws/)

https://github.com/aws-samples/crossplane-aws-blueprints

Thoughworks Tech Radar: Assess https://www.thoughtworks.com/de-de/radar/tools/crossplane

IONOS and crossplane https://docs.ionos.com/crossplane-provider/example

CRD examples

https://github.com/upbound/platform-ref-s3-website

Show status of infrastructure components

k8s event-log! :events and then scroll down to the newest events and have a (constant) look

Opt out of automatic Composition Updates in XRs/Claims

Composition Updates are applied automatically to all XRs/Claims by default. As the docs state:

If you have 10 PlatformDB XRs all using the big-platform-db Composition, all 10 of those XRs will be instantly updated in accordance with any updates you make to the big-platform-db Composition.

So if we change a Composition, all the XRs will be updated - which we can see in the Kubernetes Events:

This can be a wanted behavior - or not at all.

But be aware of this error:

cannot compose resources: cannot parse base template of composed resource "securitygrouprule-cluster-inbound": cannot change the kind or group of a composed resource from ec2.aws.upbound .io/v1beta1, Kind=SecurityGroupRule to ec2.aws.upbound.io/v1beta1, Kind=SecurityGroupIngressRule (possible composed resource template mismatch) 

See crossplane/crossplane#1909

In the first steps with Crossplane you might encounter the issue already:

cannot compose resources: cannot parse base template of composed resource "securitygrouprule-cluster-inbound": cannot change the kind or group of a composed resource from ec2.aws.upbound .io/v1beta1, Kind=SecurityGroupRule to ec2.aws.upbound.io/v1beta1, Kind=SecurityGroupIngressRule (possible composed resource template mismatch) 

Composition Revisions allow XRs to opt out of automatic updates.

If you don't want Composition to do automatic updates, you need to use Composition Revisions in your XRs/XRCs via the compositionUpdatePolicy: Manual keyword:

apiVersion: example.org/v1alpha1
kind: PlatformDB
metadata:
  name: example
spec:
  parameters:
    storageGB: 20
  # The Manual policy specifies that you do not want this XR to update to the
  # current CompositionRevision automatically.
  compositionUpdatePolicy: Manual
  compositionRef:
    name: example
  writeConnectionSecretToRef:
    name: db-conn

Troubleshooting

https://docs.crossplane.io/knowledge-base/guides/troubleshoot/

Crossplane CRD Validation

Install crossplane CLI

https://github.com/crossplane/crossplane/releases/tag/v1.15.0

Enhancements to the Crossplane CLI: New subcommands like crossplane beta validate for schema validation, crossplane beta top for resource utilization views similar to kubectl top pods, and crossplane beta convert for converting resources to newer formats or configurations.

Install crossplane CLI:

curl -sL "https://raw.githubusercontent.com/crossplane/crossplane/master/install.sh" |sh

If that produces an error, try to manually craft the download link:

curl -sfLo crossplane "https://releases.crossplane.io/stable/v1.15.0/bin/linux_amd64/crank"
sudo mv crossplane /usr/local/bin

Be sure to have the v1.15.0 version installed as a minimum, otherwise the crossplane beta validate command won't work:

crossplane --version
v1.15.0

Validate XR or Claim against Composite Resource Definition

Before using the command have a look at the command reference: https://docs.crossplane.io/latest/cli/command-reference/#beta-validate:

The crossplane beta validate command validates compositions against provider or XRD schemas using the Kubernetes API server’s validation library.

So let's grab a definition.yaml and validate a claim.yaml against it:

crossplane beta validate --cache-dir ~/.crossplane definition.yaml claim.yaml
[✓] crossplane.jonashackt.io/v1alpha1, Kind=ObjectStorage, managed-upbound-s3 validated successfully
Total 1 resources: 0 missing schemas, 1 success cases, 0 failure cases

To prevent the command from polluting our projects with .crossplane directories, we should also provide a --cache-dir ~/.crossplane flag, which will deposit the directory in the user profile folder.

Validate Managed Resources against a Crossplane provider (family)

We need to provide the provider's schemes.

For example grab a Composition and validate it against the AWS provider:

crossplane beta validate --cache-dir ~/.crossplane config/provider-aws-s3.yaml resources/bucket.yaml
package schemas does not exist, downloading:  xpkg.upbound.io/upbound/provider-aws-s3:v1.1.0
[✓] s3.aws.upbound.io/v1beta1, Kind=Bucket, crossplane-argocd-s3-bucket validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketPublicAccessBlock, crossplane-argocd-s3-bucket-pab validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketOwnershipControls, crossplane-argocd-s3-bucket-osc validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketACL, crossplane-argocd-s3-bucket-acl validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketWebsiteConfiguration, crossplane-argocd-s3-bucket-websiteconf validated successfully
Total 5 resources: 0 missing schemas, 5 success cases, 0 failure cases
Validate a full directory against XRDs or Provider schema

We can also validate a full directory:

crossplane beta validate --cache-dir ~/.crossplane provider-aws-s3/definition.yaml provider-aws-s3
[✓] crossplane.jonashackt.io/v1alpha1, Kind=ObjectStorage, managed-upbound-s3 validated successfully
[!] could not find CRD/XRD for: apiextensions.crossplane.io/v1, Kind=Composition
[!] could not find CRD/XRD for: pkg.crossplane.io/v1, Kind=Provider
[!] could not find CRD/XRD for: aws.upbound.io/v1beta1, Kind=ProviderConfig
[!] could not find CRD/XRD for: apiextensions.crossplane.io/v1, Kind=CompositeResourceDefinition
[✓] s3.aws.upbound.io/v1beta1, Kind=Bucket, crossplane-argocd-s3-bucket validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketPublicAccessBlock, crossplane-argocd-s3-bucket-pab validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketOwnershipControls, crossplane-argocd-s3-bucket-osc validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketACL, crossplane-argocd-s3-bucket-acl validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketWebsiteConfiguration, crossplane-argocd-s3-bucket-websiteconf validated successfully
Total 10 resources: 4 missing schemas, 6 success cases, 0 failure cases
crossplane beta validate --cache-dir ~/.crossplane provider-aws-s3/config/provider-aws-s3.yaml provider-aws-s3
[!] could not find CRD/XRD for: crossplane.jonashackt.io/v1alpha1, Kind=ObjectStorage
[!] could not find CRD/XRD for: apiextensions.crossplane.io/v1, Kind=Composition
[!] could not find CRD/XRD for: pkg.crossplane.io/v1, Kind=Provider
[!] could not find CRD/XRD for: aws.upbound.io/v1beta1, Kind=ProviderConfig
[!] could not find CRD/XRD for: apiextensions.crossplane.io/v1, Kind=CompositeResourceDefinition
[✓] s3.aws.upbound.io/v1beta1, Kind=Bucket, crossplane-argocd-s3-bucket validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketPublicAccessBlock, crossplane-argocd-s3-bucket-pab validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketOwnershipControls, crossplane-argocd-s3-bucket-osc validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketACL, crossplane-argocd-s3-bucket-acl validated successfully
[✓] s3.aws.upbound.io/v1beta1, Kind=BucketWebsiteConfiguration, crossplane-argocd-s3-bucket-websiteconf validated successfully
Total 10 resources: 5 missing schemas, 5 success cases, 0 failure cases
beta render: kind of Unit tests

https://docs.crossplane.io/latest/cli/command-reference/#beta-render

https://blog.crossplane.io/building-crossplane-composition-functions-to-empower-your-control-plane/

Composition Validation

To be able to validate Compositions & XRs, we need another command in the game: crossplane beta render:

https://docs.crossplane.io/latest/cli/command-reference/#validate-render-command-output

You can pipe the output of crossplane beta render into crossplane beta validate to validate complete Crossplane resource pipelines, including XRs, compositions and composition functions.

Therefore we need to use the --include-full-xr command with crossplane beta render and the - option with crossplane beta validate like that:

crossplane beta render composition.yaml --include-full-xr | crossplane beta validate config/provider-aws-s3.yaml -

Provider Upgrades

Imagine a provider beeing configured like this:

apiVersion: pkg.crossplane.io/v1
kind: Provider
metadata:
  name: provider-aws-ec2
spec:
  package: xpkg.upbound.io/upbound/provider-aws-ec2:v1.1.0
  packagePullPolicy: Always
  revisionActivationPolicy: Automatic
  revisionHistoryLimit: 1

Now if provider-aws-ec2:v1.1.1 gets released, the revisionActivationPolicy: Automatic will lead to a automatically upgraded provider.

This can be great - or it can lead to a sitation, where the new provider version doesn't work as expected (e.g. because it presents a https://docs.crossplane.io/latest/concepts/providers/#unhealthypackagerevision) - and thus beeing rolled back, also automatically. This could lead to a situation, where the provider gets unusable from time to time.

If we want to do Provider upgrades in a GitOps fashion through a git commit to a git repository, we should configure the packagePullPolicy to IfNotPresent instead of Always (which means " Check for new packages every minute and download any matching package that isn’t in the cache", see https://docs.crossplane.io/master/concepts/packages/#configuration-package-pull-policy) - BUT leave the revisionActivationPolicy to Automatic! Since otherwise, the Provider will never get active and healty! See https://docs.crossplane.io/master/concepts/packages/#revision-activation-policy), but I didn't find it documented that way!

apiVersion: pkg.crossplane.io/v1
kind: Provider
metadata:
  name: provider-aws-ec2
spec:
  package: xpkg.upbound.io/upbound/provider-aws-ec2:v1.1.0
  packagePullPolicy: IfNotPresent # Only download the package if it isn’t in the cache.
  revisionActivationPolicy: Automatic # Otherwise our Provider never gets activate & healthy
  revisionHistoryLimit: 1

Provider & Configuration Package Upgrades with Renovate

Renovate supports Crossplane by the end of November 2023:

The Renovate docs tell us how we can configure Crossplane support:

https://docs.renovatebot.com/modules/manager/crossplane/

To use the crossplane manager you must set your own fileMatch pattern. The crossplane manager has no default fileMatch pattern, because there is no common filename or directory name convention for Crossplane YAML files. The crossplane manager supports these depTypes: configuration, function, provider

So we always need to explicitely configure Renovate to let it handle our Crossplane Provider & Configuration Package updates for us!

The simplest configuration would be:

"crossplane": {
    "fileMatch": ["\\.yaml$"]
  }

It makes sense, if most of the files are for Crossplane.

Readiness checks

https://docs.crossplane.io/latest/concepts/compositions/#resource-readiness-checks

https://docs.crossplane.io/latest/concepts/compositions/#match-a-boolean

End-2-End tests

With uptest and kuttl:

https://github.com/upbound/platform-ref-s3-website/blob/main/Makefile

How does Patching work?

https://github.com/awslabs/crossplane-on-eks?tab=readme-ov-file#features

https://github.com/awslabs/crossplane-on-eks/blob/main/doc/patching-101.md

Stacks now called Packages

https://github.com/crossplane/crossplane/blob/master/design/design-doc-composition.md

Note that Crossplane packages were formerly known as 'Stacks'. This term has been rescoped to refer specifically to a package of infrastructure configuration but is frequently synonymous with packages in historical design documents and some remaining Crossplane APIs.

Others

KUTTL - KUbernetes Test TooL

https://kuttl.dev/

example: https://github.com/jonashackt/crossplane-kuttl

GitHub Actions recommendations

A good advice is to use unique bucket names based on the branch to prevent interfering jobs with each other. Otherwise a Renovate enabled Crossplane project using real Cloud resources will create multiple crashed jobs in a row.

But we also need to split out the branch name, see https://stackoverflow.com/a/73467112/4964553 otherwise we'll run into errors like:

compose resources: cannot associate composed resources with Composition resource templates: cannot get composed resource: invalid resource name "spring2024-bucket-renovate/xpkg.upbound.io-upbound-provider-aws-s3-1.x": [may not contain '/']'

Therefore I created a BUCKET_NAME variable and Split branch name step which uses the predefined `{{ github.ref_name }}' containing the branch name with often with a slash (when they are opened by Renovate for example):

name: provision-aws

on: [push]

env:
  KIND_NODE_VERSION: v1.29.2
  BUCKET_NAME: spring2024-bucket
  # AWS
  AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
  AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
  AWS_DEFAULT_REGION: 'eu-central-1'

jobs:
  crossplane-provision-aws:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout
        uses: actions/checkout@master

      # Using branch unique name for our bucket to prevent interfering jobs
      # But we need to split out the branch name, see https://stackoverflow.com/a/73467112/4964553
      # otherwise we'll run into errors like: 'compose resources: cannot associate composed resources with Composition resource templates: cannot get composed resource: invalid resource name "spring2024-bucket-renovate/xpkg.upbound.io-upbound-provider-aws-s3-1.x": [may not contain '/']'
      - name: Split branch name
        env:
          BRANCH: ${{ github.ref_name }}
        id: split
        run: echo "::set-output name=branchbucketsuffix::${BRANCH##*/}"

Later I create a inline Claim instead of using the yaml file to be able to override the bucket name dynamically:

      # Not using kubectl apply -f upbound/provider-aws-s3/claim.yaml currently to prevent jobs from interfering with each other
      - name: Create XRD, Composition & Claim to create S3 Bucket (now using Upbound official AWS Provider Families)
        run: |
          ...

          echo "### Create Claim, which should create S3 Bucket (inline here to prevent jobs from interfering with each other)"
          kubectl apply -f - <<EOF
          apiVersion: crossplane.jonashackt.io/v1alpha1
          kind: ObjectStorage
          metadata:
            namespace: default
            name: managed-upbound-s3
          spec:
            compositionRef:
              name: objectstorage-composition
            parameters:
              bucketName: "$BUCKET_NAME-${{ steps.split.outputs.branchbucketsuffix }}"
              region: eu-central-1
          EOF

          ...

Now in AWS console we can watch Renovate doing it's work without interfering with other branches:

And finally we need to use the bucket name also when we interact with AWS S3:

      - name: Upload index.html to S3 and check deployment works
        run: |
          echo "### Upload index.html to Bucket via AWS CLI"
          aws s3 sync static "s3://$BUCKET_NAME-${{ steps.split.outputs.branchbucketsuffix }}" --acl public-read

          echo "### Access S3 Bucket static website"
          curl "http://$BUCKET_NAME-${{ steps.split.outputs.branchbucketsuffix }}.s3-website.eu-central-1.amazonaws.com"

      - name: Delete index.html and remove Claim for S3 Bucket deletion
        run: |
          echo "### Delete index.html from S3 Bucket"
          aws s3 rm "s3://$BUCKET_NAME-${{ steps.split.outputs.branchbucketsuffix }}/index.html"

          ...

About

Example project showing how to get started with Crossplane to e.g. provision a AWS S3Bucket & Azure StorageAccount through K8s CRDs

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages