Skip to content

crossplane-aws-s3

Upbound’s Crossplane provider for AWS S3, version v2.9.0. You create S3 buckets and their settings as Kubernetes objects in your namespace. The provider makes them in your AWS account.

  • API reference: provider-aws-s3 v2.9.0
  • API groups: s3.aws.m.upbound.io (buckets and their settings) and aws.m.upbound.io (ProviderConfig).
  • Not there: Crossplane itself (compositions, packages) and the cluster-scoped types (ClusterProviderConfig, group s3.aws.upbound.io).
  1. Turn the plugin on: open your control plane, then Plugins → Turn on.
  2. Give your AWS credentials: a Secret with the label plugins.ctrlplane.run/copy=true.
  3. Make a ProviderConfig named default that reads the Secret.
  4. Make your buckets. Each one must point to your ProviderConfig.

Replace <namespace> with your control plane’s namespace, for example alice-dev. The site’s plugin catalog fills it in for you.

# 1. Your AWS credentials. The label copies the Secret to the operator.
apiVersion: v1
kind: Secret
metadata:
name: aws-creds
namespace: <namespace>
labels: {plugins.ctrlplane.run/copy: "true"}
stringData:
credentials: |
[default]
aws_access_key_id = <your key id>
aws_secret_access_key = <your secret key>
---
# 2. A ProviderConfig that reads the Secret.
apiVersion: aws.m.upbound.io/v1beta1
kind: ProviderConfig
metadata: {name: default, namespace: <namespace>}
spec:
credentials:
source: Secret
secretRef: {name: aws-creds, namespace: <namespace>, key: credentials}
---
# 3. A bucket. The name is the S3 bucket name: it must be unique in all of AWS.
apiVersion: s3.aws.m.upbound.io/v1beta1
kind: Bucket
metadata: {name: <namespace>-example, namespace: <namespace>}
spec:
forProvider: {region: eu-central-1}
providerConfigRef: {kind: ProviderConfig, name: default}

Apply it with kubectl apply -f, or in the site under Resources → Apply resources.

The Secret is copied to the operator within five minutes. Until then, the bucket is not ready.

Terminal window
kubectl get buckets.s3.aws.m.upbound.io

SYNCED and READY become True when the bucket exists in AWS. If not, look at the conditions:

Terminal window
kubectl describe buckets.s3.aws.m.upbound.io <name>
  • Set providerConfigRef on every resource, with kind: ProviderConfig. Without it, the provider uses a ClusterProviderConfig named default. That does not exist, so the request is refused.
  • Use credentials.source: Secret in a ProviderConfig. Other sources (IRSA, WebIdentity, PodIdentity, Upbound) use the identity of the provider pod, which is not yours. They are refused.
  • Do not set spec.endpoint in a ProviderConfig. The provider calls AWS only; a custom endpoint is refused.
  • When you delete a bucket object, the provider deletes the bucket in AWS. To keep the bucket, set spec.managementPolicies before you delete. See the API reference.
  • To turn the plugin off, delete all its objects first. Your ProviderConfig counts too.