object-storage-operator

README

Controllers

Bucket controller

This controller reconciles Buckets. It creates a cloud provider bucket in the Management Cluster Region. It currently only supports AWS S3 on CAPA management clusters and Azure Storage Container on CAPZ management clusters.

CAPZ resources

To handle an object storage on Azure, we need to create:

• a storage account
• a storage container

We choose to create a unique relation Storage Account - Storage Container to have a proper clean up when a bucket is deleted. This way, there won't have orphan storage account on Azure.

We add a lifecyle management rule on the storage account to clean old data (bucket.spec.expirationPolicy.days)

When the object storage is created, we retrieve the Access Key and create a secret in the bucket namespace containing the name of the storage account and the access key. This secret is necessary for the application desiring to use this object storage.

By default, a reclaim policy is set to reclaimPolicy: Retain that means when a Bucket CR is deleted, nothing is done. The idea is to avoid accidental Bucket CR deletions that result in data loss on the Cloud provider. However, if we need to clean up the bucket, we can set the reclaim policy to reclaimPolicy: Delete. This will remove all data on the Cloud provider.

Testing

You can run all tests with

make test-all

Unit tests

You can run only the unit tests with

make test-unit

Integration tests

You can run only the integration tests with

make test-integration

This project uses LocalStack for integration tests. Remember that you can use the aws cli targetting the local LocalStack environment with

$ aws --endpoint=http://localhost:4566 s3 list-buckets