Operator for managing S3 resources on a Ceph RadosGW installation via Kubernetes Custom Resources.
Creating an initial S3 user on the Ceph RadosGW, which will be used to create all resources via the operator.
$ radosgw-admin user create --uid=operator --display-name="Kubernetes Operator" --gen-access-keys
$ radosgw-admin --id admin caps add --caps="buckets=*;users=*;usage=*;metadata=*" --uid=operator
New RadosGW users, which will be used to logicially separate entities, can be created via the following CRD:
apiVersion: s3.hanse-merkur.de/v1alpha1
kind: User
metadata:
name: operator-test
annotations:
# The `allow-deletion` annotation enables the actual deletion of the Ceph User, aside from the resource itself
s3.hanse-merkur.de/allow-deletion: "true"
# `allow-import` allows the import of existing Ceph users under the management of the resource
s3.hanse-merkur.de/allow-import: "true"
# `skip-tenant` unsets the operator configured tenant as prefix for the Ceph user
s3.hanse-merkur.de/skip-tenant: "true"
spec:
# Unless `skip-tenant` is set the user will be named <tenant>-<userId>
userId: operator-test
# Free form text field for contact information
contactName: Operator Test
# Enable quotas for the user globally. Applied to all buckets
quotas:
enabled: true
maxObjects: 1000000
maxSize: 1024000
maxBuckets: 10
Created Ceph users have to be unique, if using the tenant unique to the operator installation.
Access Keys allow the access to the Bucket and other S3 functionality. The resource, unlike the User, is namespaced as it can access a Secret containing the access and secret keys. Access Keys are unique and can be auto-generated or use existing Kubernetes secrets.
apiVersion: s3.hanse-merkur.de/v1alpha1
kind: AccessKey
metadata:
name: operator-test
namespace: default
spec:
# References the User resource created by the previous YAML
owner: operator-test
# Free form text field for general descriptions
description: Access key for operator-test
# Name of the secret created or existing in the same namespace as the AccessKey resource
secretName: operator-test
# Using jinja2 the resoluting secret can optionally be modified to be application specific as seen in this example
template:
metadata:
labels:
hansemerkur.de/owned-by: some-group
annotations:
hansemerkur.de/cleanup: none
data:
credentials: |
[default]
aws_secret_key_id={{ data.aws_access_key_id }}
aws_secret_access_key={{ data.aws_secret_access_key }}
Buckets are namespaced resources and use the implied AccessKey to create and update the Bucket in Ceph.
apiVersion: s3.hanse-merkur.de/v1alpha1
kind: Bucket
metadata:
name: operator-test
namespace: default
annotations:
# The `allow-deletion` annotation enables the actual deletion of the Ceph Bucket, aside from the resource itself
s3.hanse-merkur.de/allow-deletion: "true"
# Without the `force-deletion` annotation the bucket will never be deleted as long as it still contains objects
s3.hanse-merkur.de/force-deletion: "true"
# `allow-import` allows the import of an existing bucket under the management of the resource
s3.hanse-merkur.de/allow-import: "true"
spec:
bucketName: operator-test
# References the previously created AccessKey
ownerAccessKey: operator-test
# A selection of private, public or custom bucket policy. In case of custom the customBucketPolicy field is expected
bucketPolicy: private
# Optional standard AWS bucket policy format in JSON
#customBucketPolicy: "{}"
# Optional AWS Lifecycle Policy in JSON format
lifeCyclePolicy: |
{
"Rules": [
{
"Status": "Enabled",
"Expiration": {
"Days": 31
},
"ID": "ExpireSnapshots"
}
]
}
# Optional bucket quotas
quotas:
enabled: true
maxObjects: 100000
maxSize: 102400