Provide more prescriptive guidance on multi-tenancy

Signed-off-by: Naadir Jeewa <[email protected]>

Author: Naadir Jeewa

docs/book/src/developer/architecture/controllers/multi-tenancy.md

@@ -1,7 +1,7 @@
 # Multi tenancy
 
-Multi tenancy in Cluster API defines the capability of an infrastructure provider to manage different credentials, each
-one of them corresponding to an infrastructure tenant.
+Multi tenancy in Cluster API defines the capability of an infrastructure provider to manage
+different credentials, each one of them corresponding to an infrastructure tenant.
 
 ## Contract
 
@@ -11,3 +11,108 @@ In order to support multi tenancy, the following rule applies:
 - Providers SHOULD deploy and run any kind of webhook (validation, admission, conversion)
   following Cluster API codebase best practices for the same release.
 - Providers MUST create and publish a `{type}-component.yaml` accordingly.
+- Providers MUST use the field of name `identityRef` to reference the multi-tenancy CR in the
+  `InfraCluster` CR.
+- Providers MUST implement `identityRef` must be of the following type:
+
+  ```go
+  type ProviderIdentityReference struct {
+    // Kind of the identity.
+    // +kubebuilder:validation:Enum=<Supported provider types>
+    Kind string
+    // Name of the identity.
+    // +kubebuilder:validation:MinLength=1
+    Name string
+  }
+  ```
+
+- A Namespace field MUST never be added to the ProviderIdentityReference type to avoid crossing
+  namespace boundaries.
+
+- Where identity types use private key material, CRs MUST implement a `secretRef` on their spec of
+  type string and only read secrets from the same namespace as the CR for namespaced scope resources
+  OR the controller namespace for cluster-scoped resources.
+
+- Providers MAY support `Secret` as a top-level
+  supported identity type, but only from the same namespace as the owning CR.
+
+- Providers MAY support the use of `identityRef` in other low-level resources, such as Load
+  Balancers.
+
+## Supported RBAC Models
+
+Providers MAY support any combination of cluster-scoped or namespace-scoped CRs as follows:
+
+### Cluster-scoped global resources for delegated access
+
+In a common use for multi-tenancy, a cloud admin will want to set up a range of identities, and then
+delegate them to individual teams. This is best done using global resources to prevent repetition.
+
+#### Cluster-scoped Contract
+
+- Cluster scoped resources MUST be named with `<Provider>Cluster<Type>Identity`. Examples:
+  - `FabrikamClusterRoleIdentity`
+  - `FabrikamClusterStaticIdentity`
+
+- Where identity types use private key material, CRs MUST
+  implement a `secretRef` on their spec of type string and
+  only read secrets from the controller namespace.
+
+- Cluster scoped resources MUST be delegated using a
+  label selector, present on the spec as:
+
+```go
+type InfraClusterIdentity struct {
+  // AllowedNamespaces is used to identify which namespaces are allowed to use the identity from.
+  // Namespaces can be selected either using an array of namespaces or with label selector.
+  // An empty allowedNamespaces object indicates that clusters can use this identity from any
+  // namespace.
+  // If this object is nil, no namespaces will be allowed (default behaviour, if this field is not
+  // provided).
+  // A namespace should be either in the NamespaceList or match with Selector to use the identity.
+  //
+  // +optional
+AllowedNamespaces *AllowedNamespaces `json:"allowedNamespaces"`
+}
+
+type AllowedNamespaces struct {
+	// AllowedNamespaces is a selector of namespaces that clusters can use this Identity from.
+  // This is a standard Kubernetes LabelSelector, a label query over a set of resources.
+  // The result of matchLabels and matchExpressions are ANDed.
+	//
+	// An empty selector indicates that resources cannot use this identity from any namespace if
+  // NamespaceList is not supplied.
+	// +optional
+	Selector metav1.LabelSelector `json:"selector"`
+}
+```
+- Providers MAY support an additional `NamespaceList` string
+  slice in the `AllowedNamespaces` struct.
+
+- Providers MUST treat nil for `AllowedNamespaces` as matching zero namespaces.
+- Providers MUST treat `{}` as matching *all* namespaces.
+- Providers SHOULD mark `NamespaceList` as deprecated from v1alpha4
+
+When Cluster API no longer supports Kubernetes versions older than Kubernetes v1.21, when the
+NamespaceDefaultLabelName [feature gate] transitioned to Beta, then:
+
+  - Providers MUST remove the `NamespaceList` field.
+  - Conversion webhooks in MUST perform the following:
+    - Translate the NamespaceList into the following match expression:
+      -  `kubernetes.io/metadata.name in (<comma-separated list of namespaces from <NamespaceList>)`
+
+### Namespaced scoped resources
+
+Namespaced scoped resources are useful most particularly when you want to allow developers to
+provision clusters on their own accounts, but may not be suitable for every use case.
+
+#### Namespace-scoped contract
+
+- Namespace scoped resources MUST be named `<Provider><Type>Identity`, e.g.:
+  - `ContosoCloudRoleIdentity`
+  - `ContosoCloudStaticIdentity`
+- Where identity types use private key material, CRs MUST
+  implement a `secretRef` on their spec of type string and
+  only read secrets from the same namespace as the CR.
+
+[feature gate]: https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/