You can label clusters the same way as any other Kubernetes object, which can help with grouping clusters and can also be leveraged by the ClusterSelector.
Starting in Kubernetes 1.7, there is alpha support for directing objects across the federated clusters with the annotation `federation.alpha.kubernetes.io/cluster-selector`. The *ClusterSelector* is conceptually similar to `nodeSelector`, but instead of selecting against labels on nodes, it selects against labels on federated clusters.
The annotation value must be JSON formatted and must be parsable into the [ClusterSelector API type](/docs/reference/federation/v1beta1/definitions/#_v1beta1_clusterselector). For example: `[{"key": "load", "operator": "Lt", "values": ["10"]}]`. Content that doesn't parse correctly will throw an error and prevent distribution of the object to any federated clusters. Objects of type ConfigMap, Secret, Daemonset, Service and Ingress are included in the alpha implementation.
The *key* is matched against label names on the federated clusters.
The *values* are matched against the label values on the federated clusters.
The possible *operators* are: `In`, `NotIn`, `Exists`, `DoesNotExist`, `Gt`, `Lt`.
The *values* field is expected to be empty when `Exists` or `DoesNotExist` is specified and may include more than one string when `In` or `NotIn` are used.
Currently, only integers are supported with `Gt` or `Lt`.