This guide explains how to use Clusters API resource in a Federation control plane.
Different than other Kubernetes resources, such as Deployments, Services and ConfigMaps, clusters only exist in the federation context, i.e. those requests must be submitted to the federation api-server.
To list the clusters available in your federation, you can use kubectl by running:
kubectl --context=federation get clusters
The --context=federation
flag tells kubectl to submit the
request to the Federation apiserver instead of sending it to a Kubernetes
cluster. If you submit it to a k8s cluster, you will receive an error saying
the server doesn't have a resource type "clusters"
If you passed the correct Federation context but received a message error saying
No resources found.
it means that you haven’t added any cluster to the Federation yet.
Creating a cluster
resource in federation means joining it to the federation. To do so, you can use
kubefed join
. Basically, you need to give the new cluster a name and say what is the name of the
context that corresponds to a cluster that hosts the federation. The following example command adds
the cluster gondor
to the federation running on host cluster rivendell
:
kubefed join gondor --host-cluster-context=rivendell
You can find more details on how to do that in the respective section in the kubefed guide.
Converse to creating a cluster, deleting a cluster means unjoining this cluster from the
federation. This can be done with kubefed unjoin
command. To remove the gondor
cluster, just do:
kubefed unjoin gondor --host-cluster-context=rivendell
You can find more details on unjoin in the kubefed guide.
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.
kubectl --context=rivendell label cluster gondor key1=value1 key2=value2
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. 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.
Here is an example ClusterSelector annotation, which will only select clusters WITH the label pci=true
and WITHOUT the label environment=test
:
metadata:
annotations:
federation.alpha.kubernetes.io/cluster-selector: '[{"key": "pci", "operator":
"In", "values": ["true"]}, {"key": "environment", "operator": "NotIn", "values":
["test"]}]'
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
.
The full clusters API reference is currently in federation/v1beta1
and more details can be found in the
Federation API reference page.