Kubernetes Cluster Federation

Kubernetes Cluster Federation (KubeFed for short) allows you to coordinate the configuration of multiple Kubernetes clusters from a single set of APIs in a hosting cluster. KubeFed aims to provide mechanisms for expressing which clusters should have their configuration managed and what that configuration should be. The mechanisms that KubeFed provides are intentionally low-level, and intended to be foundational for more complex multicluster use cases such as deploying multi-geo applications and disaster recovery.

KubeFed is currently alpha and moving rapidly toward its initial beta release.

Concepts

KubeFed is configured with two types of information:

• Type configuration declares which API types KubeFed should handle
• Cluster configuration declares which clusters KubeFed should target

Propagation refers to the mechanism that distributes resources to federated clusters.

Type configuration has three fundamental concepts:

• Templates define the representation of a resource common across clusters
• Placement defines which clusters the resource is intended to appear in
• Overrides define per-cluster field-level variation to apply to the template

These three abstractions provide a concise representation of a resource intended to appear in multiple clusters. They encode the minimum information required for propagation and are well-suited to serve as the glue between any given propagation mechanism and higher-order behaviors like policy-based placement and dynamic scheduling.

These fundamental concepts provide building blocks that can be used by higher-level APIs:

• Status collects the status of resources distributed by KubeFed across all federated clusters
• Policy determines which subset of clusters a resource is allowed to be distributed to
• Scheduling refers to a decision-making capability that can decide how workloads should be spread across different clusters similar to how a human operator would

Features

Feature	Maturity	Feature Gate	Default
Push propagation of arbitrary types to remote clusters	Alpha	PushReconciler	true
CLI utility (kubefedctl)	Alpha
Generate KubeFed APIs without writing code	Alpha
Multicluster Service DNS via external-dns	Alpha	CrossClusterServiceDiscovery	false
Multicluster Ingress DNS via external-dns	Alpha	FederatedIngress	false
Replica Scheduling Preferences	Alpha	SchedulerPreferences	true

Guides

Quickstart

1. Clone this repo:

   git clone https://github.com/kubernetes-sigs/kubefed.git
   

2. Start a kind cluster:

   kind create cluster
   

3. Deploy kubefed:

   make deploy.kind
   

You now have a Kubernetes cluster with kubefed up and running. The cluster has been joined to itself and you can test federation of resources like this:

1. Verify the KubeFedCluster exists and is ready:

   kubectl -n kube-federation-system get kubefedcluster
   

   If you're on macOS the cluster will not immediately show as ready. You need to change the API endpoint's URL first:

   ./scripts/fix-joined-kind-clusters.sh
   

2. Create a namespace to be federated:

   kubectl create ns federate-me
   

3. Tell kubefed to federate that namespace (and the resources in it):

   ./bin/kubefedctl federate ns federate-me
   

4. Create a ConfigMap to be federated:

   kubectl -n federate-me create cm my-cm
   

5. Tell kubefed to federate that ConfigMap:

   ./bin/kubefedctl -n federate-me federate configmap my-cm
   

6. Verify the FederatedConfigMap has been created and propagates properly:

   kubectl -n federate-me describe federatedconfigmap my-cm
   

User Guide

Take a look at our user guide if you are interested in using KubeFed.

Development Guide

Take a look at our development guide if you are interested in contributing.

Community

Refer to the contributing guidelines if you would like to contribute to KubeFed.

Communication channels

KubeFed is sponsored by SIG Multicluster and it uses the same communication channels as SIG multicluster.

• Slack channel: #sig-multicluster
• Mailing list

Code of Conduct

Participation in the Kubernetes community is governed by the Kubernetes Code of Conduct.