Kubernetes Cluster Federation


Kubernetes lets us to associate multiple clusters into a cluster of clusters as of side to side Cluster Federation. It permits customers to deploy and manage apps through multiple clusters running in diverse locations in the world. The determination for Cluster Federation is not merely to safeguard great availability, nonetheless to combine multiple heterogeneous clusters into a single super-cluster managed over a single management interface.

Example # 1

We can run privacy-sensitive components of our application system on-premises as joining an on-premises cluster with one running on a cloud provider’s infrastructure, whereas the non-sensitive portions may run in the cloud.

Example # 2

The other example is primarily running our application only in a small on-premises cluster. This is happen when an application’s compute supplies surpass the cluster’s volume, allowing the application fall over to a cloud-based cluster, which is routinely provisioned on the cloud provider’s substructure.

The architecture understanding

What is Kubernetes Cluster Federation?

The cluster of clusters may be related to a steady cluster where as an alternative of nodes, we have whole clusters. The federated cluster contains of a Federated Control Plane and multiple Kubernetes clusters just as a Kubernetes cluster consists of a Control Plane and multiple worker nodes. The Federated Control Plane does the similar entity, but crosswise a set of clusters instead of nodes as how the Kubernetes Control Plane achieves applications through a set of worker nodes. The Federated Control Plane comprises on three stuffs:

  • etcd for storing the federated API objects

It stocks the federated API objects

  • Federation API server

The Federation Controller Manager runs the numerous federation controllers that do operations founded on the API objects we create over the API server.

  • Federation Controller Manager

The API server is the REST endpoint all other components speak to.

Users approach the Federation API server to make federated API objects. The federation controllers look to these objects and then talk to the underlying clusters’ API servers to generate regular Kubernetes resources.

Federated API objects

We learnt that the federated API server permits us to create federated variants. Federated versions of Kubernetes resources

We would have to check the Kubernetes Cluster Federation documentation for an up-to-date list of supported federated resources. The following federated resources are often supported:

  • Namespaces
  • ConfigMaps and Secrets
  • Services and Ingresses
  • Deployments, ReplicaSets, Jobs, and Daemonsets
  • HorizontalPodAutoscalers

The Federated API server also cares the Cluster object, which represents an underlying Kubernetes cluster, the identical way a Node object signifies a worker node in a steady Kubernetes cluster. We visualize for help how federated objects relate to the objects created in the underlying clusters.

What federated resources do?

The controllers running in the Federation Controller Manager would generate steady cluster-scoped resources in all underlying Kubernetes clusters and accomplish them till the federated object is deleted when we create the object in the Federation API server for portion of the federated objects. The resources shaped in the underlying clusters are exact replicas of the federated resource for firm federated resource types. And for others, they are rather changed forms, although sure federated resources do not cause everything to be produced in the underlying clusters at all. The replicas are kept in sync with the unique federated versions. The organization is one-directional only from the federation server down to the underlying clusters. The changes would not be synced up to the Federation API server if we change the resource in an underlying cluster. For instance, a namespace with the same name would be created in all underlying clusters if we create a namespace in the federated API server. If we then create a federated ConfigMap inside that namespace, a ConfigMap with that exact name and contents would be shaped in all underlying clusters in the same namespace. It’s too spread on to Secrets, Services, and DaemonSets. The different are ReplicaSets and Deployments . They are not sightlessly copied to the underlying clusters, since that’s not what the user typically needs. Finally, if we make a Deployment with a wanted replica count of 10, we probably don’t want 10 pod replicas running in each underlying cluster. We want 10 replicas in total. Because of this, the Federation controllers generate underlying Deployments/ReplicaSets so that the sum of their desired replica counts equals the wanted replica count specified in the federated Deployment or Replica Set, when we require a desired replica count in a Deployment or Replica Set,. The replicas are extent regularly crosswise the clusters, but this may be overridden by default.


We require currently connecting to every cluster’s API server separately to get the list of pods running in that cluster. We may not list all the clusters’ pods via the Federated API server. On the other hand the federated entry resource doesn’t outcome in the formation of any entry objects in the underlying clusters. The federated Ingress controller is required for regular Ingresses.

Mansoor Ahmed

Mansoor Ahmed is Chemical Engineer, web developer, a Tech writer currently living in Pakistan. My interests range from technology to web development. I am also interested in programming, writing, and reading.