You are viewing documentation for a previous version of GKE On-Prem. View the latest documentation.


The following sections offer definitions for GKE On-Prem and Kubernetes terms. You can also refer to the Kubernetes standardized glossary.


Admin cluster

A cluster which creates user clusters and manages their control planes. All API calls to and from GKE On-Prem are handled by the admin control plane that runs in the admin cluster.

Admin control plane

The control plane running in the admin cluster.

This control plane specifically handles all Kubernetes API calls to and from GKE on-prem: when an API call is made, it goes to the admin control plane for processing, then routes to its intended destination. The admin control plane manages the full lifecycle of user clusters, including creation, upgrading, and deletion. It runs services that interact with vSphere and the Connect Agent.



From the Kubernetes standardized glossary: "A set of machines, called nodes, that run containerized applications managed by Kubernetes."

In GKE On-Prem, there are admin clusters and user clusters. Clusters can run on-premises or in the cloud.

Control plane

A cluster's controlling unit, consisting of a set of components that schedule and manage workloads, communicate with clusters, and ensure that clusters are functioning. Control planes include the etcd key-value datastore, the Kubernetes API server, the scheduler, and the controller manager. Also refer to the Kubernetes documentation for Kubernetes control plane.



See Deployment.



The command-line interface (CLI) to GKE On-Prem. You use gkectl to create and manage GKE On-Prem clusters, and to diagnose issues with clusters.


Island Mode

GKE On-Prem does not create an overlay network for cluster networking. Instead, it creates a node-to-node mesh using BGP so that Pods can reach each other within the cluster using the existing underlay network. But this network is not directly reachable from outside the cluster, as the routes are only announced between the nodes that make up the cluster. This is configuration is called Island Mode networking as it can be thought of an island within the existing on-prem network.



See node.



A virtual or physical machine in a cluster on which workloads run and that can have workloads scheduled against it. Also called a "VM" or "machine." A set of nodes is called a cluster.



A Kubernetes object that runs a containerized workload. It is the smallest deployable unit of computing in Kubernetes. Pods are usually managed by another object, like a Deplpoyment or StatefulSet. Also refer to the Kubernetes documentation for Pod.



A Kubernetes object that logically groups a set of Pods and defines a policy by which to access them. Also refer to the Kubernetes documentation for Service.


A Kubernetes objects meant for stateful applications. Pods managed by a StatefulSet get a unique, persistent identity in their cluster. Also refer to the Kubernetes documentation for StatefulSet.


User cluster

A cluster where developers' Kubernetes workloads run. User clusters are managed by an admin cluster.


Virtual machine (VM)

An emulation of a computer system with specific architecture and hardware specifications. Substitutes physical machine hardware.