OpenNebula Kubernetes Service (OneKS) simplifies the deployment and management of elastic Kubernetes clusters on OpenNebula. It provides a complete cluster lifecycle workflow, from provisioning the control plane and adding worker nodes to deploying applications, scaling the cluster, and terminating it when it is no longer needed. 

Kubernetes Management, Integrated into OpenNebula

OneKS uses Cluster API (CAPI) to provision and manage Kubernetes clusters throughout their lifecycle. Cluster API was introduced in OpenNebula 7.0 and is now heavily used by OneKS. 

Through the dedicated Kubernetes section in Sunstone, administrators can create and manage clusters, configure NodeGroups, monitor deployment logs, and retrieve kubeconfig files. 

Before deploying a cluster, OneKS requires tproxy to be correctly configured for OneGate. Since OpenNebula 7.4, OneKS also provides a pre-deployment verification function that helps identify issues and confirm that the environment is ready for Kubernetes deployment. 

From Cluster Provisioning to Application Deployment

Once the environment is ready, administrators can create a Kubernetes cluster through the OneKS interface. The deployment wizard allows them to select the OpenNebula cluster and public and private networks, choose the Kubernetes version, and select between a single-node or HA deployment. 

Worker capacity can then be added through NodeGroups, which define the size and number of worker VMs. OneKS provisions the VMs and adds them to the Kubernetes cluster. 

Once the cluster is running, administrators can retrieve its kubeconfig and use standard Kubernetes tools to deploy workloads. In the demonstration, a test application is deployed and exposed through a Kubernetes LoadBalancer service. OneKS provisions a virtual router to route traffic from the routable network to the exposed service. 

OneKS also supports scaling existing NodeGroups, allowing administrators to add worker capacity when needed. When a cluster is no longer required, it can be removed from the same management interface, initiating the deprovisioning of the cluster and its associated resources.

See OneKS in Action

The following screencast demonstrates the complete OneKS workflow, from verifying the environment and provisioning a Kubernetes cluster to adding worker nodes, deploying and exposing an application, scaling the cluster, and finally terminating it. 

What Is Coming?

OneKS is being extended with an application chart catalog that simplifies the deployment of common Kubernetes applications and platform services. Instead of requiring administrators to install and configure each component manually, the catalog provides curated applications with their supported versions, configuration options, dependencies, and deployment requirements. These applications can be deployed and managed directly on OneKS clusters, extending the service from Kubernetes cluster management to application lifecycle management.

The catalog will also include applications for AI and GPU-accelerated environments, such as the NVIDIA GPU Operator and Run:ai. On clusters with GPU-enabled worker nodes, these charts can be deployed together with their required dependencies to configure and expose GPU resources to Kubernetes and to manage and share those resources among AI workloads.

Want to explore Kubernetes on OpenNebula further? Visit our Kubernetes on OpenNebula page to learn more about OneKS and explore the Screencasts Hub for more practical OpenNebula walkthroughs.