Getting Started

Overview

Private Cloud Director lets you create and manage Kubernetes clusters from the same console you use for your virtual machines. A Kubernetes cluster consists of a control plane, which holds the cluster state and serves the Kubernetes API, and a set of worker nodes that run your applications.

Every cluster you create involves two choices:

• The infrastructure the nodes run on. With virtualized nodes, PCD provisions cluster nodes as virtual machines on one of your virtualized clusters. With physical nodes, the cluster runs on bare metal machines that you onboard yourself.
• Where the control plane runs. A Managed Control Plane is hosted for you on the PCD management plane. A Colocated Control Plane runs on your own infrastructure, and each cluster gets its own dedicated, isolated control plane. For a full comparison, see Control Plane Types.

This guide covers the virtualized-node path, which is the most common starting point. For the bare metal path, see Set up your Bare Metal Kubernetes Cluster.

In this guide, you will deploy your first Kubernetes cluster on virtualized nodes and confirm that it comes up.

Before You Begin

Ensure that:

• Virtualization is configured, with hosts onboarded and ready to create virtual machines. Review Pre-requisites before you start.
• You have a virtualized cluster with at least one host in it. Virtualized clusters without hosts cannot be selected in the wizard.
• You have an SSH key available for the node virtual machines, or are ready to create one during the flow.
• You have decided which control plane type this cluster needs.

Start the Deployment Wizard

1. Navigate to Kubernetes > Infrastructure > Clusters in the left-hand navigation menu.
2. Start the wizard:

   • If you have not created a cluster yet, select Create New Cluster with Virtualized Nodes on the getting-started page.
   • If you already have clusters, select Deploy New Cluster, then choose with Virtualized Nodes.

The Deploy A New Cluster wizard opens. It has three steps: Cluster Architecture, Servers, Storage and Network, and Configure your Cluster. If you select a Colocated control plane on the first step, an extra step, Control Plane Node Configuration, appears after it.

Each step has Back and Next. The final step has Submit.

Step 1: Choose Your Cluster Architecture

On the Cluster Architecture step, you name the cluster and select the compute infrastructure it uses.

1. In Cluster Name, enter a name that is unique among your clusters. This name identifies the cluster throughout the console.
2. Under Select the type of control plane, choose one of the two cards:

   • Managed Control Plane, hosted on the PCD management plane. There are no control plane nodes for you to size or provision, and your compute capacity stays available for worker nodes.
   • Colocated Control Plane, hosted on your own infrastructure. Each cluster gets its own dedicated, isolated control plane. Creating the cluster takes longer, because dedicated control plane nodes have to be provisioned and joined.
3. From Virtualized Cluster, select the virtualized cluster that this Kubernetes cluster attaches to. Its nodes are created as virtual machines there.
4. From SSH Key, select the key to install on the node virtual machines. If you do not have a key yet, select Create SSH Key to import one, then return to the wizard.
5. Select Next.

The control plane type is chosen when the cluster is created and cannot be changed afterwards. If you are unsure which to pick, see Control Plane Types.

Step 2: Configure Control Plane Nodes

This step applies only to a Colocated control plane. If you chose a Managed control plane, the wizard skips it and you continue at Step 3.

The Control Plane Node Configuration step configures the Control Plane Node Group, whose nodes host the Kubernetes control plane components.

1. Select the virtual machine flavor for the control plane nodes and set how many nodes to create. The group defaults to three nodes. Keep an odd number of at least three, so the control plane retains a quorum if a node is lost.
2. Select the Network and then the Subnet for these nodes. The network must be able to reach both the control plane nodes and the worker nodes.
3. Select Next.

Autoscaling does not apply to the control plane node group, so no autoscaling option is shown on this step.

Step 3: Configure Servers, Storage and Network

On the Servers, Storage and Network step, you configure the worker node groups. A node group is a logical entity that acts as a template for a set of Kubernetes worker nodes, each group with its own node count, networking, and storage properties.

For each node group:

1. Select the virtual machine flavor and the number of nodes to create from it.
2. Select the Network and then the Subnet the nodes attach to.
3. To let the cluster add and remove nodes on demand, enable Automatically scale this Node Group and set Max Nodes. PCD deploys and operates a Kubernetes cluster autoscaler for each cluster, so there is nothing to install yourself.
4. Select Next.

Node group root volumes can use either ephemeral storage or volume-backed storage. To use volume-backed storage, for example when compute nodes have limited local disk space, specify a root volume size and type in the node group settings.

For virtual machine based clusters, Kubernetes version upgrades and other operations are performed by creating a new set of nodes and recycling the old nodes.

Create enough nodes in the cluster to perform these operations safely and to reduce the impact of a single node being replaced.

Step 4: Configure Your Cluster

The Configure your Cluster step sets the Kubernetes version and the add-ons the cluster ships with.

1. From Kubernetes Version, choose a version that your workloads support.
2. From Available Images, choose the node operating system image. Only images that match the selected Kubernetes version are listed.
3. Optionally, expand Kubernetes Flags to pass custom flags through API Server Flags, Scheduler Flags, and Controller Manager Flags.
4. Under Add-Ons, enable what the cluster needs:

   • Core Add-ons are critical to the functionality of the cluster, and include CoreDNS and the Calico container network interface.
   • Load Balancers offers MetalLB.
   • Other Add-ons offers the Nvidia GPU Operator, for clusters built on GPU-backed flavors.
5. Select Submit.

PCD deploys only the MetalLB controllers on the cluster. You still have to configure MetalLB after the cluster is created. See Set up Load balancers using MetalLB.

Monitor the Deployment

The wizard shows a progress screen while the cluster is built. When the deployment finishes, select Finish to close the wizard.

Your cluster then appears in the cluster list. Select it to see its details, its nodes, and its add-ons.

Next Steps

• To connect kubectl to your new cluster, see Cluster Access - Kubeconfig.
• To add or remove worker nodes later, see Scale Clusters.
• To upgrade the cluster to a newer Kubernetes version, see Upgrade Clusters.

Related Pages

• Control Plane Types: how Managed and Colocated control planes differ, and how to choose between them
• Pre-requisites: what to have in place before you create a cluster
• Architecture and Technical Overview: the components behind a Kubernetes cluster
• Set up your Bare Metal Kubernetes Cluster: the same flow on physical nodes