Deploying Kubernetes Clusters

This guide covers creating and managing a fully managed Kubernetes cluster inside your tenant. PaaS Cloud provisions the control plane as containers and runs the worker nodes as virtual machines, so you get Kubernetes as a Service without managing any servers yourself.

Prerequisites

• Active MetaCentrum membership
• Logged into https://chloe.platform.cloud.e-infra.cz/
• A selected tenant with Kubernetes support enabled and sufficient resource quota

About Tenant Kubernetes Clusters

• The control plane runs as containers on the platform — you never manage master nodes.
• Worker nodes are virtual machines, sized using VM instance types.
• The Kubernetes version of your cluster is independent of the platform. You can choose versions from 1.28 to 1.33.
• Once created, the cluster behaves like any standard Kubernetes cluster — use kubectl, Helm, or your own CI/CD to deploy workloads into it.

Create a Cluster (Dashboard)

1. Open the Marketplace

• In the dashboard, switch to the tenant that will own the cluster using the tenant selector.
• Navigate to the Marketplace and select Kubernetes.

2. Configure the Cluster

Fill in the deployment form:

• Name — A unique identifier within the tenant. It cannot be changed after creation.
• Replicas — The number of worker nodes.
• Resource preset — The size of each worker node VM.
• Version — The Kubernetes version (1.28–1.33).

3. Deploy and Verify

• Click Submit.
• Open the cluster’s detail page and wait for its status to become Ready. Provisioning the worker VMs can take several minutes.

4. Download the Kubeconfig

• On the cluster’s detail page, open the Secrets section.
• Download the kubeconfig file to connect to your new cluster.

Create a Cluster (kubectl)

If you prefer to work declaratively, apply a Kubernetes resource with your tenant kubeconfig.

apiVersion: apps.cozystack.io/v1alpha1
kind: Kubernetes
metadata:
  name: mycluster
  namespace: tenant-myteam        # your tenant namespace
spec:
  minReplicas: 2                  # number of worker nodes
  resourcesPreset: small          # size of each worker node VM
  resources: {}                   # explicit cpu/memory — overrides resourcesPreset if set
  version: "1.32"                 # Kubernetes version (1.28–1.33)

kubectl --kubeconfig tenant-myteam.kubeconfig apply -f kubernetes-mycluster.yaml

Check the status and wait until READY is True:

kubectl --kubeconfig tenant-myteam.kubeconfig get kubernetes
kubectl --kubeconfig tenant-myteam.kubeconfig describe kubernetes mycluster

Connect to Your Cluster

Once you have the kubeconfig, use it like any other:

kubectl --kubeconfig mycluster.conf get nodes
kubectl --kubeconfig mycluster.conf get pods -A

Scale the Worker Nodes

Increase or decrease the number of worker nodes at any time.

• Dashboard — Open the cluster’s detail page, edit the Replicas value, and submit.
• kubectl — Edit minReplicas in the manifest and reapply, or edit the resource in place:

kubectl --kubeconfig tenant-myteam.kubeconfig edit kubernetes mycluster

New worker nodes are provisioned as VMs and join the cluster automatically once ready.

Deploy Applications to the Cluster

The tenant cluster is a standard Kubernetes cluster — deploy to it with kubectl, Helm, or a CI/CD pipeline.

Managed PaaS applications (such as PostgreSQL, MariaDB, or Redis) deployed in the same tenant are reachable from inside the cluster by their internal service names, for example:

mariadb-mydb.tenant-myteam.svc.cluster.local:3306
postgres-mydb-rw.tenant-myteam.svc.cluster.local:5432

This lets your cluster workloads use managed databases without exposing them externally. See Deploying Applications to provision them.

Delete a Cluster

• Dashboard — Open the cluster’s detail page and use the delete action.
• kubectl:

kubectl --kubeconfig tenant-myteam.kubeconfig delete kubernetes mycluster

Next Steps

• Deploying Applications — Provision managed databases and services your cluster can use.
• Using kubectl — Retrieve your tenant kubeconfig and authenticate.
• Exposing Services — Publish workloads running in your cluster to the internet.
• VM Instance Types — Reference for worker node sizing.