clouder k8s

Manage Kubernetes clusters — create clusters, manage node pools, and configure kubeconfig.

Commands

clouder k8s create

Create a Kubernetes cluster.

clouder k8s create <name>

Argument	Description
name (required)	Name for the Kubernetes cluster

Example:

clouder k8s create my-cluster

This provisions the necessary VMs and bootstraps a Kubernetes cluster using kubeadm. See the Kubeadm documentation for details on the bootstrap process.

---

clouder k8s ls

List Kubernetes clusters.

clouder k8s ls

Shows all clusters managed by Clouder in the current context.

---

clouder k8s kubeconfig

Download the kubeconfig for a Kubernetes cluster.

clouder k8s kubeconfig <name>

Argument	Description
name (required)	Cluster name

Example:

clouder k8s kubeconfig my-cluster

Downloads the kubeconfig file and saves it locally.

---

clouder k8s use

Set a cluster's kubeconfig as the default kubectl context.

clouder k8s use <name>

Argument	Description
name (required)	Cluster name

Example:

clouder k8s use my-cluster

# Verify
kubectl get nodes

---

clouder k8s create-nodepool

Create a node pool in a Kubernetes cluster.

clouder k8s create-nodepool <cluster_name> <nodepool_name> [OPTIONS]

Argument	Description
cluster_name (required)	Cluster to add the node pool to
nodepool_name (required)	Name for the new node pool

Option	Short	Default	Description
--flavor	-f	b2-15	Node flavor / VM size
--min		3	Minimum number of nodes
--desired		3	Desired number of nodes
--max		10	Maximum number of nodes
--roles			Comma-separated role names
--variant		default	Variant for the pool
--xpu		cpu	Compute type: cpu, gpu-cuda, qpu

Examples:

# System node pool
clouder k8s create-nodepool my-cluster \
    system --flavor Standard_D4s_v5 \
    --min 1 --desired 2 --max 10 \
    --roles system --xpu cpu

# GPU node pool for Jupyter
clouder k8s create-nodepool my-cluster \
    jupyter-gpu --flavor Standard_NC6s_v3 \
    --min 0 --desired 1 --max 5 \
    --roles jupyter --variant medium --xpu gpu-cuda

# CPU node pool for Jupyter
clouder k8s create-nodepool my-cluster \
    jupyter-cpu --flavor Standard_B4ms \
    --min 0 --desired 3 --max 10 \
    --roles jupyter --variant small --xpu cpu

---

clouder k8s update-nodepool

Update a node pool in a Kubernetes cluster (e.g., scale up or down).

clouder k8s update-nodepool <cluster_name> <nodepool_name> [OPTIONS]

Argument	Description
cluster_name (required)	Cluster name
nodepool_name (required)	Node pool name

Option	Default	Description
--min	0	Minimum nodes
--desired	0	Desired nodes
--max	10	Maximum nodes

Examples:

# Scale down to zero (cost saving)
clouder k8s update-nodepool my-cluster \
    jupyter-gpu --min 0 --desired 0 --max 5

# Scale up
clouder k8s update-nodepool my-cluster \
    system --min 2 --desired 4 --max 10

---

Example: Full Cluster Workflow

# 1. Create the cluster
clouder k8s create my-cluster

# 2. Get kubeconfig and set as default
clouder k8s kubeconfig my-cluster
clouder k8s use my-cluster

# 3. Create node pools
clouder k8s create-nodepool my-cluster \
    system --flavor Standard_D4s_v5 \
    --min 1 --desired 2 --max 10 \
    --roles system --xpu cpu

clouder k8s create-nodepool my-cluster \
    jupyter-gpu --flavor Standard_NC6s_v3 \
    --min 0 --desired 1 --max 5 \
    --roles jupyter --xpu gpu-cuda

# 4. Verify
kubectl get nodes

# 5. Scale GPU nodes to zero when idle
clouder k8s update-nodepool my-cluster \
    jupyter-gpu --desired 0

# 6. List clusters
clouder k8s ls