Kubernetes setup

Clanker provides comprehensive Kubernetes cluster management across EKS, GKE, and kubeadm. You can create clusters, deploy applications, and query resources using natural language.

Supported platforms

Amazon EKS: Managed Kubernetes on AWS
Google GKE: Managed Kubernetes on GCP
Kubeadm: Self-managed clusters on EC2/GCE instances

Create your first EKS cluster

Create an Amazon EKS cluster with worker nodes:

clanker k8s create eks my-cluster --nodes 2 --node-type t3.small

Clanker will:

Generate an execution plan
Show estimated costs and resources
Prompt for confirmation
Create the EKS cluster and node group
Update your kubeconfig automatically

Example output:

=== Kubernetes Cluster Plan ===

Cluster: my-cluster (EKS)
Region: us-east-1
Kubernetes Version: 1.29

Resources:
  - EKS Cluster Control Plane
  - Managed Node Group (2 nodes)
  - Node Type: t3.small
  - VPC with public/private subnets
  - Security groups
  - IAM roles (cluster, node)

Estimated monthly cost: ~$75
  - EKS control plane: $73/month
  - 2x t3.small nodes: ~$30/month

Do you want to create this cluster? [y/N]: y

[eks] creating cluster control plane...
[eks] creating managed node group...
[eks] waiting for nodes to be ready...

=== Cluster Created Successfully ===

Name:       my-cluster
Status:     ACTIVE
Endpoint:   https://ABC123.eks.us-east-1.amazonaws.com
Version:    1.29

Nodes:
  - ip-10-0-1-45.ec2.internal (Ready)
  - ip-10-0-2-67.ec2.internal (Ready)

Kubeconfig: ~/.kube/config

Next steps:
  kubectl get nodes
  kubectl get pods -A

Use --plan to preview the cluster configuration without creating it:

clanker k8s create eks my-cluster --nodes 3 --plan

Create a GKE cluster

Create a Google Kubernetes Engine cluster:

clanker k8s create gke my-gke-cluster \
  --gcp-project my-project \
  --nodes 2 \
  --node-type e2-standard-2 \
  --gcp-region us-central1

GKE-specific options:

--preemptible: Use preemptible VMs for cost savings
--gcp-region: Specify GCP region (default: us-central1)
--version: Kubernetes version (default: GKE default)

Create a kubeadm cluster

Create a self-managed cluster using kubeadm on EC2:

clanker k8s create kubeadm my-kubeadm-cluster \
  --workers 2 \
  --key-pair my-key \
  --node-type t3.medium

Kubeadm clusters give you full control:

Manual CNI selection (Calico, Flannel)
Custom Kubernetes versions
Direct access via SSH
Lower cost than managed services

Kubeadm workflow:

[k8s] checking SSH key configuration...
[k8s] SSH key 'my-key' exists in region us-east-1
[k8s] private key: ~/.ssh/my-key.pem

=== Kubernetes Cluster Plan ===

Cluster: my-kubeadm-cluster (kubeadm)
Control Plane: 1x t3.medium
Worker Nodes: 2x t3.medium
CNI: Calico

[k8s] launching control plane instance...
[k8s] installing kubeadm, kubectl, kubelet...
[k8s] initializing cluster with kubeadm...
[k8s] installing Calico CNI...
[k8s] launching worker nodes...
[k8s] joining workers to cluster...

=== Cluster Created Successfully ===

Control Plane:
  master-node: 10.0.1.45 (Public: 54.234.123.45)

Workers:
  worker-1: 10.0.1.67 (Public: 54.234.123.67)
  worker-2: 10.0.1.89 (Public: 54.234.123.89)

Kubeconfig: ~/.kube/kubeadm-my-kubeadm-cluster.conf

SSH access:
  ssh -i ~/.ssh/my-key.pem ubuntu@54.234.123.45

List clusters

View all clusters of a specific type:

# List EKS clusters
clanker k8s list eks

# List GKE clusters
clanker k8s list gke --gcp-project my-project

# List kubeadm clusters
clanker k8s list kubeadm

Output:

=== EKS Clusters ===

Name:     my-cluster
Status:   ACTIVE
Region:   us-east-1
Version:  1.29
Endpoint: https://ABC123.eks.us-east-1.amazonaws.com
Workers:  2

Name:     prod-cluster
Status:   ACTIVE
Region:   us-east-1
Version:  1.28
Endpoint: https://DEF456.eks.us-east-1.amazonaws.com
Workers:  5

Deploy an application

Deploy a containerized application to your cluster:

clanker k8s deploy nginx --name my-nginx --port 80 --replicas 3

This creates:

A Deployment with 3 replicas
A Service of type LoadBalancer
Automatic pod scheduling across nodes

Deployment output:

=== Deployment Plan ===

Name: my-nginx
Image: nginx
Replicas: 3
Port: 80
Namespace: default
Service Type: LoadBalancer

Do you want to deploy this application? [y/N]: y

[k8s] deploying application...
deployment.apps/my-nginx created
service/my-nginx created

=== Deployment Successful ===

Endpoint: http://a1b2c3d4e5f6-1234567890.us-east-1.elb.amazonaws.com

Next steps:
  kubectl get pods -l app=my-nginx
  kubectl logs -f deployment/my-nginx
  kubectl scale deployment/my-nginx --replicas=5

Get cluster resources

Retrieve all cluster resources as JSON or YAML:

# Get resources from specific cluster (JSON)
clanker k8s resources --cluster my-cluster

# Get resources in YAML format
clanker k8s resources --cluster my-cluster -o yaml

# Get resources from all EKS clusters
clanker k8s resources

This returns comprehensive cluster state:

Nodes (capacity, status, labels)
Pods (all namespaces)
Services (ClusterIP, LoadBalancer, NodePort)
PersistentVolumes and PersistentVolumeClaims
ConfigMaps and Secrets
Deployments, StatefulSets, DaemonSets

Monitoring and debugging

View pod logs

# Get logs from a pod
clanker k8s logs my-pod

# Logs from specific container
clanker k8s logs my-pod -c my-container

# Follow logs in real-time
clanker k8s logs my-pod -f

# Last 100 lines
clanker k8s logs my-pod --tail 100

# Logs from last 1 hour
clanker k8s logs my-pod --since 1h

# With timestamps
clanker k8s logs my-pod --timestamps

# All containers in pod
clanker k8s logs my-pod --all-containers

# Previous container (after restart)
clanker k8s logs my-pod -p

Resource metrics

# Node metrics
clanker k8s stats nodes
clanker k8s stats nodes --sort-by cpu
clanker k8s stats nodes -o json

# Pod metrics  
clanker k8s stats pods
clanker k8s stats pods -n kube-system
clanker k8s stats pods -A  # All namespaces
clanker k8s stats pods --sort-by memory

# Specific pod metrics
clanker k8s stats pod my-pod -n production
clanker k8s stats pod my-pod --containers

# Cluster-wide metrics
clanker k8s stats cluster

Example metrics output:

NAME                CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%
ip-10-0-1-45        245m         12%    1456Mi          45%
ip-10-0-2-67        189m         9%     1123Mi          35%

Natural language queries

Use k8s ask for natural language Kubernetes queries:

# Basic queries
clanker k8s ask "how many pods are running"
clanker k8s ask "list all deployments and their replica counts"
clanker k8s ask "tell me the health of my cluster"

# With specific cluster
clanker k8s ask --cluster my-cluster "show me all pods"

# Resource metrics
clanker k8s ask "which pods are using the most memory"
clanker k8s ask "show node resource usage"
clanker k8s ask "top 10 pods by cpu usage"

# Troubleshooting
clanker k8s ask "show me recent logs from nginx"
clanker k8s ask "why is my pod crashing"
clanker k8s ask "get warning events from the cluster"

# Follow-up questions (maintains context)
clanker k8s ask "show me the nginx deployment"
clanker k8s ask "now show me its logs"

How k8s ask works:

LLM analyzes your question
Determines required kubectl operations
Executes operations in parallel
Synthesizes results into markdown response
Maintains conversation context per cluster

Cluster management

Get kubeconfig

# EKS
clanker k8s kubeconfig eks my-cluster

# GKE
clanker k8s kubeconfig gke my-gke-cluster --gcp-project my-project

# Kubeadm
clanker k8s kubeconfig kubeadm my-kubeadm-cluster

Delete clusters

# Delete EKS cluster
clanker k8s delete eks my-cluster

# Delete GKE cluster
clanker k8s delete gke my-gke-cluster --gcp-project my-project

# Delete kubeadm cluster (terminates all EC2 instances)
clanker k8s delete kubeadm my-kubeadm-cluster

Deleting a cluster is permanent and will remove all workloads, data, and configuration.

Advanced configurations

Custom EKS cluster

clanker k8s create eks production-cluster \
  --nodes 5 \
  --node-type m5.large \
  --version 1.29 \
  --profile prod-aws

GKE with preemptible nodes

clanker k8s create gke staging-cluster \
  --gcp-project my-project \
  --nodes 3 \
  --node-type e2-standard-4 \
  --preemptible \
  --gcp-region us-west1

Multi-zone kubeadm

clanker k8s create kubeadm multi-zone-cluster \
  --workers 3 \
  --node-type t3.large \
  --key-pair my-key \
  --version 1.28

Best practices

Use managed services

Prefer EKS/GKE for production workloads. They provide automatic upgrades, security patches, and HA control planes.

Right-size nodes

Start with smaller node types (t3.medium) and scale up based on actual usage. Monitor with k8s stats.

Plan before creating

Use --plan to review cluster configuration and costs before provisioning.

Namespace isolation

Deploy applications to dedicated namespaces for better organization and resource quotas.

Troubleshooting

Cluster creation fails

Common issues:Insufficient IAM permissions

# Verify your AWS user has EKS permissions
aws iam get-user --profile my-profile

VPC quota exceeded

# Check VPC limits
aws ec2 describe-account-attributes --attribute-names vpc-max-elastic-ips

Invalid instance type for region

# List available instance types
aws ec2 describe-instance-type-offerings --location-type availability-zone

Cannot connect to cluster

If kubectl commands fail:

# Verify kubeconfig is set
echo $KUBECONFIG

# Re-fetch kubeconfig
clanker k8s kubeconfig eks my-cluster

# Test connection
kubectl cluster-info

# Check AWS credentials
aws sts get-caller-identity --profile my-profile

Nodes not joining cluster

For kubeadm clusters:

# SSH to master node
ssh -i ~/.ssh/my-key.pem ubuntu@<master-ip>

# Check cluster status
kubectl get nodes

# View join token
sudo kubeadm token list

# SSH to worker and check logs
ssh -i ~/.ssh/my-key.pem ubuntu@<worker-ip>
sudo journalctl -u kubelet -f

Next steps

Monitoring resources

Set up comprehensive cluster monitoring

Kubernetes debugging

Debug pods, services, and networking issues

Multi-environment

Manage dev, staging, and prod clusters

Cost optimization

Optimize Kubernetes costs with right-sizing

Documentation Index

​Supported platforms

​Monitoring and debugging

​View pod logs

​Resource metrics

​Natural language queries

​Cluster management

​Get kubeconfig

​Delete clusters

​Advanced configurations

​Custom EKS cluster

​GKE with preemptible nodes

​Multi-zone kubeadm

​Best practices

Use managed services

Right-size nodes

Plan before creating

Namespace isolation

​Troubleshooting

​Next steps

Monitoring resources

Kubernetes debugging

Multi-environment

Cost optimization

Supported platforms

Monitoring and debugging

View pod logs

Resource metrics

Natural language queries

Cluster management

Get kubeconfig

Delete clusters

Advanced configurations

Custom EKS cluster

GKE with preemptible nodes

Multi-zone kubeadm

Best practices

Troubleshooting

Next steps