Multi-Cloud Kubernetes Operator

Guide for provisioning and managing Kubernetes clusters on budget-friendly cloud providers, optimized for hackathon projects and learning environments.

Supported Providers

Provider	Min Cost/Month	Free Tier	Setup Time	Complexity	Best For
OCI OKE	$0	Yes (Always Free)	30-60 min	High	Zero-cost learning
Hetzner K3s	€4-8 ($4.50-9)	No	5-10 min	Medium	Best price/performance
DigitalOcean DOKS	$12+	$200 credit	3-5 min	Low	Fastest setup

Quick Provider Selection

Decision flowchart:

Need truly free?
├─ Yes → Oracle OKE (but allow 1-2 hours for setup)
└─ No, can pay → ┐
                  ├─ Need fastest setup? → DigitalOcean DOKS
                  └─ Need cheapest paid? → Hetzner K3s

Hackathon recommendations:

24-48 hour hackathon: DigitalOcean (speed > cost)
Week-long hackathon: Hetzner (best value)
Multi-week project: Oracle OKE (free but complex)
Team learning: DigitalOcean (consistent experience)

Provider-Specific Guides

Each provider has detailed documentation in the references/ directory:

DigitalOcean DOKS - Easiest setup, $200 free credit
Hetzner K3s - Best price/performance, €4-8/month
Oracle OKE - Truly free tier, complex networking

Universal Kubernetes Operations

These commands work identically across all providers once kubectl is configured:

Context Management

# List available contexts
kubectl config get-contexts

# Switch context
kubectl config use-context <context-name>

# View current context
kubectl config current-context

Basic Operations

# View cluster info
kubectl cluster-info
kubectl get nodes
kubectl get pods --all-namespaces

# Deploy application
kubectl create deployment nginx --image=nginx:latest
kubectl expose deployment nginx --port=80 --type=LoadBalancer

# Scale deployment
kubectl scale deployment nginx --replicas=3

# View logs
kubectl logs <pod-name>
kubectl logs -f <pod-name>  # Follow logs

# Delete resources
kubectl delete deployment nginx
kubectl delete service nginx

Helm Package Management

Commands verified via Context7 (/websites/helm_sh):

# Add repository
helm repo add bitnami https://charts.bitnami.com/bitnami

# Update repositories
helm repo update

# Search charts
helm search repo bitnami

# Install chart
helm install my-release bitnami/mysql

# List releases
helm list

# Uninstall release
helm uninstall my-release

Cost Management Best Practices

Pre-Deployment Checklist

Estimated cluster runtime (hours/days)
Calendar reminder set for deletion
Team knows teardown procedure
Backup/export strategy for data

During Hackathon

# Monitor cluster resources
kubectl top nodes
kubectl top pods --all-namespaces

# Avoid expensive resources:
# - LoadBalancer services (use NodePort for dev)
# - Large persistent volumes
# - High replica counts

Post-Hackathon Cleanup

Critical: Delete clusters immediately after hackathon ends.

# DigitalOcean
doctl kubernetes cluster delete <cluster-name> --force

# Hetzner (requires config file, protect_against_deletion: false)
hetzner-k3s delete --config cluster_config.yaml

# Oracle OKE
oci ce node-pool delete --node-pool-id <id> --force
oci ce cluster delete --cluster-id <id> --force

Common Failures & Solutions

Authentication Issues

DigitalOcean:

# Symptom: "invalid token"
# Solution:
doctl auth init  # Re-enter token
# Or regenerate at: https://cloud.digitalocean.com/account/api/tokens

Hetzner:

# Symptom: "API token invalid"
# Solution:
curl -H "Authorization: Bearer $HETZNER_TOKEN" \
  'https://api.hetzner.cloud/v1/server_types'
# If fails, regenerate token in Hetzner Console

Oracle OCI:

# Symptom: "NotAuthenticated"
# Solution:
oci setup config  # Reconfigure
oci iam region list  # Verify

Connection Issues

kubectl can't connect:

# Refresh kubeconfig
# DigitalOcean:
doctl kubernetes cluster kubeconfig save <cluster-name>

# Hetzner:
export KUBECONFIG=$(pwd)/kubeconfig

# Oracle:
oci ce cluster create-kubeconfig --cluster-id <id> --file ~/.kube/config

# Verify context
kubectl config current-context
kubectl get nodes -v=6  # Verbose debugging

Provisioning Stuck

Cluster stuck "provisioning":

Wait 10-15 minutes (normal for OKE, DOKS)
Check provider console/dashboard
If stuck >15 minutes, delete and recreate
For OCI: Check Always Free quota limits

Nodes not joining:

# Check node status
kubectl get nodes
kubectl describe node <node-name>

# Provider-specific debugging:
# - DigitalOcean: Check droplet status in console
# - Hetzner: SSH into node, check `systemctl status k3s`
# - Oracle: Check compute instance status, security lists

Resource Limits

Quota exceeded:

DigitalOcean: Contact support or upgrade account
Hetzner: No limits on standard instances
Oracle: Always Free has strict limits, try different region

Multi-Cloud Portability Patterns

Application Deployment (provider-agnostic)

# deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: webapp
spec:
  replicas: 2
  selector:
    matchLabels:
      app: webapp
  template:
    metadata:
      labels:
        app: webapp
    spec:
      containers:
      - name: webapp
        image: nginx:latest
        ports:
        - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: webapp
spec:
  selector:
    app: webapp
  ports:
  - port: 80
    targetPort: 80
  type: LoadBalancer  # Works on all providers

Deploy:

kubectl apply -f deployment.yaml
# Same command works on DOKS, K3s, OKE

Storage Considerations

Each provider has different storage classes:

# List available storage classes
kubectl get storageclass

# DigitalOcean: digitalocean-block-storage
# Hetzner: hcloud-volumes
# Oracle: oci, oci-bv

# Use default for portability:
kubectl get storageclass | grep default

Load Balancer Costs

Cost implications:

DigitalOcean: $12/month per LoadBalancer
Hetzner: €5.39/month per LoadBalancer
Oracle: 1 free 10Mbps LB, then paid

Cost-saving alternative (dev/test):

# Use NodePort instead of LoadBalancer
kubectl expose deployment webapp --port=80 --type=NodePort

# Access via node IP:port
kubectl get nodes -o wide
kubectl get svc webapp  # Check NodePort (e.g., 30080)
# Access: http://<node-ip>:<node-port>

Provider-Specific Notes

DigitalOcean DOKS

✅ Pros: Managed control plane, automatic upgrades, easy setup ⚠️ Cons: Higher cost, minimum $12/month 📖 Full guide: references/digitalocean-doks.md

Hetzner K3s

✅ Pros: Lowest cost, fast provisioning, K3s simplicity ⚠️ Cons: DIY management, requires config file, European regions only 📖 Full guide: references/hetzner-k3s.md

Oracle OKE

✅ Pros: Truly free tier, managed Kubernetes ⚠️ Cons: High setup complexity, networking prerequisites, capacity issues 📖 Full guide: references/oracle-oke.md

Documentation Grounding

All CLI commands in this skill are verified via Context7 MCP server documentation:

doctl: /digitalocean/doctl (Context7)
hetzner-k3s: /vitobotta/hetzner-k3s (Context7)
oci-cli: /oracle/oci-cli (Context7)
helm: /websites/helm_sh (Context7)
kubectl: K3s documentation (Context7)

Commands marked "⚠️ UNVERIFIED" require checking official documentation:

OCI OKE Documentation

Safety Warnings

⚠️ ALWAYS:

Set calendar reminders to delete clusters
Verify deletion in provider console
Check for orphaned resources (load balancers, volumes)
Monitor costs daily during multi-day hackathons

⚠️ NEVER:

Hardcode credentials in config files
Commit API tokens to version control
Leave clusters running after hackathon
Use production credentials for testing

Quick Reference Commands

Provider CLI verification:

# Check installed CLIs
doctl version
hetzner-k3s version
oci --version
kubectl version --client
helm version

# Check authentication
doctl account get
curl -H "Authorization: Bearer $HETZNER_TOKEN" https://api.hetzner.cloud/v1/server_types
oci iam region list

Universal cleanup:

# List all clusters
doctl kubernetes cluster list
# (Hetzner: check Hetzner Console)
oci ce cluster list --compartment-id <id>

# Delete pattern
<provider-delete-command> <cluster-name> --force

Additional Resources

Official Documentation:

Cost Calculators:

multi-cloud-kubernetes-operator

SKILL.md.claude/skills/multi-cloud-kubernetes-operator/SKILL.md