Skip to main content

​
The Challenge

Development and testing on Kubernetes is slow and expensive:
  • Shared dev clusters cause conflicts - Teams step on each other’s deployments
  • Slow environment provisioning - Waiting days for new clusters blocks productivity
  • Expensive infrastructure - Running dedicated clusters per developer or PR is costly
  • CI/CD bottlenecks - Limited test environments create queues and slow pipelines
  • Production parity is hard - Dev environments don’t match production configurations

​
How vCluster Solves It

vCluster provides lightweight, ephemeral Kubernetes clusters that:
  • Spin up in seconds - Create isolated environments on-demand
  • Run on shared infrastructure - Maximize density to reduce costs
  • Match production - Test with the same Kubernetes version and configurations
  • Integrate with CI/CD - Automate cluster creation and teardown in pipelines
  • Enable parallel testing - Run multiple test suites simultaneously without conflicts

​
Real-World Examples

​
CI/CD Pipeline Testing

Scanmetrix achieved 99% faster deployments by using vCluster for ephemeral test environments. Each PR gets its own isolated Kubernetes cluster.

​
Development Environments

Ada improved developer productivity 10x by giving each developer their own on-demand vCluster. Developers can experiment freely without affecting others.

​
Integration Testing

Run integration tests in parallel across multiple isolated clusters. Aussie Broadband achieved 99% faster provisioning for test environments.

​
Ephemeral Dev Clusters

Lightweight configuration optimized for fast startup: 
controlPlane:
  backingStore:
    database:
      embedded:
        enabled: true  # Use embedded SQLite for speed
  statefulSet:
    resources:
      requests:
        cpu: 100m
        memory: 128Mi
      limits:
        memory: 1Gi
  coredns:
    embedded: true  # Faster DNS resolution

sync:
  fromHost:
    nodes:
      enabled: false  # Use pseudo nodes for maximum density

policies:
  resourceQuota:
    enabled: true
    quota:
      requests.cpu: 4
      requests.memory: 8Gi
      count/pods: 50

​
CI/CD Testing

Automated creation and cleanup: 
controlPlane:
  statefulSet:
    persistence:
      volumeClaim:
        enabled: false  # No persistence needed for ephemeral clusters
  
sync:
  fromHost:
    nodes:
      enabled: false
    storageClasses:
      enabled: auto

policies:
  resourceQuota:
    enabled: true
    quota:
      requests.cpu: 8
      requests.memory: 16Gi
      limits.cpu: 12
      limits.memory: 24Gi

​
Integration Testing

Parallel test execution: 
controlPlane:
  distro:
    k8s:
      enabled: true
      version: v1.31.0  # Match production version
  backingStore:
    database:
      embedded:
        enabled: true

integrations:
  metricsServer:
    enabled: true  # For testing HPA and metrics

policies:
  networkPolicy:
    enabled: true  # Test network policies
    workload:
      publicEgress:
        enabled: true

​
Preview Environments

Per-branch environments for PR reviews: 
controlPlane:
  ingress:
    enabled: true
    host: pr-${PR_NUMBER}.preview.example.com
    annotations:
      cert-manager.io/cluster-issuer: letsencrypt-prod

sync:
  fromHost:
    nodes:
      enabled: false
  toHost:
    ingresses:
      enabled: true

integrations:
  certManager:
    enabled: true

​
Best Practices

​
1. Automate Cluster Lifecycle in CI/CD

Create and destroy clusters automatically: 
# .github/workflows/pr-preview.yml
name: PR Preview

on:
  pull_request:
    types: [opened, synchronize, closed]

jobs:
  preview:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      
      - name: Install vCluster CLI
        run: |
          curl -L -o vcluster "https://github.com/loft-sh/vcluster/releases/latest/download/vcluster-linux-amd64"
          chmod +x vcluster
          sudo mv vcluster /usr/local/bin/
      
      - name: Create vCluster
        if: github.event.action != 'closed'
        run: |
          vcluster create pr-${{ github.event.pull_request.number }} \
            --namespace pr-environments \
            --values .vcluster-ci.yaml \
            --connect=false
      
      - name: Deploy Application
        if: github.event.action != 'closed'
        run: |
          vcluster connect pr-${{ github.event.pull_request.number }} --namespace pr-environments
          kubectl apply -f ./k8s/
      
      - name: Delete vCluster
        if: github.event.action == 'closed'
        run: |
          vcluster delete pr-${{ github.event.pull_request.number }} --namespace pr-environments

​
2. Use Resource Quotas

Prevent runaway test environments: 
policies:
  resourceQuota:
    enabled: true
    quota:
      requests.cpu: 4
      requests.memory: 8Gi
      requests.storage: 50Gi
      count/pods: 50
      count/services: 20
      services.loadbalancers: 0  # Disable for dev environments

​
3. Match Production Kubernetes Version

Test against the same version used in production: 
controlPlane:
  distro:
    k8s:
      enabled: true
      version: v1.31.0  # Match production

​
4. Enable Fast Cleanup

Use labels for bulk deletion: 
# Create with labels
vcluster create test-${BUILD_ID} \
  --namespace ci-tests \
  --labels ci=true,build=${BUILD_ID}

# Clean up old test clusters
kubectl delete vcluster -n ci-tests -l ci=true,build!=${BUILD_ID}

​
5. Share Storage Classes

Reuse host cluster storage for faster provisioning: 
sync:
  fromHost:
    storageClasses:
      enabled: true  # Share host storage classes

​
6. Optimize for Fast Startup

controlPlane:
  backingStore:
    database:
      embedded:
        enabled: true  # Faster than external databases
  coredns:
    embedded: true  # Reduce pod count
  statefulSet:
    persistence:
      volumeClaim:
        enabled: false  # No persistence for ephemeral clusters
    probes:
      startupProbe:
        periodSeconds: 2  # Faster startup detection

​
7. Implement Naming Conventions

Make it easy to identify and clean up environments: 
# Development
vcluster create dev-${USER}-${FEATURE_NAME}

# CI/CD
vcluster create ci-pr-${PR_NUMBER}

# Testing
vcluster create test-${COMMIT_SHA}

​
8. Set Automatic Cleanup Policies

Delete inactive clusters automatically (requires vCluster Platform): 
# Via vCluster Platform
sleep:
  afterInactivity: 2h
  deleteAfter: 24h  # Auto-delete after 1 day of sleep

​
CI/CD Integration Examples

​
GitHub Actions

- name: Run Integration Tests
  run: |
    vcluster create test-${{ github.run_id }} -n ci-tests --wait
    vcluster connect test-${{ github.run_id }} -n ci-tests
    kubectl apply -f ./test-fixtures/
    make integration-test
    vcluster delete test-${{ github.run_id }} -n ci-tests

​
GitLab CI

integration_tests:
  stage: test
  script:
    - vcluster create test-$CI_PIPELINE_ID -n ci-tests --wait
    - vcluster connect test-$CI_PIPELINE_ID -n ci-tests
    - kubectl apply -f ./test-fixtures/
    - make integration-test
  after_script:
    - vcluster delete test-$CI_PIPELINE_ID -n ci-tests

​
Jenkins

pipeline {
  agent any
  stages {
    stage('Test') {
      steps {
        sh '''
          vcluster create test-${BUILD_NUMBER} -n ci-tests --wait
          vcluster connect test-${BUILD_NUMBER} -n ci-tests
          kubectl apply -f ./test-fixtures/
          make integration-test
        '''
      }
    }
  }
  post {
    always {
      sh 'vcluster delete test-${BUILD_NUMBER} -n ci-tests'
    }
  }
}

​
Cost Optimization

​
Maximize Density

Run dozens of virtual clusters on a single host cluster: 
sync:
  fromHost:
    nodes:
      enabled: false  # Pseudo nodes for maximum density

controlPlane:
  coredns:
    embedded: true  # Reduce pod overhead
  statefulSet:
    resources:
      requests:
        cpu: 50m
        memory: 64Mi

​
Sleep Inactive Clusters

For long-running dev environments, enable sleep mode (requires vCluster Platform): 
sleep:
  afterInactivity: 1h

​
Use Spot Instances

Run test clusters on cheap spot/preemptible nodes: 
kubectl label nodes spot-node-1 spot-node-2 workload=ci-tests

controlPlane:
  statefulSet:
    scheduling:
      nodeSelector:
        workload: ci-tests
      tolerations:
        - key: spot
          operator: Exists

​
Performance Tips

  1. Use embedded database for dev/test - Faster startup than external databases
  2. Disable persistence for CI/CD - Ephemeral clusters don’t need PVCs
  3. Use embedded CoreDNS - Reduces pod count and startup time
  4. Pre-warm host cluster - Keep base images cached on nodes
  5. Parallelize tests - Run multiple test suites in separate vClusters simultaneously

Build docs developers (and LLMs) love

Get started for freeTalk to us