Skip to main content

​
What is vCluster?

vCluster creates fully functional virtual Kubernetes clusters that run inside namespaces of a host cluster. Each virtual cluster has its own API server, runs on shared or dedicated infrastructure, and gives you flexible tenancy options—from simple namespaces to fully dedicated clusters.
Trusted by the industry: Over 40 million virtual clusters have been deployed by companies like Adobe, CoreWeave, Atlan, and NVIDIA.
Virtual clusters provide complete Kubernetes API isolation while sharing the underlying infrastructure of the host cluster. Think of them as lightweight Kubernetes clusters that give you all the power of a full cluster without the operational overhead.

​
Why Virtual Clusters?

Traditional Kubernetes multi-tenancy relies on namespaces, which have significant limitations. Virtual clusters solve these problems by providing:

Cluster-Scoped Resources

Use CRDs, namespaces, cluster roles, and other cluster-scoped resources without host cluster conflicts

True Isolation

Complete separation with dedicated control planes and API servers for each tenant

Cost Efficiency

Much cheaper than full clusters—runs as a single pod with shared host resources

Ease of Use

Created in seconds with a single command, works in any Kubernetes cluster

Cluster-Wide Permissions

Grant users admin access inside their vCluster while limiting host cluster permissions

Version Flexibility

Test different Kubernetes versions inside a single host cluster

​
Key Concepts

​
Control Plane Isolation

Each vCluster runs its own Kubernetes control plane components:
  • API Server: Dedicated API endpoint for complete API isolation
  • Controller Manager: Manages Kubernetes controllers within the virtual cluster
  • Data Store: Embedded etcd or external database (PostgreSQL, MySQL, RDS)
This isolation ensures that tenants interact only with their own virtual cluster, never directly with the host cluster.

​
Resource Syncing

vCluster uses bidirectional resource synchronization to connect virtual and host clusters:
1

Virtual Cluster Creation

Users create resources (pods, services, secrets) in the virtual cluster through its API server
2

Syncing to Host

The syncer component watches for resources and syncs them to the host cluster namespace
3

Physical Execution

Workloads run on the host cluster’s nodes, using its CNI, CSI, and infrastructure
4

Syncing from Host

Status updates and events flow back from host to virtual cluster automatically
This approach gives users a complete Kubernetes experience while leveraging the host’s platform stack.
By default, vCluster syncs pods, services, secrets, configmaps, and persistent volume claims. You can configure additional resources like ingresses, network policies, and custom resources.

​
Shared vs. Isolated Infrastructure

vCluster supports multiple deployment architectures with different levels of isolation:
ArchitectureNode IsolationCNI/CSI IsolationBest For
Shared Nodes❌❌Development, testing, maximum density
Dedicated Nodes✅❌Production workloads, compliance
Private Nodes✅✅Full isolation, GPU workloads
Standalone✅✅Bare metal, edge, no host cluster

​
Architecture Overview

vCluster follows a lightweight architecture that maximizes efficiency: The vCluster control plane runs as a single StatefulSet in a host namespace. Users interact with the virtual cluster’s API server, completely isolated from the host. The syncer ensures workloads execute on host infrastructure.
For more detailed architecture information, see the Architecture Overview section.

​
Main Benefits

​
Multi-Tenancy Without Compromise

Provide teams with full Kubernetes clusters instead of limited namespace access:
  • Admin privileges inside their vCluster
  • Ability to install cluster-scoped operators and CRDs
  • Complete isolation from other tenants
  • Minimal permissions required on the host cluster

​
Rapid Environment Provisioning

Create isolated Kubernetes environments in seconds: 
# Create a vCluster in under 30 seconds
vcluster create my-vcluster --namespace team-x
Perfect for:
  • CI/CD pipelines requiring clean Kubernetes environments
  • Development and testing workflows
  • Preview environments for pull requests
  • Temporary sandbox environments

​
Cost Optimization

Consolidate multiple teams or projects onto fewer clusters:
  • 70% cost reduction reported by Fortune 500 companies
  • Share expensive infrastructure (GPU nodes, storage, load balancers)
  • Sleep mode for inactive clusters to save resources
  • Higher density than traditional cluster-per-tenant approaches

​
GPU and AI/ML Workloads

Optimize GPU utilization across teams:
  • Maximize GPU utilization without sacrificing isolation
  • Support multiple schedulers (Karpenter, Volcano, YuniKorn)
  • Private nodes with full CNI/CSI control for AI workloads
  • Auto-scaling with Auto Nodes feature

​
Common Use Cases

GPU Cloud Providers

Launch managed Kubernetes for GPU customers with isolated, production-grade environments

Internal GPU Platform

Self-service GPU access for AI/ML teams with maximum utilization

Multi-Tenancy

Give teams full cluster access while maintaining security boundaries

Development & Testing

Rapid provisioning of isolated test environments for CI/CD

Bare Metal Kubernetes

Run Kubernetes on bare metal with zero VMs and strong isolation

Cost Savings

Consolidate clusters and reduce Kubernetes infrastructure costs by 70%

​
How vCluster Compares

vCluster offers a unique approach compared to other multi-tenancy solutions:
FeatureNamespacesvClusterSeparate Clusters
IsolationLimitedStrongComplete
Cluster-scoped resources❌✅✅
Admin permissions❌✅✅
CostLowLowHigh
Provisioning timeSecondsSecondsMinutes/Hours
Operational overheadLowLowHigh
Kubernetes versionSame as hostFlexibleAny
vCluster bridges the gap between namespaces and full clusters, giving you the best of both worlds.

​
What’s New

vCluster continues to evolve with cutting-edge features:
Tailscale-powered overlay networks and automated network isolation for hybrid infrastructures with vCluster VPN and Netris integration.
Run vCluster without a host cluster—deploy the control plane directly on bare metal or VMs for the highest level of isolation.
Karpenter-powered dynamic autoscaling for private nodes with automatic provisioning and deprovisioning.
External nodes with full CNI/CSI isolation join the virtual cluster directly with their own networking stack.
Multiple scheduler support for AI/ML workloads and fine-grained namespace synchronization.

​
Getting Started

Ready to create your first virtual cluster? Here’s what comes next:

Quickstart

Get your first vCluster running in under 5 minutes

Installation

Detailed installation instructions for all platforms

Architecture Deep Dive

Understand how vCluster works under the hood

Configuration

Learn how to configure vCluster for your needs

​
Community and Support

Join thousands of vCluster users:
  • GitHub: Star the project and contribute
  • Slack: Join 5K+ community members
  • Documentation: Comprehensive guides and references
  • Blog: Latest updates and best practices
vCluster requires Kubernetes 1.18+ on the host cluster and Helm 3.10.0+ for installation.