Skip to main content

Build and operate

vClustervClusterAn open-source software product that creates and manages clusters within Kubernetes infrastructure. vCluster provides tenant isolation capabilities while reducing infrastructure costs.Related: Cluster, Control plane cluster lets you provision isolated clusters on your existing infrastructure without a separate physical cluster per customer. This section maps the common production architectures to concrete implementation paths. Choose the one that matches what you are building and follow it from initial design to a running, operated platform.

These build-and-operate guides differ from the service pages. A service page, such as Kubernetes as a service, covers one thing you offer your tenants. A build-and-operate guide covers how one kind of organization runs vCluster from design through Day 2 operations, combining several services.

Need to evaluate vCluster first?

Start with a quick start to prove the deployment model in your environment, then return here to plan production.

Decide whether tenants share hardware first

Shared and private nodes are a security boundary decision, locked at deployment time. Before you pick a path below, run through Decide whether tenants share hardware to confirm which one your offering requires.

What are you building?​

Find the goal that matches your platform, then choose the desired outcome. The linked production path takes you from architecture decisions through Day 2 operations.

Offer AI or Kubernetes infrastructure to customers​

Desired outcomeStart hereWorker-node guidance
Serve models through managed endpoints while your team operates the Kubernetes infrastructureInference providerPrivate nodes for dedicated or customer-controlled serving. Shared nodes may suit provider-owned models with API-only access
Give customers isolated, GPU-backed Kubernetes environments through your productAI cloud providerPrivate nodes for external customer workloads
Give each enterprise customer a dedicated cluster stack for node-level, regulated, or sovereign isolationSaaS and B2B providerPrivate nodes for node-level customer isolation

Build a platform for internal teams​

Desired outcomeStart hereWorker-node guidance
Standardize governed GPU environments across production, development, and experiment tiersEnterprise AI factoryPrivate nodes for production. Shared nodes for trusted development and experiment tiers
Give trusted engineering teams long-running, isolated Kubernetes environmentsTeam and CI clusters as a serviceShared nodes for trusted team workloads only
Give each CI/CD pipeline an ephemeral, isolated Kubernetes environment with automatic cleanupAdd CI clustersShared nodes for trusted pipeline workloads only

Operate compute across providers or locations​

Desired outcomeStart hereWorker-node guidance
Manage GPU capacity from multiple compute sources through one operations layerCompute aggregatorPrivate nodes for dedicated customer capacity at each site
Manage customer workloads at distributed edge sites from a central control planeEdge providerPrivate nodes for dedicated customer capacity at edge sites

Not sure where to start? Decide whether tenants share hardware explains how to choose between shared and private nodes. If no build-and-operate guide fits, review Architecture.

What production-ready means​

A production vCluster platform delivers:

  • Tenant isolationTenant IsolationThe capability to host multiple separate users, teams, or workloads on the same infrastructure while providing strong isolation between them. vCluster delivers tenant isolation through dedicated control planes, isolated resource namespaces, and optional private nodes per tenant.Related: Cluster: every customer or team sees only their own cluster and workloads. Node level isolation requires private nodes
  • Repeatable provisioning: new clusters deploy from a defined template, not from manual steps. When each environment also needs the same ordered set of applications, use Stacks to standardize that software layer
  • A worker node model matched to your isolation needs: Private nodes suit customers with cluster, scheduler, or workload execution access. Shared nodes suit trusted internal teams or a provider-operated offering where you own all workloads. A dedicated node pool scopes placement within shared nodes but doesn't provide private-node isolation. See Decide whether tenants share hardware.
  • Governed access: who can create, access, and administer clusters, enforced through Platform policies. Where separate organizations share the platform, a tenant is the outer boundary that determines what each one can use
  • Durable control planes: HA, data store, and backup procedures defined before customers depend on the system
  • Operational readiness: monitoring, upgrade, restore, and incident response procedures documented and tested

Connect quick start results to production paths​

Each quick start validates a specific deployment model. Use this table to connect what you proved to the production path that extends it.

If you completedYou have provenProduction paths to consider
Docker (vind)Local development, testing, or CI cluster behaviorUse vind to validate cluster behavior, then choose a production path for deployment.
Shared nodesClusters hosted on an existing Kubernetes clusterTeam and CI clusters as a service or Enterprise AI factory (shared tier)
Private nodesClusters with dedicated worker nodesInference provider, AI cloud provider, Enterprise AI factory (production tier), Compute aggregator, SaaS and B2B provider, or Edge provider
StandaloneControl plane cluster on bare metal or VMsInference provider, AI cloud provider, Compute aggregator, Enterprise AI factory (on-premises), or Edge provider

Day 2 operations reference​

Common operations that apply across all paths.

OperationRead next
Monitor Platform and cluster workloadsFleet Observability, other architectures
Back up and restore clustersSnapshots, restore, Velero
Back up and restore PlatformBack up Platform, Restore Platform, Platform database
Upgrade Platform and clustersUpgrade vCluster, upgrade Platform, lifecycle policy
Rotate certificatesCertificate rotation
Manage private worker nodesManage private nodes, Auto Nodes
Scale and recover the platformPlatform HA, multi-region Platform
Troubleshoot incidentsvCluster troubleshoot, debug commands, Platform troubleshooting