Skip to main content
Version: v0.36

Sleep tenant clusters

Enterprise
Available in these plansFreeDevProdScale
Auto Sleep

Free, Dev, Prod, and Scale are vCluster Platform license plans. Open source does not need a license or a Platform connection. See Compare open source and free tiers.

Sleep reduces the compute resources that idle tenant clusters consume. You can pause and resume a tenant clusterTenant clusterA fully isolated Kubernetes environment provisioned for a single tenant. Each tenant cluster has its own API server, controller manager, and resource namespace, backed by a virtualized control plane hosted on a control plane cluster. From the tenant's perspective it behaves exactly like a standard Kubernetes cluster.Related: Control plane cluster, Tenant cluster on demand, or configure auto sleep to respond to inactivity, schedules, and supported HTTP traffic.

Choose a sleep method​

MethodBehaviorUse when
Manual pause and resumeThe vcluster CLI pauses or resumes the tenant cluster when you run a command.You need an on-demand sleep trigger instead of inactivity or schedules.
Standalone auto sleepThe tenant cluster scales down its workloads but keeps its control plane running to detect activity and wake the workloads.The tenant cluster isn't managed by vCluster Platform.
Platform-managed auto sleepThe platform agent manages sleep and wakeup and can scale down the tenant cluster control plane.You need control plane sleep, Platform-proxied activity detection, Platform web-routing integrations, or auto-delete.

When a tenant cluster with auto sleep connects to a compatible platform agent, the agent takes over sleep management automatically. See How auto sleep is managed for the version requirements and the workloads-only exception.

Pre-production workloads

Auto sleep is intended for pre-production workloads. Sleeping interrupts workloads, and recreated Pods lose temporary filesystem data.

What happens during standalone auto sleep​

Without a platform agent, the tenant cluster control plane remains running. Native auto sleep scales down and restores the following workload types:

  • Deployments
  • StatefulSets
  • ReplicaSets
  • ReplicationControllers
  • DaemonSets, which sleep by using a node selector that doesn't match a node

Native auto sleep deletes Pods that don't have a workload controller. Because it deletes the Pod object from the tenant cluster API, wakeup doesn't restore these Pods.

You can exempt selected workloads from auto sleep. For example, keep an ingress controller running so it can receive traffic that wakes the tenant cluster.

What happens during platform-managed auto sleep​

The platform sleeps the tenant cluster's namespaces on the control plane cluster. It scales down Deployments, StatefulSets, and standalone ReplicaSets, prevents new physical Pods from starting, and can scale down the tenant cluster control plane.

The tenant cluster's Kubernetes API objects remain in its data store. After the control plane and syncerSyncerA component in vCluster that synchronizes resources between the tenant cluster and the control plane cluster, enabling tenant clusters to function while maintaining isolation.Related: vCluster, Tenant cluster wake, vCluster recreates physical workload Pods, including Pods without a Kubernetes workload controller. Recreated Pods lose temporary filesystem data and can receive new IP addresses.

What wakes a tenant cluster​

Depending on the deployment and configuration, the following activity can wake a tenant cluster or reset its inactivity timer:

  • Kubernetes API requests routed through vCluster Platform
  • Tenant cluster API requests handled by standalone auto sleep
  • Requests to supported Ingress, Istio, or Gateway API endpoints
  • A configured wakeup schedule
  • A manual wakeup or CLI connection

HTTP activity detection depends on the routing integration. Safe activity-based auto sleep with Gateway API requires a controller that supports HTTPRoute request mirroring.

Next steps​