Skip to content

ResourceSupervisor#

The ResourceSupervisor is a namespace-scoped custom resource that enables self-service hibernation of workloads within a single namespace. It allows application teams or namespace owners to define schedules for scaling down and restoring Deployments and StatefulSets without requiring cluster-wide permissions.

✅ Only affects the namespace in which it is created.

Supported Workloads#

  • Deployment
  • StatefulSet

These resources are scaled to 0 replicas during sleep and restored to their original replica count during wake.

Ignored Namespaces#

While ResourceSupervisor is namespace-scoped and only acts within its own namespace, the Hibernation Operator still respects global exclusion rules. A namespace (including the one containing the ResourceSupervisor) will be ignored if it has:

  • The annotation:
hibernation.stakater.com/exclude: "true"

🔒 The operator never modifies workloads in excluded namespaces—even if a ResourceSupervisor exists there.

Supported Modes#

1. Hibernation with Cron Schedule (Sleep + Wake)#

Define both sleepSchedule and wakeSchedule to automatically cycle workloads on and off.

apiVersion: hibernation.stakater.com/v1beta1
kind: ResourceSupervisor
metadata:
  name: nightly-hibernation
  namespace: my-app-staging  # ← Must match target namespace
spec:
  schedule:
    sleepSchedule: "0 20 * * *"   # Sleep daily at 8:00 PM UTC
    wakeSchedule: "0 8 * * *"     # Wake daily at 8:00 AM UTC

🕒 Cron Format: Standard Unix cron (minute hour day month weekday). Timezone is UTC.

2. Permanent Sleep (Manual Wake)#

Omit wakeSchedule to keep workloads asleep indefinitely. Workloads will only wake when:

  • The ResourceSupervisor is deleted, or
  • A wakeSchedule is added later
apiVersion: hibernation.stakater.com/v1beta1
kind: ResourceSupervisor
metadata:
  name: pause-for-maintenance
  namespace: demo-env
spec:
  schedule:
    sleepSchedule: "0 12 * * *"   # Sleep today at noon UTC, never wake
    # wakeSchedule: omitted → stay asleep

💡 This is useful for temporary freezes (e.g., during security reviews or budget pauses).

3. Immediate Sleep (No Schedule)#

To sleep immediately, create a ResourceSupervisor with an empty schedule block:

apiVersion: hibernation.stakater.com/v1beta1
kind: ResourceSupervisor
metadata:
  name: sleep-now
  namespace: ci-preview-pr123
spec:
  schedule: {}
  • Workloads are scaled to 0 as soon as the CR is created.
  • They remain asleep until the CR is deleted.

🚀 Ideal for ephemeral environments (e.g., CI/CD preview namespaces).

Status Tracking#

The operator updates the CR’s status to reflect current state:

status:
  currentStatus: sleeping      # or "running", "error"
  nextReconcileTime: "2025-02-01T08:00:00Z"

Use kubectl describe or kubectl get -o yaml to monitor:

kubectl get resourcesupervisor nightly-hibernation -n my-app-staging -o jsonpath='{.status}'

Key Notes#

  • ❌ No argocd field: ResourceSupervisor does not support ArgoCD integration. Use ClusterResourceSupervisor for AppProject-based hibernation.
  • ❌ No namespaces field: It only affects its own namespace—you cannot target other namespaces.
  • ✅ Safe by default: Only Deployments and StatefulSets are scaled; all other resources are untouched.
  • ✅ Stateful restoration: Original replica counts are stored in the CR’s status and restored accurately—even after operator restarts.

Example: CI/CD Preview Namespace#

A CI pipeline creates a namespace pr-456 and immediately hibernates it to save costs:

# pr-456-hibernation.yaml
apiVersion: hibernation.stakater.com/v1beta1
kind: ResourceSupervisor
metadata:
  name: auto-sleep
  namespace: pr-456
spec:
  schedule: {}

Apply it:

kubectl apply -f pr-456-hibernation.yaml

Workloads sleep instantly. When the PR is merged, the pipeline deletes the namespace (or the CR), waking workloads briefly before cleanup.