Skip to content

Suspend Scheduled Shutdown for Testing

Your base plan hibernates a service every night from 20:00 to 07:00 — reliable, cost-effective, and automated.

Tonight is different: the QA team is running a long load-test suite that must run until morning. You do not want to edit the base plan (it is shared infrastructure-as-code), and you do not want to remember to re-enable it tomorrow. A suspend exception carves out exactly one night.

Goal

  • Keep the base plan untouched and reusable after the test.
  • Suppress the entire Wednesday-night cycle (20:00 → 07:00) so the environment stays awake.
  • Prevent an accidental hibernation start just before the test window with leadTime.

Base Plan

apiVersion: hibernator.ardikabs.com/v1alpha1
kind: HibernatePlan
metadata:
  name: nightly-service
  namespace: hibernator-system
spec:
  schedule:
    timezone: "Asia/Jakarta"
    offHours:
      - start: "20:00"
        end: "07:00"
        daysOfWeek: ["MON", "TUE", "WED", "THU", "FRI", "SAT", "SUN"]
  execution:
    strategy:
      type: Parallel
  behavior:
    mode: Strict
    retries: 2
  targets:
    - name: service-nodegroups
      type: eks
      connectorRef:
        kind: CloudProvider
        name: aws-dev
      parameters:
        clusterName: service-cluster
        nodeGroups: []
        awaitCompletion:
          enabled: true

The Exception

apiVersion: hibernator.ardikabs.com/v1alpha1
kind: ScheduleException
metadata:
  name: qa-test-night
  namespace: hibernator-system
spec:
  planRef:
    name: nightly-service
  type: suspend
  validFrom: "2026-08-12T00:00:00Z"
  validUntil: "2026-08-13T23:59:59Z"
  leadTime: "30m"
  windows:
    - start: "20:00"
      end: "07:00"
      daysOfWeek: ["WED"]

kubectl apply -f qa-test-exception.yaml

How It Works

  • type: suspend — prevents hibernation during the specified windows. The base schedule is carved out, not replaced.
  • leadTime: "30m" — no new hibernation cycle may start within 30 minutes before the suspension window begins. Here: no shutdown starts from 19:30 onward on Wednesday, so the test runner is not interrupted mid-setup.
  • The window 20:00 → 07:00 on Wednesday matches the base window exactly — the entire night is suppressed. Thursday night resumes normally with zero manual steps.
  • The exception expires automatically at validUntil and transitions to Expired.

LeadTime is optional

For an emergency stop (active incident), omit leadTime so suppression takes effect immediately. See the emergency example.

Verification

# Exception state: Pending → Active → Expired
kubectl get schedex qa-test-night -n hibernator-system

# On Wednesday at 20:00, confirm the plan stays Active
kubectl get hibernateplan nightly-service -n hibernator-system \
  -o jsonpath='{.status.phase}'
# Active (not Hibernating)

# Check exception history on the plan
kubectl get hibernateplan nightly-service -n hibernator-system \
  -o jsonpath='{.status.exceptionReferences}' | jq

Variations

  • Partial evening only — if the team only needs until 23:59, use end: "23:59". After midnight the base schedule resumes and the plan hibernates 00:00–07:00.
  • Recurring testing window — e.g., every Tuesday and Thursday: add both days to daysOfWeek and extend validFrom/validUntil to cover the recurring period.
  • Suspend without leadTime — for unplanned incidents where you need immediate suppression.
  • Compose with an extend — a suspend and an extend can coexist on the same plan: see Composing Multiple Exceptions.

Next Steps