Skip to content

Holiday Schedule Replacement (replace + executionOverride)

Your company shuts down between Christmas and New Year. The office environment (office-plan: EKS node groups plus the reporting database) normally hibernates weeknights only — but during the shutdown week you want something completely different:

  1. Stay hibernated almost 24/7, waking only for a daily 10:00–12:00 maintenance window when internal automation runs.
  2. Run those cycles carefully — sequentially, tolerating individual failures, with minimal retries. Nobody is watching dashboards on December 26.

type: replace swaps the entire schedule; executionOverride swaps the strategy and behavior. The base plan goes back to normal automatically when the exception expires.

Goal

  • For 2026-12-24 → 2027-01-01: ignore the base schedule; hibernate 12:00 → 10:00 (next day), every day.
  • During exception cycles: Sequential strategy, BestEffort, retries: 1.
  • After Jan 1: resume the base weeknight schedule with zero manual steps.

Base Plan (unchanged)

apiVersion: hibernator.ardikabs.com/v1alpha1
kind: HibernatePlan
metadata:
  name: office-plan
  namespace: hibernator-system
spec:
  schedule:
    timezone: "America/New_York"
    offHours:
      - start: "20:00"
        end: "06:00"
        daysOfWeek: ["MON", "TUE", "WED", "THU", "FRI"]

  execution:
    strategy:
      type: Parallel
      maxConcurrency: 2

  behavior:
    mode: Strict
    retries: 2

  targets:
    - name: office-nodegroups
      type: eks
      connectorRef:
        kind: CloudProvider
        name: aws-production
      parameters:
        clusterName: office-cluster
        nodeGroups: []
        awaitCompletion:
          enabled: true
          timeout: "10m"

    - name: reporting-db
      type: rds
      connectorRef:
        kind: CloudProvider
        name: aws-production
      parameters:
        selector:
          instanceIds: [office-reporting]
        snapshotBeforeStop: true
        awaitCompletion:
          enabled: true
          timeout: "15m"

(Uses the aws-production CloudProvider from Dependency-Ordered Shutdown.)

The Exception

apiVersion: hibernator.ardikabs.com/v1alpha1
kind: ScheduleException
metadata:
  name: year-end-shutdown
  namespace: hibernator-system
spec:
  planRef:
    name: office-plan
  type: replace
  validFrom: "2026-12-24T00:00:00Z"
  validUntil: "2027-01-01T23:59:59Z"
  windows:
    - start: "12:00"
      end: "10:00"
      daysOfWeek: ["MON", "TUE", "WED", "THU", "FRI", "SAT", "SUN"]
  executionOverride:
    strategy:
      type: Sequential
    behavior:
      mode: BestEffort
      retries: 1

kubectl apply -f year-end-exception.yaml

One window, not two

"Off all day except 10:00–12:00" is a single window: start: "12:00", end: "10:00". Because end < start, the window crosses midnight — hibernation runs continuously from noon to 10 AM the next morning, every day. No gaps, no midnight edge cases.

How It Works

  • type: replace — the base schedule is ignored while the exception is active. Only the exception windows drive cycles.
  • executionOverride is a FULL replacement of strategy and behavior for exception cycles. Here: targets run one at a time (Sequential), a single failure does not abort the cycle (BestEffort), and each target gets only one retry.
  • Omitting a field inside executionOverride keeps the base behavior for that aspect (e.g., no strategy = base strategy).
  • executionOverride is only valid on extend and replace — not on suspend.
  • When validUntil passes, the exception transitions to Expired and the base schedule resumes by itself. Cycle intent for any in-flight cycle stays locked via the plan snapshot (see Event-Mode Overrides).

Daily timeline during the shutdown week:

Time What happens
00:00–10:00 Hibernated (window active since yesterday noon)
10:00 Window ends → WakingUp → Active
10:00–12:00 Awake — automation does its work
12:00 Window opens → Hibernating → Hibernated
12:00–24:00 Hibernated, straight through to next day's 10:00 wakeup

Verification

# Watch the exception lifecycle: Pending → Active → Expired
kubectl get schedex year-end-shutdown -n hibernator-system

# Confirm the override is driving current cycles
kubectl get hibernateplan office-plan -n hibernator-system \
  -o jsonpath='{.status.appliedExceptionOverride}{"\n"}{.status.planSnapshot.execution.strategy.type}{"\n"}'
# year-end-shutdown
# Sequential

# After expiry — exception history is preserved on the plan
kubectl get hibernateplan office-plan -n hibernator-system \
  -o jsonpath='{.status.exceptionReferences}' | jq

On the first weekday after January 1, verify the base schedule has resumed (plan follows the 20:00→06:00 MON–FRI window again).

Variations

  • Full-stop holiday (zero awake time) — a single 00:00–23:59 window for all seven days keeps everything down for the entire period; see the holiday example.
  • Only soften behavior, keep base schedule — use type: extend with an executionOverride and no extra practical windows... or better, decide which schedule you actually want first; replace exists precisely so you do not have to reason about unions.
  • Ramp into the holiday — compose this replace with an extend for the week before it: see Composing Multiple Exceptions.

Next Steps