Dry Run Guide

Dry run mode lets you observe what Hybernate would do without it taking any action. The operator evaluates all policies, emits events, and updates predictions, but never modifies your workloads.

Enabling Dry Run

Per Workload

apiVersion: hybernate.io/v1alpha1
kind: ManagedWorkload
metadata:
  name: my-api
  namespace: staging
spec:
  target:
    kind: Deployment
    name: my-api
  idlePolicy:
    cpuIdleThreshold: 10
    memoryIdleThreshold: 10
    gracePeriod: "5m"
  prediction:
    confidence: 85
  dryRun: true  # Observe only

Via WorkloadPolicy (for auto-managed workloads)

apiVersion: hybernate.io/v1alpha1
kind: WorkloadPolicy
metadata:
  name: staging-policy
  namespace: staging
spec:
  mode: auto-manage
  dryRun: true  # Auto-created ManagedWorkloads inherit this

What Happens in Dry Run

The operator runs its full evaluation pipeline:

Action	Dry Run Behavior
Idle detection	Signals are checked, grace period is tracked, and events are emitted, but the workload is not paused
Scaling	Forecast engine proposes targets and constraints are evaluated, but replicas are not changed
Pause expiry	Expiry is detected, but the workload is not resumed or destroyed
Cost tracking	Costs are accumulated normally (resource usage is real regardless of management)
Prediction engine	Data points are observed and confidence builds normally
Events	All events are emitted with a [DRY RUN] prefix
Status	Phase and conditions update to reflect what would happen

Observing Dry Run Results

Events

kubectl describe managedworkload my-api -n staging

Look for events like:

[DRY RUN] Idle confirmed — would pause workload (grace period elapsed, all signals confirm)
[DRY RUN] Scale — would scale from 5 to 3 replicas (prediction: low demand)

Status

The status reflects the evaluated state:

kubectl get managedworkload my-api -n staging -o yaml

status:
  phase: Running  # Stays Running because no action was taken
  prediction:
    dailyPhase: DailyActive
    dailyConfidence: 87
    weeklyPhase: Observing
    weeklyConfidence: 0

Recommended Workflow

1. Deploy with dryRun: true. Observe events and status for a few days.
2. Check prediction confidence. Wait until the forecast engine reaches DailyActive and confidence exceeds your threshold.
3. Review events. Confirm the operator would have made the right decisions.
4. Disable dry run. Flip to false to enable automation.

kubectl patch managedworkload my-api -n staging \
  --type merge -p '{"spec":{"dryRun":false}}'

When to Use Dry Run

• First time deploying a ManagedWorkload
• After changing idle or scale policies
• When onboarding a new namespace via WorkloadPolicy
• Before moving from suggest to auto-manage mode
• In production environments where you want to validate before acting