Skip to main content

How It Works

Cluster Agent​

A lightweight agent runs inside each cluster:

  • Monitors workloads and schedules.
  • Executes scale-down/scale-up actions.
  • Requires access to metrics server.

Scheduling​

  • Define a recurring 7-day calendar (nights, weekends, custom).
  • Assign workloads with the ZombieMode label.
  • At the scheduled time, replicas scale down to zero (or minimal value).

Shutdown Behavior​

  • Ingress and TLS remain intact.
  • Requests hit services but return 503 (no pods running) - this indicates Zombie Mode is working!

Wake-up Behavior​

  • Replicas restore automatically on schedule.
  • CI/CD updates are applied even while asleep.
  • Manual overrides or snoozes can wake workloads early.

Deterministic Savings​

A nights + weekends schedule equals ~68% of uptime removed, i.e. ~68% cost savings on compute.