Skip to content
Docs
Release candidateThese docs are for Napwake 0.2.0-rc.9, a release candidate, and may change before it's released. Read the docs for the latest release.

Napwake docs

Non-prod environments that sleep when nobody's using them, and wake when someone does.

Pauses idle Kubernetes workloads in staging, preview and dev, and wakes them on the first request.

It never deletes a workload or its data.

InstallQuickstart
devasleepstagingwaking uppreviewawake
A request has reached staging while it was asleep. The doorman holds it at the door while staging wakes, then lets it through. Dev stays asleep. Preview is already awake.

See what’s idle first

Section titled “See what’s idle first”

A team that works 10 hours a day, five days a week, uses staging 50 of 168 hours. The other 70%, it’s idle. kubectl napwake scan shows your cluster’s own number before you install anything.

The scan report for an example cluster. It leads with 19 of 31 workloads running for nobody and $691 that could be freed a month. Beside it, the cluster drawn as a hotel at night, twice: now, with most rooms lit, and with Napwake, the rooms lit for nobody dark.
Terminal window
kubectl napwake scan
  • It runs with your kubeconfig and installs nothing in the cluster.
  • It shows what's idle right now as rooms lit for nobody, and what each costs at your nodes' list prices.
  • With Prometheus, it replays the last week of each workload's CPU and rollouts, and says what pausing would have freed.

One file to send your team. It loads nothing from elsewhere and prints to PDF.

Open the example report Scan, in the plugin's docs

Four doors

Section titled “Four doors”
Asleep · paused
Scaled to zero. Its Services are routed to the doorman, which wakes it on the first request.
Waking up · resuming
A request, an activity annotation, or autoResume ahead of demand woke it. Its replicas are restored and its dependencies woken.
Awake · running
Its pods are Ready and held requests passed through. With no activity for idleAfter, and no forecast of demand, it pauses again.
Needs you · needs attention
A note pinned on the door. Something stops Napwake doing its job, and kubectl napwake status says what to fix.

How it works

Section titled “How it works”

Label a Deployment, StatefulSet or namespace napwake.dev/managed: "true" and Napwake creates a ManagedWorkload for it, with its settings from annotations. Then it keeps an activity clock: CPU, deploys, activity annotations, and Prometheus.

  1. Awake · running

    Running

    Activity clock: CPU, deploys, activity annotations, Prometheus

  2. No activity for idleAfter, and no forecast of demand
  3. Asleep · paused

    Paused

    Scaled to zero, its Services routed to the doorman

  4. A request, an activity annotation, or autoResume ahead of demand
  5. Waking up · resuming

    Resuming

    Replicas restored, dependencies woken

Pods Ready, held requests passed through

What it does

Section titled “What it does”

  • Opt in with a label

    Label a workload or namespace napwake.dev/managed: "true". Settings are annotations in your Git, so Argo CD and Flux work as today.

  • Activity clock

    Napwake records each workload's last activity: CPU over cpuThreshold, a deploy, an activity annotation or a Prometheus query. After idleAfter with none, it pauses. No learning period.

  • Demand forecasting

    A Holt-Winters double seasonal model learns each workload's day and week from up to four weeks of Prometheus. It scores itself from week two and acts when accurate enough: eager, balanced or cautious. It re-learns after drift, never from a holiday.

  • Safe by default

    Napwake never pauses what it can't measure. A confident forecast can hold a pause off. dryRun pauses nothing and shows what would. Anyone else's scale-up is a wake. If a GitOps tool undoes a pause, Napwake tells you how to fix it.

  • Cost tracking

    Napwake tracks each workload's cost, what pausing frees, and what came off the bill, from each node pool's spare capacity. Freed capacity is money once an autoscaler removes nodes. kubectl napwake nodes shows what keeps them. How to let them go.

  • Metrics and alerts

    Prometheus metrics for operator health, lifecycle transitions, wakes on request and forecast confidence, with alerting rules.

Quick example

Section titled “Quick example”
deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-api
  namespace: dev
  labels:
    napwake.dev/managed: "true"
  annotations:
    napwake.dev/dry-run: "true"
    napwake.dev/idle-after: "1h"

Napwake watches my-api and records each time it’s active: CPU above cpuThreshold (10% of its requests by default), a deploy, or an activity annotation, and, with a Prometheus query, its request rate. In dry-run it only measures. Once you run kubectl napwake enable my-api -n dev, it pauses the Deployment after an hour with no activity, and wakes it when a request reaches its Service.

For every setting, see Opting in. For settings annotations don’t cover, write a ManagedWorkload yourself.

Get started