Skip to content
/ nozzle Public

A friendly serverless or binary controller to shutdown your Kubernetes workload and reduce your cluster power consumtion

License

Notifications You must be signed in to change notification settings

fjudith/nozzle

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Nozzle

Reduce Kubernetes workload on schedule or traffic decision basis and restore workload on-demand via Web UI or http trigger.

Introduction

This project has been initially created during the Sustainable Digital Challenge event organized by APIdays to demonstrate the vision of an aggressive compute resource saving strategy.

Do you keep the lights on, once you left the room ?

This simple sentense resumes the resoning on how over estimating availability leads to uncouscious a wast of natural resources.

Nozzle promotes sustainability by simply shuting down eligible resources like:

  • Non-production environments managed by developer
  • Application services required to run only on working hours

How it works

Nozzle has been initially designed to run as function inside Kubernetes. The following sequence diagram describe the steps that functions acheive to securely reduce and restore the workload managed by this orchestrator.

UML

Design requirements

The project has been design using the following requirements to ensure the product sustainability.

  • Store only when required
  • Search for reuse before creating something new
  • Aims for the least footprint
  • Constrain the volume of data transfer and resource allocation
  • Limit the usage of Javascript frameworks
  • Scale-to-zero when possible
  • Compliance with the 12 Factor

Implementations roadmap

Serverless Kubeless stable
Serverless OpenFaaS stable
Serverless Fission stable
Serverless Nuclio stable
Workflow Argo Worflow, Argo Events stable
Binary Golang not started

Screenshots

The follwing screeshots show the Nozzle impact on the microservices-demo from Weaveworks, which by default supports 14 microservices.

User gets the following reactive landing page after noozle downscaled the application.

Rescaler landing page

Once clicked on the rescale button, the user gets the logs related to the rescaling of the workload associated to its application.

From a storage efficiency point-of-view Noozle restores the application from Kubernetes resource annotations. By defualt the last known Kubernetes resources configuration is backed up inside the native kubectl.kubernetes.io/last-applied-configuration annotation. If not present, Nozzle generates its own annotation containing the minimal amount of date. The following annotation are generated depending on the resource type:

  • Deployment & Statefulset: replicas.nozzle.io/last-known-configuration
  • Ingress: rules.nozzle.io/last-known-configuration

Rescaler landing page logs

Resource consumption after user request to rescale.

Rescaling footprint

User have to refresh it brower tab to regain access to its application

Rescaling footprint

This last picture shows the resource consumtion after noozle downscaled the application and deployed the rescaler.

Grafana footprint

Message formats

  • Replicas JSON: {"namespace": str, "name": str, "kind": str, "replicas": int, "selector": { dict }}
  • Ingress JSON: {"namespace": str, "name": str, "rules": { dict }}

About

A friendly serverless or binary controller to shutdown your Kubernetes workload and reduce your cluster power consumtion

Resources

License

Stars

Watchers

Forks

Packages

No packages published