https://github.com/premiscale/premiscale

The PremiScale autoscaling controller for Kubernetes
https://github.com/premiscale/premiscale

autoscaling infrastructure kubernetes

Last synced: 11 months ago
JSON representation

The PremiScale autoscaling controller for Kubernetes

• Host: GitHub
• URL: https://github.com/premiscale/premiscale
• Owner: premiscale
• License: other
• Created: 2022-09-03T22:50:58.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2025-04-10T18:33:05.000Z (over 1 year ago)
• Last Synced: 2025-04-10T19:55:23.073Z (over 1 year ago)
• Topics: autoscaling, infrastructure, kubernetes
• Language: Python
• Homepage: https://premiscale.com
• Size: 6.24 MB
• Stars: 1
• Watchers: 0
• Forks: 0
• Open Issues: 12
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE
  • Code of conduct: CODE_OF_CONDUCT.md

Awesome Lists containing this project

README

          
# PremiScale

![GitHub Release](https://img.shields.io/github/v/release/premiscale/premiscale?include_prereleases&sort=semver&display_name=release&link=https%3A%2F%2Fgithub.com%2Fpremiscale%2Fpremiscale%2Freleases)

![Business License 1.1](https://img.shields.io/badge/License-Business_Source_1.1-red)

PremiScale is a controller that brings autoscaling of virtual and physical infrastructure to local, self-hosted and private data centers, with a particular focus on integrating with the Kubernetes [cluster autoscaler](https://github.com/kubernetes/autoscaler).

## Architecture

PremiScale uses [libvirt](https://libvirt.org/) to connect to hosts and manage lifecycles of virtual machines. The Libvirt daemon provides a rich API for interacting with hypervisors, hosts and virtual machines.

The controller can be configured to run in two different modes, including `kubernetes` and `standalone` modes. In either configuration, the controller aims to start only relevant processes for both data collection and managing virtual machines and hardware.

### Standalone

In `standalone` mode, the controller starts its own time series data collection process.



  



### Kubernetes

Starting the controller in `kubernetes` mode starts relevant components of the controller that allow it to interface with the cluster autoscaler.



  



Note that, in this configuration, the controller does not require a time series database. State is still reconciled, but the time series signal comes from the cluster autoscaler instead.

## Configuration

The controller is configured in a couple ways, including its command line interface, environment variables (as indicated in the help text below), and through the required config file (all versions of which are documented [here](https://github.com/premiscale/premiscale/tree/master/src/premiscale/config/docs) in this repository).

## Installation

This project is intended to be deployed to Kubernetes, whether you intend to integrate with the cluster autoscaler or not.

See the Helm [chart README](https://github.com/premiscale/premiscale/tree/master/helm/premiscale) for an overview of controller installation.