Dynamic AutoScaling of GitHub Runners

This article describes Dgraph Labs' transition from a self-hosted on-prem TeamCity setup to GitHub Actions for their CI/CD needs. They were facing challenges with high compute costs, maintenance efforts, and configuration time, which led them to explore GitHub Actions.

With GitHub Actions, they were able to reduce their compute costs, maintenance efforts, and configuration time. However, they still faced challenges with idle runners and manual attach/detach of runners.

To solve these challenges, they built an in-house solution called "Dynamic AutoScaling of GitHub Runners." Their architecture includes VM images, SSM Parameter Store, and an orchestrator, which monitors GitHub events and dynamically scales up or down based on the job queue and available runner count. They also provide a financial analysis showing the significant cost savings they achieved after enabling Dynamic AutoScaling of GitHub Runners.