A practical guide to setting up autoscalable GitLab Runners on AWS EC2 using the Fleeting plugin, eliminating always-on instances and reducing CI infrastructure costs. The setup uses a single managing EC2 instance that orchestrates ephemeral worker instances via AWS Auto Scaling Groups. Key steps covered include choosing between instance and docker-autoscaler executors, configuring IAM permissions, installing the Fleeting plugin, writing config.toml with idle policies and scaling rules, building custom AMIs with Packer, and managing Launch Templates with Terraform. The guide also discusses pros and cons, including cost savings and job isolation benefits versus the pain of updating the managing runner's config.
Table of contents
Goals: automated scaling, lower costsWorkflow and GitLab + AWS setup overviewImplementationUseful considerationsKey takeaways with pros & cons161 Impressions