GitLab.com Shared Runners use Autoscaling

What this post added

This post announces the introduction of autoscaling for GitLab.com Shared Runners, powered by GitLab Runner 1.1. This feature aims to reduce build queue times and machine allocation times. The autoscaling configuration uses the `docker+machine` executor, provisioned on Digital Ocean 4GB instances with CoreOS and Docker. Key aspects of the autoscaling setup include distributed caching, a Docker registry mirror, and a security focus on using fresh machines for each build. The post also details changes to the default services available to runners, removing pre-configured databases and requiring explicit declaration in `.gitlab-ci.yml`.