Introduction
Battleship runs your GitHub Actions jobs on a machine reserved for you, at a flat monthly price. When the machine is full, jobs wait in the queue.
Your own machine for GitHub Actions
Battleship gives your organisation a dedicated machine that runs GitHub Actions jobs. Your workflows stay in your repository. A job moves to the machine when you change its runs-on.
GitHub keeps your workflows, scheduling, secrets, checks and job logs. Battleship supplies the machine and starts a new runner for each job.
How a job runs
- GitHub queues a job whose
runs-onnames a label of your machine, for examplebattleship-ubuntu-latest. - GitHub sends a signed
workflow_jobevent. Battleship checks its signature, installation, repository and labels. - Battleship starts a new disposable virtual machine on your machine and registers the official GitHub Actions runner in it for one job.
- GitHub assigns the job to that runner. Steps, logs and results appear in GitHub as usual.
- When the job ends, the virtual machine is deleted. The next job gets a new one.
GitHub decides which matching runner takes which job. A runner started for one job can take another queued job with the same labels.
A flat price and a queue
The machine has a fixed amount of CPU, memory and disk, and its price does not change with the number of jobs or minutes. See pricing for current plans.
When every runner slot is in use, new jobs stay queued in GitHub and start as earlier jobs finish. They do not fail because the machine is busy. GitHub cancels a job that has been queued for more than 24 hours. Read Queueing and capacity for the order in which jobs start.
Jobs opt in by label
Only jobs that name a matching label run on the machine. A job that says runs-on: ubuntu-latest keeps running on GitHub-hosted runners. To move a job back, restore its previous runs-on.
Runners are Linux x64. Keep Windows, macOS, arm64 and GPU jobs on GitHub-hosted runners.
What is new
See the changelog for what is new.
Start with the quickstart.
Machines explains the ship classes and what the dashboard shows, and Usage explains how to read a machine's history.