Run GitHub runners on a node
A node can also run your GitHub Actions jobs. You declare a pool on the node: which repository, which GitHub credential, and how many runners. The node then installs that many self-hosted runners, registers them on the repository and keeps them running. Each runner takes one job at a time and shares the machine's memory, so more runners leave less for each job.
Who can do this: requires the org admin role, and a plan whose GitHub Runners limit is not 0. Without either, the node has no GitHub Runners tab. The tab is also not offered on a Jaah Cloud node, one the operator allocated, or a node on a machine shared with other organizations. Set up GitHub Runners and Create pool also need the permission to edit nodes, and Remove runners the permission to delete nodes.
You need a GitHub credential first: a fine-grained personal access token with Administration read and write and Actions read on the repository. A Connect credential is offered in the list, but it can't carry Administration, so the runners fail to register after you save. See Credentials.
Set up the pool
-
In the sidebar, select Directory, then the Machines tab.
-
Select the node's name. Its details open on the Overview tab.
-
Select the GitHub Runners tab. A node without a pool says No GitHub Runners.
-
Select Set up GitHub Runners.
-
In Repository, pick the project whose GitHub repository the runners register on.
-
In GitHub credential, pick the token described above.
-
In Runners, enter how many runners the node keeps, from 1 to 32.
-
Select Create pool. The pool shows Applying while the node installs the runners, then Running, with how many are up out of how many you asked for, the repository, and when the node last reported.
Each runner then has its own row: Idle, or Busy with the workflow and job it is running. To change the count later, edit Runners and select Apply.
Pause, resume or remove the runners
The buttons sit on the pool's line; hover one for its name.
- Pause — after each current job (⏸): each runner finishes the job it is on, then stops taking jobs. The pool reads Draining while jobs finish, then Paused. Nothing running is killed.
- Resume (▶): the runners take jobs again.
- Remove runners (✕): asks Remove the GitHub Actions runners from and the node's name. Select Remove runners. Each runner is unregistered from GitHub and deleted from the machine once its current job ends, so this can take as long as the longest job. When none are left, the tab says Runners removed.
A pool also pauses itself when the machine's disk is nearly full (Paused (disk)) or its memory is critical (Paused: memory critical), and resumes on its own when usage drops.
A pool only runs while its node is online: see Read a node's status. Every state a runner and its pool can show is in Machines; what a node is, in How nodes work.