# Gitea Actions self-hosted runners — Ansible

Provisions `act_runner` instances as **systemd services**. One host can run runners for
multiple projects (e.g. `gatehouse-api` + `gatehouse-ui`); each project gets its own
runner name, install dir, label set, repo URL, and registration token.

## Layout

```
ansible.cfg                 inventory + ssh defaults
inventory.ini               stage / prod host groups (set ansible_host)
group_vars/all.yml          pinned act_runner version + sha256, gitea_instance URL
host_vars/<host>.yml        runner_env + per-project `runners` matrix
tasks/install_project.yml   reads token, expands count -> N instances
tasks/install_one_runner.yml  download/register/service one runner
install-runner.yml          the playbook
```

## Runner naming

- name:    `<host>-<project>-<env>-<N>`  e.g. `stage01-gatehouse-api-stage-1`
- dir:     `/home/gitea-runner/act-runner-<project>-<N>`
- labels:  `project_spec.labels` (comma-separated, e.g. `self-hosted,linux,stage`)
- service: `gitea-runner-<project>-<N>.service`

## Before first run — pin act_runner version

1. Check [act_runner releases](https://gitea.com/gitea/act_runner/releases) for the latest stable version.
2. Download `act_runner-<version>-linux-amd64.sha256` from the release page.
3. Update `act_runner_version` and `act_runner_sha256` in `group_vars/all.yml`.

## Tokens

Registration tokens are **per-repo and expire ~1h after minting**. Exported as env
vars on the control node, named by each project's `token_env` in `host_vars`.

Mint each from the repo's **Settings → Actions → Runners → Create new runner token**,
then run the playbook within the hour.

## Run

```bash
cd deploy/ansible

# 1. Edit inventory.ini (set ansible_host / ansible_user) and host_vars/*.yml.
# 2. Fill in act_runner_version + act_runner_sha256 in group_vars/all.yml.
# 3. Export fresh tokens (leading space keeps them out of shell history):
 export RUNNER_TOKEN_GATEHOUSE_API=xxx
 export RUNNER_TOKEN_GATEHOUSE_UI=yyy

ansible-playbook --syntax-check install-runner.yml
ansible-playbook install-runner.yml --check --diff --limit stage   # dry run
ansible-playbook install-runner.yml --limit stage                  # apply
```

## Verify

```bash
# on the host
systemctl list-units 'gitea-runner-*'
ls /home/gitea-runner/                # act-runner-gatehouse-api-1, ...
```

Each repo's **Settings → Actions → Runners** should list the runners as **Idle** with
the expected labels. Idempotent: re-running skips already-registered runners
(`creates: .runner` guards re-registration; binary download is checksum-gated).

## Migrating from GitHub Actions runners

Stop and remove the old runners on any existing host before re-running:

```bash
sudo su - github-runner -c "cd ~/actions-runner-*/; ./svc.sh stop && ./svc.sh uninstall"
sudo userdel -r github-runner   # optional — removes home dir too
```

## Notes

- Bump `act_runner_version` + `act_runner_sha256` together in `group_vars/all.yml`.
- Labels in `host_vars/<host>.yml` must match the `runs-on:` values in workflow files.
- `deploy/deploy.sh` needs the runner user to have passwordless sudo for nginx reload —
  add that sudoers drop-in separately (not yet automated here). Update the username from
  `github-runner` to `gitea-runner` in that drop-in.
- `prod01.ansible_host` in `inventory.ini` is still `CHANGE_ME` — fill in before running prod.