# Prometheus alerting rules for github-actions-runner-exporter.
# Load with `rule_files:`, or paste the group into a PrometheusRule's spec.groups.
# Thresholds are starting points; tune `for:` and the ratios to your runner fleet.
groups:
  - name: github-actions-runner-exporter
    rules:
      - alert: GithubRunnerOffline
        expr: max by (runner, os) (github_runner_up) == 0
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: "Self-hosted runner {{ $labels.runner }} is offline"
          description: The runner is registered but not connected to GitHub. Jobs that need it will queue.

      - alert: GithubRunnersAllOffline
        expr: count(github_runner_up) > 0 and sum(github_runner_up) == 0
        for: 5m
        labels:
          severity: critical
        annotations:
          summary: All self-hosted runners are offline
          description: No self-hosted runner can pick up jobs; every workflow that targets them is stuck.

      - alert: GithubRunnersSaturated
        # Every online runner busy for a long stretch: jobs are probably queueing.
        expr: sum(github_runner_busy) >= sum(github_runner_up) and sum(github_runner_up) > 0
        for: 30m
        labels:
          severity: warning
        annotations:
          summary: All online self-hosted runners have been busy for 30 minutes
          description: New jobs are likely waiting in the queue. Consider adding runners.

      - alert: GithubRunnerExporterStale
        expr: max(github_runners_up) == 0
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: Runner status is stale
          description: The exporter's last runner poll failed and a cached result is being served. Check the token and GitHub API reachability.

      - alert: GithubOrgStatsStale
        expr: max(github_org_up) == 0
        for: 30m
        labels:
          severity: warning
        annotations:
          summary: Org and repo stats are stale
          description: The org stats poll keeps failing; CI, PR and Dependabot metrics are no longer updating.

      - alert: GithubApiRateLimitLow
        expr: max(github_rate_limit_remaining) / max(github_rate_limit_limit) < 0.10
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: "GitHub API rate limit at {{ $value | humanizePercentage }} remaining"
          description: The exporter (or anything else sharing the token) is close to the limit, and polls will start failing. Raise the cache TTLs or use a separate token.

      - alert: GithubWorkflowFailing
        # The latest completed run of the workflow, on whatever branch ran last.
        expr: max by (repo, workflow, url, conclusion) (github_repo_ci_last_run_conclusion{conclusion=~"failure|timed_out"}) == 1
        for: 1h
        labels:
          severity: info
        annotations:
          summary: "{{ $labels.repo }}: {{ $labels.workflow }} last run {{ $labels.conclusion }}"
          description: "Latest run: {{ $labels.url }}"

      - alert: GithubDependabotCriticalAlerts
        expr: sum by (repo) (github_repo_dependabot_alerts_open{severity="critical"}) > 0
        for: 1h
        labels:
          severity: warning
        annotations:
          summary: "{{ $labels.repo }} has {{ $value }} open critical Dependabot alert(s)"
          description: A dependency has a known critical vulnerability with an available advisory.