Skip to content

runner-types.md

Latest commit

 

History

History
55 lines (43 loc) · 3.15 KB

runner-types.md

File metadata and controls

55 lines (43 loc) · 3.15 KB

Runner Types: Integration vs Default

Modern CI pipelines in this repository classify runners into two categories to control cost and coverage. Understanding the difference allows you to target expensive resources only when necessary and keep feedback cycles fast.

Default ("standard") runners

  • Environment – GitHub-hosted virtual machines such as ubuntu-latest or windows-latest.
  • Use cases – Linting, unit tests and any step that can run on a clean ephemeral image.
  • Configuration – In requirements.json omit runner_type or set it to standard. Workflows simply use runs-on: ubuntu-latest or similar.
  • Characteristics – High concurrency, minimal boot time and no persistent state. Ideal for rapid validation.

Integration runners

  • Environment – Long-lived machines with preinstalled tooling such as LabVIEW, g-cli and hardware drivers. They may be self-hosted or specialized GitHub images.
  • Use cases – End‑to‑end scenarios that interact with external systems, require licensed software or need deterministic state.
  • Configuration – Tag the runner with runner_type: "integration" in requirements.json and reference the runner by label in workflows. Integration entries often set skip_dry_run: true to force real execution.
  • Characteristics – Limited availability and higher cost; jobs are serialized to protect shared resources. Treat these runners as scarce infrastructure.

Declaring runner types

The requirements.json file defines which runner each test or requirement targets:

{
  "runners": {
    "ubuntu-latest": {
      "runner_label": "ubuntu-latest",
      "runner_type": "integration"
    },
    "windows-latest": {
      "runner_label": "windows-latest"
      /* implicit runner_type: "standard" */
    }
  },
  "requirements": [
    {
      "id": "REQ-009",
      "runner": "ubuntu-latest",
      "tests": ["Build.Workflow"]
    }
  ]
}

The summarizer partitions results by runner_type, producing artifacts such as summary-integration.md alongside summary-standard.md. This separation keeps integration evidence distinct from fast feedback produced on default runners.

Recommendations for CI/CD engineers

  1. Minimize integration usage. Start with standard runners and move tests to integration environments only when they require external dependencies or state.
  2. Isolate heavy workflows. Place integration jobs in separate stages or repositories to avoid blocking quick validation paths.
  3. Protect self-hosted runners. Apply concurrency limits and explicit needs chains so multiple integration jobs do not compete for the same hardware.
  4. Audit runner labels. Keep requirements.json synchronized with the actual fleet of self-hosted machines. Stale labels lead to idle jobs.
  5. Document expectations. When adding new requirements or workflows, update docs/requirements.md so the Runner and Runner Type columns reflect the intended infrastructure.

By explicitly classifying jobs, teams can scale the project efficiently—routine tasks remain fast on default runners while integration tests validate real‑world behavior without overloading scarce resources.