How to Route Cheap Jobs to Cheaper Runners

Running a five-second lint check on an 8-core runner pays for cores it never uses; match the runner to the job.

Set runs-on per job so lightweight checks use the standard runner and only genuinely heavy jobs claim larger machines.

Steps

• Classify jobs as light (lint, format) or heavy (build, integration tests).
• Point light jobs at the smallest runner and heavy jobs at larger ones.
• Review quarterly; jobs grow and their right size changes.

Workflow

jobs:
  lint:
    runs-on: ubuntu-latest        # small, cheap
    steps:
      - uses: actions/checkout@v4
      - run: npm run lint
  integration:
    runs-on: ubuntu-latest-8-cores  # larger, only where it pays off
    steps:
      - uses: actions/checkout@v4
      - run: npm run test:integration

Gotchas

• Custom runner labels must be defined in org or repo runner settings first.
• Do not over-shard; each extra job pays fresh setup overhead.
• A larger runner idling on a light job is pure waste; the rate is billed per minute regardless of load.

Related guides