lifecycle

Cleanup and outcome handling in raw YAML means smearing if: always(), if: failure(), and steps.<id>.outcome guards across your steps and remembering to give each guarded step an id. The lifecycle hooks — ensure, on-success, on-failure, and on-abort — declare those steps next to the work they guard, and Actio stamps in the right condition (and any needed id) at compile time. The same vocabulary scales up to a workflow-level finally:.

Lifecycle hooks react to an outcome (cleanup, log upload, paging). To recover from a failure — retry on another runner, or a catch step that changes the result — see fallback.

Step and job hooks

At step scope the hook steps are spliced in right after the target step; at job scope they are appended to the job. Either way the hook key is stripped from the output.

name: CI
on: [push]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - run: ./start-db.sh
        ensure:
          - run: ./stop-db.sh
      - run: ./build.sh
        on-failure:
          - run: ./upload-logs.sh

name: CI
on:
  - push
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - run: ./start-db.sh
      - run: ./stop-db.sh
        if: always()
      - run: ./build.sh
        id: actio_test_step_2
      - run: ./upload-logs.sh
        if: "!cancelled() && steps.actio_test_step_2.outcome == 'failure'"

name: CI
on: [push]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - run: ./start-db.sh
        ensure:
          - run: ./stop-db.sh
      - run: ./build.sh
        on-failure:
          - run: ./upload-logs.sh

name: CI
on:
  - push
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - run: ./start-db.sh
      - run: ./stop-db.sh
        if: always()
      - run: ./build.sh
        id: actio_test_step_2
      - run: ./upload-logs.sh
        if: "!cancelled() && steps.actio_test_step_2.outcome == 'failure'"

./stop-db.sh gets if: always(). The guarded ./build.sh step is given a synthetic id so the on-failure step can key off its outcome before any continue-on-error remapping.

Workflow teardown (finally)

finally: declares jobs that run after the rest of the workflow, gated on overall outcome. Actio injects needs: over every real job plus the matching outcome guard.

name: Deploy
on: [push]
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - run: ./deploy.sh
finally:
  on-failure:
    page:
      runs-on: ubuntu-latest
      steps:
        - run: ./page-oncall.sh

name: Deploy
on:
  - push
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - run: ./deploy.sh
  page:
    runs-on: ubuntu-latest
    needs: 
      - deploy
    if: failure()
    steps:
      - run: ./page-oncall.sh

name: Deploy
on: [push]
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - run: ./deploy.sh
finally:
  on-failure:
    page:
      runs-on: ubuntu-latest
      steps:
        - run: ./page-oncall.sh

name: Deploy
on:
  - push
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - run: ./deploy.sh
  page:
    runs-on: ubuntu-latest
    needs: 
      - deploy
    if: failure()
    steps:
      - run: ./page-oncall.sh

Gotchas

See the lifecycle hooks and finally entries in the syntax reference for the full guard table.