Run a Workflow After Another One Finishes

Start a workflow when a run of another workflow completes, and choose which runs it reacts to.

Use a run.completed trigger when one workflow must start after another workflow's run ends. Shipfox sends this event from the built-in shipfox source each time a run attempt succeeds, fails, or is cancelled.

The example below starts a deploy workflow after each successful build run in the api project. To react to other runs, keep the trigger and change the filter.

Before you begin

You need:

A synced upstream workflow, and its configuration path.
The name of the upstream workflow's project.

Start a workflow after a successful run

Create .shipfox/workflows/deploy.yml. Replace api with the upstream project's name and .shipfox/workflows/build.yml with the upstream workflow's configuration path. This is a complete workflow:

# yaml-language-server: $schema=https://www.shipfox.io/docs/workflow.schema.json
name: Deploy after build
runner: shipfox

triggers:
  after_build:
    source: shipfox
    event: run.completed
    filter: >-
      event.project.name == 'api'
      && event.workflow.path == '.shipfox/workflows/build.yml'
      && event.run.origin == 'synced'
      && event.run.status == 'succeeded'

jobs:
  deploy:
    steps:
      - env:
          UPSTREAM_WORKFLOW: ${{ event.workflow.name }}
          UPSTREAM_RUN: ${{ event.run.number }}
        run: echo "Deploying after $UPSTREAM_WORKFLOW run $UPSTREAM_RUN"

The trigger receives events only once the workflow is synced.

Every workflow that reacts to another run needs three things:

Subscribe to the event. Set source: shipfox and event: run.completed.
Select the runs. Write a filter that names the project, the workflow, the run origin, and the status you want. Without a filter, the trigger matches every run in the workspace, including its own runs.
Read the upstream run. Read event.run, event.workflow, and event.project in the jobs. The Shipfox events reference lists every field.

In your own workflow, change the filter values and the deploy step.

The downstream run uses its own project's default branch. It doesn't check out the upstream run's commit. Read event.run.ref and event.run.commit if your jobs need them.

Avoid trigger loops

Shipfox doesn't stop a workflow from triggering itself. A trigger that matches its own runs starts a new run each time one ends. Two workflows that trigger each other loop the same way.

Keep every run.completed filter safe with these three conditions:

Filter on event.project.name. A configuration path, such as .shipfox/workflows/build.yml, can exist in every project of the workspace.
Filter on event.run.origin. Dev runs also send run.completed. Filter on 'synced' to ignore them.
Exclude the workflow's own path when the filter doesn't name one upstream workflow.

For example, this trigger reacts to failed runs of any workflow in the api project, except its own runs:

triggers:
  on_failure:
    source: shipfox
    event: run.completed
    filter: >-
      event.project.name == 'api'
      && event.workflow.path != '.shipfox/workflows/notify-failures.yml'
      && event.run.origin == 'synced'
      && event.run.status == 'failed'

This fragment replaces the triggers map of a workflow stored at .shipfox/workflows/notify-failures.yml in the api project.

Adapt it to your workflow

To...	Change this
Pass a value from the upstream run	Declare workflow outputs in the upstream workflow, then read `event.run.outputs.<name>`. See Publish workflow outputs.
React to failed or cancelled runs	Compare `event.run.status` with another status. Read `event.run.status_reason` to tell a timeout from a failed job. See Shipfox events.
React to runs from one trigger	Add a condition on `event.run.trigger.source` or `event.run.trigger.event`.
Skip reruns	Add `event.run.attempt == 1`. Each rerun sends its own `run.completed` event.
Continue a run instead of starting one	Use `run.completed` in a listening job's `on` matcher. See Listening fields.

Verify the downstream run

Start a run of the upstream workflow that matches your filter.
Wait for the upstream run to end.
Open the downstream workflow's project. Confirm that a new run of the downstream workflow started.
Open its log and confirm that it names the upstream run.

In the example, run Build in the api project. After it succeeds, a Deploy after build run starts. Its log contains Deploying after Build run and the number of the build run.

If no run starts, open Settings → Events and filter by source shipfox. Follow Troubleshoot events that do not start a run.

Was this page helpful?

Edit this page on GitHub