wait-other-jobs

This GitHub Action waits for all jobs by default even if they run in other workflows.
You can also choose to wait for specific jobs.
If any of the jobs fail, this action fails too.

Usage

It works with zero or little configuration.

jobs:
  your_job:
    # Enabling these permissions are required in private repositories
    # permissions:
    #   contents: read
    #   checks: read
    #   actions: read
    runs-on: ubuntu-24.04
    steps:
      - uses: kachick/wait-other-jobs@v3.8.1
        timeout-minutes: 15 # Recommended to be enabled with your appropriate value for fail-safe use

You can configure the token, polling interval, allow/deny lists, and early-exit behavior as shown below.

with:
  warmup-delay: 'PT30S' # default 'PT10S'
  minimum-interval: 'PT300S' # default 'PT15S'
  retry-method: 'exponential_backoff' # default 'equal_intervals'
  early-exit: 'false' # default 'true'
  skip-same-workflow: 'true' # default 'false'
  wait-list: |
    [
      {
        "workflowFile": "ci.yml",
        "jobName": "test",
        "eventName": "${{ github.event_name }}"
      },
      {
        "workflowFile": "release.yml",
        "optional": true
      }
    ]
  skip-list: |
    [
      {
        "workflowFile": "pages.yml"
      }
    ]

Full list of the options

NAME	DESCRIPTION	TYPE	DEFAULT	OPTIONS
github-api-url	The Github API endpoint. Override for Github Enterprise usage.	string	${{ github.api_url }}	https://api.github.com, https://ghe-host.example.net/api/v3
github-token	The GITHUB_TOKEN secret. You can use PAT if you want.	string	${{ github.token }}
warmup-delay	Wait this interval before first polling	string	PT10S	ISO 8601 duration format
minimum-interval	Wait for this or a longer interval between each poll to reduce GitHub API calls	string	PT15S	ISO 8601 duration format
retry-method	How to wait for next polling	string	equal_intervals	exponential_backoff, equal_intervals
early-exit	Stop polling as soon as one job fails	bool	true
attempt-limits	Stop polling if reached to this limit	number	1000
wait-list	Wait only for these jobs	string	[]	JSON Array
skip-list	Wait for all jobs except these	string	[]	JSON Array
skip-same-workflow	Skip jobs defined in the same workflow which using this action	bool	false
dry-run	Avoid requests for tests	bool	false

Guide for option syntax and reasonable values

• ISO 8601 duration format
• Trailing commas are not allowed in JSON
• GitHub API Limit: At least we should consider about GITHUB_TOKEN, that is allowed 1000 per hour per repository.
  Roughly calculating for long jobs, setting the minimum-interval larger than or equal PT4S would be safer.
  • Primary Limit
  • Secondary Limit

Schema of wait-list and skip-list

Lists should be given with JSON array, do not use both wait-list and skip-list together

• Each item should have a "workflowFile" field, and they can optionally have a "jobName" field.
• If no jobName is specified, all the jobs in the workflow will be targeted.
• wait-list:
  • If the checkRun for the specified name is not found, this action raise errors by default.
    You can disable this validation with "optional": true or use the startupGracePeriod that described in following section
  • Wait for all event types by default, you can change with "eventName": "EVENT_NAME_AS_push".

Required GITHUB_TOKEN permissions

In public repositories, they are satisfied by default

permissions:
  contents: read # Since v2
  checks: read
  actions: read

Support for Github Enterprise

To run this action in your Github Enterprise (GHE) instance you need to override github-api-url:

with:
  github-api-url: 'https://ghe-host.example.net/api/v3'

outputs.<output_id>

These outputs are for testing and debugging only. The schema is not defined.

• parameters
  Parsed values from with and some context.

• dump
  A file path for collected resources which keeps fields than logged.

Examples

I'm using this action for auto-merging bot PRs and wait for deploy.
See the docs for further detail.

Deadlocks

When using this action in multiple jobs within the same repository, be careful to avoid deadlocks.
The skip-list, wait-list and skip-same-workflow options cover this use case.

If you changed job name from the default, you should set skip-list or roughly use skip-same-workflow

jobs:
  your_job: # This will be used default job name if you not specify below "name" field
    name: 'Changed at here'
    runs-on: ubuntu-24.04
    steps:
      - uses: kachick/wait-other-jobs@v3.8.1
        with:
          skip-list: |
            [
              {
                "workflowFile": "this_file_name.yml",
                "jobName": "Changed at here"
              }
            ]
        timeout-minutes: 15

You need to consider similar problems when using matrix, because GitHub does not provide enough context.

• https://github.com/orgs/community/discussions/8945
• https://github.com/orgs/community/discussions/16614

However you can set prefix for jobMatchMode to create small skip-list to avoid this problem.

jobs:
  your_job:
    strategy:
      matrix:
        os:
          - ubuntu-24.04
          - ubuntu-22.04
    runs-on: ${{ matrix.os }}
    steps:
      - uses: kachick/wait-other-jobs@v3.8.1
        with:
          skip-list: |
            [
              {
                "workflowFile": "this_file_name.yml",
                "jobMatchMode": "prefix",
                "jobName": "${{ github.job }}"
              }
            ]
      - run: make test

Startup grace period

This action only checks the status of jobs at each polling time.
Use this option when a job may start with a short delay after this action starts.

Example using a wait-list.

with:
  wait-list: |
    [
      {
        "workflowFile": "might_be_triggered_after_0-4_minutes.yml",
        "optional": false,
        "startupGracePeriod": { "minutes": 5 }
      }
    ]

This action starts immediately but ignores the job missing in the first 5 minutes.

• No need to extend warmup-delay
• Disable optional, because it is needed to check
• Set enough value for startupGracePeriod for this purpose.
  It must be parsible by TC39 - Temporal.Duration
  e.g
  • "PT3M42S" # ISO 8601 duration format
  • { "minutes": 3, "seconds": 42 } # key-value for each unit

If you're not using wait-list, you need to handle this pattern with warmup-delay.

Alternative candidates

gh commands, such as gh pr checks and gh run watch, should be useful if your requirements are simple.

Limitations

• If any workflow starts many jobs as 100+, this action does not support it.
  Because of nested paging in GraphQL makes complex. See related docs for further detail.

License

The scripts and documentation in this project are released under the MIT License