The Ansible Development Cycle

The Ansible development cycle happens on two levels. At a macro level, the team plans releases and tracks progress with roadmaps and projects. At a micro level, each PR has its own lifecycle.

Macro development: roadmaps, releases, and projects

If you want to follow the conversation about what features will be added to Ansible for upcoming releases and what bugs are being fixed, you can watch these resources:

Micro development: the lifecycle of a PR

Ansible accepts code through pull requests (“PRs” for short). GitHub provides a great overview of how the pull request process works in general. The ultimate goal of any pull request is to get merged and become part of Ansible Core. Here’s an overview of the PR lifecycle:

  • Contributor opens a PR
  • Ansibot reviews the PR
  • Ansibot assigns labels
  • Ansibot pings maintainers
  • Shippable runs the test suite
  • Developers, maintainers, community review the PR
  • Contributor addresses any feedback from reviewers
  • Developers, maintainers, community re-review
  • PR merged or closed

Automated PR review: ansibullbot

Because Ansible receives many pull requests, and because we love automating things, we’ve automated several steps of the process of reviewing and merging pull requests with a tool called Ansibullbot, or Ansibot for short.

Ansibullbot serves many functions:

  • Responds quickly to PR submitters to thank them for submitting their PR
  • Identifies the community maintainer responsible for reviewing PRs for any files affected
  • Tracks the current status of PRs
  • Pings responsible parties to remind them of any PR actions for which they may be responsible
  • Provides maintainers with the ability to move PRs through the workflow
  • Identifies PRs abandoned by their submitters so that we can close them
  • Identifies modules abandoned by their maintainers so that we can find new maintainers

Ansibot workflow

Ansibullbot runs continuously. You can generally expect to see changes to your issue or pull request within thirty minutes. Ansibullbot examines every open pull request in the repositories, and enforces state roughly according to the following workflow:

  • If a pull request has no workflow labels, it’s considered new. Files in the pull request are identified, and the maintainers of those files are pinged by the bot, along with instructions on how to review the pull request. (Note: sometimes we strip labels from a pull request to “reboot” this process.)
  • If the module maintainer is not $team_ansible, the pull request then goes into the community_review state.
  • If the module maintainer is $team_ansible, the pull request then goes into the core_review state (and probably sits for a while).
  • If the pull request is in community_review and has received comments from the maintainer:
    • If the maintainer says shipit, the pull request is labeled shipit, whereupon the Core team assesses it for final merge.
    • If the maintainer says needs_info, the pull request is labeled needs_info and the submitter is asked for more info.
    • If the maintainer says needs_revision, the pull request is labeled needs_revision and the submitter is asked to fix some things.
  • If the submitter says ready_for_review, the pull request is put back into community_review or core_review and the maintainer is notified that the pull request is ready to be reviewed again.
  • If the pull request is labeled needs_revision or needs_info and the submitter has not responded lately:
    • The submitter is first politely pinged after two weeks, pinged again after two more weeks and labeled pending action, and the issue or pull request will be closed two weeks after that.
    • If the submitter responds at all, the clock is reset.
  • If the pull request is labeled community_review and the reviewer has not responded lately:
    • The reviewer is first politely pinged after two weeks, pinged again after two more weeks and labeled pending_action, and then may be reassigned to $team_ansible or labeled core_review, or often the submitter of the pull request is asked to step up as a maintainer.
  • If Shippable tests fail, or if the code is not able to be merged, the pull request is automatically put into needs_revision along with a message to the submitter explaining why.

There are corner cases and frequent refinements, but this is the workflow in general.

PR labels

There are two types of PR Labels generally: workflow labels and information labels.

Workflow labels
  • community_review: Pull requests for modules that are currently awaiting review by their maintainers in the Ansible community.
  • core_review: Pull requests for modules that are currently awaiting review by their maintainers on the Ansible Core team.
  • needs_info: Waiting on info from the submitter.
  • needs_rebase: Waiting on the submitter to rebase.
  • needs_revision: Waiting on the submitter to make changes.
  • shipit: Waiting for final review by the core team for potential merge.
Information labels
  • backport: this is applied automatically if the PR is requested against any branch that is not devel. The bot immediately assigns the labels backport and core_review.
  • bugfix_pull_request: applied by the bot based on the templatized description of the PR.
  • cloud: applied by the bot based on the paths of the modified files.
  • docs_pull_request: applied by the bot based on the templatized description of the PR.
  • easyfix: applied manually, inconsistently used but sometimes useful.
  • feature_pull_request: applied by the bot based on the templatized description of the PR.
  • networking: applied by the bot based on the paths of the modified files.
  • owner_pr: largely deprecated. Formerly workflow, now informational. Originally, PRs submitted by the maintainer would automatically go to shipit based on this label. If the submitter is also a maintainer, we notify the other maintainers and still require one of the maintainers (including the submitter) to give a shipit.
  • pending_action: applied by the bot to PRs that are not moving. Reviewed every couple of weeks by the community team, who tries to figure out the appropriate action (closure, asking for new maintainers, and so on).
Special Labels
  • new_plugin: this is for new modules or plugins that are not yet in Ansible.

Note: new_plugin kicks off a completely separate process, and frankly it doesn’t work very well at present. We’re working our best to improve this process.

Human PR review

After Ansibot reviews the PR and applies labels, the PR is ready for human review. The most likely reviewers for any PR are the maintainers for the module that PR modifies.

Each module has at least one assigned maintainer, listed in the BOTMETA.yml file.

The maintainer’s job is to review PRs that affect that module and decide whether they should be merged (shipit) or revised (needs_revision). We’d like to have at least one community maintainer for every module. If a module has no community maintainers assigned, the maintainer is listed as $team_ansible.

Once a human applies the shipit label, the committers decide whether the PR is ready to be merged. Not every PR that gets the shipit label is actually ready to be merged, but the better our reviewers are, and the better our guidelines are, the more likely it will be that a PR that reaches shipit will be mergeable.

Making your PR merge-worthy

We don’t merge every PR. Here are some tips for making your PR useful, attractive, and merge-worthy.

Changelogs

Changelogs help users and developers keep up with changes to Ansible. Ansible builds a changelog for each release from fragments. You must add a changelog fragment to any PR that changes functionality or fixes a bug. You don’t have to add a changelog fragment for PRs that add new modules and plugins, because our tooling does that for you automatically.

We build short summary changelogs for minor releases as well as for major releases. If you backport a bugfix, include a changelog fragment with the backport PR.

Creating a changelog fragment

A basic changelog fragment is a .yaml file placed in the changelogs/fragments/ directory. Each file contains a yaml dict with keys like bugfixes or major_changes followed by a list of changelog entries of bugfixes or features. Each changelog entry is rst embedded inside of the yaml file which means that certain constructs would need to be escaped so they can be interpreted by rst and not by yaml (or escaped for both yaml and rst if that’s your desire). Each PR must use a new fragment file rather than adding to an existing one, so we can trace the change back to the PR that introduced it.

To create a changelog entry, create a new file with a unique name in the changelogs/fragments/ directory. The file name should include the PR number and a description of the change. It must end with the file extension .yaml. For example: 40696-user-backup-shadow-file.yaml

A single changelog fragment may contain multiple sections but most will only contain one section. The toplevel keys (bugfixes, major_changes, and so on) are defined in the config file for our release note tool. Here are the valid sections and a description of each:

major_changes
Major changes to Ansible itself. Generally does not include module or plugin changes.
minor_changes
Minor changes to Ansible, modules, or plugins. This includes new features, new parameters added to modules, or behavior changes to existing parameters.
deprecated_features
Features that have been deprecated and are scheduled for removal in a future release.
removed_features
Features that were previously deprecated and are now removed.
bugfixes
Fixes that resolve issues. If there is a specific issue related to this bugfix, add a link in the changelog entry.
known_issues
Known issues that are currently not fixed or will not be fixed.

Most changelog entries will be bugfixes or minor_changes. When writing a changelog entry that pertains to a particular module, start the entry with - [module name] - and include a link to the related issue if one exists.

Here are some examples:

bugfixes:
  - win_updates - fixed issue where running win_updates on async fails without any error

minor_changes:
  - lineinfile - add warning when using an empty regexp (https://github.com/ansible/ansible/issues/29443)

bugfixes:
  - copy module - The copy module was attempting to change the mode of files for
 remote_src=True even if mode was not set as a parameter.  This failed on
 filesystems which do not have permission bits.

You can find more example changelog fragments in the changelog directory for the 2.6 release. You can also find documentation of the format, including hints on embedding rst in the yaml, in the reno documentation.

Once you’ve written the changelog fragment for your PR, commit the file and include it with the pull request.

Backporting merged PRs

All Ansible PRs must be merged to the devel branch first. After a pull request has been accepted and merged to the devel branch, the following instructions will help you create a pull request to backport the change to a previous stable branch.

We do not backport features.

Note

These instructions assume that:

  • stable-2.9 is the targeted release branch for the backport
  • https://github.com/ansible/ansible.git is configured as a git remote named upstream. If you do not use a git remote named upstream, adjust the instructions accordingly.
  • https://github.com/<yourgithubaccount>/ansible.git is configured as a git remote named origin. If you do not use a git remote named origin, adjust the instructions accordingly.

  1. Prepare your devel, stable, and feature branches:

    git fetch upstream
git checkout -b backport/2.9/[PR_NUMBER_FROM_DEVEL] upstream/stable-2.9

  2. Cherry pick the relevant commit SHA from the devel branch into your feature branch, handling merge conflicts as necessary:

    git cherry-pick -x [SHA_FROM_DEVEL]

  3. Add a changelog fragment for the change, and commit it.

  4. Push your feature branch to your fork on GitHub:

    git push origin backport/2.9/[PR_NUMBER_FROM_DEVEL]

  5. Submit the pull request for backport/2.9/[PR_NUMBER_FROM_DEVEL] against the stable-2.9 branch

  6. The Release Manager will decide whether to merge the backport PR before the next minor release. There isn’t any need to follow up. Just ensure that the automated tests (CI) are green.

Note

The choice to use backport/2.9/[PR_NUMBER_FROM_DEVEL] as the name for the feature branch is somewhat arbitrary, but conveys meaning about the purpose of that branch. It is not required to use this format, but it can be helpful, especially when making multiple backport PRs for multiple stable branches.

Note

If you prefer, you can use CPython’s cherry-picker tool (pip install --user 'cherry-picker >= 1.3.2') to backport commits from devel to stable branches in Ansible. Take a look at the cherry-picker documentation for details on installing, configuring, and using it.