Copyright 2014 Hewlett-Packard Development Company, L.P.

This work is licensed under a Creative Commons Attribution 3.0 Unported License. http://creativecommons.org/licenses/by/3.0/legalcode

Separate puppet-module projects

Rather than having all puppet modules in a single project, each should be in its own project.

Problem Description

The openstack-infra/config project is a monolithic ecosystem of puppet modules. This makes it hard to pick-and-choose specific modules to use downstream such as asterisk or graphite when the larger ecosystem is not wanted, and also makes it easier to accidentally write changes in the wrong puppet module. Some modules, such as jenkins or gerrit, should be generic and reusable across deployments while others (the openstack_project module and possibly others in the future) are very site-specific. Making the site-specific modules a downstream consumer of the generic modules will more clearly delineate where site-specific changes should be made and make the generic modules more easily re-consumable.

Proposed Change

Put each (or at least most) puppet modules in their own project.


  • Leave openstack-infra/config as is (monolithic)

  • Create several projects, each with a logical group of modules (e.g. logstash + log_processor + elasticsearch + kibana)


The openstack_project module and manifests/site.pp will remain in config. The devstack_host, remove_nginx, and salt modules will be deleted. All other modules will be put in their own project.


Primary assignee(s):
  • jesusaurus

  • nibalizer

Additional assignee(s):
  • bodepd

Gerrit Topic

Use Gerrit topic “module-split” for all patches related to this spec.

git-review -t module-split

Work Items

We will need to create integration tests to ensure that api compatibility is maintained between puppet modules. This will be difficult to create without having multiple modules, but is important to have before actively developing the new modules. After breaking out the first module we should create an integration test to make sure that changes to it do not break openstack-infra/config and then we can continue with the rest of the modules.

The following process must be done for each module separately:

  1. Freeze changes to a specific module within system-config. You should solicit a request in the #openstack-infra IRC channel for a core member to shepherd the changes.

  2. Isolate the module’s history with git-subtree:

    git subtree split --prefix=modules/$MODULE --branch=$MODULE
    • git-subtree will do the right thing by re-rooting the resulting code at modules/$MODULE but will require git >= (and even then may be in a separate contrib package)

  3. Create a temporary project (e.g. personal github repo) to host the new base repo

  4. Push the new branch to the temporary project created above

    git remote add github https://github.com/$YOU/puppet-$MODULE.git
    git push github $MODULE:master
  5. Add a new gerrit project for the module in project-config (using the temporary project as upstream)

    • Follow example here: https://review.openstack.org/#/c/131302/

    • Remember to enable storyboard: For example: https://review.openstack.org/#/c/135420/

    • Amend the commit message with a reference to system-config patch from the next step.

      This patch should only be approved after arranging for a
      system-config core to ensure there are no outstanding changes
      and to approve the related module removal patch within a short time.
      Needed-By: <system-config patch Change-Id>
    • IMPORTANT: Once this patch merges, the project in the temporary repo will be pulled into -infra. e.g. http://git.openstack.org/cgit/openstack-infra/puppet-$MODULE/, and all open changes that affect files in the module will then have to be resubmitted to the new repo.

  6. Modify system-config/modules.env to install the module from the new gerrit project and add the new project to the puppet integration tests. Remove the old module from system-config

    • We should continuously deploy the master branch

    • Include in commit message a reference to the project-config patch done in previous step

      Depends-On: <project-config patch Change-Id>
    • Follow example here: https://review.openstack.org/#/c/131305/

  7. Propose a review to add some of the files that are needed by the module:

    • After the project-config patch merges, you can clone the new repo and submit the following changes for review.

    • .gitreview

    • Rakefile

      require 'rubygems'
      require 'puppetlabs_spec_helper/rake_tasks'
      require 'puppet-lint/tasks/puppet-lint'
      PuppetLint.configuration.fail_on_warnings = true
    • README.md

      # OpenStack $module Module
      This module installs and configures $module
    • metadata.json

        "name": "openstackci-$module",
        "version": "0.0.1",
        "author": "Openstack CI",
        "summary": "Puppet module for $module",
        "license": "Apache 2.0",
        "source": "https://git.openstack.org/openstack-infra/puppet-$module.git",
        "project_page": "http://docs.openstack.org/infra/system-config/",
        "issues_url": "https://storyboard.openstack.org/#!/project/$lookup-module-id",
        "dependencies": []

    # Note that determining dependencies may not be immediately obvious, we must count on the code review process to ensure that we’ve done this right.

    # Note that the Modulefile is deprecated and we should be using metadata.json exclusively now.

  8. When dependent puppet-module splits are completely ready to merge, a core reviewer will commit to approving them in the appropriate order or coordinate with another reviewer to take over.

  9. Lather, rinse, and repeat


  • openstack-infra/puppet-accessbot

  • openstack-infra/puppet-asterisk

  • openstack-infra/puppet-bugdaystats

  • openstack-infra/puppet-bup

  • openstack-infra/puppet-cgit

  • openstack-infra/puppet-drupal

  • openstack-infra/puppet-elastic_recheck

  • openstack-infra/puppet-elasticsearch

  • openstack-infra/puppet-exim

  • openstack-infra/puppet-gerrit

  • openstack-infra/puppet-gerritbot

  • openstack-infra/puppet-github

  • openstack-infra/puppet-graphite

  • openstack-infra/puppet-iptables

  • openstack-infra/puppet-jeepyb

  • openstack-infra/puppet-jenkins

  • openstack-infra/puppet-kibana

  • openstack-infra/puppet-lodgeit

  • openstack-infra/puppet-log_processor

  • openstack-infra/puppet-logrotate

  • openstack-infra/puppet-logstash

  • openstack-infra/puppet-mailman

  • openstack-infra/puppet-mediawiki

  • openstack-infra/puppet-meetbot

  • openstack-infra/puppet-mysql_backup

  • openstack-infra/puppet-nodepool

  • openstack-infra/puppet-openstackid

  • openstack-infra/puppet-packagekit

  • openstack-infra/puppet-pip

  • openstack-infra/puppet-planet

  • openstack-infra/puppet-recheckwatch

  • openstack-infra/puppet-redis

  • openstack-infra/puppet-releasestatus

  • openstack-infra/puppet-remove_nginx

  • openstack-infra/puppet-reviewday

  • openstack-infra/puppet-salt

  • openstack-infra/puppet-snmpd

  • openstack-infra/puppet-ssh

  • openstack-infra/puppet-ssl_cert_check

  • openstack-infra/puppet-statusbot

  • openstack-infra/puppet-storyboard

  • openstack-infra/puppet-subversion

  • openstack-infra/puppet-sudoers

  • openstack-infra/puppet-tmpreaper

  • openstack-infra/puppet-ulimit

  • openstack-infra/puppet-unattended_upgrades

  • openstack-infra/puppet-unbound

  • openstack-infra/puppet-user

  • openstack-infra/puppet-zuul



DNS Entries



Each new module will have its own documentation.




  • Unit tests: We currently only lint and syntax-check the modules in config. They should also have rspec-beaker and server-spec tests written for them (even if we don’t move them to their own project).

  • Integration tests: We need to test that changes to the new projects do not break config (such as with changes to a class’s parameter list).

Developer Impact

By migrating from a single project to many projects, developers will no longer be able to atomically change multiple modules at the same time. This means that changes that touch multiple modules will have to be made in a backwards-compatible way with soft dependencies between changes (such as two changes mentioning each other in their commit messages). Requiring backwards-compatible changes will also make it easier for downstream consumers to use the modules.