This repository is archived and Perforce will no longer be updating this repository. For more information, see this Puppet blog post.
- Overview
- Runtime requirements
- Building puppet-agent
- Branches in puppet-agent
- License
- Code Owners
- Running Tests
The puppet agent is a collection of software that is required for puppet and its dependencies to run. This includes puppet, facter, and other Puppet software, but also vendored dependencies like ruby, curl, openssl, and more.
This repository contains configuration to build puppet-agent and the facter gem for all of Puppet's supported platforms using vanagon, a packaging utility.
The full list of software components built into the puppet agent and the facter gem can be found in their project definitions, and each of the components has its own configuration in the components directory.
Components that are not developed by Puppet (like ruby, curl, or openssl) are built separately into a tarball and consumed here in the puppet-runtime component. See the puppet-runtime project for more information and a full list of the vendored dependencies it provides.
pxp-agent is built separately into a tarball and consumed here in the pxp-agent component. See the pxp-agent-vanagon project for more information.
Ruby and bundler are required to build puppet-agent. The Gemfile specifies all of the necessary ruby libraries to build a puppet-agent package. Additionally, puppet-agent requires a VM to build within for each desired package.
The location of Vanagon in the Gemfile can be overridden with the environment variable VANAGON_LOCATION. Can be set prior to bundle install or updated with bundle update.
0.3.14- Specific tag from the Vanagon git repohttps://github.com/puppetlabs/vanagon#version- Remote git location and version (can be a ref, branch or tag)file:///workspace/vanagon- Absolute file pathfile://../vanagon- File path relative to the project directory
By default, headers and other files that aren't needed in the final puppet-agent package will be removed as part of the cleanup component. If you'd like to keep these files in the finished package, set the DEV_BUILD environment variable to some non-empty value. Note that this will increase the size of the package considerably.
If you wish to build puppet-agent yourself:
- First, build the puppet-runtime for your target platform and agent version.
- Run
bundle installto install required ruby dependencies. - Update the
locationandversionin the puppet-runtime component json file as follows:locationshould be a file URL to your local puppet-runtime output directory, for example:file:///home/you/puppet-runtime/outputversionshould be the version of puppet-runtime that you built; You can find this value at the top level of the json metadata file produced by the build in your puppet-runtime output directory.
- You can disable the packaging of pxp-agent by setting
NO_PXP_AGENTENV variable. If you want to build an agent package that also contains pxp-agent you need to update thelocationandversionin the pxp-agent component json file as follows:
locationshould be a file URL to your local pxp-agent- output directory, for example:file:///home/you/pxp-agent-vanagon/outputversionshould be the version of pxp-agent that you built; You can find this value at the top level of the json metadata file produced by the build in your pxp-agent output directory.- You also may need to change the source URIs for some other components. We recognize this is less than ideal at this point, but we wanted to err on the side of getting this work out in public rather than having everything perfect. If you have your own mirror of the components of puppet-agent, you can also use a rewrite rule. See the Vanagon README for an example.
-
Now use vanagon to build the puppet-agent. Run the following:
bundle exec build <project-name> <platform> <vm-hostname>
Where:
- project name is a project from configs/projects (this can be
puppet-agent), - platform is a platform supported by vanagon and defined in the
configs/platforms directory (for example,
el-7-x86_64), and - the vm hostname is the hostname of a vm matching the desired platform. The current user must be able to ssh into that vm as root (vanagon has facilities to provide an ssh key beyond what is listed in .ssh/config).
- project name is a project from configs/projects (this can be
Tracking branch (main + stable):
- some components may reference tags if they’re slow moving (ruby, openssl)
- some components reference SHAs promoted by a CI pipeline (generally puppet-agent#main pipelines track components' main branches, and likewise for stable)
Guidelines on Merging Between Branches
- stable should be merged to main regularly (e.g. per commit), as is done for component repos; no PR needed
- main should be merged to stable as-needed; typically this is done when a component merges its main to stable, and there are matching changes needed in puppet-agent
Generally, no PR is needed for routine merges from stable to main, but a PR is advised for other merges. Use your judgment of course, and put up a PR if you want review.
Note that for all merges from main or stable, the merge should pick up:
- changes outside of config/components
- changes that bumped to a tag inside config/components
But never:
- changes that bumped to a SHA inside config/components
Here's a sample snippet used for a stable -> main merge:
git merge --no-commit --no-ff stable
for i in {hiera,facter,puppet,pxp-agent,cpp-pcp-client}; do git checkout main -- configs/components/$i.json;done
git commit -m "(maint) Restore promoted components refs after merge from stable"
File issues in the Puppet Agent (PA) project on the Puppet Labs Jira site. Issues with individual components should be filed in their respective projects.
Puppet agent is licensed under the Apache-2.0 license.
See CODEOWNERS