Jenkins Infra

This repository is the r10k control repository for the Jenkins project's own infrastructure.

Structure

See the Jenkins infrastructure project for overview of the project's infrastructure and the services being managed by this repository. A non exhaustive list of services is available here.

Implementation notes

The services are managed r10k and Puppet, configuration files are available inside this repository.
There are multiple types of service deployments:
- The majority of services run as containers inside Kubernetes, and are NOT managed here (ref. jenkins-infra/charts)
- Some services like ci.jenkins.io run inside virtual machines provisioned in cloud providers
- the other services are running on bare metal machines provided by sponsors
There are Puppet templates for all services. Configuration options are defined by Hiera and stored in hieradata. See hieradata/common.yaml for the most of the settings.
Not all services are fully configured with Configuration-as-Code. For example, Jenkins masters (buildmaster template) rely on configurations being supplied from Jenkins home directories.

Containerized services

All containerized services are stored in separate repositories (Plugin Site, IRC Bot, etc.). They have their own release cycles and maintainers. This repo just manages and configures the deployments.

See this page for service repository links.
Service images are hosted inside the jenkinsciinfra DockerHub account.
Usually there is a Continuous Delivery pipeline configured for services inside their repositories.
Image versions are defined in the hieradata/common.yaml file by the *:image_tag variables. Services can be updated by submitting a pull request with the version update.

Local development

The amount of testing that can be done locally is as follows:

bundle install - To get the necessary gems to run tests locally, if you're unfamiliar with Ruby development you may want to use RVM to create an isolated Ruby environment
./check - Will run the rspec-puppet unit tests and the puppet-lint style validation. If you intend to run the rspec-puppet over and over, use rake spec_standalone to avoid re-initializing the Puppet module fixtures every time.

Vagrant-based testing

Running server spec tests

We're using serverspec for on-machine acceptance testing. Combined with Vagrant, this allows us to create an acceptance test per-role which provisions and tests an entire Puppet catalog on a VM.

Pre-requisites for Vagrant

Install Vagrant
Install Vagrant plugins: vagrant plugin install vagrant-serverspec
Run the ./vagrant-bootstrap script locally to make sure your local environment is prepared for Vagranting

To launch a test instance, vagrant up ROLE where ROLE is one of the defined roles. You can rerun puppet and execute tests with vagrant provision ROLE repeatedly while the VM is up and running. To just rerun serverspect without puppet, vagrant provision --provision-with serverspec ROLE. When it's all done, deprovision the instance via vagrant destroy ROLE.

Updating dependencies

For reasons that Tyler will hopefully clarify at some point, this module maintains the list of Puppet module dependencies in Puppetfile and .fixtures.yml. They need to be kept in sync. When you modify them, you can have the local environment reflect changes by running bundle exec rake resolve.

Branching model

The default branch of this repository is staging which is where pull requests should be applied to by default.


+----------------+
| pull-request-1 |
+-----------x----+
             \
              \ (review and merge, runs acceptance tests)
staging        \
|---------------o--x--x--x---------------->
                          \
                           \ (manual merge, auto-deploys to prod hosts)
production                  \
|----------------------------o------------->

The branching model is a little different than what you might be familiar with. We merge pull requests into a special branch called staging where we can run Puppet acceptance tests from. Once somebody has code reviewed a pull request it can be merged into staging.

When a infra project team member is happy with the code in staging they can create a merge from staging to production. Once something has been merged to production, it will be automatically deployed to production hosts.

Installing agents

For installing agents refer to the installing agents section of the PuppetLabs documentation.

Adding a new branch/environment

"Dynamic environments" are in a bit of flux for the current version (3.7) of Puppet Enterprise that we're using. An unfortunate side-effect of this is that creating a branch in this repository is not sufficient to create a dynamic environment that can be used via the Puppet master.

The enable an environment, add a file on the Puppet master: /etc/puppetlabs/puppet/environments/my-environment-here/environment.conf with the following:

modulepath = ./dist:./modules:/opt/puppet/share/puppet/modules
manifest = ./manifests/site.pp

Contributing

See this page for the overview and links.

Channels:

#jenkins-infra on the Freenode IRC network
INFRA project in JIRA.
infra@lists.jenkins-ci.org

smerle33 / jenkins-infra