Skip to content

openshift-eng/aos-cd-jobs

Repository files navigation

aos-cd-jobs

This repository backs Jenkins jobs on a couple of Jenkins masters.

Jenkins pipeline definitions under scheduled-jobs/

Scheduled pipeline definitions are stored in this directory so they are not indexed by the process described below and turned into a branch on the multi-branch pipeline. This is done to facilitate enabling and disabling the jobs without needing to change the source code on the repository, and any job that requires that should be under this directory.

Job Name Description
build/ose Runs build/ose daily. Presently used to build 3.6 for daily integration test environments.
build/t-th Runs build/ose every Tuesday and Thursday for particular builds of OCP.

Jenkins pipeline definitions under jobs/

An internal Continuous Infrastructure Jenkins instance indexes Jenkinsfiles in the branches of this repository. The branches are automatically generated from the Jenkinsfiles that live under the jobs/ directory on the master branch. The job responsible for generating, updating and removing the branches can be found in the Jenkinsfile at the root directory. The branch update job is configured to be executed periodically, but can be manually triggered in jenkins.

The scripts used by the job described above are pruner.py, which removes branches for jobs that no longer exist, and updater.py, which creates/updates branches for existing jobs. A "job" is any directory under the jobs/ directory which contains a Jenkinsfile. Every branch is an orphan (doesn't contain any history) and its contents are the contents of the master branch with the corresponding directory under jobs/ copied to the root directory and the jobs/ directory removed.

As an example, the contents of the root and jobs/build/openshift-scripts directories in master are currently:

├── build-scripts
│   └── …
├── Jenkinsfile
├── jobs
│   …
│   └── build
│       └── openshift-scripts
│           ├── Jenkinsfile
│           ├── README.md
│           └── scripts
│               └── merge-and-build-openshift-scripts.sh
…
└── README.md

The final contents of the build/openshift-scripts branch, after the execution of the job, will be:

├── build-scripts
│   └── …
├── Jenkinsfile
├── README.md
…
└── scripts
    └── merge-and-build-openshift-scripts.sh

Note that the files Jenkinsfile and README.md in the master branch exist both in the root directory and in the job directory. Because of the sequence of steps described above, the former will be overwritten by the latter.

Jobs under the jobs/build/ directory are indexed at the aos-cd-builds grouping. Some jobs are described below.

Job Name Description
build/ocp Main build task for OCP 3.7. Also builds openshift-ansible 3.7 and all OCP images.
build/ose Main build task for OCP <=3.6. Also builds openshift-ansible artifiacts and jenkins images.
build/openshift-scripts Builds RPMs and container images for the OpenShift Online team.
build/refresh-images
build/scan-images Scans the images for CVEs using openscap.
sprint/stage-to-prod Promote RPMs from the staging repositories to the production repositories (Copies files from latest/ in the enterprise online-stg repo to online-prod/lastest. Also copies files from libra rhel-7-libra-stage to libra's latest online-prod in a new directory based on the day's date.).
sprint/control Send out messages about dev/stage cut to engineering teams.
package-dockertested Tests new Brew builds of Docker and tags them into a mirror repo for use by the CI systems.
starter/operation Run specific operations on starter clusters.
starter/upgrade Runs an openshift-ansible based upgrade on a starter cluster.

Jenkins Job Builder configuration under jjb/

Jenkins Job Builder definitions under the jjb/ directory are not currently used to underpin any jobs, but were an investigation into how the JJB system was used by the AOS CI team to build and support CI jobs for the openshift-ansible repository.

Continuous Upgrade job configuration under continuous-upgrade/

Continuous Upgrade job is using Jenkins Job Builder framework to continuously upgrade an Openshift cluster.

To be able to generate XML configuration of continuous-upgrade jobs you need to install jenkins-jobs tool. After installing the tool run continuous-upgrade/generate-jobs.py to re-generate XMLs of the jobs.

To push the changes in any of the jobs to the server use:

sjb/push-update.sh continuous-upgrade/generated/continuous-upgrade_JOB_NAME.xml

Custom XML Generator configuration under sjb/

A custom XML generator lives under the sjb/ directory. This generator is meant to be a tightly scoped tool that would help us bridge the gap between monolithic scripts inside of Freestyle Jenkins Jobs and segmented Jenkins Pipelines driven by source- controlled Groovy scripts and libraries.

The generator understands a small set of actions, each of which is underpinned by a Python module under sjb/actions/. A configuration YAML file is read in by sjb/generate.py and used to generate a set of input variables to the Jinja job template XML. Jobs can depend on a parent to reuse configuration. Documentation on the YAML syntax can be found at syntax.md.

A typical workflow for a developer making changes to the job would look like:

  • make edits to a configuration file under sjb/config/
  • run sjb/generate.sh
  • commit changes
  • run sjb/push-update-automatic.sh once changes are approved and merged into master

Your local environment needs Python dependencies installed to run sjb/generate.sh - this can be done via the command $ pip install -r sjb/requirements.txt. You will also need pip, which comes bundled with most Python distributions.

In order to test a job, it is necessary to copy a configuration file under sjb/config to a new YAML file with a different name, then re-generate XML and use the following command to push only your test job up to the server:

sjb/push-update.sh sjb/generated/YOUR_TEST_JOB.xml

Cleanup of these jobs post-test is still manual.

If changes are being made to the files under sjb/ in this repository, it is not enough to copy a job configuration and run it to test the changes. Instead, it will be necessary to mark the copied job as syncing a pull request for aos-cd-jobs using the type field on the repository as per the spec. Then, when running your copied job, configure it at run-time to merge in your pull request by entering in your pull request number in the appropriate parameter field in the Jenkins UI when starting the job.

Push Credentials

Note: the sjb/push-update{,-automatic}.sh scripts expect $USERNAME and $PASSWORD to be set as envars when they are run. $USERNAME is your user with which you log in to the Jenkins master at ci.openshift. $PASSWORD is a Jenkins API token you have to generate through the Jenkins UI. As a logged-in user, click your username in the upper right hand of the UI. After the account page loads, click "Configure" on the right hand side, and after the configuration page loads, you will see an option to generate a new token. Copy this to your password store, since it is only displayed for copy/pasting when you first generate it. The $USERNAME and $PASSWORD are used for basic auth against the server on push actions.

Pull Request approvers under approvers/

In order to ensure that pull requests are only merged during phases of a sprint where they are appropriate, all [merge] jobs now call out to an approver on the Jenkins master that will determine if the pull request should merge into the specific branch and repo that it targets.

When running [merge] on a PR, developers will optionally be able to add [severity: value] extensions, where value can take:

  • none ( [merge] )
  • bug ( [merge][severity: bug] )
  • blocker ( [merge][severity: blocker] )
  • low-risk ( [merge][severity: lowrisk] )

The lowrisk severity is special in that all approvers other than the closed_approver.sh, will allow merges with it. Developers should use this tag when they are making changes to code in the repository that does not make up any part of the shipped product and therefore does not have any chance of impacting deployments.

There will be four possible designations for any branch in your repo:

Pull Request Severity
None Bug Blocker Low-Risk
Branch Stage
Open ✔️ ✔️ ✔️ ✔️
DevCut ✔️ ✔️ ✔️
StageCut ✔️ ✔️
Closed

Consulting an Approver

In order to determine if a pull request should merge, consult the approve.sh script on the Jenkins master on which the job runs:

approve.sh "${REPO}" "${TARGET_BRANCH}" "${MERGE_SEVERITY:-"none"}"

Configuring Branch Status

To configure a branch status, run the configure_approver job on the ci.dev Jenkins master. This job will configure the approver you ask for as well as propagate the changes to the ci.openshift server. The job runs the configure_approver script:

for repo in ${REPOSITORIES}; do
    for branch in ${BRANCHES}; do
        configure_approver.sh "${repo}" "${branch}" "${STAGE}"
    done
done

list_approvers.sh

Approver Design

Approvers are configured by creating a symbolic link at ~jenkins/approvers/openshift/${REPO}/${TARGET_BRANCH}/approver for the approver that is requested for that branch. The approvers are the closed_approver.sh, open_approver.sh, devcut_approver.sh, and stagecut_approver.sh scripts in this repository under approvers/.

Developer Workflow

Development on approver scripts in this repository is fairly straightforward. When your changes are ready and have been merged, run the push.sh script to deploy your changes to the Jenkins masters. You will need to have your SSH config set up for the ci.openshift and ci.dev.openshift hosts in order for this script to work.