Revision as of 08:25, 31 July 2020

Zuul Based CI

Goals

Bring CI infrastructure based on Zuul for projects hosted on pagure.io and src.fedoraproject.org.
- Gerrit and Github are supported as well and we could for specific cases provide CI for projects hosted there.
Propose jobs and workflow of jobs around Pull Requests for Fedora packages (distgits on src.fedoraproject.org).

For Github projects, Packit should be also considered.

News

June 22, 2020

Support for F30 branch have been removed due to F30 EOL.
pagure.io/fedora-infra/ansible is using Zuul to validate incoming Pull Request.
More and more Python distgit repositories validated by Zuul.
A first Haskell distgit repository is using Zuul: https://src.fedoraproject.org/rpms/ghc-http2/pull-request/1
Full list of Fedora repositories attached to Zuul: https://fedora.softwarefactory-project.io/zuul/projects.

January 24, 2020

Devconf.cz talk: CI/CD for Fedora packaging with Zuul
- Slides with speaker notes: File:CI CD for Fedora packaging with Zuul - final - with notes.pdf

January 14, 2020

Thanks to an update of the Zuul's Pagure driver on softwarefactory-project.io, the zuul user no longer needs to be in the collaborator **admin** group but only in the **commit** group. The project config helper tool has been updated.

November 13, 2019

Introduction of the service on the Fedora CI mailing list https://lists.fedoraproject.org/archives/list/ci@lists.fedoraproject.org/thread/CXDD2VYR6PHTR76JCJ5H5VEBIRG7YL37/

October 22, 2019 - Service is ready to be beta tested

This wiki page contains the process to attach a repository from src.fedoraproject.org or pagure.io to Zuul.
Current known Issue: Sometime, src.fedoraproject.org, does not call Zuul event hooks (https://pagure.io/fedora-infrastructure/issue/8320). Then Zuul is unable to react and run jobs from Pagure events. This does not happen with pagure.io. This is a blocker for CI which needs to be reliable in order to gain confidence, I've pinged infra folks in that issue to help us debug it. UPDATE: "it was DNS" - resolved January 10, 2020

August 28, 2019 - Created Taiga EPIC

Created the Taiga EPIC with the stories to achieve in the goal to provide Pagure PR Zuul jobs for Fedora distgit: https://teams.fedoraproject.org/project/ci/epic/14

Flock 2019

Talk: Pagure CI based on Zuul:
- slides: File:Slides-flock2019.pdf
- video: not published yet

What is Zuul/Nodepool

Zuul [1] is the CI and gating system from the Open Infrastructure Project [2]. It is able to scale and handles by default features such as artifacts sharing between jobs and cross Git repositories testing. You can see Zuul in action here [3].

Below is a list of features proposed by Zuul and its companion Nodepool:

Event-driven pipelines based on Code-Review or Pull-Request workflow: jobs can be triggered automatically when a PR is submitted, changed, approved, merged, or when the repository is tagged.
CI-as-code: jobs are defined as YAML + Ansible playbooks, pipeline definitions as YAML files. Zuul reads and loads those definitions directly from Git repositories.
Support for jobs inheritance, jobs dependencies, jobs chaining (with artifacts sharing).
Speculative testing of new jobs before merging: jobs will be run as they are submitted to make sure they behave as expected.
Cross repositories dependencies: a jobs' workspace can include unmerged patches from other projects if specified
Cross provider: a jobs' workspace can include unmerged patches from other projects even when hosted on different provider like Github and Pagure.
Parallel job run, only capped by resources available or predefined quotas
Automated jobs resources lifecycle management: resources like VMs or containers needed by a given job can be defined in-repository, spawned on demand at a job's start, and destroyed when the job is finished, or held for debugging
Job resources support of OpenStack, OpenShift, K8S, Static nodes, AWS.
Well-defined, reproducible job environments to eliminate flakiness
Speculative testing before merging (gating): if several patches are about to land at the same time, they are tested on the repository's future state.

Until now, Zuul was only able to listen to Gerrit or Github events. Recently a new driver [4] allow Zuul to interface with Pagure as well. Pagure, Zuul and Nodepool could therefore combine into a very efficient CI/CD stack.

How to Zuul attach a Pagure repository on Zuul

Configure the repository for Zuul

In the project settings:

In Project Options
- Web-hooks:
  - For a repository hosted on pagure.io add the following URL into the Web-hooks field: https://softwarefactory-project.io/zuul/api/connection/pagure.io/payload
  - For a repository hosted on src.fedoraproject.org add the following URL into the Web-hooks field: https://softwarefactory-project.io/zuul/api/connection/src.fedoraproject.org/payload
- "User & Groups"
  - Add "zuul" as "ticket" collaborator
- (For gating, optional)
  - Minimum score to merge pull-request: 0 or -1
  - Open metadata access to all: False
  - In "Users & Groups" setting: add "zuul" as "commit" collaborator
  - In "Tags" setting: add the "gateit" tag

This helper script could be used to ease the setup: https://pagure.io/fedora-project-config/blob/master/f/tools/project-settings-helper

Add the repository into the Zuul configuration

Open a Pull Request on https://pagure.io/fedora-project-config

Edit resources/fedora.yaml and add the repository such as:

For a repository hosted on pagure.io:

resources:
  projects:
    Fedora-Zuul-CI:
      ...
      source-repositories:
        ...
        - repository-name

For a repository hosted on src.fedoraproject.org:

resources:
  projects:
    Fedora-Zuul-CI:
      ...
      source-repositories:
        ...
        - repository-name:
            connection: src.fedoraproject.org
            zuul/include: []

Once the Pull Request is accepted and merged, the repository should be available in the Zuul project list: https://fedora.softwarefactory-project.io/zuul/projects

Attach packaging jobs for a distgit repository on src.fedoraproject.org

Note that all repositories with a name that match the regexp '^rpms/.*' will get, by default, the jobs defined by the build, lint and test templates attached. Thus, if the default configuration is sufficient, then no need to edit the zuul.d/projects.yaml file as explained at the end of this section.

We have managed to provide some standard Pull Request's jobs and workflow. This idea is to reduce the amount of manual steps to propose packaging changes to Fedora by taking advantage of CI.

We propose a set of templates (a template can be seen as a workfow of jobs reacting to Pull Request events). Templates are available here https://pagure.io/fedora-zuul-jobs/blob/master/f/zuul.d/templates.yaml.

Templates configure jobs through three types of pipeline (not the same as Jenkins/Gitlab pipeline concept):

check: Jobs within that pipeline are triggered when a Pull Request is Opened or Updated
gate: Jobs within that pipeline are triggered when a Pull Request is approved (prior to be merged and closed)
promote: Jobs with that pipeline are triggered when a Pull Request is closed and merged

Conditions for a Pull Request to be approved:

All check pipelines's jobs succeeded
The Pull Request received the metadata tag: 'gateit'

Available jobs are (https://pagure.io/fedora-zuul-jobs/blob/master/f/zuul.d/jobs.yaml):

rpm-scratch-build: Runs a scratch build on Koji (Koji target based on the PR's branch) and retrieves artifacts on the test node (rpms).
rpm-build: Runs a regular build on Koji (Koji target based on the PR's branch)
rpm-linter: Runs a rpmlint on artifacts (rpms) passed from the parent job
rpm-rpminspect: Runs rpminspect on artifacts (rpms) passed from the parent job. The job also finds the latest build on the related Koji tag and passes it to the rpminspect job so you get the rpminspect diff.
rpm-test: Executes distgit embedded tests tests/tests.yml on the related Fedora node (rawhide VM for master branch, Fedora 30 VM for f30 branch, ...). It is compatible with the Fedora standard-test-interface as STI dependencies are available on the node.

Available templates are (https://pagure.io/fedora-zuul-jobs/blob/master/f/zuul.d/templates.yaml):

build: In the check pipeline, runs the rpm-scratch-build job in the check pipeline.
lint: In the check pipeline, runs, once the parent job rpm-scratch-build is done, the rpm-linter + rpm-rpminspect jobs against the rpms built by the parent job.
test: In the check pipeline, runs, once the parent job rpm-scratch-build is done, the rpm-test jobs against the rpms built by the parent job.
publish: In the gate pipeline, run a noop job (always succeed), to make the gate succeed. This makes Zuul to merge the Pull Request. Then in the promote pipeline, runs the rpm-build job (regular koji build).

To attach a template to a repository you need to open a Pull Request on https://pagure.io/fedora-zuul-jobs-config.

Edit zuul.d/projects.yaml to add:

- project:
    name: repository
    templates:
      - build
      - lint
      - test
      - publish

repository should be like "rpms/python-gear".

Once the Pull Request is merged, the jobs of the chosen templates are attached to the repository.

Currently f30, f31 and master branches are supported. Only the x86_64 arch is supported.

Sequence diagrams of the PR workflow

These diagrams show interactions between components involved in a workflow where build, lint, test, publish templates are used.

Pull Request created/updated/rechecked workflow

Pull Request approved workflow

Example of PR managed with that workflow

https://src.fedoraproject.org/rpms/nodepool/pull-request/6

Nodepool nodes

Jobs are configured to execute on Fedora VMs or containers. The default jobs we provide are configured to use the right node based on the PR target branch (especially *rpm-test* and *rpm-install-test*). For instance the rpm-test job for the master branch execute on a rawhide VM.

Fedora rawhide cloud image (VM) https://softwarefactory-project.io/cgit/config/tree/nodepool/virt_images/cloud-fedora-rawhide.yaml
Fedora 32 cloud image (VM) https://softwarefactory-project.io/cgit/config/tree/nodepool/virt_images/cloud-fedora-32.yaml
Fedora 31 cloud image (VM) https://softwarefactory-project.io/cgit/config/tree/nodepool/virt_images/cloud-fedora-31.yaml

Here the container for fedora-32: https://softwarefactory-project.io/cgit/config/tree/containers/fedora-32 Containers are used to run linter jobs for distgit' PRs.

Fedora's projects using Zuul

Projects registered into the Fedora's Zuul are listed here: projects list

Advanced topics

In-project Zuul configuration (.zuul.yaml)

Zuul is very flexible about where configuration elements are defined. Each piece of a Zuul configuration is managed in a git repository and the full configuration can be spread across an unlimited amount of git repositories.

For distgit projects on src.fedoraproject.org

For every distgit projects we believe that the Zuul pipeline should be similar, that's why be default we do not allow package maintainer to add custom Zuul configurations in-distgit. But for specific cases a maintainer may need custom and non generic jobs and project's pipeline.

If there is a need for such custom in-distgit Zuul configuration then Zuul must be told to read that configuration from the distgit. To do so:

Open a Pull Request on https://pagure.io/fedora-project-config

Edit resources/fedora.yaml and add modify such as:

resources:
  projects:
    Fedora-Zuul-CI:
      ...
      source-repositories:
        ...
        - repository-name:
            connection: src.fedoraproject.org
            zuul/include:
              - project
              - job

This will let Zuul load jobs and project stanza from the .zuul.yaml stored in the distgit repository.

Then is the distgit repository, open a Pull Request with a .zuul.yaml such as:

 - project:
     templates:
       - build
       - lint
       - test
     check:
       jobs:
         - noop

The noop special job should run as long as other jobs from the templates. From there you can write a custom job to replace noop. This project pipeline allows to keep the default Zuul jobs for a distgit project but also to add custom jobs to the default pipeline.

In the .zuul.yaml add:

 - job:
     name: my-job
     description: My custom job
     run: ci/run.yaml

And in a file called ci/run.yaml write the job Ansible playbook:

 - hosts: all
   tasks:
     - name: List project directory on the test system
       command: ls -al {{ansible_user_dir}}/{{zuul.project.src_dir}}

For a source project on pagure.io

The default configuration lets Zuul load all Zuul configuration elements.

Then is the repository, open a Pull Request with a .zuul.yaml such as:

 - project:
     check:
       jobs:
         - noop

The noop special job should run as long as other jobs from the templates. From there you can write a custom job to replace noop.