Skip to content

Contributing#

Pull Request Checklist#

  1. Make sure your change does not break idempotency tests. See Testing (or let CI run the tests for you if you are certain it is idempotent.) If a task cannot be made idempotent, add the tag molecule-idempotence-notest.
  2. Unless a change is small or doesn't affect users, create an issue on GitHub.
  3. Add a changelog update.
  4. Write an excellent Commit Message. Make sure you reference and link to the issue.
  5. Push your branch to your fork and open a Pull request across forks.
  6. Add GitHub labels as appropriate.

Testing#

The tests can be run as they are on GitHub Actions with tox, or they can run with various options using molecule directly.

Requirements#

Install Docker, and add yourself to the group that is authorized to administer containers, and log out and back in to make the permissions change take effect. The authorized group is typically the "docker" group:

gpasswd --add "$(whoami)" docker

NOTE: Docker containers can differ from bare-metal or VM OS installs. They can have different packages installed, they can run different kernels, and so on.

Using Molecule#

  1. Install molecule, molecule-docker, ansible and ansible-lint. It is recommended that you do so with pip in a virtualenv for Python 3.6+.

  2. Install collection dependencies

ansible-galaxy collection install -r ./requirements.yml

  1. Run molecule commands.

Test the default scenario, release-static, on all hosts (linux distros.)

molecule test

Test a specific scenario.

molecule test --scenario-name source

Test all scenarios on all hosts.

molecule test --all

Use debug for increased verbosity.

molecule --debug test

Create and provision, but don't run tests or destroy.

molecule create
Run the main test phase "converge" after creating and provisioning.
molecule converge

Explanation of Different Molecule Scenarios#

The molecule scenarios have names like release-static, which we will refer to as "prefix-suffix".

There is a 3 by 3 matrix of them.

The prefixes are:

  1. release - Install stable release from PyPI
  2. source - Install from git checkout. Different codepaths; preflight check does not exist.
  3. packages - Install from distro (RPM) packages. Even more different codepaths; compatibility checks (preflight) is the job of the repo maintainer.

The suffixes are:

  1. static - Statically specify the roles in the "roles:" syntax in the main playbook, such as in the example playbooks.
  2. cluster - Runs as few roles as possible per host, creating a pulp cluster, rather than a set of independent pulp hosts. Catches any issues. Also, run roles 1 at a time via include_role: under tasks: in the main playbook. Catches undeclared dependencies between roles, and other "dynamic" include errors. This covers use cases such as users running a 3rd party role, setting vars, and later running our role.
  3. upgrade - Upgrade an existing Pulp 3.y container, to test upgrading Pulp 3.y. Roles are applied statically. Depends on said containers existing on a registry, and having been built manually (See instructions below))

release-static is symlinked to default, so that commands like molecule test will use it.

There are other (intentional) differences between tests:

  1. static - These include using unix sockets for the webserver to connect to pulp-api & pulp-content. The remainder use TCP connections (upgrade because that's what older installs only did, cluster because containers communicate via TCP.)
  2. The release-upgrade scenario uses its own converge.yml playbook instead of the default one used by all other scenarios. This playbook upgrades a Pulp installation multiple times to ensure that all upgrade scenarios work correctly.
  3. The source-static scenario defines paths different than the default for the following variables: pulp_media_root, pulp_settings.working_directory, pulp_user_home, pulp_install_dir, pulp_config_dir and developer_user_home.
  4. Misc tests test different distros. This is intended to ensure that many distros are tested, but that tests do not take too long.

In order for cluster scenarios to work on your local system, you must do the following to enable container networking: 1. Run firewall-cmd --zone=public --add-masquerade --permanent (assuming your firewall zone is public). 1. Run firewall-cmd --permanent --zone=public --add-rich-rule='rule family=ipv4 source address=172.27.0.0/16 accept' (assuming your firewall zone is public). 1. Create /etc/sysctl.d/10-allow_docker_networking.conf with and reboot:

net.bridge.bridge-nf-call-iptables=0
net.bridge.bridge-nf-call-arptables=0
net.bridge.bridge-nf-call-ip6tables=0

To test both webserver solutions we testing apache as webserver with

  • packages-cluster
  • source-cluster
  • release-static

All others scenarios use nginx as a webserver.

Docs Testing#

Cross-platform:

pip install mkdocs pymdown-extensions mkdocs-material mike mkdocs-git-revision-date-plugin

Then:

mkdocs serve
Click the link it outputs. As you save changes to files modified in your editor, the browser will automatically show the new content.

NFS & SELinux testing#

Here are example commands for testing the functionality prescribed for the variable pulp_selinux_remount_data_dir:

vagrant up --no-destroy-on-error pulp3-sandbox-centos9-stream
(hit ctrl-c while ansible-galaxy is running, or early on in the ansible run)
vagrant ssh pulp3-sandbox-centos9-stream
sudo -i
dnf install -y nfs-utils
systemctl enable --now nfs-server.service
mkdir -p /data/var/lib/pulp/pulpcore_static/
mkdir -p /var/lib/pulp/pulpcore_static/
echo '/data/var/lib/pulp 127.0.0.1(rw,no_root_squash)' >> /etc/exports
exportfs -a
echo '127.0.0.1:/data/var/lib/pulp /var/lib/pulp nfs defaults,_netdev,nosharecache 0 0' >> /etc/fstab
echo '127.0.0.1:/data/var/lib/pulp/pulpcore_static /var/lib/pulp/pulpcore_static,nosharecache nfs defaults,_netdev,context="system_u:object_r:httpd_sys_content_rw_t:s0" 0 0' >> /etc/fstab
mount -a
exit
exit
vagrant provision pulp3-sandbox-centos9-stream

Creating Molecule Upgrade Test Containers#

  1. Identify a version from 3.14 or older (because the release-upgrade tests 1st test upgrading to 3.14 before the current version. This should be revisited in the future.)
  2. Identify how you will build the old version. Preferably using the old version of the installer. But if necessary, modifying the version variables with the current installer.
  3. Modify molecule/release-static/molecule.yml to test the desired distro if it is not in it. Add a host_vars if you do this.
  4. Modify molecule/release-static/group_vars (or host_vars) so that the webserver is nginx rather than apache, and so that pulp_container is not installed. Also modify the version variables (including pulpcore_version) if need be. Install release versions of pulpcore and plugins.
  5. molecule create && molecule converge
  6. If that failed, be prepared to put workarounds on the branch. Such as modifying prereq_pip_packages.
  7. docker exec -it <container-name> /bin/bash
  8. dnf clean all
  9. rm -rf /var/lib/pulp/.cache
  10. exit
  11. docker commit <container-name <tag-name> with a tag like "quay.io/pulp/pulp_installer-pip-ci-f36:3.13.0"
  12. Modify molecule.yml for release-upgrade and source-upgrade (or packages-upgrade)
  13. Test the image locally (molecule create -s release-upgrade && molecule converge -s release-upgrade)
  14. docker push <tag-name>
  15. If this is the 1st image of a new distro, go to https://quay.io/organization/pulp, set the descritpion of the repo based on existing ones, and set the repo to public.
  16. Create a PR.

Specifying Supported Distros#

We want to support the entire family of RHEL clones, even though we only test CentOS and Rocky in our CI.

We also want to workaround (undocumented) limitation of Ansible 2.9 not recognizing Rocky as a member of the "Red Hat" family.

Therefore, in order to support the entire family of RHEL clones, add logic like the following

when:
  - __pulp_os_family == 'RedHat
  - ansible_facts.distribution_major_version|int in [7,8,9,10]


Last update: 2023-01-23