Skip to content

Paternoster allows you to run Ansible playbooks like ordinary Python or Bash scripts.

License

Notifications You must be signed in to change notification settings

Uberspace/paternoster

Repository files navigation

Paternoster Build Status

Paternoster enables ansible playbooks to be run like normal bash or python scripts. It parses the given parameters using python's argparse and the passes them on to the actual playbook via the ansible API. In addition it provides an automated way to run commands as another user, which can be used to give normal shell users special privileges, while still having a sleek and easy to understand user interface.

Ansible 2.1.x to 2.10.x as well as python 2.7 to 3.8 is supported and tested automatically. We recommend using ansible 2.8+ and python 3.6+.

Once everything is set up, a paternoster script can be used like this:

$ create-user --help
usage: create-user [-h] -u USERNAME [-v]

Create a user.

required arguments:
  -u USERNAME, --username USERNAME
                        name of the user to create

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         run with a lot of debugging output
$ create-user -u luto
creating user luto

The script looks like a normal ansible playbook, except for a few additions. Firstly, it uses a different shebang-line, which kicks off paternoster instead of ansible. Secondly, there is a special play at the beginning of the playbook, which contains the configuration for parameter parsing and other features.

#!/usr/bin/env paternoster

- hosts: paternoster
  vars:
    description: Create a user.
    parameters:
      - name: username
        short: u
        help: "name of the user to create"
        type: paternoster.types.restricted_str
        required: yes
        type_params:
          regex: "^[a-z]+$"

- hosts: localhost
  tasks:
    - debug: msg="creating user {{ param_username }}"

For more information on how to develop scripts using paternoster, please refer the the corresponding sub-document: doc/script_development.md.

Privilege Escalation

Paternoster also provides an automated way to run commands as another user. To use this feature, set the become_user to the desired username. This causes paternoster to execute itself as the given user using sudo. For this to work a sudoers-config has to be created by the developer.

Please refer to the Deployment section of this document for further details.

Deployment

Python-Module

The python module can be installed using pip: pip install paternoster.

Note that this is the only distribution packaged by us. We do not and cannot check the content of all other, following methods.

Fedora

Paternoster is also available as a Fedora package.

dnf install paternoster

RHEL/CentOS

Paternoster is also available in EPEL for RHEL7, RHEL8, CentOS7 and CentOS8.

yum install epel-release
yum install paternoster

AUR

Paternoster is also available as an AUR package for arch linux.

sudo

If you are planning to let users execute certain commands as root, a few changes to your sudo-configuration are needed. boils down to:

ALL ALL = NOPASSWD: /usr/local/bin/your-script-name

This line allows any user to execute the given command as root.

Please refer to the sudoers(5)-manpage for details.

Notes

  • This library makes use of the tldextract-module. Internally this relies on a list of top level domains, which changes every so often. Execute the tldextract --update-command as root in a cronjob or similar to keep the list up to date.

Library-Development

Most tasks can be achieved by writing scripts only. Therefore, the library does not need to be changed in most cases. Sometimes it might be desirable to provide a new type or feature to all other scripts. To fulfill these needs, the following section outlines the setup and development process for library.

Setup

To get a basic environment up and running, use the following commands:

virtualenv venv --python python2.7
source venv/bin/activate
python setup.py develop
pip install -r requirements.txt
pre-commit install --overwrite --install-hooks

This project uses Python 2.7, because Python 3.x is not yet supported by ansible. All non-ansible code is tested with python 3 as well.

Vagrant

Most features, where unit tests suffice can be tested using a virtualenv only. If your development relies on the sudo-mechanism, you can spin up a [vagrant VM][] which provides a dummy uberspace-add-domain-script as well as the library-code in the /vagrant-directory.

vagrant up
vagrant ssh

And inside the host:

Last login: Wed Aug  3 17:23:02 2016 from 10.0.2.2
[vagrant@localhost ~]$ uberspace-add-domain -d a.com -v

PLAY [test play] ************** (...)

If you want to add your own scripts, just add the corresponding files in vagrant/files/scripts. You can deploy it using the following command: ansible-playbook vagrant/site.yml --tags scripts. Once your script has been deployed, you can just edit the source file to make further changes, as the file is symlinked, not copied.

Tests

Linter

To lint the source code, you can run:

tox -e lint

Unit Tests

The core functionality of this library can be tested using the tox- command. If only Python 2.x or 3.x should be tested, the -e parameter can be used, like so: tox -e py36-ansible23, tox -e py27-ansible22. New tests should be added to the paternoster/test-directory. Please refer to the pytest-documentation for further details.

NOTE: you might need to install the the proper "devel" package, for the Python versions you want to test.

Integration Tests

Some features (like the become_user function) require a correctly setup Linux environment. They can be tested using the provided ansible playbooks in vagrant/tests.

The playbooks can be invoked using the run_integration_tests.py-utility:

$ ./vagrant/run_integration_tests.py --file test_variables.yml
=== running test_variables.yml with ansible>=2.1,<2.2
=== running test_variables.yml with ansible>=2.2,<2.3
=== running test_variables.yml with ansible>=2.3,<2.4
$ ./vagrant/run_integration_tests.py ansible22 --file test_variables.yml
=== running test_variables.yml with ansible>=2.2,<2.3
$ ./vagrant/run_integration_tests.py --help
usage: run_integration_tests.py [-h] [--file FILE]
                                [{ansible21,ansible22,ansible23,all}]

Run paternoster integration tests.

(...)

Boilerplate

A typical ansible playbook for a system-test might look like this:

- name: give this test a proper name
  hosts: all
  tasks:
    - include: drop_script.yml
      vars:
        ignore_script_errors: yes
        script_params: --some-parameter
        playbook: |
          - hosts: all
            tasks:
              - debug: msg="hello world"
        script: |
          #!/bin/env python2.7
          # some python code to test

    - assert:
        that:
          - "script.stdout_lines[0] == 'something'"

Most of the heavy lifting is done by the included drop_script.yml-file. It creates the required python-script & playbook, executes it and stores the result in the script-variable. After the execution, all created files are removed. After the script has been executed, the assert-module can be used to check the results.

There are several parameters to control the behavior of drop_script.yml:

Name Optional Description
script no the content of a python script to save as /usr/local/bin/uberspace-unittest and execute
playbook yes (default: empty) the content of a playbook to save as /opt/uberspace/playbooks/uberspace-unittest.yml
ignore_script_errors yes (default: false) whether to continue even if python script has a non-zero exitcode
script_params yes (default: empty) command line parameters for the script (e.g. "--domain foo.com")

Releasing a new version

Assuming you have been handed the required credentials, a new version can be released as follows.

  1. adapt the version in setup.py, according to semver
  2. commit this change as Version 1.2.3
  3. tag the resulting commit as v1.2.3
  4. push the new tag as well as the master branch
  5. update the package on PyPI:
rm dist/*
python setup.py sdist bdist_wheel
twine upload dist/*

License

All code in this repository (including this document) is licensed under the MIT license. The logo (both the png and svg versions) is licensed unter the CC-BY-NC-ND 4.0 license.