pipelines

Build pipelines for automation, deployment, testing...

Pipelines

.. image:: https://cloud.githubusercontent.com/assets/919180/20129399/425a0c2a-a68a-11e6-82ef-b252424a4b48.png :align: center :alt: Pipeline UI screenshot

Pipelines is a simple tool with a web UI to manage running tasks. It supports running tasks manually through a Web UI or automatically via webhooks.

Pipelines is composed of three components:

  • Web UI : User interface that allows users to run tasks, inspect logs and show the task details. It also support username/password authentication.
  • Pipelines API : This is the backend for the Web UI and webh
  • Pipelines library : This component includes core logic for running pipelies such as reading task definitions, handling logging and running the pipelines.

Pipelines is primarily developed to run on Linux / MacOS. Windows support is not available at the moment.

Installation

Requirements:

  • python 2.7 (other versions not tested)
  • pip

Base install

.. code-block:: bash

pip install pipelines

Or get the latest dev version from Github <https://github.com/Wiredcraft/pipelines>_ and run pip install . from within the cloned repo. Or run pip directly from git pip install git+git://github.com/Wiredcraft/[email protected].

Configuration

Pipelines runs solely on files. No database is currently required. All the pipelines, the logs of each run and various temporary files are stored under the workspace folder.

Workspace is a folder that needs to be specified when running pipelines.

.. code-block:: bash

mkdir ~/pipelines_workspace

Drop your pipelines files (see format below) directly at the root of this folder.

Run standalone

Start the API with the following:

.. code-block:: bash

pipelines server --workspace ~/pipelines_workspace --username admin --password admin

You may want to specify a different binding IP address (default: 127.0.0.1) or different port (defaut: 8888). Refer to the pipelines --help for additional parameters.

You can now access pipelines at http://localhost:8888

How to run as a daemon

Create a dedicated user to run pipelines

.. code-block:: bash

# Create a pipelines user
useradd -m -d /var/lib/pipelines -s /bin/false pipelines

# Create the workspace folder (optional)
mkdir /var/lib/pipelines/workspace
chown -R pipelines. /var/lib/pipelines

# Create a SSH key pair (optional)
sudo -u pipelines ssh-keygen

You may want to rely on supervisord to run the API.

.. code-block:: bash

# Ubuntu / Debian
apt-get install supervisor

# CentOS / RedHat (to confirm)
yum install supervisord

Copy and adapt de config file from etc/supervisor/pipelines.conf to /etc/supervisor.

.. code-block:: bash

# Update and reload supervisord
supervisorctl reread
supervisorctl update
supervisorctl start pipelines

Access the web interface at http://localhost:8888

Additionaly you may want to use nginx as reverse proxy as well. See sample config from etc/nginx.

Authentication

Static authentication “”“”“”“”“”`

You can define a static admin user by specifying the following options when running pipelines:

.. code-block:: bash

--username ADMIN_USER
--password ADMIN_PASS

Github Oauth “”“”“”

This is an experimental feature

You can add oauth support from Github to allow teams to access pipelines. You will need to set it by using environment variables for the Oauth Apps, and the --github-auth to limit teams access.

.. code-block:: bash

GH_OAUTH_KEY=my_oauth_app_key \
GH_OAUTH_SECRET=my_super_secret \
pipelines server [--options] --github-auth=MY_ORG/MY_TEAM

You can create Oauth Key/Secret in Github Oauth Applications <https://github.com/settings/developers>_

Note: If you use Github Oauth, you will not be able to use static authentication.

Pipelines file format

Pipeline definition file uses YAML syntax. A few examples below. Pipelines files are meant to be put at the root of your workspace.

Simple example

This is a very basic pipeline definition. Save it in your workspace within a .yaml file (e.g. WORKSPACE/example-pipeline.yaml). It does … nothing really useful TBH.

.. code-block:: yaml

---
# Pipeline definitions are standard yaml and you can include comments inside

# Variables are exposed to all actions through {{ varname }} syntax.
vars:
    code_branch: dev

# Triggers define the automated ways to run the task. In addition to manual execution 
# through the UI, only webhook is supported for now.
triggers:
    - type: webhook

# Actions are the steps that are run for this pipeline. The default action plugin is bash, 
# but you can use others by defining the "type" field.
actions:
    - 'echo "Starting task for {{ code_branch }}"'
    - name: 'My custom name step'
      type: bash
      cmd: "echo 'less compact way to define actions'"
    - 'ls -la /tmp'

Vars

The vars section of the pipeline definition defines variables that will then be available in any of the actions.

.. code-block:: yaml

vars:
    my_var: something

actions:
    - echo {{ my_var }}

You can then use the variables as seen above.

Note:

  • You may have to quote " your vars to respect the YAML format.

Prompts

You can prompt users to manually input fields when they run the pipeline through the web-UI. To do this add a prompt section to your pipeline definition. The prompt fields will override the variables from the vars section.

You can alternatively provide a list of acceptable values; the prompt will then appear as a select field and let you choose from the available values

.. code-block:: yaml

vars:
    # This is the default value when triggered and no prompt is filled (e.g. via webhook)
    my_var: default_no_prompt

prompt:
    # This is the default value when triggered via the web UI
    my_var: default_with_prompt

    # This will appear as a select field
    my_var_from_select:
        type: select
        options:
            - value1
            - value2

actions:
    # This will display:
    #    "default_no_prompt" when call via webhook
    #    "default_with_prompt" when call via UI but keeping the default
    #    "other" when call via UI and "other" is inputted by the user
    - echo {{ my_var }}

    # Depending on the selected value, will display value1 or value2
    - echo {{ my_var_from_select }}

Actions

Default actions use the bash plugin and will execute command as if they were shell commands.

Other actions can be used by specifying another type. Supported types currently are:

  • bash: run bash command.
  • python: write inline script or run python script inside a virtualenv.
  • slack: send message to Slack.

bash “”

See example above.

python “””

The python plugin allows to run python scripts or inline python code.

.. code-block:: yaml

actions:
  - type: python
    script: |
      import json
      a = {'test': 'value', 'array': [1,2,3]}
      print json.dumps(a, indent=2)
  - type: python
    virtualenv:  /opt/venvs/my_env
    file: '/tmp/some_script.py'

Explanation of the fields:

  • script: inline python code to be run against the python interpreter.
  • file: run a python script.
  • virtualenv: run the python code (inline or file) inside a virtualenv.

Note:

  • The path of either virtualenv folder or file need to exist and be on the server. It is currently set relatively to the CWD where the Pipelines api / UI is running from.

slack “”`

The slack plugin allows sending messages over to Slack <https://slack.com>_ (e.g. pipelines execution status).

.. code-block:: yaml

vars:
    slack_webhook: https://hooks.slack.com/services/SOME/RANDOM/StrIng
    name: some_name

actions:
    - type: slack
      message: 'Deployment finished for project {{ name }}.'
      always_run: true

Explanation of fields:

  • type: tells Pipelines to execute the action through the slack plugin.
  • always_run: ensure the action is run all the time - even if a former action failed.
  • message: is the message to send to Slack.

Note:

  • The slack plugin require a slack_webhook vars defined in the vars section of the pipeline.

Slack Hooks URL are defined via the Incoming WebHooks <https://slack.com/apps/A0F7XDUAZ-incoming-webhooks>_ app (Slack API details here <https://api.slack.com/incoming-webhooks>_).

Triggers

Webhooks “”“”

If you want to run your pipeline by triggering it through a webhook you can enable it in the triggers section.

.. code-block:: yaml

triggers:
    - type: webhook

If you open the web-UI you can see the webhook URL that was generated for this pipeline in the “Webhook” tab. You can for example configure GitHub repository <https://developer.github.com/webhooks/creating/>_ to call this url after every commit.

You can access the content of the webhook content in the actions in the webhook_content variable; e.g. echo {{ webhook_content.commit_id }}

Note:

  • You need to send the message via POST as application/json Content-Type.
  • Documentation is coming to explain how to use the content of the data sent through the hook.

Advanced Templates

Pipelines uses Jinja2 <http://jinja.pocoo.org/docs/2.9/templates/>_ to do variables replacement. You can use the whole set of builtin features from the Jinja2 engine to perform advanced operations.

.. code-block:: yaml

prompt:
    stuff:
        type: select
        options:
            - good
            - bad

actions:
    - name: Print something
      type: bash
      cmd: |
          {% if stuff == 'good' %}
            echo "Do good stuff..."
          {% else %}
            echo "Do not so good stuff..."
          {% endif %}

    - name: Use builtin filters
      type: bash
      # Will display 'goose' or 'base'
      cmd: echo {{ stuff | replace('d', 'se') }}

Dirty line by line setup

TODO: Make a real setup script / one-liner script … and not Debian only …

  • apt-get update
  • apt-get upgrade
  • apt-get install python-pip git
  • pip install virtualenv
  • virtualenv pipelines
  • source pipelines/bin/activate
  • pip install pipelines
  • mkdir ~/pipelines_workspace
  • pipelines server --workspace ~/pipelines_workspace --username admin --password admin

Docker

Note: Not heavily tested.

.. code-block:: bash

docker run -d boratbot/pipelines

Roadmap

No definitive roadmap for the moment, mainly focusing on having a lean code base (heavy refactoring to come).

Among the possible features:

  • Improved web UI & features
  • Better webhook management
  • Better management of the tasks
  • CLI
  • Toolbar
  • Improved Auth
  • etc.

Related Repositories

luigi

luigi

Luigi is a Python module that helps you build complex pipelines of batch jobs. It handles dependency resolution, workflow management, visualization etc. It also comes with Hadoop support built in. ...

tpot

tpot

A Python tool that automatically creates and optimizes machine learning pipelines using genetic programming. ...

resty

resty

Little command line REST client that you can use in pipelines (bash or zsh). ...

wercker

wercker

The Wercker CLI can be used to execute pipelines locally for both local development and easy introspection. ...

ramda-cli

ramda-cli

:ram: A CLI tool for processing data with functional pipelines ...