vps-deploy 0.4.0a4

Common commands for deployment on a VPS.

This Python package contains a number of Fabric 1.* and 2.* deployment functions that we use across several Python/Django projects.

Fabric is a very handy tool for automating deployment processes. Unlike Salt or Ansible, you’re writing code from the beginning than diving into the Turing Tarpit of templated YAML. Also unlike Salt, Ansible or shell scripts, Fabric neatly coordinates local and remote SSH tasks. It might not necessarily be your best choice for provisioning, but as a general tool for automatic custom workflows, it is exceptional.

Installation (recommended Fabric 2.*)

If Fabric 2 is available as an operating system package:

$ sudo apt install fabric invoke
$ pip install --user vps-deploy

Otherwise:

$ pip install --user fabric invoke vps-deploy

Installation (legacy Fabric 1.*)

Fabric 1.* is recommended only for supporting existing projects. For new projects, use Fabric 2.*.

Since Fabric 1.* is Python 2-only and you’re probably using a Python 3 virtual env, you’ll need to install Fabric and VPS Deploy as a –user package:

# System package for Fabric, user package for vps-deploy:
$ sudo apt install fabric
$ pip install --user vps-deploy

# All in user packages:
$ pip install --user fabric~=1.14 vps-deploy

Alternately, while Fabric 1.* remains available for some operating systems, you could also install it that way.

Getting started

To get started, create a fabfile.py in your top-level project directory. For a standard Python/Django project based on Nginx, uWSGI Emperor, Memcached and PostgreSQL, it might look something like this:

from fabric import task
from invoke.collection import Collection

from vps_deploy import django_fabric2 as df2

hosts = ['[e-mail]']

@task(hosts=hosts)
def deploy(c):
    df2.transfer_files_git(c)
    df2.init(c)
    df2.grep_for_pdb(c)
    df2.lint(c)
    df2.prepare_virtualenv(c)
    df2.prepare_django(c)
    df2.fix_permissions(c,
        # Sentry needs access to Git repo.
        read=[
            '.git', 'deploy', 'env', 'project'],
        read_write=['project', 'project/collected_static/CACHE'],
    )
    df2.reload_uwsgi(c)
    df2.flush_memcached(c)
    df2.update_nginx(c)

ns = Collection(
    deploy,
    task(df2.download_postgres_db),
    task(df2.mirror_postgres_db),
    task(df2.mirror_media),
)
ns.configure({
    # Built-in Fabric config.
    'run': {
        'echo': True,
        # Needed so local commands work. Can also use FABRIC_RUN_REPLACE_ENV.
        'replace_env': False,
        # Needed for Guix. Can also use FABRIC_RUN_SHELL.
        # 'shell': '/run/current-system/profile/bin/bash',
    },

    # Our custom project config.
    'env': {
        'branch': 'master',
        'app_user': 'www-data',
        'db_name': 'exampledb',
        'user': 'exampledeployuser',
        'project_dir': '/srv/exampleproject',
        'media_dir': 'project/media',
        'virtualenv': '/srv/venvs/exampleproject-django-py38',
        'site_name': 'examplesite',
        'requirements': 'requirements/production.txt',
        'settings': 'project.settings.live',
        'uwsgi_conf': 'deploy/uwsgi.ini',
        'nginx_conf': 'deploy/nginx.conf',
        'python': '/usr/bin/python3.8',
    },
})

Deploying

To make a deployment:

fab –prompt-for-sudo-password deploy

This depends on a few things being already set up, such as SSH access to the server and having the server-site software and accounts set up. Those tend to to be better handled with configuration management tools like Salt or Ansible (and potentially triggered by Fabric!).