Vagga in Docker

Status:	PoC / tracking issue

This is a prototype which brings vagga as the first-class tool to OS X and (possibly) windows through docker's layer of compatibility.

Also see Running in Gitlab

Installation

Currently it's (you need python >= 3.4):

$ pip3 install git+http://github.com/tailhook/vagga-docker
[ .. snip .. ]
$ vagga
Available commands:
    run

WARNING: ensure that docker that you're running is docker for mac, and not any other docker running by docker-machine or vagrant. Probably do this by running:

printenv DOCKER_HOST

If this returns non-empty string run:

unset DOCKER_HOST

But this only cleans environment for current shell. So you must fix your docker host where it is defined.

For the reference, error usually looks like this:

Config not found in path "/work"

(while the config actually there, basically it means docker doesn't see the volume from your host system)

Short FAQ

Why is it in python? For a quick prototype. It will be integrated into vagga as soon as is proven to be useful. Or may be we leave it in python if it would be easier to install.

So should I try this version or wait it integrated in vagga? Definitely you should. The integrated version will work the same.

Why this uses `docker run` instead of API? We don't want to reimplement whole tty handling that is done by docker. If it is proven to be bad descision we will revisit it later. Also we use API for other things that are not running anything.

Is there any difference between this and vagga on linux? There are two key differences: you need to export ports that you want to be accessible from the host system. And we keep files of a container filesystem inside a docker volume (the reasons are here) However, you can export some part of the filesystem that is non-sensible for ownership semantics, like this:

containers:
  django:
    setup:
    - !Alpine v3.3
    - !Py3Install ['Django >=1.9,<1.10']
    _expose-dirs:
    - /usr/lib/python3.5

commands:
  run: !Command
    description: Start the django development server
    container: django
    _expose-ports: [8080]
    run: python3 manage.py runserver

Please report if you find any other differences using the tool. Ah, but exact text of some error messages may differ, don't be too picky :)

Why `_expose-ports` and `_expose-dirs` are underscored? This is a standard way to add extension metadata or user-defined things in vagga.yaml. We will remove the underscores as soon as integrate it into main code. Fixing underscores isn't going to be a big deal.

Will linux users add `_expose-ports` and `_expose-dirs` for me? Frankly, currently probably now. But it's small change that probably noone will need to delete. In the future we want to apply seccomp filters to allow to bind only exposed ports on linux too. And expose-dirs will be used to filter directories that will not be optimized for disk usage, so IDE see them as a normal directory.

(It's also cool project to detect expose-dirs in vagga metadata and add them to the project files automatically. But I'm not IDE guy, so I'm not sure if this is possible, or viable)

What will be changed when we integrate this into vagga? We will move more operations from docker into host system. For example list of commands will be executed by mac os. Also vagga _list, some parts of vagga _clean and so on. But we will do our best to keep semantics exactly the same.

Running in Gitlab

To run a container on gitlab CI's docker infrastructure, use image tailhook/vagga:v0.8.1, like this:

image: tailhook/vagga:v0.8.1

test:
  script:
  - vagga test

Or check out example-project

LICENSE

This project has been placed into the public domain.