Run tests in a docker container
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Paul Belanger a842b2db46
Add ubuntu-xenial docker image
2 years ago
doc/source Removed utf-8 coding 2 years ago
dockerfiles/infra Add ubuntu-xenial docker image 2 years ago
dox Merge "Replace assertEqual(None, *) with assertIsNone in tests" 2 years ago
tools Add ubuntu-xenial docker image 2 years ago
.coveragerc Change ignore-errors to ignore_errors 3 years ago
.gitignore Ignore dox 4 years ago
.gitreview Update .gitreview for new namespace 3 years ago
.mailmap Initial commit 4 years ago
.testr.conf Initial commit 4 years ago
CONTRIBUTING.rst Workflow documentation is now in infra-manual 4 years ago
HACKING.rst Initial commit 4 years ago
LICENSE Initial commit 4 years ago Initial commit 4 years ago
README.rst Update doc for 1.3 bootdocker 4 years ago
babel.cfg Initial commit 4 years ago
bindep.txt Ensure we can build docker images 2 years ago
dox.yml Allow parent section for dox.yaml 4 years ago
requirements.txt Remove argparse from requirements 3 years ago
setup.cfg Clean up build and write a README 4 years ago Clean up build and write a README 4 years ago
test-requirements.txt Add tests for tox_ini 4 years ago
tox.ini Remove tox py26 and pypi 2 years ago



dox is a tool for using docker containers to run local tests, inspired by tox and virtualenv for python. There are two elements to its configuration:

  • What commands should be run?
  • In what image should they be run?

If there is a dox.yml file, you're set. You want to specify what image to use and what commands to run. You win:

image: ubuntu:trusty
commands: |
  pip install . -r test-requirements.txt
  python test

You might either not be willing to commit to dox as a way of life yet, or you may want to use dox in a project that similarly has not done so.

What commands should be run

dox.yml wins.

If there is a tox.ini file, the commands specified in the base [testenv] will be used.

If there is a .travis.yml file, the script section will be used.

If there are none of those things, dox will do its best to infer what should be done. Examining the directory can often provide hints if you haven't been too clever. For instance, if you have a Gruntfile, you probably want to run grunt. If you have a Makefile, then make && make test is probably your bag. If you have a, you probably want to run autotools first. If you have a file, python test is a likely choice (although in that case, you probably haven't done it right because setuptools support for this is quite awful.)

After all of that, if we still can't figure out what you want - it's probably easiest to just edit a file called dox.yml and put in a section telling us what to do.

In what image should they be run

Again, dox.yml wins, and thanks for making things easy!

If there is a tox.ini file, and it contains a [docker] section, the value in "image" will be used:


If there is not an image key in the docker section or an image key in the dox.yml but there is a Dockerfile in the repo, an new image will be built using the Dockerfile and the test commands will be run inside of the image.

If all of that fails, tests are going to run in a bare ubuntu image. Good luck!

Image Caching

Every run actually has two images associated with it: The base image and the test image.

The image referenced above is the base image. When you execute dox, it will search docker for the named image. If it exists, it will do nothing. If it does not, it will either pull it, or, it will build it from the Dockerfile. In the case of building from a Dockerfile, dox will make an image named for the source directory. So if you're in ~/src/openstack/nova, it'll make an image called "dox/nova/base".

The test image is an image made by dox for the test run. Similar to the base run, it'll have a generated name, such as "dox/nova/test", and similar to the base image, it will get generated once and then left alone.

If you want to regenerate the test image, you can run dox with the -r option. If you want to regenerate/repull everything, you can run it with the --rebuild-all option.

The reasoning behind this is that the base image should really be the substrata which doesn't have a lot to do with the repo itself ... it shouldn't really expect to change much based on day to day changes in the repo. The test image on the other hand is built a bit more based on the repo itself. So, for instance, in the base image you might want to do things like:

apt-get install libxml-dev

and in the test image, you'd want things like:

pip install -U requirements.txt

Neither change frequently, but the second it more likely to change day to day than the first.

Additional information

Regardless, dox will mount the current source dir as a volume at /src in the container and will run commands in that context.

dox will attempt to reuse containers. Since the source is bind-mounted into the container, things that might be expensive like copying source dirs or re-installing the source into the system can be minimized.

Boot2Docker support

To get support for non Linux OSes that doesn't support natively docker there is a tool called boot2docker which allows you to run a remote docker server in a VirtualBox VM and the client running on the non Linux desktop.

Volume support started to be supported by boot2docker since docker version 1.3. You will need to be at least on that version to have volume mount support.

If you username id is different from the docker username on the boot2docker vm, you will need to specify it on the command line, like this:

dox --user-map=docker:1000:10

If you use VirtualBox guest additions to mount your osx onto the boot2docker vm, example /Users/your_name/openstack/oslo-incubator as /home/your_name/openstack/oslo-incubator, then you can add the following mapping:

dos --path-map=/Users/your_name:/home/your_name


It is possible to specify multiple images to be used in a dox run. Images can be provided on the command line, via the dox.yml file, or the tox.ini file.

For the command line, images should be provided via the --images option, in a comma-separated list.

The tox.ini file should also use a comma-separated list.

The dox.yml file should list images in an array.

The same prep and command instructions will be executed on each image sequentially.

dox does not currently allow for multiple images executing different tasks at this time. However, it is a goal to allow for such test scenarios in the future.