christianlight/tutor
6
0
Fork 0
mirror of https://github.com/ChristianLight/tutor.git synced 2024-11-06 05:17:54 +00:00
Code Issues Releases Activity
1a5c6da972
tutor/docs/dev.rst
Régis Behmo f9402f7879 feat: run all services as unprivileged containers 
With this change, containers are no longer run as "root" but as unprivileged
users. This is necessary in some environments, notably some Kubernetes
clusters.

To make this possible, we need to manually fix bind-mounted volumes in
docker-compose. This is pretty much equivalent to the behaviour in Kubernetes,
where permissions are fixed at runtime if the volume owner is incorrect. Thus,
we have a consistent behaviour between docker-compose and Kubernetes.

We achieve this by bind-mounting some repos inside "*-permissions" services.
These services run as root user on docker-compose and will fix the required
permissions, as per build/permissions/setowner.sh These services simply do not
run on Kubernetes, where we don't rely on bind-mounted volumes. There, we make
use of Kubernete's built-in volume ownership feature.

With this change, we get rid of the "openedx-dev" Docker image, in the sense
that it no longer has its own Dockerfile. Instead, the dev image is now simply
a different target in the multi-layer openedx Docker image. This makes it much
faster to build the openedx-dev image.

Because we declare the APP_USER_ID in the dev/docker-compose.yml file, we need
to pass the user ID from the host there. The only way to achieve that is with a
tutor config variable. The downside of this approach is that the
dev/docker-compose.yml file is no longer portable from one machine to the next.
We consider that this is not such a big issue, as it affects the development
environment only.

We take this opportunity to replace the base image of the "forum" image. There
is now no need to re-install ruby inside the image. The total image size is
only decreased by 10%, but re-building the image is faster.

In order to run the smtp service as non-root, we switch from namshi/smtp to
devture/exim-relay. This change should be backward-compatible.

Note that the nginx container remains privileged. We could switch to
nginxinc/nginx-unprivileged, but it's probably not worth the effort, as we are
considering to get rid of the nginx container altogether.

Close #323.
2021-10-25 16:26:04 +02:00

232 lines
11 KiB
ReStructuredText
Raw Blame History

.. _development:
Open edX development
====================
In addition to running Open edX in production, Tutor can be used for local development of Open edX. This means that it is possible to hack on Open edX without setting up a Virtual Machine. Essentially, this replaces the devstack provided by edX.
The following commands assume you have previously launched a :ref:`local <local>` Open edX platform. If you have not done so already, you should run::
tutor local quickstart
In order to run the platform in development mode, you **must** answer no ("n") to the question "Are you configuring a production platform".
Note that the local.overhang.io `domain <https://dnschecker.org/#A/local.overhang.io>`__ and its `subdomains <https://dnschecker.org/#CNAME/studio.local.overhang.io>`__ all point to 127.0.0.1. This is just a domain name that was setup to conveniently access a locally running Open edX platform.
Once the local platform has been configured, you should stop it so that it does not interfere with the development environment::
tutor local stop
Finally, you should build the ``openedx-dev`` docker image::
tutor dev dc build lms
This ``openedx-dev`` development image differs from the ``openedx`` production image:
- The user that runs inside the container has the same UID as the user on the host, in order to avoid permission problems inside mounted volumes (and in particular in the edx-platform repository).
- Additional python and system requirements are installed for convenient debugging: `ipython <https://ipython.org/>`__, `ipdb <https://pypi.org/project/ipdb/>`__, vim, telnet.
- The edx-platform `development requirements <https://github.com/edx/edx-platform/blob/open-release/lilac.master/requirements/edx/development.in>`__ are installed.
Since the ``openedx-dev`` is based upon the ``openedx`` docker image, it should be re-built every time the ``openedx`` docker image is modified.
Run a local development webserver
---------------------------------
::
tutor dev runserver lms # Access the lms at http://local.overhang.io:8000
tutor dev runserver cms # Access the cms at http://studio.local.overhang.io:8001
Running arbitrary commands
--------------------------
To run any command inside one of the containers, run ``tutor dev run [OPTIONS] SERVICE [COMMAND] [ARGS]...``. For instance, to open a bash shell in the LMS or CMS containers::
tutor dev run lms bash
tutor dev run cms bash
To open a python shell in the LMS or CMS, run::
tutor dev run lms ./manage.py lms shell
tutor dev run cms ./manage.py cms shell
You can then import edx-platform and django modules and execute python code.
To collect assets, you can use the ``openedx-assets`` command that ships with Tutor::
tutor dev run lms openedx-assets build --env=dev
.. _bind_mounts:
Sharing directories with containers
-----------------------------------
It may sometimes be convenient to mount container directories on the host, for instance: for editing and debugging. Tutor provides different solutions to this problem.
Bind-mount from the "volumes/" directory
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Tutor makes it easy to create a bind-mount from an existing container. First, copy the contents of a container directory with the ``bindmount`` command. For instance, to copy the virtual environment of the "lms" container::
tutor dev bindmount lms /openedx/venv
This command recursively copies the contents of the ``/opendedx/venv`` directory to ``$(tutor config printroot)/volumes/venv``. The code of any Python dependency can then be edited -- for instance, you can then add a ``import ipdb; ipdb.set_trace()`` statement for step-by-step debugging, or implement a custom feature.
Then, bind-mount the directory back in the container with the ``--volume`` option::
tutor dev runserver --volume=/openedx/venv lms
Notice how the ``--volume=/openedx/venv`` option differs from `Docker syntax <https://docs.docker.com/storage/volumes/#choose-the--v-or---mount-flag>`__? Tutor recognizes this syntax and automatically converts this option to ``--volume=/path/to/tutor/root/volumes/venv:/openedx/venv``, which is recognized by Docker.
.. note::
The ``bindmount`` command and the ``--volume=/...`` option syntax are available both for the ``tutor local`` and ``tutor dev`` commands.
Manual bind-mount to any directory
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The above solution may not work for you if you already have an existing directory, outside of the "volumes/" directory, which you would like mounted in one of your containers. For instance, you may want to mount your copy of the `edx-platform <https://github.com/edx/edx-platform/>`__ repository. In such cases, you can simply use the ``-v/--volume`` `Docker option <https://docs.docker.com/storage/volumes/#choose-the--v-or---mount-flag>`__::
tutor dev run --volume=/path/to/edx-platform:/openedx/edx-platform lms bash
Override docker-compose volumes
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The above solutions require that you explicitly pass the ``-v/--volume`` to every ``run`` or ``runserver`` command, which may be inconvenient. Also, these solutions are not compatible with the ``start`` command. To address these issues, you can create a ``docker-compose.override.yml`` file that will specify custom volumes to be used with all ``dev`` commands::
vim "$(tutor config printroot)/env/dev/docker-compose.override.yml"
You are then free to bind-mount any directory to any container. For instance, to mount your own edx-platform fork::
version: "3.7"
services:
lms:
volumes:
- /path/to/edx-platform/:/openedx/edx-platform
cms:
volumes:
- /path/to/edx-platform/:/openedx/edx-platform
lms-worker:
volumes:
- /path/to/edx-platform/:/openedx/edx-platform
cms-worker:
volumes:
- /path/to/edx-platform/:/openedx/edx-platform
This override file will be loaded when running any ``tutor dev ..`` command. The edx-platform repo mounted at the specified path will be automatically mounted inside all LMS and CMS containers. With this file, you should no longer specify the ``-v/--volume`` option from the command line with the ``run`` or ``runserver`` commands.
.. note::
The ``tutor local`` commands loads the ``docker-compose.override.yml`` file from the ``$(tutor config printroot)/env/local/docker-compose.override.yml`` directory.
Common tasks
------------
Setting up a development environment for edx-platform
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Following the instructions :ref:`above <bind_mounts>` on how to bind-mount directories from the host above, you may mount your own `edx-platform <https://github.com/edx/edx-platform/>`__ fork in your containers by running either::
# Mount from the volumes/ directory
tutor dev bindmount lms /openedx/edx-platform
tutor dev runserver --volume=/openedx/edx-platform lms
# Mount from an arbitrary directory
tutor dev runserver --volume=/path/to/edx-platform:/openedx/edx-platform lms
# Add your own volumes to $(tutor config printroot)/env/dev/docker-compose.override.yml
tutor dev runserver lms
If you choose any but the first solution above, you will have to make sure that your fork works with Tutor.
First of all, you should make sure that you are working off the ``open-release/lilac.2`` tag. See the :ref:`fork edx-platform section <edx_platform_fork>` for more information.
Then, you should run the following commands::
# Run bash in the lms container
tutor dev run [--volume=...] lms bash
# Compile local python requirements
pip install --requirement requirements/edx/development.txt
# Install nodejs packages in node_modules/
npm install
# Rebuild static assets
openedx-assets build --env=dev
To debug a local edx-platform repository, add a ``import ipdb; ipdb.set_trace()`` breakpoint anywhere in your code and run::
tutor dev runserver [--volume=...] lms
XBlock and edx-platform plugin development
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In some cases you will have to develop features for packages that are pip-installed next to edx-platform. This is quite easy with Tutor. Just add your packages to the ``$(tutor config printroot)/env/build/openedx/requirements/private.txt`` file. To avoid re-building the openedx Docker image at every change, you should add your package in editable mode. For instance::
echo "-e ./mypackage" >> "$(tutor config printroot)/env/build/openedx/requirements/private.txt"
The ``requirements`` folder should have the following content::
env/build/openedx/requirements/
private.txt
mypackage/
setup.py
...
You will have to re-build the openedx Docker image once::
tutor images build openedx
You should then run the development server as usual, with ``runserver``. Every change made to the ``mypackage`` folder will be picked up and the development server will be automatically reloaded.
Loading custom edx-platform settings
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
By default, tutor settings files are mounted inside the docker images at ``/openedx/edx-platform/lms/envs/tutor/`` and ``/openedx/edx-platform/cms/envs/tutor/``. In the various ``dev`` commands, the default ``edx-platform`` settings module is set to ``tutor.development`` and you don't have to do anything to set up these settings.
If, for some reason, you want to use different settings, you will need to define the ``TUTOR_EDX_PLATFORM_SETTINGS`` environment variable.
For instance, let's assume you have created the ``/path/to/edx-platform/lms/envs/mysettings.py`` and ``/path/to/edx-platform/cms/envs/mysettings.py`` modules. These settings should be pretty similar to the following files::
$(tutor config printroot)/env/apps/openedx/tutor/lms/development.py
$(tutor config printroot)/env/apps/openedx/tutor/cms/development.py
Alternatively, the ``mysettings.py`` files can import the tutor development settings::
# Beginning of mysettings.py
from .tutor.development import *
You should then specify the settings to use on the host::
export TUTOR_EDX_PLATFORM_SETTINGS=mysettings
From then on, all ``dev`` commands will use the ``mysettings`` module. For instance::
tutor dev runserver lms
Running edx-platform unit tests
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
It's possible to run the full set of unit tests that ship with `edx-platform <https://github.com/edx/edx-platform/>`__. To do so, run a shell in the LMS development container::
tutor dev run lms bash
Then, run unit tests with ``pytest`` commands::
# Run tests on common apps
unset DJANGO_SETTINGS_MODULE
unset SERVICE_VARIANT
export EDXAPP_TEST_MONGO_HOST=mongodb
pytest common
pytest openedx
# Run tests on LMS
export DJANGO_SETTINGS_MODULE=lms.envs.tutor.test
pytest lms
# Run tests on CMS
export DJANGO_SETTINGS_MODULE=cms.envs.tutor.test
pytest cms
.. note::
Getting all edx-platform unit tests to pass on Tutor is currently a work-in-progress. Some unit tests are still failing. If you manage to fix some of these, please report your findings in the `Tutor forums <https://discuss.overhang.io>`__.
View Git Blame Copy Permalink