Simple ML platform - current iteration just collects data from public APIs and displays/stores, for local use.
Project skeleton originally generated with (including this README) https://github.com/pydanny/cookiecutter-django
LICENSE: BSD
ml relies extensively on environment settings which will not work with Apache/mod_wsgi setups. It has been deployed successfully with both Gunicorn/Nginx and even uWSGI/Nginx.
For configuration purposes, the following table maps the 'ml' environment variables to their Django setting:
Environment Variable | Django Setting | Development Default | Production Default |
---|---|---|---|
DJANGO_CACHES | CACHES (default) | locmem | redis |
DJANGO_DATABASES | DATABASES (default) | See code | See code |
DJANGO_DEBUG | DEBUG | True | False |
DJANGO_SECRET_KEY | SECRET_KEY | CHANGEME!!! | raises error |
DJANGO_SECURE_BROWSER_XSS_FILTER | SECURE_BROWSER_XSS_FILTER | n/a | True |
DJANGO_SECURE_SSL_REDIRECT | SECURE_SSL_REDIRECT | n/a | True |
DJANGO_SECURE_CONTENT_TYPE_NOSNIFF | SECURE_CONTENT_TYPE_NOSNIFF | n/a | True |
DJANGO_SECURE_FRAME_DENY | SECURE_FRAME_DENY | n/a | True |
DJANGO_SECURE_HSTS_INCLUDE_SUBDOMAINS | HSTS_INCLUDE_SUBDOMAINS | n/a | True |
DJANGO_SESSION_COOKIE_HTTPONLY | SESSION_COOKIE_HTTPONLY | n/a | True |
DJANGO_SESSION_COOKIE_SECURE | SESSION_COOKIE_SECURE | n/a | False |
DJANGO_DEFAULT_FROM_EMAIL | DEFAULT_FROM_EMAIL | n/a | "ml <[email protected]>" |
DJANGO_SERVER_EMAIL | SERVER_EMAIL | n/a | "ml <[email protected]>" |
DJANGO_EMAIL_SUBJECT_PREFIX | EMAIL_SUBJECT_PREFIX | n/a | "[ml] " |
The following table lists settings and their defaults for third-party applications:
Environment Variable | Django Setting | Development Default | Production Default |
---|---|---|---|
DJANGO_AWS_ACCESS_KEY_ID | AWS_ACCESS_KEY_ID | n/a | raises error |
DJANGO_AWS_SECRET_ACCESS_KEY | AWS_SECRET_ACCESS_KEY | n/a | raises error |
DJANGO_AWS_STORAGE_BUCKET_NAME | AWS_STORAGE_BUCKET_NAME | n/a | raises error |
DJANGO_SENTRY_DSN | SENTRY_DSN | n/a | raises error |
DJANGO_SENTRY_CLIENT | SENTRY_CLIENT | n/a | raven.contrib.django.raven_compat.DjangoClient |
DJANGO_SENTRY_LOG_LEVEL | SENTRY_LOG_LEVEL | n/a | logging.INFO |
DJANGO_MAILGUN_API_KEY | MAILGUN_ACCESS_KEY | n/a | raises error |
DJANGO_MAILGUN_SERVER_NAME | MAILGUN_SERVER_NAME | n/a | raises error |
The steps below will get you up and running with a local development environment. We assume you have the following installed:
- pip
- virtualenv
- PostgreSQL
First make sure to create and activate a virtualenv, then open a terminal at the project root and install the requirements for local development:
$ pip install -r requirements/local.txt
Create a local PostgreSQL database:
$ createdb ml
Run migrate
on your new database:
$ python manage.py migrate
You can now run the runserver_plus
command:
$ python manage.py runserver_plus
Open up your browser to http://127.0.0.1:8000/ to see the site running locally.
To create a normal user account, just go to Sign Up and fill out the form. Once you submit it, you'll see a "Verify Your E-mail Address" page. Go to your console to see a simulated email verification message. Copy the link into your browser. Now the user's email should be verified and ready to go.
To create an superuser account, use this command:
$ python manage.py createsuperuser
For convenience, you can keep your normal user logged in on Chrome and your superuser logged in on Firefox (or similar), so that you can see how the site behaves for both kinds of users.
To run the tests, check your test coverage, and generate an HTML coverage report:
$ coverage run manage.py test $ coverage html $ open htmlcov/index.html
If you'd like to take advantage of live reloading and Sass / Compass CSS compilation you can do so with a little bit of prep work.
Make sure that nodejs is installed. Then in the project root run:
$ npm install
If you don't already have it, install compass (doesn't hurt if you run this command twice):
gem install compass
Now you just need:
$ grunt serve
The base app will now run as it would with the usual manage.py runserver
but with live reloading and Sass compilation enabled.
To get live reloading to work you'll probably need to install an appropriate browser extension
This app comes with Celery.
To run a celery worker:
cd ml
celery -A ml.taskapp worker -l info
Please note: For Celerys import magic to work, it is important where the celery commands are run. If you are in the same folder with manage.py, you should be right.
In development, it is often nice to be able to see emails that are being sent from your application. For this purpose, a Grunt task exists to start an instance of maildump which is a local SMTP server with an online interface.
Make sure you have nodejs installed, and then type the following:
$ grunt start-email-server
This will start an email server. The project is setup to deliver to the email server by default. To view messages that are sent by your application, open your browser to http://127.0.0.1:1080
To stop the email server:
$ grunt stop-email-server
The email server listens on 127.0.0.1:1025
Sentry is an error logging aggregator service. You can sign up for a free account at http://getsentry.com or download and host it yourself. The system is setup with reasonable defaults, including 404 logging and integration with the WSGI application.
You must set the DSN url in production.
It's time to write the code!!!
N.B. The integration tests will not run on Windows.
To install the test runner:
$ pip install hitch
To run the tests, enter the ml/tests directory and run the following commands:
$ hitch init
Then run the stub test:
$ hitch test stub.test
This will download and compile python, postgres and redis and install all python requirements so the first time it runs it may take a while.
Subsequent test runs will be much quicker.
The testing framework runs Django, Celery (if enabled), Postgres, HitchSMTP (a mock SMTP server), Firefox/Selenium and Redis.
It is possible to deploy to Heroku, to your own server by using Dokku, an open source Heroku clone or using docker-compose.
Run these commands to deploy the project to Heroku:
heroku create --buildpack https://github.com/heroku/heroku-buildpack-python
heroku addons:create heroku-postgresql:hobby-dev
heroku pg:backups schedule --at '02:00 America/Los_Angeles' DATABASE_URL
heroku pg:promote DATABASE_URL
heroku addons:create heroku-redis:hobby-dev
heroku addons:create mailgun
heroku config:set DJANGO_SECRET_KEY=`openssl rand -base64 32`
heroku config:set DJANGO_SETTINGS_MODULE='config.settings.production'
heroku config:set DJANGO_AWS_ACCESS_KEY_ID=YOUR_AWS_ID_HERE
heroku config:set DJANGO_AWS_SECRET_ACCESS_KEY=YOUR_AWS_SECRET_ACCESS_KEY_HERE
heroku config:set DJANGO_AWS_STORAGE_BUCKET_NAME=YOUR_AWS_S3_BUCKET_NAME_HERE
heroku config:set DJANGO_MAILGUN_SERVER_NAME=YOUR_MALGUN_SERVER
heroku config:set DJANGO_MAILGUN_API_KEY=YOUR_MAILGUN_API_KEY
heroku config:set PYTHONHASHSEED=random
git push heroku master
heroku run python manage.py migrate
heroku run python manage.py check --deploy
heroku run python manage.py createsuperuser
heroku open
You need to make sure you have a server running Dokku with at least 1GB of RAM. Backing services are added just like in Heroku however you must ensure you have the relevant Dokku plugins installed.
cd /var/lib/dokku/plugins
git clone https://github.com/rlaneve/dokku-link.git link
git clone https://github.com/luxifer/dokku-redis-plugin redis
git clone https://github.com/jezdez/dokku-postgres-plugin postgres
dokku plugins-install
You can specify the buildpack you wish to use by creating a file name .env containing the following.
export BUILDPACK_URL=<repository>
You can then deploy by running the following commands.
git remote add dokku [email protected]:ml
git push dokku master
ssh -t [email protected] dokku redis:create ml-redis
ssh -t [email protected] dokku redis:link ml-redis ml
ssh -t [email protected] dokku postgres:create ml-postgres
ssh -t [email protected] dokku postgres:link ml-postgres ml
ssh -t [email protected] dokku config:set ml DJANGO_SECRET_KEY=RANDOM_SECRET_KEY_HERE
ssh -t [email protected] dokku config:set ml DJANGO_SETTINGS_MODULE='config.settings.production'
ssh -t [email protected] dokku config:set ml DJANGO_AWS_ACCESS_KEY_ID=YOUR_AWS_ID_HERE
ssh -t [email protected] dokku config:set ml DJANGO_AWS_SECRET_ACCESS_KEY=YOUR_AWS_SECRET_ACCESS_KEY_HERE
ssh -t [email protected] dokku config:set ml DJANGO_AWS_STORAGE_BUCKET_NAME=YOUR_AWS_S3_BUCKET_NAME_HERE
ssh -t [email protected] dokku config:set ml DJANGO_MAILGUN_API_KEY=YOUR_MAILGUN_API_KEY
ssh -t [email protected] dokku config:set ml DJANGO_MAILGUN_SERVER_NAME=YOUR_MAILGUN_SERVER
ssh -t [email protected] dokku run ml python manage.py migrate
ssh -t [email protected] dokku run ml python manage.py createsuperuser
When deploying via Dokku make sure you backup your database in some fashion as it is NOT done automatically.
Warning
Docker is evolving extremely fast, but it has still some rough edges here and there. Compose is currently (as of version 1.4) not considered production ready. That means you won't be able to scale to multiple servers and you won't be able to run zero downtime deployments out of the box. Consider all this as experimental until you understand all the implications to run docker (with compose) on production.
Run your app with docker-compose
Prerequisites:
- docker (tested with 1.8)
- docker-compose (tested with 0.4)
Before you start, check out the docker-compose.yml file in the root of this project. This is where each component of this application gets its configuration from. It consists of a postgres service that runs the database, redis for caching, nginx as reverse proxy and last but not least the django application run by gunicorn. Since this application also runs Celery, there are two more services with a service called celeryworker that runs the celery worker process and celerybeat that runs the celery beat process.
All of these services except redis rely on environment variables set by you. There is an env.example file in the root directory of this project as a starting point. Add your own variables to the file and rename it to .env. This file won't be tracked by git by default so you'll have to make sure to use some other mechanism to copy your secret if you are relying solely on git.
By default, the application is configured to listen on all interfaces on port 80. If you want to change that, open the docker-compose.yml file and replace 0.0.0.0 with your own ip. If you are using nginx-proxy to run multiple application stacks on one host, remove the port setting entirely and add VIRTUAL_HOST=example.com to your env file. This pass all incoming requests on nginx-proxy to the nginx service your application is using.
Postgres is saving its database files to /data/ml/postgres by default. Change that if you wan't something else and make sure to make backups since this is not done automatically.
To get started, pull your code from source control (don't forget the .env file) and change to your projects root directory.
You'll need to build the stack first. To do that, run:
docker-compose build
Once this is ready, you can run it with:
docker-compose up
To run a migration, open up a second terminal and run:
docker-compose run django python manage.py migrate
To create a superuser, run:
docker-compose run django python manage.py createsuperuser
If you need a shell, run:
docker-compose run django python manage.py shell_plus
Once you are ready with your initial setup, you wan't to make sure that your application is run by a process manager to survive reboots and auto restarts in case of an error. You can use the process manager you are most familiar with. All it needs to do is to run docker-compose up in your projects root directory.
If you are using supervisor, you can use this file as a starting point:
[program:ml] command=docker-compose up directory=/path/to/ml redirect_stderr=true autostart=true autorestart=true priority=10
Place it in /etc/supervisor/conf.d/ml.conf and run:
supervisorctl reread supervisorctl start ml
To get the status, run:
supervisorctl status
If you have errors, you can always check your stack with docker-compose. Switch to your projects root directory and run:
docker-compose ps
to get an output of all running containers.
To check your logs, run:
docker-compose logs
If you want to scale your application, run:
docker-compose scale django=4 docker-compose scale celeryworker=2
Don't run the scale command on postgres or celerybeat