Deploying your Will
Will's happy to run on your machine all day, every day. But he really shines out in production where he can run independently of anything you do with your local machine. Here are some best practices for deploying will.
Deploy on Heroku
Heroku is our recommended platform for deployment because it's simple, easy, and free. That's a tough combination to beat. You're in no way locked in to running will on heroku, though - run him wherever you want!
Step 1: Set up your heroku app, and a redis addon.
Assuming you have the heroku toolbelt installed and all set up, it's as easy as this:
heroku create our-will-name heroku addons:add rediscloud
A note on rediscloud: you can also use redistogo, openredis, or anyone else. We use rediscloud and like them.
Step 2: Add all the needed environment variables:
You'll want to take all the variables that live in your virtualenv's
postactivate file that feed into your will environment, and provide them to your heroku app.
At minimum, that's
heroku config:set \ WILL_PUBLIC_URL="http://our-will-name.herokuapp.com" \ # Slack WILL_SLACK_API_TOKEN="lkasjflkaklfjlasfjal1249814" # Hipchat WILL_HIPCHAT_USERNAMEfirstname.lastname@example.org' \ WILL_HIPCHAT_PASSWORD='asj2498q89dsf89a8df' \ WILL_HIPCHAT_V2_TOKEN='asdfjl234jklajfa3azfasj3afa3jlkjiau' \ # Rocket.Chat WILL_ROCKETCHAT_USERNAMEemail@example.com'\ WILL_ROCKETCHAT_PASSWORD='12o312938asfjilasdlfjasdlkfj'\ WILL_ROCKETCHAT_URL='https://heywill.rocket.chat'\
Finally, for will's schedule to be correct, you need to set him to the time zone you want:
heroku config:set TZ="America/Los_Angeles"
Note: You don't have to worry about setting the
HTTPSERVER_PORT. Will auto-detects those and takes care of it.
Step 3: Deploy!
You're all set to deploy your will. Ready? Go!
git push heroku
You app is now up on heroku. Finally, turn it on!
heroku scale web=1
That should be it - head over to http://our-will-name.herokuapp.com, and you should see will's smiling face.
If so, pop into chat, and give a:
Deploying future updates
Just commit your updates, then
git push heroku
Simple. For best-practices, see our continuous deployment recommendations below.
Deploy in Docker
Will is packaged with a Dockerfile and docker-compose files to allow deploying in a container with redis.
You should have docker already installed; additionally, the instructions require docker-compose. If you're not using docker-compose, you can still also the pre-defined images from Docker Hub.
Step 1: Configure your container's environment variables
In Will's docker directory, update the default.env file with your environment's settings. At a minimum, this should contain:
# For Slack WILL_SLACK_API_TOKEN # For Hipchat WILL_HIPCHAT_USERNAME WILL_HIPCHAT_PASSWORD WILL_HIPCHAT_V2_TOKEN # Rocket.Chat WILL_ROCKETCHAT_USERNAME WILL_ROCKETCHAT_PASSWORD WILL_ROCKETCHAT_URL
Note, we've pre-defined the redis url and the HTTP Server port; if you update these values, make sure you update the docker-compose file accordingly.
Step 2: Deploy your plugins and templates
If you have any custom templates, create directories for your plugins and templates, and load your plugins and templates they'll be mounted when the container starts up giving your containerized bot access to your templates.
mkdir ./plugins ./templates
Step 3: Build will.
Run build your docker image locally.
Step 4: Start the container
Start your image from docker-compose using:
or to run the container in the background,
docker-compose up -d
Roll your Own Container
Will docker file(s) are part of the main repository; this lets developers/teams build on the containers to embed configuration files, custom plugins etc... For now Will containers are represented in python 2.7 and python 3 flavors.
Will is built on the alpine distribution(s) of python docker images. This was done to provide the smallest container footprint; yet, nothing is perfect so a few things need to be added. Will-base provides the foundation to perform these base modifications.
The heywill/will image is the container with Will executing as the container process. A build args are available for branch based builds from a repository. To build from a specific repo and branch use,
--build-arg repo=<repo_name> branch=<branch_name>
otherwise skoczen/will is the default repo, while master is the default branch.
Deploy Everywhere Else
Will is Just Python
That says it all. Will is just python, and you can deploy him anywhere you have python, an open port, and access to redis.
In your chosen deploy environment and setup, you'll want to do a couple things:
Replicate your environment settings
At minimum, that's:
export WILL_PUBLIC_URL="http://our-will-name.herokuapp.com" export WILL_REDIS_URL='redis://some-domain.com/7/' export WILL_HTTPSERVER_PORT='80' # Slack WILL_SLACK_API_TOKEN="lkasjflkaklfjlasfjal1249814" # Hipchat export WILL_HIPCHAT_USERNAMEfirstname.lastname@example.org' export WILL_HIPCHAT_PASSWORD='asj2498q89dsf89a8df' export WILL_HIPCHAT_V2_TOKEN='asdfjl234jklajfa3azfasj3afa3jlkjiau' # Rocket.Chat WILL_ROCKETCHAT_USERNAMEemail@example.com' WILL_ROCKETCHAT_PASSWORD='12o312938asfjilasdlfjasdlkfj' WILL_ROCKETCHAT_URL='https://heywill.rocket.chat'
If you have more than 30 chat rooms, you must also set the V1 token to avoid hipchat rate limits:
You'll also need to set any environment variables for your plugins.
Call run_will with something that handles restarts and crashes.
The command to kick off will is just:
You can run that with whatever supervisory process you'd like. Will's very, very stable, but having something that handles crashes is still a pretty good idea.
Ah, you say, but I have this crazy python-twisted-zope-node monolith that self-repairs mars and can do everything. Why do I need the shell at all?
Answer, crazy monolith mars person: you don't. Will is Just Python.
from will.main import WillBot bot = WillBot() bot.bootstrap()
and you're good.
Will's default storage backend is Redis, but he supports some others if you can't run Redis.
To change the backend, just set
config.py and then supply any other needed settings for the new storage backend. The currently supported backends are:
redis- The default Redis backend
couchbase- A Couchbase backend
file- Keeps the settings as files on a local filesystem
Couchbase requries you set
COUCHBASE_URL in your config.
You are also required to have the python Couchbase client (and thus, libcouchbase) installed. If you are installing for development you can use
pip install -r requirements.couchbase.txt to pull in the Couchbase client. See the Python Couchbase client repo for more info.
File requires you set
FILE_DIR in your config to point to an empty directory.
Will's default pubsub backend is Redis, and support for ZeroMQ and a pure-python backend is on the way.
To change the backend, just set
config.py and then supply any other needed settings for the new backend. The currently supported backend is:
redis- The default Redis backend
In this section, we describe how we deploy and host will, in the hopes that others come forward and share what's working for them, too. The more good practices, the better.
We host on heroku, using rediscloud for redis, mailgun for email, and things have been peachy. We deploy via CD using CircleCI. Details on that below.
Use Continuous Deployment
Our stack is set up so that any pushes on will's master branch have tests run on CircleCI, and if they pass, a new version is deployed to heroku immediately. This has been delightful. Even though will has very, very minimal tests, we generally catch if things are horribly broken, and it's meant that adding new functionality to will takes minutes, sometimes seconds of developer time.
Continuous Deployment has dramatically changed how we build and use will - instead of talking about "what if will did...", generally, people just implement it, push it, and play with it for real. It's been a great place to be. It might be for you too.
That's it in getting your will up and running! But maybe you're one of those people who wants to pitch in and make will even better. Awesome. Learn how to improve will.