How to run HuBot in Docker on AWS EC2 Container Services

It all began when I was too lazy to check the website of our local canteen to find out todays lunch menu. Ok, it’s only a couple of clicks and not actual “work” but hey…

I have already played around with HuBot in the past and I’m also very excited about Docker so the idea raised to combine these things. And because we are in the cloud now it has to be hosted on AWS as well. Meanwhile AWS published the new EC2 Container Service (ECS) in Ireland so I had everything I needed. Let me share my experiences (and some code) with you.

Big Picture

The idea is to have all relevant sources in one GitHub repository. With each commit a build should be triggered and when the tests are green it should automatically deployed to AWS EC2 Container Service (ECS).

GitHub -> CircleCI -> Docker Hub -> AWS ECS

Sources are available on my GitHub repository.

Run HuBot in Docker container

Installing HuBot is quite easy. You can find a good documentation here. With Docker it’s also very easy. First of all we need to define all requirements:

FROM ubuntu

RUN apt-get update
RUN apt-get -y install expect redis-server nodejs npm
RUN ln -s /usr/bin/nodejs /usr/bin/node

RUN npm install -g coffee-script
RUN npm install -g yo generator-hubot

Create your own user

Normally you would run “yo hubot” to create a new instance but that leads to the exception Error: EACCES, permission denied ‘/root/.config/configstore/insight-yo.yml’

The reason is that yo doesn’t run as root user. So we have to create our own user and switch to it.

# Create hubot user
RUN	useradd -d /hubot -m -s /bin/bash -U hubot

# Log in as hubot user and change directory
USER	hubot
WORKDIR /hubot

# Install hubot
RUN yo hubot --owner="You" --name="HuBot" --description="HuBot on Docker" --defaults

Adding custom scripts

The next step is to load some scripts. HuBot supports three different ways to install custom scripts.

Built-in scripts: HuBot is shipped with some built-in scripts which are disabled by default. All you need is a hubot-scripts.json file where you define all the scripts which should be enabled. This file has to be added to your container
```
 ADD hubot-scripts.json /hubot/
```
Community scripts: Scripts from the community can be installed by npm. In addition these scripts have also be enabled in a external-scripts.json file which needs also be added to the container.
```
 RUN npm install hubot-standup-alarm --save && npm install
 ADD external-scripts.json /hubot/
```
Custom scripts: That’s the easiest and fastest way to add your self written scripts without publishing them to npm. Just add the (coffee) script file to your container.
```
 ADD hubot-leitwerk.coffee /hubot/scripts/
```

Integrate with Slack

We use Slack as chat tool. Luckily there is an adapter and the installation is also easy. At the end of our Dockerfile when we start HuBot we have to define the desired adapter. RUN npm install hubot-slack –save && npm install CMD bin/hubot -a slack

Let it run

To run HuBot first build the container and start it. In addition you have to give the slack adapter a security token. As I don’t want to share that token with all of you it’s not part of the Dockerfile but I pass it as environment variable when I start the container instance.

# Build the container
docker build -t hubot .

# And run it
docker run -e HUBOT_SLACK_TOKEN=xxx -d tatsu

To be continued

Running HuBot inside a Docker container was just the beginning for me. It’s amazing how fast I could start and run it w/out any experience with Docker or nodejs.

The next blog post will show how I deploy the container to AWS.