rotterdam

A standalone app that simplifies automation of your [Continuous Integration / Quality Assurance / End to End Testing] environments.

---

Rotterdam watches redis for changes to a set of docker container configurations. It then executes the running/ stopping/ updating of these containers seamlessly.

Rotterdam also has the ability to serve containers on virtualhosts with hipache.

This means that you can deploy isolated builds of your application to public urls easily from git / jenkins/ whatever, and then share them with your QA team.

Installation

Rotterdam is available via npm. Install it globally using the command:

npm install -g rotterdam

Rotterdam's only dependencies are node, redis and docker. You should be able to install redis using your distro's package manager. On ubuntu installing everything you need to run rotterdam is as easy as:

apt-get -y install software-properties-common python-software-properties
add-apt-repository -y ppa:rwky/redis
apt-get -y update
apt-get install -y redis-server git nodejs npm docker.io

(You may need to run these using sudo)

Redis conf

For watching for changes to redis keyspaces we require that redis has the following enabled:

redis-cli CONFIG SET notify-keyspace-events KEA

Running Rotterdam

We (mammal) prefer to run rotterdam via pm2. But you're free to use something like supervisor or an upstart script.

Once you've got pm2 installed you should be able to start rotterdam by running:

pm2 start rotterdam

Alternatively if you'd just like to test it out the rotterdam bin file should be in your $PATH so just calling rotterdam will have the same effect.

Configuration

All rotterdam config is done via env variables. The defaults are:

KEY_PREFIX='rotterdam:'
CONTAINER_KEY_PREFIX=$KEY_PREFIX'containers:'
CHANGES_QUEUE_KEY=$KEY_PREFIX'changesqueue:'
 
REDIS_HOST='127.0.0.1'
REDIS_PORT=6379
 
DOCKER_SOCKET='/var/run/docker.sock'
 

Container definitions

Containers to be run can be defined as follows:

{
  "name": "{{ the display name }}",
  "createOptions": {
    // Anything from http://docs.docker.com/reference/api/docker_remote_api_v1.15/#create-a-container
  },
  "startOptions": {
    // Anything from http://docs.docker.com/reference/api/docker_remote_api_v1.15/#start-a-container
  },
  "vhosts": {
    "{{ the port on localhost to map the vhost to }}": [
      "{{ the vhost minus http:// }}"
    ]
  }
}

Some example containers:

rotterdam:containers:redis

{
  "name": "redis",
  "createOptions": {
    "Image": "repo/redis",
    "PortBindings": {
      "6379/tcp": []
    },
    "Volumes": {
      "/var/lib/redis": {}
    }
  },
  "startOptions": {
    "Binds": [
      "/var/lib/redis:/var/lib/redis:rw"
    ],
    "Tty": false
  }
}

rotterdam:containers:webapp

{
  "name": "webapp",
  "createOptions": {
    "Env": [
      "PORT=3232"
    ],
    "Image": "repo/webapp:tag",
    "PortBindings": {
      "3232/tcp": [
        {
          "HostIp": "0.0.0.0",
          "HostPort": "49153"
        }
      ]
    }
  },
  "startOptions": {
    "Links": [
      "redis:redis"
    ]
  },
  "vhosts": {
    "3232": [
      "webapp.com"
    ]
  }
}

Practical Example

With Jenkins

We pass new build groups to rotterdam from our jenkins tasks that are triggered on our staging branches.

Pushes to our staging branch run a jenkins task that does the following:

1. Builds the latest image for the web app
2. Runs the unit tests against the local image
3. Pushes the passing image to our private docker repo
4. Pushes the container configs for:
   • The docker app (including a vhost unique to this build)
   • A CouchDB process
   • A Redis process
   • A Worker process
5. It then polls the unique vhost until it returns a 200 response
6. It then runs the E2E tests via Sauce Labs pointing at the public vhost.
7. It then reports the results back to the developer who pushed the last commit.

The build is then accessible for the developer to explore and run manual tests against.

Contributing

To help with contributing we've provided a Vagrantfile to automatically provision a box containing a shared volume of your local rotterdam clone.

First run vagrant up, then vagrant ssh into the box and run node /opt/rotterdam/index.js.

This will have same same effect as running the rotterdam bin file above.

Roadmap

1. Web interface
2. Make Hipache configurable
3. On startup calculate image dependecies and start the non-dependent ones first
4. Container Grouping