Building reproducible dev environments with Docker @grahamgilchrist 

Scenario 1 - The old project 

Scenario 2 - The new member of staff Image credit: Paul Inkles https://www.flickr.com/photos/dumfstar/8553473662 

Scenario 3 - The missing documentation 

Scenario 4 - Different hardware, different OS 

The problem ● Large number of projects and variations on environments ● Different hardware, operating systems and software versions trying to run same project ● Environment setup in installation guide is forgotten or becomes out of date ● All these things take up your time to maintain, often fixing the same issue on multiple machines 

How do we keep the environments consistent and automate this boring job? Image credit: Steve Jurvetson https://www.flickr.com/photos/jurvetson/9690512888/ 

We want a solution that: 1. Stores environment config versioned with project code 2. Makes everyone use the same environment 3. Is reproducible from scratch 4. Automate the build process 

+ docker compose 

Advantages of Docker over VM ● Run speed ● Build speed (image and build step caching) ● Disk space - economical re-use of base system images etc. ● Re-use of common services between projects ● Large resource of pre-built community images setup via the docker hub image registry 

Typical web project > docker-compose up project docker-compose.yml 

python docker-compose.yml python: build: python/ command: manage.py runserver environment: DATABASE_URL: 'postgres://dbname@db/postgres' volumes: - ./python:/usr/src/app links: - db ports: - "8000:8000" db: image: postgres:9.3 ports: - "5432:5432" 

python Dockerfile FROM python:3.4.3 # We add libmemcached from debian repositories RUN apt-get update && \ apt-get install -y \ libmemcached-dev \ && rm -rf /var/lib/apt/lists/* RUN mkdir -p /usr/src/app WORKDIR /usr/src/app COPY requirements.txt /usr/src/app/ RUN pip install -r requirements.txt 

NodeJS docker-compose.yml grunt: build: node/ volumes: - ./web:/usr/src/app command: grunt bower: image: blinkmobile/bower:1.3.12 environment: bower_allow_root: true bower_analytics: false working_dir: /data volumes: - ./web:/data 

How is it working out for us? Much higher bus factor Saving time Image credits: Tom Page https://www.flickr.com/photos/tompagenet/6935847473/ Steve Collis https://www.flickr.com/photos/swampa/8105520673/ 

Outstanding issues ● Mac OS - not native ○ slow for disk access via volume mounts ● SSH issues ○ rsync, ssh ○ installing packages via github ssh auth ● Learning curve for 'the docker way' ○ Separation concepts are hard and totally new to most devs. ○ Hard to get people to configure it themselves 

Future improvements Identical dev / Deployment setup Issues on Mac OS Image credit: Steve Jurvetson https://www.flickr.com/photos/jurvetson/3327872958 

Thanks! Final thoughts @grahamgilchrist