Docker coolness

Obviously Docker is wicked ahsome and I should get acquainted with it. I've also been intending to get better acquainted with Redis so I'm borrowing some stuff from the docs pertaining to that.

First things first. Do not just go "sudo apt-get install docker.io", or you'll end up with version 1.0.1 which is horribly buggy. Follow the correct instructions:

wget -qO- https://get.docker.com/ | sh

The install process will offer you an option to do some usermod stuff so you can run Docker as non-root. My experience thusfar has been that that doesn't work because it still stores the containers and images where only root can touch them, so I ignore that option.

Intriguing things about Docker

• I can chop up a project into microservices and implement them in separate Docker containers. When I want to combine them, I can use docker-compose.
• I can spin up lots of containers on my dev machine, and pretend I'm working with some big complicated VPC on AWS.
• Theoretically, I can tar up images I've built on my dev machine using docker save, scp the tar file up to my server-in-the-cloud, and unpack it there using docker load. I haven't tried this yet, but it sounds like it would make deployment pretty doggone easy.

Most companies I've worked at in the past few years have a collection of git repositories, and each server in their production site uses some combination of these repositories. It wouldn't be too hard to create a Docker image for each server, and be able to run the whole kittenkaboodle on my dev machine. It may be necessary to tinker with ssh permissions to clone all the git repositories, but that's a soluble problem, and being able to emulate the production system on my dev machine should be very valuable.

The docker community

There is a registry of existing Docker images that people have put together. You can find images for all sorts of popular web tools: Redis, Celery, Postgres, nginx, MongoDB, MySQL, Ubuntu, NodeJS, and with a bit of digging, more interesting things like Neo4J, Jena/Fuseki, GROMACS, POV-Ray and a Raspberry Pi toolchain. (A couple more RPi hacks: running Docker on the RPi and building a RPi cluster using docker.)

If I'm understanding this correctly, I could put together a shopping list of tools that I want to use, pull down the appropriate docker images, connect them with docker-compose, and just add my own application code.

A guy at work is excited about CoreOS, a collection of tools that build upon docker to handle all kinds of deployment concerns. He specifically said I should be looking at etcd and fleet. I find the CoreOS stuff a bit overwhelming so I think I'll work on that later.