Where I have the choice, rather than running applications natively, I've decided to package them in containers and run them in isolation. There are a few advantages to this approach:
- They can easily be reconstructed and then run on any operating system with support for Docker
- Easy to maintain different application versions
- The isolation means you never get conflicting dependencies
- Docker is extremely fast to start
Drawbacks:
- Containers use a lot more disk space than installing the applications natively (this can be mitigated to an extent with minimal operating systems like Alpine)
- Dockerfile maintenance can be clunky (especially since Docker doesn't have a multiple inheritance system)
- Docker itself introduces some complexity where there otherwise wouldn't be
To simplify Dockerfile development, these containers are intended to be run in interactive mode. For that reason, I like to maintain the shell prompt from my host machine, so my system is to use base containers that install ZSH and UTF-8 locales. When running the container, I can mount in my zshrc and the oh-my-zsh directory. Due to a lack of a multiple inheritance system in Docker, there's a lot of duplication in the Dockerfiles, but I've decided to accept this limitation. The reality is the container definitions don't get updated that much.
A makefile is available for building the images in the suite.
Each image can be built with an individual target:
make ansibleThe entire suite can be built:
make allUsing clean targets, the same follows for removing images. There's a clean-base-images target that will also remove the base operating system images.
The following table lists the images and their versions:
| Name | Latest Version | Base |
|---|---|---|
| Ansible | 2.1.1.0 | ZSH Debian Jessie |
| AWS CLI | 1.11.57 | ZSH Python 2.7 |
| Azure CLI | 0.10.6 | Microsoft Azure CLI |
| Openshift Origin Client Tools | 1.2.1 | ZSH CentOS |
| Terragrunt | 0.11.0 | ZSH Debian Jessie |