Getting into Containers

Sat, Nov 14, 2015

For a while now I’ve been experimenting with Docker which provides a useful level of abstraction over Containers. I’m still finding my feet in this exciting new world of cheap sort-of-virtualisation but so far I’m really enjoying the isolation they provide, allowing for explicit bundling of dependencies and repeatable builds. You can probably see how interested I’ve been from my github history these past months!

One thing I have found is that I don’t really trust other people’s builds all that much and have been preferring to roll my own in most cases, which has the benefit of forcing me to learn the tool but obviously contributes to the existing problem that we already have lots of images purportedly doing the same thing - just have a look at the number of images for Caddy to get idea of this problem.

The problem is two-fold: the first half is that builds are not necessarily trustworthy because they aren’t signed - this is currently being addressed by the Docker team. The second half is more subtle though; with the tech being so new people are all doing their own thing in their own way; lots of images require volumes mounting with configuration being provided that way, while others prefer to take it from a central configuration provided by Consul or etcd. Personally I’m erring towards stdin and command-line arguments where possible, largely because I don’t like the volume solution and don’t have any experience with the enterprise-level key-value stores.

Watching the trends and best-practises should prove very interesting indeed.