2 minute read

Lots of pretty in-depth discussions about configuration management, or at least configuration management tooling, this week. Lots of differing opinions, solid theory, innovation in low level systems and more to mull over.


Devops Weekly is sponsored by Brightbox Cloud - launch cloud servers in multiple UK datacentres in seconds…


A fantastic two part blog post about bringing continuous deployment and configuration management together. Lots of theory, discussion of how containers and microservices fit in and points about operators and developers prioritising seemingly different things.

Building software systems out of many small services is becoming increasingly common. This post delves into what a microservices architecture is, why it’s an important pattern and also what it might look like.

Configuration Management isn’t stupid, but it should be according to this next post. Some tools introduce (rather than address) complexity, determinism and reproducibility in systems and this post argues that better ones are needed to fix the problem.

And another post riffing on why current configuration management tools aren’t perfect, focusing on the Nix package management tool as a potential solution.

Introducing devops practices into an organisation needs some understanding of how to bring about change. This next slide deck is a nice introduction to being a change agent, with lots of tips about what to expect and how to get people on-board.

A description of the human moments experienced during a complex systems failure, and a case made for these being consistent between typical web operations work and other domains. The better we understand error, hopefully the better we’ll be able to deal with it.

Vagrant has a new release out with lots of new features. As well as Hyper-V support and rsync for shared folders, Vagrant now has an inbuilt box hosting service called Vagrant Cloud.

A detailed look at why it’s difficult to get at memory usage details from within an LXC container, along with how this might be resolved in the future.

A nice example of analysing nginx logs with pandas and matplotlib. The ipython notebook formats nicely shows off the results as it goes along, including nicely formatted graphs of the results.


A postmortem analysis, conducted in an open, blameless way, is the best way to learn from outages (and other failures and near-misses). April 15th at #ChefConf will feature a full-day workshop with Dave Zwieback, covering the theory and practice of conducting and participating in awesome postmortems.


Gunnery is described as a remote execution frontend. It’s a web based interface for executing commands on groups of servers, and features nice integration with ant, phing, fabric, capistrano, make and puppet out of the box.