Devops Weekly A weekly slice of devops news
Issue 33 - 7th August
In a quick break from the news and releases, I have to give special mention this week to John Willis for his latest addition to the music of devops: http://www.youtube.com/watch?v=pebIr4F-vjQ. Right, on with the regular programme.
Many of the abstracts for PuppetConf have been released. I’m particularly looking forward to the Noah talk, and some of the notes on Luke Kanies' keynote sound interesting; a move into application management and how puppet can be smaller, faster, and easier to integrate with.
My experience with Chef has been that you really only get the benefits if you understand all the different moving parts and how they tie together. I came across a useful representation of this architecture on the much improved Chef wiki this week, specifically everything mapped out using UML.
I’m guessing most of the readers of this newsletter spend a lot of their time on the command line. So I’m always happy when I find people talking about user interface issues and design guidelines for command line applications. Next time you’re building a little application try and remember some of the guidelines on this post.
It’s always interesting to see what technologies and movements other people find interesting, and the Thoughworks technology radar is a nice example. Infrastructure as code, Continuous deployment and Devops ideas feature prominently alongside lots of other things I’m fond of that I want to see more people using.
A good argument that reaching some areas of the wider systems administration ecosystem with devops ideas might be much harder that you might think. The reluctance for some to write code in any form is pointed out, as is the scarcity of people who make good systems programmers.
The idea of Polyglot programmers has been popular for a while, with more new languages seemingly becoming popular overnight. This post from Heroku talks about what that means for your typical deployment and operating infrastructure. The table comparing different languages and the duplication of tools is interesting, as is the call for more language-neutral interfaces.
On the face of it EC2 reserved instances look like a great way of reducing the cost of running in EC2. This post however argues that a hidden cost of reserved instances is the lack of flexibility. If you’re just using EC2 to bring up a relatively stable number and type of instance then reserved instances can work but you’re not getting the full benefit of the cloud environment. If your cluster is much more dynamic then reserved instances might not save you money.
An hour long video from Damon Edwards talking devops and system architecture with Ernest Mueller and Peco Karayanev of National Instruments. It should be particularly interesting to anyone working inside a large organisation as lots of the conversation is about how to make positive changes and get everyone involved in delivery and operations.
I’m really interested in the fight between continuous integration and feature branches. It feels to me that we need more discussion from both sides on this issue, partly because I see the same thing pointed out in the Thoughtworks technology radar - that as developers we’re branching and merging more because the latest tools make it easy to do so. Martin Fowlers piece on one of the problems with a branch heavy approach makes interesting reading.
Carrenza need some help with the Sport Relief 2012 platform build, and are looking for contract or permanent Chef lovers. email@example.com if you’re feeling charitable.
(This section is sponsored, if you’re interested in featuring a job get in touch at firstname.lastname@example.org)
The idea of writing tests for configuration management tools is talked about a lot, and chef spec looks an interesting approach if you’re using Chef. For a new project the documentation is also great.
Porkshop describes itself as a simple network-based system information server. Install it on your servers and it exposes information about the host over a simple HTTP interface. Plugins look simple enough to write too.
Bunraku is a simple status panel for puppet, hooking into the puppet reporting facility and storing information in Redis. It’s also looks like it would make a good starting point if you wanted to build you’re own report processing tool.