ACCU
Continuous
Delivery – Compared by
Continuous Delivery is one of those techniques that I know what it is in theory but aren’t sure what everyone actually does in practice. One obvious way to find out more about it would be to buy “the” book on the topic – Continuous Delivery. Instead I chose a different route – to listen to various people describe how they’re doing it.
The venue was the Mozilla offices near
Rather than a single speaker or set of lightning talks the format was 3 speakers each giving closer to a 15 minute presentation on how they are executing Continuous Delivery in their organisation. With a large audience clearly wanting to extract every last ounce of information from the presenters there was a barrage of questions after each segment that they deftly handled.
One of my goals was to find out what the size of a “feature” is so that when I see the stream of tweets from companies boasting at the number of features they’ve pushed into production today I have some frame of reference. Luckily I didn’t have to wait too long as first up was Chris O’Dell describing how the process works at 7Digital. Her team seems to have adopted many of the modern practices, such as pair programming and Kanban to create a very fluid environment. This wasn’t plain sailing and they clearly still have some baggage to address for some time yet. The most useful part of the presentation was the slides that walked through a “day in the life” of a change, from development right through to deployment. Chris left you with a feeling that they walk-the-walk as well as talking-the-talk.
Next up was Ed himself with Cosmin Onea to explain the 1E approach. This is a company that seems to have to deal with a little more “paperwork” for the final hurdle which I can definitely empathise with. They had a video similar to Chris’s earlier slides that presented the change workflow, but this time in a style more reminiscent of a production-line. Naturally there were many parallels to the first talk, with automated build, testing and deployment the common thread. A desire to avoid feature branches and focus on small well defined tasks was also apparent.
The final speaker was Robert Chatley who instead chose to talk about how he has tackled cloud based deployments. I suspect a fair portion of the audience is working on systems that are, or intend to, run in the cloud and so this was probably highly relevant for them. The crux of the presentation was about how to keep that purity of deploying exactly what you’ve tested, which in his case extended right down into the OS image on which the tests were run. He also provided some tips on how you might go about switching over from one version of the software to the next which requires a fair bit of thought when you have a considerable number of nodes to upgrade and whilst maintaining a minimum level of capacity.
Nice though the offices of Mozilla are with its alluring fridges, it was time to switch to a more traditional watering hole and continue the discussion with proper ale. Not unsurprisingly I discovered there is quite a bit of latitude in the definition of Continuous Delivery and that for some this only extends from DEV through to UAT. Sadly there is often some heavy-handed “review” process in place to bring the pipeline to an eventual standstill. Even so, I came away positive knowing that my team and I are heading in the right direction and that there is still a wealth of tooling to explore to help extract every last ounce out of the process.
Chris Oldwood
30 October 2012
Bio
Chris started out as a bedroom coder in the
80s, writing assembler on 8-bit micros. These days it’s C++ and C# on Windows
in big plush corporate offices. He is also the commentator for the
Godmanchester Gala Day Duck Race and can be contacted via