DevOpsDays London March 2013
Right, time to do the “quick brain dump” after DevOpsDays London, before I forget too much of it.
I think my main takeaway, and this is something I started realising a while ago, is that in contrast to my initial opinions when I first heard the term DevOps, is that it is, ultimately, based around culture. I was so far off the mark when I thought it was mainly about installing a selection of puppet, chef, cfengine, statsd, graphite, logstash, travis, jenkins, teamcity, etcetera, etcetera, and then things would fall into place that it’s funny. To mangle a quote from the weekend, use tools, but don’t be a tool.
Obviously, a main driver behind getting the right culture going, is communication, and I was encouraged to see that IRC got a mention in almost all sessions/talks. Old school communication for the win. I do wonder, though, if it is the best medium for company wide communication. IRC is possibly a wee bit techie for certain audiences I think. Although I missed the BizDevOps open space, I have a feeling that the next step for the DevOps style movement is going to be to try and extend the cultural shift to whole company communications and openness. Not that we’re entirely there now with ops and dev, but if we can start bringing business operations onboard as well, and getting in on the conversation, using the various tools, then we can start making the world a happier, brighter place 😉
There were things to take away from all of the talks, but the two real killer talks for me were Gene Kim’s talk on “How Can We Better Sell DevOps?” (slides at slideshare), which had the real takeaway point of making people realise that “This isn’t an ops/IT problem, this is a business problem”, and the first ever public presentation (not sure I believe that, too polished 😉 ) from Sam Eaton, entitled “DevOps in the hell of a thousand different platforms” (slides and notes at speakerdeck) – superb talk, and definitely one I will be watching again when (if?) the videos appear online.
Apart from the “has been on my to-do list forever” item of getting logstash up and running (is there anything more frustrating that waiting for a project to kick off?), I’ve added ‘epic loadbalancer testing’ to my must-do list, after seeing Daniel Pope sneak a lightning talk into the ignite section of the day. Daniel has promised to write up a blog post. Essentially, he used honeyd and farpd to mock up the services behind the loadbalancer, allowing things like haproxy and nginx configs to be tested without them needing to connect to the back end. Clever stuff, and right at the top of my list to get set up and start using.