I traveled to Raleigh, North Carolina recently to participate in their 2017 DevOpsDays event, which was two days of thought-provoking presentations, Open Spaces style discussions, and lots and lots of great hallway conversations. The event was very well organized; a big thanks to the conference committee, sponsors, attendees, and speakers!
I presented a talk, too: What Should I Instrument, and How Should I Do It? I'm embedding the slides below. Here's the description of the talk:
I’ve been there, you’ve been there: you're building a service and you want it to be easy to observe in production. You think, “I can just sprinkle some StatsD on it,” but how, exactly? What things are worth emitting to StatsD? What types of metrics should they be? Or perhaps you're considering logging, or tracing. You know that it's important to log things well, but ... you get the point. Good instrumentation sounds like something we all should do, but we have so few truly good examples of instrumentation that delivers signal and avoids noise. In fact, many of the most familiar examples are anti-patterns. And what about the culture? What works at Etsy doesn't always work elsewhere, unless the culture is there too. In this talk I'll share what I’ve learned about how to add good instrumentation and avoid bad stuff, and the boundary between instrumentation and culture. This is a sequel to my popular "what should I monitor, and how should I do it?" talk.
I think the conference was recording the main stage presentations, and if they post those publicly we'll share them here and on @VividCortex. Cheers!