- Instrumenting distributed systems for operational visibility: An article from the aws builders library (brilliant resource!) talks about how they think about logging for a service and its dependencies. 1 log line per event (eg client request) that gets passed around in a process collecting info about the task (input parameters, timings, context, cache hits / misses, etc) and is emitted at the end of the request. Correlation ids if you’re passing between microservices (bleh! :)). Metrics are mines from logs at as a separate task from collection. Log collection is of prime importance.
- Retrospectives anti-patterns: Gives us 3 things to look out for: Wheel of fortune, In the soup, and Loudmouth. The first is about not going through the distinct phases of a retro that allow you to get at value. (Symptom treatment or getting lucky.) The second is about focusing on stuff you can’t control. And the third is about someone who can dominate a group by being chatty or telling long stories, etc.
- 5 retro stages
- Setting the Stage Intro, theme, follow up on previous action points
- Gather Data Collect data about how the team is working together
- Generate Insights Find the causes behind the issues, understand and learn
- Decide what to do Find a few action points or experiments to try in the near future
- Close the Retrospective Make sure there is a concrete plan for the action points, outro
- A book with more of this stuff
