I’ll write down a few thoughts across the 3 days worth of talks that are things I really want to reinforce in my own head. I’m unsure these are all ideas that I should be driving towards, but they are making me think right now.

  • Use histogram type visualization as a way to retain more data (closer to raw but not quite) that can show more info about how a service works
  • Traces should be considered first class citizens in a monitoring platform beside logs and metrics
  • Instrument your code!
  • Slis, slos, rates vs raw counts (# of matching events / all events)
  • Ben Rockwood is using playwright with github actions to do production monitoring
    • Every 30m kick off e2e testing with a client in a couple different regions
    • Pump results into a discord channel
    • Mondoohq/simmons
    • 3m time to run the full suite
  • Github actions is an automation tool (so is Jenkins!)

Live streams from the event

And a few slides that I like a lot

Definition of observability