microxchg 2017

The overall feedback from other participants was that the niveau declined compared to previous years. Not able to judge that, I could still not find a single other company that went fully reactive which surprised me about a microservice conference. Nonetheless there were some great talks.

Some were reassuring (Reactivtity rocks, REST and synchronicity leads to the problems we are trying to avoid).

I better understood the dangers of reusability and will from now on push for only making things reusable after the third semantically identical usage.

From the unexpectedly great “Monitoring Bash Microservices At Scale” talk I learned quite something

  • What we did with terraform (wrapping it in a helper bash script to protect against invalid operations and reduce api surface has been taken much further by their terraform_helper.
  • github.com/weaveworks/kubediff diffs running k8s with local config
  • We should use more RED (rate, errors, duration) as base metrics
  • This principle we will definitely adopt: a tools to programmatically create Grafana dashboards. What they call github.com/weaveworks/grafanalib we will do in JS or Go and use it for a domain-specific reusable widget/graph library to be reused by all our dozens of microservices
  • Others did what we did with @style/kbomb (although we took things a bit further) github.com/errordeveloper/kubegen

I also learned words for the different types of monoliths that we have to protect against:

  • application monolith
  • joined at the DB
  • monolithic releases (coupled)
  • monolithic thinking (standardization)

And, I got inspired to use checklists for many many things from now on to reduce cognitive load (therefore making it harder to make mistakes) for loads of stuff and visibly improve processes

Finally Daniel Bryant gave some book recommendations which I will check out: “Production-Ready Microservices” and “John P. Kotter – Leading Change”.