How to Get a Service Mesh Into Prod without Getting Fired
summary
Four practical lessons for getting a service mesh into production: use Linkerd or conduit as examples, start with one service or one environment and start small, and prioritize human buy-in by mapping stakeholders, incentives and worries. Case studies include a 4,000-person education company that runs parallel service versions to survive nine‑month feature freezes (successful in prod >1 year) and a multi‑billion dollar bank that hit complex rollout bugs. Expect blame, add rich diagnostics, and have rollback and post‑mortem plans.
AI-written summary · about
chapters (27)
- 0:00Intro
- 0:10The Situation
- 0:44William Morgan
- 1:34What this talk is about
- 2:19Are you really going to get fired
- 2:41Realistic failure scenarios
- 3:55Scenario 1 You cant convince your peers
- 4:19Audience Participation Game
- 4:55The Problem
- 5:45Good Problem Bad Problem
- 7:00Solving the Right Problem
- 7:21The Incentive
- 7:47Morgans 4th Law
- 8:07Examples
- 9:19Case Study
- 10:39Case Study 2
- 11:13Case Study 3
- 12:33Lessons Learned
- 14:06Metrics
- 14:45Mesh Creators
- 15:10Example Problems
- 16:01You Cant Avoid This
- 16:47Failure Scenario 4
- 18:03Lessons from Failure
- 18:55Summary
- 19:37Whats Next
- 20:20Separating Concerns
related talks
25mMaking microservices a behaviour not a buzzword
55mA Decade of Progress: The Evolution of Production Engineering and Reliability
23mFrom Data Strategy to a Scalable Data Mesh – How to Successfully Navigate the Journey
53mHatching software development ecosystems (the gordian knot)
30mPlatform as a Product
41mMe, My Organization and My-croservices
41mThe six obstacles of moving to microservices and how to deal with them
32mHow to enable organizations to go faster