How to build an operator that doesn't break prod
summary
Explains how OpenShift Dedicated SREs maintain dozens of real operators, uses operator sdk plus wrapped clients for AWS RDS and Kubernetes objects, and shows how splitting the reconcile loop into small, idempotent reconcile steps prevents race conditions. Demonstrates tests-first development: early unit tests force small functions, make external layers mockable, and form a CI safety net so changes can be rolled to many clusters with confidence. wrap external dependencies
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:44Real world example
- 2:14What does the operator do
- 2:57What does the BBQ operator do
- 3:18The reconcile loop
- 4:05What are operators
- 5:39Which operators are the best
- 7:06reconcile loop
- 8:43Operators are pack animals
- 10:02Operators are production software
- 11:51Architectural boundaries
- 13:14Reconciling
- 15:18Writing tests
related talks
35mBeyond Reconcile Loop - Patterns for Kubernetes Operator Development
43mLessons learned writing 10+ Kubernetes Operators
32mTestable Kubernetes Operators
29mPackage Operator: Large-Scale Day-2 Operations
31mKubernetes Operators within Reach of Any Developer
24mKubernetes Operators with Go
36mLessons learned writing 10+ Kubernetes Operators - Lars Francke and Jannik Heyl
39mWriting a Kubernetes/Openshift Operators Using Golang