Managing CRDs and Operators in Practice
summary
Runs and maintains more than 100+ CRDs created by 10+ teams where ~70% come from upstream packages, and introduces a company CRD standard that enforces naming, labels and categories. Builds a resource metadata repository for CRD manifests, developer contact info and CI review; gates changes with Jenkins dry‑runs and syntax checks. Adds a rollout-control CRD and service‑account mounting to manage operator rollouts, pause/partition semantics, and Helm value conventions; reports these pieces reduce manual errors and make operator onboarding repeatable.
AI-written summary · about
chapters (19)
- 0:00Intro
- 1:06$ whoami
- 1:26Alibaba's Journey to Cloud Native 0
- 2:58CRD and operators at scale
- 3:35Why CRD and operators
- 5:10Who are building CRD and operator
- 8:16Problems: learn to write operators
- 10:12Solutions: speed up learning
- 11:03Problems improper naming
- 12:06Solutions: CRD standards
- 13:24common pitfalls and best practice
- 15:58Problems: manual access control
- 18:24Solutions: resource meta repo
- 19:36Solutions: CI/CD based access review
- 21:12test & review items
- 23:21kubeconfig & webhook issuer
- 24:28Problems: integration with CICD pipeline
- 27:55Solutions: common rollout control
- 30:52lessons and takeaways
related talks
28mOperators and Helm It Takes Two to Tango
38mTo Crd, or Not to Crd, That is the Question
34mExtending Kubernetes 101
43mLessons learned writing 10+ Kubernetes Operators
21mOperators for Beginners
29mKubernetes Reloaded AKA How to Avoid the Dev Plateau with CRDs
45mKubernetes Operators: Operating Cloud Native services at scale
19mSimplify Kubernetes Operator Development With a Modular Design Pattern