Deep Dive Into API Machinery
Talkat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 14:25· Room 1AB - San Diego Convention Center, Upper Level·
summary
Details on marking CRD fields immutable, strict defaulting/validation/equality and new list semantics in v1.16 (lists-as-set and lists-as-map with key fields) and how atomic vs merge strategies change updates. Server-side apply stores per-field ownership in object metadata, detects conflicts (returns a 409 with owner and timestamp), and currently adds measurable CPU and object-size overhead. API Priority and Fairness uses flow schemas, priority levels with assured concurrency shares, queueing via shuffle sharding, and options to drop or queue excess requests.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:51Immutability
- 2:56Defaulting
- 3:53Equality
- 5:04Normalization
- 6:00Request Pipeline
- 6:56Imutability
- 8:10Merging
- 9:07ServerSide Apply
- 11:37Southside
- 12:04Tracking Intent
- 12:33Workflows
- 13:11Tracking Ownership
- 15:32Merge
- 16:19Missing Features
- 20:59Example Configurations
- 22:49Unions
related talks
38mDeep Dive: API Machinery SIG
30mSIG API Machinery: Project Updates and Release Planning
29mSIG API Machinery in the Era of AI: Updates
29mBoxcutter - or what we learnt about managing objects with Server-Side-Apply in K8
10mAPI Management in the CRD World: What Linkerd Has Learned
30mAPI Priority and Fairness: Kube-APIServer Flow-control Protection
30mAPI Evolution With CRDs: Best Practices For Authoring & Fuzz Testing APIs
19mKeynote: CRDs, No Longer 2nd Class Thing!