Planes, Raft, and Pods: A Tour of Distributed Systems Within Kubernetes [B]
summary
Examines etcd + Raft to show how Kubernetes enforces consistency via leader election, terms and majority quorum (e.g., a 5-node cluster needs 3 to stay writable), then digs into controller mechanics like replica set reconciliation and the scheduler’s filtering and ranking functions. Uses kubectl, the API server, kubelet and Docker Hub to deploy a Hello World Go app: a Deployment creates a ReplicaSet then three pods; the scheduler assigns all three to one node and kubelet pulls the image and starts the containers successfully.
AI-written summary · about
chapters (30)
- 0:00Intro
- 0:24About Bo
- 1:18Whats a distributed system
- 2:16Kubernetes
- 14:59Leaders
- 15:15Followers
- 16:28Heartbeat
- 16:38Follow Rule
- 16:58Game of Thrones
- 17:46Rights
- 18:45Simultaneous Right
- 19:40Cap Theorem
- 20:05Raft Paper
- 20:36Controllers
- 20:50Replica Set
- 21:30Deployment Controller
- 22:25Scheduler
- 22:49Scheduling Algorithm
- 23:03Predicates
- 26:07Ranking
- 26:51Ranking Functions
- 28:51Hello World Go
- 29:06Pod Selection
- 29:30QCuttle
- 30:18Warning
- 31:27Scheduling
- 32:57Downloading Images
- 34:01Starting a Container
- 34:36We did it
- 35:02Recap
related talks
28mKubernetes Plane Aerodynamics: Breaking the Architectural...
36mUnderstanding Kubernetes Through Real-World Phenomena and Analogies
30mWho Let the Pods Out? Extending Kubernetes with Custom Controllers and CRDs
23mBeyond Block Diagrams: Different Ways of Understanding K8s Architecture
44mA Practical Guide to Container Scheduling
32mYAML Your Cloud! Managing Your Hosted Resources With Kubernetes
32mKubernetes as a platform for platforms
41mAbstractions on Abstractions, Are You Building for Kubernetes?