Managing Large-Scale Kubernetes Clusters Effectively and Reliably
summary
Covers Ant Financial’s approach to automating management of very large Kubernetes deployments (10k+ nodes per data center) across multiple data centers. Describes fully automated node lifecycle management including failure detection and recovery using NPD, Autoscalers and a custom Operator; automated cluster deployment and upgrades; and a Prometheus + Operator based risk prevention/control system for automatic fault detection and isolation.
AI-written summary · about
chapters (13)
- 0:00Start
- 0:17About Me
- 0:39Agenda
- 2:21Background: Cluster Scale
- 4:41Motivation
- 6:38Design Concept: Imperative
- 8:26Design Concept: Declarative
- 10:13Design Concept: Self-recovery
- 12:15Cluster-Operator
- 13:43Machine-Operator
- 16:08Node Final-state Management
- 18:56Fault Self-recovery
- 24:51Risk Prevention
related talks
37mManaging Large-Scale Kubernetes Clusters Effectively and Reliably
34mA story of managing kubernetes cluster with 15k nodes & various workloads
24mDynamic Big Data in the Cloud with Kubernetes
41mKubernetes the Very Hard Way
15mDemo: Resource(Cost) optimization on planet scale Kubernetes clusters.
35mAutoscaling Kubernetes Deployments: A (Mostly) Practical Guide
35mExtending Deployment for Internet Financial Mission-Critical Scenarios
28mScaling in Kubernetes Safely on On-Prem KaaS Across 1,300+ Clusters and 40,000+ Nodes