Panel: Improving and Managing Kubernetes at Scale
Panelat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 15:20· Room 6C - San Diego Convention Center, Upper Level·
summary
Multiple companies report concrete limits and fixes: control plane and etcd benchmarking shows storage and API server as bottlenecks; an 8k nodes -> 40k pods: 5m to 30s experiment cuts launch time by suppressing pod events; deployment models include public shared pools and dedicated clusters with API gateway frontends; runtime tweaks include virtual kubelet, sidecar controllers, injected API servers (virtual clusters), CPU sets and two‑level scheduling for collocation; largest fleets claim >6,000 nodes; >100k containers per cluster and require staged upgrades with week‑long baking.
AI-written summary · about
chapters (4)
- 0:00Start
- 18:06Testing against Api Server
- 21:59What Is the Biggest Pain Point You Guys Face When Adopting Kubernetes and What Is the Status of each of the Companies
- 26:45How Do We Use Kubernetes Extension Points To Fit in an Already Rich Infrastructure Ecosystem
related talks
36mUnderstanding Scalability and Performance in the Kubernetes Master
48mScale and Performance Testing of Kubernetes: Answers for Specific Applications
50mScale Kubernetes to Manage 60K Nodes Through Architectural Extension
15mDemo: Resource(Cost) optimization on planet scale Kubernetes clusters.
32mScaling Kubernetes Application to 1M Users
28mScaling in Kubernetes Safely on On-Prem KaaS Across 1,300+ Clusters and 40,000+ Nodes
38mHow Can Pod Start-up Be Accelerated on Nodes in Large Clusters?
03mFrom 5 to 1,300+ Clusters: Declarative Scaling for Private Kubernetes