GASP! My Container Cluster Just Went Down and I Don’t Know Why!
summary
Claim: enterprise clusters keep a north‑south central load balancer for single‑point traffic control rather than full service‑mesh east‑west. Implementation: host agents watch Docker events and perform dynamic registration of IP‑per‑pod endpoints into a scalable control‑plane API that can handle thousands of add/remove calls. Built: a disaggregated controller plus one‑proxy‑per‑node data plane (Avi controller + proxies) deployed as containers, with analytics, TLS throughput, autoscaling and percent‑based blue/green traffic shifting — proven in PayPal production.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:48Introduction
- 1:13Different Types of applications
- 2:06Load Balancing Techniques
- 6:04LB in a traditional VM world
- 7:08LB in a cluster-manager world
- 9:23Cluster Manager World: Challenges
- 17:04Architecture to meet these requirements
- 18:31Modern Distributed Architecture
related talks
33mHow Container Clusters, like Kubernetes, ...
29mControl Plane, Service, or Both? – Argo CD Multi-Cluster Architectures
35mAcing Distributed Container Management
27mInfrastructure in the New World of Containers...
35mMake Load-Balancing Great Again with !
24mFrom VMs To Containers: A DevOps Journey
35mScaling clusters declaratively with Kubernetes and Docker realisation
44mPreventing Controller Sprawl From Taking Down Your Cluster - When a Scalable Pattern Stops Being Scalable