Music Matched to Different Ways to Scale in K8s, What Could Go Wrong?
summary
Explains how Kubernetes scaling actually behaves: cluster autoscaler (1,000 nodes × 30 pods test) auto-adds and removes nodes; HPA on requests-per-second with threshold 10 uses custom metrics to add replicas; ACI/virtual nodes spillover to virtual couplet runs containers without VMs so a cluster can scale pods into Azure Container Instances instantly. Notes VPA conflicts with HPA, recommends admission controllers to inject node affinities, and demos scaling to a dozen+ ACI pods from a small node pool.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:41Cluster autoscaler
- 2:26Vertical pod autoscaler
- 4:37Horizontal pod autoscaler
- 7:21Virtual node architecture
- 11:33Use anti-affinity or affinities to schedule to specific kinds of nodes
- 12:47Use tolerations & admission controllers to be
related talks
17mScaling Made Simple: Implementing Autoscaling for Optimal K8s Performance
31mScaling Your Cluster (Both Ways)
34mIntroduction to Autoscaling
35mSIG Autoscaling Intro
13mUsing Virtual Nodes to Auto-Scale Your Application on Kubernetes
35mAutoscaling Kubernetes Deployments: A (Mostly) Practical Guide
07mAutoscaling with virtual node and Azure Kubernetes Service (AKS)
08mGetting Creative with HPA Scaling in Kubernetes
