A Tale of Autoscale: Increasing Resiliency while Reducing On-Call Toil

on YouTube ↗14m

" summary

A team with no SRE experience converts stateless ingest services on Kubernetes to autoscale, choosing CPU-based HPA over VPA because it ships with clusters, setting per-region min/max replicas and adding a detector for max-reached to alert on sizing limits. Implementing HPA against CPU metrics and conservative thresholds reduces sizing alerts by 75% reduction, cuts manual paging, and prevents runaway cloud costs while keeping downstream services stable.

AI-written summary · about

" chapters (10)

1. 0:00Introduction
2. 0:48Autoscaler rise and on-call
3. 2:56Incidents to reduce toil
4. 4:20Choosing Kubernetes horizontal pod autoscaling
5. 6:28Configuring HPA automation
6. 8:15HPA thresholds and limits
7. 10:25HPA YAML and sizing
8. 11:50Monitoring max replicas
9. 12:11Results and HPA adoption
10. 12:54Summary

" related talks

• 31mFailing to Autoscale?Erwin Daria · SREday 2023 London
• 35mSIG Autoscaling IntroMarcin Wielgus, Solly Ross · KubeCon + CloudNativeCon Europe 2018
• 42mMaking Resilient, Self-Adaptive, And Autonomous Kubernetes ClustersViktor Farcic · Craft Conference 2019
• 51mHorizontal Autoscaling with KubernetesRob Richardson · KCD Los Angeles 2023
• 36mIntro: Autoscaling SIGMarcin Wielgus · KubeCon + CloudNativeCon Europe 2019
• 29mKubernetes Autoscaling: Beyond One Size Fits AllJohn Platt · KubeDay Colombia 2024
• 37mDeep Dive into AutoscalingMarcin Wielgus, Vivek Bagade · KubeCon + CloudNativeCon North America 2019
• 17mScaling Made Simple: Implementing Autoscaling for Optimal K8s PerformanceEli Birger · KubeDay Israel 2023