Horizontal Pod Autoscaler Reloaded - Scale on Custom Metrics

on YouTube ↗30m

" summary

HPA v2 adds custom metrics so Kubernetes can scale on per-pod, object, or external signals rather than only CPU; demo uses Prometheus adapter and the Metrics Server, shows HPA polling every 50s and computing replicas by total utilization ÷ target, and scales a demo service from 3 to 6 replicas when HTTP QPS exceeds a 1 request/second target. External metrics (Stackdriver) and object metrics with label selectors are supported in 1.10 and work as shown.

AI-written summary · about

" chapters (13)

1. 0:00Introduction to scaling
2. 1:00Scaling approaches
3. 2:13Kubernetes auto-scaling
4. 3:28HPA v1 vs v2 features
5. 6:06Basic HPA demo
6. 7:35How HPA works
7. 9:15Custom metrics with pods
8. 10:37Troubleshooting HPA
9. 13:03The Prometheus adapter
10. 16:00Object metrics scaling
11. 18:55External metrics
12. 21:00Future plans and closing
13. 22:13Q&A session

" related talks

• 51mHorizontal Autoscaling with KubernetesRob Richardson · KCD Los Angeles 2023
• 29mKubernetes Autoscaling: Beyond One Size Fits AllJohn Platt · KubeDay Colombia 2024
• 35mIntro: SIG AutoscalingMarcin Wielgus, Maciej Pytel · KubeCon + CloudNativeCon China 2018
• 08mGetting Creative with HPA Scaling in KubernetesDrew Rogers · Philly ETE 2023
• 24mSIG-Autoscaling: Introduction + HPA/VPA UpdatesMarcin Wielgus, Joe Burnett · KubeCon + CloudNativeCon North America 2021
• 25mAutoscaling Kubernetes Workloads Based On MetricsRakshit Gondwal · KCD Mumbai 2023
• 34mScale your microservices by MicroProfile Metrics on KubernetesKenji Kazumura · EclipseCon 2022
• 35mAutoscaling Kubernetes Deployments: A (Mostly) Practical GuideNatalie Serrino · KubeCon + CloudNativeCon Europe 2022