Predictive autoscaling patterns in Kubernetes
summary
Shows a working Vertical Pod Autoscaler setup that reads Kubernetes metrics and automatically adjusts pod requests and limits so a memory‑stressed demo app avoids crash loops by restarting with higher resources. Describes VPA components recommender, updater, admission plugin and modes auto/initial/off, shows min/max caps to prevent runaway allocations, warns against HPA on the same resource, and recommends pairing VPA with Cluster Autoscaler so recommendations don’t leave pods Pending.
AI-written summary · about
chapters (8)
- 0:00Introduction and speakers
- 1:29Resource limits challenge
- 5:01Consequences of wrong limits
- 7:49Vertical Pod Autoscaler solution
- 10:59VPA modes and architecture
- 13:25Combining VPA and Cluster Autoscaler
- 15:13Live demo setup
- 23:20Q&A session
