Deep Dive: SIG Autoscaling
summary
Vertical Pod Autoscaler ships as a CRD, runs a recommender that writes lower/upper/target/uncapped target recommendations from metrics, and applies sizes via an admission plugin (updater kills pods to recreate them with new requests); example: recommender might want 5 GB but uses a 3 GB spec limit. Cluster Autoscaler (GA since Kubernetes 1.8) supports GCP, AWS, Azure, Alibaba, imports scheduler predicates to simulate a greedy bin‑packing scale up and uses node utilization and pod relocatability (below ~50%) to scale down.
AI-written summary · about
chapters (11)
- 0:00Start
- 0:44Horizontal Auto Scaling
- 1:25Vertical Auto Scaling
- 3:41Specification
- 10:30Recommendations
- 14:21Cluster Autoscaler
- 15:22The Scheduler
- 15:59Predicate Functions
- 16:49Priority Function
- 25:50Events on Ports
- 27:43Is There any Conflict if We Use Horizontal Autoscaler in Vertical Scale at the Same Time
