An overview of autoscaling features in Kubernetes, covering horizontal, vertical, and cluster autoscalers, and upcoming features like the new HPA v2 stable API and support for alternative recommenders in VPA.
- SIG Auto Scaling is responsible for horizontal and vertical polynomial scaling, cluster proportional system component auto scaling, and cluster autoscaler.
- Horizontal Pod Autoscaler (HPA) scales in and out based on pre-defined metrics, while Vertical Pod Autoscaler (VPA) adjusts resources for workloads that don't necessarily scale horizontally.
- Cluster autoscaler adjusts the number of nodes in any of the node pools that are configured.
- The new HPA v2 stable API and support for alternative recommenders in VPA are upcoming features.
- The autoscaling community can be accessed through Slack channels and weekly meetings.
The speaker gave an example of a cluster with two workloads, one configured to scale horizontally and the other configured to scale horizontally and vertically. The HPA adjusts the number of pods based on pre-defined metrics, while the VPA adjusts resources for workloads that don't necessarily scale horizontally. The cluster autoscaler adds capacity and brings it back in, which is an important part of the autoscaling system.