Y55. Chapter III kubernetes from entry to proficiency -- HPA controller and metrics server (28)

11.HPA Controller and metrics-server

11.1 HPA Automatic telescopic pod Number

kubectl autoscale Automatic control at k8s Running in a cluster pod Number ( Horizontal auto retraction ), It needs to be set up in advance pod Scope and trigger conditions .

k8s from 1.1 The version began to add the name HPA(Horizontal Pod Autoscaler) The controller , For implementation based on pod China Resources (CPU/Memory) The utilization rate is adjusted pod Realization of automatic expansion and contraction function of , Earlier versions can only be based on Heapster Component implementation for CPU Utilization as trigger condition , But in k8s 1.11 Version in use Metrices Server Complete data collection , Then the collected data is passed through API（Aggregated API, Summary API）, for example metrics.k8s.io、custom.metrics.k8s.io、external.metrics.k8s.io, Then provide the data to HPA Controller to query , To achieve a resource utilization based on pod The purpose of expansion and contraction .

 The control manager defaults to every 15s（ Can pass –horizontal-pod-autoscaler-sync-period modify ） Inquire about metrics Resource usage of 
 The following three types are supported metrics The index type ：
	-> predefined metrics（ such as Pod Of CPU） Calculated in terms of utilization 
	-> Self defined Pod metrics, At original value （raw value） Method of calculation 
	-> Self defined object metrics
 Two kinds of support metrics A query ：
	->Heapster
	-> Self defined REST API
 Support more metrics