Open Infrastructure Days UK 2019

Currently, more and more Kubernetes adaptors have an increasing number of k8s clusters. Kubernetes federation is the feature that allow an easy management of multiple clusters, with a consistency of resources running across clusters and cross cluster discovery. However, why is Kube Fedetation still in it’s early stages and can it be used in a production environment? Vertical pod autoscaler is a mechanism that automatically computes the resource requests for an application. It is a easy install CRD, that recently moved to a beta version (v0.3.0), and is composed of 3 parts (recommender, updater and admission controller). How VPA can be used to provide better usage of cluster resources? Horizontal pod autoscaler is responsible to provide a mechanism that will automatically adjust the amount of replicas required for an application base on a specific metric or resource. Since Kubernetes v1.10, external metrics are supported by the HPA object, which exposes any outside-cluster metrics to the autoscaler. Hence, at this point in time, HPA in a combination with cluster autoscaler can expand the size of the cluster on any mereics provided by the metrics APIs provice (e.g. Datadog cluster agent). Why it’s important to delegate the horizontal scaling to HPA and how in combination with cluster autoscaler, k8s can scale on any internal or external metrics?