cost-manager does not currently have a well-defined roadmap, however here we describe some Kubernetes cost optimisations that could be automated as cost-manager controllers in the future:
- Generating VerticalPodAutoscalers: The Vertical Pod Autoscaler (VPA) automatically adjusts Kubernetes Pod resource requests based on actual usage and can help reduce over-provisioning. By automatically generating VPA resources for all cluster workloads (being careful to consider the limitations when using VPA together with HPA) operators can make sure that workloads are only requesting the resources that they need
- Garbage collecting disks and load balancing infrastructure: PersistentVolumeClaims and Services can be used to automatically provision cloud resources, however if the cluster is deleted without first deleting these resources then the cloud resources can become orphaned. By using metadata on these resources, a controller can be used to automatically detect orphaned resources provisioned by Kubernetes and clean them up to save costs
- Scheduled cluster scale down: There are many cases where cluster workloads do not need to be running all the time (e.g. CI infrastructure or development clusters). On a schedule, ResourceQuotas can be used to limit Pods in all Namespaces (except for the cost-manager Namespace) and then all Pods deleted to allow the cluster to scale down. To scale back up, the ResourceQuotas can simply be deleted