In an effort to optmize as much as possible the resources being used by the infrastrucutre components running on our Reference Solution Kubernetes platforms, we’ve considerably reduced the memory used by the Cluster Autoscaler by optimizing its configuration. This means that the autoscaler will now run more reliably, and that there’ll be a bit more memory available for other workloads running on the K8s clusters.
Changes have been rolled out to all EKS clusters already.