https://slnotes.blog.castrillon.ch/k8s/scaling/Recent content in Scaling on SL NotebookHugo -- gohugo.ioSun, 13 Sep 2026 17:49:54 +0200https://slnotes.blog.castrillon.ch/k8s/scaling/vpa/Sun, 13 Sep 2026 17:49:54 +0200https://slnotes.blog.castrillon.ch/k8s/scaling/vpa/VPA stands for Vertical pod autoscaler and is a solution to adjust pod resource requests and limits automatically to optimally use the infrastructure Info VPA needs to be installed on top of the cluster as it does not come by default. This can be done by installing: CRDs RBAC resources Deployments (one per component) pulling them from the official repo. Process In here: The recommender evaluates current and historical usage and creates a recommendation The updater then identifies pods that do not comply with the recommendation (have too much, too little) and evicts them This causes the deployment/statefulset/replicaset to create a new pod,which is injected the recommended values by the Admission Controller Attention VPA by default depends on the metrics server being available.https://slnotes.blog.castrillon.ch/k8s/scaling/hpa/Sun, 13 Sep 2026 17:02:30 +0200https://slnotes.blog.castrillon.ch/k8s/scaling/hpa/HPA stands for Horizontal Pod Autoscaler and its kubernetes built-in solution for scaling a replica set (adding more pods) based on resource consumption. Process Attention HPA by default depends on the metrics server being available. As mentioned in the monitoring section, this is an extra step In here: The hpa is aware of the limit declared by the deployment and is configured with a threshold Hpa monitors metrics servers periodically to check on resource consumption If it goes over the limit, it will add a pod When it goes down it will remove a pod Manually scaling The same process could be done manually by observing the metrics and then scaling with kubectl scale .