I upgraded the OS packages on all nodes 1. Cordoned and Drained a node 2. apt update && apt upgrade 3. reboot 4. repeat on next node the last one was rke-node04-prd now this is the node with the least load I would still like to balance the workloads to it edit: all other nodes have a higher cpu usage now because elasticsearch is reindexing

If you want to change how the kubernetes scheduler chooses a node, you could try reconfiguring the scheduler. https://kubernetes.io/docs/concepts/scheduling-eviction/kube-scheduler/#kube-scheduler-implementation Although if nodes are getting overloaded, it might be simpler to set higher resource requests in the pod specs.