Pods migration from one node to another
0
I currently have an unmanaged node group consisting of 2 worker nodes. Node1 has pods running and has enough CPU capacity to run node2's pods. Is there a way in AWS to automatically migrate pods from node2 to node1 and scale down node2? So whenever a worker node has sufficient CPU capacity, it will be filled and the other node will be scaled down to optimize the costs of an unnecessary node?
1 Answers
0
Just wanted to understand how are you scaling instances ? Are you not leveraging Cluster Autoscaler or Karpenter which takes care of scale in and out automatically. If none of this is implemented, you can try safely draining the node. https://kubernetes.io/docs/tasks/administer-cluster/safely-drain-node/
answered a month ago
Relevant questions
Pods migration from one node to another
asked a month agoEKS static IPs for managed node group nodes
Accepted Answerasked 2 years agoadd node into node group
asked a month agoEKS node capacity settings with Karpenter
asked 7 months agoEKS Worker-node join
asked 3 years agoforce auto scaling group to scale in by terminating k8s pods ungracefully
asked a year agoEKS pods get node IP address instead of IP from Calico IP Pool
Accepted Answerasked 3 months agoHow to create EKS cluster with dedicated host node group
Accepted Answerasked 7 months agoCoreDNS can't resolve on public subnet in VPC with public and private subne
asked 3 years agoGuarantee response to specific node? (Microservices intercommunication)
asked 3 years ago
Just to clarify my question: we have a k8s cluster for development purposes, and we have many apps (each app running on a pod) running on different nodes. our strategy is to first run an instance and fill it with as many applications as possible, then scale up a new instance if needed. Until there everything is working fine. Sometimes we need to stop an app (eg the app is running on node1), and now we have node1 which can handle an additional pod. My question is if Cluster Autoscaler or Karpenter can move a pod from node2 to node1 so if node2 is empty of pods can be scaled down?