kube-aws-autoscaler
kube-aws-autoscaler copied to clipboard
hjacobs
Simple, elastic Kubernetes cluster autoscaler for AWS Auto Scaling Groups
For larger clusters (100+ nodes), scaling down (say to 20 nodes) - takes an order of several hours due to a conservative nature of scale-down step in autoscaler. It would...
Upscaling or node replacement poses the risk of race conditions between application pods and required system pods (DaemonSet). We have not yet figured out a good way of postponing application...
Add a new option to filter ASGs which should be considered by name. This was proposed in #24
The autoscaler should ignore every `Pending` pod if its resource requests exceed the capacity of the biggest node. Such pods will never get scheduled (without changing the EC2 instance type)...
In my mind it is not always necessary to have a buffering enabled for the whole cluster, furthermore buffering it is not a guarantee of a smooth scale up in...
The Kube AWS Autoscaler currently relies on the `unschedulable` node property, but this will change in future Kubernetes releases. See https://github.com/kubernetes/kubernetes/issues/29178 and https://github.com/kubernetes/kubernetes/pull/31647 for deprecation of `kubelet` flag `--register-schedulable`