Changelog

General

• Allow cordoning nodes before deleting them in scale-down
• Fix bug where a node that becomes ready after 2 mins can be treated as unready

AWS

• Set maxAsgNamesPerDescribe to 100
• Reduce memory usage while generating ec2
• Properly support AWS EC2 IMDSv2
• Improve misleading log

Azure

• Don't proactively decrement Azure cache for unregistered nodes

Cluster API

• Fix sample manifest

Packet

• Fix writing to a nil map in cloud-provider code
• Consider equinixmetal:// a valid prefix for providerID in the Node spec
• Make controller node label configurable

Images

• k8s.gcr.io/autoscaling/cluster-autoscaler:v1.20.1
• k8s.gcr.io/autoscaling/cluster-autoscaler-arm64:v1.20.1
• k8s.gcr.io/autoscaling/cluster-autoscaler-amd64:v1.20.1