Changelog

General

• Subtracting mirror pod with Daemonsets when calculating node utilisation in --ignore-daemonsets-utilization flag
• Added Issue templates for Bug Report and Feature Request
• Add max pods per node config to build node
• Bugfix: Expander Priority warns misleading log
• Make arch-specific releases use separate images instead of tags on the same image
• Added towca as an approver
• Reference to additionalLabels in deployment template for charts fixed
• Added Per nodegroup scale-down config
• Updated Chart testing action
• Define clusterapi in cloudprovider layer
• Follow WG Naming Recommendations on Master -> Control Plane
• Implement round-robin for SchedulerBasedPredicateChecker
• Allow Daemonsets to safely evict on empty nodes
• Migrated Test and Verify to using Git Actions
• Cleanup Makefile
• Helm Chart Image Update (v1.18.1 → v1.20.0)
• Fix: Set different hostname label for upcoming nodes
• Update Index for CA Chart
• Final removal of ca-chart chart post deprecation
• Allow name of cluster-autoscaler status ConfigMap to be specificed
• Added prometheus rule support for CA Chart
• Allow for specifying automountServiceAccountToken on cluster-autoscaler service account for CA Chart
• Simplify vendor update script
• Add arch specific cluster-autoscaler targets to gitignore
• Add a metric for unregistered nodes removed by CA
• Fix bug where a node that becomes ready after 2 mins can be treated as unready. Deprecated LongNotStarted
• Release leader election lock on shutdown
• Support for reporting authorization errors during scale up
• adds status-config-map-name to values.yaml of helm chart
• Log failed scale up metric based on string value of AutoscalerErrorType
• Separate and refactor custom resources logic
• Set cluster autoscaler-specific user agent
• Added OVHcloud as a new cloud provider
• Added Linode as a new cloud provider
• Added Bizflycloud as a new cloud provider
• Added Hetzner Cloud as a new cloud provider

GCE

• Reading Ephemeral Storage from Instance Template and allow scale from zero of nodepools for pods requesting ephemeral storage
• Added flag to block boot disk being used as ephemeral storage when picking from instance template
• Added additional OS Distributions
• Added price for e2-highcpu-32
• Fixing Kernel Memory Usage estimation
• expose IP_SPACE_EXHAUSTED error
• Add A2 machine family pricing
• additional memory reservation for low memory machines

AWS

• Added new instance types: r5b, m5zn, d3, d3en
• Reduce logging when creating placeholder instances
• Don't pile up successive full refreshes during AWS scaledowns
• Support new GPU type - T4
• Add Fetch Error Output to Warning Log
• Add gjtempleton to cloud provider OWNERS approvers
• Allow generic labels to be overwritten by tags coming from the ASG

Huaweicloud

• Cleanup unused SDK

Azure

• Updated Instance type list
• add missing call to fetch autodiscovered nodegroups

ClusterApi

• Fix sample manifest
• Switch from v1beta1 to v1 for rbac
• Add services into role of example file
• Bump clusterapi sample suggested version from 1.18.1 to 1.20.0
• CAPI Management cluster informer to watch single namespce

Linode

• Implement new GetOptions method for linode provider

OVHcloud

• Add node group autoscaling options

Magnum

• Enable magnum provider scale to zero

Images:

• k8s.gcr.io/autoscaling/cluster-autoscaler:v1.21.0
• k8s.gcr.io/autoscaling/cluster-autoscaler-arm64:v1.21.0
• k8s.gcr.io/autoscaling/cluster-autoscaler-amd64:v1.21.0