Changelog

Note: This is the first release to have an official s390x architecture image. This is also supported under the multi-arch image.

General

• Support for a new cloud provider - VolcEngine - #5637
• Add pod namespace when logging why scaleup wasn't possible - #5740
• A new "parallel-scale-up" flag, that makes the scale-up process run parallel across multiple node groups. The default behavior is unchanged, node groups are scaled synchronously, one by one. Before using the flag make sure that the cloud provider you're using is thread-safe and ready for concurrent actions. - #5731
• Add cluster-autoscaler name and version to the user agent - #5873
• When using the --balance-similar-node-groups flag of the Cluster Autoscaler, more logging is now available at verbosity levels 2 and 3 to help diagnose balancing failures. - #5835
• Fixed an issue where balancing nodes between zones (--balance-similar-node-groups flag) didn't work correctly in some edge cases like very large scale-ups and node groups being close to max size. - #5750
• Fix similar node group calculation for non-existing node groups - #5760
• Bugfix - Fix race condition between CA fetching list of scheduled pods and unscheduled pods - #5705
• Bugfix - Snapshot actuation status before refreshing cache - #5777
• Bugfix - Skip checking non-existing similar node groups on "safe to scale-up" criteria, similarly to how we don't check it for non-existing primary node groups. - #5824
• Bugfix - When comparing PodSpec semantically, drop projected volumes for init containers #5852

AliCloud

• Support RRSA auth - #5340

AWS

• Allow k8s.io/cluster-autoscaler/resources/: tags on AWS managed nodegroups (in addition to ASGs) to indicate availability of custom resources or ephemeral storage in a scale-from-zero scenario - #5596
• Add support for Trn1n.32xl instance - #5735
• Track lifecycle and skip terminating already terminating instances - #5411
• Add preview EC2 instance type p4de.24xlarge - #5773

Azure

• Fast nodegroup backoff on failed provisioning - #5548

Brightbox

• Add support for building new nodes using an image name as well as an ID - #5764

Civo

• Support pool wise config - #5743

Cluster API

• Fix the buildTemplateLabels method for the ClusterApi provider - #5697

GCE

• Add pricing info for G2 - #5719
• Bugfix - fixes a possible map write race condition in GCE cloud provider - #5800
• Bugfix - Fix zonal GCE outage breaking CA when only some of the zones are failed - #5853

Hetzner

OCI

• Enhance oci implementation to include oci node pools - #5577
• Added OCI Cloud Provider Workload Identity support - #5717

Magnum

• Change magnum to use control-plane role as renamed - #5776

VolcEngine

• Now supported - #5637

Vultr

• Refactored Vultr node delete - logging for nodes missing provider and node IDs changed - #5709

Images

• k8s.gcr.io/autoscaling/cluster-autoscaler:v1.28.0
• k8s.gcr.io/autoscaling/cluster-autoscaler-arm64:v1.28.0
• k8s.gcr.io/autoscaling/cluster-autoscaler-amd64:v1.28.0
• k8s.gcr.io/autoscaling/cluster-autoscaler-s390x:v1.28.0