Workload cluster release aws-33.0.0 for CAPA

WARNING: This release enables the OwnerReferencesPermissionEnforcement admission controller by default. This means that only users with delete permission to an object can change its metadata.ownerReferences, and only users with update permission to the finalizers subresource of the referenced owner can change metadata.ownerReferences[x].blockOwnerDeletion. If you have workloads that need to modify these fields, please ensure that the necessary RBAC permissions are in place before upgrading to this release.

Example:

- apiGroups: ["<group>"]
  resources: ["<resource>", "<resource>/finalizers"]
  verbs: ["delete", "..."]  # Add any additional verbs your use case requires

Changes compared to v32.0.0

Components

  • cluster-aws from v5.0.0 to v6.0.0
  • Flatcar from v4230.2.2 to v4230.2.3
  • Kubernetes from v1.32.9 to v1.33.5

cluster-aws v5.0.0…v6.0.0

Added

  • Add standard tags to IRSA infrastructure.
  • Expose value to configure terminationGracePeriod in the karpenter node pools.

Changed

  • Chart: Update cluster to v4.2.0.
  • The container registry passed as value to default apps is set to gsoci.azurecr.io, regardless of the cluster region. The mirroring feature of containerd will make sure the right registry is used.
  • Switch to HelmReleases to install karpenter and karpenter-crossplane-resources charts.
  • Bump flux HelmReleases api version to v2.
  • Reduce heartbeat timeout for ASG lifecycle hooks to from 30 minutes to 3 minutes since aws-node-termination-handler-app (NTH) can now send heartbeats
  • Configure the following startupTaints to help karpenter ignore pending Pods due to these taints that will be removed after the node starts, avoiding unnecessary instance provisioning:
    • node.cluster.x-k8s.io/uninitialized:NoSchedule
    • node.cilium.io/agent-not-ready:NoSchedule
    • ebs.csi.aws.com/agent-not-ready:NoExecute
  • Include cilium ENI mode pod CIDRs in the NodePort Services security group ingress rules.

Removed

  • Removed capi-node-labeler app. From now on, the worker nodes won’t have the node-role.kubernetes.io/worker or node.kubernetes.io/worker labels.

Apps

  • aws-ebs-csi-driver from v3.0.5 to v3.2.0
  • aws-nth-bundle from v1.2.2 to v1.3.0
  • cert-exporter from v2.9.9 to v2.9.12
  • cert-manager from v3.9.2 to v3.9.3
  • cilium from v1.3.0 to v1.3.1
  • cloud-provider-aws from v1.32.3 to v1.33.2-1
  • cluster-autoscaler from v1.32.2-gs1 to v1.33.1-2
  • coredns from v1.27.0 to v1.28.2
  • etcd-defrag from v1.0.8 to v1.2.1
  • etcd-k8s-res-count-exporter from v1.10.7 to v1.10.9
  • k8s-audit-metrics from v0.10.6 to v0.10.8
  • Added karpenter v1.3.0
  • Added karpenter-crossplane-resources v0.4.0
  • node-exporter from v1.20.5 to v1.20.7
  • observability-bundle from v2.2.2 to v2.3.2
  • vertical-pod-autoscaler from v6.0.1 to v6.1.1
  • vertical-pod-autoscaler-crd from v4.0.1 to v4.1.1

aws-ebs-csi-driver v3.0.5…v3.2.0

Changed

  • Configure gsoci.azurecr.io as the default container image registry.
  • Set default updateStrategy.rollingUpdate.maxUnavailable to 25% in DaemonSet to speed up rolling update.

aws-nth-bundle v1.2.2…v1.3.0

Changed

  • Upgrade aws-nth-crossplane-resources to v1.3.0, fixing support for multiple OIDC providers in the NTH IAM role as required for cleanup of migrated vintage clusters, and supporting heartbeat sending
  • Upgrade aws-node-termination-handler-app to v1.23.0, enabling heartbeats by default and upgrading to upstream application version v1.25.2 which fixes a resource leak bug relevant to heartbeat sending
  • Upgrade aws-nth-crossplane-resources to v1.1.1, supporting multiple OIDC providers in the NTH IAM role as required for cleanup of migrated vintage clusters

cert-exporter v2.9.9…v2.9.12

Changed

  • Go: Update dependencies.
  • Chart: Add value to toggle creation of Daemonset resources.
  • Go: Update dependencies.

cert-manager v3.9.2…v3.9.3

Changed

  • Fix missing targetPort in cainjector-service

cilium v1.3.0…v1.3.1

Changed

cloud-provider-aws v1.32.3…v1.33.2-1

Changed

  • Chart: Update to upstream v1.33.2.

cluster-autoscaler v1.32.2-gs1…v1.33.1-2

Changed

  • Chart: Grant access to VolumeAttachments. (#345)
  • Update Kyverno API to v2 for policy exceptions
  • Chart: Update to upstream v1.33.1.

coredns v1.27.0…v1.28.2

Changed

  • Update coredns image to 1.13.1.
  • Add value to toggle creation of controlplane deployment.
  • Update coredns image to 1.13.0.

etcd-defrag v1.0.8…v1.2.1

Changed

  • Chart: Update dependency ahrtr/etcd-defrag to v0.34.0. (#62)
  • Chart: Update dependency ahrtr/etcd-defrag to v0.33.0. (#60)
  • Update Kyverno API to v2 for policy exceptions
  • Chart: Update dependency ahrtr/etcd-defrag to v0.32.0. (#57)

etcd-k8s-res-count-exporter v1.10.7…v1.10.9

Changed

  • Go: Update dependencies.
  • Update Kyverno API to v2 for policy exceptions
  • Go: Update dependencies.

k8s-audit-metrics v0.10.6…v0.10.8

Changed

  • Go: Update dependencies.
  • Update Kyverno API to v2 for policy exceptions
  • Go: Update dependencies.

karpenter v1.3.0

Changed

  • Updated karpenter to 1.7.1

karpenter-crossplane-resources v0.4.0

Changed

  • Add iam:ListInstanceProfiles for release 1.7.1

node-exporter v1.20.5…v1.20.7

Changed

  • Go: Update dependencies.
  • Update Kyverno API to v2 for policy exceptions
  • Go: Update dependencies.

observability-bundle v2.2.2…v2.3.2

Added

  • Add KSM metrics for cloudnative-pg Cluster objects

Changed

  • Update alloy-app to 0.15.0
    • Bumps alloy to 1.11.0

Fixed

  • Update alloy-app to 0.15.1
    • Bumps alloy to 1.11.2

vertical-pod-autoscaler v6.0.1…v6.1.1

Changed

  • Chart: Update Helm release vertical-pod-autoscaler to v11.1.1. (#375)
  • Chart: Update Helm release vertical-pod-autoscaler to v11.1.0. (#372)

vertical-pod-autoscaler-crd v4.0.1…v4.1.1

Changed

  • Chart: Sync to upstream. (#166)
  • Chart: Sync to upstream. (#164)