[gardener]

⚠️ Breaking Changes

• [DEVELOPER] Deployment specific information from ControllerRegistrations was outsourced to a new resource type ControllerDeployment. If you use Gardener's /hack scripts and utilities, please make sure to run make generate after you have updated your Gardener dependencies to this version, as it updates the example registration in your extension. Otherwise, please make sure to adjust your example ControllerRegistration manifest to the new structure. (#3995, @timuthy)

✨ New Features

• [USER] The shoot resource now support optional columns that are shown only when kubectl get is used with the -o wide flag. (#4032, @vpnachev)
  • The optional columns are: SEED, PURPOSE, GARDENER VERSION, APISERVER, CONTROL, NODES and SYSTEM.
  • New mandatory columns REGION and STATUS are introduced.
• [USER] Add the resource ShootExtensionStatus to the resource group core.gardener.cloud which is intended to hold the information of the ProviderStatus field from multiple Gardener Extension resources from the Seed cluster. (#3878, @danielfoehrKn)
• [OPERATOR] A new resource ControllerDeployment was added to the core.gardener.cloud API group. A ControllerDeployment is supposed to contain specifications about how extension controllers are deployed to seed clusters. Specifying deployments via ControllerRegistrations directly is therefore deprecated. Instead, we recommend to use a dedicated ControllerDeployment resource and reference it in a ControllerRegistration via spec.deployment.DeploymentRefs[].Name. Please consult the documentation for more information. (#3995, @timuthy)
  • As ControllerDeployments may contain sensitive information, we recommend to add this type to the Gardener API server's EncryptionConfiguration. This happens automatically if you deploy Gardener via our pre-configured Helm charts (/charts/gardener).
• [OPERATOR] Introduce compressed etcd backups for faster upload and download of etcd snapshots. (#3953, @shreyas-s-rao)

🐛 Bug Fixes

• [OPERATOR] A bug generally preventing ManagedSeed deletion has been fixed. (#4045, @rfranzke)
• [OPERATOR] Fix an issue where the gardenlet overwrites changes from hvpa for the aggregate-prometheus. (#3998, @wyb1)

🏃 Others

• [OPERATOR] Fixed an issue that prevented the update and deletion of managed seeds on soil clusters. (#4070, @stoyanr)
• [OPERATOR] Gardener administrators are now allowed to access certificatesigningrequests. (#4059, @ialidzhikov)
• [OPERATOR] Some shoot clusters try to wrongly configure the kube-apiserver and it fails to start. Since some time, Gardener shows the last 10 log lines for such kube-apiservers in the `shoot..status.lastOperation.description', but error message with the useful information might not be in those lines. Therefore, for shoots clusters running on k8s <1.19 now also the first 1KiB of logs will be shown. (#4050, @vpnachev)
• [OPERATOR] Adjust hvpa limitsRequestsGapScaleParams for prometheus (#4049, @wyb1)
• [OPERATOR] Apply a mitigation that will prevent gardenlet to panic under certain circumstances. (#4046, @ialidzhikov)
• [OPERATOR] istio images are now pulled from the official mirrors to Google Container Registry to prevent any potential dockerhub rate limit issues in environments that use the default images (don't specify any imagevector overwrite). (#4022, @ialidzhikov)
• [OPERATOR] Increase limits for metrics-server. This is a temporary fix until we have non-circular auto-scaling for metrics-server. (#4017, @amshuman-kr)
• [OPERATOR] Disable etcd scale down for shoot clusters with purpose "production". This avoids multiple etcd restarts during the shoot's maintenance time window if VPA recommendation for scale down is inappropriate. (#4016, @amshuman-kr)
• [DEVELOPER] The unknown github label area/operations has been removed from the github issues templates. (#4013, @vpnachev)

📰 Noteworthy

• [OPERATOR] The .gardener.garden.identity value (deprecated with v1.11.0, removed with v1.22.0) is added again and will be passed to the Helm chart values of ControllerInstallations. It's still deprecated and planned to be removed in a future version, hence, Gardener operators have to make sure to update affected provider extensions accordingly. (#4021, @rfranzke)

[autoscaler]

🐛 Bug Fixes

• [DEVELOPER] Avoids panics when VM type isn't found during scale from zero (gardener/autoscaler#78, @prashanth26)
• [DEVELOPER] Fetches the VM from the correct map for MCM provider Azure and hence doesn't panic anymore (gardener/autoscaler#78, @prashanth26)

🏃 Others

• [OPERATOR] Bug fix: Allow scaling down of machine with already lowered priority (gardener/autoscaler#75, @prashanth26)

[etcd-backup-restore]

🏃 Others

• [USER] Added CLI flags --auto-compaction-mode and --auto-compaction-retention to configure auto-compaction for embedded etcd. Default values: auto-compaction-mode="periodic" and auto-compaction-retention="30m" (gardener/etcd-backup-restore#315, @ishan16696)
• [OPERATOR] Added a new metric etcdbr_snapshotter_failure used as a blanket metric to capture any snapshotter error. (gardener/etcd-backup-restore#326, @aaronfern)
• [OPERATOR] etcdbr-compression specification can be configured through helm-charts. (gardener/etcd-backup-restore#307, @ishan16696)
• [OPERATOR] Added CLI-flags (max-call-send-message-size, max-request-bytes and max-txn-ops) to enable restoration for delta snapshots with large amount of data (large number of events or events with large data). (gardener/etcd-backup-restore#282, @abdasgupta)

📰 Noteworthy

• [USER] Move bootstrap script out of the helm chart, as it will now be part of the etcd-custom-image. (gardener/etcd-backup-restore#327, @shreyas-s-rao)
• [USER] Add support for snapshot compression/decompression. Compression and compression policy can be configured through flags: --compress-snapshots and --compression-policy respectively. Supported compression policies currently are gzip (default), lzw and zlib. Snapshot compression is disabled by default. (gardener/etcd-backup-restore#293, @ishan16696)

[etcd-druid]

✨ New Features

• [USER] Configure auto-compaction policy for etcd and backup sidecar's embedded etcd via Etcd resource via .spec.sharedConfig.autoCompactionMode and .spec.sharedConfig.autoCompactionRetention. (gardener/etcd-druid#157, @ishan16696)
• [OPERATOR] Etcd bootstrap script now resides in the custom etcd image instead of being mounted as a configmap. (gardener/etcd-druid#162, @shreyas-s-rao)

🏃 Others

• [OPERATOR] If an etcd StatefulSet remains pending, warning events of unbound PVCs are now added to the .status.lastError of the etcd resource. This makes it easier for operators to spot potential issues. (gardener/etcd-druid#146, @timuthy)
• [OPERATOR] Snapshot compression specification can be configured through helm-charts as well as etcd resource spec configuration file. (gardener/etcd-druid#138, @ishan16696)

📰 Noteworthy

• [USER] Moved parts of reconciliation for ETCD status that deals with Statefulset to a separate controller. (gardener/etcd-druid#133, @abdasgupta)

Docker Images

admission-controller: eu.gcr.io/gardener-project/gardener/admission-controller:v1.23.0
apiserver: eu.gcr.io/gardener-project/gardener/apiserver:v1.23.0
controller-manager: eu.gcr.io/gardener-project/gardener/controller-manager:v1.23.0
scheduler: eu.gcr.io/gardener-project/gardener/scheduler:v1.23.0
gardenlet: eu.gcr.io/gardener-project/gardener/gardenlet:v1.23.0
seed-admission-controller: eu.gcr.io/gardener-project/gardener/seed-admission-controller:v1.23.0
landscaper-gardenlet: eu.gcr.io/gardener-project/gardener/landscaper-gardenlet:v1.23.0