If you are installing Tanzu CLI using the artifacts published as part of this release, please follow the instructions.
-
AWS classy clusters now populate os-name information from config file and for all possible machineDeployments. (#2839, @tenczar)
-
Add AWS EBS CSI Driver and AzureDisk CSI Driver (#2875, @jeffwubj)
-
Add ClusterClass based cluster creation validation check at runtime based on feature-flag and feature-gate (#2804, @anujc25)
-
Add Object propagation controller to the tkg-clusterclass package (#2749, @anujc25)
-
Add node role selector to kapp controller (#2864, @maralavi)
-
Addons-manager "--cluster-delete-timeout" is exposed to templates as tanzuAddonsManager.deployment.clusterDeleteTimeout (#2996, @adduarte)
-
Adds the following variables to support PCI Passthrough on TKG on vSphere without Supervisor:
VSPHERE_WORKER_PCI_DEVICES, if set will configure PCI passthrough on all worker machines
VSPHERE_WORKER_PCI_DEVICES must be specified in the format <vendor_id>:<device_id>
e.g. VSPHERE_WORKER_PCI_DEVICES: "10DE:1EB8"
VSPHERE_CONTROL_PLANE_PCI_DEVICES, if set will configure PCI passthrough on all control plane machines
VSPHERE_CONTROL_PLANE_PCI_DEVICES must be specified in the format <vendor_id>:<device_id>
e.g. VSPHERE_CONTROL_PLANE_PCI_DEVICES: "10DE:1EB8"
VSPHERE_IGNORE_PCI_DEVICES_ALLOW_LIST, if set to true will allow other PCI devices to be passed through
other than NVIDIA T4 (i.e. "10DE:1EB8")
VSPHERE_CONTROL_PLANE_CUSTOM_VMX_KEYS, if set will set custom VMX keys on all control plane machines
Keys must be set in the form =,=
WORKER_ROLLOUT_STRATEGY configures the MachineDeployment rollout strategy. If set to OnDelete,
on updates, the existing worker machines will be deleted first before the replacement worker machines are created. (#2948, @randomvariable)
-
Bug fix for regression caused by https://github.com/vmware-tanzu/tanzu-framework/pull/2926/ (#2974, @vijaykatam)
-
Bump Cluster API and providers (CAPI/CAPD to 1.1.5, CAPV provider to 1.3.1, CAPZ provider to 1.4.0) (#2951, @ykakarap)
-
Bump cert-manager to v1.7.2. (#2490, @ridavid2002)
-
Bump kapp-controller to 0.38.3 (#2719, @ridavid2002)
-
Bump kapp-controller to version 0.38.4 (#2880, @shivaani0505)
-
Bumping Antrea component to version 1.5.3 (#2792, @knabben)
-
CAPI controller deployed during management cluster creation will activate ClusterTopology featuregate only if needed (#2994, @vuil)
-
Change default AVI_CONTROLLER_VERSION to 20.1.3. (#2931, @HanFa)
-
Enable leader election for Addons Controller (#2787, @HL-EverGreen)
-
Featuregates controller reconciles on Namespace resource changes. (#2984, @sathyanarays)
-
Fix incorrect namespace used with IsClusterClassBased during management cluster upgrade (#2859, @12345lcr)
-
Fixed object-propagation-controller copying cluster owned MachineTemplate
and KubeadmConfigTemplate
resources to other namespaces (#3007, @anujc25)
-
Fixes bug where passing a ccluster input file to create dry-run would error expecting a legacy config file. (#2899, @tenczar)
-
Kapp-controller .sh file update (#2901, @shivaani0505)
-
Legacy-style Clusters (with tanzuKubernetesRelease
label) now have UpdatesAvailable status condition. If the condition is True, its message looks similar to the one in "classy" Clusters: a space-separated TKR version list, looking like this: [v1.21.3+vmware.1-tkg.3 v1.22.0+vmware.1-tkg.1 v1.22.0+vmware.1-tkg.2]
(#2997, @imikushin)
-
Modifying machinehealthchecks on classy clusters is not supported. The API will return an error when performing these operations on classy clusters. (#2693, @tenczar)
-
Package-based-lcm: change cluster pause webhook behavior to not pause for legacy TKGS tkr's (#2935, @maralavi)
-
Package-based-lcm: set maxUnavailable to 1 in addons-manager deployment (#2962, @ggpaue)
-
Package-based-lcm: Add AntreaNSXPodRoutingEnabled to CPI CRD for uTKG routablePod (#2929, @lubronzhan)
-
Package-based-lcm: Fix stuck addons finalizer on deleted clusters (#2848, @vijaykatam)
-
Package-based-lcm: Verify addons installation and deletion as part of tkg clusterclass tests (#2913, @shyaamsn)
-
Package-based-lcm: add common functions to test addon packages (#2743, @ggpaue)
-
Package-based-lcm: fix memory leaks in packageinstallstatus_controller (#3004, @vijaykatam)
-
Package-based-lcm: updates node pool API to properly support all node pool customization values, e.g vmClass, storageClass (#2900, @tenczar)
-
Package-based-lcm: updates node pool API to support clusterclass based clusters doc (#2445, @tenczar)
-
Package-based-lcm: when feature flag is set, legacy style cluster creation should still work (#2898, @vuil)
-
Remove feature-gate validation for TKGS cluster and make it based on vSphere version (#2986, @anujc25)
-
Show core and additional package install statuses in clusterbootstrap status message field. (#2781, @codegold79)
-
Show plugin version as installed version and status based on discovered plugin version (#2903, @anujc25)
-
Support air-gapped scenario for context-scoped plugins to allow registry override (#2937, @anujc25)
-
Support configuring multiple default trusted registry for plugin downloads during build-time using TANZU_PLUGINS_ALLOWED_IMAGE_REPOSITORIES
(#2977, @anujc25)
-
The TKR Resolver Cluster webhook now honors the custom imageRepository
cluster variable:
if present, its host
field value is written into the resolved TKR_DATA values'
kubernetesSpec imageRepository fields. (#3013, @imikushin)
-
Updates scale API to support clusterclass based clusters. (#2999, @tenczar)
-
Use fslock to acquire lock when reading/writing the tanzu config file for update (#2882, @anujc25)
-
When cluster's CNI implementation is Antrea, AKO will use NodePortLocal routing mode by default (#2938, @XudongLiuHarold)