vSphere

KKP fully supports VMware vSphere as a tier 1 provider. It enables automatic provisioning of user cluster nodes and storage management by integrating vSphere CCM and vSphere CSI.

The Kubernetes vSphere driver contains bugs related to detaching volumes from offline nodes. See the Volume detach bug section for more details.

VM Images

When creating worker nodes for a user cluster, the user can specify an existing image. Defaults may be set in the seed configuration spec.datacenters.EXAMPLEDC.vsphere.endpoint.

Supported Operating Systems

  • Ubuntu 20.04 ova
  • Ubuntu 22.04 ova
  • Flatcar (Stable channel) ova

Importing the OVA

  1. Go into the VSphere WebUI, select your datacenter, right click onto it and choose “Deploy OVF Template”
  2. Fill in the “URL” field with the appropriate url
  3. Click through the dialog until “Select storage”
  4. Select the same storage you want to use for your machines
  5. Select the same network you want to use for your machines
  6. Leave everything in the “Customize Template” and “Ready to complete” dialog as it is
  7. Wait until the VM got fully imported and the “Snapshots” => “Create Snapshot” button is not grayed out anymore

Importing the QCOW2

  1. Convert it to vmdk: qemu-img convert -f qcow2 -O vmdk Ubuntu-7-x86_64-GenericCloud.qcow2 Ubuntu-7-x86_64-GenericCloud.vmdk
  2. Upload it to a Datastore of your vSphere installation
  3. Create a new virtual machine that uses the uploaded vmdk as rootdisk

Modifications

Modifications like network, disk size, etc. must be done in the ova template before creating a worker node from it. If user clusters have dedicated networks, all user clusters therefore need a custom template.

VM Folder

During creation of a user cluster Kubermatic Kubernetes Platform (KKP) creates a dedicated VM folder in the root path on the Datastore (Defined in the seed cluster spec.datacenters.EXAMPLEDC.vsphere.datastore). That folder will contain all worker nodes of a user cluster.

Credentials / Cloud-Config

KKP utilises provider credentials to create and manage infrastructure on the respective cloud provider. For vSphere, permissions are needed to manage VMs, storage, networking and tags.

The vSphere provider allows to split permissions into two sets of credentials:

  1. Credentials passed to the vSphere Cloud Controller Manager (CCM) and CSI Storage driver. These credentials are currently inherited into the user cluster and should therefore be individual per user cluster. This type of credentials can be passed when creating a user cluster or setting up a preset.
  2. Credentials used for creating and managing infrastructure (VMs, tags, networks). This set of credentials is not shared with the user cluster and is kept on the seed cluster. This type of credentials can either be passed in the Seed configuration (.spec.datacenters.EXAMPLEDC.vpshere.infraManagementUser) for all user clusters created in this datacenter or individually while creating a user cluster.

If such a split is not desired, one set of credentials used for both use cases can be provided instead. Providing two sets of credentials is optional.

Permissions

The vsphere users has to have to following permissions on the correct resources. Note that if a shared set of credentials is used, roles for both use cases need to be assigned to the technical user which will be used for credentials.

Cloud Controller Manager (CCM) / CSI

Note: Below roles were updated based on vsphere-storage-plugin-roles for external CCM which is available from kkp v2.18+ and vsphere v7.0.2+

For the Cloud Controller Manager (CCM) and CSI components used to provide cloud provider and storage integration to the user cluster, a technical user (e.g. cust-ccm-cluster) is needed. The user should be assigned all roles listed below:

Role k8c-ccm-storage-vmfolder-propagate
  • Granted at VM Folder and Template Folder, propagated
  • Permissions
    • Virtual machine
      • Change Configuration
        • Add existing disk
        • Add new disk
        • Add or remove device
        • Remove disk
    • Folder
      • Create folder
      • Delete dolder
$ govc role.ls k8c-ccm-storage-vmfolder-propagate
Folder.Create
Folder.Delete
VirtualMachine.Config.AddExistingDisk
VirtualMachine.Config.AddNewDisk
VirtualMachine.Config.AddRemoveDevice
VirtualMachine.Config.RemoveDisk
Role k8c-ccm-storage-datastore-propagate
  • Granted at Datastore, propagated
  • Permissions
    • Datastore
      • Allocate space
      • Low level file operations
$ govc role.ls k8c-ccm-storage-datastore-propagate
Datastore.AllocateSpace
Datastore.FileManagement
Role Read-only (predefined)
  • Granted at …, not propagated
    • Datacenter
$ govc role.ls ReadOnly
System.Anonymous
System.Read
System.View

Infrastructure Management

For infrastructure (e.g. VMs, tags and networking) provisioning actions of KKP in the scope of a user cluster, the following roles have to be added to the existing user (if a single set of credentials is used) or an additional technical user (e.g. cust-infra-user-cluster) is needed that has the following roles attached:

Role k8c-user-vcenter
  • Granted at vcenter level, not propagated
  • Needed to customize VM during provisioning
  • Permissions
    • CNS
      • Searchable
    • Profile-driven storage
      • Profile-driven storage view
    • VirtualMachine
      • Provisioning
        • Modify customization specification
        • Read customization specifications
    • vSphere Tagging
      • Assign or Unassign vSphere Tag
      • Assign or Unassign vSphere Tag on Object
      • Create vSphere Tag
      • Create vSphere Tag Category
      • Delete vSphere Tag
      • Delete vSphere Tag Category
      • Edit vSphere Tag
      • Edit vSphere Tag Category
      • Modify UsedBy Field For Category
      • Modify UsedBy Field For Tag
$ govc role.ls k8c-user-vcenter
Cns.Searchable
InventoryService.Tagging.AttachTag
InventoryService.Tagging.CreateCategory
InventoryService.Tagging.CreateTag
InventoryService.Tagging.DeleteCategory
InventoryService.Tagging.DeleteTag
InventoryService.Tagging.EditCategory
InventoryService.Tagging.EditTag
InventoryService.Tagging.ModifyUsedByForCategory
InventoryService.Tagging.ModifyUsedByForTag
InventoryService.Tagging.ObjectAttachable
StorageProfile.View
System.Anonymous
System.Read
System.View
VirtualMachine.Provisioning.ModifyCustSpecs
VirtualMachine.Provisioning.ReadCustSpecs
Role k8c-user-datacenter
  • Granted at datacenter level, not propagated
  • Needed for cloning the template VM (obviously this is not done in a folder at this time)
  • Permissions
    • Datastore
      • Allocate space
      • Browse datastore
      • Low level file operations
      • Remove file
    • vApp
      • vApp application configuration
      • vApp instance configuration
    • Virtual Machine
      • Change Configuration
        • Change CPU count
        • Change Memory
        • Change Settings
      • Edit Inventory
        • Create from existing
    • vSphere Tagging
      • Assign or Unassign vSphere Tag on Object
$ govc role.ls k8c-user-datacenter
Datastore.AllocateSpace
Datastore.Browse
Datastore.DeleteFile
Datastore.FileManagement
InventoryService.Tagging.ObjectAttachable
System.Anonymous
System.Read
System.View
VApp.ApplicationConfig
VApp.InstanceConfig
VirtualMachine.Config.CPUCount
VirtualMachine.Config.Memory
VirtualMachine.Config.Settings
VirtualMachine.Inventory.CreateFromExisting
  • Role k8c-user-cluster-propagate
    • Granted at cluster level, propagated
    • Needed for upload of cloud-init.iso (Ubuntu) or defining the Ignition config into Guestinfo (CoreOS)
    • Permissions
      • AutoDeploy
        • Rule
          • Create
          • Delete
          • Edit
      • Folder
        • Create folder
      • Host
        • Configuration
          • Storage partition configuration
          • System Management
        • Local operations
          • Reconfigure virtual machine
        • Inventory
          • Modify cluster
      • Resource
        • Assign virtual machine to resource pool
        • Migrate powered off virtual machine
        • Migrate powered on virtual machine
      • vApp
        • vApp application configuration
        • vApp instance configuration
      • vSphere Tagging
        • Assign or Unassign vSphere Tag on Object
$ govc role.ls k8c-user-cluster-propagate
AutoDeploy.Rule.Create
AutoDeploy.Rule.Delete
AutoDeploy.Rule.Edit
Folder.Create
Host.Config.Storage
Host.Config.SystemManagement
Host.Inventory.EditCluster
Host.Local.ReconfigVM
InventoryService.Tagging.ObjectAttachable
Resource.AssignVMToPool
Resource.ColdMigrate
Resource.HotMigrate
VApp.ApplicationConfig
VApp.InstanceConfig
Role k8c-network-attach
  • Granted for each network that should be used (distributed switch + network)
  • Permissions
    • Network
      • Assign network
    • vSphere Tagging
      • Assign or Unassign vSphere Tag on Object
$ govc role.ls k8c-network-attach
InventoryService.Tagging.ObjectAttachable
Network.Assign
System.Anonymous
System.Read
System.View
Role k8c-user-datastore-propagate
  • Granted at datastore / datastore cluster level, propagated
  • Also provides permission to create vSphere tags for a dedicated category, which are required by KKP seed controller manager
  • Please note below points about tagging.

Note: If a category id is assigned to a user cluster, KKP would claim the ownership of any tags it creates. KKP would try to delete tags assigned to the cluster upon cluster deletion. Thus, make sure that the assigned category isn’t shared across other lingering resources.

Note: Tags can be attached to machine deployments regardless if the tags are created via KKP or not. If a tag was not attached to the user cluster, machine controller will only detach it.

  • Permissions
    • Datastore
      • Allocate space
      • Browse datastore
      • Low level file operations
    • vSphere Tagging
      • Assign or Unassign vSphere Tag on an Object
$ govc role.ls k8c-user-datastore-propagate
Datastore.AllocateSpace
Datastore.Browse
Datastore.FileManagement
InventoryService.Tagging.ObjectAttachable
System.Anonymous
System.Read
System.View
Role k8c-user-folder-propagate
  • Granted at VM Folder and Template Folder level, propagated
  • Needed for managing the node VMs
  • Permissions
    • Folder
      • Create folder
      • Delete folder
    • Global
      • Set custom attribute
    • Virtual machine
      • Change Configuration
      • Edit Inventory
      • Guest operations
      • Interaction
      • Provisioning
      • Snapshot management
    • vSphere Tagging
      • Assign or Unassign vSphere Tag
      • Assign or Unassign vSphere Tag on an Object
      • Create vSphere Tag
      • Delete vSphere Tag
$ govc role.ls k8c-user-folder-propagate
Folder.Create
Folder.Delete
Global.SetCustomField
InventoryService.Tagging.CreateTag
InventoryService.Tagging.DeleteTag
InventoryService.Tagging.AttachTag
InventoryService.Tagging.ObjectAttachable
System.Anonymous
System.Read
System.View
VirtualMachine.Config.AddExistingDisk
VirtualMachine.Config.AddNewDisk
VirtualMachine.Config.AddRemoveDevice
VirtualMachine.Config.AdvancedConfig
VirtualMachine.Config.Annotation
VirtualMachine.Config.CPUCount
VirtualMachine.Config.ChangeTracking
VirtualMachine.Config.DiskExtend
VirtualMachine.Config.DiskLease
VirtualMachine.Config.EditDevice
VirtualMachine.Config.HostUSBDevice
VirtualMachine.Config.ManagedBy
VirtualMachine.Config.Memory
VirtualMachine.Config.MksControl
VirtualMachine.Config.QueryFTCompatibility
VirtualMachine.Config.QueryUnownedFiles
VirtualMachine.Config.RawDevice
VirtualMachine.Config.ReloadFromPath
VirtualMachine.Config.RemoveDisk
VirtualMachine.Config.Rename
VirtualMachine.Config.ResetGuestInfo
VirtualMachine.Config.Resource
VirtualMachine.Config.Settings
VirtualMachine.Config.SwapPlacement
VirtualMachine.Config.ToggleForkParent
VirtualMachine.Config.Unlock
VirtualMachine.Config.UpgradeVirtualHardware
VirtualMachine.GuestOperations.Execute
VirtualMachine.GuestOperations.Modify
VirtualMachine.GuestOperations.ModifyAliases
VirtualMachine.GuestOperations.Query
VirtualMachine.GuestOperations.QueryAliases
VirtualMachine.Interact.AnswerQuestion
VirtualMachine.Interact.Backup
VirtualMachine.Interact.ConsoleInteract
VirtualMachine.Interact.CreateScreenshot
VirtualMachine.Interact.CreateSecondary
VirtualMachine.Interact.DefragmentAllDisks
VirtualMachine.Interact.DeviceConnection
VirtualMachine.Interact.DisableSecondary
VirtualMachine.Interact.DnD
VirtualMachine.Interact.EnableSecondary
VirtualMachine.Interact.GuestControl
VirtualMachine.Interact.MakePrimary
VirtualMachine.Interact.Pause
VirtualMachine.Interact.PowerOff
VirtualMachine.Interact.PowerOn
VirtualMachine.Interact.PutUsbScanCodes
VirtualMachine.Interact.Record
VirtualMachine.Interact.Replay
VirtualMachine.Interact.Reset
VirtualMachine.Interact.SESparseMaintenance
VirtualMachine.Interact.SetCDMedia
VirtualMachine.Interact.SetFloppyMedia
VirtualMachine.Interact.Suspend
VirtualMachine.Interact.TerminateFaultTolerantVM
VirtualMachine.Interact.ToolsInstall
VirtualMachine.Interact.TurnOffFaultTolerance
VirtualMachine.Inventory.Create
VirtualMachine.Inventory.CreateFromExisting
VirtualMachine.Inventory.Delete
VirtualMachine.Inventory.Move
VirtualMachine.Inventory.Register
VirtualMachine.Inventory.Unregister
VirtualMachine.Provisioning.Clone
VirtualMachine.Provisioning.CloneTemplate
VirtualMachine.Provisioning.CreateTemplateFromVM
VirtualMachine.Provisioning.Customize
VirtualMachine.Provisioning.DeployTemplate
VirtualMachine.Provisioning.DiskRandomAccess
VirtualMachine.Provisioning.DiskRandomRead
VirtualMachine.Provisioning.FileRandomAccess
VirtualMachine.Provisioning.GetVmFiles
VirtualMachine.Provisioning.MarkAsTemplate
VirtualMachine.Provisioning.MarkAsVM
VirtualMachine.Provisioning.ModifyCustSpecs
VirtualMachine.Provisioning.PromoteDisks
VirtualMachine.Provisioning.PutVmFiles
VirtualMachine.Provisioning.ReadCustSpecs
VirtualMachine.State.CreateSnapshot
VirtualMachine.State.RemoveSnapshot
VirtualMachine.State.RenameSnapshot
VirtualMachine.State.RevertToSnapshot

The described permissions have been tested with vSphere 8.0.2 and might be different for other vSphere versions.

Datastores and Datastore Clusters

Datastores in vSphere are an abstraction for storage. A Datastore Cluster is a collection of datastores with shared resources and a shared management interface.

In KKP Datastores are used for two purposes:

  • Storing the VMs files for the worker nodes of vSphere user clusters.
  • Generating the vSphere cloud provider storage configuration for user clusters. In particular to provide the default-datastore value, that is the default datastore for dynamic volume provisioning.

Datastore Clusters can only be used for the first purpose as it cannot be specified directly in vSphere cloud configuration.

There are three places where Datastores and Datastore Clusters can be configured in KKP:

  • At datacenter level (configured in the Seed CRD)) it is possible to specify the default Datastore that will be used for user clusters dynamic volume provisioning and workers VMs placement in case no Datastore or Datastore Cluster is specified at cluster level.
  • At Cluster level it is possible to provide either a Datastore or a Datastore Cluster respectively with spec.cloud.vsphere.datastore and spec.cloud.vsphere.datastoreCluster fields.
  • It is possible to specify Datastore or Datastore Clusters in a preset than is later used to create a user cluster from it.

These settings can also be configured as part of the “Advanced Settings” step when creating a user cluster from the KKP dashboard.

Known Issues

Volume Detach Bug

After a node is powered-off, the Kubernetes vSphere driver doesn’t detach disks associated with PVCs mounted on that node. This makes it impossible to reschedule pods using these PVCs until the disks are manually detached in vCenter.

Upstream Kubernetes has been working on the issue for a long time now and tracking it under the following tickets:

Internal Kubernetes endpoints unreachable

Symptoms

  • Unable to perform CRUD operations on resources governed by webhooks (e.g. ValidatingWebhookConfiguration, MutatingWebhookConfiguration, etc.). The following error is observed:
Internal error occurred: failed calling webhook "webhook-name": failed to call webhook: Post "https://webhook-service-name.namespace.svc:443/webhook-endpoint": context deadline exceeded
  • Unable to reach internal Kubernetes endpoints from pods/nodes.
  • ICMP is working but TCP/UDP is not.

Cause

On recent enough VMware hardware compatibility version (i.e >=15 or maybe >=14), CNI connectivity breaks because of hardware segmentation offload. cilium-health status has ICMP connectivity working, but not TCP connectivity. cilium-health status may also fail completely.

Solution

sudo ethtool -K ens192 tx-udp_tnl-segmentation off
sudo ethtool -K ens192 tx-udp_tnl-csum-segmentation off

These flags are related to the hardware segmentation offload done by the vSphere driver VMXNET3. We have observed this issue for both Cilium and Canal CNI running on Ubuntu 22.04.

We have two options to configure these flags for KKP installations:

References