website/docs/getting-started-guides/vsphere.md

11 KiB

approvers title
erictune
jbeda
VMware vSphere

This page covers how to get started with deploying Kubernetes on vSphere and details for how to configure the vSphere Cloud Provider.

  • TOC {:toc}

Getting started with the vSphere Cloud Provider

Kubernetes comes with vSphere Cloud Provider, a cloud provider for vSphere that allows Kubernetes Pods to use vSphere Storage.

Deploy Kubernetes on vSphere

To deploy Kubernetes on vSphere and use the vSphere Cloud Provider, see Kubernetes-Anywhere.

Detailed steps can be found at the getting started with Kubernetes-Anywhere on vSphere page.

vSphere Cloud Provider

vSphere Cloud Provider allows Kubernetes to use vSphere-managed storage. It supports:

  • Services such as de-duplication and encryption with vSAN, QoS, high availability and data reliability.
  • Policy based management at granularity of container volumes.
  • Volumes, Persistent Volumes, Storage Classes, dynamic provisioning of volumes, and scalable deployment of Stateful Apps with StatefulSets.

For more detail visit vSphere Storage for Kubernetes Documentation.

Documentation for how to use vSphere managed storage can be found in the persistent volumes user guide and the volumes user guide.

Examples can be found here.

Enable vSphere Cloud Provider

If a Kubernetes cluster has not been deployed using Kubernetes-Anywhere, follow the instructions below to enable the vSphere Cloud Provider. These steps are not needed when using Kubernetes-Anywhere, they will be done as part of the deployment.

Step-1 Create a VM folder and move Kubernetes Node VMs to this folder.

Step-2 Make sure Node VM names must comply with the regex [a-z](([-0-9a-z]+)?[0-9a-z])?(\.[a-z0-9](([-0-9a-z]+)?[0-9a-z])?)*. If Node VMs do not comply with this regex, rename them and make it compliant to this regex.

Node VM names constraints:

  • VM names can not begin with numbers.
  • VM names can not have capital letters, any special characters except . and -.
  • VM names can not be shorter than 3 chars and longer than 63.

Step-3 Enable disk UUID on Node virtual machines.

The disk.EnableUUID parameter must be set to "TRUE" for each Node VM. This step is necessary so that the VMDK always presents a consistent UUID to the VM, thus allowing the disk to be mounted properly.

For each of the virtual machine nodes that will be participating in the cluster, follow the steps below using govc tool

  • Set up the govc environment

    export GOVC_URL='vCenter IP OR FQDN'
    export GOVC_USERNAME='vCenter User'
    export GOVC_PASSWORD='vCenter Password'
    export GOVC_INSECURE=1
    
  • Find Node VM Paths

    govc ls /datacenter/vm/<vm-folder-name>
    
  • Set disk.EnableUUID to true for all VMs

    govc vm.change -e="disk.enableUUID=1" -vm='VM Path'
    

Note: If Kubernetes Node VMs are created from template VM then disk.EnableUUID=1 can be set on the template VM. VMs cloned from this template, will automatically inherit this property.

Step-4 Create and assign Roles to the vSphere Cloud Provider user and vSphere entities.

Note: if you want to use Administrator account then this step can be skipped.

vSphere Cloud Provider requires the following minimal set of privileges to interact with vCenter. Please refer vSphere Documentation Center to know about steps for creating a Custom Role, User and Role Assignment.

Roles Privileges Entities Propagate to Children
manage-k8s-node-vms Resource.AssignVMToPool
System.Anonymous
System.Read
System.View
VirtualMachine.Config.AddExistingDisk
VirtualMachine.Config.AddNewDisk
VirtualMachine.Config.AddRemoveDevice
VirtualMachine.Config.RemoveDisk
VirtualMachine.Inventory.Create
VirtualMachine.Inventory.Delete
Cluster,
Hosts,
VM Folder
Yes
manage-k8s-volumes Datastore.AllocateSpace
Datastore.FileManagement
System.Anonymous
System.Read
System.View
Datastore No
k8s-system-read-and-spbm-profile-view StorageProfile.View
System.Anonymous
System.Read
System.View
vCenter No
ReadOnly System.Anonymous
System.Read
System.View
Datacenter,
Datastore Cluster,
Datastore Storage Folder
No

Step-5 Create the vSphere cloud config file (vsphere.conf). Cloud config template can be found here.

This config file needs to be placed in the shared directory which should be accessible from kubelet container, controller-manager pod, and API server pod.

vsphere.conf for Master Node:

[Global]
        user = "vCenter username for cloud provider"
        password = "password"
        server = "IP/FQDN for vCenter"
        port = "443" #Optional
        insecure-flag = "1" #set to 1 if the vCenter uses a self-signed cert
        datacenter = "Datacenter name" 
        datastore = "Datastore name" #Datastore to use for provisioning volumes using storage classes/dynamic provisioning
        working-dir = "vCenter VM folder path in which node VMs are located"
        vm-name = "VM name of the Master Node" #Optional
        vm-uuid = "UUID of the Node VM" # Optional        
[Disk]
    scsicontrollertype = pvscsi

Note: vm-name parameter is introduced in 1.6.4 release. Both vm-uuid and vm-name are optional parameters. If vm-name is specified then vm-uuid is not used. If both are not specified then kubelet will get vm-uuid from /sys/class/dmi/id/product_serial and query vCenter to find the Node VM's name.

vsphere.conf for Worker Nodes: (Only Applicable to 1.6.4 release and above. For older releases this file should have all the parameters specified in Master node's vSphere.conf file).

[Global]
        vm-name = "VM name of the Worker Node"

Below is summary of supported parameters in the vsphere.conf file

  • user is the vCenter username for vSphere Cloud Provider.

  • password is the password for vCenter user specified with user.

  • server is the vCenter Server IP or FQDN

  • port is the vCenter Server Port. Default is 443 if not specified.

  • insecure-flag is set to 1 if vCenter used a self-signed certificate.

  • datacenter is the name of the datacenter on which Node VMs are deployed.

  • datastore is the default datastore to use for provisioning volumes using storage classes/dynamic provisioning.

  • vm-name is recently added configuration parameter. This is optional parameter. When this parameter is present, vsphere.conf file on the worker node does not need vCenter credentials.

    Note: vm-name is added in the release 1.6.4. Prior releases does not support this parameter.

  • working-dir can be set to empty ( working-dir = ""), if Node VMs are located in the root VM folder.

  • vm-uuid is the VM Instance UUID of virtual machine. vm-uuid can be set to empty (vm-uuid = ""). If set to empty, this will be retrieved from /sys/class/dmi/id/product_serial file on virtual machine (requires root access).

    • vm-uuid needs to be set in this format - 423D7ADC-F7A9-F629-8454-CE9615C810F1

    • vm-uuid can be retrieved from Node Virtual machines using following command. This will be different on each node VM.

      cat /sys/class/dmi/id/product_serial | sed -e 's/^VMware-//' -e 's/-/ /' | awk '{ print toupper($1$2$3$4 "-" $5$6 "-" $7$8 "-" $9$10 "-" $11$12$13$14$15$16) }'
      
  • datastore is the default datastore used for provisioning volumes using storage classes. If datastore is located in storage folder or datastore is member of datastore cluster, make sure to specify full datastore path. Make sure vSphere Cloud Provider user has Read Privilege set on the datastore cluster or storage folder to be able to find datastore.

    • For datastore located in the datastore cluster, specify datastore as mentioned below

      datastore = "DatastoreCluster/datastore1"
      
    • For datastore located in the storage folder, specify datastore as mentioned below

      datastore = "DatastoreStorageFolder/datastore1"
      

Step-6 Add flags to controller-manager, API server and Kubelet to enable vSphere Cloud Provider.

  • Add following flags to kubelet running on every node and to the controller-manager and API server pods manifest files.
--cloud-provider=vsphere
--cloud-config=<Path of the vsphere.conf file>

Manifest files for API server and controller-manager are generally located at /etc/kubernetes/manifests.

Step-7 Restart Kubelet on all nodes.

  • Reload kubelet systemd unit file using systemctl daemon-reload
  • Restart kubelet service using systemctl restart kubelet.service

Note: After enabling the vSphere Cloud Provider, Node names will be set to the VM names from the vCenter Inventory.

Known issues

Please visit known issues for the list of major known issues with Kubernetes vSphere Cloud Provider.

Support Level

For quick support please join VMware Code Slack (kubernetes) and post your question.

IaaS Provider Config. Mgmt OS Networking Docs Conforms Support Level
Vmware vSphere Kube-anywhere Photon OS Flannel docs Community (@abrarshivani), (@kerneltime), (@BaluDontu), (@luomiao), (@divyenpatel)

If you identify any issues/problems using the vSphere cloud provider, you can create an issue in our repo - VMware Kubernetes.

For support level information on all solutions, see the Table of solutions chart.