Skip to content

lamw/vsphere-with-tanzu-nsxt-automated-lab-deployment

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

37 Commits
 
 
 
 
 
 

Repository files navigation

vSphere with Tanzu using NSX-T Automated Lab Deployment

Table of Contents

Description

Similar to other "VMware Automated Lab Deployment Scripts" (such as here, here and here), this script makes it very easy for anyone with VMware Cloud Foundation 4 (for vSphere 7.0 deployments) or VMware Tanzu (for vSphere 7.0U1 deployments) licensing to deploy vSphere with Kubernetes/Tanzu in a Nested Lab environment for learning and educational purposes. All required VMware components (ESXi, vCenter Server, NSX Unified Appliance and Edge) are automatically deployed and configured to allow enablement of vSphere with Kubernetes. For more details about vSphere with Kubernetes, please refer to the official VMware documentation here.

Below is a diagram of what is deployed as part of the solution and you simply need to have an existing vSphere environment running that is managed by vCenter Server and with enough resources (CPU, Memory and Storage) to deploy this "Nested" lab. For a complete end-to-end example including workload management enablement (post-deployment operation) and the deployment of a Tanzu Kubernetes Grid (TKG) Cluster, please have a look at the Sample Execution section below.

You are now ready to get your K8s on! 😁

Changelog

  • 11/07/2023

  • Updated for vSphere 8.0 and NSX 4.1.1 due to API changes since vSphere 7 and NSX 3

  • Added a few checks to allow reuse of existing objects like vCenter VDS, VDPortGroup, StoragePolicy, Tag and TagCategory, NSX TransportNodeProfile.

  • Added FAQ to create multiple Clusters, and using the same VDS/VDPortGroup, This allow Multi Kubernetes Cluster High-Availability with vSphere Zone and Workload Enablement. Please see this blog post

  • Added a few pause in the usecase where we deploy only a new cluster to allow Nested ESXi to boot and fully come online (180s) and before VSAN Diskgroup creation (30s).

  • Added FTT configuration for VSAN allowing 0 redundancy and to use only one node demo lab VSAN Cluster.

  • $hostFailuresToTolerate = 0

  • Added pause to the script to workaround blog post without babysitting for AMD Zen DPDK FastPath capable owner CPU.

  • $NSXTEdgeAmdZenPause = 0

  • Added -DownloadContentOnDemand option in TKG Content Library to prevent the download in advance of 250GB for each cluster and reduce to a few GB.

  • Added T0 VRF Gateway Automated Creation with Static route like the Parent T0 (Note: an uplink segment '$NetworkSegmentProjectVRF' is connected to parent T0)

  • Added Project and VPC Automated Creation with there respective IP Blocks/Subnets.

  • FAQ Updated for the recent changes

  • Download links updated

  • 02/09/2023

    • Allow additional NSX-T Edge nodes
  • 02/03/2023

    • Fix issue #29
  • 03/08/2021

    • Changes to better support vSphere 7.0 Update 1 & NSX-T 3.1.x
    • Added TKG Content Library
    • Minor misc. revisions
  • 02/21/2021

    • Verified support for vSphere 7.0 Update 1 & NSX-T 3.1
    • Fix T0 Interface creation due to API changes with NSX-T
  • 04/27/2020

    • Enable minimum vSphere with K8s Deployment. Please see this blog post for more details.
  • 04/13/2020

    • Initial Release

Requirements

  • vCenter Server running at least vSphere 6.7 or later

    • If your physical storage is vSAN, please ensure you've applied the following setting as mentioned here
  • Resource Requirements

    • Compute

      • Ability to provision VMs with up to 8 vCPU
      • Ability to provision up to 116-140 GB of memory
      • DRS-enabled Cluster (not required but vApp creation will not be possible)
    • Network

      • Single Standard or Distributed Portgroup (Native VLAN) used to deploy all VMs
        • 6 x IP Addresses for VCSA, ESXi, NSX-T UA and Edge VM
        • 5 x Consecutive IP Addresses for Kubernetes Control Plane VMs
        • 1 x IP Address for T0 Static Route
        • 32 x IP Addresses (/27) for Egress CIDR range is the minimum (must not overlap with Ingress CIDR)
        • 32 x IP Addresses (/27) for Ingress CIDR range is the minimum (must not overlap with Egress CIDR)
        • All IP Addresses should be able to communicate with each other
    • Storage

      • Ability to provision up to 1TB of storage

      Note: For detailed requirements, plesae refer to the official document here

  • VMware Cloud Foundation Licenses

  • Desktop (Windows, Mac or Linux) with latest PowerShell Core and PowerCLI 12.0 Core installed. See instructions here for more details

  • vSphere 7 & NSX-T OVAs:

FAQ

  1. What if I do not have a VMware Cloud Foundation 4 License?

  2. Can I reduce the default CPU, Memory and Storage resources?

    • You can, but it is highly recommended to leave the current defaults for the best working experience. For non-vSphere with Kubernetes usage, you can certainly tune down the resources. For vSphere Pod usage, it is possible to deploy the NSX-T Edge with just 4 vCPU, however if you are going to deploy TKG Clusters, you will need 8 vCPUs on the NSX-T Edge for proper functionality. For memory resources, you can reduce the ESXi VM memory to 16GB but if you intend to deploy K8s application/workloads, you will want to keep the default. For NSX-T memory, I have seen cases where system will become unresponsive and although you can probably tune it down a bit more, I would strongly suggest you keep the defaults unless you plan to do exhaustive testing to ensure there is no negative impact.

    UPDATE (04/27/20): Please see this blog post for more details.

  3. Can I just deploy vSphere (VCSA, ESXi) and vSAN without NSX-T and vSphere with Kubernetes?

    • Yes, simply search for the following variables and change their values to 0 to not deploy NSX-T components or run through the configurations

      $setupTanzuStoragePolicy = 0
      $deployNSXManager = 0
      $deployNSXEdge = 0
      $postDeployNSXConfig = 0
      $setupTanzu = 0
      
  4. Can I just deploy vSphere (VCSA, ESXi), vSAN and NSX-T but not configure it for vSphere with Kubernetes?

    • Yes, but some of the NSX-T automation will contain some configurations related to vSphere with Kubernetes. It does not affect the usage of NSX-T, so you can simply ignore or just delete those settings. Search for the following variables and change their values to 0 to not apply the vSphere with Kubernetes configurations

      $setupTanzu = 0
      
  5. Can i just deploy additional vSphere Cluster?

    • Yes, simply changes following variables: new $NestedESXiHostnameToIPs, customize like $NewVCVSANClusterName = "Workload-Cluster-1" and $vsanDatastoreName = "vsanDatastore-1"

      $VAppName = "Nested-vSphere-with-Tanzu-NSX-T-Lab-qnateilb" # "Nested-vSphere-with-Tanzu-NSX-T-Lab-$random_string" reuse the $VAppName for 2nd and 3rd cluster deployments
      
       $deployVCSA = 0
       $deployNSXManager = 0
       $deployNSXEdge = 0
      
          $runHealth=$true
          $runCEIP=$false
          $runAddVC=$false
          $runIPPool=$false
          $runTransportZone=$false
          $runUplinkProfile=$false
          $runTransportNodeProfile=$true
          $runAddEsxiTransportNode=$true
          $runAddEdgeTransportNode=$false
          $runAddEdgeCluster=$false
          $runNetworkSegment=$false
          $runT0Gateway=$false
          $runT0StaticRoute=$false
          $registervCenterOIDC=$false
      
  6. Can the script setup T0 VRF Gateway?

    • Yes, simply
      # T0 VRF Gateway
      $VRF = "0" #any integer from 0-4094 
      $NetworkSegmentProjectVRF = "$ProjectName-$VRF"
      $NetworkSegmentVlanProjectVRF = "$VRF" #any integer from 0-4094 or the variable $VRF if it's an integer
      $NetworkSegmentProjectVRFSubnetGw = "192.168.2.177/28"
      
      $T0GatewayVRFName = "T0-$ProjectName-$VRF-Gateway"
      $T0GatewayVRFInterfaceAddress = "192.168.2.190" # should be a routable address on the vrf's vlan
      $T0GatewayVRFInterfacePrefix = "28"
      $T0GatewayVRFInterfaceStaticRouteName = "T0-$ProjectName-$VRF-Static-Route"
      $T0GatewayVRFInterfaceStaticRouteNetwork = "0.0.0.0/0"
      $T0GatewayVRFInterfaceStaticRouteAddress = "192.168.2.177"
      
      # Which T0 to use for the Project External connectivity : $T0GatewayName or $T0GatewayVRFName
      $ProjectT0 = $T0GatewayVRFName
      
  7. Can the script setup NSX 4.x Projects?

    • Yes, simply
      # Project ,Public Ip Block, Private Ip Block
      $ProjectName = "Project-160"
      $ShortId = "PRJ160"
      $ProjectPUBipblockName = "VRF-160-192-168-1-160-27"
      $ProjectPUBcidr = "192.168.1.160/27" # Maximum 5 Public Ip Block per Project
      $VpcProjectPRIVipblockName = "VRF-160-10-10-160-0-23"
      $VpcProjectPRIVcidr = "10.10.160.0/23"
      
  8. Can the script setup NSX 4.1.1 VPC?

    • Yes, simply
      # VPC, Public Subnet, Private Subnet
      $VpcName = "VPC-160"
      $VpcPublicSubnet = "192-168-1-176-28"
      $VpcPublicSubnetIpaddresses = "192.168.1.176/28" # Must be subset of Project Public cidr, and can't use the first or last subnet block size !
      $VpcPublicSubnetSize = 16 # Minimum 16
      $VpcPrivateSubnet = "10-10-160-0-24"
      $VpcPrivateSubnetIpaddresses = "10.10.160.0/24" # Must be subset of Project Private cidr
      $VpcPrivateSubnetSize = 256 # Minimum 16
      
  9. Can the script be pause to workaround (see Changelog) for AMD Zen before NSX Edge is used?

    • Yes, simply set the following variable to 1, apply the workaround, wait a few minutes for the NSX Edge to reboot, then hit anyKey for the script to continue.
      $NSXTEdgeAmdZenPause = 1
      
  10. Can i deploy Multiple Zone Kubernetes Cluster High Availability

    • Yes, simply deploy 3 cluster with a single VDSwitch, then see this blog post and refer to the official VMware documentation here
  11. Can the script deploy two NSX-T Edges?

    • Yes, simply append to the configuration to include the additional Edge Hostname IP which will be brought into the Edge Cluster during configuration. The limit 10 Edge Nodes Per Cluster
  12. How do I enable vSphere with Kubernetes after the script has completed?

    • Please refer to the official VMware documentation here with the instructions
  13. How do I troubleshoot enabling or consuming vSphere with Kubernetes?

  14. Is there a way to automate the enablement of Workload Management to a vSphere Cluster?

Configuration

Before you can run the script, you will need to edit the script and update a number of variables to match your deployment environment. Details on each section is described below including actual values used in my home lab environment.

This section describes the credentials to your physical vCenter Server in which the Tanzu lab environment will be deployed to:

$VIServer = "mgmt-vcsa-01.cpbu.corp"
$VIUsername = "[email protected]"
$VIPassword = "VMware1!"

This section describes the location of the files required for deployment.

$NestedESXiApplianceOVA = "C:\Users\william\Desktop\Tanzu\Nested_ESXi7.0_Appliance_Template_v1.ova"
$VCSAInstallerPath = "C:\Users\william\Desktop\Tanzu\VMware-VCSA-all-7.0.0-15952498"
$NSXTManagerOVA = "C:\Users\william\Desktop\Tanzu\nsx-unified-appliance-3.0.0.0.0.15946738.ova"
$NSXTEdgeOVA = "C:\Users\william\Desktop\Tanzu\nsx-edge-3.0.0.0.0.15946738.ova"

Note: The path to the VCSA Installer must be the extracted contents of the ISO

This section defines the number of Nested ESXi VMs to deploy along with their associated IP Address(s). The names are merely the display name of the VMs when deployed. At a minimum, you should deploy at least three hosts, but you can always add additional hosts and the script will automatically take care of provisioning them correctly.

$NestedESXiHostnameToIPs = @{
    "tanzu-esxi-7" = "172.17.31.113"
    "tanzu-esxi-8" = "172.17.31.114"
    "tanzu-esxi-9" = "172.17.31.115"
}

This section describes the resources allocated to each of the Nested ESXi VM(s). Depending on your usage, you may need to increase the resources. For Memory and Disk configuration, the unit is in GB.

$NestedESXivCPU = "4"
$NestedESXivMEM = "24" #GB
$NestedESXiCachingvDisk = "8" #GB
$NestedESXiCapacityvDisk = "100" #GB

This section describes the VCSA deployment configuration such as the VCSA deployment size, Networking & SSO configurations. If you have ever used the VCSA CLI Installer, these options should look familiar.

$VCSADeploymentSize = "tiny"
$VCSADisplayName = "tanzu-vcsa-3"
$VCSAIPAddress = "172.17.31.112"
$VCSAHostname = "tanzu-vcsa-3.cpbu.corp" #Change to IP if you don't have valid DNS
$VCSAPrefix = "24"
$VCSASSODomainName = "vsphere.local"
$VCSASSOPassword = "VMware1!"
$VCSARootPassword = "VMware1!"
$VCSASSHEnable = "true"

This section describes the location as well as the generic networking settings applied to Nested ESXi VCSA & NSX VMs

$VMDatacenter = "San Jose"
$VMCluster = "Cluster-01"
$VMNetwork = "SJC-CORP-MGMT"
$VMDatastore = "vsanDatastore"
$VMNetmask = "255.255.255.0"
$VMGateway = "172.17.31.253"
$VMDNS = "172.17.31.5"
$VMNTP = "pool.ntp.org"
$VMPassword = "VMware1!"
$VMDomain = "cpbu.corp"
$VMSyslog = "172.17.31.112"
$VMFolder = "Tanzu"
# Applicable to Nested ESXi only
$VMSSH = "true"
$VMVMFS = "false"

This section describes the configuration of the new vCenter Server from the deployed VCSA. Default values are sufficient.

$NewVCDatacenterName = "Tanzu-Datacenter"
$NewVCVSANClusterName = "Workload-Cluster"
$NewVCVDSName = "Tanzu-VDS"
$NewVCDVPGName = "DVPG-Management Network"

This section describes the Tanzu Configurations. Default values are sufficient.

# Tanzu Configuration
$StoragePolicyName = "tanzu-gold-storage-policy"
$StoragePolicyTagCategory = "tanzu-demo-tag-category"
$StoragePolicyTagName = "tanzu-demo-storage"
$DevOpsUsername = "devops"
$DevOpsPassword = "VMware1!"

This section describes the NSX-T configurations, the defaults values are sufficient with for the following variables which ust be defined by users and the rest can be left as defaults. $NSXLicenseKey, $NSXVTEPNetwork, $T0GatewayInterfaceAddress, $T0GatewayInterfaceStaticRouteAddress and the NSX-T Manager and Edge Sections

# NSX-T Configuration
$NSXLicenseKey = "NSX-LICENSE-KEY"
$NSXRootPassword = "VMware1!VMware1!"
$NSXAdminUsername = "admin"
$NSXAdminPassword = "VMware1!VMware1!"
$NSXAuditUsername = "audit"
$NSXAuditPassword = "VMware1!VMware1!"
$NSXSSHEnable = "true"
$NSXEnableRootLogin = "true"
$NSXVTEPNetwork = "Tanzu-VTEP" # This portgroup needs be created before running script

# Transport Node Profile
$TransportNodeProfileName = "Tanzu-Host-Transport-Node-Profile"

# Transport Zones
$TunnelEndpointName = "TEP-IP-Pool"
$TunnelEndpointDescription = "Tunnel Endpoint for Transport Nodes"
$TunnelEndpointIPRangeStart = "172.30.1.10"
$TunnelEndpointIPRangeEnd = "172.30.1.20"
$TunnelEndpointCIDR = "172.30.1.0/24"
$TunnelEndpointGateway = "172.30.1.1"

$OverlayTransportZoneName = "TZ-Overlay"
$OverlayTransportZoneHostSwitchName = "nsxswitch"
$VlanTransportZoneName = "TZ-VLAN"
$VlanTransportZoneNameHostSwitchName = "edgeswitch"

# Network Segment
$NetworkSegmentName = "Tanzu-Segment"
$NetworkSegmentVlan = "0"

# T0 Gateway
$T0GatewayName = "Tanzu-T0-Gateway"
$T0GatewayInterfaceAddress = "172.17.31.119" # should be a routable address
$T0GatewayInterfacePrefix = "24"
$T0GatewayInterfaceStaticRouteName = "Tanzu-Static-Route"
$T0GatewayInterfaceStaticRouteNetwork = "0.0.0.0/0"
$T0GatewayInterfaceStaticRouteAddress = "172.17.31.253"

# Uplink Profiles
$ESXiUplinkProfileName = "ESXi-Host-Uplink-Profile"
$ESXiUplinkProfilePolicy = "FAILOVER_ORDER"
$ESXiUplinkName = "uplink1"

$EdgeUplinkProfileName = "Edge-Uplink-Profile"
$EdgeUplinkProfilePolicy = "FAILOVER_ORDER"
$EdgeOverlayUplinkName = "uplink1"
$EdgeOverlayUplinkProfileActivepNIC = "fp-eth1"
$EdgeUplinkName = "tep-uplink"
$EdgeUplinkProfileActivepNIC = "fp-eth2"
$EdgeUplinkProfileTransportVLAN = "0"
$EdgeUplinkProfileMTU = "1600"

# Edge Cluster
$EdgeClusterName = "Edge-Cluster-01"

# NSX-T Manager Configurations
$NSXTMgrDeploymentSize = "small"
$NSXTMgrvCPU = "6" #override default size
$NSXTMgrvMEM = "24" #override default size
$NSXTMgrDisplayName = "tanzu-nsx-3"
$NSXTMgrHostname = "tanzu-nsx-3.cpbu.corp"
$NSXTMgrIPAddress = "172.17.31.118"

# NSX-T Edge Configuration
$NSXTEdgeDeploymentSize = "medium"
$NSXTEdgevCPU = "8" #override default size
$NSXTEdgevMEM = "32" #override default size
$NSXTEdgeHostnameToIPs = @{
    "tanzu-nsx-edge-3a" = "172.17.31.116"
}

Once you have saved your changes, you can now run the PowerCLI script as you normally would.

Logging

There is additional verbose logging that outputs as a log file in your current working directory vsphere-with-tanzu-nsxt-lab-deployment.log

Sample Execution

In this example below, I will be using a single /24 native VLAN (172.17.31.0/24) which all the VMs provisioned by the automation script will be connected to. It is expected that you will have a similar configuration which is the most basic configuration for POC and testing purposes.

Hostname IP Address Function
tanzu-vcsa-3.cpbu.corp 172.17.31.112 vCenter Server
tanzu-esxi-7.cpbu.corp 172.17.31.113 ESXi
tanzu-esxi-8.cpbu.corp 172.17.31.114 ESXi
tanzu-esxi-9.cpbu.corp 172.17.31.115 ESXi
tanzu-nsx-edge.cpbu.corp 172.17.31.116 NSX-T Edge
tanzu-nsx-ua.cpbu.corp 172.17.31.118 NSX-T Unified Appliance
n/a 172.17.31.119 T0 Static Route Address
n/a 172.17.31.120 to 172.17.31.125 K8s Master Control Plane VMs
n/a 172.17.31.140/27 Ingress CIDR Range
n/a 172.17.31.160/27 Egress CIDR Range

Note: Make sure Ingress/Egress CIDR ranges do NOT overlap and the IP Addresses within that block is not being used. This is important as the Egress CIDR will consume at least 15 IP Addresses for the SNAT of each namespace within the Supervisor Cluster.

Lab Deployment Script

Here is a screenshot of running the script if all basic pre-reqs have been met and the confirmation message before starting the deployment:

Here is an example output of a complete deployment:

Note: Deployment time will vary based on underlying physical infrastructure resources. In my lab, this took ~40min to complete.

Once completed, you will end up with your deployed vSphere with Kubernetes Lab which is placed into a vApp

Enable Workload Management

To consume the vSphere with Kubernetes capability in vSphere 7, you must enable workload management on a specific vSphere Cluster, which is currently not part of the automation script. The instructions below outline the steps and configuration values used in my example. For more details, please refer to the official VMware documentation here.

Step 1 - Login to vSphere UI and click on Menu->Workload Management and click on the Enable button

Step 2 - Select the Workload Cluster vSphere Cluster which should automatically show up in the Compatible list. If it does not, then it means something has gone wrong with either the selected configuration or there was an error durig deployment that you may have missed.

Step 3 - Select the Kubernetes Control Plane Size which you can use Tiny

Step 4 - Configure the Management Network by selecting the DVPG-Management-Network distributed portgroup which is automatically created for you as part of the automation. Fill out the rest of the network configuration based on your enviornment

Step 5 - Configure the Workload Network by selecting the Tanzu-VDS distributed virtual switch which is automatically created for you as part of the automation. After selecting a valid VDS, the Edge Cluster option should automatically populate with our NSX-T Edge Cluster called Edge-Cluster-01. Next, fill in your DNS server along with both the Ingress and Egress CIDR values (/27 network is required minimally or you can go larger)

Step 6 - Configure the Storage policies by selecting the tanzu-gold-storage-policy VM Storage Policy which is automatically created for you as part of the automation or any other VM Storage Policy you wish to use.

Step 7 - Finally, review workload management configuration and click Finish to begin the deployment.

This will take some time depending on your environment and you will see various errors on the screen, that is expected. In my example, it took ~26 minutes to complete. You will know when it is completely done when you refreshed the workload management UI and you see a Running status along with an accessible Control PLane Node IP Address, in my case it is 172.17.31.129

Note: In the future, I may look into automating this portion of the configuration to further accelerate the deployment. For now, it is recommended to get familiar with the concepts of vSphere with Kubernetes by going through the workflow manually so you understand what is happening.

Create Namespace

Before we can deploy a workload into Supervisor Cluste which uses vSphere Pods, we need to first create a vSphere Namespace and assign a user and VM Storage Policy.

Step 1 - Under the Namespaces tab within the workload management UI, select the Supervisor Cluster (aka vSphere Cluster enabled with workload management) and provide a name.

Step 2 - Click on Add Permissions to assign both the user [email protected] and [email protected] which was automatically created by the Automation or any other valid user within vSphere to be able to deploy workloads and click on Edit Storage to assign the VM Storage Policy tanzu-gold-storage-policy or any other valid VM Storage Policy.

Step 3 - Finally click on the Open URL under the Namespace Status tile to download kubectl and vSphere plugin and extract that onto your desktop.

Deploy Sample K8s Application

Step 1 - Login to Control Plane IP Address:

./kubectl vsphere login --server=172.17.31.129 -u [email protected] --insecure-skip-tls-verify

Step 2 - Change context into our yelb namespace:

./kubectl config use-context yelb

Switched to context "yelb".

Step 3 - Create a file called enable-all-policy.yaml with the following content:

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
 name: allow-all
spec:
 podSelector: {}
 ingress:
 - {}
 egress:
 - {}
 policyTypes:
 - Ingress
 - Egress

Apply the policy by running the following:

./kubectl apply -f enable-all-policy.yaml

networkpolicy.networking.k8s.io/allow-all created

Step 3 - Deploy our K8s Application called Yelb

./kubectl apply -f https://raw.githubusercontent.com/lamw/vmware-k8s-app-demo/master/yelb-lb.yaml

service/redis-server created
service/yelb-db created
service/yelb-appserver created
service/yelb-ui created
deployment.apps/yelb-ui created
deployment.apps/redis-server created
deployment.apps/yelb-db created
deployment.apps/yelb-appserver created

Step 4 - Access the Yelb UI by retrieving the External Load Balancer IP Address provisioned by NSX-T and then open web browser to that IP Address

./kubectl get service

NAME             TYPE           CLUSTER-IP    EXTERNAL-IP     PORT(S)        AGE
redis-server     ClusterIP      10.96.0.69    <none>          6379/TCP       43s
yelb-appserver   ClusterIP      10.96.0.48    <none>          4567/TCP       42s
yelb-db          ClusterIP      10.96.0.181   <none>          5432/TCP       43s
yelb-ui          LoadBalancer   10.96.0.75    172.17.31.130   80:31924/TCP   42s

Deploy Tanzu Kubernetes Cluster

Step 1 - Create a new subscribed vSphere Content Library pointing to https://wp-content.vmware.com/v2/latest/lib.json which contains the VMware Tanzu Kubernetes Grid (TKG) Images which must be sync'ed before you can deploy a TKG Cluster.

Step 2 - Navigate to the Workload-Cluster and under Namespaces->General click on Add Library to associate the vSphere Content Library we had just created in the previous step.

Step 3 - Create a file called tkg-cluster.yaml with the following content:

apiVersion: run.tanzu.vmware.com/v1alpha1
kind: TanzuKubernetesCluster
metadata:
  name: tkg-cluster-1
  namespace: yelb
spec:
  distribution:
    version: v1.16.8
  topology:
    controlPlane:
      class: best-effort-xsmall
      count: 1
      storageClass: tanzu-gold-storage-policy
    workers:
      class: best-effort-xsmall
      count: 3
      storageClass: tanzu-gold-storage-policy
  settings:
    network:
      cni:
        name: calico
      services:
        cidrBlocks: ["198.51.100.0/12"]
      pods:
        cidrBlocks: ["192.0.2.0/16"]
    storage:
      defaultClass: tanzu-gold-storage-policy

Step 4 - Create TKG Cluster by running the following:

./kubectl apply -f tkg-cluster.yaml

tanzukubernetescluster.run.tanzu.vmware.com/tkg-cluster-1 created

Step 5 - Login to TKG Cluster specifying by running the following:

./kubectl vsphere login --server=172.17.31.129 -u [email protected] --insecure-skip-tls-verify --tanzu-kubernetes-cluster-name tkg-cluster-1 --tanzu-kubernetes-cluster-namespace yelb

Step 6 - Verify the TKG Cluster is ready before use by running the following command:

./kubectl get machine

NAME                                           PROVIDERID                                       PHASE
tkg-cluster-1-control-plane-2lnfb              vsphere://421465e7-bded-c92d-43ba-55e0a862b828   running
tkg-cluster-1-workers-p98cj-644dd658fd-4vtjj   vsphere://4214d30f-5fd8-eae5-7b1e-f28b8576f38e   provisioned
tkg-cluster-1-workers-p98cj-644dd658fd-bjmj5   vsphere://42141954-ecaf-dc15-544e-a7ef2b30b7e9   provisioned
tkg-cluster-1-workers-p98cj-644dd658fd-g6zxh   vsphere://4214d101-4ed0-97d3-aebc-0d0c3a7843cb   provisioned

Step 7 - Change context into tkg-cluster-1 and you are now ready to deploy K8s apps into a TKG Cluster provisioned by vSphere with Kubernetes!

./kubectl config use-context tkg-cluster-1

Network Topology

Here is view into what the networking looks like (Network Topology tab in NSX-T UI) once this is fully configured and workloads are deployed.You can see where the T0 Static Route Address is being used to connect both vSphere Pods (icons on the left) and Tanzu Kubernetes Grid (TKG) Clusters (icons on the right).

About

vSphere with Tanzu using NSX-T Automated Lab Deployment

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published