nutanixdev/anthos-on-ahv

Anthos on AHV automation with Nutanix Calm or Terraform

Anthos clusters on AHV

---

NOTE

The CentOS 8.2.2004 image used in this repo is EOL since December 31, 2021. The automation scripts in this repo use CentOS vault repo to download the packages (archived/unsupported packages)

---

What to expect

This repo provides two automation methods for deploying Anthos clusters on AHV. You can choose between a Calm blueprint or a Terraform file.

Overview

The characteristics for the Kubernetes cluster are:

• Anthos model: bare metal

• Anthos versions:

  • Supported: 1.6.x, 1.7.x, 1.8.x and 1.9.x

  • Unsupported: 1.10.x

• Type: hybrid - https://cloud.google.com/anthos/clusters/docs/bare-metal/1.9/installing/install-prep#hybrid_cluster_deployment

• Number of virtual machines: 6 (Total resources: 24 vCPU / 192 GB memory / 768 GB storage )

  • 1 x Admin workstation

  • 3 x Control plane

  • 2 x Worker nodes

• Virtual machine OS: CentOS 8.2 GenericCloud - https://cloud.centos.org/centos/8/x86_64/images/CentOS-8-GenericCloud-8.2.2004-20200611.2.x86_64.qcow2

• High availability: yes

• Load balancing: yes (bundled)

• Ingress: yes

• Persistent storage: yes (Nutanix CSI)

• Proxy: no

• KubeVirt: no

• OpenID Connect: no

• Application logs/metrics: no

• Scale Out/In: yes

• Upgrade Anthos version: yes

• Decommission Anthos cluster: yes

Architecture

This diagram represents a typical hybrid model deployment for Anthos on AHV. Other Anthos deployment models are supported too.

Prerequisites

• Nutanix:

  • Cluster:

    • AHV: 20201105.1045 or later

    • AOS: 5.19.1 or later

    • iSCSI data service IP configured

    • VLAN network with AHV IPAM configured

    • Prism Central: 2020.11.0.1 or later

• Google Cloud:

  • A project with Owner role

  • The project must have enabled Monitoring - https://console.cloud.google.com/monitoring

  • A service account - https://console.cloud.google.com/iam-admin/serviceaccounts/create

    • Role: Project Owner

    • A private key: JSON

• Networking:

  • Internet connectivity

  • AHV IPAM: Minimum 6 IP addresses available for the virtual machines

  • Kubernetes:

    • Control plane VIP: One IP address in the same network than virtual machines but not part of the AHV IPAM

    • Ingress VIP: One IP address in the same network than virtual machines but not part of the AHV IPAM. This IP must be part of the load balancing pool

    • Load balancing pool: Range of IP addresses in the same network than virtual machines but not part of the AHV IPAM. The Ingress VIP is included in this pool

    • Pods network: CIDR network with enough IP addresses, usually /16 and not sharing the same network than virtual machines or Kubernetes Services. If your containerized application must communicate with a system out of the Kubernetes cluster, make sure then this network doesn't overlap either with the external system network

    • Services network: CIDR network with enough IP addresses, usually /16 and not sharing the same network than virtual machines or Kubernetes Pods. If your containerized application must communicate with a system out of the Kubernetes cluster, make sure then this network doesn't overlap either with the external system network

• Credentials:

  • (Calm only) Operating system: you need a SSH key. It must start with ---BEGIN RSA PRIVATE KEY---. To generate one in a terminal:

    ssh-keygen -m PEM -t rsa -f <keyname>

  • Prism Element: an account, local or Active Directory, with User Admin role. This is for the CSI plugin configuration

Variables

Calm	Terraform	Description
NTNX_PC_IP	n/a	If you are using a remote Prism Central instance, use the IP address of that instance. This is used to expand the OS disk via PC API
OS_DISK_SIZE	*_vm_config	The minimum OS disk size MUST be 128GB, recommended by Google is 256GB
NTNX_CSI_URL	ntnx_csi_url	Nutanix CSI Driver URL. Minimum supported version is 2.3.1
NTNX_PE_IP	ntnx_pe_ip	The Prism Element VIP address is needed for the CSI plugin to create persistent volumes via the API. This VIP doesn't have to be the one where the Anthos cluster will run. You can choose any VIP of any of your clusters from where you want to get persistent storage
NTNX_PE_PORT	ntnx_pe_port	Prism Element port. Default is 9440
NTNX_PE_DATASERVICE_IP	ntnx_pe_dataservice_ip	Data service is required to allow iSCSI connectivity between the Kubernetes pods and the volumes created by CSI plugin
NTNX_PE_STORAGE_CONTAINER	ntnx_pe_storage_container	This is the Nutanix Storage Container where the requested Persistent Volume Claims will get their volumes created. You can enable things like compression and deduplication in a Storage Container. The recommendation is to create at least one storage container in Prism Element well identified for Kubernetes usage. This will facilitate the search of persistent volumes when the environment scales
ANTHOS_CLUSTER_NAME	anthos_cluster_name	Anthos cluster name
ANTHOS_VERSION	anthos_version	Anthos cluster version. Supported: 1.6.x, 1.7.x, 1.8.x and 1.9.x (default 1.9.4) - Unsupported: 1.10.x
ANTHOS_SERVICES_NETWORK	anthos_services_network	This is the network for your services. Preferably do not overlap with other networks. CIDR format: XXX.XXX.XXX.XXX/XX
ANTHOS_PODS_NETWORK	anthos_pods_network	This is the network for your pods. Preferably do not overlap with other networks. CIDR format: XXX.XXX.XXX.XXX/XX
ANTHOS_CONTROLPLANE_VIP	anthos_controlplane_vip	This is the IP address for Kubernetes API. Format: XXX.XXX.XXX.XXX
ANTHOS_INGRESS_VIP	anthos_ingress_vip	This is the IP address for Kubernetes Ingress. This address MUST be within the load balancing pool. Format: XXX.XXX.XXX.XXX
ANTHOS_LB_ADDRESSPOOL	anthos_lb_addresspool	This is the IP address range for Load Balancing. Format: XXX.XXX.XXX.XXX-YYY.YYY.YYY.YYY
KUBERNETES_SERVICE_ACCOUNT	kubernetes_service_account	This K8s SA is for Google Cloud Console so the K8s cluster can be managed in GKE. This service account will have cluster-admin role for Google Cloud Marketplace to work. Default is google-cloud-console
PYTHON_ANTHOS_GENCONFIG	python_anthos_genconfig	This script is hosted externally and produce an Anthos configuration file for cluster creation with user provided inputs during launch. DO NOT CHANGE default value unless you will host the script in an internal repository
(credential)	ntnx_pe_username	Username for Prism Element cluster for the CSI plug-in to work
(credential)	ntnx_pe_password	Password for Prism Element cluster
(credential)	google_application_credentials_path	Local path to the GCP service account JSON file
(service spec)	subnet_name	Subnet used for Anthos deployment
(service replica count)	amount_of_anthos_worker_vms	Amount of Anthos worker VMs. Changing this value will result in scale-up or scale-down of the cluster
(credential)	admin_vm_username	Username used for Anthos installation. Default: nutanix