Create a Cluster on Alauda OS

On this path, creates clusters whose nodes run Alauda OS, an image-based, read-only operating system. On Huawei DCS, Huawei Cloud Stack, and VMware vSphere, the nodes are virtual machines. On Bare Metal, they are physical servers.

Architecture

Cluster API objects on the global cluster describe a cluster on Alauda OS. Two providers run on global and act on those objects:

  • The Alauda Container Platform Cluster Bootstrap Provider turns machines into Kubernetes nodes with kubeadm and runs the control plane.
  • An infrastructure provider creates the machines from Alauda OS images. There is one for each IaaS platform. On Bare Metal, the provider does not create virtual machines. It reinstalls the physical servers that you registered.

Cluster API Architecture

What ac Does and What the Platform Does

You create a cluster with ac adm alpha cluster create. ac renders the Cluster API and provider objects from one install-config.yaml, checks them, and applies them to the global cluster. From then on, the providers on global create the machines and keep the cluster running. For node changes, see Manage Nodes on Alauda OS Clusters. For version changes, see the upgrade documentation.

Nodes Are Replaced, Not Changed in Place

A Kubernetes or Alauda OS version change replaces the nodes one at a time. On Huawei DCS, Huawei Cloud Stack, and VMware vSphere, each node is replaced with a new virtual machine. On Bare Metal, the existing hosts are reinstalled, and a host that is removed from the cluster returns to its pool. To change supported OS settings, use Machine Configuration.

A cluster with a single control plane node is for testing and cannot be upgraded.

DataWhen a node is replaced
Data on the node's own disks on Huawei DCS, Huawei Cloud Stack, and VMware vSphere, such as hostPath volumes and local PersistentVolumes, except data on pool-managed persistent disksNot kept. Store data that must survive on CSI storage or on a pool-managed persistent disk.
Data on pool-managed persistent disks (Huawei DCS, Huawei Cloud Stack, VMware vSphere), such as the platform data in /var/cpaasKept when the replacement node takes over the pool entry of the node it replaces.
Data on the disks of a Bare Metal host: the platform data in /var/cpaas on the system disk, and the data on other disks, such as the raw disks that LVM Storage (TopoLVM) usesKept. The provider does not wipe the disks of a host when it reinstalls the host.
Other data on the system disk of a Bare Metal hostNot guaranteed. The reinstall writes the Alauda OS image to the system disk again. Store data that must survive on CSI storage.
External storage, such as volumes provided through CSINot affected.

Fixed at Creation

Decide the following before you create the cluster, because you cannot change them on an existing cluster:

  • The control plane endpoint mode and address. See Plan the Control Plane Endpoint.
  • The image registry of the cluster, which is either the registry of the global cluster or a dedicated registry. See Choose the Image Registry for a Workload Cluster.
  • On Huawei DCS, Huawei Cloud Stack, and VMware vSphere, the layout of an existing persistent disk, such as its size and, on Huawei DCS, its datastore. On VMware vSphere, the addresses and disks of a pool slot also cannot be changed once a machine takes the slot. Plan the disks of each node pool before you create the cluster.

Supported Providers

ProviderMachinesControl plane endpointNotable for planning
Bare MetalPhysical servers that register once from an installation ISO built by the platform. You list them in a pool for each node groupSelf-built VIP (default), or an external load balancerPrepare the OS and register the hosts once, independently of any cluster. Then create the pools from available hosts before you run ac.
Huawei DCSVirtual machines cloned from a DCS VM templateExternal load balancer (default), or a Self-built VIPEach node has a fixed address and hostname. You reserve them; DCS does not. The Kubernetes version comes from the VM template.
Huawei Cloud StackECS instances in an existing VPC, subnet, and security groupAn HCS ELB that the provider createsEach node pool is in one availability zone. The ELB uses one VIP, two L4 addresses, and two L7 addresses.
VMware vSphereVirtual machines cloned from a vSphere VM template, each with a static IP address and hostname from install-config.yaml (DHCP is not used)External load balancer (default), or a Self-built VIPEvery node group needs its own machine config pool.

Control Plane Endpoint

For what to prepare for each endpoint mode, see Plan the Control Plane Endpoint.

Install the Providers on the global Cluster

Install two providers on the global cluster: the provider for your IaaS platform and the Cluster Bootstrap Provider. Follow Install the Infrastructure Provider. A global cluster installed with ac already runs the Cluster Bootstrap Provider and the provider of its own platform.

Before you upgrade Kubernetes on an Alauda OS global cluster, upload and install the provider for that IaaS platform on the global cluster.

Cluster Creation Workflow

Follow these steps to create a workload cluster on Alauda OS. Prepare describes the first three steps in the same order.

  1. Install the Provider: Install the provider for your IaaS platform, and the Cluster Bootstrap Provider, on the global cluster.
  2. Infrastructure Resources: Prepare what the provider requires, as listed on its Prerequisites page. With ac, this is mainly the credential Secret and the IaaS objects that install-config.yaml names. On Bare Metal, this means the OS images, the registrations, and the registered hosts; see Prepare Alauda OS for Hosts and Register Hosts.
  3. Plan the Cluster: Read Prepare for the decisions that apply to every provider, such as the control plane endpoint mode and the image registry of the cluster. Also check the following before you create the cluster:
    • Node Labels and Taints: If you plan dedicated node pools (monitoring, logging, ingress), read Infra Nodes before you create the pools. It explains how to declare role labels and taints on a pool.
    • CPU Baseline: For x86_64 cluster nodes that use ACP-provided Alauda OS images, the underlying CPUs must support the x86-64-v2 ISA baseline. For details, see OS Support Matrix.
  4. Create a Cluster: Create the cluster, including its first worker nodes, with ac adm alpha cluster create from one install-config.yaml.

Each provider has its own pages for these steps:

If a failure is not provider-specific, see Troubleshooting.

ScenarioDocumentation
Install the global cluster on Alauda OSInstalling the global Cluster
Plan disaster recovery for the global clusterCreate a Global Cluster Disaster Recovery Pair
Learn how Kubernetes and Alauda OS are upgraded on each providerWho Upgrades Kubernetes
Upgrade the global clusterUpgrade the global Cluster
Upgrade workload clustersUpgrade Workload Clusters
Manage cluster nodesManage Nodes on Alauda OS Clusters
Configure machines (OS or kernel level)Machine Configuration
Look up Cluster API and provider resourcesCluster API Reference and Provider APIs

Prepare

Bare Metal

Huawei DCS

Huawei Cloud Stack

VMware vSphere

Creation Reference

Troubleshooting