Install on Alauda OS
The global cluster is the control plane of the platform. It runs the platform services and manages every workload cluster. This section describes how to install the global cluster when its nodes run Alauda OS. To install it on a non-Alauda OS such as Ubuntu or RHEL, see Install on Non-Alauda OS.
How the Installation Works
You describe the cluster in one install-config.yaml file and run ac adm alpha global create on a temporary bootstrap node. The installation then proceeds as follows:
acchecksinstall-config.yaml, the registry, and the infrastructure.- It starts a temporary bootstrap cluster on the bootstrap node and installs the Cluster Bootstrap Provider and the infrastructure provider into it. All images come from the external registry.
- The providers create the machines of the
globalcluster, and the control plane is initialized. - The platform is installed on the
globalcluster. - Handover: the
globalcluster takes over its own machines and manages them from then on. acinstalls the two providers on theglobalcluster, writes the cluster's kubeconfig, and removes the bootstrap cluster.
After the installation, the global cluster runs the Cluster Bootstrap Provider and the infrastructure provider of its own platform. You no longer need the bootstrap node.
On Huawei DCS, VMware vSphere, and Huawei Cloud Stack, ac stores the infrastructure credential in the global-credential Secret on the global cluster. When you change the IaaS password, update that Secret.
Keep the installation directory, <install-dir>. It holds the only copy of install-config.yaml, the rendered manifests, output/kubeconfig, and the logs. Copy the directory to a safe place before you reclaim the bootstrap node. See Installation Directory.
Disaster Recovery
A disaster recovery pair consists of two global installations, a primary and a standby. You install each one from its own bootstrap node with its own install-config.yaml. Global Cluster Disaster Recovery runs on the standby and copies the data of the primary to it.
- The two sides share the platform domain, the user-provided certificate of the platform access address, the encryption key, and, on Bare Metal, the ServiceAccount signing key.
- Each side keeps its own control plane VIP, registry configuration, and infrastructure credentials.
- To fail over, you switch the platform domain manually; see Fail Over to the Standby global Cluster.
- Metrics and logs are not replicated. Before you install either side, plan where workload clusters send them; see Observability Placement.
See Create a Global Cluster Disaster Recovery Pair.
Deployment Path
- Prepare. Meet the common prerequisites, then the prerequisites of your provider.
- Decide whether you need disaster recovery. Read Create a Global Cluster Disaster Recovery Pair before you write the first
install-config.yaml. You cannot add disaster recovery to an installedglobalcluster without technical support. - Install the
globalcluster on your provider. For a disaster recovery pair, install both sides as that page describes, then start the synchronization. - Create workload clusters. See Create on Alauda OS.
For the installation stages and the failures that every provider shares, see Troubleshooting. For the common install-config.yaml fields and the create and destroy commands, see Installation Reference.
Not Covered Here
- Installing providers on the
globalcluster for other infrastructure: see Prepare. - Upgrading the
globalcluster, including Kubernetes and Alauda OS: see Upgrade the global Cluster. - Upgrading Kubernetes and Alauda OS on workload clusters: see Who Upgrades Kubernetes.