Prerequisites
Complete the following before you install a global cluster on any provider. Each provider adds its own prerequisites on its prerequisites page.
This installation path applies when the nodes of the global cluster run Alauda OS on Bare Metal, Huawei DCS, VMware vSphere, or Huawei Cloud Stack. For a non-Alauda OS such as Ubuntu or RHEL, use Install on Non-Alauda OS instead.
TOC
Common PrerequisitesBootstrap NodeHardwareOperating SystemNetworkPrepare the Core Package and the RegistryPlan the Control Plane EndpointEndpoint Modes by ProviderExternal Load BalancerCheck the Listener Before CreationSelf-built VIPNext StepsCommon Prerequisites
First prepare the resources that every installation needs, as listed in Prepare Resources: the installation packages, the external platform image registry with the payload uploaded, the platform domain name, and your own certificate if you use one. This installation path does not support the platform built-in Registry.
On Alauda OS, also prepare the following:
- A bootstrap node that meets the Bootstrap Node requirements. A disaster recovery deployment needs one bootstrap node on each side.
- The Alauda OS image published for this release. Prepare it for the nodes as the prerequisites of your provider describe.
- A stable control plane endpoint on port
6443. The endpoint is an external load balancer or a Self-built VIP; see Plan the Control Plane Endpoint. On Huawei Cloud Stack, the provider creates an ELB; see its prerequisites. - On Huawei DCS, VMware vSphere, and Huawei Cloud Stack, a static IPv4 address and a hostname for each node. The provider assigns them to the virtual machines that it creates. See the prerequisites of your provider.
- For x86_64 nodes, CPUs that support the
x86-64-v2ISA baseline. See OS Support Matrix.
If you use your own certificate for the platform access address, put its files on the bootstrap node: a PEM certificate and private key, or a PKCS#12 bundle. install-config.yaml refers to them by absolute paths.
Bootstrap Node
The bootstrap node is a temporary machine. It runs ac and, during the installation, the bootstrap cluster. It never joins the global cluster. You can reclaim it after the installation; before you do, see Installation Directory for what to keep.
Hardware
On arm64, use at least 6 cores and 12 GB of memory. The recommended size is 8 cores and 16 GB.
Operating System
Use any 64-bit Linux distribution with Bash and root access.
Prepare the bootstrap node so that it meets the following conditions. ac starts the bootstrap cluster with the setup.sh script of the Core Package, and setup.sh checks these conditions first:
- You run
acasroot. - Swap is disabled.
swapon -slists no swap, and/etc/fstabhas no uncommented line that containsswap. - If AppArmor is installed, as on Ubuntu, it is disabled. The
apparmorservice is neither active nor enabled, and/sys/module/apparmor/parameters/enabledis notY. - No container runtime is installed.
containerdanddockerdare not on thePATH.setup.shinstalls thecontainerdthat the bootstrap cluster uses. /etc/cnidoes not exist./etc/resolv.confexists and has at most twonameserverentries. None of them is a loopback address, except127.0.0.53.
If a condition is not met, setup.sh stops before it changes the host. Fix the condition and run create again, as the output of ac shows. See Common Failures.
Network
The bootstrap node needs a static IP address that does not change during the installation.
Allow the network flows that the Communication Matrix lists for the bootstrap node and the global nodes in these sections: Installation, Infrastructure Providers, and Image Registry.
Prepare the Core Package and the Registry
ac runs on the bootstrap node and reads the extracted Core Package there. Run these steps on the bootstrap node, or on the bootstrap node of each side for a disaster recovery pair:
- Upload the Core Package to the bootstrap node and extract it. In this documentation,
<extract-dir>is the extractedinstaller/directory. - Copy all packages of ACP Common Aligned Extensions into
<extract-dir>/plugins/, as described in Download. - If you have not uploaded the payload to the external registry yet, upload it from
<extract-dir>as described in Prepare the Target-Version Payload. For a disaster recovery pair with one registry per side, upload it to both registries.
The Core Package contains ac at <extract-dir>/res/<arch>/ac, where <arch> is amd64 or arm64. The installation pages run it from there. Keep it in place; you do not need to install it.
Plan the Control Plane Endpoint
Every cluster needs a stable endpoint for the Kubernetes API. Select the endpoint mode before you create the cluster.
Do not change the control plane endpoint after the cluster is created. The endpoint is part of the API server certificates, the kubeconfigs, and the connection from the global cluster.
Endpoint Modes by Provider
If you do not choose a mode, the default applies. You set the mode and its values in install-config.yaml; see the template of your provider: Bare Metal, Huawei DCS, or VMware vSphere.
On VMware vSphere, a cluster can use a Self-built VIP only if it was created with one. A cluster whose provider was upgraded from an earlier release keeps its external load balancer. See Control Plane Endpoint Modes.
External Load Balancer
Provision the load balancer before you create the cluster. It must meet the External Load Balancer requirements.
Check the Listener Before Creation
Run this command on a node of the global cluster, or, for a global installation, on the bootstrap node:
Self-built VIP
- Reserve an unused IPv4 address in the same Layer 2 network as the control plane interface that will hold the VIP.
- Use a VRID from
1to255that is unique in that Layer 2 domain. - Allow VRRP traffic and gratuitous ARP updates between the control plane nodes and the surrounding network; see Self-built VIP.
- Use at least three control plane nodes for high availability.
- Use an IP address, not a domain name, for the Self-built VIP.
- Do not maintain a second backend list for the same VIP by hand. The provider updates the backend list when control plane nodes are replaced.
- If you name the VIP interface, use a Linux interface name of at most 15 characters, such as
eth0. If you do not name one, the VIP is held on the interface that has the node address.
On Huawei DCS, the port group allows VIP ownership, gratuitous ARP, and VRRP traffic.
On VMware vSphere, the following additional rules apply:
- The VIP must be IPv4. IPv6 and dual-stack endpoints are not accepted.
- The VIP must not equal any node address, primary or additional. It must also be outside every external IPAM pool range, because the provider does not detect a conflict with an external IPAM pool.
- On nodes with more than one NIC, name the VIP interface explicitly. You cannot change it after the cluster is created.
- The dvPortGroup security policy and any NSX-T distributed firewall rule allow VRRP (IP protocol
112, multicast224.0.0.18) and gratuitous ARP between the control plane nodes.
Next Steps
You now have:
- The external registry address and credentials.
- The platform access address.
- The user-provided certificate, if you use one.
- The control plane endpoint.
Continue with the overview of your provider. It describes how the installation works on that provider and lists the remaining steps in order. Then complete the prerequisites of your provider and create the cluster:
For a primary and a standby global cluster, read Create a Global Cluster Disaster Recovery Pair before you write either install-config.yaml.