Deploying the Platform

Procedure

1. Decompress the main installation package to the /root/cpaas-install directory on the first master node of the global platform. (You can also decompress it to another directory. The directory for storing decompressed files must have at least 100 GiB space. After the deployment is complete, you can delete the directory.)
2. Run the following commands on the first master node to decompress the package and go to the installation directory:

   tar -xvf < Installation package file path, for example, installer-v3.6.0.tar> -C /root/cpaas-install # Decompress the installation package.
   cd /root/cpaas-install/installer

3. Select the global cluster network and run the corresponding command.
   • To use a Kube-OVN overlay network to deploy the global cluster, run the following command:

     bash setup.sh

   • To use a Calico network to deploy the global cluster, run the following command:

     bash setup.sh --network-mode calico

4. Enter the URL in the address box of the browser to access the platform deployment page based on the command output.
   

   It takes about 5 minutes to wait for the minialauda to be ready.

   Figure 1 Basic settings
   
   Figure 2 Advanced settings
   

   Table 1 Parameters

   Configuration	Parameter	Description	Remarks
   System Settings	Kubernetes Version	Select the Kubernetes component version.	To ensure proper function dependency, it is important to carefully consider the version of Kubernetes selected for the service mesh. The product baseline document can provide information on function dependency that may be helpful in making this decision.
   	Container Runtime	Select the container runtime version. containerd is used by default, helping you quickly set up a container. containerd containerd has a shorter tracing and fewer components. It is more stable and occupies fewer node resources.	None
   	Cluster Endpoint	API server address exposed by the Kubernetes cluster where global is in. Determine the deployment architecture by referring to "Capacity Planning" in Alauda Cloud Native Success Platform Installation Guide. There are two ways to configure cluster HA. You can use Keepalived or a load balancer: Keepalived: Enable on-premises VIP (virtual router ID) and set IP/Domain to the applied VIP. Once the cluster is successfully created and using the on-premises VIP, the access address will be <VIP>:6443. Load balancer: Disable the on-premises VIP and set IP/Domain to the IP address or domain name of the load balancer (such as the F5, IaaS load balancer, or HAProxy) outside the cluster. To quickly experience the platform's functions, you can deploy a single node and use it as both the master and compute node. You can disable the on-premises VIP and set IP address/domain name to the IP address of a prepared node. If you plan on adding master nodes to the cluster after it is created, it is recommended that you enter the IP address or domain name of the load balancer outside the cluster in the designated text box. This will make future scale-out easier.	Note: To enable the on-premises VIP, Keepalived needs the host network to support Virtual Router Redundancy Protocol (VRRP) and all nodes to be in the same subnet as the VIP. To ensure HA of the cluster IP address, you can use an on-premises VIP. However, you will need to contact the platform network administrator or O&M personnel beforehand to apply for the VIP and virtual router ID. It is important to note that the virtual router ID must be unique within the same subnet.
   	GPU Type	Select the corresponding GPU as needed.	Make sure that the node has the GPU driver installed.
   	Hardware architecture	Select x86 or Arm.	It specifies the hardware architecture for the cluster. Once selected, only nodes with the corresponding architecture can be added to the cluster.
   	Platform Access Address	Access address of the platform, which is used for communications between the global cluster and the service cluster. By default, it is the same as the cluster IP address. You can configure it by referring to the following: If you need both internal and external network access, enter an internal network address. Other access addresses of the platform can be added in the advanced settings on this installation page. If you need DR, the address must be a domain name. If you need a valid certificate, enter the domain name and configure a self-signed certificate in the Certificate parameter or upload a certificate issued by the certification authority.	The platform is deployed using the HTTPS protocol by default. If you want to use the insecure HTTP protocol, you can enable the option in the advanced settings.
   Image Repository	Image Repository	Repository for storing platform component images. Platform Deployment: A built-in image repository is created during deployment of global. All component images on the platform are pulled from this repository. External: You need to enter necessary details of an external image repository. In DR environments, you also need to ensure that the repository IP address is accessible to all DR nodes. It is recommended to set up an image repository beforehand to store platform component images. For more information, contact technical support.	None
   Container Network	Cluster Network Protocol	Implementing IPv4/IPv6 dual-stack can effectively address the issue of limited IPv4 network address resources. Before enabling dual-stack, make sure that all nodes have correctly configured IPv6 network addresses. Once a cluster has been created with dual-stack enabled, it cannot be reverted back to IPv4 single stack. In the following scenarios, it is recommended that you enable dual-stack: Your applications need to provide access services to users with IPv6 terminals. You need to analyze and process the access sources that use IPv6 terminals to access services provided by your applications. Your application system needs to use IPv6 to access other systems (such as the database system) or application systems.	The global platform uses the Kube-OVN overlay network by default. Make sure that the container network and host network belong to different network CIDR blocks to avoid any exceptions during system deployment.
   	Default Subnet	Default subnet CIDR block, which is the cluster CIDR block
   	Service CIDR	IP address range used by the Kubernetes ClusterIP Services. It must not overlap with the default subnet's network CIDR block.
   	Join CIDR	IP address range used for communications between nodes and container groups in Kube-OVN overlay transmission mode. It cannot overlap with the default subnet or service CIDR block.
   Node Settings	Network Interface Card	Node network interface used by the cluster network add-on. If this parameter is not specified, the system automatically obtains the network interface corresponding to the default route of the node.	None
   	Node Name	Select Use IP or Use Hostname.	When selecting a hostname as the node name, you need to ensure that the hostname is unique in the cluster.
   	Global Cluster Platform Node Isolation	If you need both the global cluster to run platform components and the global cluster to run service applications, you can enable this function to prevent platform components and service applications from preempting resources. Once you enable this function, you must configure the Application Deployable for any node you add. This configuration ensures that only platform components can be deployed on the node, and that platform components (excluding certain DaemonSets) will not be scheduled to nodes without Application Deployable configured.	None
   	Add Node	Only one or three master nodes are supported. If there are three master nodes in the cluster, the global cluster is an HA cluster. Application Deployable: For a controller node: When you enabled Application Deployable, related functions are forcibly disabled. In this case, service applications cannot be deployed on this node. When you disabled Application Deployable, related functions are disabled by default, but you can still determine whether to deploy service applications on this node. For a compute node: When you enabled Application Deployable, related functions are forcibly disabled. In this case, service applications cannot be deployed on this node. When you disabled Application Deployable, related functions are forcibly enabled, so that service applications can be deployed on this node. When adding a master node or compute node, if the GPU node is enabled, you need to manually install the GPU driver and container runtime. When the Kube-OVN overlay network is used to deploy the platform, if the network interface name of a node is specified, the node uses the configured network interface.	For details about how to select a deployment architecture, see "Capacity Planning" in Alauda Cloud Native Success Platform Installation Guide. After you click Add Node, the platform checks the availability of the node. If the verification fails, adjust the configuration following instructions and add the node again.
   Log monitoring settings	Monitoring Type	It is recommended to use Prometheus for monitoring. When choosing a node for setting up the monitoring service, it is advisable to select a non-master node. If VictoriaMetrics is selected, you must configure the number of VictoriaMetrics agent instances, that is, the number of VMAgents. You are advised to add one. A maximum of three can be added.	To properly configure the monitoring node, you must meet the requirements listed in "Hardware Configuration" in Alauda Cloud Native Success Platform Installation Guide. You can deploy HA monitoring components in scenarios that demand HA. The monitoring component is typically deployed on a small scale by default. If you need to make changes to the deployment, you can disable the monitoring component in platform management after deployment, redeploy the component, and select nodes to deploy monitoring components of varying scales.
   	Log Storage Node	Deploy components like Elasticsearch on the node. It is recommended to choose a non-master node for setting up the log service. The default log service has limited capacity, so if you require a larger log scale, contact the relevant personnel.	Only one or three log nodes can be selected. If more log nodes are required, see Changing the Number of Nodes Where Elasticsearch Runs.
   Other settings	Other Access Addresses	You can enter multiple IP addresses or domain names.	When entering IP addresses, make sure they can be forwarded to the cluster IP address. When entering domain names, make sure they have been resolved to the cluster IP address.
   	Pod Number Limit	Maximum number of pods on each node. The default value is 110. If all-in-one deployment architecture is used, you can increase the maximum number of pods on a node to 255 manually to ensure that there are enough pod IP addresses.	None
   	Product	Select the name of the product to be deployed.	None
   	extension parameters	It is not recommended to manually configure extension parameters. Doing so may result in the cluster becoming unavailable, and it cannot be modified after creation. If you need to make changes, contact technical support for assistance.	None

   After setting these preceding parameters, click Start Deployment in the lower right corner.

Verifying the Platform

To check whether the global platform has been successfully deployed, run the following command on the master node of the Kubernetes cluster where the global component is running:

#Check whether the Sentry component is successfully deployed. Run the following command to search for the chart that fails to be deployed:
kubectl get apprelease --all-namespaces
#Check whether all the pods are running properly. If there is a problem, run the following command to search for the failed pod:
kubectl get pod --all-namespaces | awk '{if ($4 != "Running" && $4 != "Completed")print}' | awk -F'[/ ]+' '{if ($3 != $4)print}'

Accessing the Platform

Using different browsers to access the platform may cause the platform UI to display incorrectly or certain functions to be unavailable due to browser compatibility issues. To avoid these issues, use the recommended browser versions listed below:

• Google Chrome 93 or later
• Firefox 92 or later

1. After the deployment and installation are complete, click the access button on the browser page to go to the platform portal. You can download and check the deployment list.
2. Platform maturity setting (Alpha function switch setting): <Platform access address >/console-platform/feature-gate

Deleting the Installer

Run docker rm -f minialauda-control-plane on the host where the platform is deployed to delete the installer. After the installer is deleted, the deployment list cannot be downloaded.

You do not need to manually delete the container. The system automatically deletes the installer 10 minutes later.