Web cluster environment preparation

It is known that load balancing Nginx does not support high concurrency on Windows Server, so for solutions that use Nginx as load balancing, if the concurrency demand is below 300, the cluster can be deployed normally on both Linux systems and Windows Server, and if the concurrency demand is above 300, it is not recommended to deploy the cluster on Windows Server, otherwise concurrency problems will occur.

(1) JDK version requirements: JDK 1.8 and minor versions need to be above JDK8u102.

(2) A set of clusters does not support a server to deploy multiple clusters of node projects.

(3) Need to confirm that the projects communicate properly with each other and with the external database.

Notes.

• Linux system users: you can use the cluster environment detection tool to automatically detect the system environment, or use the ping statement to detect.

• Windows system users: use ping or telenet statement to detect.

Linux systems require the top user (root) to install FTP, otherwise it cannot be done, but this requirement can be ignored if FTP is not installed.

Both Redis and Nginx rely on the GCC compiler for deployment on Linux systems, so please note that if you choose a Linux server, please ensure that your system has a GCC compiler environment

Query command: gcc -v, if you do not have a gcc environment, you need to install it yourself, network installation command：yum install gcc gcc-c++

Clusters can be deployed in both public and intranet environments, although a good communication environment between the servers in the cluster is required, as follows.

Currently, a two-node cluster base solution (Nginx+Tomcat+Redis+ftp) is more commonly used, and we use this as an example to illustrate the requirements for the server.

5.1 Number of Servers

For a cluster scenario with two nodes deployed, it is recommended to prepare three servers with the following architecture.

5.2 Memory Configuration

5.3 Disk Space

Deploying a cluster project requires at least 40G of free space in the root directory of each machine, and if the template resource file is large, choose a server with larger disk space according to the actual situation.

Notes：

• 1) Web container, file server and state server contain many kinds, please open the ports by yourself according to the selection.

• 2）If the default port conflicts with other projects, you need to modify the port first, and then open the corresponding port.

• 3) Open port 21 and any port larger than 1024, and then open ftp passive mode through passive.

• 4) For docker containers, or environments with strict port restrictions between servers, you need to open ports between servers of nodes for inter-node communication.

• Select TCP protocol is 7800/7810/7820/7830/7840/7850/7860/7870 eight ports

• Select UDP protocol is 45588~65536 random ports

Notes：

A large number of calculations are done by the database, and BI memory consumption is mainly related to result set size and concurrent performance.

If virtual machines are used, the number of cores and logical cores are kept consistent.

The number of nodes increases concurrent performance improvement is more obvious, support expansion nodes, the bottleneck of maximum concurrency is mainly in the configuration read.

Edit user concurrency: users who do datasets/dashboards.

Total user concurrency: editing + preview users. ( For example, 150 concurrent users under 8 cores and 16G memory. Means maximum 150 users of which editing users are maximum 20 other are preview users )