Deploying a Cluster on CentOS 7 and Red Hat Enterprise Linux 7

Note!
YMatrix uses vectorized execution to improve query performance. Deployment will fail if the host (physical or virtual machine) does not support the AVX2 instruction set.
If you are deploying on a virtual machine, ensure that the underlying physical host supports AVX2 and that AVX2 is enabled in the VM configuration.

Operating System and Platform Requirements

Operating System	Supported CPU Architecture
CentOS 7 Red Hat Enterprise Linux 7	x86_64

Example Cluster Configuration

This document uses a three-node cluster as an example:

Master node: mdw
Two segment nodes: sdw1 and sdw2

Server Installation

The server installation process consists of six steps: checking basic system information, preparation, installing the database RPM package, installing Python dependencies, deploying the database, and post-installation configuration.

1. Check Basic Server Information

Before starting the installation, check the basic system information. This is a good practice, as understanding the server helps you plan the cluster deployment effectively.

Step	Command	Purpose
1	`free -h`	View memory usage
2	`df -h`	View disk space
3	`lscpu`	View number of CPU cores
4	`cat /etc/system-release`	View OS version
5	`uname -a`	Display all kernel information in the following order: kernel name; hostname; kernel release; kernel version; hardware architecture; processor type (non-portable); hardware platform (non-portable); OS name. Fields `-p` and `-i` are omitted if unavailable.
6	`tail -11 /proc/cpuinfo`	View CPU details

2. Installation Preparation

2.1 Copy RPM Package

Copy the RPM package to all nodes in the cluster.

~ scp <local_path> <username>@<server_ip>:<server_path>

2.2 Install Dependencies

Note!
Perform the following steps as the root user or with sudo privileges on all nodes.

YMatrix requires Python 3.6. Install it and set it as the default Python version:

# yum install centos-release-scl
# yum install rh-python36
# scl enable rh-python36 bash

2.3 Modify System Settings

Disable the firewall:

# systemctl stop firewalld.service
# systemctl disable firewalld.service

Disable SELinux by editing /etc/selinux/config and setting SELINUX=disabled:

# sed s/^SELINUX=.*$/SELINUX=disabled/ -i /etc/selinux/config
# setenforce 0

Ensure each node has a persistent hostname. If not already set, use the following command. In this example, the master node is named mdw:

# hostnamectl set-hostname mdw

Set hostnames for the two segment nodes as well (sdw1, sdw2):

# hostnamectl set-hostname sdw1

# hostnamectl set-hostname sdw2

Ensure all nodes can resolve each other by hostname and IP address. Add entries to /etc/hosts to map hostnames to local network interface addresses:

# vim /etc/hosts

For example, all three nodes should have the following entries in /etc/hosts:

192.168.100.10 mdw
192.168.100.11 sdw1
192.168.100.12 sdw2

3. Install Database RPM Package

Note!
You must install the database RPM package using the root user or sudo on all nodes. System dependencies will be installed automatically.

By default, the package installs under /opt/matrixdb6.

Run the following command:

# yum install matrixdb6_6.2.0+enterprise-1.el7.x86_64.rpm

After successful installation, the supervisord and MXUI processes start automatically. These background services provide the web-based management interface and process control.

If you need to customize service ports, modify the configuration file after RPM installation. This step is required only on the Master node:

# vim /etc/matrixdb6/defaults.conf

4. Deploy the Database

YMatrix provides a graphical deployment tool. The remote web interface is accessible via ports 8240 and 4617. After installation, these ports are open by default on all nodes. The MXUI process provides the web interface.

Note!
If the graphical interface is unavailable, refer to Command-Line Installation.

Open a browser and navigate to the installation wizard URL. The server whose IP address you enter will become the Master node of your cluster (in this example, mdw):

http://<IP>:8240/

On the first page of the wizard, enter the superuser password. You can view the default password using the following command:

On the second page, select Multi-Node Deployment, then click Next.

Proceed with the four-step multi-node deployment process.

Step 1: Add Nodes
Click the "Add" button.

Enter the IP addresses, hostnames, or FQDNs of sdw1 and sdw2, then click Comfirm and Next.

Step 2: Configure Cluster Parameters
Enable Data Mirroring to create backup segments. This is recommended in production environments for high availability. The system automatically suggests the largest available disk and an appropriate number of segments based on system resources. Adjust according to your use case. The configured cluster topology is displayed visually. Click Next after confirmation.

Step 3: Set Data and etcd Storage Paths
Select a storage path for etcd on each server. The etcd cluster will be created on an odd number of servers to ensure leader election consistency and avoid ties.

If you enable Use Data Disk for etcd, be aware of the associated risks.

Step 4: Execute Deployment
Review the configuration summary. Click Deploy to begin.

The system automatically deploys the cluster and displays detailed progress. When all steps complete successfully, deployment is finished.

Deployment completed.

5. Post-Installation Configuration

By default, YMatrix allows remote connections. If you did not select Allow Remote Access to Database during installation, manually add a line like the following to the pg_hba.conf file to allow all users from any IP to connect with password authentication:

host  all       all   0.0.0.0/0  md5

After making changes, switch to the mxadmin user and reload the configuration:

# su - mxadmin
$ mxstop -u

Use the following commands to start, stop, restart, or check the status of the YMatrix cluster:

$ mxstart -a
$ mxstop -af
$ mxstop -arf 
$ mxstate -s

Command	Purpose
`mxstop -a`	Stop the cluster. (Stops only after active sessions end.)
`mxstop -af`	Forcefully stop the cluster immediately.
`mxstop -arf`	Restart the cluster. Waits for currently running SQL statements to finish. (Stops only after active sessions end.)
`mxstate -s`	Check cluster status

← Previous

Onboard: Management

CentOS 8, Red Hat 8, etc.

English Русский 简体中文