Step‑by‑Step Guide to Deploy UCloud’s Free USDP Big Data Platform on CentOS

Background

Cloudera and Hortonworks merged, ending free community versions of CDP and HDP. UCloud released a free, one‑stop USDP (UCloud Data Platform) that supports HDFS, Kudu, Elasticsearch and other components.

Environment Preparation

USDP requires a Manager Node with a MySQL instance and Worker Nodes running the Agent. A minimum of three CentOS 7.2‑7.6 nodes (8 CPU, 32 GB RAM, 500 GB data disk) is needed.

Download and Extract USDP

Download the free package (≈43 GB) from https://s3-cn-bj.ufileos.com/jungle111111/usdp-1.0.0.0/install/usdp-free-1.0.0.tar.gz , then extract it to /opt/usdp-srv/ on the repair node.

[root@node1 usdp-1.0.0.0]# ll

Directory Structure

agent : USDP distributed client

bin : start/stop scripts

config : configuration files

jmx_exporter : monitoring exporter

recommend : service templates

repair : initialization scripts and packages

repository : service resource packages

scripts : auxiliary scripts

server : USDP manager

sql : metadata initialization SQL

templated : service configuration templates

verify : certificate storage

versions : package version info

Repair Module Configuration

Edit repair.properties to set YUM source, NMAP, NTP, MySQL hosts, passwords, and node counts. Edit repair-host-info.properties to list each node’s IP, SSH port, password, and hostname.

# Set the YUM source host IP
yum.repo.host.ip=10.23.110.136
# NMAP host
nmap.server.ip=10.23.110.136
nmap.server.port=22
nmap.server.password=abcd123456
# MySQL host
mysql.ip=10.23.110.136
mysql.host.ssh.port=22
mysql.host.ssh.password=abcd123456
mysql.password=abc123456
repair.host.num=3
repair.log.dir=./logs

Initialize the Cluster

Run the one‑click repair script on the repair node:

cd /opt/usdp-srv/usdp/repair/sbin
bash repair.sh initAll
source /etc/profile

The script installs required packages (JDK, Python, MySQL), distributes configuration files, and reports SUCCESS for each component.

Configure MySQL for USDP

Update /opt/usdp-srv/usdp/config/application-server.yml with the datasource URL, driver, username and password.

datasource:
  type: com.zaxxer.hikari.HikariDataSource
  driver-class-name: com.p6spy.engine.spy.P6SpyDriver
  url: jdbc:p6spy:mysql://node1:3306/db_udp?useUnicode=true&characterEncoding=utf-8&useSSL=false
  username: root
  password: abc123456

Start USDP Manager Service

On the manager node, execute:

cd /opt/usdp-srv/usdp/
bin/start-udp-server.sh

Successful start is indicated by UDP Server is running with: 10691.

Web UI and Cluster Creation

Access http://10.23.110.136, set an admin password, import the free license (generated from the hardware ID), and use the wizard to create a cluster. Choose at least three nodes, select a recommended component set (e.g., Recommendation B), configure HDFS and Hive defaults, and start deployment. Progress reaches 100 % when finished.

Monitoring and Alerts

USDP aggregates JMX, HTTP, and custom metrics into Prometheus and provides Grafana dashboards. Pre‑defined alert templates can notify via WeChat, DingTalk, email, or webhook, and users can create custom alerts.

Cluster Usage Example

Log in as the hadoop user and launch Hive:

su hadoop
/srv/udp/1.0.0.0/hive/bin/hive

Create a table and insert a row:

create table iteblog_test_usdp_hive (id int, name string, age int);
insert into iteblog_test_usdp_hive values (1,'iteblog',100);
select * from iteblog_test_usdp_hive;

The default execution engine is Tez, which can be changed via the USDP UI.

Conclusion

USDP offers a fully automated, free alternative to CDH for deploying a production‑grade Hadoop ecosystem, dramatically reducing manual effort and error risk.