Deploying StarRocks 3 on ARM Architecture Using Docker

StarRocks is a next‑generation, high‑performance MPP database that supports the MySQL 5.7 protocol, standard SQL, and integrates with Hive on Hadoop, making it a versatile big‑data solution.

The article compares X86 and ARM architectures, noting that StarRocks BE nodes rely on the AVX2 instruction set on X86, so ARM deployment requires either Docker‑based deployment or recompiling without AVX2.

Deployment steps are detailed as follows:

1. Pull and run the Docker image

# 拉取镜像
docker pull starrocks/artifacts-centos7:latest

# 启动镜像
docker run -p 9030:9030 -p 8030:8030 -p 8040:8040 --privileged=true -itd --name starrocks-art starrocks/artifacts-centos7

2. Export configuration files from the image

# 进入镜像内
docker exec -it starrocks-art sh
mkdir starRocks
cp -r be_artifacts/be starRocks/
cp -r fe_artifacts/fe starRocks/
cp -r broker_artifacts/apache_hdfs_broker starRocks/

# 文件压缩
tar -zcvf starRocks.tar.gz starRocks/

# 文件导出
docker cp artifacts-art:/release/starRocks.tar.gz /usr/local/

3. FE node deployment (node01)

# 复制 Hive 与 HDFS 配置文件
cp /opt/hadoop3/etc/hadoop/core-site.xml /opt/starRocks/fe/conf/
cp /opt/hadoop3/etc/hadoop/hdfs-site.xml /opt/starRocks/fe/conf/
cp /opt/hive/conf/hive-site.xml /opt/starRocks/fe/conf/

# 创建 meta 目录并修改 fe.conf
cd /opt/starRocks/fe
mkdir meta
vim conf/fe.conf
# 添加以下配置
meta_dir=/opt/starRocks/fe/meta
priority_networks=node01/24
enable_udf=true

# 指定 Java 11
export JAVA_HOME=/{JAVA 安装目录}/zulu11.66.19-ca-jdk11.0.20.1-linux_aarch64
sed -i '108s/.*/JAVA_VERSION=11/' start_fe.sh

# 启动 FE
cd /opt/starRocks/fe/bin
./start_fe.sh --daemon

4. BE node deployment (node02‑node04)

# 复制配置文件
cp /opt/hadoop3/etc/hadoop/core-site.xml /opt/starRocks/be/conf/
cp /opt/hadoop3/etc/hadoop/hdfs-site.xml /opt/starRocks/be/conf/
cp /opt/hive/conf/hive-site.xml /opt/starRocks/be/conf/

# 配置 be.conf
cd /opt/starRocks/be
mkdir data
vim conf/be.conf
# 示例配置
priority_networks=node02/24
load_process_max_memory_limit_percent=50
mem_limit=75%
storage_root_path=/opt/starRocks/be/data

# 指定 Java 11 并启动 BE
vim bin/start_be.sh
export JAVA_HOME=/{JAVA 安装目录}/zulu11.66.19-ca-jdk11.0.20.1-linux_aarch64
cd /opt/starRocks/be/bin
./start_be.sh --daemon

5. Service monitoring and log cleanup scripts

# 自动删除 3 天前的日志 (FE)
#!/bin/sh
find /opt/starRocks/fe/log/ -mtime +3 -name "fe.*" -exec rm -rf {} \;

# 自动删除 3 天前的日志 (BE)
#!/bin/sh
find /opt/starRocks/be/log/ -mtime +3 -name "be.*" -exec rm -rf {} \;

# FE 监控脚本
#!/bin/sh
pidFE=$(ps aux | grep 'StarRocksFE' | grep -v grep | awk '{print $2}')
if [[ -z $pidFE ]]; then
  sh /opt/starRocks/fe/bin/start_fe.sh --daemon
fi

# BE 监控脚本
#!/bin/sh
pidBE=$(ps aux | grep 'starrocks_be' | grep -v grep | awk '{print $2}')
if [[ -z $pidBE ]]; then
  sh /opt/starRocks/be/bin/start_be.sh --daemon
fi

After all FE and BE nodes are started, the article shows how to verify the deployment using the MySQL client, add backend nodes with ALTER SYSTEM ADD BACKEND , set the root password, create a sample database, and optionally create an external Hive catalog.

In conclusion, the guide demonstrates a complete, Docker‑based method to run StarRocks 3 on ARM platforms, addressing the lack of an official ARM release and confirming that the setup works as expected.