在学习hadoop课程中，讲师介绍了hadoop的单机以及集群部署方式，由于本地资源限制，只有一台虚拟机，所以考虑使用docker的方式实现分布式集群搭建。

如上图：
需要在主节点启动NameNode,Secondary namenode,Resource Manager
从节点启动：DataNode，Node Manager

准备

1. 下载centos镜像: docker pull centos
2. 启动一个容器：docker run -td --name base -i centos bash
3. 进入容器，安装jdk及hadoop，方法和主机一样
4. 容器中安装ssh：yum install openssh-server openssh-clients
5. 设置免密登录： ssh-keygen -t rsa cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys
6. 保存容器修改：docker commit base hadoop:4

配置

新建目录结构如下，只需要创建两个文件：docker-compose.yml以及Dockerfile，其他文件夹是启动后自动生成的

docker-compose.yml

version: '2.0'
services:
  hadoop01:
    build: .
    container_name: "hadoop01"
    volumes:
     - ./hadoop01:/data/hadoop_repo
    tty: true
    privileged: true
    hostname: hadoop01
    ports:
     - "9870:9870"  
     - "8088:8088"
  hadoop02:
    image: "hadoopdockercluster_hadoop01"
    container_name: "hadoop02"
    tty: true
    privileged: true
    hostname: hadoop02
    volumes:
     - ./hadoop02:/data/hadoop_repo
  hadoop03:
    image: "hadoopdockercluster_hadoop01"
    container_name: "hadoop03"
    tty: true
    privileged: true
    hostname: hadoop03
    volumes:
     - ./hadoop03:/data/hadoop_repo

Dokerfile

FROM hadoop:4
ENV JAVA_HOME=/opt/jdk1.8.0_181
ENV HADOOP_HOME=/opt/hadoop-3.2.0-cluster
ENV PATH=.:$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin:$JAVA_HOME/bin
CMD ["/usr/sbin/sshd","-D"]

启动

docker-compose build
docker-compose up -d
docker exec -it hadoop01 bash
格式化HDFS： hdfs namenode -format
启动： start-all.sh

验证

查看容器列表

前台访问：

hdfs: http://192.168.56.104:9870/

yarn: http://192.168.56.104:8088/