网上部署glusterfs集群，都是采用登录到主机，执行命令的方式。如果了解ansible，使用ansible批量管理主机，部署glusterfs集群更方便。

测试环境

有4台虚拟机，所有机器都安装centos7.4系统。

tmaster，tslave1，tslave2作为glusterfs服务器端节点，phn作为glusterfs客户端节点。

10.101.19.12 tmaster
10.101.19.13 tslave1
10.101.19.14 tslave2
10.101.17.54 phn

还有一台ansible节点，用来管理其他主机。（假设你已经装好了ansible节点）

ansible的hosts文件

[clusterfs] 10.101.19.12 10.101.19.13 10.101.19.14 10.101.17.54  [clusterfsserver] 10.101.19.12 10.101.19.13 10.101.19.14  [clusterfsclient] 10.101.17.54

部署过程

1. 在ansible机器上中执行，初始化主机环境

# ping所有主机
ansible clusterfs -m ping

# 修改时区（可选）
ansible clusterfs -m file -a 'path=/etc/localtime src=/usr/share/zoneinfo/Asia/Shanghai state=link'
ansible clusterfs -a 'date -R'

# 修改hosts文件
ansible clusterfs -m lineinfile -a 'path=/etc/hosts regexp=tmaster$ line="10.101.19.12 tmaster"'
ansible clusterfs -m lineinfile -a 'path=/etc/hosts regexp=tslave1$ line="10.101.19.13 tslave1"'
ansible clusterfs -m lineinfile -a 'path=/etc/hosts regexp=tslave2$ line="10.101.19.14 tslave2"'
ansible clusterfs -m lineinfile -a 'path=/etc/hosts regexp=phn$ line="10.101.17.54 phn"'

ansible clusterfs -a "ping tmaster -c 1"
ansible clusterfs -a "ping phn -c 1"

2. 在ansible机器上中执行，配置glusterfs服务器节点

# 为clusterfs服务器节点，安装软件
ansible clusterfsserver -m package -a 'name=centos-release-gluster state=present'
ansible clusterfsserver -m package -a 'name=glusterfs state=present'
ansible clusterfsserver -m package -a 'name=glusterfs-server state=present'
ansible clusterfsserver -m package -a 'name=glusterfs-fuse state=present'
ansible clusterfsserver -m package -a 'name=glusterfs-rdma state=present'

# 为clusterfs服务器节点，启动服务
ansible clusterfsserver -m service -a 'name=glusterd enabled=yes state=started'
ansible clusterfsserver -a 'systemctl status glusterd'

# 为clusterfs服务器节点，创建数据目录
ansible clusterfsserver -m file -a 'path=/opt/gluster/data state=directory'

3. 在master节点上执行，创建集群，和添加volume等
gluster peer probe tmaster 
gluster peer probe tslave1 
gluster peer probe tslave2
gluster peer status

gluster volume info
gluster volume create models replica 3 tmaster:/opt/gluster/data tslave1:/opt/gluster/data tslave2:/opt/gluster/data force
gluster volume info
gluster volume start models

4. 在ansible机器上中执行，为客户端节点，安装软件等
ansible clusterfsclient -m package -a 'name=glusterfs state=present'
ansible clusterfsclient -m package -a 'name=glusterfs-fuse state=present'

为客户端节点，挂载卷
ansible clusterfsclient -m file -a 'path=/opt/gfsmount  state=directory'
ansible clusterfsclient -m mount -a 'path=/opt/gfsmount  state=mounted fstype=glusterfs src=tmaster:models'
ansible clusterfsclient -a "df -h"

5. 在10.101.17.54客户端测试
time dd if=/dev/zero of=/opt/gfsmount/hello bs=100M count=1 

参考

http://blog.csdn.net/phn_csdn/article/details/75153913